Strongly Coupled Heavy and Light Quark Thermal Motion from

AdS/CFT
A. K. Mes1* , R. W. Moerman2, 3 , J. P.Shock1 , W. A. Horowitz2

1 Department of Mathematics and Applied Mathematics, University of Cape Town, Private Bag X3,
Rondebosch 7701, South Africa
2 Department of Physics, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa
3 Department of Physics, Astronomy and Mathematics, University of Hertfordshire, Hatfield,
Hertfordshire, AL10 9AB, United Kingdom
* msxale002@[Link]

November 22, 2021

Abstract
arXiv:2008.09196v3 [hep-th] 19 Nov 2021

We give a pedagogical presentation of heavy and light probe quarks in a thermal plasma using
the AdS/CFT correspondence. Three cases are considered: external heavy and light quarks
undergoing Brownian motion in the plasma, and an external heavy quark moving through
the plasma with a constant velocity. For the first two cases we compute the mean-squared
transverse displacement of the string’s boundary endpoint. At early times the behaviour
is ballistic, while the late time dynamics are diffusive. We extract the diffusion coefficient
for both heavy and light quarks in an arbitrary number of dimensions and comment on the
relevance for relativistic heavy ion phenomenology.

Contents
1 Introduction 3
1.1 Structure of the Article 4

I Theory 5
2 The Gauge/String Duality 5
2.1 Anti-de Sitter Spacetime as a limit of D3-brane Geometry 6
2.2 The AdS/CFT Conjecture 6

3 Brownian Motion: the Dynamics of Langevin’s Model 8

3.1 The Non-retarded Langevin Model 8
3.2 The Generalised Langevin Model 11

II Calculations in the Bulk 15

4 Analysing Heavy Quark Brownian Motion 15
4.1 Polyakov String Equations of Motion 16
4.2 Leading Order String Behaviour 19
4.2.1 Test Strings in R1,1 19
4.2.2 Test Strings in AdS3 -Schwarzschild 20
4.3 Fluctuating Test String Dynamics 23
4.3.1 Nambu-Goto String Equations of Motion for Transverse Fluctuations 23
4.3.2 Calculating the Heavy Quark’s Mean-Squared Displacement s2 (t) in AdS3 -Schwarzschild 26
4.3.3 The Limiting Cases of s2 (t) 32
4.4 Generalising to AdSd -Schwarzschild 34

5 Analysing Light Quark Brownian Motion 35

5.1 Leading Order String Behaviour 35
5.1.1 Parameterising the Limp Noodle: The Bars et al. method 36

1
 5.1.2 Test Strings in R1,1 39
5.1.3 Test Strings in AdS3 -Schwarzschild 43
5.2 Fluctuating Test String Dynamics 44
5.2.1 Calculating the Light Quark’s Mean-Squared Displacement s2 (t) in AdS3 -Schwarzschild 45
5.2.2 The Limiting Cases of s2 (t) 46
5.3 Generalising to AdSd -Schwarzschild 51
5.3.1 Expanding the AdS3 -Schwarzschild Transverse Equations of Motion and its Solution
in the Near-Horizon Region 53
5.3.2 Expanding the AdS3 -Schwarzschild Metric in the Near-Horizon Region 54
5.3.3 The Diffusion Constant in AdSd -Schwarzschild 57

6 Drag Force in AdS/CFT 61

6.1 Test Strings in AdS5 -Schwarzschild 61
6.2 Generalising to AdSd -Schwarzschild 67

7 Conclusions and Future Outlook 69

Appendix A Polyakov String Equations of Motion 72

Appendix B The Virasoro Constraints and String Equations of Motion in Isother-

mal Coordinates 73

Appendix C Energy of a Test String in an AdS-Schwarzschild Background 77

Appendix D Nambu-Goto String Equations of Motion for Transverse Fluctuations 80

Appendix E Canonical Commutation Relations and Normalised Basis 82

Appendix F Leading Order Contributions of the Near-Horizon Tortoise Coordinate 84

Appendix G Accessing the Mathematica Code 86

Bibliography 87

List of Figures
1 The spacetime around Nc D3-branes in type IIB theory 7
2 A fundamental open string of length `0 used as a probe in an AdS black hole background 15
3 Snapshots of a fundamental open string of length `0 used as a probe in an AdS black hole
background, falling as time increases from left to right 35
4 Light-cone coordinate lattice structure imposed on the worldsheet parameter space in order
to determine the general string solution using the Bars et al. method 36
5 The functions (i) F (z) and (ii) f (z) over the range (−6σf , 6σf ) 41
6 The worldsheet parameter space and embedding functions in R1+1 42
7 The worldsheet parameter space in light-cone coordinates and isothermal embedding func-
tions in R1+1 43
8 The worldsheet parameter space and embedding functions in AdS3 -Schwarzschild 44
9 A fundamental open string used as a probe in an AdS black hole background to model an
infinitely massive quark moving through a thermal plasma at a constant velocity v on the
boundary 62

List of Tables
1 The AdS/CFT dictionary - translating between quantities in the bulk and boundary theories 7

Note on Notation
Throughout this article, lower-case Latin alphabet indices (e.g. a, b, c, d) index over the string’s worldsheet
parameter space, typically over the standard parameter variables (τ, σ) or the light-cone parameter variables

2
 (σ + , σ − ). Lower-case Greek alphabet indices (e.g. µ, ν, λ, γ) are used to index over the spacetime, typically
in anti-de Sitter spacetime over the temporal t, radial r, and transverse directions X I . The subspace of
anti-de Sitter spacetime spanned by the transverse directions are always indexed by I, J. In some instances
the Greek alphabet indices are enumerated – for example, the isothermal coordinate X µ , where µ ∈ {0, 1}
indexes over the first and second directions in a two-dimensional subspace of spacetime.

Note on Details

Some of the details which may prove useful to students have been pruned from this paper. They can still
be found at arXiv:2008.09196 [hep-th], if desired.

Note on Coloured Links

In this article, citations are declared in dark green; while links to page numbers, figures, tables, sections
and equations are designed to be less noticeable and are given in dark gray.

1. Introduction

It is both experimentally pertinent and theoretically interesting to examine the movement of a quark un-
dergoing Brownian motion in a hot plasma. The main aim of this work is to present an instructional or
pedagogical approach to using the AdS/CFT correspondence to explore the dynamical behaviour of probe
heavy and light quarks immersed in a thermal plasma.

The plasma in question, the quark-gluon plasma (QGP), is created by colliding lead and gold nuclei at high
energies in the Large Hadron Collider (LHC) at CERN; and the Relativistic Heavy Ion Collider (RHIC)
in Brookhaven National Laboratory (BNL). Before the first results from RHIC in 2000 it was expected
that at high energy densities QCD asymptotic freedom would result in a weakly-coupled system exhibiting
gas-like behaviour. However, the experimental results [1] indicated that the produced QGP (a deconfined
state of mater consisting of quarks and gluons) does not expand isotropically and behaves, in-fact, as a
strongly-coupled medium [2, 3]. The thermal plasma expands anisotropically in its azimuthal direction.
The momentum anisotropy of the measured particles is known as elliptic flow 1 . This discovery did not
definitively settle the question as to whether the QGP is weakly- or strongly-coupled. There is evidence
to support both. Weak coupling techniques from perturbative quantum chromodynamics (pQCD) have
been successful in predicting the distributions of high transverse momentum observables [6, 7, 8]; while low
transverse momentum observables described by near-ideal relativistic hydrodynamics [9, 10, 11, 12] can be
understood within a strong coupling paradigm [13, 14]. Further, jet suppression [15, 16] and heavy quark
energy loss studies [17] support the theory of a strongly-coupled thermal plasma.

At strong coupling, the usual perturbation techniques are no longer applicable and many quantities in
gauge theories become difficult to compute. In 1974, ’t Hooft [18] postulated that the generalisation of
the quantum chromodynamic SU (3) gauge group (where Nc = 3 is the number of quark colours in the
theory) would be an SU (Nc ) gauge group, where the large Nc limit is taken (Nc → ∞ while λ := gY M Nc
is kept fixed and large). The Yang-Mills coupling is denoted by gY M , and the quantity λ is known as the
’t Hooft coupling. The large Nc limit provides an approximation to compute the gauge theory at strong
coupling2 . Inspired by ’t Hooft’s idea, a promising approach in studying the strong coupling limit of non-
abelian quantum field theories (such as QCD) was formulated in 1999 by Juan Maldacena [21]. Known as
the anti–de Sitter/Conformal Field Theory (AdS/CFT) correspondence, it is a study of the duality between
bulk gravitational physics of a given d-dimensional spacetime and a (d − 1)-dimensional gauge theory on its
boundary. Considering the phenomenologically relevant case, at finite temperature the equivalence exists
between type IIB string theory on ten-dimensional spacetime approximated by Einstein’s general relativ-
ity on a five-dimensional non-compact anti-de Sitter spacetime and a five-dimensional sphere (AdS5 × S 5 )

1 For two comprehensive reviews on discoveries relating to the hydrodynamic description of relativistic heavy-ion collisions
(specifically collective flow and viscosity), see [4, 5].
2 In the large N limit, a topological factor N ζ (where ζ is the Euler characteristic) is assigned to each Feynman graph. Summing
c c
over these graphs can be thought of as summing over the worldsheets of the supposed ‘QCD string’ (the QCD string dual is
unknown and is approximated by the N = 4 SYM string dual) to yield the partition function for the large Nc theory [19, 20].

3
 (referred to as either the gravity or bulk theory); and N = 4 supersymmetric Yang-Mills theory (SYM) on
four-dimensional Minkowski spacetime [20] (referred to as either the gauge or boundary theory). Quantities
which are difficult to compute in a strongly-coupled gauge theory can be calculated in a weakly-coupled
gravity theory, and translated back via the AdS/CFT dictionary (see table (1), page 7)3 . After its initial
discovery, generalisations of the AdS/CFT correspondence followed and the field is now grouped under
gauge/string dualities. In the succeeding two decades since the initial publications of [21, 24, 25, 26, 27]
there has been much interest (and success) in using the gauge/string duality to determine properties of the
given gauge theory’s plasma state.

Propelled by experiments at RHIC and the LHC, theoretical developments in this regard have been inspired
by an early calculation [28] which found the ratio of shear viscosity to volume entropy density (η/s) in
the strong coupling regime using the AdS/CFT correspondence. This ratio is experimentally measurable,
and data provided by RHIC supported the theoretical result [29, 30]. Building on this, Kovtun et. al.
[31, 32] discovered the shear viscosity/entropy density ratio is universal for a large class of these QFTs and,
further, that this ratio is a universal lower bound for the viscosity in general4 . This discovery was essential
in understanding the elliptic flow that had been observed at RHIC, and sparked interest in the link between
string theory and relativistic heavy-ion phenomenology5 . Exploring this link, two pertinent examples this
article will focus on are (i) the dissipative and diffusive behaviour of a massive quark moving through a field
theory plasma explained by studying trailing strings in the AdS spacetime [35, 36, 22, 37, 38, 39, 40, 41];
and (ii) heavy and light quarks undergoing Brownian motion studied by examining transverse fluctuations
on open strings in an AdS black hole background [42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52].

Motivated by these past studies, this article seeks to better understand the fluctuating energy loss of light
and heavy quarks in a thermal medium. Specifically, heavy and light probe quarks in the gauge theory
are considered to undergo Brownian motion6 , and the mean-squared transverse displacement of the string’s
boundary endpoint s2 (t; d) (equivalent to the mean-squared displacement travelled by the external quark
which is initially at rest in the thermal plasma) is computed. Functioning as a pedagogical work for these
types of gauge/string calculations, this article aims to present a consistent framework7 – providing insights
into published results and including proofs of some of the more vague statements in the literature. In terms
of original advancement, previous generalisations of s2 (t; 3) in AdS3 -Schwarzschild to AdSd -Schwarzschild
for the light quark’s case are challenged and the method in which to generalise to s2 (t; d) correctly given for
the first time. Notwithstanding this, the other main contribution of the article to this field is presenting a
definitive, consolidated theoretical derivation – from which further research, including analytical calculations
with different test string configurations or numerical analysis confirming previous analytic results, can be
undertaken.

1.1. Structure of the Article

This article is organised in two parts. The first part provides some useful theoretical background needed to
understand the bulk calculations in Part II. The central results of this article are given in the second part.

In Part I:

First, the gauge/string duality is introduced in section (2). Particular attention is paid to the throat con-
struction of anti-de Sitter spacetime as a limit of D3-brane geometry, and the justification of the AdS/CFT
conjecture. In section (3) the basic theory of Brownian motion and the non-retarded and retarded Langevin
models are explored. The first and second fluctuation-dissipation theorems are derived. The mean-squared

3 There is a hitherto unaddressed assumption at play here – the theory of quantum chromodynamics is not exactly equivalent to
N = 4 supersymmetric Yang-Mills theory. The theories are however analogous: while the matter fields and quantum dynamics
of the theories differ, their gauge fields and tree-level interactions are common [22]. Qualitatively the plasmas of QCD and
N = 4 SYM share many properties, such as Debye screening and finite spatial correlation lengths [23]. Hence, for studies in
the high temperature regime it is assumed that the plasma of N = 4 supersymmetric Yang-Mills is similar to the quark-gluon
plasma of QCD.
4 In addition, see Buchel et al.’s contributions [33, 34].
5 A worthwhile review on the topic is given by Casalderrey-Solana et al. [14].
6 Brownian motion [53] concerns the ceaseless, random motion of a given particle undergoing microscopic collisions with the

constituent particles of the fluid it is immersed in. The motion is responsible for the dissipative nature of a system and its
approach to thermal equilibrium. Any particle suspended in a finite temperature fluid undergoes this motion, and as such a
probe quark immersed in the QGP behaves the same way. The AdS/CFT correspondence can be used to study the behaviour
a probe quark exhibits while interacting with the strongly-coupled thermal plasma.
7 A necessary pursuit considering how versatile the field is – publications on gauge/string dualities span all physics arXivs [20].

4
 displacement s2 (t) is calculated and its behaviour in the early and late time limits are discussed. At early
times, the Brownian particle’s behaviour is proportional to time and the motion√is expected to be ballistic
s(t) ∼ t; while, at late times, the Brownian particle’s motion is diffusive, s(t) ∼ t.

In Part II:

To begin Part II, section (4) focuses on calculating an external heavy quark’s diffusion coefficient in the bulk
AdS3 -Schwarzschild spacetime. The dual description of the on-mass-shell heavy quark exhibiting Brown-
ian motion in the thermal plasma – an open string attached at the anti-de Sitter boundary8 and hanging
towards the horizon undergoing transverse fluctuations – is used to compute the mean-squared transverse
displacement of the string’s boundary endpoint s2 (t; 3). The diffusion coefficient is extracted at late times.
Similarly, in section (5) an off-mass-shell light quark exhibiting Brownian motion in the thermal plasma is
modelled as an open string undergoing transverse fluctuations (initially stretched between the AdS bound-
ary and just above the horizon) whose AdS boundary endpoint is released to fall at the local speed of light
in the presence of a space filling D7-brane [54]. Initially this section focuses on calculating an external
light quark’s diffusion coefficient in the bulk AdS3 -Schwarzschild spacetime. Then, importantly, the results
are generalised to AdSd -Schwarzschild by expanding and truncating the near-horizon metric. Further, in
section (6), an infinitely massive probe quark moving through the thermal plasma with a constant velocity
is considered. An open string trailing out behind the quark (arcing down into AdS-Schwarzschild) is used
to model the situation. From this, the drag force on the test string is calculated in the bulk and rewritten
– via the AdS/CFT correspondence – in terms of relevant quantities in the gauge theory9 . The Langevin
equation is then used to extricate the friction coefficient, and (by means of the Einstein-Sutherland rela-
tion) the diffusion coefficient for a heavy quark in AdSd -Schwarzschild. Section (7) concludes the article
and proposes avenues for future research.

There are a number of appendices to this article. Appendix (Appendix A) derives the string equations of
motion by calculating the functional derivative of the Polyakov action with respect to the string worldsheet
coordinates and setting this variation to zero. In an analogous fashion, the string equations of motion
for the transverse fluctuations are found in appendix (Appendix D) by varying the Nambu-Goto action.
The Virasoro constraints and string equations of motion in isothermal coordinates are given in appendix
(Appendix B); while a derivation of the energy of a test string in an AdS-Schwarzschild background can
be found in appendix (Appendix C). Appendix (Appendix E) aims to fix the normalization constant
Aω , thereby completely determining the general solution for the transverse equations of motion. As the
penultimate, appendix (Appendix F) calculates the leading order contributions of the tortoise coordinate
in the near-horizon region. Appendix (Appendix G) provides details for accessing the GitHub repository
where the latest annotated Mathematica code notebooks supporting the analytical analysis of open string
evolution presented here, can be found. The authors believe it is important to make the Mathematica
notebooks available for reference purposes and in the spirit of open collaboration.

Part I
Theory
2. The Gauge/String Duality

As recognised by ’t Hooft [18], understanding gauge theories with an SU (Nc ) gauge group in the large Nc
limit perhaps offers the best course to illuminating the strong coupling behaviour of QCD. In the late 1900’s
it was suspected that string theory could describe the large Nc limit [55]. Maldacena [21] first made the
suggestion that a (d − 1)-dimensional, conformally invariant field theory in the limit of large Nc corresponds
to string theory and supergravity on anti-de Sitter spacetime in d dimensions times by a d-dimensional
spherical compact manifold (AdS×S). This idea, based on the holographic principle10 and formally known
as the AdS/CFT correspondence, was developed in the months that followed by [24, 25, 26, 27], to name a
few. The AdS/CFT correspondence is an example of gauge/string dualities which have subsequently been

8 The boundary of a D7-brane stretching from r = ∞ to the string’s boundary endpoint [54].
9 Historically, the drag force calculation was the first application of AdS/CFT to understand the behaviour of a probe heavy
quark in the quark-gluon plasma (see [22, 35]).
10 Holographic as proposed by ’t Hooft, Susskind and Thorn in [56, 57, 58] respectively. In this sense, a holographic theory

encodes a theory in d dimensions by a theory in (d − 1) dimensions.

5
 developed into a powerful tool in understanding strongly-coupled systems11 .

2.1. Anti-de Sitter Spacetime as a limit of D3-brane Geometry

The solution of low energy, type IIB string theory containing D3-branes prompted the formulation of the
gauge/string dualities. Solving the supergravity equations of motion yields the spacetime metric sourced
by Nc Dp-branes [59, 60, 61]. Specifically, the spacetime for the extremal D3-brane12 is

ds2 = H −1/2 −dt2 + d~x2 + H 1/2 dr2 + r2 dΩ25 ,

(2.1)

where the D3-brane is extended along the spatial coordinates ~x = (x1 , x2 , x3 ); H(r) = 1 + l4 /r4 is known
as the warping factor ; and the second term metric describes the y−directions transverse to the D3-brane
written in spherical coordinates (where r2 = y12 + y22 + ... + y62 is the radial coordinate). The parameter l
is interpreted as the characteristic length scale of the range of the Nc D3-branes’ gravitational effects [14].
In the limit r  l, H ' 1 and the metric Eq.(2.1) reduces to Minkowski spacetime13 . In the limit r  l,
H ' l4 /r4 (which corresponds to a series expansion around r/l = 0 to leading order). Hence, in this limit
the metric Eq.(2.1) becomes

r2
l2 2
ds2 = ds2AdS5 + l2 dΩ25 , where ds2AdS5 = −dt2
+ d~
x 2
+ 2 dr . (2.2)
l2 r

Now l is identified as the radius of curvature of AdSd and S d (see table (1), page 7); ds2AdS5 is the metric
of five-dimensional anti-de Sitter spacetime14,15 ; and dΩ25 is the metric on S 5 with unit radius. Therefore,
the spacetime sourced from a stack of D3-branes corresponds to ten-dimensional Minkowski spacetime far
away from the branes (see blue patch, figure (1)), while a throat geometry of the form AdS5 × S 5 becomes
apparent close to the branes (see red patch, figure (1)).

Generalising to non-zero finite temperature T systems by exciting the degrees of freedom on the D3-brane,
adapts the anti-de Sitter spacetime part of the metric. Specifically, Eq.(2.2) becomes

r2 l2
ds2 = ds2AdS5 -Sch + l2 dΩ25 , ds2AdS5 -Sch = 2 2
dr2 ,


where −hdt + d~
x + (2.3)
l2 r2 h
4
where h(r) = 1 − rH /r4 is known as the blackening factor. The first term of the metric in Eq.(2.3) describes
an AdS5 spacetime with a Schwarzschild black hole horizon at r = rH 16,17 .

A stack of Nc D3-branes can be described by (i) closed strings propagating in a curved spacetime geometry,
which – in the low energy limit – becomes closed IIB string theory in AdS5 × S 5 (as discussed above); but
also (ii) open strings attached to a hyper-plane in flat spacetime, where the low-energy limit is given by
N = 4, SU (Nc ) supersymmetric Yang-Mills theory [63, 14]. The hypothesis that the two descriptions are
equivalent was the first example of the AdS/CFT conjecture.

2.2. The AdS/CFT Conjecture

The best example of a gauge/string duality remains the equivalence between type IIB string theory in an
AdS5 ×S 5 background to N = 4, SU (Nc ) SYM theory on the four-dimensional Minkowski spacetime bound-
ary of AdS5 . As has been seen in subsection (2.1), if the above duality is considered at finite temperature,
the addition of a black hole in the anti-de Sitter spacetime is necessary. The spacetime is then referred to
as the anti-de Sitter-Schwarzschild spacetime (or AdS-Schwarzschild).

11 A duality between two theories describes a situation where one theory is in the strong coupling limit, while the other theory is
in the weak coupling limit [20]. In order to study a strongly-coupled gauge theory, the equivalent weakly-coupled gravitational
string theory (i.e. a theory where the curvature of the spacetime is small) can instead be used.
12 Only the ground state of the N D3-branes is being considered.
c
13 A small correction proportional to l4 /r 4 is present.
14 The defining characteristic of AdS is a spacetime described by a constant negative curvature.
15 The five-dimensional anti-de Sitter spacetime metric Eq.(2.2) represents a Poincaré chart of AdS spacetime. There is an
5
alternative formulation of anti-de Sitter spacetime in terms of global coordinates which provides a global coordinate chart (see,
for example, [62]).
16 The Schwarzschild black hole is the solution to the Einstein equation with no matter fields or cosmological constant.
17 The radial position of the black-brane horizon r
H is proportional to the temperature T (see table (1), page 7).

6
 Figure 1: The spacetime around Nc D3-branes in type IIB theory, where l is the radius of curvature of AdSd and
S d (adapted from [63, 14]). See page 6 for details.

The AdS/CFT correspondence claims that the partition functions of the gauge and gravitational theories
are equivalent, i.e. Zgauge = ZAdS [20]. In doing so the conjecture specifies a dictionary between two
physical theories which appear to be very different (these relations are summarised in the reference table
(1)). In practice, the strong/weak coupling nature of the duality means that it is difficult to explicitly test
the validity of the conjecture18 . There are, however, several contiguous tests – coupling-independent prop-
erties of these theories that can be compared to investigate the duality. Specifically for AdS5 × S 5 /N = 4
supersymmetric Yang-Mills, tests include (i) comparing the global symmetries of the theories; (ii) matching
the coupling-independent correlation functions (these are normally protected from quantum corrections and
related to anomalies); (iii) comparing the spectrum of chiral operators; and (iv) examining the qualitative
behaviour of the theory deformed by chiral operators19 [27].

AdSd (d − 1)-dimensional Gauge Theory Description

√
l α0 λ1/4 Radius of curvature of AdSd and S d
√
`s α0 ≡ λ−1/4 l Fundamental string length scale
0
T0 1/2 π α String tension
(l/`s )4 λ ’t Hooft coupling20
rH 4π l2 T /(d − 1) Radial position of the black hole horizon
(d − 1) rH /4π l2 T ≡ 1/β Temperature of the gauge theory21
rc ≡ (rs + `0 ) 2 π α0 (Mrest + ∆m) Minimal radius of D7-brane22
T0 rH ∆m(T ) Thermal rest mass shift
T0 (rc − rH ) Mrest (T ) Static thermal mass of external particle23

Table 1: The AdS/CFT dictionary - translating between quantities in the bulk and boundary theories (adapted
from Herzog et al. [35])

18 With contemporary knowledge, pertubative computations in λ can mostly only be achieved in the field theory, while in the
string theory only pertubative calculations in 1/λ are possible. Therefore, comparing the correlation functions of these theories
is in principle not realisable [27].
19 Qualitative tests include the existence of confinement in finite temperature gauge theories [64], and examining how the theory

behaves on its moduli space [65, 66, 67].

7
 The main advantage of the AdS/CFT correspondence is that otherwise intractable problems are able to
be solved by their mapping onto the equivalent dual. Applications of the correspondence include gluon
scattering amplitudes calculations in strongly-coupled N = 4 supersymmetric Yang-Mills [69]; and the
study of holographic superconductivity and critical phenomena [70, 71, 72]24 .

3. Brownian Motion: the Dynamics of Langevin’s Model

Discovered by Robert Brown in 1827 [53] the ceaseless and irregular motion which small pollen particles un-
dergo when suspended in water, became known as Brownian motion. Brown postulated that this movement
had a physical rather than biological origin. In 1905 Albert Einstein [74]25 combined statistical mechanics
and the diffusion equation, to arrive at a theoretical explanation of the phenomenon. His famous formula
for the particle’s mean-squared displacement, hs2 (t)i, provided an observable for experimental physicists to
measure. Independently, Marian Smoluchowski [75]26 derived a similar result based on combinatorics and
kinetic theory’s mean-free-path approximation [76].

In a mathematical context, Louis Bachelier wrote his 1900 doctoral thesis on modelling prices on the Paris
stock exchange as the limit of a random walk [77]27 . Neither Einstein nor Bachelier rigorously classified
Brownian motion as a stochastic process. This was taken up by the American mathematician, Norbert
Wiener, who ultimately proved the existence of Brownian motion by defining it as a stochastic process in
1923 [78]. Wiener’s output includes a series of papers starting in 1918 giving the mathematical definition and
properties of the physical process of Brownian motion abstracted as a stochastic process. Hence, Wiener
processes (essentially the same concept as Brownian motion but emphasising the mathematical aspects)
developed separately to the school of thought emphasising the physical aspects.

Turning back towards the explanations of physicists, Einstein and Smoluchowski failed to take into account
the inertia of the particle undergoing Brownian motion. Applying Newtonian dynamics to a Brownian
particle, Paul Langevin [79]28 arrived at a useful phenomenological model in 1908.

3.1. The Non-retarded Langevin Model

Consider the following: a non-relativistic particle of mass m, undergoing Brownian motion in one spatial
dimension, can be described by the Langevin equation

ṗ(t) = −γ0 p(t) + F (t) ,

(3.1)
⇒ m v̇(t) = −m γ0 v(t) + F (t) ,

where ṗ ≡ dp/dt, γ0 is a constant known as the friction coefficient, and v is the velocity of the particle.
Eq.(3.1) is a stochastic differential equation and can be thought of as consisting of two parts: i). a sys-
tematic term −mγ0 v corresponding to the friction force, and ii). a fluctuating part F (t) corresponding
to a random force [80]. It is worthwhile to note that although it is simpler to understand the Langevin
equation in terms of a friction force and a random force – on a microscopic scale both forces are caused by

20 The ’t Hooft coupling [18] is specifically defined in the phenomenologically relevant AdS5 /N = 4 SYM case. In this case: λ ≡
2 √
gYM Nc , where Nc is the number of colours and the Yang-Mills coupling is related to the string coupling by: gYM = 2 π gs .
4
However, this article follows [42] in using the same terminology to refer to (l/`s ) for general d dimensions.
21 This is equivalent to the Hawking temperature of the black hole.
22 A UV cut-off surface imposed near the boundary in order to consider external particles of finite mass in the gauge theory.
23 In the AdS /N = 4 SYM case, this is the free energy of a quark at rest in N = 4 SYM plasma. In the limit of zero temperature
5
it is equal to the QCD Lagrangian quark mass, mq [68].
24 See [20] for a extensive review on systems with phase transitions in AdS/CFT, and [73] for an introduction to holographic

methods for condensed matter physics.

25 Original publication: Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden

Flüssigkeiten suspendierten Teilchen, Annalen der physik 322.8 (1905): 549-560. Translated from German.
26 Original publication: Zarys kinetycznej teoriji ruchów browna i roztworów metnych, Rozprawy i Sprawozdania z Posiedzen

Wydzialu Matematyczno-Przyrodniczego Akademii Umiejetnosci 3 (1906): 257-282. Reprinted in German: Zur kinetischen
theorie der brownschen molekularbewegung und der suspensionen, Annalen der physik 326.14 (1906): 756-780. Translated
from German.
27 Original publication: Théorie de la spéculation, Annales scientifiques de l’École normale supérieure. Vol. 17. (1900). Trans-

lated from French.

28 Original publication: Sur la théorie du mouvement brownien, Compt. Rendus 146 (1908): 530-533. Translated from French.

8
 the particle’s collision with the constituent modules of the immersive fluid [42].

Concerning the statistical properties of F (t), three assumptions are made. First, the average value of the
random force at time t, over a particle ensemble with the same initial velocity, vanishes

hF (t)i = 0 , (3.2)

where an ensemble is defined as a large number of similar, but independent particles. Second, the autocor-
relation of the random force is related to a constant, κ0 , which measures the fluctuation strength

hF (t) F (t0 )i = κ0 δ(t − t0 ) , (3.3)

where δ(x) is the Dirac delta function. This assumption implies

Z ∞
hF (t) F (t0 )i dt0 = κ0 . (3.4)
−∞

Lastly, F (t) is assumed to be a Gaussian process29 .

To find the solutions of Eq.(3.1) – a first order linearly separable differential equation – the variables are
separated, each side is multiplied with respect to eγ0 t , and integrated between initial time t0 and t

m v̇(t) eγ0 t + m γ0 v(t) eγ0 t = F (t) eγ0 t

Z t  Z t  Z t
0 0 t 0 1 0
⇒ v̇(t0 ) eγ0 t dt0 + v(t0 ) eγ0 t − v̇(t0 ) eγ0 t dt0 = F (t0 ) eγ0 t dt0
t0 t 0 t0 m t0 (3.5)
Z t
1 0
⇒ v(t) = v0 e−γ0 t + F (t0 ) e−γ0 (t−t ) dt0 ,
m t0

where the initial velocity of the Brownian particle is denoted by v(t0 ) = v0 , γ0 ∈ R+ and it is assumed
that t ≥ 0. Unless otherwise stated, the system is thought of as having an initial time t0 = 0, and being in
equilibrium at t = ∞. As time progresses the velocity of the Brownian particle depends on the exponential
decay of the initial velocity (the first term in Eq.(3.5)) and the extra velocity caused by the random force
(the second term in Eq.(3.5)).

Since, on average, the random force vanishes (Eq.(3.2)) the average of the velocity (Eq.(3.5)) is

hv(t)i = v0 e−γ0 t , t ≥ 0. (3.6)

Due to friction, the mean velocity decreases exponentially. Using Eq.(3.5), the autocorrelation function of
the velocity is given by
Z t Z t0
0 0 1 0 000
) −γ0 (t−t00 )
hv(t)v(t )i = v02 e− γ0 (t+t ) + 2 dt00
dt000 hF (t00 ) F (t000 )i e−γ0 (t −t e , (3.7)
m t0 t0

where the two cross-terms which appeared when multiplying v(t)v(t0 ) are both first order in the noise and,
as such, disappear when averaging over the noise. In the limit t = t0 , Eq.(3.7) reduces to
Z t Z t
2 −2 γ0 t 1 0 0 00
2
hv (t)i = v0 e + 2 dt dt00 hF (t0 ) F (t00 )i e−γ0 (t−t ) e−γ0 (t−t ) . (3.8)
m t0 t0

29 The central limit theorem can be employed to justify this assumption. The random force F (t) is thought of as resulting
from the superposition of many identically distributed random functions, since the Brownian particle has undergone multiple
collisions [81].

9
 Using the assumption Eq.(3.3), the mean-squared velocity becomes
κ0
hv 2 (t)i = v02 e−2 γ0 t + 1 − e−2γ0 t


2
(3.9)
2 m γ0

Taking the limit as t → ∞, the value of the mean-squared velocity at equilibrium is

κ0
lim hv 2 (t)i = . (3.10)
t→∞ 2 m2 γ0

Assuming the fluid is in thermodynamic equilibrium at temperature T , the average energy of the particle
takes on its equipartion value30,31 hEi = T /2 [81]. Since hE(t)i := mhv 2 (t)i/2, Eq.(3.10) becomes

κ0 = 2 γ0 m T , (3.11)

which describes the relationship between the magnitude of the random force κ0 and the friction coefficient
γ0 . Using assumption Eq.(3.4), Eq.(3.11) can be rewritten into the form
Z ∞
1
γ0 = hF (t) F (t0 )i dt0 . (3.12)
2 m T −∞

This equation is recognised as the second fluctuation-dissipation theorem [82]. The theorem states that the
random fluctuations of the Brownian particle have the same origin as the dissipative frictional force acting
on the Brownian particle as it moves in the medium. In order to calculate the position of the Brownian
particle, x(t) := v(t)t, Eq.(3.5) can be integrated between the initial time t0 and t

Z t Z t Z t Z t00
0 0 −γ0 t0 0 1 00
−t0 )
v(t ) dt = v0 e dt + F (t0 ) e−γ0 (t dt0 dt00
t0 t0 m t0 t0
(3.13)
t 0
1 − eγ0 (t−t ) 0
Z
v0 1
⇒ x(t) = x0 + (1 − e−γ0 t ) + F (t0 ) dt ,
γ0 m t0 γ0

The average of the displacement (remembering Eq.(3.2)) is

v0
hx(t)i = x0 + (1 − e−γ0 t ) . (3.14)
γ0

The displacement s(t) := x(t) − x0 , is easily read off from Eq.(3.13)

t 0
1 − eγ0 (t−t ) 0
Z
v0 1
s(t) = (1 − e−γ0 t ) + F (t0 ) dt . (3.15)
γ0 m t0 γ0

In order to calculate the particle’s mean-squared displacement, hs2 (t)i, Eq.(3.15) can be squared and aver-
aged,
Z t Z t
v2 1 0
00

hs2 (t)i = 02 (1 − e−γ0 t )2 + 2 2 dt0 dt00 hF (t0 )F (t00 )i 1 − eγ0 (t−t ) 1 − eγ0 (t−t )
γ0 m γ 0 t0 t0

t
v02
Z
κ0 0
2
= (1 − 2 e−γ0 t + e−2γ0 t ) + 2 2 dt0 1 − eγ0 (t−t ) (3.16)
γ02 m γ0 t0

v02 −γ0 t −2γ0 t κ0  −γ0 t −2γ0 t


= (1 − 2 e + e ) + 2γ 0 t − 3 + 4 e − e .
γ02 2 m2 γ03

30 The equipartition theorem of classical statistical mechanics relates the average energy of a system to its temperature.
31 The Boltzmann constant is taken to be kB = 1.

10
 Note that, in the second line, the two cross-terms vanish and the assumption Eq.(3.3) has been used.
Further, a second average of hs2 (t)i over all possible initial velocities needs to be taken. This yields,
T 1  
hs2 (t)i = 2 2 γ0 t − 1 + e−γ0 t , (3.17)
m γ0

where the relation for κ0 (Eq.(3.11)), and the equipartition of energy (hv02 i = T /m) is used.

The Einstein-Sutherland relation defines the diffusion coefficient

T
D := . (3.18)
γ0 m

Hence, Eq.(3.17) becomes

2D  
hs2 (t)i = γ0 t − 1 + e−γ0 t . (3.19)
γ0

This result was first derived in 1917 by Dutch physicist Leonard Ornstein [83]. The early and late time
limits of the mean-squared displacement hs2 (t)i should be examined.

(i) In the limit of early times t  γ10 : a Taylor expansion of e−γ0 t can be performed.

γ02 t2
  
2D T 2
∴ hs2 (t)i ≈ γ0 t − 1 + 1 − γ0 t + = D γ0 t2 = t , (3.20)
γ0 2 m

where the last equality comes from using the Einstein-Sutherland relation Eq.(3.18). In this limit,
known as the ballistic regime, the Brownian particle’s behaviour is proportional to time, s(t) ∼ t. This
behaviour is expected since the initial behaviour of the Brownian particle – before it is bombarded by
a substantial number
p of fluid particles – is inertial (with a velocity determined by the equipartition of
energy v(t) = T /m ) [42].

(ii) In the limit of late times t  γ10 : e−γ0 t → 0 as t → ∞, and γ0 t becomes the dominant term.

∴ hs2 (t)i ≈ 2 D t . (3.21)

Eq.(3.21) is the famous equation

√ found by Einstein in his 1905 paper [74]. In this limit, referred to
as the diffusive regime, s(t) ∼ t and the particle experiences a random walk. The Brownian particle
deviates from its initial course due to numerous collisions with the constituent fluid particles that
cause the Brownian particle to ‘forget’ its early time behaviour [42].

The cross-over time32 between the diffusive regime and the ballistic regime is denoted by
1
trelax ∼ , (3.22)
γ0
which represents the time it takes for a Brownian particle that had some initial velocity at t = t0 to ther-
malize in the medium.

3.2. The Generalised Langevin Model

The Langevin model can be generalised to include retardation effects. The retarded Langevin equation33 is
given by [82, 84]
Z t
ṗ(t) = − dt0 γ(t − t0 ) p(t0 ) + F (t) + K(t) , (3.23)
−∞

32 In this article, the cross-over time is interchangeably referred to as the relaxation time.
33 The retarded Langevin Equation is also known as the generalised Langevin Equation.

11
 where the memory kernel γ(t − t0 ) allows the friction term to depend on the past trajectory of the Brownian
particle, and K(t) is an external force which acts on the system. The system is taken to be in equilibrium
in the limit t0 → −∞. The non-retarded Langevin Equation Eq.(3.1) only holds if the Brownian particle
is taken to have infinite mass with respect to the constituent fluid particles. The generalisation of the
Langevin equation to Eq.(3.23) fixes two physical problems, (i) the friction is no longer considered to be
instantaneous, (ii) a correlation can exist between random forces at different times. The random force in
Eq.(3.23) is taken to satisfy

hF (t)i = 0 and hF (t) F (t0 )i = κ(t − t0 ) , (3.24)

where κ(t) is an unspecified function. As in the non-retarded case, F (t) is assumed to be a Gaussian process.

The retarded Langevin equation is Fourier transformed34 in order to analyse its behaviour.
Fourier transforming both sides of Eq.(3.23), and using the linearity and derivative properties of the Fourier
Transform gives
hZ t i
−i ω p(ω) + F dt0 γ(t − t0 ) p(t0 ) = F (ω) + K(ω) . (3.25)
−∞

The convolution property would be useful in calculating the second term in Eq.(3.25). However, the asym-
metric bounds of this integral (t0 ∈ [−∞, t]) are inconvenient. Hence, the memory kernel’s bounds are
redefined by introducing the causal memory kernel

γ̃(t) = Θ(t) γ(t) , (3.26)

where Θ(t) is the Heavy-side function. Notice that while γ(t − t0 ) is defined only for t > 0, γ̃(t − t0 ) is
defined for all t.

Rewriting Eq.(3.25) in terms of the causal memory kernel yields

hZ ∞ i
⇒ −i ω p(ω) + F dt0 γ̃(t − t0 ) p(t0 ) = F (ω) + K(ω)
−∞ (3.27)
⇒ −i ω p(ω) + p(ω) γ̃(ω) = F (ω) + K(ω) ,

where, in the second line, the symmetric bounds of the integral ensured that the convolution property can
be applied. Note that in Eq.(3.27)
Z ∞ Z ∞ Z ∞
γ̃(ω) = dt γ̃(t) eiωt = dt Θ(t) γ(t) eiωt = dt γ(t) eiωt := γ[ω] , (3.28)
−∞ −∞ 0

is actually the Fourier-Laplace transform of γ(t), while

Z ∞
p(ω) := dt p(t) eiωt , (3.29)
−∞

F (ω) and K(ω) are the standard Fourier transforms of p(t), F (t) and K(t) respectively. By rearranging
Eq.(3.27) into a simpler form

F (ω) + K(ω)
p(ω) = , (3.30)
γ[ω] − i ω

34 TheFourier and Inverse Fourier Transform are defined here as

Z ∞ Z ∞
G(ω) := F [ g(t) ] = g(t) eiωt dt and g(t) := F −1 [ G(ω) ] = G(ω) e−iωt dt
−∞ −∞

where w = −2πf .

12
 it is now easy to take the statistical average. Remembering the assumptions regarding the random force
(Eq.(3.24)), yields
1
hp(ω)i = µ(ω) K(ω), where µ(ω) := . (3.31)
γ[ω] − i ω

The quantity µ(ω) is known as the admittance, and describes the system’s response to an external pertur-
bation. Measuring the response hp(ω)i to an external force on the system K(ω), the admittance µ(ω) can
be calculated and hence γ[w] determined.

The first fluctuation-dissipation theorem relates the admittance to the autocorrelation of the equilibrium
velocity [82, 81]. Following Eq.(3.7) the autocorrelation function of the equilibrium velocity for the retarded
Langevin equation Eq.(3.23) is

Z t Z 0
1 00 0
hv(t)v(0)i = dt0 dt00 hF (t0 ) F (t00 )i eγ t e−γ (t−t )
m2 −∞ −∞
Z t Z 0
κ0 00 0
(3.32)
= dt0 dt00 δ(t0 − t00 ) eγ t e−γ (t−t )
m2 −∞ −∞

T −γ |t|
= e ,
m

where |t| is the absolute value of t; the random force correlation function is chosen, for simplicity, to have the
form κ(t − t0 ) = κ0 δ(t − t0 ); and Eq.(3.11) has been used in the third line. The Fourier-Laplace transform
of Eq.(3.32) is

Z ∞
T 1
hv(t)v(0)i eiωt dt =
0 m γ[ω] − iω
(3.33)
Z ∞
m iωt
⇒ µ(ω) = hv(t)v(0)i e dt ,
T 0

where the definition of the admittance is recognised. Eq.(3.33) is known as the first fluctuation-dissipation
theorem [82].

The power spectrum

Z ∞
IO (ω) = dt hO(t0 ) O(t0 + t)i eiωt , (3.34)
−∞

is defined for a quantity O(t). The Wiener-Khintchine theorem35 states that a stationary random process’
autocorrelation function has a spectral decomposition described by the process’ power spectrum. The
theorem allows one to compare the autocorrelation function and the power spectrum

hO(ω) O(ω 0 )i = 2 π δ(ω + ω 0 ) IO (ω) , (3.35)

of a quantity O(t).

As previously mentioned, γ(t) and κ(t) are related to each other by the second fluctuation-dissipation
theorem. However they can be individually determined by considering two cases where different external
forces are applied on the system:

(i) the external force is periodic with frequency ω (i.e. K(t) = K0 e−iωt ), and Eq.(3.31) becomes

hp(t)i = µ(ω) K0 e−iωt . (3.36)

35 The Wiener-Khintchine theorem was proven for a deterministic function by Norbert Wiener in 1930 [85].

13
 The memory kernel of the system γ(t) can be determined by measuring the response hp(t)i to K(t).
(ii) the external force K(t) is absent, and Eq.(3.30) becomes

F (ω)
p(ω) = . (3.37)
γ[ω] − i ω

In order to determine γ(t) and κ(t) individually, p(ω) should first be squared and averaged to find the
relation between the power spectrum of p (Ip (ω)) and the power spectrum of F (IF (ω)). This yields,

IF (ω)
Ip (ω) = , (3.38)
| γ[ω] − i ω |2

where, the Wiener-Khintchine theorem has been applied to both sides. The measured response’s
autocorrelation hp(ω)p(ω 0 )i defines the power spectrum Ip (ω) through Eq.(3.35). Since γ(t) and Ip (ω)
are known, IF (ω) can be determined from Eq.(3.38). Using IF (t) the function κ(t) is then calculated
from Eq.(3.24) and Eq.(3.35), i.e.

κ(t − t0 )
IF (t) = . (3.39)
2 π δ(t + t0 )

For the generalised Langevin model, the relaxation time Eq.(3.22) becomes
1 1
trelax = R ∞ = = µ(ω = 0) , (3.40)
0
dt γ(t) γ[ω = 0]

where Eq.(3.28) and Eq.(3.31) have been used in the above simplification.

In order to gain an intuition regarding the relaxation time for the generalised Langevin model, consider an
example where the memory kernel γ(t) is sharply peaked around t = 0, i.e. the friction is approximated
as being instantaneous. The retarded effect of the friction term in Eq.(3.23) is ignored, and the integral
becomes
Z ∞ Z ∞
0 0 0 1
dt γ(t − t ) p(t ) ≈ dt0 γ(t0 ) p(t) = p(t) , (3.41)
0 0 t relax

where Eq.(3.40) is used in the final equality. Inserting Eq.(3.41) into the generalised Langevin equation
Eq.(3.23), returns the non-retarded Langevin equation Eq.(3.1). Hence the interpretation of trelax remains
the same as for the non-retarded Langevin model: it is the time it takes for the Brownian particle to ther-
malize in the medium [42].

Another relevant time scale, the correlation (or microscopic) time, tc , is defined. The correlation time is
considered to be the width of the random force correlator function. Specifically,
Z ∞
κ(t)
tc = dt . (3.42)
0 κ(0)

The quantity tc measures the time duration of a single scattering process, i.e. it indicates how long the
random force is correlated for. In most cases trelax  tc , however this does not necessarily hold for Brownian
motion dual to AdS black holes.

On a final note for this subsection, the Langevin Model can also be generalised to d spatial dimensions. In
Eq.(3.1), the momentum p(t) and random force F (t) become d-component vectors, while the assumptions
Eq.(3.2) and Eq.(3.3) generalise to

hFi (t)i = 0 and hFi (t) Fj (t0 )i = κ0 δi j δ(t − t0 ) , (3.43)

14
 where i, j = (1, 2, ... , n). The fluctuation-dissipation theorem is independent of d; as such κ0 is still given
by Eq.(3.11). The mean-squared displacement behaves like hs2 (t)i ≈ 2 d D t in the late time limit t  γ10 ,

where the diffusion coefficient D is still given by the Einstein-Sutherland relation Eq.(3.18).

This concludes a brief foray into some of the salient theory about Brownian motion36 . It is this motion
which is responsible for the dissipative nature of a system and its approach to thermal equilibrium. Any
particle suspended in a finite temperature fluid undergoes Brownian motion and as such a probe quark
immersed in a thermal plasma behaves the same way. The AdS/CFT correspondence can be used to study
the Brownian behaviour a probe quark exhibits while interacting with the strongly-coupled thermal plasma
by modelling the external quark as a test string in the bulk theory. This is done in sections (4) and (6) for
the heavy quark case, and section (5) for the light quark case.

Part II
Calculations in the Bulk
4. Analysing Heavy Quark Brownian Motion
In the previous section the theory of Brownian motion (specifically the Langevin model) was reviewed. An
external test quark in a thermal plasma is governed by these equations. In the AdS/CFT context, this is the
gauge theory – or boundary – side. The following three sections focus on the gravitational – or bulk – side,
in which a probe string is set up in an anti-de Sitter black hole background37 and the transverse fluctuations
on this string (resulting from its proximity to the Schwarzschild black hole) are examined. The AdS/CFT
correspondence is used to equate the two descriptions. Section (4) begins this incursion into the bulk theory
by modelling an on-mass-shell quark of large finite mass, identified by particle physicists as an on-mass-shell
heavy quark, as an open probe string stretched between the boundary and a Schwarzschild black hole in AdS
spacetime38 . The presence of the black hole results in thermal fluctuations in the transverse X I directions
on the string. The resultant random movement of the string’s endpoint on the boundary corresponds to the
heavy quark undergoing Brownian motion. This set-up is depicted in figure (2).

Figure 2: A fundamental open string of length `0 used as a probe in an AdS black hole background. The string
starts at the boundary of the anti-de Sitter spacetime and hangs down to a stretched horizon (rs = (1 + ) rH
where 0 <   1) placed just above the Schwarzschild black hole horizon. This figure is adapted from [42].

The stretched horizon depicted in figure (2) is introduced to regulate an infrared divergence. Similarly, a UV

36 For more detailed reading on the topic see [80, 86] who provide insightful early reviews, while [87, 81, 88] can be useful in
understanding some of the more modern developments in the field.
37 As discussed in section (2), the bulk theory is in AdS × S d . However, the physics on the compact S d space corresponds to
d
a set of scalar and fermion fields rotated among each other in N = 4 SYM. In this article, the rotational supersymmetric
Yang-Mills charges are ignored. The calculations can be considered to take place in AdSd and at a point on S d .
38 Throughout this article the test strings considered are in the probe approximation, i.e. the string’s backreaction on the

background is taken to be negligible. Further, it is assumed that no B-field is present in the background.

15
 cut-off is imposed near the boundary to ensure the mass of the probe particle is finite39 . Dirichlet boundary
conditions will be imposed on the fixed string endpoint attached to the stretched horizon, while Neumann
boundary conditions will be imposed at the boundary. Using this setup, we will explicitly compute the
mean distance squared s2 (t) travelled by the endpoint of the string, which represents the position of the
heavy quark in the thermal medium. We will see that at late times s2 (t) is diffusive, growing linearly with
time,

s2 (t) = 2 D t, (4.1)

with a diffusion coefficient that exactly respects Einstein’s fluctuation-dissipation theorem for non-relativistic
motion [42],

T
D= , (4.2)
γm

where the friction coefficient γ will be determined in section (6).

4.1. Polyakov String Equations of Motion

The Brink-Di Vecchia-Howe-Deser-Zumino action [89, 90] – or Polyakov Action [91] for short – describes
the leading order dynamics of the fundamental probe string and is given by
Z Z
1 2 1 p
SP := 0
d σ LP = − 0
d2 σ −det (γab ) γ ab gab , (4.3)
4πα M 4πα
where the slope parameter α0 is related to the string tension T0 [92]; LP is the Lagrangian density; the
worldsheet parameter space is denoted by M (with coordinates (t, σ) ∈ [0, tf ] × [0, σf ] = M) and the
spacetime by N . The induced worldsheet metric is given by

gab := ∂a X µ ∂b X ν Gµν , (4.4)

Gµν is the spacetime metric, and γab is an auxiliary worldsheet metric. From here onwards, the determinant
of γab will simply be denoted by γ. Notice that a gauge choice has been made: the static gauge is used. The
reparameterization offered by the static gauge separates the time and space coordinates [92], identifying the
τ parameter with the time coordinate, τ = t.

The string worldsheet is embedded into the target spacetime by the mapping functions X µ : M →
N ; (t, σ) → X µ (t, σ). The canonical conjugate momentum densities Πaµ (t, σ) are easily determined once
the functional derivative of the Polyakov Action with respect to the derivatives of these embedding func-
tions has been calculated40 . Specifically,
Z
δSP
δ(∂a X µ ) SP := d2 σ δ(∂a X µ (t, σ)) . (4.5)
M δ(∂a X µ (t, σ))

Using the definition of the Polyakov Action Eq.(4.3),

1 √
Z  
δ(∂a X µ ) SP = d2 σ δ(∂a X µ ) − −γ γ ab
G µν ∂b X ν
, (4.6)
M 2πα0

where the factor of two arises from symmetry (the variation of ∂a X µ and ∂b X ν gives the same the re-
sult). Hence the canonical momentum densities, defined as the variation of the action with respect to the

39 Following how this terminology used in [42], the terms infrared (IR) and ultraviolet (UV) are understood to be with respect
to the boundary energy. In the bulk theory infrared means near the horizon, while ultraviolet means near the boundary.
40 The functional derivative (sometimes referred to as the variational or Frécet derivative) compares the change in a functional

to the change in a function that the functional depends on. The functional derivative of functional J with respect to function
f (evaluated at point x) is defined as
Z b
δJ
δf J = δf (x) dx .
a δf (x)

16
 derivatives of the embedding functions, is given by
δ SP 1 √
Πaµ (t, σ) := = − −γ γ ab Gµν ∂b X ν . (4.7)
δ(∂a X µ (t, σ)) 2πα0

The energy-momentum tensor Tab can be defined by the variation of the Polyakov Action with respect to
the auxiliary worldsheet metric [93],
1 δSP
Tab := −4 π √ . (4.8)
−γ δγ ab

Before calculating Tab , note that determining the functional derivative of the auxiliary worldsheet metric
gives the two useful relations41

δ det(γab ) ≡ δγ = γ γ ab δγab = γ γab δγ ab ,

and
√ 1 γ 1√
γab δγ ab = − −γ γab δγ ab .



δ −γ = − √ δγ = − √
2 −γ 2 −γ 2
√

From the definition of the Polyakov Action Eq.(4.3) and the relation for δ −γ , the functional derivative
of the action with respect to the auxiliary worldsheet metric can be simplified to

1
Z  √ √ 
δγ SP = − d2 σ δ( −γ) γ ab Gµν ∂a X µ ∂b X ν + −γ δγ ab Gµν ∂a X µ ∂b X ν
4πα0 M

√
Z  
1 1  
= − d2 σ −γ δγ ab Gµν − γab γ cd ∂c X µ ∂d X ν + ∂a X µ ∂b X ν (4.9)
4πα0 M 2
Z
δSP
≡ d2 σ δγ ab ,
M δγ ab

where the final line is simply the definition of the functional derivative. Hence, the energy-momentum tensor
Tab (Eq.(4.8)), becomes

1 h 1 √  1
cd µ ν µ ν
i
Tab = −4 π √ − −γ G µν − γ ab γ ∂c X ∂d X + ∂a X ∂b X
−γ 4πα0 2
(4.10)
1 1 
= 0 − γab γ cd gcd + gab ,
α 2

where the definition of the induced worldsheet metric (Eq.(4.4)) is used in the last line. Requiring by the
principle of least action that the energy-momentum tensor vanishes, implies
1
gab = γab γ cd gcd . (4.11)
2

Eq.(4.11) sets the auxiliary metric γab proportional to the induced metric gab at every point of the worldsheet.
If the proportionality constant is defined to be positive (the notions of timelike and spacelike vectors defined
by γab and gab should agree); the metrics are said to be conformal to each other. This proportionality
constant is denoted by f 2 . Writing,

41 To prove the first relation make use of the matrix property δ det(A) = det(A) Tr A−1 δA , where A is an n × n matrix.

17
γab = f 2 gab

⇒ det(γab ) = f 4 det(gab )
1 1 1 (4.12)
⇒ (−γ)− 2 = 2
(−g)− 2
f
γab gab
⇒ √ = √ ,
−γ −g

a constraint equation is obtained. From Eq.(4.12) it is evident that the Polyakov action has an additional
symmetry to the Nambu-Goto string action (an action defined in terms of the induced metric gab ) [94, 95],
since the auxiliary worldsheet metric can be re-scaled arbitrarily. Indeed, the Polyakov action is locally
scale invariant on the string worldsheet. This is known as conformal/Weyl invariance: under a local change
of scale all angles are kept fixed on the worldsheet, while the length of the lines may change.

Due to this additional symmetry, another gauge choice presents itself. Choosing the conformal gauge
γ ab = η ab (which restricts the choice of worldsheet to one with vanishing Euler characteristic), the constraint
Eq.(4.12) simplifies to
gab
η ab = √ , (4.13)
−g
since η ab = ηab , and η = det(ηab ) = −1. Hence, with this choice of gauge restricting the auxiliary metric γab ,
the induced worlsheet metric gab becomes conformally flat. In the conformal gauge, the energy-momentum
tensor Eq.(4.10) becomes
1  1
cd µ ν µ ν

Tab = Gµν − η ab η ∂ c X ∂ d X + ∂a X ∂b X . (4.14)
α0 2
The worldsheet parameter space coordinate system can be changed to light-cone coordinates, where the
light-cone coordinates are defined as
1 1
σ ± = √ τ ± σ and ∂± =



∂τ ± ∂σ . (4.15)
2 2

The energy-momentum tensor can be written in terms of light-cone coordinates. This is easily done after
explicitly writing out the components of Tab (Eq.(4.14)),
1 0 0

Tτ τ = Tσσ = 0
Gµν Ẋ µ Ẋ ν + X µ X ν , and
2α
(4.16)
1 0
Tτ σ = Tστ = 0 Gµν Ẋ µ X ν ,
α
0 0 0
where Ẋ µ = ∂τ X µ and X µ = ∂σ X µ . Further, X µ Ẋ ν = Ẋ µ X ν since Gµν is a diagonal matrix. In
light-cone coordinates the components of the energy-momentum tensor are
1 1
T++ = (Tτ τ + Tτ σ ) and T−− = (Tτ τ − Tτ σ ) , (4.17)
2 2

which is easily proven from Eq.(4.14) using the definition Eq.(4.15) and the light-cone metric (where η+− =
η−+ = −1, η++ = η−− = 0). Inserting Eq.(4.16) into Eq.(4.17), yields

1 1 h1 i
∂τ X µ ∂τ X ν + ∂σ X µ ∂σ X ν ± ∂τ X µ ∂σ X ν


T±± = 0
Gµν
2 α 2
(4.18)
1
= 0 Gµν ∂± X µ ∂± X ν ,
α

As previously mentioned, due to energy conservation, the energy-momentum tensor vanishes: Tab = 0. In

18
 light-cone coordinates T++ = T−− = 0. Hence, Eq.(4.18) becomes

Gµν ∂± X µ ∂± X ν = 0 , (4.19)
which are known as the Virasoro constraint equations.

Finally, varying the Polyakov Action with respect to the string worldsheet coordinates and setting this
functional variation to zero, yields the string equations of motion in the static gauge

0 = ∂a Πaµ − Γα ν a a
µν ∂a X Πα =: ∇a Πµ , (4.20)
where the Christoffel symbols are defined as
1 αγ
Γα


µν := G ∂µ Gνγ + ∂ν Gµγ − ∂γ Gµν . (4.21)
2

The details of this derivation are given in appendix (Appendix A). The boundary conditions chosen in
order to satisfy Eq.(4.20) are
σ=σf
δX µ Πaµ σ=0
= 0. (4.22)

This becomes a Dirichlet boundary condition at the stretched horizon (r = rs )42

δX µ (t, 0) = 0 , (4.23)

and a Neumann boundary condition at the boundary endpoint43

Πaµ (t, σf ) = 0 . (4.24)

Eq.(4.20) is the equations of motion describing the leading order string behaviour. Considering additional
behaviour, small transverse fluctuations for example, would results in supplementary equations of motion
for X I (where I denote the transverse directions). This is precisely the focus of subsection (4.3).

The general, leading order solution to the string equations of motion is found by solving Eqs.(4.19, 4.20)
with respect to the boundary conditions Eqs.(4.23, 4.24) and the relevant initial conditions. The metric
Gµν , the initial conditions, and hence the string solution will depend on the geometry of the spacetime the
test string is set up in. For example, in R1,1 the string’s worldsheet parameter space M (with coordinates
(t, σ) ∈ [0, tf ] × [0, σf ] = M) is embedded into the target spacetime N = R1,1 by the function
µ
µ
XMink (t, σ) = t, x(t, σ) , (4.25)

where x(t, σ) still needs to be determined. This is the aim of the following two subsections, where the general
solutions for the test string in an R1,1 and an AdS3 -Schwarzschild background respectively are found44 .

4.2. Leading Order String Behaviour

4.2.1. Test Strings in R1,1
µ
The embedding functions45 , XMink , given in Eq.(4.25) are found by solving the Virasoro constraints and
1,1
string equations of motion in R . For simplicity, a square parameter space is chosen i.e. tf = σf . Working

42 The stretched horizon is defined by rs = (1 + ) rH where 0 <   1. In the limit  → 0 implementing a Neumann boundary
condition here instead of a Dirichlet condition would be equivalent. This is done by de Boer et al. [42].
43 From the bulk perspective, this endpoint terminates on a space filling flavour D7-brane which is introduced in order to ensure

the test quarks in the field theory have a finite mass [54].
44 A note to the reader: a more interesting embedding occurs when studying the off-mass-shell light quark, since the parameter

space is divided into two separate regions. This will be explored in detail in subsection (5.1).
45 Referred to interchangeably throughout this article as the leading order solution to the string equations of motion.

19
 in the static gauge, a Dirichlet Boundary condition at σ = 0 is imposed (corresponding to the fixed initial
position of the string endpoint)

x(t, 0) = x0 ∈ R , (4.26)
and a Neumann Boundary condition at σ = σf is imposed (enforcing zero flux through the other string
endpoint)

∂σ x(t, σ) σ=σf
= 0. (4.27)

For initial conditions, take the string to be static at t = 0,

∂t x(t, σ) t=0
= 0, (4.28)

and stretched between x0 and x0 + `0 in the x spatial direction,

σ
x(0, σ) = x0 + `0 , `0 ∈ R+ . (4.29)
σf

The total energy and momentum of the string are given by

Z Z
τ
E = − dσ Π t and p = dσ Πτx , (4.30)

where Πaµ are the canonical momentum densities defined in Eq.(4.7) [92]. In flat space the total energy of
the static string46 – which would be equal to the mass of a heavy test quark in the boundary theory47 – is
given by
Z σf
1 `0
E = − dσ η τ b ηtν ∂b X ν = − . (4.31)
2πα0 0 2πα0

where it has been recognised that σf = `0 (which follows from the Virasoro constraints in flat space [52]).
The static gauge choice (τ = t), and X 0 = t (Eq.(4.25)) are also used. Therefore,

`20
E2 ≡ m2q = , (4.32)
4π 2 α02

where E 2 = m2 since the string is initially static – i.e. the total momentum vanishes. The length of the
string and the magnitude of the quark’s mass are directly proportional. This same relationship holds for a
test string in an AdS-Schwarzschild background, and indeed Eq.(4.32) remains true. For proof of this see
appendix (Appendix C).

Since the leading order dynamics are that of a static string, the initial condition Eq.(4.29) holds for all
µ
t ∈ R+ . Hence, the embedding functions XMink are somewhat trivially48 given by
µ
µ
XMink (t, σ) = t, x0 + σ , (4.33)

with coordinates (t, σ) ∈ [0, σf ] × [0, σf ] = M.

4.2.2. Test Strings in AdS3 -Schwarzschild

Consider a test string in AdS3 -Schwarzschild set up as in figure (2), page 15. Examining the leading order
behaviour of the string (no transverse fluctuations are present), the equations of motion with respect to the

46 Static in terms of the string’s configuration, not in terms of gauge choice.

47 Supposing for a moment that the gauge/string duality postulated the existence of such a boundary theory to R1,1 spacetime.
48 As (t, σ) sweep out a square region of parameter space (t, σ) ∈ [0, σ ] × [0, σ ] = M, the embedding functions map out a
f f
rectangular region (square if x0 = 0) of target spacetime.

20
 µ
boundary and initial conditions can be solved in order to find the embedding functions XAdS3 -Sch
(t, σ). The
AdSd -Schwarzschild metric in d dimensions is given by49

r2  2
~ I2 + l dr ,
 2
ds2d = 2
−h(r; d) dt2 + dX 2
(4.34)
l r h(r; d)

where t ∈ [0, ∞) is the temporal coordinate, r ∈ [0, ∞) is the radial coordinate, and the transverse spatial
~ I = X 2 , X 3 , ...., X (d−1) ∈ Rd−2 . Further, l ∈ R+ is the curvature radius of


directions are denoted by X
AdSd and S d , and the blackening factor of the Schwarzschild black hole situated at the horizon, h(r; d), is
given by
 d−1
rH
h(r; d) := 1 − . (4.35)
r
The function h(r; d) ∈ [0, 1], where h(r; d) = 0 at the stretched horizon and h(r; d) = 1 at the AdS-
Schwarzschild boundary. The radial position of the black-brane horizon is denoted by rH ∈ R+ . The
Hawking temperature of the black-brane in AdSd -Schwarzschild – corresponding to the temperature of the
thermal plasma in the boundary theory – is

1 (d − 1) rH
T ≡ = . (4.36)
β 4 π l2

In order to consider whether r is a good radial coordinate to use to find the embedding functions, the causal
structure of the spacetime is briefly explored using probe light rays approaching the horizon (r = rH ).
Considering a null geodesic (ds2d = 0) along the radial direction (X ~ I = 0), the AdSd -Schwarzschild metric
Eq.(4.34) becomes
 2 −1
dt r
=± h(r; d) , (4.37)
dr l2

i.e. as a probe light ray approaches the black hole event horizon (r = rH ), dt/dr → ±∞50 . There appears
to be singular behaviour at r = rH (as the event horizon is approached, movement in the radial direction
with respect to the coordinate time t becomes less and less successful), but it’s actually highly dependent on
the chosen coordinate system. For instance, the singular behaviour can be assuaged if the time coordinate
is replaced with a coordinate which moves appropriately slowly along the null geodesic. Defining t := ±r∗
along this null geodesic, Eq.(4.37) can be integrated with respect to r to yield
Z
2 1
r∗ (d) = l dr 2 , (4.38)
r h(r; d)

where r∗ is known as the tortoise coordinate [96]. Using Eq.(4.35) to solve the integral51 yields

l2
  r d−1 
1 d H
r∗ (d) = − 2 F1 1, ; ; , (4.39)
r d−1 d−1 r

where 2 F1 is the Gaussian hypergeometric function52 . Using the tortoise coordinate r∗ as the new radial
coordinate presents an advantage – the horizon can be approached at the relevant rate (dt/dr∗ remains
finite)53 . Hence, to proceed in solving the string equations of motion, the coordinate set (t, r∗ ) is chosen
since it allows the correct boundary conditions for the fluctuations at the black hole horizon to be specified.
The tortoise coordinate, however, can only be inverted for d = 3. Therefore it is only trivial to solve the
string equations of motion to find the embedding functions of the test string in AdS3 -Schwarzschild. For

49 As mentioned in section (2), the anti-de Sitter-Schwarzschild spacetime metric Eq.(4.34) represents a Poincaré chart of AdSd
spacetime. The global AdS-Schwarzschild spacetime yields two black hole solutions at the same temperature: a black hole
with a specific heat which is negative, and a black hole in thermal equilibrium exhibiting Hawking radiation. In order to access
both solutions, global AdS-Schwarzschild geometry with a compact spatial boundary would need to be considered [42] – which
is beyond the current scope of this work.
50 Think of a series of light-cones drawn from each point along the trajectory of the light ray approaching the horizon – as the

light ray approaches r = rH , the light-cones close up [96].

51 The integral Eq.(4.38) was solved using Mathematica [97], and specifically an integration package called Rubi [98].
52 See footnote 65, page 32, for the definition of the Gaussian hypergeometric function.
53 The cost of this coordinate replacement is that the horizon surface is pushed to infinity, i.e. at r = ∞ [96].
∗

21
 d = 3, Eq.(4.39) becomes

1 l2 l2
   
r − rH r
r∗ (3) = ln = coth−1 − , (4.40)
2 rH r + rH rH rH

which can be easily inverted to find r in terms of r∗ (3):

r r 
H ∗
r = −rH coth , (4.41)
l2
where r∗ (3) is given by r∗ for concision. In d = 3 dimensions, the metric Eq.(4.34) is given by54

r2 − rH2
l2 r2
ds23 = − 2
dt2 + 2 2 dr2 + 2 dx2 . (4.42)
l r − rH l

This, incidentally, is also the metric for the non-rotating BTZ black hole [99]. The AdS3 -Schwarzschild
metric Eq.(4.42) can be written in terms of the coordinate set (t, r∗ ) using Eq.(4.41) and its differential,
2
rH r r 
H ∗
dr = 2
csch2 dr∗ . (4.43)
l l2

Inserting Eqs.(4.41, 4.43) into the metric Eq.(4.42)55 , yields

2 2
rH 2 rH r∗
  rH 2 rH r∗
 
ds23 = 2 2
dx2 .


csch −dt + dr∗ + coth (4.44)
l2 l2 l2 l2

From Eq.(4.44), notice that the AdS3 -Schwarzschild metric in the (t, r∗ ) coordinate system is conformally
flat, i.e. the chosen coordinates (t, r∗ ) are a set of isothermal coordinates.

As discussed previously, to regulate an infrared divergence at the horizon, the stretched horizon rs =
(1 + ) rH (where 0 <   1) is introduced. At the stretched horizon, a Dirichlet boundary condition
(comparable to Eq.(4.23)) for the test string in AdS3 -Schwarzschild is implemented
r
XAdS 3 -Sch
(t, 0) = rs , (4.45)
while a Neumann boundary condition (comparable to Eq.(4.24)) is implemented at the boundary endpoint

r
∂σ XAdS3 -Sch
(t, σ) σ=σf
= 0. (4.46)

The boundary conditions Eq.(4.45) and Eq.(4.46) are rewritten in terms of the coordinate set (t, r∗ ). The
Dirichlet condition Eq.(4.45) becomes
r∗
XAdS 3 -Sch
(t, 0) = rs∗ , (4.47)
where rs∗ is defined using Eq.(4.40),

l2
 
rs
rs∗ = coth−1 − , (4.48)
rH rH

and the Neumann condition Eq.(4.46) becomes

2
rH r
H

0 = 2
csch2 2 XAdS
r∗
3 -Sch
r∗
(t, σ) ∂σ XAdS 3 -Sch
(t, σ) σ=σf
l l (4.49)
r∗
= ∂σ XAdS3 -Sch
(t, σ) σ=σ ,
f

54 In AdS -Schwarzschild there

is only one transverse direction, refered to as x.
3
55 Mathematica is used to help
simplify the trigonometry. This computation (and others in this section) are explicitly shown in
Mathematica Notebook [b] ([Link]). For access, see appendix (Appendix G).

22
 where Eq.(4.43) is used in the first line, and the last line follows since csch(x) 6= 0 , ∀ x ∈ R. The boundary
conditions in the conformally flat description of AdS3 -Schwarzschild (Eqs.(4.47, 4.49)) are respectively
analogous to the boundary conditions for the test string in R1+1 , Eqs.(4.26, 4.27). Hence, the embedding
functions for the test string in the conformally flat description of AdS3 -Schwarzschild are the direct analogue
of Eq.(4.33). In (t, r∗ ) coordinates, the embedding functions are
µ
µ
XAdS 3 -Sch
(t, σ) = t, rs∗ + σ , 0 , (4.50)

with coordinates (t, σ) ∈ [0, σf ] × [0, σf ] = M. Using the inverse tortoise transformation Eq.(4.41), the
embedding functions Eq.(4.50) can be written in terms of the original (t, r) coordinates
 r  µ
µ H
XAdS 3 -Sch
(t, σ) = t, −rH coth (rs∗ + σ) + σ , 0 , (4.51)
l2

where the position of the fixed string endpoint attached to the stretched horizon rs∗ is given by Eq.(4.48)
and the length of the string in tortoise coordinates is given by

l2
 
rs + `0
σf = coth−1 − − rs∗ . (4.52)
rH rH

Making the identification

 
rH (rs∗ + σ)
r = −rH coth , (4.53)
l2

leads to agreement between Eq.(4.51), and the stretched string of de Boer et al.’s [42, 46] calculations.
Finally, notice that comparing Eq.(4.41) with Eq.(4.53), yields r∗ ≡ rs∗ + σ.

4.3. Fluctuating Test String Dynamics

In this subsection transverse fluctuations of the test string in AdS3 -Schwarzschild are studied56 , in order to
explore (using the AdS/CF T correspondence) the heavy quark’s Brownian motion induced by the thermal
noise in the plasma. The fluctuations are chosen to be small so as to not affect the leading order radial
solution (discovered in subsection (4.2.2)). de Boer et al. [42] outlays a semiclassical treatment of the test
string’s transverse motion in AdS3 -Schwarzschild by quantizing these transverse fluctuations and relating
the behaviour of the modes on the string to the dynamics of the boundary endpoint.

The following subsection (4.3.1) derives the equations of motion for small transverse fluctuations on the test
string in AdSd -Schwarzschild. Following the method first laid out in [42], subsection (4.3.2) solves these
equations of motion to find the general solution, and from there the mean-squared displacement of the test
string’s boundary endpoint, s2 (t).

4.3.1. Nambu-Goto String Equations of Motion for Transverse Fluctuations

The standard Nambu-Goto action57 can be used to describe the dynamics of the transverse fluctuations on
the test string
√
Z Z
1 2 1
SN G := 0
d σ LN G = − 0
d2 σ −g , (4.54)
2πα M 2πα M

where LN G is the Lagrangian density, g := det (gab ) and the induced worldsheet metric gab is given by
Eq.(4.4).

Take the embedding functions (given in AdS3 -Schwarzschild by Eq.(4.51)) to be the leading order static
string solution in a (1 + 1)-dimensional subspace of spacetime spanned by the temporal and radial directions

56 Later, in subsection (4.4), this will be generalised to AdSd -Schwarzschild.

57 Using Eq.(4.12) the equivalence between the Polyakov action Eq.(4.3) and the Nambu-Goto action Eq.(4.54) is apparent.

23
(t, r), denoted by X0µ (t, σ) (where X0µ : [0, tf ]×[0, σf ] → Rd−1,1 ; (t, σ) → X0µ (t, σ)). Consider the addition of
transverse fluctuations X I (where I ∈ (2, 3, ... , d − 1)) to X0µ (t, σ). The effective action for the transverse
fluctuations appears as a correction to the leading order Nambu-Goto action. It follows that to find the
effective action for the transverse fluctuations (and from this the equations of motion) the Nambu-Goto
action needs to be expanded about X0µ (t, σ), in terms of X I ,
Z
1 2 (2) (4)

(6)

SN G = d σ LN G + SN G + S N G + O S N G , (4.55)
2πα0 M X0µ

where
Z Z
(2) 1 (2) (4) 1 (4)
SN G := d2 σ LN G and SN G := d2 σ LN G . (4.56)
2πα0 M 2πα0 M

In order to explicitly calculate the expansion in Eq.(4.55), the determinant of the induced worldsheet metric
is needed. This is derived in appendix (Appendix C) and is given below

r02
 
1
g := det (gab ) = Grr Gtt r02 + GII ẊI2 + GII XI0 2 , (4.57)
Gtt Grr
0

where Ẋ = ∂τ X and X = ∂σ X; I indexes over the transverse directions XI = X 2 , X 3 , ...., X (d−1) ;
and the AdSd -Schwarzschild spacetime metric Gµν is the given explicitly by Eq.(C.2). Using Eq.(4.57) the
Nambu-Goto action Eq.(4.54) becomes

Z r
1 2 02

12 1 1
SN G =− d σ −r Grr Gtt 1+ GII ẊI2 + 02 GII XI0 2
2πα0 M Gtt r Grr

√ √
Z Z  
1 1 GII 2 GII
=− d2 σ −g X0µ
− d2 σ −g X0µ
ẊI + 02 X0 2
2πα0 M 4πα0 M Gtt r Grr I (4.58)
X0µ

2
√
Z 
1 GII 2 GII
+ 2
d σ −g X0µ
Ẋ + 02 X0 2 ,
16πα0 M Gtt I r Grr I
X0µ

2
where a Taylor expansion up to quadratic order (i.e. (1 + x)1/2 ≈ 1 + x2 − x8 ) was first performed. Then it
is noticed from Eq.(4.57) that to leading order the determinant of the induced worldsheet metric is simply
given by

g X0µ
= r02 Grr Gtt . (4.59)

The quadratic correction in the expansion of the Nambu-Goto action (defined in Eq.(4.56)) is the effective
action for the transverse fluctuations. Matching terms between Eq.(4.55) and Eq.(4.58), the quadratic
correction is given by

√
Z  
(2) 1 GII 2 GII
SN G = − d2 σ −g X0µ
ẊI + 02 X0 2
4πα0 M Gtt r Grr I X0µ
√  (4.60)
−g
Z
1 2 I J
=− d σ GIJ ∂a X ∂a X ,
4πα0 M gab X0µ

where a, b indexes over the static gauge worldsheet coordinates (t, σ), the spacetime metric (Eq.(C.2)) is
recognised as a diagonal matrix, and to leading order the induced worldsheet metric (see appendix (Appendix
C)) is given by
   
Gtt + GII Ẋ 2 GII ẊI XI0 Gtt 0
gab X µ =   =  . (4.61)
0
GII XI0 ẊI r02 Grr + GII XI0 2 X µ 0 r02 Grr
0

24
 (2)
Hence, the effective action for the transverse fluctuations SN G is given by

√
Z
(2) 1
d2 σ −g g ab GIJ X µ ∂a X I ∂b X J


SN G = − 0
4πα M 0

(4.62)
∂ 2 LN G
Z
1 2
≡ d σ ∂a X I ∂b X J ,
4πα0 M



∂ ∂a X I ∂ ∂b X J X0µ

where I, J index over the transverse directions (2, 3, ... , d − 1). Comparing Eq.(4.62) with Eq.(4.3), notice
that if the auxiliary worldsheet metric is chosen to be the leading order induced worldsheet metric (γab =
gab |X0µ ) the action for transverse fluctuations Eq.(4.62) can be understood as an effective Polyakov Action.
(4)
It remains to be proven whether or not the quartic term SN G is significant and contributes to the effective
(4)
action of the transverse fluctuations. The quartic term SN G (defined in Eq.(4.56)) is given by
1
Z  √ 2
(4)
d2 σ −gg ab GIJ X µ ∂a X I ∂b X J .


SN G = 0
(4.63)
16πα M 0

Comparing the quadratic action term Eq.(4.62) with the quartic action term Eq.(4.63), yields

(2) 1  (4)  21
SN G = − √ SN G . (4.64)
πα0
(4)
Hence the quartic correction term SN G will significantly contribute
√ to the effective action for the transverse
fluctuations when the test string is within a distance of α0 from the black-brane horizon. Therefore the
(2)
quadratic action SN G (Eq.(4.62)) can be considered to solely contribute to the effective action for the
transverse
√ fluctuations, as long as the fluctuations are small and the test string is in the region further than
α0 away from the black-brane horizon, where
√
α0 = λ−1/4 l , (4.65)

√
in AdSd -Schwarzschild. Eq.(4.65) defines the fundamental string length scale α0 is terms of the radius of
curvature of AdSd spacetime l and the ’t Hooft coupling λ.

The canonical momentum densities conjugate to the transverse coordinates X I are defined as
(2)
δ SN G 1 √
ΠaI := −g g ab GIJ ∂b X J ,


I
= − 0 X0µ
(4.66)
δ(∂a X ) 2πα

where the factor of two arises from symmetry (the variation of ∂a X I and ∂b X J gives the same the result).
(2)
Varying the effective action for the transverse fluctuations SN G with respect to the transverse string world-
sheet coordinates X I and setting this functional variation to zero, yields the equations of motion

0 = ∂a ΠaI = ∇a ΠaI . (4.67)

These equations of motion are analogous to Eq.(4.20); but in Eq.(4.67) the Christoffel symbols (defined in
Eq.(4.21)) are identically zero (Γα
IJ = 0) due to the symmetry between each of the transverse directions. The
details of this derivation are given in appendix (Appendix D). The chosen boundary conditions (analogous
to Eq.(4.22)) are
σ=σf
ΠaI δX I σ=0
= 0, (4.68)

which ensures the correct boundary terms vanish in the derivation of the string equations of motion for the

25
transverse fluctuations (Eq.(4.67)).

4.3.2. Calculating the Heavy Quark’s Mean-Squared Displacement s2 (t) in AdS3 -Schwarzschild
In AdS3 -Schwarzschild, the string boundary endpoint’s mean-squared transverse displacement is given by
Z ∞
β2 dω 1
2
fω (σf )fω∗ (σf ) 2 − e−iωt − eiωt .


s (t) = √ βω
4π λ 0 ω e − 1
2

A detailed derivation of this equation is given in this subsection.

The leading order static solution to the string equations of motion is given by Eq.(4.51) in AdS3 -Schwarzschild.
Following subsection (4.3.1), this solution is renamed X0µ (t, σ). Adding non-zero fluctuations in the trans-
verse x-direction, the test string solution becomes

µ  r  µ
µ µ H
XAdS 3 -Sch
(t, σ) = X0 (t, σ) + 0, 0, X(t, σ) = t, −rH coth (rs∗ + σ) , X(t, σ) . (4.69)
l2

In AdS3 -Schwarzschild, the equations of motion for the transverse fluctuations, Eq.(4.67), become
!
√ 2
 
ab r
0 = ∂a −g g ∂b X(t, σ) , (4.70)
l2 X µ
0

where the definition transverse momentum densities Eq.(4.66) are used in the first line, and the spacetime
metric Eq.(C.2) in the second line. To leading order the induced worldsheet metric is given by Eq.(4.61).
For d = 3, this metric simplifies to
     2 2  
r −r
gtt gtσ Gtt 0 − l2 H 0
gab X µ :=   =   =   2  , (4.71)
0
gσt gσσ X µ 0 r02 Grr 0 r02 r2 −rl
2
0 H

0
where Eq.(4.42) is used, and r = ∂σ r. Hence, its inverse is given by

     2
 
l
g tt g tσ 1 r02 Grr 0 − r2 −r 2 0
g ab X0µ
:=   =
  =  H  , (4.72)
r 2 −rH2

g σt g σσ det gab |X0µ 0 Gtt 0 1
X0µ r 02 l2

where det gab |X0µ is given in Eq.(4.59). The indices in Eq.(4.70) can be expanded
! !
√ r2 √ r2
   
σσ tt
0 = ∂σ −g g ∂σ X(t, σ) + ∂t −g g 2 ∂t X(t, σ)
l2 X0µ l X0µ
(4.73)
r2 − rH 2
 
1 1 2 2
= − ∂t2 X(t, σ) + 2


4 2
∂σ r r − rH ∂σ X(t, σ) ,
l r ∂σ r ∂σ r

where Eqs.(4.59, 4.72) are used in the second line. The transverse equations of motion can be fully converted
into spacetime variables (t, r). The derivative operator ∂σ first needs to be converted into ∂r using Eq.(4.53).
The differential of Eq.(4.53) is
2
 
rH 2 rH (rs∗ + σ)
∂σ = 2 csch ∂r
l l2
(4.74)
r2 − rH
2
= ∂r .
l2

Using Eq.(4.74), the transverse equations of motion Eq.(4.73) become

r2 − rH 2
0 = − ∂t2 X(t, r) + ∂r r2 r2 − rH
2



∂r X(t, r) . (4.75)
l4 r 2

26
 Further, the transverse equations of motion can be completely written in terms of worldsheet parameter
coordinates (t, σ) using Eqs.(4.53, 4.74), which yields
1  r
H
 
0 = − ∂t2 X(t, σ) + 2 rH

∂σ coth2 2 (rs∗ + σ) ∂σ X(t, σ) . (4.76)
coth l2 (rs∗ + σ) l

To proceed in solving the linear and homogeneous partial differential equation Eq.(4.75) – or Eq.(4.76) in
(t, σ) coordinates – the method laid out in de Boer et al. [42] is followed. Explicitly, consider a mode
expansion in X,
X(t, r) = fω (r) e−iωt , (4.77)

where X(t, r) is a separable eigenmode solution to Eq.(4.75) which oscillates in time with a well-defined
angular frequency ω ∈ R+ . Inserting Eq.(4.77) into Eq.(4.75), it can been seen that fω (r) satisfies the
ordinary differential equation

r2 − rH 2
0 = ω 2 fω (r) + ∂r r2 r2 − rH
2



∂r fω (r) . (4.78)
l4 r 2

Defining the dimensionless quantity

l2 ω
ν := , (4.79)
rH
the ordinary differential equation (Eq.(4.78)) can be rewritten as

r2 (r2 − r2 )
 
ν 2 + H 8 2 H ∂r r2 r2 − rH 2



∂r fω (r) = 0 . (4.80)
l r

This second order ODE has two linearly independent solutions58

 ±iν/2
1 r ± irH ν r − rH
fω(±) (r) = . (4.81)
1 ± iν r r + rH

In solving Eq.(4.80) the initial condition near the horizon is chosen to be

I do not see ini cond followed  ±iν/2
r − rH
fω(±) (r) → as r → rH , (4.82)
r + rH

i.e. the normalization at the black-brane horizon (r = rH ) is taken to be 1 ±1 iν r ± ir

r
Hν
= 1. For convenience,
the solutions Eq.(4.81) can be partially rewritten in terms of the tortoise coordinate r∗ . Rearranging
Eq.(4.53) yields

l2
 
r
rs∗ + σ = coth−1 −
rH rH
 ±iν/2 (4.83)
±iω (rs∗ +σ) r − rH
⇒ e = ,
r + rH

where the definition for ν, Eq.(4.79), is used in the second line. Inputting Eq.(4.83) into Eq.(4.81) gives
1 r ± irH ν ±iω (rs∗ + σ)
fω(±) (r) = e , (4.84)
1 ± iν r

58 This is consistent with Eq.(2.36) in de Boer et al.’s paper [42].

27
 where the initial condition near the horizon Eq.(4.82) becomes

fω(±) (r) → e±iω (rs∗ + σ) as r → rH . (4.85)

(±)
From Eq.(4.85) it is apparent that fω (r) denotes out-going (+) and in-falling (−) basis modes59 . The
general solution to the ODE, fω (r), is a linear combination of these modes

fω (r) = fω(+) (r) + Bω fω(−) (r) , (4.86)

where the constant Bω measures the difference in phase between the (+) and (−) modes, and is yet to be
determined. Imposing a Neumann boundary condition in the radial direction at the boundary endpoint,

0 = ∂r fω (r) r=rs +`o

1 r + irH ν

r − rH
iν/2
1 r − irH ν

r − rH
−iν/2 ! (4.87)
= ∂r + Bω ,
1 + iν r r + rH 1 − iν r r + rH
r=rs +`o

enables Bω to be fixed. Note that Eqs.(4.81, 4.86) have been used in the second line. Hence, Eq.(4.87) is
solved60 to find Bω :

   iν
i+ν r ν − irH r − rH
Bω =
−i + ν r ν + irH r + rH
r=rs +`o
(4.88)
   iν
1 − iν 1 + ir̃0 ν r̃0 − 1
= ,
1 + iν 1 − ir̃0 ν r̃0 + 1

where the dimensionless quantity

rs + `0
r̃0 := , (4.89)
rH

has been defined in the second line. The coefficient Bω can be partially rewritten in terms of the tortoise
coordinate r∗ . To this end Eq.(4.52) is rearranged,

l2
 
rs + `0
rs∗ + σf = coth−1 −
rH rH
 iν (4.90)
r̃0 − 1
⇒ ei2 ω (rs∗ +σf ) = ,
r̃0 + 1

where the definition for ν, Eq.(4.79), and Eq.(4.89) is used in the final line. Inputting Eq.(4.90) into
Eq.(4.88), yields
  
1 − iν 1 + ir̃0 ν
Bω = ei2 ω (rs∗ + σf ) . (4.91)
1 + iν 1 − ir̃0 ν

The Eqs.(4.84, 4.86, 4.91) determine fω (r) completely. The linear superposition of fω (r) for all frequencies

59 The mode fω+ (r) is outgoing at the horizon; while fω− (r) is a mode which is reflected at the boundary and – after experiencing
a phase shift – falls back towards the horizon.
60 Mathematica is used (see Mathematica Notebook [b]: [Link]).

28
ω ∈ R+ , yields the general solution for X(t, r),
Z ∞
dω
Aω fω (r) e−iωt aω + fω∗ (r) eiωt aω∗ ,
 
X(t, r) := (4.92)
0 2π

where the mode expansion Eq.(4.77) is used, and (aω , aω∗ ) are Fourier coefficients. The constant Aω is fixed
by demanding the normalization of the appropriate basis.

To quantize the theory, the scalar field X(t, σ) and its canonically conjugate momentum P t (t, σ) are pro-
moted to operators and suitable commutation relations are imposed. From Eq.(4.92), the position operator
is given by
Z ∞
dω
Aω fω (σ) e−iωt âω + fω∗ (σ) eiωt â†ω ,
 
X̂(t, σ) := (4.93)
0 2π

and from Eq.(4.66) the conjugate momentum operator is given by

1 √
P̂ t (t, σ) := − −g g tt GII


0 X0µ
∂t X̂(t, σ)
2πα
Z ∞
1 √ dω
tt
(−i ω) Aω fω (σ) e−iωt âω − fω∗ (σ) eiωt âω∗

 
=− −g g GII X µ
2πα0 0
0 2π
2
 Z ∞ (4.94)
i rH rH (rs∗ + σ) dω
2
ω Aω fω (σ) e−iωt âω − fω∗ (σ) eiωt âω†
 
=− 0 2
coth 2
2πα l l 0 2π
∞
i r2
Z
dω
ω Aω fω (σ) e−iωt âω − fω∗ (σ) eiωt âω† ,
 
=−
2πα0 l2 0 2π

where Eq.(4.93) is used in the second line, r∗ ≡ rs∗ + σ is used in the fourth line, and the identification
Eq.(4.53) in the final line. The metric entries in the fourth line follow from using the identification Eq.(4.53)
to convert the spacetime metric as well as the induced metric, its determinant and its inverse into (t, r∗ )
coordinates. From the spacetime metric Eq.(C.2),

r2
 
2 rH (rs∗ + σ)
GII = GII X µ = H coth , (4.95)
0 l2 l2

while the induced metric, its determinant and its inverse (Eqs.(4.71, 4.59, 4.72)) are given by
   2   
r
Gtt 0 − lH2 csch2 rH (rls∗2 + σ) 0
gab ≡ gab X µ =   =  2
  , (4.96)
rH 2 rH (rs∗ + σ)
0
0 r02 Grr 0 l2 csch l2

r4
 
rH (rs∗ + σ)
= − H csch4



g := det gab ≡ det gab | X0µ , (4.97)
l4 l2

  2   
1 r02 Grr 0 − rl2 sinh2 rH (rls∗2 + σ) 0
g ab ≡ g ab X0µ
=
  =  H   ,
det gab |X0µ l2 rH (rs∗ + σ)
0 Gtt 0 2
rH
sinh2 l2
(4.98)

in the preferred isothermal set of coordinates (t, r∗ ). In order to fix the normalization constant Aω canonical
commutation relations between X̂(t, σ) and P̂ t (t, σ) are enforced:

29
 h i δ(σ − σ 0 )
X̂(t, σ), nt P̂ t (t, σ 0 ) = i δ(σ, σ 0 ) = i √ ,
Σ g̃|Σ
(4.99)
h i h i
0
X̂(t, σ), X̂(t, σ ) = 0 = nt P̂ t (t, σ), nt P̂ t (t, σ 0 ) ,
Σ Σ

where Σ is a Cauchy hypersurface in the xµ = (t, r)µ part of spacetime that is chosen to be a constant√ time
surface61 , g̃ is the induced metric on Σ, and nµ is the future pointing normal to Σ (where nµ = δµt / −g̃tt ).

Additionally, canonical creation and annihilation commutation relations on the Fourier coefficient operators
(âω , â†ω ) are enforced:

h i
âω , â†ω0 = 2πδ(ω − ω 0 ) ,
Σ
h i h i (4.100)
âω , âω0 = 0 = â†ω , â†ω0 .
Σ Σ

Demanding consistency between the sets of commutation relations Eq.(4.99) and Eq.(4.100), requires the
normalization constant to be defined as
r
l πα0 β
Aω := = √ , (4.101)
rH ω 2 πω λ1/4

where the second equality follows from using the definition of the AdS radius of curvature l (Eq.(4.65)), and
2
the relation l = βrH /2π from Eq.(4.36). The derivation of Eq.(4.101) is the subject of appendix (Appendix
E).

With the constants Bω and Aω (Eqs.(4.91, 4.101) respectively) found, the general solution for transverse
equations of motion (Eq.(4.92)) is completely determined. The displacement of the boundary endpoint as
the test string undergoes these transverse fluctuations can now be calculated. The position of the endpoint
of the test string on the boundary is defined by the operator

X̂end (t) := X̂(t, σf ) . (4.102)

At the Hawking temperature, the transverse fluctuations on the string are excited. Assume – as was done
by de Boer et al. [42] – that these excitations are purely thermal and are therefore described by the
Bose-Einstein distribution
2π δ(ω − ω 0 )
hâ†ω âω0 i = . (4.103)
eβω − 1

The Bose-Einstein distribution applies when (i) quantum effects are important, (ii) particles are indistin-
guishable, and (iii) are bosons (particles which do not obey the Pauli exclusion principle). The string
boundary endpoint’s mean-squared transverse displacement, s2 (t), is defined as62

2
s2 (t) := h: X̂end (t) − X̂end (0) :i = h:X̂end
2 2
(t):i + h:X̂end (0):i − 2h:X̂end (t)X̂end (0):i . (4.104)

In order to determine s2 (t), begin by calculating the expectation value of the position of the boundary
endpoint at two different times. Using Eq.(4.93),

61 Giving initial conditions on this hypersurface determines the future (and past) evolution uniquely.
62 The mean-squared transverse displacement, s2 (t), is a gauge and coordinate independent quantity, i.e. it doesn’t matter which
coordinate system s2 (t) is calculated in. Note that the position of the horizon is still dependent on the choice of coordinates
(as discussed in subsection (5.1.3) when motivating for the use of (t, r∗ ) coordinates).

30
h:X̂end (t1 )X̂end (t2 ):i
Z ∞
β2 dω dω 0 1 h ∗ −iωt1 +iω 0 t2 † ∗ iωt1 −iω 0 t2 †
i
= √ √ f (σ )f
ω f ω 0 (σ f )e h:â â
ω ω 0 :i + f ω (σ )f
f ω 0 (σ f )e h:âω ωâ 0 :i
4π λ 0 (2π)2 ωω 0
Z ∞ 0 0
β2 dω dω 0 1 h e−iωt1 +iω t2 eiωt1 −iω t2 i
= √ √ δ(ω 0 − ω)fω (σf )fω∗0 (σf ) βω0 + δ(ω − ω 0 )fω∗ (σf )fω0 (σf ) βω
4π λ 0 2π ω ω0 e −1 e −1
∞
β2
Z
dω 1 1 h i
= √ βω
fω (σf )fω∗ (σf )e−iω(t1 −t2 ) + fω∗ (σf )fω (σf )eiω(t1 −t2 ) ,
4π λ 0 2π ω e − 1
(4.105)

where :âω† âω0 : = :âω0 âω† : = âω† âω0 is the normal ordering operator, and the definition of the Bose-Einstein
distribution (Eq.(4.103))
R∞ is used in the third equality. The final line follows from the property of the Dirac
delta function 0 dk e−kx δ(k − a) = e−ax . Using the complex conjugate property Re(z) = (z + z̄)/2,
Eq.(4.105) simplifies to
Z ∞
β2 dω 1 
∗ −iω(t1 −t2 )

h:X̂end (t1 )X̂end (t2 ):i = √ Re f ω (σ f )f ω (σ f )e . (4.106)
4π 2 λ 0 ω eβω − 1

The normal ordering operator is necessary to remove the logarithmic ultraviolet divergences. Had it not
been used, extra terms would arise in the calculation of Eq.(4.105). Specifically,

hX̂end (t1 )X̂end (t2 )i

Z ∞
β2 dω dω 0 1 h ∗ −iωt1 +iω 0 t2

† 0

= √ √ f (σ
ω f ω )f 0 (σ f )e hâ ω ω
0 â i + 2π δ(ω − ω )
4π λ 0 (2π)2 ωω 0
0
i
+ fω∗ (σf )fω0 (σf )eiωt1 −iω t2 hâω† âω0 i

∞ 0
β2 dω dω 0 1 e−iωt1 +iω t2
Z 
0
= √ √ δ(ω 0 − ω)fω (σf )fω∗0 (σf ) βω0 + δ(ω − ω 0 )fω (σf )fω∗0 (σf )e−iωt1 +iω t2
4π λ 0 2π ωω 0 e −1
0
eiωt1 −iω t2

+ δ(ω − ω 0 )fω∗ (σf )fω0 (σf ) βω
e −1
Z ∞
β2
 
dω 1 
∗ −iω(t1 −t2 )
 1
∗ −iω(t1 −t2 )
= √ Re f (σ
ω f ω)f (σf )e + f (σ )f
ω f ω (σ f )e ,
4π 2 λ 0 ω eβω − 1 2
(4.107)

where the relation hâω âω†0 i = hâω†0 âω i + 2π δ(ω − ω 0 ) results from the commutation relation Eq.(4.100).
Following [42], the term 21 fω (σf )fω∗ (σf )e−iω(t1 −t2 ) is interpreted as a logarithmic UV divergence. Since
this divergence stems from the zero-point energy and exists at zero temperature, it can be regularised by
instituting normal ordering of the oscillators aω , âω† (as was done in Eq.(4.105)). At specific times, the
regularized correlator (Eq.(4.105)) becomes
Z ∞
β2 dω 1 
∗ −iωt

h:X̂end (t)X̂end (0):i = √ Re fω (σ f )fω (σ f )e , (4.108)
4 π 2 λ 0 ω eβω − 1
Z ∞
β2 dω 1  
h:X̂end (0)X̂end (0):i = h:X̂end (t)X̂end (t):i = √ βω
Re fω (σf )fω∗ (σf ) . (4.109)
4 π2 λ 0 ω e − 1

Inputting Eqs.(4.108, 4.109) into Eq.(4.104), the string boundary endpoint’s mean-squared transverse dis-

31
 placement can be calculated
Z ∞
β2 dω 1 
∗ ∗ −iωt

s2 (t) = √ Re 2f ω (σ f )f ω (σ f ) − 2f ω (σ f )f ω (σ f )e
4π 2 λ 0 ω eβω − 1
Z ∞ (4.110)
β2 dω 1  ∗ −iωt iωt


= √ fω (σ f )f ω (σ f ) 2 − e − e ,
4π 2 λ 0 ω eβω − 1
where in the second line the complex conjugate property Re(z) = (z + z̄)/2 is again used.

4.3.3. The Limiting Cases of s2 (t)

In this subsection the behaviour of s2 (t), Eq.(4.110), in the asymptotic early and late time limits is consid-
ered. Inputting63 Eqs.(4.84, 4.86, 4.91) to calculate fω (σf )fω∗ (σf ), yields
∞ 2 πtν
4β 2 dν 1 + ν 2 sin β
Z
s2 (t) = √ , (4.111)
π2 λ 0 ν 1 + r̃02 ν 2 e2πν − 1

where the change of variables was performed using Eqs.(4.36, 4.79). The integral in Eq.(4.111) is tricky to
analytically evaluate. Following the insights of de Boer et al. [42], the integral can be broken into two parts,

sin2 πνt 2 πνt

" ! !#
∞ ∞
4β 2 r̃2 − 1 dν sin β
Z Z
dν 1 β 1
s (t) = √ 4 0 2
2
+4 2
π2 λ r̃0 0 ν 1 + r̃02 ν 2 e2πν − 1 r̃0 0 ν e2πν − 1
(4.112)
β2 r̃02 − 1
 
1
= √ I1 + I2 ,
π2 λ r̃02 r̃02

where the integrals I1 and I2 have been defined as,

Z ∞
dν 1 sin2 πνt
β
Z ∞
dx 1 sin2 kx
2
I1 := 4 2 2 2πν
= 4 2 2 x−1
, (4.113)
0 ν 1 + r̃0 ν e −1 0 x 1 + a x e
and
2 πνt
∞
dx sin2 kx
Z ∞
dν sin β
Z
2
I2 := 4 = 4 . (4.114)
0 ν e2πν − 1 0 x ex − 1

The change of variables in Eqs.(4.113, 4.114) is made by defining the new variables
r̃0 t
x := 2πν , a := , k := . (4.115)
2π β
The integral I1 is solved by deforming the contour on the complex x plane. The solution is given by [42],

           
1 1 1 1 k k −ak k π |k|
− a
 1
I1 = ψ 1+ +ψ 1− + e Ei −
a + e Ei − 1−e cot
2 2πa 2πa 2 a a 2 2a

!
1 1
; e−2π|k| 1 1 −2π|k|


e−2π|k| 2 F1 1, 1 + 2πa ; 2 + 2πa 2 F1 1, 1 − 2πa ; 2 − 2πa ; e
+ 1 + 1
2 2πa + 1 1 − 2πa
 
2a sinh(πk)
+ ln ,
k
(4.116)

where Ei(z) is the exponential integral64 ; 2 F1 (a, b; c; z) is the Gaussian hypergeometric function65 ; and

63 Mathematica is used to simplify the algebra (see Mathematica Notebook [b]: [Link]).
R∞
64 Forreal, non-zero values of z, the exponential integral is defined as Ei(z) = − −z dt (e−t /t).
65 The hypergeometric function is a special function defined for |z| < 1 by the hypergeometric series,
2 F1 (a, b; c; z) =

32
 ψ(z) is the digamma function66 . The integral I2 can be performed analytically,

∞ Z ∞  
X dx kx
I2 = 4 sin2 e−nx
n=1 0 x 2

∞  !
Y k2 (4.117)
= ln 1+ 2
n=1
n
 
sinh (πk)
= ln ,
πk

P∞
where, in the first line, the geometric series 1/(1 − x) = n=0 xn is used; and in the second line the integral
was performed using Mathematica [97].

From Eq.(4.117) it appears that β = t/k naturally defines a cross-over time scale between the dynamics
of early and late times67 . The integrals I1 and I2 can be examined in the limits (t  β) and (t  β) to
yield the early and late time behaviour respectively. Since the probe particle in the boundary theory is an
on-mass-shell heavy quark, the assumption r̃0  1 (equivalently a  1) can be made. This is essentially
the statement that the large mass of the external particle translates to the distance in the radial direction
between the test string’s boundary endpoint and the stretched horizon being large68 .

(I) Asymptotic Early Time Dynamics

In the early time limit t  β (equivalently k  1  a), the integrals I1 and I2 can be simplified. Defining
a temporary variable f := 1/a, and series expanding the solution to I1 (Eq.(4.116)) around f = 0 yields

   
2 sinh(kπ) 1 2
− ln 1 − e−2kπ + ln(f ) + ln k πf + O f 2



I1 = − kπ +
f 2
(4.118)
k2 π
 
1
= +O ,
2a a2

where, in the second line, the expression is rewritten in terms of the variable a and appropriately truncated
at leading order.

For the I2 integral, notice there is no a-dependence – hence a series expansion in 1/a yields
I2 = O a0 . Therefore, in the early time limit, s2 (t) (Eq.(4.112)) becomes

β2
 2
r̃0 − 1 π 2 t2 t2
     
1 1 0 1
s2 (t) |(tβ) √ + O + O (r̃0 ) = √ + O , (4.119)
π2 λ r̃02 β 2 r̃0 r̃02 r̃02 r̃0 λ r̃02

where the variables have been changed using Eq.(4.115). In terms of the relevant field theoretic quantities,
Eq.(4.119) can be converted using Eq.(C.13). Specifically, s2 (t) = (T t2 )/mq , where mq is the mass of the
probe quark and T is the temperature of the thermal plasma. Since s2 (t) ∼ t2 , the early time dynamics
exhibit ballistic behaviour (see Eq.(3.20))69 .

(II) Asymptotic Late Time Dynamics

At asymptotically late times, t  β (equivalently k  a  1), the integrals I1 and I2 have the same

P∞ (a)k (b)k z k
k=0 (c)k k!
.
66 The 1 dΓ(z)
digamma function is defined in terms of derivatives of the gamma function: ψ(z) := Γ(z) dz
. The function is meromor-
phic, and defined on the complex numbers C.
67 More specifically, the cross-over time (defined in Eq.(3.42)) is given by t ∼ β r̃ ∼ (α0 m )/(l2 T 2 ), where Eqs.(4.65, C.13)
c 0 q
have been used in the second approximation.
68 The assumption r̃ → ∞ would correspond to an infinitely massive quark, such as the probe particle considered in section (6).
0
69 The result for s2 (t) in the early time limit given in Eq.(4.119) agrees exactly with de Boer et al. [42], Eq.(3.6) (after recognising

r̃0 = ρc , and making use of Eq.(4.65)).

33
 x0
behaviour. To prove this, take the integral I1 (Eq.(4.113)) and make the change of variables x := k,
0 0
∞
dx0 sin2 x2 ∞
dx0 sin2 x2
Z Z
1 (βt)
I1 = 4 2 0 0 −→ 4 0 ≡ I2 , (4.120)
0 x0 1 + ak2x e xk − 1 0 x0 e xk − 1

a2 x0
where it is noticed that the term k2  1 can be neglected in the late time limit. Expanding the solution
to I2 (Eq.(4.117)), yields

I1 |βt ≡ I2 = ln sinh πk − ln πk

= ln eπk − ln 2 − ln πk




(4.121)

= πk + O(ln k) ,

where the second line follows from trigonometric identity sinh(x) = (ex − e−x )/2, and in the third line the
term e−πk /2 → 0 since k  1 at late times.

Therefore, in the late time limit, s2 (t) (Eq.(4.112)) becomes

β2 r̃02 − 1
        
πt t 1 πt t βt t
s2 (t) |(tβ) √ + O ln + 2 + O ln = √ + O ln , (4.122)
π2 λ r̃02 β β r̃0 β β π λ β

where the variables have been changed using Eq.(4.115). Since s2 (t) ∼ t, the late time dynamics exhibit
diffusive behaviour70 . Concretely, at late times s2 (t) = 2 D t, where D is the diffusion coefficient (Eq.(3.21)).
Comparing this to Eq.(4.122), the diffusion coefficient in AdS3 -Schwarzschild is extricated,
β
D = √ . (4.123)
2π λ
As will be shown in section (6), the drag, or friction, experienced by a slowly moving heavy quark in AdS3
— which is identical to the dissipation of momentum down the string into the black hole horizon — is given
by √
dp 2 π λT 2
= −γ p , and γ = (4.124)
dt m
One can see that the fluctuation-dissipation theorem, which connects the drag and diffusion of a probe
particle in a thermal medium [42], is exactly respected:

T
D= . (4.125)
γm

4.4. Generalising to AdSd -Schwarzschild

For the off-mass-shell light quark case – as will be seen in subsection (5.3) – generalising the diffusion
coefficient to AdSd -Schwarzschild71 relies on the observation that the behaviour of an arbitrary virtuality
light quark at asymptotically late times is encoded in the small virtuality light quark case. Since s2 (t)
in the latter case can easily be calculated for any dimensions d ≥ 3 in AdSd -Schwarzschild, the diffusion
AdSd
coefficient DLQ can be determined. Unfortunately, this method can not be applied to the on-mass-shell
72
heavy quark case . The avenue left is to determine, in AdSd -Schwarzschild, the friction coefficient of
the medium plasma and relate this – via the Einstein-Sutherland equation (Eq.(3.18)) – to the diffusion
coefficient. This is precisely what is done in section (6), where the friction coefficient in AdSd -Schwarzschild
is found by first calculating the drag force on an infinitely massive, heavy quark moving through the thermal

70 The result for s2 (t) in the late time limit given in Eq.(4.122) agrees exactly with de Boer et al. [42], Eq.(3.6) (where Eqs.(4.36,
4.65) have been used to show the equivalence).
71 Remember that a general d dimensions in the bulk theory corresponds to a (d − 1)-dimensional thermal plasma in the boundary

theory.
72 To talk about the small virtuality/small mass limit of a massive, heavy probe quark, would be nonsensical.

34
 plasma with a constant velocity73 in the bulk theory. As will be seen, the diffusion coefficient extracted
from this method74 , agrees exactly with the the diffusion coefficient given in Eq.(4.123) for d = 3.

5. Analysing Light Quark Brownian Motion

A light quark probe in a thermal medium on the boundary starts at t = 0 as an off-mass-shell particle,
radiates energy as it travels through the thermal medium and finally stops radiating as it becomes an on-
mass-shell particle. Dual to this light quark in the boundary theory, a test string in the bulk can be thought
of as half the yo-yo string in R1,1 . The latter is a string which, at a time t = 0, stretches a length of 2L
along the x-axis; as time progresses the string shrinks till, at a time t = L, it is a point; then it begins to
expand again and the process repeats. The yo-yo solution is a well-studied classical solution of the open
string [100]. There is plenty of recent work [101, 102, 103, 104, 105, 15] modelling gluons and light quarks
as half the initial yo-yo string contracting to a point after a time t, in order to calculate energy loss (see
also [106, 52] for other works using this string set-up). This idea of a falling string is precisely what is used
in order to study the off-mass-shell light quark in this article. Such a string set-up is refered to, following
the convention of [52], as the Limp Noodle.

If the test string is placed in an AdS-Schwarzschild background, the presence of the black hole results in
thermal fluctuations in the transverse X I directions on the string. The resultant random movement of the
string endpoint on the boundary corresponds to the light quark undergoing Brownian motion. This set-up
is depicted in figure (3).

Figure 3: A fundamental falling open string of length `0 used as a probe in an AdS black hole background.
Initially, at t = ti , the string starts at the boundary of the anti-de Sitter spacetime and hangs down to a stretched
horizon (rs = (1 + ) rH where 0 <   1) placed just above the Schwarzschild black hole horizon. As time evolves
ti < t2 < t3 , the string endpoint at the boundary is released and the string collapses. Its trajectory as it falls (grey,
dashed line) is predicted by [101, 102, 103, 104, 105, 15].

An initial static set-up of the Limp Noodle such that one of its endpoints is held fixed at the boundary
for an asymptotically long time, allowing the string to be fully populated by thermal modes emanating
from the black hole horizon, is then equivalent to the leading order static stretched string for the heavy
quark described in subsection (4.2). The two situations diverge once the boundary endpoint of the string is
released, and the endpoint starts to fall at the local speed of light. Information, however, is also restricted to
travel at the local speed of light. Hence, all parts of the string below the falling boundary endpoint can not
know whether the boundary endpoint has remained fixed (as in the case of the heavy quark’s test string)
or been released to fall until the endpoint crashes through it. Intuitively, it is therefore expected that all
parts of the string except the falling boundary endpoint are described by a solution (embedding functions)
identical to Eq.(4.33) in R1,1 and Eq.(4.51) in AdS3 -Schwarzschild, while the stretched endpoint falls at the
local speed of light. This will be confirmed in the following subsection.

5.1. Leading Order String Behaviour

The basics of string theory discussed in subsection (4.1) – leading to the Virasoro constraints (Eq.(4.19)) and
the string equations of motion (Eq.(4.20)) – hold for the light quark’s test string (Limp Noodle). In order

73 Implies that the probe quark is under the influence of an external force.
74 These calculations predominantly follow the work of [22, 35].

35
to solve Eqs.(4.19, 4.20) (with respect to the boundary conditions Eqs.(4.23, 4.24) and the relevant initial
conditions) to yield the embedding functions requires the correct partitioning of the worldsheet parameter
space. This partitioning was trivial in the heavy quark’s test string case, since its leading order behaviour
is static for all t. It is however more complicated in the Limp Noodle case, considering that the boundary
string endpoint is allowed to fall at the local speed of light for t > 0. A method in which to partition the
worldsheet in order to find the solution to the string equations of motion was given by Itzhak Bars in 1994
[107], and expounded upon a year later, in 1995, by himself and a collaborator Jürgen Schulze [108]. The
Bars et al. method is summarised in the following subsection.

5.1.1. Parameterising the Limp Noodle: The Bars et al. method

The Bars et al. method can be used to solve the classical motion of a relativistic string in any (1 + 1)-
dimensional spacetime. Originally, the method was used by Bars et al. to study the motion of folded strings
in curved spacetime. However the Limp Noodle can be interpreted as half a folded (or yo-yo) string. The
application of the Bars et al. method to parameterise the Limp Noodle and find the general solution to the
string equations of motion was first proposed by Moerman et al. in [52]. The Bars et al. method can be
broken into the following steps:

(i) Define lattice-like patches on the worldsheet using the light-cone coordinates.

Figure 4: Lattice structure imposed on the worldsheet parameter space using the light-cone coordinates σ + and
σ − . This structure is used in the Bars et al. method [107, 108] in order to determine the general solution for a
moving open string on a manifold.

(ii) Naive solutions to the string equations of motion come in four classes, referred to as: A, B, C, and
D. They can be defined only on patches of the worldsheet, since the full solution needs to satisfy
two conditions: (a) string solutions must be periodic in σ, and (b) the global time coordinate must
be an increasing function of τ and σ (in order to exclude anti-strings from appearing in the same
solutions as strings). In the static gauge the latter condition is trivially satisfied. Hence to build the
general solution from the solution classes A, B, C, D, assign the pattern of solutions to ensure forward
propagation and impose periodicity at a fixed value of τ .
(iii) Complete solutions are obtained by matching boundary conditions at the boundary of the patches.

(iv) Define a transfer operation matrix to determine future evolution. The propagation of the solution into
the future is performed by using this matrix to matching the boundary conditions while increasing τ .

The leading order dynamics of the Limp Noodle are constrained to the (1 + 1)-dimensional subspace of
AdSd -Schwarzschild described by the set of coordinates (x0 , x1 ), where x0 denotes the temporal direction
and x1 denotes the radial direction. The metric of any (1 + 1)-dimensional subspace (where one of these
dimensions is the temporal dimension) can be transformed into a conformally flat metric [107, 108]. In
order to perform this transformation the string field in the target spacetime is redefined by the coordinate
0
reparametrization xµ (τ, σ) = xµ (y µ (τ, σ)). This yields the new spacetime metric in terms of isothermal
0
coordinates y µ (τ, σ), which can be related to the old spacetime metric by
0 0
∂y µ ∂y ν
Gµν (x) = Gµ0 ν 0 (y) , (5.1)
∂xµ ∂xν

36
 0
where the isothermal coordinates in the static gauge y µ (t, σ), are given by
0 1   0 1  
y 0 = √ x0 + x1 and y 1 = √ x0 − x1 . (5.2)
2 2

In this isothermal coordinate frame the metric is conformally flat

0 0
ds2 = Gµ0 ν 0 (y) dy µ dy ν , (5.3)

and the new spacetime metric is found by inverting Eq.(5.1)

∂xµ ∂xν ∂xµ ∂xν
Gµ0 ν 0 (y) = 0 0 Gµν (x) = G ηµν , (5.4)
µ
∂y ∂y ν ∂y µ0 ∂y ν 0

where the second equality follows from choosing the conformal gauge, and G = G(y 0 , y 1 ) ∈ R+ is some non-
zero coordinate dependent scalar function. The metric can be calculated component-wise. For example,

∂x0 ∂x0 ∂x1 ∂x1

     1   1 
1  1 
G00 00 (y) = G − 00 + = G − √ √ + √ √ = 0, (5.5)
∂y ∂y 00 ∂y 00 ∂y 00 2 2 2 2

where the definition of the isothermal coordinates Eq.(5.2) have been used. Similarly G00 10 (y) = −G,
G10 00 (y) = −G, and G10 10 (y) = 0. Hence Gµ0 ν 0 (y) has the explicit form
 
0 −1
Gµ0 ν 0 (y) = G   = G η̃µ0 ν 0 , (5.6)
−1 0

0 0
where η̃µν is the light-cone metric, and G := G(y 0 , y 1 ) ∈ R+ is defined as some non-zero coordinate
dependent scalar function. Using Eq.(5.6), the metric Eq.(5.3) becomes
0 0 0 0

ds2 = −G dy 0 dy 1 + dy 1 dy 0 . (5.7)

0
The string mapping functions Y µ (σ + , σ − ) are given by
0 1   0 1  
Y 0 = √ X 0 + X 1 and Y 1 = √ X 0 − X 1 . (5.8)
2 2

Note that since the coordinates are isothermal, Y 0 = −Y1 and Y 1 = −Y0 . The canonically conjugate
0
momentum densities Eq.(4.7) can be rewritten in terms of the isothermal embedding functions, Y µ . In the
conformal gauge, the canonical momentum densities are given by

1
Πaµ (t, σ) = − η ab Gµν ∂b X ν
2πα0
(5.9)
1
⇒ Π±
µ (t, σ) = G ηµν ∂∓ X ν ,
2πα0

∂
if the light-cone coordinate frame is chosen (where ∂∓ := ∂σ ∓ ). Specifically,
1 1
Π±
0 (t, σ) = − G ∂∓ X 0 and Π±
1 (t, σ) = G ∂∓ X 1 . (5.10)
2πα0 2πα0
0
Hence the conjugate momentum densities can be defined in terms of the string mapping functions Y µ (σ + , σ − )

37
 1  ±  1 0
Π±
00 (t, σ) := √ Π 0 + Π±
1 ≡ − 0
G ∂∓ Y 1 , and
2 2 πα
(5.11)
1   1 0
Π±
10 (t, σ) := √ Π± ±
0 − Π1 ≡ − 0
G ∂∓ Y 0 .
2 2 πα

Combining the components in Eq.(5.11), renaming the indices µ0 = µ, and recognising the raising and
lowering property of the isothermal coordinates, leaves
1
Π±
µ (t, σ) = G ∂∓ Y(1−µ) . (5.12)
2 πα0

The Virasoro constraint equations and the string equations of motion, Eqs.(4.19, 4.20), can also be rewritten
0
in terms the new string mapping functions Y µ (σ + , σ − ):

∂± Y 0 ∂± Y 1 = 0 ,

G∂− Y 1 + ∂− G∂+ Y 1 = (∂0 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0 ,






∂+ (5.13)

G∂− Y 0 + ∂− G∂+ Y 0 = (∂1 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0 ,






∂+

where ∂µ := ∂Y∂ µ for µ = (0, 1). Since, the derivation of Eq.(5.13) is somewhat tedious, it has been relegated
to appendix (Appendix B).

There are four classes of solutions to the set of equations Eqs.(5.13). Namely,

A: Y 0 (σ + , σ − ) = YA0 (σ + ) , Y 1 (σ + , σ − ) = YA1 (σ − )

B: Y 0 (σ + , σ − ) = YB0 (σ − ) , Y 1 (σ + , σ − ) = YB1 (σ + )
(5.14)
Y 0 (σ + , σ − ) = YC0 = c1 , Y 1 (σ + , σ − ) = fC αC (σ + ) + βC (σ − ); YC0


C:

Y 0 (σ + , σ − ) = fD αD (σ + ) + βD (σ − ); YD0 , Y 1 (σ + , σ − ) = YD1 = c2 ,


D:

where c1,2 are constants, and YA0 (σ + ), YA1 (σ − ), YB0 (σ − ), YB1 (σ + ), αC (σ + ), βC (σ − ), αD (σ + ), and βD (σ − )
are arbitrary functions [107, 108]. The string solutions A, B, C and D given in Eq.(5.14) may be verified
by direct substitution into Eqs.(5.13). This is easy for classes A and B, since these classes exist for any
metric. Taking the derivative with respect to the light-cone coordinates σ± , yields

A : ∂− YA0 (σ + ) = 0 , ∂+ YA1 (σ − ) = 0



(5.15)
B : ∂+ YB0 (σ + ) = 0 , ∂− YB1 (σ − ) = 0 ,

which is used, together with the chain rule, to easily show that the A and B classes of string solutions
satisfy the Virasoro constraints and the string equations of motion Eqs.(5.13). This is harder to prove for
classes C and D. These classes are metric dependent solutions since the functions fC and fD are metric
dependent. These functions are calculated by inverting the following relations that depend on the target
spacetime metric,
Z
C : Y 0 (σ + , σ − ) = YC0 , ds G YC0 , s = αC (σ + ) + βC (σ − )


fC
Z (5.16)
1 + − −
YD1 , s, YD1 +


D: Y (σ , σ ) = ds G = αD (σ ) + βD (σ ) .
fD

Taking the derivatives of Eq.(5.16) with respect to the light-cone coordinates σ± and making use of Eq.(5.14),

38
 ∂± Y 0 (σ + , σ − ) = 0, G ∂± Y 1 (σ + , σ − ) = ∂± αC (σ + ) + βC (σ − )


C:
(5.17)
∂± Y 1 (σ + , σ − ) = 0, G ∂± Y 0 (σ + , σ − ) = ∂± αD (σ + ) + βD (σ − ) .


D:

Using Eq.(5.17), the C and D classes of string solutions can be easily proven to satisfy the Virasoro con-
straints and string equations of motion Eqs.(5.13) by direct substitution. For the C class, the Virasoro
constraints and second equation of motion are trivially satisfied. It remains to be checked whether the first
equation of motion is satisfied. To this end, Eq.(5.17) can be substituted into the first equation of motion
in Eq.(5.13)

0 = ∂+ G∂− Y 1 + ∂− G∂+ Y 1 = ∂+ ∂− βC (σ − ) + ∂− ∂+ αC (σ + ) = 0 .





(5.18)

Similarly for class D, the Virasoro constraints and the first equation of motion are trivially satisfied, while
the second equation of motion is verified by substituting Eq.(5.17) into Eq.(5.13).

It is not immediately obvious that the classes of solutions presented in Eq.(5.14) form a complete set. Com-
pleteness was shown in Bars et al. [108] by using the curved spacetime approach based on G/H gauged
WZW models. Note that the pattern of solutions defined on the patches of the worldsheet is metric inde-
pendent and spacetime dimension independent [107]. Hence, in order to find the general string solution for
the Limp Noodle in AdSd -Schwarzschild, start by finding the pattern of solutions defined on the patches of
the worldsheet for the Limp Noodle in flat space. This is done in the next subsection.

Prior to this, some intuition regarding the patch-parametrization of the Limp Noodle can be developed.
At a fold in a string or a string endpoint, the determinant of the induced worldsheet metric gab (Eq.(4.4))
vanishes. Hence the string endpoint’s trajectory in the target spacetime is a null geodesic (the point travels
at the local speed of light). For classes C and D – by virtue of either Y 0 or Y 1 reducing to a constant
over the patch of parameter space where the class is a solution – the induced worldsheet metric reduces
to zero, gab = 0. From this it can be deduced that all points in C or D parameter space patches map to
the trajectory of the string endpoint (or fold) in target spacetime. This is a many-to-one mapping: the
string’s endpoint in target spacetime is represented several times on the worldsheet [107]. The string (sans
the endpoints) behaves like a massive state, hence it is expected that motion of the string is described by
classes A and B – ever-present, metric independent solutions.

5.1.2. Test Strings in R1,1

In order to study the Limp Noodle in R1,1 the static gauge ansatz is chosen

X µ : [0, tf ] × [0, σf ] → R1,1 , (5.19)

(t, σ) → xµ (t, σ) = (t, x(t, σ))µ . (5.20)

The initial test string set-up is the same as in the heavy quark case. Eqs.(4.26, 4.27) are chosen as the
boundary conditions for the position of the fixed endpoint at the horizon and the boundary endpoint falling
in the x(t, σ) direction respectively. For initial conditions Eqs.(4.28, 4.29) are chosen i.e. at t = 0 the string
is static and stretches between x0 and x0 + `0 . Set tf to be the minimum time it takes for a string to shrink
from its initial length `0 to a point x0 .

In flat space the total energy of the string, which represents the mass of an initially off-mass-shell light
quark, is given by Eq.(4.32)75 . The equation of motion for x(t, σ) can be easily adapted from Eqs.(B.8,
B.14), since the spacetime is R1+1 (therefore G = 1 and (∂0 G) = (∂1 G) = 0). Hence, the non-trivial string
equation of motion reduces to

∂+ ∂− X 1 = 0 −→ ∂+ ∂− x(t, σ) = 0 in R1+1 , (5.21)

75 The derivation of the total energy of the test string in an AdS-Schwarzschild background is given in appendix (Appendix C).

39
which is simply the wave equation. The general solution to Eq.(5.21) is a sum of continuous analytic
functions of σ + and σ − (i.e. left and right movers)
1
f1 (σ + ) + f2 (σ − ) .


x(t, σ) = (5.22)
2

Using the boundary condition Eq.(4.26) to redefine f (z) := f1 (z) = 2 x0 − f2 (z), Eq.(5.22) becomes
1
f (σ + ) − f (σ − ) .


x(t, σ) = x0 + (5.23)
2

The Neumann boundary condition Eq.(4.27) becomes

1 
∂+ f (σ + ) + ∂− f (σ − )

0 = σ=σf
, (5.24)
2

which follows from the fact that x0 ∈ R and the definition Eq.(4.15), since ∂σ f (σ + ) = (∂+ − ∂− ) f (σ + ) =
∂+ f (σ + ). To simplify Eq.(5.24) let z := t − σf , then

σ − |σ=σf = t − σf = z and σ + |σ=σf = t + σf = z + 2 σf . (5.25)

d
Using Eq.(5.25) and defining F (z) := f (z), Eq.(5.24) becomes
dz

F (z + 2σf ) = −F (z) , (5.26)

which shows that F (z) is anti-periodic with an interval of 2 σf . The Virasoro constraint equations Eq.(4.19)
can also be rewritten in terms of F (z)

0 = − ∂± t ∂± t + ∂± x(t, σ) ∂± x(t, σ)
h  1
i2 (5.27)
f (σ + ) − f (σ − )



= − ∂± t ∂± t + ∂± x0 + .
2

The first line follows from Eq.(5.20), and Eq.(5.23) was used in the final line. Eq.(5.27) breaks up into two,
(+) and (−), constraint equations. For the (+) equation, Eq.(5.27) becomes

    2
1 1 1 +
0 = − + ∂+ f (σ )
2 2 2 (5.28)
⇒ F (σ + )2 = 1 .

Similarly, for the (−) equation F (σ − )2 = 1. Hence the Virasoro constraint equations become

F (σ ± )2 = 1 , (5.29)

which implies F (σ − )2 |σ=σf = 1 or equivalently F (z)2 = 1. What remains to be determined is the sign of
F (z). This is given by the anti-periodicity condition Eq.(5.26) which results in
 

z+σf
F (z) = − 1 2σf , (5.30)

where bxc is the floor operation which returns the largest integer less than or equal to x. The function F (z),
plotted in figure (5), is a step function. The definition of F (z) as the derivative of f (z) with respect to z,

40
implies that f (z) can be found through integrating Eq.(5.30). But F (z) is a discrete function – performing
this integration involves integrating each segment of F (z) separately. Since it is already known that F (z)
is alternating (+1) or (−1) in each 2 σf interval, and the integral of a constant is a straight monotonic
function; all that remains to be determined is if the function f (z) is monotonically increasing or decreasing
on each interval. Hence f (z) can be defined by
 

z+σ f
f (z) = − 1 2σf
([(z + σf ) mod 2σf ] − σf ) + σf , (5.31)

 z+σf 
where − 1 2σf defines the domain of each segment to be integrated, and ([(z + σf ) mod 2σf ] − σf )

yields 0 if z ∈ z1 σf (where z1 is an even integer) and −σf if z ∈ z2 σf (where z2 is an odd integer). The
function f (z) is a triangular wave function and is plotted in figure (5).

(i) The function F(z) (ii) The function f(z)

1 2 σf

σf
-5 σf -3 σf -σf σf 3 σf 5 σf

-1
-5 σf -3 σf -σf σf 3 σf 5 σf

Figure 5: The functions (i) F (z) and (ii) f (z) over the range [−6σf , 6σf ], given by Eq.(5.30) and Eq.(5.31)
respectively.

Now that a specific form of the function f (z) has been obtained, x(t, σ) (Eq.(5.23)) is completely known.
For example, at t = σf , Eq.(5.23) becomes
1
x(σf , σ) = x0 + (f (σf + σ) − f (σf − σ)) = x0 , (5.32)
2

where the second equality follows from using the definition Eq.(5.31). In studying the Limp Noodle, the
primary interest is in the behaviour of the string as it shrinks from a length of `0 to a single point x0 . From
Eq.(5.32) this clearly is achieved in a t = σf amount of time; hence the choice tf = σf is made. The
worldsheet parameter space is therefore given by M = [0, σf ] × [0, σf ], i.e. the parameter space is square.

Given the insight of [107, 108, 106] the parameter space M can be minimally partitioned into two exclusive
triangular regions


M1 := (t, σ) ∈ M|σ ∈ [0, σf − t]
 (5.33)
M2 := (t, σ) ∈ M|σ ∈ (σf − t, σf ] .

In order to satisfy the boundary and initial conditions two classes of solutions exist, one in each of the two
parameter space regions: the upper triangular region (M2 ) and the lower triangular region (M1 ). The
embedding functions of the Limp Noodle in R1+1 is therefore given by
  µ
µ
 σ, if (t, σ) ∈ M 1

XMink (t, σ) = t, x0 +  . (5.34)
σ − t, if (t, σ) ∈ M2 
f

The Limp Noodle embedding functions satisfy the given boundary and initial conditions. The boundary
conditions can be checked first: for σ = 0, (t, σ) ∈ M1 and the Dirichlet condition Eq.(4.26) is satisfied;

41
 while for σ = σf , (t, σ) ∈ M2 and the Neumann condition Eq.(4.27) is satisfied. The initial conditions
are also fulfilled. For t = 0, (t, σ) ∈ M1 and the initial condition Eq.(4.28) is satisfied. Further the initial
condition Eq.(4.29) is satisfied, since σf = `0 is a requirement of the Virasoro constraints in flat space.
µ
From Eq.(5.34), observe that the functions XMink (t, σ) map the parameter space M1 region to an extended,
space-like region in the target spacetime; while the parameter space M2 region is mapped to a null geodesic
in the target spacetime. Null geodesics describe the trajectories of massless ‘point-like’ objects, and as such
the worldsheet is considered to be wrapped up to a point along this geodesic. Hence, the string solution
on M1 describes the entire string sans the falling endpoint, while the string solution on M2 describes the
falling endpoint. This is evident from figure (6)76 .

Figure 6: Left: The worldsheet parameter space, where M1 is the blue region and M2 is the red region
µ
(Eq.(5.33)). Right: The embedding functions XMink (t, σ), Eq.(5.34), where M1 is mapped to a space-like region
(blue region) and M2 is mapped to a null geodesic (red, dashed line) in the target spacetime.

µ µ
XMink (M2 ) lies on the boundary of XMink (M1 ), and momentum flows between these two regions in the
target spacetime. In order to continue to satisfy the Neumann boundary condition Eq.(4.27), momentum
flows out of the M1 region at σ = σf − t, but immediately stops at M2 since any momentum flowing out
of the string endpoint would change the physics in the boundary theory77 .

If Eq.(5.34) is converted into isothermal coordinates in terms of light-cone parameter space (σ + , σ − ),

µ
YMink (σ + , σ − ) defined on the region M1 is a Bars et al. [107, 108] class A solution78 and the solution
defined on the region M2 is a class C solution, as expected. To see this, use the definition Eq.(5.8) to define
the isothermal embedding functions on M1 ,

0 1  
Y 0 = √ t + (x0 + σ)
2
 
1 1 + −

1 + −


= √ √ σ +σ + x0 + √ σ − σ (5.35)
2 2 2
1 0
= σ + + √ x0 =: Y 0 σ + ,


2

76 Details regarding figures (6)-(8) can be found in Mathematica Notebook [a] ([Link]) – see appendix
(Appendix G) for access.
77 An infinite force is required to stop the flow of momentum abruptly at the M /M divide, and yet the string has a finite
1 2
energy density (since the string has a finite tension). This discrepancy stems from the fact that strings are fundamentally
one-dimensional objects and are not made up of d = 0 constituent objects.
78 See subsection (5.1.1) for details.

42
where the definition of the light-cone coordinates (Eq.(4.15)) is used in the second line. Similarly,

0 1 0
Y 1 = σ − − √ x0 =: Y 1 σ − .


(5.36)
2

Comparing Eqs.(5.35, 5.36) against the four classes of Bars et al. solutions (Eq.(5.14)), it is obvious that
µ
the isothermal embedding functions YMink (σ + , σ − ) defined on the region M1 are a class A solution. This
is depicted in figure (7).

The isothermal embedding functions on M2 are given by

0 1   1 0
Y 0 = √ t + (x0 + σf − t) = √ (x0 + σf ) =: YC0 , (5.37)
2 2

and

0 1   0
Y 1 = √ t − (x0 + σf − t) = − x0 + σf − σ + + σ − =: Y 1 σ + , σ − .




(5.38)
2

Matching Eqs.(5.37, 5.38) with four classes of Bars et al. solutions (Eq.(5.14)) it is easily apparent that the
µ
embedding functions YMink (σ + , σ − ) defined on the region M2 are a class C solution. Figure (7) displays
this graphically.

Figure 7: Left: The worldsheet parameter space in light-cone coordinates (σ + , σ − ), where M1 is the blue region
and M2 is the red region. The parameter space regions M1 and M2 are plotted from Eq.(5.33) which is converted
µ
to (σ + , σ − ) coordinates using Eq.(4.15). Right: The isothermal embedding functions YMink (σ + , σ − ), Eqs.(5.35,
5.36, 5.37, 5.38), where M1 is mapped to a space-like region (blue region) and M2 is mapped to a null geodesic
(red, dashed line) in the target spacetime.

5.1.3. Test Strings in AdS3 -Schwarzschild

In the previous subsection, the general embedding functions for the Limp Noodle in R1+1 , Eq.(5.34), were
discovered. From this the embedding functions for the Limp Noodle in AdS3 -Schwarzschild can be found
using the same minimal partitioning of the parameter space and the determined pattern of solution classes

43
on the patches of the parameter space.

In order to trivially write down the general solution for the Limp Noodle in AdS3 -Schwarzschild a set of
isothermal coordinates is needed. As was shown in subsection (4.2.2), the tortoise coordinate r∗ together
with the temporal coordinate t forms such a set of coordinates in AdSd -Schwarzschild.

The Dirichlet and Neumann conditions, Eqs.(4.45, 4.46), are chosen as boundary conditions for the Limp
Noodle in AdS3 -Schwarzschild. The boundary conditions can be rewritten in terms of the isothermal (t, r∗ )
coordinate set. In subsection (4.2.2), it was noticed that these boundary conditions in the conformally flat
description of AdS3 -Schwarzschild, Eqs.(4.47, 4.49), are respectively analogous to the boundary conditions
for the Limp Noodle (and the heavy quark test string) in R1+1 , Eqs.(4.26, 4.27). Hence, the embedding
functions of the Limp Noodle in the conformally flat description of AdS3 -Schwarzschild are the direct
analogue of Eq.(5.34). In (t, r∗ ) coordinates, the embedding functions are
   µ
µ
 σ, if (t, σ) ∈ M1 
XAdS 3 -Sch
(t, σ) = t, rs∗ + , 0 , (5.39)
σ − t, if (t, σ) ∈ M 
f 2

where the M1 and M2 parameter space regions are given by Eq.(5.33). Using the inverse tortoise trans-
formation Eq.(4.41), the embedding functions for the AdS3 -Schwarzschild Limp Noodle (Eq.(5.39)) can be
rewritten in (t, r) coordinates
   µ
 −r coth rH (r + σ)
, if (t, σ) ∈ M 
µ H l2 s∗ 1
XAdS 3 -Sch
(t, σ) = t, , 0 , (5.40)
−r coth rH (r + σ − t)
, if (t, σ) ∈ M 
H l2 s∗ f 2

where the position of the fixed string endpoint attached to the stretched horizon rs∗ is given by Eq.(4.48)
and the length of the string in tortoise coordinates is given by Eq.(4.52). Notice that the identification
Eq.(4.53) still holds. The embedding functions are plotted in figure (8).

Figure 8: Left: The worldsheet parameter space, where M1 is the blue region and M2 is the red region
µ
(Eq.(5.33)). Right: The embedding functions XAdS3-Sch (t, σ), Eq.(5.40), where M1 is mapped to a space-like
region (blue region) and M2 is mapped to a null geodesic (red, dashed line) in the target spacetime.

5.2. Fluctuating Test String Dynamics

As discussed previously, the boundary endpoint of the Limp Noodle falls at the local speed of light – the
same speed at which information propagates. Hence, any part of the string below the falling endpoint does
not know if the string is stretched like an on-mass-shell quark or its endpoint is falling like an off-mass-shell
quark. Therefore, the transverse fluctuations of all parts of the Limp Noodle except the falling endpoint

44
 are described by the same transverse equations of motion as those of the on-mass-shell heavy quark’s test
string and its solution (discussed in subsections (4.3.1) and (4.3.2)).

The leading order solution to the string equations of motion – renamed X0µ (t, σ) – is given by Eq.(5.40) in
AdS3 -Schwarzschild. Adding non-zero fluctuations in the transverse x-direction, the Limp Noodle solution
becomes

µ  r  µ
µ H
XAdS3 -Sch
(t, σ) = X0µ (t, σ) + 0, 0, X(t, σ) = t, −rH coth (rs∗ + σ) , X(t, σ) . (5.41)
M1 l2

The transverse string equations of motion are supplied in Eq.(4.76); and its solution is given by Eq.(4.93)
in AdS3 -Schwarzschild. Continuity of the string solution in spacetime means that it is unnecessary to find
the explicit string solution on the M2 parameter space region which maps onto the falling string endpoint.

5.2.1. Calculating the Light Quark’s Mean-Squared Displacement s2 (t) in AdS3 -Schwarzschild
Unlike in the case of the heavy quark’s test string where the boundary endpoint is fixed in the radial
direction (subsection (4.3.2)); for the Limp Noodle the position of the boundary endpoint is falling at the
local speed of light and is defined by the operator

X̂end (t) := X̂(t, σf − t) . (5.42)

As in subsection (4.3.2), in order to determine s2 (t) – the falling string endpoint’s mean-squared transverse
displacement defined in Eq.(4.104) – begin by calculating the expectation value of the position of the free
endpoint at two different times. Using Eq.(4.93),

∞
β2 dω dω 0 1 h
Z
0
h:X̂end (t1 )X̂end (t2 ):i = √ 2
√ fω (σf − t1 )fω∗0 (σf − t2 )e−iωt1 +iω t2 h:âω âω†0 :i
4π λ 0 (2π) ωω 0

0
i
+ fω∗ (σf − t1 )fω0 (σf − t2 )eiωt1 −iω t2 h:âω† âω0 :i

∞ 0
β2 dω dω 0 e−iωt1 +iω t2
Z 
1
= √ √ δ(ω 0 − ω)fω (σf − t1 )fω∗0 (σf − t2 ) βω0 (5.43)
4π λ 0 2π ω ω0 e −1
0
eiωt1 −iω t2

+ δ(ω − ω 0 )fω∗ (σf − t1 )fω0 (σf − t2 ) βω
e −1
∞
β2
Z
dω 1 
∗ −iω(t1 −t2 )

= √ Re fω (σ f − t1 )f ω (σ f − t 2 )e ,
4π 2 λ 0 ω eβω − 1

where :âω† âω0 : = :âω0 âω† : = âω† âω0 is the normal ordering operator79 , and the definition of the Bose-Einstein
distribution (Eq.(4.103)) R ∞ is used in the third equality. The final equality follows from the property of the
Dirac delta function 0 dk e−kx δ(k − a) = e−ax ; and the complex conjugate property Re(z) = (z + z̄)/2
has also been used. At specific times, the regularized correlator Eq.(5.43) becomes
Z ∞
β2 dω 1
Re fω (σf − t)fω∗ (σf )e−iωt ,


h:X̂end (t)X̂end (0):i = √ βω
(5.44)
4 π2 λ 0 ω e − 1

∞
β2
Z
dω 1
Re fω (σf )fω∗ (σf ) ,


h:X̂end (0)X̂end (0):i = √ (5.45)
4 π2 λ 0 ω eβω − 1

∞
β2
Z
dω 1
Re fω (σf − t)fω∗ (σf − t) .


h:X̂end (t)X̂end (t):i = √ (5.46)
4 π2 λ 0 ω eβω − 1

79 Normal ordering of the annihilation and creation operators âω , âω† is used to remove logarithmic UV divergences.

45
 Inputting Eqs.(5.44, 5.45, 5.46) into the definition Eq.(4.104), the string falling endpoint’s mean-squared
transverse displacement can be calculated as

∞
β2
Z
dω 1
s2 (t) = Re fω (σf − t)fω∗ (σf − t) + fω (σf )fω∗ (σf ) − 2fω (σf − t)fω∗ (σf )e−iωt


√ βω
4π 2 λ 0 ω e −1
2 Z ∞
β dω 1 2
= √ fω (σf − t) − fω (σf )eiωt ,
4π 2 λ 0 ω eβω − 1
(5.47)

where in the second line the complex conjugate property Re(z) = (z + z̄)/2 is again used, and in the final
line another complex conjugate property |z|2 = z z̄ is used.

5.2.2. The Limiting Cases of s2 (t)

The mass (or virtuality) of the probe, off-mass-shell light quark80 in the boundary theory determines the
length of the test string `0 in the bulk. Specifically81 ,

r̃0 rH ≡ rc := (rs + `0 ) ⇐⇒ 2 π α0 (Mrest + ∆m) , (5.48)

where Mrest (T ) is the static thermal mass of external particle82 and ∆m(T ) is the thermal rest mass shift.
In the boundary theory this corresponds to the virtuality of the probe light quark being much less than the
temperature of the thermal plasma, i.e. Q2  T 2 . Equivalently in the dual gravitational picture,

λ `20 1 4λ `20
Q2  T 2 =⇒ 2 4
 2 =⇒ 2 2  1 =⇒ `0  rH , (5.49)
4π l β (d − 1) rH

for fixed λ. In the second equality, the Hawking temperature Eq.(4.36) and the definition of the quark mass
(or virtuality) Eq.(4.32) is used; and in the third equality, Eq.(4.36) is used again. In the small virtuality
limit, the minimum radius of the space filling D7-brane is small, i.e. r̃0 → 1 + , where 0 <   1 and r̃0 is
defined in Eq.(4.89). In juxtaposition, the large virtuality limit refers r̃0 → ∞.

In the current subsection the string falling endpoint’s mean-squared transverse displacement, s2 (t), is exam-
ined in a number of different limiting cases. Videlicet, the small virtuality case and the arbitrary virtuality
case. In both of these cases, the asymptotically early time and the asymptotically late time limit is ex-
plored. For the arbitrary virtuality case, a further limit of small or large virtuality can be considered. It is
expected that the large virtuality, early time limit will correspond exactly with the early time limit of the
on-mass-shell heavy quark – since, at early times the light quark’s test string is static (i.e. the boundary
endpoint has not yet begun to fall).

Small Virtuality
In this subsection the behaviour of s2 (t), Eq.(5.47), in the small virtuality limit is considered; and after
obtaining a suitable expression for s2small (t), the asymptotic early and late time dynamics are explored. The
defining condition for the small virtuality limit is given in Eq.(5.49). This equation implies83

80 An off-mass-shell quark does not satisfy Einstein’s energy and momentum relation E 2 = (pc)2 + (2mc)2 . In the case of the
light quark it starts at t = 0 as an off-mass-shell particle (corresponding to the initial static string), radiates energy as it travels
through the thermal medium (string contracts as the boundary endpoint falls at the local speed of light), and finally stops
radiating as it becomes an on-mass-shell particle.
81 See table (1), page 7, for a comprehensive dictionary between quantities in the boundary theory and the bulk theory.
82 In the AdS /N = 4 SYM case, this is the free energy of a quark at rest in N = 4 SYM plasma. In the limit of zero temperature
5
it is equal to the QCD Lagrangian quark mass, mq [68].
83 Explicitly, for the case σ = σ ,
f

         
rH (rs∗ + σ) rH (rs∗ + σf ) rs + `0 rs (→0)
coth2 = coth2 = coth2 coth−1 − ≈ coth2 coth−1 − −→ 1
l2 l2 rH rH

where the second equality follows from using the definition for σf , Eq.(4.52); the approximation follows since `0  rH ; and
in taking the limit the definition of the stretched horizon rs = (1 + ) rH (where 0 <   1) is used. The limit follows since,

46
  
2 rH (rs∗ + σ)
coth → 1, since `0  rH , (5.50)
l2

where σ ∈ [0, σf ]. Using Eq.(5.50) the equations of motion in AdS3 -Schwarzschild, Eq.(4.76), simplifies to
the wave equation

− ∂t2 X(t, σ) + ∂σ2 X(t, σ) = 0 . (5.51)

The plane wave solutions to Eq.(5.51) are the near-horizon solutions described in Eq.(4.85). Hence in the
small virtuality limit the in-falling (+) and out-going (−) modes of the general solution fω (σ) are given by,

fω(±) (σ) = e±iω (rs∗ + σ) . (5.52)

The general solution fω (σ) is a linear combination of the in-falling and out-going modes (defined in
Eq.(4.86)). The coefficient Bω is fixed in the same way as in subsection (4.3.2): by imposing a Neumann
boundary condition at the falling string endpoint in the radial direction at t = 0,

∂σ fω (σ) σ=σf
= 0. (5.53)

Using Eqs.(4.86, 5.52), Eq.(5.53) can be easily solved in Mathematica84 to yield

Bω = ei2ω (rs∗ + σf ) . (5.54)

Hence, the general solution fω (σ) is given by,

fω (σ) = fω(+) (σ) + Bω fω(−) (σ)

(5.55)
= 2eiω (rs∗ + σf ) cos ω (σ − σf ) ,

where Eqs.(5.52, 5.54) and the trigonometric identity cos(x) = (eix +e−ix )/2 are used in the second line. In
order to calculate the falling string endpoint’s mean-squared transverse displacement in the small virtuality
limit, Eq.(5.55) is substituted into Eq.(5.47):

∞
β2
Z 2
dω 1
s2small (t) := s2 (t) 2eiω(rs∗ +σf ) cos ωt − 2eiω (rs∗ +σf ) eiωt


Q2 T 2
= √ βω
4π 2 λ 0 ω e −1
∞
β2
Z
dω 1
2 − e−2iωt − e2iωt


= √ βω − 1
(5.56)
2
4π λ 0 ω e
Z ∞
β2 dω 1
= √ βω
sin2 (ωt) ,
π λ 0 ω e −1
2

where the complex conjugate property |z|2 = z z̄ is used in the second line, and the trigonometric identities
cos(x) = (eix + e−ix )/2 and sin2 (x) = (1 − cos(2x))/2 are used in the final line. The integral in Eq.(5.56)

coth−1 (−(1 + )) → −∞ as  → 0; and coth2 (x) → 1 as x → −∞. Further, for the case σ = 0,
    
rH (rs∗ + σ) 2 rH rs∗
 
−1 rs (→0)
coth2 = coth = coth 2
coth − −→ 1
l2 l2 rH

where the first equality follows from the definition for rs∗ , Eq.(4.48).
84 See
Mathematica Notebook [b]: [Link].

47
 can be solved analytically. Introducing a change of variables
t
x := βω and k := , (5.57)
β

Eq.(5.56) becomes

∞ Z ∞
β2 X dx
s2small (t) = √ sin2 (kx) e−nx
π2 λ n=1 0 x

∞  !
β2 Y 4k 2 (5.58)
= √ ln 1+ 2
4 π2 λ n=1
n

β2
  
β 2πt
= √ ln sinh ,
4 π2 λ 2πt β

P∞ n
where in the second line the geometric series 1/(1 − x) = n=0 x is used; and in the third line the integral
was performed using Mathematica85 . In the final line, Mathematica was again used, followed by a change
of variables using Eq.(5.57). Notice from Eq.(5.58) that β naturally defines a cross-over time scale between
the dynamics of early and late times86 , and that the cross-over time is independent of the length of the
string `0 .

(I) Asymptotic Early Time Dynamics

Since, in the early time limit t  β, Eq.(5.58) can be expanded in powers of k = t/β. For small k a Taylor
expansion (about k = 0) can be performed. Explicitly, ln(sinh(x)/x) = x2 /6 − x4 /180 + x6 /2835 + O(x8 )
(where x = 2kπ). Therefore, Eq.(5.58) becomes
"  4 #  4
2 (tβ) β2 2π 2 t2 t t2 t
ssmall (t) −→ √ + O = √ + O . (5.59)
4 π2 λ 3 β2 β 6 λ β

Since s2small (t) ∼ t2 , the early time dynamics exhibit ballistic behaviour87 (see Eq.(3.20)).

(II) Asymptotic Late Time Dynamics

At asymptotically late times (t  β) the integral in Eq.(5.58) can be expanded in powers of 1/k = β/t.
Specifically,

β2 h  1 i
s2small (t) =



√ ln sinh 2 π k + ln
4 π2 λ 2πk

β2
  
2πt β
+ ln 1 − e−4 π k + ln


= √ (5.60)
4 π2 λ β 4πt

β2
 
(βt) βt β  −4 π t

−→ √ + √ ln + O β2 e β ,
2π λ 4 π2 λ 4πt

where the second line follows from trigonometric identity sinh(x) = (ex − e−x )/2; and, in the final line, the
expansion ln (1 − x) = −x − x2 /2 − x3 /3 + O(x4 ) (where x = e−4 π k ) is performed, since e−4 π k → 0 in
the late time regime (k  1).

Since s2small (t) ∼ t, the late time dynamics exhibit diffusive behaviour88 . From Eq.(3.21) it is expected that

85 SeeMathematica Notebook [b]: [Link].

86 Seefootnote 67, page 33.
87 The early time limit of the on-mass-shell heavy quark also displays ballistic behaviour (see Eq.(4.119)).
88 The late time limit of the on-mass-shell heavy quark also displays diffusive behaviour (see Eq.(4.122)).

48
 s2small (t) = 2 D t at late times, where D is the diffusion coefficient. Comparing this to Eq.(5.60), it is easy
to see that the diffusion coefficient is given by

AdS3 β
DLQ = √ . (5.61)
4π λ

The diffusion coefficient for the on-mass-shell heavy quark is given by Eq.(4.123). By comparing Eq.(5.61)
with Eq.(4.123), notice that an off-mass-shell small virtuality light quark which is initially at rest in a
strongly-coupled thermal plasma has a diffusion coefficient related to that of a massive on-mass-shell heavy
quark by

AdS3 1 AdS3
DLQ = D . (5.62)
2 HQ

The factor of 1/2 may arise through the differences in partitioning the worldsheet for the heavy and light
quark test strings. This is explored further in subsection (6.2).

Arbitrary Virtuality

(I) Asymptotic Early Time Dynamics

Now consider a probe light quark with arbitrary virtuality. The behaviour of s2 (t), Eq.(5.47), at asymptot-
ically early times is considered. Since t  β, Eq.(5.47) can be expanded in powers of k = t/β,

2 (tβ)
fω (σf − t) − fω (σf )eiωt fω (σf ) − fω (σf ) eiωt fω∗ (σf ) − fω∗ (σf ) e−iωt



−→
2
2 − e−iωt − eiωt


= fω (σf )
2 (5.63)
ω 2 t2 + O(ωt)4


= fω (σf )

1 + ν2
ω 2 t2 + O(ωt)4 ,


= 4 2
1 + r̃0 ν

where the first line follows since in the early time regime t  β implies fω (σf − t) ≈ fω (σf ); and in the
third line eiωt and e−iωt are Taylor expanded. The final line follows89 from using Eqs.(4.84, 4.86, 4.91) to
2
calculate fω (σf ) = fω (σf )fω∗ (σf ). Inputting Eq.(5.63) into Eq.(5.47), yields
Z ∞
4 t2 ν 1 + ν2
s2 (t) tβ = √ dν 2πν , (5.64)
λ 0 e − 1 1 + r̃0 ν 2

where a change of variables was performed using Eqs.(4.36, 4.79). Eq.(5.64) is evaluated by breaking the
integral into two parts and recognising that these terms are the series expansion to O(k)0 (around k = 0,
where k = t/β) of the second derivatives with respect to k of the integrals I1 and I2 given in Eq.(4.113) and
Eq.(4.114) respectively. To clarify, the second derivative of I1 and I2 are given by

∞
sin2 kx
Z

dx 2
I1 = 4
0 x(1 + a2 x2 ) ex − 1
Z ∞ (5.65)
∂ 2 I1 ν 1
⇒ = 4π dν 2πν ,
∂k 2 k=0 0 e − 1 1 + r̃02 ν 2

and,

89 Mathematicais used to simplify the algebra. This computation (and others in this section) are explicitly shown in Mathematica
Notebook [b] ([Link]). For access, see appendix (Appendix G).

49
 ∞
dx sin2 kx
Z

2
I2 = 4
0 x ex − 1
Z ∞ (5.66)
∂ 2 I2 ν
⇒ = 4π dν ,
∂k 2 k=0 0 e2πν − 1

where x = 2πν, a = r̃0 /(2π) and k = t/β. Explicitly rewriting Eq.(5.64) in terms of ∂k2 I1 and ∂k2 I2 yields

∞
4 t2 r̃02 − 1
Z  
2 ν 1 1 ν
s (t) tβ
= √ dν 2 2πν − 1 1 + r̃ 2 ν 2
+ 2 2πν
λ
0 r̃0 e 0 r̃0 e −1
(5.67)
t2
 2
r̃0 − 1 ∂ 2 I1 1 ∂ 2 I2

= √ + .
π λ r̃02 ∂k 2 k=0 r̃02 ∂k 2 k=0

The solution to the integrals I1 and I2 are given by Eqs.(4.116, 4.117) respectively. Taking the second
derivative with respect to k followed by a series expansion to O(k)0 (around k = 0) for each of these
solutions, yields

         
r̃0 2π π 1 1


∂ 2 I1 2 −π 2 ln 2π + π 2 ln r̃0 + π 3 cot r̃0 − π2 ψ 1 + r̃0 − π2 ψ 1 − r̃0 − 2π 2 ln(2π)
= ,
∂k 2 k=0 r̃02
(5.68)

where it was recognised in the early time case90 that |k| = k, and ψ(z) is the digamma function. In addition,

∂ 2 I2 π2
= . (5.69)
∂k 2 k=0 3

Inputting Eq.(5.68) and Eq.(5.69) into Eq.(5.67), yields91

t2
       
2 2 2

π 1 1
s (t) tβ
= √ r̃0 + 6 1 − r̃0 2 ln(r̃0 ) − π cot +ψ 1+ +ψ 1− . (5.70)
6r̃04 λ r̃0 r̃0 r̃0

The limit of small or large virtuality can now be considered.

(i) Small Virtuality Limit (r̃0 → 1 + )

Using Mathematica to take the limit of Eq.(5.70) as r̃0 → 1 + , yields

(r̃0 →1+) t2
s2 (t) tβ
−→ √ + O() . (5.71)
6 λ

This agrees exactly with Eq.(5.59), i.e. taking the early time limit followed by the small virtuality limit
is equivalent to taking small virtuality limit followed by the early time limit – a necessary consistency
check.
(ii) Large Virtuality Limit (r̃0 → ∞)
Since r̃0  0, define a variable y = 1/r̃0 (where y  0) so that a Taylor expansion in y can be
performed. Rewriting Eq.(5.70) in terms of y and performing a series expansion (around y = 0) in

90 For
the physically relevant solution t ≥ 0, the early time limit t  β implies β > 0. Hence, k = t/β ≥ 0, and |k| = k.
91 Eq.(5.70)
agrees with Eq.(3.42) in Moerman et al.’s [52] if the the digamma function is rewritten in terms of the harmonic
numbers. Specifically, ψ(n) = H(n−1) − γE , where γE is the Euler-Mascheroni constant and n is a positive integer. It is worth
remarking that this form was avoid here since 1 + r̃1 and 1 − r̃1 are not necessarily positive integers.



0 0

50
 Mathematica, yields
2
t2

(r̃0 →∞) 1
s2 (t) tβ
−→ √ +O . (5.72)
r̃0 λ r̃0

As expected, the large virtuality early time behaviour corresponds exactly with the early time limit
of the on-mass-shell heavy quark (i.e. the static string solution at early times, Eq.(4.122)).

(II) Asymptotic Late Time Dynamics

The behaviour of s2 (t), Eq.(5.47), at asymptotically late times is considered. Defining two dimensionless
quantities
β 1
u := = and z := ωt , (5.73)
t k
the late time dynamics of s2 (t) can be examined. At asymptotically late times t  β, or equivalently
u  1. The integral Eq.(5.47) can then be rewritten
Z ∞
β2 dz 1 2
s2 (t) = √ fz/t (σf − β/u) − fz/t (σf )eiz . (5.74)
4 π 2 λ 0 z ezu − 1

Using Eqs.(4.84, 4.86, 4.91) to simplify the integrand, yields

2
 ∗ ∗

fz/t (σf − β/u) − fz/t (σf )eiz = fz/t (σf − β/u) − fz/t (σf )eiz fz/t (σf − β/u) − fz/t (σf )eiz
(5.75)
(u1) 2
−→ 4 sin (ωt) + O (β/t) ,

where the final line is written in terms of ω, β and t using Eq.(5.73). Further, it has also been recognised
that at asymptotically late times the test string’s falling endpoint is in the near-horizon region, therefore
r ≈ rH . Substituting Eq.(5.75) into Eq.(5.74), yields
Z ∞
β2 dω 1
s2 (t) tβ = √ βω − 1
sin2 (ωt) ≡ s2small (t) . (5.76)
2
π λ 0 ω e

Therefore, the late time dynamics of an arbitrary virtuality light quark in a thermal plasma is diffusive
s2small (t) ∼ t, and the diffusion coefficient is given by Eq.(5.61). This statement can be made due to a key
insight of Moerman et al. [52] – that the integral in Eq.(5.76) is identical to s2small (t), the string falling
endpoint’s mean-squared transverse displacement in the limit of small virtualities (see subsection (5.2.2)).
This allows them to conclude that the behaviour of arbitrary virtuality quarks at asymptotically late times
(i.e. in the near-horizon region) is encoded in the small virtuality case. This is intuitive, since the length of
the Limp Noodle is arbitrarily short at asymptotically late times92 . Since s2small (t; d) can be solved for any
dimensions d ≥ 3 in AdSd -Schwarzschild, the late time behaviour (and therefore the diffusion coefficient)
of Limp Noodles with arbitrary length (equivalently, light quarks of arbitrary virtuality) in any number
of transverse spatial dimensions are able to be determined. This generalisation to AdSd -Schwarzschild is
presented in the following subsection.

5.3. Generalising to AdSd -Schwarzschild

To determine s2small (t; d) the AdSd -Schwarzschild metric (Eq.(4.34)) is examined in the near-horizon limit93 ,
which surmounts to letting r = (1+˜ )rH , where (0 <  ≤ ˜  1) and series expanding each term individually
around ˜ = 0.

92 Comparing Eq.(5.50) with Eq.(E.6) the limit of small virtualities appears to be equivalent to the near-horizon limit.
93 As discussed in subsection (2.1), ‘dropping the 1’ from the function H(r) in the AdSd × S d metric corresponds to being in the
near-horizon region (or in the ‘deep-throat’) of AdSd × S d spacetime – see figure (1), page 7. Here, a further limit is considered
– letting r = (1 + ˜)rH and studying the near-horizon geometry of the black-brane metric – which is the near-horizon limit
refered to in this subsection.

51
 Setting r = (1 + ˜)rH , the AdSd -Schwarzschild metric Eq.(4.34) becomes

2 2
l2
  1 d−1 
rH 2 ~2 rH 1
ds2d 2
= 2 (1 + ˜) dXI − 2 (1 + ˜) 1 − dt2 + 2 ,
d˜ (5.77)
l l 1 + ˜ (1 + ˜)2 1 − (1/(1 + ˜))d−1

where each term is now able to be expanded in ˜ (around ˜ = 0). Specifically94 ,

2 2 2 2 2
 
rH ~ I2 = rH 2rH ˜ rH ˜ ~ I2 ,
(1 + ˜)2 dX + + dX (5.78)
l2 l 2 l 2 l 2

2 2 !
2 2 d2 − 5d + 4 rH
  1 d−1 
rH 2 (d − 1)rH ˜ ˜ 3
(1 + 
˜ ) 1 − dt2 −→ − + ) dt2 ,
+ O (˜ (5.79)
l2 1 + ˜ l2 2l2

!
l2 1 l2 (d − 4)l2 d2 − 14d + 36 l2 ˜ 2
2 −→
d˜ + + + O (˜
) d˜ 2 . (5.80)
(1 + ˜)2 1 − (1/(1 + ˜))d−1 (d − 1)˜
 2(d − 1) 12(d − 1)

~2
The question now arises as how to truncate the series expansions (Eqs.(5.78)-(5.80)), such that the dt2 , dXI
2
and d˜ terms in the metric are all to the same O(˜ ). To help clarify this task of consistent truncation,
there are actually only two options available:
(i) assume dt2 ∼ O(˜ ~ 2 ∼ O(˜
)0 and dX )0 , such that the AdSd -Schwarzschild metric becomes
I

2 2 2
l2
 
ds2d =
rH
+
2rH ˜
dX~ I2 − (d − 1)rH ˜ dt2 + 2 + O (˜
d˜
2
) , (5.81)
l2 l2 l2 (d − 1)˜


(ii) or, assume dt2 ∼ O(˜ ~ 2 ∼ O(˜

)0 and dX ), such that the AdSd -Schwarzschild metric becomes
I

2 2
rH ~ I2 − (d − 1)rH ˜ dt2 + l2 2
ds2d = dX 2 + O (˜
d˜ ) . (5.82)
l2 l2 (d − 1)˜


These are the only options, since if higher order terms in the dt2 or d˜2 series are included, the different signs
in Eqs.(5.79, 5.80) will result in a metric which is not conformally flat in (t, r∗ ) coordinates. The leading
2 series expansion (Eq.(5.80)), therefore – to be consistent –
order metric includes only the first term in the d˜
all terms in the leading order metric must be of O(˜ ). One of the ramifications of this is that it can not be as-
sumed that dX ~ I ∼ O(˜ ~ 2 term in the metric would then disappear.
) (or higher) at leading order, since the dX I

To decide between options (i) or (ii), establish which option ensures that t, X ~ I , ˜ are on equal footing, and
~
that the dt, dXI , d˜  terms scale the same way as ˜ → λ ˜. This will ensure the resulting near-horizon AdSd -
Schwarzschild metric has directions which encode transverse fluctuations95 . For option (ii), considering the
scaling ˜ → λ ˜, the d˜2 term in Eq.(5.82) scales as

l2 l2
2 → λ
d˜ 2 .
d˜
(d − 1)˜
 (d − 1)˜


Under the same transformation, in order for the dt2 term in Eq.(5.82) to scale as
2 2
−(d − 1)rH ˜ 2 −(d − 1)rH ˜ 2
2
dt → λ 2
dt ,
l l

94 In )2 disappear when d = 3.

Eq.(5.79), all terms greater than O (˜
95 Itis clear that the near-horizon AdSd -Schwarzschild metric needs to have transverse directions since these are present in
the near-horizon AdS3 -Schwarzschild metric (which was found directly from inverting the tortoise coordinate in d = 3, and
calculating the transverse string equations of motion without series expanding the parameters – see subsection (5.2)).

52
 ~ 2 term in Eq.(5.82) to scale as
the temporal variable needs to scale like t → t. Similarly, in order for the dX I

2 2
rH ~ I2 → λ rH dX ~ I2 ,
2
dX
l l2
√
the transverse variable needs to scale like X ~I → λ X ~ I 96 . A consistent way to transform t and X ~ I , such
~
that the dt, dXI , d˜  terms in the AdSd -Schwarzschild metric scale the same way under the transformation
˜ → λ ˜, can not be found for option (i). This strongly suggests that option (ii) is the correct way to truncate
the near-horizon AdSd -Schwarzschild metric.

Intuitive arguments aside, a rigorous mathematical approach to proving dt2 ∼ O(˜ )0 and dX~ 2 ∼ O(˜)
I
can be applied in the d = 3 case. Since in AdS3 -Schwarzschild the tortoise coordinate is invertible, the
transverse string equations of motion and its solution can be exactly found. The equations of motion and
its solution can be series expanded in ˜ in the near-horizon region. Order by order these should match
with the equations of motion and its solution derived from the near-horizon AdS3 -Schwarzschild (found
by expanding the metric in ˜ and assuming dt2 ∼ O(˜ )0 and dx2 ∼ O(˜))97 . If this method of matching
terms proves to be consistent, the order of ˜ of the temporal and transverse directions will be known, and
this remains true for a general number of dimensions, d. The near-horizon AdSd -Schwarzschild metric will,
therefore, also be correctly known. The remainder of the subsection will follow the standard method of
finding the equations of motion for the transverse fluctuations, its solution, the string falling endpoint’s
mean-squared transverse displacement s2 (t), and – finally – the diffusion coefficient for the light quark in
AdSd -Schwarzschild.

5.3.1. Expanding the AdS3 -Schwarzschild Transverse Equations of Motion and its Solution in the Near-
Horizon Region
The AdS3 -Schwarzschild metric, the resulting transverse equation of motion, and its solution are given by
Eqs.(4.42, 4.75, 4.77) respectively98 . In the near-horizon region (which corresponds to setting r = (1+˜
)rH ),
Eq.(4.42) becomes
2 2
rH rH l2
ds23 = − 
˜ (2 + 
˜ ) dt2
+ (1 + 
˜ 2
) dx2
+ 2 ,
d˜ (5.83)
l2 l2 ˜ (2 + ˜)
which is equivalent to Eq.(5.77) when d = 3; Eq.(4.75) becomes

2
rH ˜ (2 + ˜)
0 = − ∂t2 X(t, ˜) + ∂˜ ˜ (1 + ˜)2 (2 + ˜) ∂˜X(t, ˜)


4
l (1 + ˜) 2

(˜ + 2) h
= − ∂t2 Xreg (t, ˜) + − l2 ω l2 ω(˜


 + 2) − 2irH ˜(3˜
 + 1)(˜  + 5) Xreg (t, ˜)
4l4 (˜
 + 1)
i
+ 4rH ˜ il2 ω(˜ 2 2
 + 2)∂˜2 Xreg (t, ˜) ,


 + 1)(˜
 + 2) + 4rH ˜(˜
 + 2) + 2rH ∂˜Xreg (t, ˜) + 4rH ˜ (˜
 + 1)(˜
(5.84)

where r = (1 + ˜)rH and ∂˜ = rH ∂r are used in the first line; and Eq.(4.77) becomes
2ω
 il2rH
i l2 ω + rH ˜ + rH −iωt

˜ il2 ω
X(t, ˜) = 2
e =: ˜− 2rH Xreg (t, ˜) , (5.85)
˜ + 2 (˜
 + 1) (rH + i l ω)

(+)
where fω (r) = fω (r) (which is defined in Eq.(4.81)), and the regular part of the solution is given by
Xreg (t, ˜). The regular equation of motion (Eq.(5.84)) can now be series expanded in ˜ (around ˜ = 0).
However, in order to perform this expansion, an ansatz for the expansion of Xreg (t, ˜) is required. Up to

96 Accepting that t and X~ I scale differently isn’t far-fetched considering this is the near-horizon region of the black-brane.
97 This is done in detail Mathematica Notebook [c] ([Link]) – see appendix (Appendix G) for access.
98 The full solution to the transverse equation of motion is given by Eq.(4.86), where the constant B is defined in Eq.(4.91)
ω
(±) (+) (−)
and the modes fω (r) are given in Eq.(4.81). The linearly independent solutions fω (r) or fω (r) are sufficient to work
(+) (−)
with; since, if an ODE is solved by fω (r) or fω (r), it will also be solved by the superposition of those solutions, fω (r).
(+) (−)
Therefore, in order to simplify the algebra, take fω (r) = fω (r) (equivalently fω (r) = fω (r) could have been chosen).

53
 O(˜)2 , the simple ansatz Xreg (t, ˜) = x0 (t) + ˜ x1 (t) + ˜2 x2 (t) can be used. The regular equation of motion
then becomes

2 !
ω −l2 ω + 5irH 2rH + il2 ω
−x000 (t) 2 00


0 = − ω x0 (t) + ˜ x0 (t) + x1 (t) − x1 (t)
l2 l4



2 !
2 ω l2 ω − 2irH −l4 ω 2 + 9il2 rH ω + 14rH
2
4rH + il2 ω 00
+ ˜ − x0 (t) + x1 (t) + ) 3 ,
x2 (t) − x2 (t) + O(˜
4l2 l4 l4
(5.86)

where x00 (t) = d2 x/dt2 . Series expanding the regular part of the solution (Eq.(5.85)) in ˜ (around ˜ = 0),
yields


!
2 2 2 4 2 2 2
il2 ω il ω l ω − 5irH l ω l ω − 11il rH ω − 34rH
Xreg (t, ˜) = e−itω 2− 2rH 1 − ˜ − 2 (l2 ω − ir ) ˜2 + O(˜
)3
4 rH (l2 ω − irH ) 32 rH H (5.87)
=: x0 (t) + ˜ x1 (t) + ˜2 x2 (t) + O(˜
) 3 .

As a consistency check, notice that the solution Eq.(5.87) solves the transverse equation of motion Eq.(5.86)
at each respective order of ˜.

5.3.2. Expanding the AdS3 -Schwarzschild Metric in the Near-Horizon Region

The transverse equation of motion and its solution which have been derived from the AdS3 -Schwarzschild
metric and then series expanded to yield the near-horizon limit (Eqs.(5.86, 5.87)) should match exactly with
the transverse equation of motion and its solution derived from the series expanded, near-horizon AdS3 -
Schwarzschild metric. Going on physical intuition, it seems like a good starting point would be to assume
dt2 ∼ O(˜)0 and dx2 ∼ O(˜ ) and check if the transverse equation of motion and its solution derived from
expanding this metric in the near-horizon limit match with those found in the previous subsection. To this
end, assuming dt2 ∼ O(˜ )0 and dx2 ∼ O(˜ ) results in an AdS3 -Schwarzschild metric
2 2
rH 2rH ˜ 2 l2 2 2
ds23 = dx2
− dt +  + O (˜
d˜ ) , (5.88)
l2 l2 2˜


where Eq.(5.82) for d = 3 has been used. The near-horizon tortoise coordinate, defined by setting r =
(1 + ˜)rH in the definition of the tortoise coordinate Eq.(4.39), can also be expanded and truncated around
˜ = 0. To leading order it is sufficient to take99

l2
˜∗ := 2 ln(˜
) , (5.89)
2 rH

where r∗ = rH ˜∗ , and the number of dimensions d = 3 has been set posthumously. Barring an integration
constant of O(˜ )0 (specifically −l2 /(2rH ) ln(2)), Eq.(5.89) agrees with expanding and truncating Eq.(4.40)
after setting r = (1 + ˜)rH 100 . Inverting and calculating the differential yields
2 2 2
2 rH
˜∗ 2 rH 2 rH
˜∗
˜ = e l2 , and d˜
= e l2 d˜
∗ . (5.90)
l2

Eqs.(5.89, 5.90) are used to transform the expanded and truncated metric into (t, ˜∗ ) coordinates. The
metric Eq.(5.88) becomes

99 It’s proven that the leading order tortoise coordinate is given by Eq.(5.89) in appendix (Appendix F).
100 See Mathematica Notebook [c] ([Link]) for details: access in appendix (Appendix G).

54
 2
rH 2 rH2
˜∗ r2
ds23 = 2 −dt2 + rH
2
2∗ + H dx2


e l2 d˜
l2 l2
(5.91)
r2 2 rH2 r∗ r2
=2 H −dt2 + dr∗2 + H dx2 ,


2
e l
l l2
where the final line is a conformally flat description of near-horizon AdS3 -Schwarzschild in (t, r∗ ) coordi-
nates. It follows from the near-horizon definition r = (1 + ˜)rH , and Eqs.(5.89, 5.90) that

l2
 
 2 rH r∗
 r
r = rH 1 + e l2 , and r∗ = ln −1 . (5.92)
2 rH rH

As discussed in subsection (5.1.3), in the conformally flat (t, r∗ ) coordinate system the boundary conditions
for the Limp Noodle in AdS3 -Schwarzschild are given by Eqs.(4.47, 4.49), which are respectively analogous
to the boundary conditions for the Limp Noodle in R1+1 , Eqs.(4.26, 4.27). Hence, the leading order solution
for the Limp Noodle in AdS3 -Schwarzschild in conformal (t, r∗ ) coordinates can be written down (Eq.(5.39)).
Converting then to (t, r) coordinates, the string solution for the d = 3 Limp Noodle in the near-horizon
limit at leading order is given by
   2r
  µ
rH 1 + e l2H (rs∗ + σ) , if (t, σ) ∈ M 1

µ
XAdS (t, σ) = t, , 0 , (5.93)
3 -Sch  r 1 + e 2 lr2H (rs∗ + σf − t) ,
 
H if (t, σ) ∈ M2 

where Eq.(5.92) is used. The position of the fixed string endpoint attached to the stretched horizon is given
by
l2 r
s
 l2
rs∗ = ln −1 or, equivalently ˜s∗ = 2 ln() , (5.94)
2 rH rH 2 rH

since the stretched horizon is defined as rs = (1 + )rH , where 0 <  ≤ ˜  1. The length of the string in
tortoise coordinates is given by

l2 l2
r + `  
s 0
 `0
σf = ln − 1 − rs∗ or, equivalently σf = ln  + − rH ˜s∗ , (5.95)
2 rH rH 2 rH rH

and the M1 and M2 parameter space regions are still given by Eq.(5.33). Using the definition of the near-
horizon limit (r = (1 + ˜)rH ) and r∗ = rH ˜∗ , the leading order string solution Eq.(5.93) can be converted
into (t, ˜) coordinates. Specifically,
  2r  µ
e l2H (rH ˜s∗ + σ) , if (t, σ) ∈ M1 
µ
XAdS -Sch (t, σ) = t, 2 r H
, 0 . (5.96)
3  e l2 (rH ˜s∗ + σf − t) , if (t, σ) ∈ M  2

2 rH
Therefore, the identification ˜ = e l2 (rH ˜s∗ + σ) is made. The equation of motion for the transverse
fluctuations in the near-horizon limit of AdS3 -Schwarzschild can be derived from Eq.(4.70),

√
 
1
−g g ab GIJ ∂b X J


0 = ∂a − 0 X0µ
2πα
! (5.97)
√ 2
 
ab rH
= ∂a −g g ∂b X(t, ˜) ,
l2 X0µ

55
 where the spacetime metric in the near-horizon limit Gµν is given explicitly by
 2

2 rH ˜
− l2 0 0
 
2
Gµν =  0
 l ,

(5.98)
2 ˜ 0
 2

rH
0 0 l2

where µ, ν index over the directions (t, ˜, x). The explicit entries of the induced worldsheet metric are
calculated from Eq.(4.4)101 . In the near-horizon limit, for d = 3, the leading order induced metric in (t, ˜)
coordinates is given by
     
2 r 2 ˜
gtt gtσ Gtt 0 − lH 2 0
gab X µ :=   =   =  , (5.99)
0 02 02 l2
gσt gσσ X µ 0 ˜ G˜˜ 0 ˜ 2 ˜
0

where Eq.(5.98) is used, and ˜0 = ∂σ ˜. Hence, its inverse is given by
     2

ab
g tt
g tσ
1 ˜02
G 
˜ 
˜ 0 − 2 rl2 ˜ 0
g X µ :=   =
  =  H , (5.100)
0
g σt g σσ X µ det gab |X0µ 0 Gtt 0 2 ˜
0 l2 ˜02

where det gab |X0µ is calculated from Eq.(5.99),

l2 2 r2
  
= ˜02 G˜˜ Gtt = ˜02 2 02
− 2H ˜ = −rH


g|X0µ ≡ det gab |X0µ ˜ . (5.101)
2 ˜ l

Expanding the indices of Eq.(5.97) using Eqs.(5.99)-(5.101), the transverse equation of motion become

! !
2
√ √ r2
   
σσ rH tt
0 = ∂σ −g g ∂σ X(t, σ) + ∂t −g g H ∂t X(t, σ)
l2 X0µ l2 X0µ

2
l2
       2 
0 rH 2 ˜ 0 rH
= ∂σ rH ˜ ∂σ X(t, σ) + ∂t rH ˜ − 2 ∂t X(t, σ)
l2 l ˜02
2 2 rH ˜ l2
(5.102)
2
 
2 4 rH 1 1
= − ∂t X(t, σ) + 4 ˜ ∂σ ˜ ∂σ X(t, σ)
l ∂σ ˜ ∂σ ˜
2
4 rH
= − ∂t2 X(t, ˜) +


4
˜ ∂˜ ˜ ∂˜X(t, ˜) .
l

where the last line is written completely in terms of near-horizon spacetime coordinates (t, ˜) by differenti-
2 rH
(rH ˜s∗ + σ)
ating the identification ˜ = e l2 ,
rH 2 rH2 rH


rH ˜s∗ + σ
∂σ = 2 e l ∂˜ = 2 ˜ ∂˜ . (5.103)
l2 l2

In terms of Xreg (t, ˜) the equation of motion is given by

4rH ˜ rH ˜ ∂˜2 Xreg (t, ˜) + rH + il2 ω ∂˜Xreg (t, ˜)
0 = −∂t2 Xreg (t, ˜) + − ω 2 Xreg (t, ˜) , (5.104)
l4

where the definition of the regular solution in Eq.(5.85) has been used. In order to check whether, at
leading order, the regular equation of motion agrees with the series expanded regular equation of motion

101 This
calculation is similar to Eq.(C.4) where the induced metric is calculated explicitly in (t, r) coordinates (see appendix
(Appendix C)).

56
 (Eq.(5.86)), the leading order ansatz Xreg (t, ˜) = x0 (t) should be used. The leading order contribution to
Eq.(5.104) then becomes

0 = −x000 (t) − ω 2 x0 (t) , (5.105)

which agrees exactly with the leading order contribution of Eq.(5.86). This proves that the expanded
near-horizon metric is equivalent to the near-horizon metric at leading order in AdS3 -Schwarzschild. The
expanded near-horizon metric given in Eq.(5.88) is therefore correct, and dt2 ∼ O(˜ )0 and dx2 ∼ O(˜). As
2
a final check, the calculations in this subsection are repeated assuming dt ∼ O(˜) and dx2 ∼ O(˜
0
)0 . The
equation of motion found is,

2
4 rH ˜
0 = − ∂t2 X(t, ˜) +


∂˜ ˜ (1 + 2 ˜) ∂˜X(t, ˜)
l4 (1 + 2 ˜)


4rH2 4rH ˜ ∂˜Xreg (t, ˜) il2 ω(2˜ + 1) + 4rH ˜ + rH
= − ∂t2 Xreg (t, ˜) +4
˜2 ∂˜2 Xreg (t, ˜) +
l l4 (2˜
 + 1)

(5.106)
ωXreg (t, ˜) l2 ω(2˜  + 1) − 4irH ˜
−
l2 (2˜
 + 1)
 
LO 4irH ˜
−→ − x000 (t) + −ω + 2 ωx0 (t) ,
l (2˜  + 1)

which does not match the leading order contribution of Eq.(5.86).

5.3.3. The Diffusion Constant in AdSd -Schwarzschild

In the previous two subsections it was proven that in order for a near-horizon expansion of the metric to
√ order metric is taken to O(˜
be consistent, the leading ) – where dt2 ∼ O(˜ )0 and dx2 ∼ O(˜
) (equivalently
t ∼ O(˜)0 and x ∼ O( ˜)). Generalising to d dimensions this holds true, where XI (instead of x) denotes
the transverse directions in AdSd -Schwarzschild. Hence the leading order, near-horizon AdSd -Schwarzschild
metric is given by Eq.(5.82).

From the near-horizon metric, which is conformal in (t, r∗ ) coordinates, the leading order string solution
followed by the transverse equations of motion can be found. A similar method to subsection (5.3.2) is
followed. The near-horizon tortoise coordinate, defined by setting r = (1 + ˜)rH in the definition of the
tortoise coordinate Eq.(4.39), can be expanded and truncated around ˜ = 0. To leading order it is sufficient
to take102

l2 ln(˜)
˜∗ := 2 (d − 1) , (5.107)
rH

where r∗ = rH ˜∗ . Inverting and calculating the differential yields

2 2 2
rH
(d−1) ˜∗ rH rH
(d−1) ˜∗
˜ = e l2 , and d˜
= 2
(d − 1) e l2 d˜
∗ . (5.108)
l

The definition of the near-horizon tortoise coordinate, its inverse and differential (Eqs.(5.107, 5.108)) can
be used to convert the metric Eq.(5.82) into (t, ˜∗ ) coordinates

2
rH rH2
(d−1) ˜∗ r2 ~ 2
ds2d = −dt2 + rH
2
2∗ + H


(d − 1) e l2 d˜ dXI
l2 l2
(5.109)
r2 rH r∗
(d−1) 2 r2 ~ 2
= H dr∗2 + H


2
(d − 1) e l2 −dt + dXI .
l l2

102 It’s proven that the leading order tortoise coordinate is given by Eq.(5.107) in appendix (Appendix F).

57
 From the final line in Eq.(5.109) notice that the metric – like the near-horizon AdS3 -Schwarzschild metric
(Eq.(5.91)) – is conformal in (t, r∗ ) coordinates. It follows from the near-horizon definition r = (1 + ˜)rH ,
and Eqs.(5.107, 5.108) that

l2
 
 rH r∗
(d−1)
 r
r = rH 1 + e l 2
, and r∗ = ln −1 , (5.110)
(d − 1) rH rH

which reduces down to Eq.(5.92) for d = 3. From the metric Eq.(5.109) the method laid out in subsection
(5.1.3) can be followed in order to write down the leading order solution for the Limp Noodle in AdSd -
Schwarzschild in conformal (t, r∗ ) coordinates, which is given by Eq.(5.39)103 . Converting then to (t, r)
coordinates, the string solution for the Limp Noodle in the near-horizon limit at leading order is given by
   r
  µ
rH 1 + e lH2 (d−1) (rs∗ + σ) , if (t, σ) ∈ M1 
µ
XAdSd -Sch (t, σ) = t, , 0 , (5.111)
 r 1 + e rlH2 (d−1) (rs∗ + σf − t) ,
 
H if (t, σ) ∈ M 2


where Eq.(5.110) is used104 . The position of the fixed string endpoint attached to the stretched horizon is
given by

l2 r
s
 l2
rs∗ = ln −1 or, equivalently ˜s∗ = ln() , (5.112)
(d − 1) rH rH (d − 1) rH

since the stretched horizon is defined as rs = (1 + )rH , where 0 <  ≤ ˜  1. The length of the string in
tortoise coordinates is given by

l2 l2
r + `  
s 0
 `0
σf = ln − 1 − rs∗ or, equivalently σf = ln  + − rH ˜s∗ ,
(d − 1) rH rH (d − 1) rH rH
(5.113)

and the M1 and M2 parameter space regions are still given by Eq.(5.33). Using the definition of the near-
horizon limit (r = (1 + ˜)rH ) and r∗ = rH ˜∗ , the leading order string solution Eq.(5.111) can be converted
into (t, ˜) coordinates. Specifically,
  r  µ
e lH2 (d−1) (rH ˜s∗ + σ) , if (t, σ) ∈ M 
µ 1
XAdS d -Sch
(t, σ) = t, rH , 0 , (5.114)
 e l2 (d−1) (rH ˜s∗ + σf − t) , if (t, σ) ∈ M  2

rH
where ˜s∗ and σf are defined in Eqs.(5.112, 5.113). Therefore, the identification ˜ = e l2 (d−1) (rH ˜s∗ + σ)
is made. The equations of motion for the transverse fluctuations in the near-horizon limit of AdSd -
Schwarzschild can be derived from Eq.(4.67), and are equivalent to Eq.(5.97) with the generalised spacetime
metric in the near-horizon limit (Gµν ) given by
 r2 
− H2 (d − 1) ˜ 0 0 0 0 0
 l 
l2 1

 0 (d−1) ˜ 0 0 0 0

 2 
rH

 0 0 l2 0 0 0

Gµν =  2
rH ..
, (5.115)
.
 
 0 0 0 l2 0
 
 .. .. 

 0 0 0 . . 0

2
rH
0 0 0 0 0 l2

where µ, ν index over (t, r, 2, ... , d − 1). The explicit entries of the induced worldsheet metric are calculated

103 This equation – derived from the worldsheet parameter space partitioning of the test string in R1,1 , Eq.(5.34) – is independent
of d, the number of spacetime dimensions.
104 It will be much harder to find next-to-leading order corrections to the string solution in the near-horizon limit, since if the

)2 terms) it will no longer be conformal in (t, r∗ ) coordinates.

near-horizon metric includes higher order terms (O(˜

58
 from Eq.(4.4). In the near-horizon limit, the leading order induced metric in (t, ˜) coordinates is given by
     2 
rH
gtt gtσ Gtt 0 − (d − 1) ˜ 0
gab X µ :=   =   =  l2 , (5.116)
0 02 02 l2 1
gσt gσσ X µ 0 ˜ G˜˜ 0 ˜ (d−1) ˜
0

where Eq.(5.115) is used, and ˜0 = ∂σ ˜. Hence, its inverse is given by
     
tt tσ 02 l2
g g 1 ˜ G ˜˜ 0 − (d−1)r 2 ˜ 0
ab
g X µ :=   =
  =  H  , (5.117)
0
g σt g σσ X µ det gab |X0µ 0 Gtt 0 (d−1) ˜
0 l2 ˜02

where det gab |X0µ is calculated from Eq.(5.116),

l2 2
  
1 rH
g|X0µ ≡ det gab |X0µ = ˜02 G˜˜ Gtt = ˜02 2 02


− (d − 1) 
˜ = −rH ˜ . (5.118)
(d − 1) ˜ l2

Expanding the indices of Eq.(5.97) using Eqs.(5.116)-(5.118), the transverse equation of motion become

! !
2
√ √ r2
   
rH
0 = ∂σ −g g σσ ∂σ X I (t, σ) + ∂t −g g tt H ∂t X I (t, σ)
l2 X0µ l2 X0µ

2
l2
       2 
0 (d − 1) ˜ rH I 0 1 rH I (5.119)
= ∂σ rH ˜ ∂σ X (t, σ) + ∂t rH ˜ − 2  ∂t X (t, σ)
l2 ˜02 l2 (d − 1)rH ˜ l2
2
rH
= − ∂t2 X I (t, ˜) + (d − 1)2 ˜ ∂˜ ˜ ∂˜X I (t, ˜) ,


l 4

where X I denotes the transverse direction (I ∈ (2, 3, ... , d − 1)); and the last line is written completely in
rH
terms of near-horizon spacetime coordinates (t, ˜) by differentiating the identification ˜ = e l2 (d−1) (rH ˜s∗ + σ) ,

rH rH2


(d−1) rH ˜s∗ + σ
 = (d − 1)
d˜ el dσ
l2
rH rH2 (d−1)


rH ˜s∗ + σ (5.120)
⇒ ∂σ = (d − 1) el ∂˜
l2
rH
⇒ ∂σ = (d − 1) 2 ˜ ∂˜ .
l

Eq.(5.120) can also be used to convert the equations of motion Eq.(5.119) to parameter space coordinates
(t, σ),

2
l2 l2
 
rH 1 1
0 = − ∂t2 I 2
X (t, σ) + 4 (d − 1) ˜ ∂σ ˜ I
∂σ X (t, σ)
l (d − 1) rH ˜ (d − 1) rH ˜ (5.121)
= − ∂t2 I
X (t, σ) + ∂σ2 I
X (t, σ) .

The near-horizon string equations of motion for the transverse fluctuations on the Limp Noodle is given
by wave equation. Remember that the small virtuality limit of the d = 3 Limp Noodle also yielded
the wave equation as the transverse equation of motion (Eq.(5.51))105 . Since this is true, repeating the

1
± (d−1) il2 ω
105 The (±) r
solution to Eq.(5.121) is therefore given by Eq.(5.52), or alternatively by fω (˜
) = ˜ H , where the identification
rH
(d−1) (rH 
˜s∗ + σ)
˜ = e l2 is used to convert from parameter space coordinates (t, σ) to near-horizon spacetime coordinates
(+) (−)
(t, ˜). Notice when fω (˜) is used as the solution (as in subsection (5.3.1)), fω (˜) is used in the definition of Xreg (t, ˜),
Eq.(5.85).

59
 calculations laid out in subsection (5.2.2) is sufficient to calculate s2 (t; d). The only change is that β scales
with the number of spacetime dimensions, so Eq.(5.56) needs to be adapted by adding in the appropriate
dimensionally dependent factor. In order to calculate this, notice that β enters the calculation of s2small (t)
in subsection (5.2.2) through the use of the normalization constant Aω . For d ≥ 3, Eq.(4.101) is generalised
r
l πα0 (d − 1)β (d − 1)
Aω (d) := = √ ≡ Aω , (5.122)
rH ω 4 πω λ1/4 2

where the second equality follows from using the definition of the AdS radius of curvature l (Eq.(4.65)) and
the Hawking temperature (Eq.(4.36)). Using the generalised normalization constant Aω (d), the expectation
value of the position of the boundary endpoint at two different times (Eq.(5.43)) becomes106

(d − 1)2 β 2 ∞ dω
Z
1 
∗ −iω(t1 −t2 )

h:X̂end (t1 ; d)X̂end (t2 ; d):i = √ Re f ω (σf − t1 )f ω (σ f − t 2 )e
16π 2 λ 0 ω eβω − 1
(5.123)
(d − 1)2
≡ h:X̂end (t1 )X̂end (t2 ):i ,
4
while the string falling endpoint’s mean-squared transverse displacement, Eq.(5.47), becomes

∞
(d − 1)2 β 2
Z
dω 1 2
s2 (t; d) = √ βω
fω (σf − t) − fω (σf )eiωt
16π 2 λ 0 ω e −1
(5.124)
(d − 1)2 2
≡ s (t) ,
4

due to the factors of Aω Aω0 appearing in the calculations of both h:X̂end (t1 ; d)X̂end (t2 ; d):i and s2 (t; d).

In subsection (5.2.2), the general solution to the wave equation, fω (σ), is given by Eqs.(5.52)-(5.55). In
order to calculate the falling string endpoint’s mean-squared transverse displacement in the small virtuality
limit, Eq.(5.55) is substituted into Eq.(5.124). Specifically,

∞
(d − 1)2 β 2
Z 2
dω 1
s2 (t; d) = 2eiω(rs∗ +σf ) cos ωt − 2eiω (rs∗ +σf ) eiωt


√ βω
16π 2 λ 0 ω e −1
∞
(d − 1)2 β 2
Z
dω 1
= √ sin2 (ωt)
4π 2 λ 0 ω eβω − 1
(5.125)
2
(d − 1) 2
≡ ssmall (t)
4
(d − 1)2 β 2
  
β 2πt
= √ ln sinh ,
16π 2 λ 2πt β

where Eq.(5.56) is used in the third line, and Eq.(5.58) in the final line. In order to calculate the diffusion
coefficient, the late time dynamics of Eq.(5.125) must be explored. Expanding in powers of 1/k = β/t yields

(d − 1)2 β 2 h  1 i
s2 (t; d) =



√ ln sinh 2πk + ln
16π 2 λ 2πk
(5.126)
2
(d − 1)2 β 2
 
(βt) (d − 1) βt β  −4πt

−→ √ + √ ln + O β2 e β ,
8π λ 16π 2 λ 4πt

where the calculation steps in Eq.(5.60) have been followed closely. Since s2 (t; d) ∼ t, the late time dynamics
exhibit diffusive behaviour. From Eq.(3.21) it is expected that s2small (t) = 2 D t at late times, where D is

106 Note that the Bose-Einstein distribution Eq.(4.103) does not carry any dimensional dependence.

60
 the diffusion coefficient. Comparing this to Eq.(5.126) it is easy to see that the diffusion coefficient is given
by

AdSd (d − 1)2 β
DLQ (d) = √ . (5.127)
16π λ
For d = 3, this reduces down to Eq.(5.61) as expected.

Moerman et al. in [52] present an expression for the light quark diffusion coefficient in d dimensions which
agrees with Eq.(5.127). However, their method of expanding the near-horizon AdSd -Schwarzschild metric
(described in subsection (3.3) of [52]) is inconsistent. The near-horizon AdSd -Schwarzschild metric given in
Eq.(3.48) of [52] is dimensionally incorrect107,108 . Using the near-horizon tortoise coordinate Eq.(5.107) (no
attempt is made to explain how the order of truncation is consistent with the leading order near-horizon
metric109 ), the metric in Eq.(3.48) of [52] is converted to (t, ˜∗ ) coordinates, and found to be conformal in
this system110 . Hence, the leading order string solution can be written down in (t, ˜∗ ) coordinates, from
which Moerman et al. proceeded to finding the transverse equations of motion – this, coincidentally, being
the wave equation. Therefore, although there is agreement between the final results, Eq.(5.127) and the
diffusion coefficient in d dimensions derived by Moerman et al., the authors conclude that the calculations
in this subsection arriving at Eq.(5.127), present the first complete, consistent derivation of the light quark
diffusion coefficient in d dimensions.

6. Drag Force in AdS/CFT

Besides the test string set-ups depicted in figures (2) and (3) (see section (4), page 15, and section (5),
page 35, respectively), there are many other string set-ups – illuminating various aspects or properties
of the thermal plasma in the boundary theory – that could be considered. One such set-up is a trailing
string in AdS5 -Schwarzschild which models an infinitely massive probe quark111 moving with a constant
velocity v in a N = 4 SYM thermal plasma112 . This was considered by Gubser [22] and Herzog et al. [35]
independently in 2006, who aimed to approximately explain jet-quenching – the phenomenon whereby energy
loss is experienced by high-energy quarks travelling through the quark-gluon plasma113 . The succeeding
subsection follows the calculations of [22, 35] in using the AdS/CFT correspondence to calculate the drag
force experienced by a massive probe quark in the N = 4 SYM thermal medium. The string set-up used is
depicted in figure (9). Further, in subsection (6.2), the set-up is generalised to AdSd -Schwarzschild.

6.1. Test Strings in AdS5 -Schwarzschild

Consider a static open string in AdS5 -Schwarzschild, with one endpoint attached to the boundary at r = ∞
and the other allowed to hang down towards the horizon. Since the string is infinitely long (which models
an infinitely massive quark in the boundary theory), the transverse fluctuations that the string experiences
due to the presence of the black hole horizon never reach the boundary, i.e. the infinitely massive quark
does not experience Brownian motion in the thermal medium. Regardless of its length, this test string is
still a classical relativistic string and, as such, is described by the standard Polyakov action (Eq.(4.3)) or
Nambu-Goto action (Eq.(4.54)). It is convenient to use the Nambu-Goto action here114 .

107 Since 2 , l2 dx2

rH 2 −1 , and  2 ) are defined as dimensionless quantities; the first
d−2 and dt all go like Energy ˜ (as well as d˜
term in the metric Eq.(3.48) of [52] goes like Energy −2 , the second term also goes like Energy −2 , but the third term is
dimensionless. In juxtaposition, all terms in the correct, leading order near-horizon AdSd -Schwarzschild metric given in this
subsection (Eq.(5.82)) consistently go like Energy −2 .
108 A note on terminology:  in [52] corresponds to ˜ in this article.
109 The paper [52] simply states ”Expanding (and truncating) each term in the AdS metric, Eq.(2.32), to lowest non-vanishing
d
order in ”. There does not appear to have been an attempt to examine if each term is consistently taken to the same order
of  (˜
 – see footnote 108).
110 The correct near-horizon AdS -Schwarzschild metric (Eq.(5.82)) is actually conformal in (t, r ) coordinates (as shown in
d ∗
Eq.(5.109)).
111 Due to its infinite mass such a quark would not undergo Brownian motion in thermal plasma.
112 The constant velocity is measured with respect to the rest frame of the plasma.
113 Around the same time, studies examining trailing strings in the AdS spacetime in order to investigate the dissipative and

diffusive behaviour of a massive quark moving through a field theory plasma were common (see [36, 37, 38, 39, 40, 41]).
114 The Nambu-Goto action given in Eq.(4.54) is defined in the String Frame. In the Einstein Frame, the Nambu-Goto action is

constructed with an added dilaton factor eφ/2 ,

√
Z
E 1
SN G := − 0
d2 σ eφ/2 −g ,
2πα M

61
 Figure 9: A fundamental open string used as a probe in an AdS black hole background to model an infinitely
massive quark moving through a thermal plasma at a constant velocity v on the boundary. One endpoint of the
test string is attached to the boundary of anti-de Sitter spacetime at infinity (r = ∞), from which the string hangs
down to a stretched horizon (rs = (1 + ) rH where 0 <   1) placed just above the Schwarzschild black hole
horizon. Left: A string which trails behind its boundary endpoint moving with velocity v in the X 2 direction. This
is the physical solution. Energy flows from the boundary, down the string, and towards the horizon. Right: A
string which trails in front of its boundary endpoint moving with velocity v in the X 2 direction. This solution is
unphysical. Energy would need to be flowing away from the horizon, up the string, and towards the boundary.

From Eq.(4.34), the AdS5 -Schwarzschild metric is given by

r2  2 2
~ 2 + l dr ,
  r 4
H
ds25 = 2
−h(r) dt2 + dX I 2
where h(r) = 1 − ∈ [0, 1] (6.1)
l r h(r) r

is the blackening factor of the Schwarzschild black hole situated at the horizon; t ∈ [0, ∞) is the temporal
coordinate; r ∈ [0, ∞) is the radial coordinate; and the transverse spatial directions along which the D3-
~ I = X 2 , X 3 , X 4 ∈ R3 . Further, l ∈ R+ is the curvature radius of AdSd


brane is extended are denoted by X
and S d .

Following [22, 35], a gauge choice is made where τ = t and σ = r are the worldsheet coordinates. The target
spacetime metric Gµν – where µ, ν index over the spacetime directions (t, r, 2, 3, 4) – is given explicitly in
Eq.(C.2); and the induced worldsheet metric gab , defined in Eq.(4.4), takes the form Eq.(C.3). The explicit
entries of the metric gab , its inverse g ab , and its determinant g := det(gab ) are given by Eqs.(C.4, C.5, C.6)
respectively. Using the target spacetime metric Gµν and remembering the gauge choice τ = t and σ = r,
the determinant of the induced metric becomes
r4
 
1
g=− 1 − ẊI2 + 4 h(r) XI02 , (6.2)
h(r) l

where Ẋ ≡ ∂τ X = ∂t X and X 0 ≡ ∂σ X = ∂r X (in both definitions the second equality follows from the
worldsheet gauge choice). Inputting Eq.(6.2), into the definition of the Nambu-Goto action Eq.(4.54) yields
r
r4
Z
1 2 1 2
SN G = − d σ 1 − Ẋ + h XI02 , (6.3)
2πα0 M h I l4

where the notational adjustment h(r) ≡ h is used for brevity. The Lagrangian density, also defined in

where g := det (gab ) and the induced worldsheet metric gab is given by Eq.(4.4). In [22] the Einstein Frame is chosen for
the proceeding calculations. The phenomenologically relevant case of AdS5 -Schwarzschild is considered. In this spacetime the
dilaton factor can be ignored, since eφ/2 = H (5−d)/4 (where H is the warping factor and eφ/2 disappears when d = 5). It is
worth a footnote to remark, however, that the Einstein Frame is related to the String Frame by a conformal rescaling of the
metric, and that the frames are considered interchangeable when describing the physics of massless modes on the string [109].
The calculations in this article are in the String Frame.

62
 Eq.(4.54), is therefore given by
r
1 2 r4
L=− 1− ẊI + 4 h XI02 . (6.4)
h l

To proceed, consider movement of the test string with a constant velocity v in one of the transverse directions,
the X 2 direction115 . Following [22], an ansatz to describe the late-time behaviour of the string is made.
Specifically116 ,
X 2 (t, r) = vt + ξ(r) + o(t) , (6.5)

where, at late times, all other motions are damped out and the o(t) term disappears. The ansatz Eq.(6.5)
relies on the assumption that steady state behaviour of the string’s motion is achieved at late times. Sub-
stituting this ansatz into the Lagrangian density Eq.(6.4) – remembering that X 2 is the only direction in
which the string is considered to move – yields
r
v2 r4
L = − 1− + 4 h ξ 02 , (6.6)
h l

where ξ(r) ≡ ξ is implied, and ξ 0 = ∂r ξ. In order to determine ξ(r), the Euler-Lagrange equations are
derived from the Lagrangian density (Eq.(6.6)), and are given by
∂ ∂L ∂L
− = 0, (6.7)
∂t ∂ξ 0 ∂ξ

where the quantity πξ can be defined as

∂L
πξ :=
. (6.8)
∂ ∂r ξ

Since the Lagrangian density (Eq.(6.6)) does not depend on ξ (i.e. ∂L/∂ ξ = 0), the Euler-Lagrange
equations become

∂L
πξ := = C
∂ ξ0
− 21 (6.9)
r4 0 v2 r4 02

= − 4 hξ 1 − + 4 hξ ,
l h l

where Eq.(6.6) is used in the second line. Hence, the equation of motion of the string is πξ being a constant
of motion. Solving Eq.(6.9) yields a defining relation for ξ 0 (r). Mathematica’s Solve function is used117 , to
find
s
0 l4 h − v2
ξ (r) = ± πξ 4 . (6.10)
r h h − rl44 πξ2

Finding ξ(r) will involve integrating Eq.(6.10) with respect to r. However, ξ(r) is required to be real
everywhere, and the square root in Eq.(6.10) is not necessarily always real. The only variable which can be
4
adjusted is πξ . Defining πξ such that (h − v 2 )/(h − rl 4 πξ2 ) is always positive (thereby ensuring the square
root is always real) is equivalent to setting the appropriate boundary conditions. Since h(r) ∈ [0, 1], at some
intermediate radius h = h∗ the numerator (h − v 2 ) changes sign. In order for the square root to be real

115 Due to a symmetry of anti-de Sitter spacetime all transverse directions are identical. Therefore any of the transverse directions
can be chosen – considering motion in either the X 2 , X 3 or X 4 directions would be equivalent.
116 To clarify the notation: the mapping function for the second transverse direction will be denoted X 2 , while squaring this

mapping function (if necessary) will be denoted (X 2 )2 .

117 Mathematica is used to help simplify the algebra. See Mathematica Notebook [d] ([Link]) for details on the calculations

presented in section (6): access in appendix (Appendix G).

63
 everywhere the denominator must change sign at the same point, i.e. at h = h∗ ,

l4 2
h− π = h − v2 = 0 . (6.11)
r4 ξ

Rearranging the first equality leaves

r4 2
πξ2 = v . (6.12)
l4

While the second equality in Eq.(6.11) is rearranged to

4
rH
r4 = . (6.13)
1 − v2

where the definition of h(r) (Eq.(6.1)) is used. Inputting Eq.(6.13) into Eq.(6.12) results in an appropriate
expression for πξ ,
2
v rH
πξ = ± √ . (6.14)
1 − v 2 l2

Plugging this defining relation for πξ into Eq.(6.10) guarantees that the square root in Eq.(6.10) is always
positive, and ξ(r) is everywhere real. The (±) in Eq.(6.14) determines if the quark is moving in the (+) or
(−) X 2 direction. Consider the case where πξ > 0 for the remainder of the calculation. Eq.(6.10) becomes,
2 2
rH l
ξ 0 (r) = ± v 4 . (6.15)
r4 − rH

An integrable expression for ξ 0 has finally been found. Integrating Eq.(6.15) with respect to r yields118

l2 v
   r 
−1 r r + rH
ξ(r) = ∓ tan + ln , (6.16)
2 rH rH r − rH

where ln(−x) = ln(x) + i π for x > 0 is used and it is recognised that i π can be absorbed into the integra-
tion constant, which (if the calculation was entirely rigorous) would be present in Eq.(6.16). The function
ξ(r) can take on two values: (i) for (ξ 0 (r) > 0, ξ(r) < 0) the string trails behind its boundary endpoint
which is moving with a velocity v in the (+) X 2 direction, and (ii) for (ξ 0 (r) < 0, ξ(r) > 0) the string
trails in front of its boundary endpoint which is moving with a velocity v in the (+) X 2 direction. The ‘tail
wagging the dog’ scenario depicted in (ii) is an unphysical solution [35]; however both of these cases are
depicted in figure (9), page 62.

Using the expression for ξ 0 (r) given in Eq.(6.15) the Lagrangian density (Eq.(6.6)) becomes
p
L = − 1 − v2 . (6.17)

√
Since the definition of the Nambu-Goto action (Eq.(4.54)) implies L = − −g, the determinant of the
induced metric becomes
g = − 1 − v2 .


(6.18)

To calculate the flow of momentum down the string dp2 /dt, the canonical momentum densities Πaµ (t, σ)
(Eq.(4.7)), and the string equations of motion (Eq.(4.20)) are needed. The conserved charges associated
with the momentum densities Πaµ are defined [96, 15] for a general curve γ on the worldsheet, as
Z
γ
pµ := dσ b ˜ab Πaµ , (6.19)
γ

118 Mathematica is used to easily integrate (see Mathematica Notebook [d]: [Link]).

64
 where ˜ab is the Levi-Civita symbol119 , and gab is the induced metric on the worldsheet in curved spacetime
(Eq.(4.4)). Since the string moves only in the X 2 direction by construction, the string’s transverse mo-
mentum densities are only non-zero in this direction. Using Eq.(6.19) the X 2 component of the spacetime
momentum flowing over some general time interval I of length ∆t is
Z
dp2
∆t = pI2 = dt Πr2 , (6.20)
dt I

where the gauge choice τ = t and σ = r is applied, and ˜rt = +1 from the definition of the Levi-Civita symbol.
The radius at which the integral in Eq.(6.20) is evaluated does not matter, since the momentum densities
Πaµ (t, σ) are conserved (Eq.(4.20)). For the physical solution where the string trails behind its boundary
endpoint (ξ 0 (r) > 0, ξ(r) < 0), the quantity dp2 /dt is identified as the drag force and the orientation of
the integral is chosen such that the drag force is negative (i.e. acts to oppose the motion of the string).
Therefore, dp2 /dt = Πr2 . According to the definition of the momentum densities Eq.(4.7),

dp2 1 √
:= Πr2 = − −γ γ rb G2 ν ∂b X ν
dt 2πα0
(6.21)
1 p
= − 1 − v 2 g rb G2 ν ∂b X ν ,
2πα0

where the constraint equation, Eq.(4.12), and Eq.(6.18) is used in the second line. Expanding the implied
summation over repeated indices in Eq.(6.21) yields

dp2 1 p
1 − v 2 G2 ν g rr ∂r X ν + g rt ∂t X ν


= − 0
dt 2πα
1 1
G2 2 Gtt + G2 2 ∂t X 2 ∂t X 2 ∂r X 2 + −G2 2 ∂r X 2 ∂t X 2 ∂t X 2




= 0
√
2πα 1 − v 2
(6.22)
1 1
= √ G2 2 Gtt ∂r ξ
2πα0 1 − v 2
2
1 rH v
= − 0 2
√ ,
2πα l 1 − v2

where, in the second line, it was recognised that G2 ν = G2 2 is the only non-zero possibility since Gµν
is a diagonal matrix, and Eq.(C.5) was used to find the explicit entries of the inverse induced metric g ab
(remembering that the gauge choice τ = t and σ = r must be made). Further, Eq.(6.18) was used and it
was noted that, by construction, the only non-zero transverse direction was the X 2 direction. In the third
line the commutative property of partial derivatives and the ansatz Eq.(6.5) was used. The final line follows
from inputting the target spacetime metric entries (Eq.(C.2)) and the defining relation for ξ 0 (r), Eq.(6.15)
(where ξ 0 (r) > 0).

Comparing Eq.(6.14) (where πξ < 0) and Eq.(6.22), notice that πξ is understood to be a momentum density,
since
1
Πr2 = πξ . (6.23)
2πα0

1
Eq.(6.23) follows from Eq.(4.7) which gives Πσ2 := 2πα0 ∂(∂σ X ) L;
2 and from the worldsheet gauge choice

119 The Levi-Civita symbol is defined in two dimensions as

 
+1 ,

 if (a, b) = (1, 2)


 
˜ab = −1 , if (a, b) = (2, 1) .

 

0 ,

if a = b 


65
 1 1
σ = r, from which the momentum density becomes Πr2 = 2πα 0 ∂(∂r ξ) L = 2πα0 πξ (where the ansatz
Eq.(6.5) is used in the first equality and Eq.(6.8) in the second equality).

2
For the phenomenologically relevant AdS5 /N = 4 SYM case, the ‘t Hooft coupling is λ ≡ gYM Nc , where
√
Nc is the number of colours and the Yang-Mills coupling is related to the string coupling by: gYM = 2 π gs .
The fundamental string length scale Eq.(4.65) becomes

l2
α0 = p 2
. (6.24)
gYM Nc

Using Eq.(6.24) and the definition of Hawking temperature (Eq.(4.36)) with d = 5, the drag force (Eq.(6.22))
can be expressed in terms of phenomenologically relevant gauge theory variables,
p
dp2 π gY2 M Nc 2 v
= − T √ . (6.25)
dt 2 1 − v2

Following the formal frameworki of Gubser [22], in the gauge theory the momentum p2 of the probe quark
can be theoretically120 related to its mass m through
v
p2 (t) = √ m, (6.26)
1 − v2

√
where the 1/ 1 − v 2 ≡ γ is identified as the Lorentz factor from special relativity. Eq.(6.26) holds regardless
of the dimension of the spacetime. Inserting Eq.(6.26) into Eq.(6.25) yields

dp2 π T2 p2 (t)
q
= − gY2 M Nc , (6.27)
dt 2 m
which is a linearly separable differential equation and easily integrable to find p2 (t). Specifically,

t 2 m
p2 (t) = p2 (0) e− t0 , where t0 = . (6.28)
gY2 M Nc T 2
p
π

The expression for the drag force experienced by a probe quark in a thermal plasma found in Eq.(6.27)
agrees with the central result of Gubser [22], and Herzog et al. [35]. Agreement with the later rests on the
identification ξ 0 = x0Herzog /l2 and g = −gHerzog /l4 .

Recall the Langevin Model discussed in section (3) describes a non-relativistic particle of mass m, undergoing
Brownian motion in one spatial dimension. The non-retarded Langevin equation Eq.(3.1) is applicable if
– as in the case presented in this section – the Brownian particle is taken to have infinite mass with
respect to the constituent fluid particles. Since the infinitely massive probe quark travels at a constant
velocity v under the influence of the external force K(t), Newton’s First Law of Motion ensures that the
random force F (t) present in the system must be of equal and opposite magnitude (K(t) = −F (t)). The
Langevin equation is adjusted accordingly and becomes ṗ(t) = −γ0 p(t), where γ0 is the friction coefficient.
It is therefore apparent, from Eq.(6.27), that in the non-relativistic limit v  1 the friction coefficient in
AdS5 -Schwarzschild is given by

π T2
q
γ0AdS5 = gY2 M Nc . (6.29)
2m

Using the Einstein-Sutherland relation Eq.(3.18) an expression for the diffusion coefficient can be obtained.
Specifically,
AdS5 2 1 2β
DHQ = p = √ , (6.30)
πT 2
gY M Nc π λ

2
where the second equality follows from the ‘t Hooft coupling λ ≡ gYM Nc , and the definition of Hawking
temperature Eq.(4.36).

120 Both the momentum p2 and the mass m of the external quark are, in fact, infinite.

66
 6.2. Generalising to AdSd -Schwarzschild
The result for the drag force experienced by a probe quark in a thermal plasma found in Eq.(6.27) can
be generalised by repeating the calculation presented in subsection (6.1) in AdSd -Schwarzschild. The AdS-
Schwarzschild metric in d dimensions is given by Eq.(4.34), where the blackening factor of the Schwarzschild
black hole situated at the horizon, h(r; d), is defined in Eq.(4.35). Explicitly, the target spacetime metric
Gµν is given by Eq.(C.2), where µ, ν index over the spacetime directions (t, r, 2, ..., d − 1). As in subsec-
tion (6.1), the reparameterization offered by the gauge choice separates the temporal and radial coordinates
such that the worldsheet parameter space coordinates become τ = t and σ = r, and the mapping functions
between the worldsheet and the target spacetime are specified by X µ (t, r).

The Hawking temperature of the black-brane in AdS-Schwarzschild scales with the number of dimensions
of the spacetime and – in AdSd -Schwarzschild – is given by Eq.(4.36). Unchanged as the calculation is
generalised to AdSd , the dynamics of the string are still described by the Nambu-Goto action (Eq.(4.54));
and the induced metric gab , its inverse g ab , and its determinant g := det(gab ) are still given by Eqs.(C.4,
C.5, C.6) respectively. Because of this, Eqs.(6.2)-(6.10) all remain true in AdSd -Schwarzschild.

Consider movement of the test string in AdSd -Schwarzschild, again with a constant velocity v in one of the
transverse directions, the X 2 direction. Following the same reasoning as for the AdS5 case, Eq.(6.10) can be
solved to yield an expression for πξ . Since the blackening factor is now given by Eq.(4.35), πξ will depend
on the number of dimensions of the spacetime. Rearranging the first equality in Eq.(6.11) leaves

r4 2
πξ2 = v . (6.31)
l4

While, for AdSd -Schwarzschild, the second equality in Eq.(6.11) is rearranged to

rH
r = , (6.32)
(1 − v 2 )1/(d−1)

where h = h(r; d) (Eq.(4.35)) is used. Inputting Eq.(6.32) into Eq.(6.31) results in an appropriate expression
for πξ ,
2
v rH
πξ = ± . (6.33)
(1 − v 2 )2/(d−1) l2

Plugging this generalised relation for πξ into Eq.(6.10) guarantees that the square root in Eq.(6.10) is always
positive, and ξ(r) is everywhere real. The (±) in Eq.(6.33) determines if the quark is moving in the (+) or
(−) X 2 direction. Again consider the case where πξ > 0. In AdSd -Schwarzschild, Eq.(6.10) becomes121
r
2 3 d−4

2
2 − d−1 r 4 (rH (v 2 −1) r d + r rHd
)
l rH r v 1−v − 4
−r r d+4 + r 5 v 2 r d (1−v 2 ) d−1 +r 5 r d
ξ 0 (r) = ±
H H H
d
, (6.34)
rH rd − r rH

which reduces down to the familiar Eq.(6.15) for d = 5. The function ξ 0 (r) can take on two values: (i) for
(ξ 0 (r) > 0, ξ(r) < 0) the string trails behind its boundary endpoint which is moving with a velocity v in
the (+) X 2 direction, and (ii) for (ξ 0 (r) < 0, ξ(r) > 0) the string trails in front of its boundary endpoint
which is moving with a velocity v in the (+) X 2 direction. Using the expression for ξ 0 (r) in Eq.(6.34) and
the definition for h(r; d) in Eq.(4.35), the Lagrangian density (Eq.(6.6)) becomes
v

√ r4 rH (v 2 − 1) rd + r rH d
u
u
L := − −g = − t 4 . (6.35)
−rH rd+4 + rH 5 v 2 r d (1 − v 2 )− d−1 + r 5 r d
H

To proceed with calculating the drag force dp2 /dt note that the canonical momentum densities Πaµ (t, σ)
(Eq.(4.7)), the string equations of motion (Eq.(4.20)), and the definition of the conserved charges associated
with the momentum densities Eq.(6.19) all remain the same in AdSd -Schwarzschild. As in the d = 5 case,
the string moves only in the X 2 direction by construction, and its transverse momentum densities are only

121 Mathematica is used to help simplify the algebra (see Mathematica Notebook [d]: [Link]).

67
 non-zero in this direction. The X 2 component of the spacetime momentum flowing over some general time
interval I of length ∆t, is still given by Eq.(6.20) in AdSd -Schwarzschild. For the physical solution where the
string trails behind its boundary endpoint (ξ 0 (r) > 0, ξ(r) < 0), dp2 /dt = Πr2 . According to the definition
of the momentum densities Eq.(4.7),

dp2 1 √
:= Πr2 = − −g g rb G2 ν ∂b X ν
dt 2πα0
v
− 21
4 2 d d
u
1 t u r rH (v − 1) r + r rH
= 4
 G2 2 Gtt ∂r ξ (6.36)
2πα0 −rH r d+4 + rH5 v 2 r d (1 − v 2 )− d−1 + r 5 r d
H

2
1 rH v
= − 0 2 ,
2 π α l2 (1 − v 2 ) d−1

where the constraint equation Eq.(4.12) is used in the first line; while in the second line, the summation over
repeated indices is expanded and Eq.(6.35) is used. Following a similar calculation to Eq.(6.22), recognize
that G2 ν = G2 2 is the only non-zero possibility, Eq.(C.5) can be used to find the explicit entries of the
inverse induced metric g ab and the ansatz Eq.(6.5) can be used to simplify the derivative terms. The final
line follows from inputting the target spacetime metric entries (Eq.(C.2)) and the defining relation for ξ 0 (r),
Eq.(6.34) (where ξ 0 (r) > 0).

The drag force given in Eq.(6.36) reduces down to Eq.(6.22) when d = 5 – a necessary consistency

check.
Using the definition of Hawking temperature (Eq.(4.36)) to insert a unit factor T 2 × 1/T 2 , the drag force
can be rewritten
dp2 8 π l2 T 2 v
= − 2 , (6.37)
dt (d − 1)2 α0 (1 − v 2 ) d−1

where the momentum p2 (t) is given by Eq.(6.26), which holds regardless of the dimension of the spacetime.
From Eq.(6.37), in the non-relativistic limit v  1, the friction coefficient in AdSd -Schwarzschild is given
by
8 π l2 T 2
γ0AdSd = . (6.38)
(d − 1)2 α0 m

Using the Einstein-Sutherland relation Eq.(3.18) an expression for the diffusion coefficient in general d
dimensions can be obtained. Specifically,

AdSd (d − 1)2 α0 (d − 1)2 β

DHQ (d) = = √ , (6.39)
8 π l2 T 8π λ

where the second equality follows from the fundamental string length scale Eq.(4.65), and the definition
of Hawking temperature Eq.(4.36). For d = 3, complete agreement is found between this equation and
the diffusion coefficient found in the case of a finite mass heavy quark undergoing Brownian motion in the
thermal plasma (subsection (4.4), Eq.(4.123)) – a comforting consistency check for this work122 .

The heavy quark diffusion coefficient in general d dimensions (Eq.(6.39)) can be compared to the light quark
diffusion coefficient in general d dimensions (Eq.(5.127)), to find

AdSd 1 AdSd
DLQ (d) = D (d) , (6.40)
2 HQ

which agrees with Eq.(5.62) for d = 3. The factor of 1/2 in Eq.(6.40) may arise through differences in
partitioning the worldsheet for the heavy and light quark test strings. Moerman et al. postulated in
[52] that by introducing a factor a which determines the fraction of the local speed of light the boundary

122 Eq.(6.39) also agrees with the result given in Eq.(3.10) of [42]; the results from subsection (3.3) in [35]; the results from section
(3) in [22]; and the result – for d = 5 – from section (V) in [38]. The rigorous reader might also be interested in [36], a topical
study which computed an ultra-relativistic quark’s transverse momentum diffusion.

68
 endpoint of the test string falls at, a general expression can be found for the mean-squared displacement of
said endpoint (and therefore the diffusion coefficient) which interpolates between the heavy and light quark
results123 . Explicitly, the string falling endpoint’s mean-squared transverse displacement is given by

2
2aβ 3
       
1 (d − 1)β π(a + 1)t π(a − 1)t 2πat
s2small (t; a; d) = √ ln sinh2 2
sinh csch ,
λ 4π π 3 (a2 − 1)2 t3 β β β
(6.41)

which, at late times, becomes124

   
β

 4 ln 2πt , if a = 0 
  0
(d − 1)2 βt (d − 1)2 β 2
 
(βt)
h a i   3
  β
s2small (t; a; d) −→ √ 1− + √ ln 3
aβ
2 2 3 , if 0 < a < 1 + O .
4π λ 2 16π 2 λ    −1) t
 4π (a 
 t
 ln β ,
 
if a = 1 
4πt
(6.42)

The diffusion coefficient for a quark in a (d−1)-dimensional thermal plasma can be extracted from Eq.(6.42).
Specifically,
h a i (d − 1)2 β
D(a; d) = 1 − √ . (6.43)
2 8π λ

When a = 0, Eq.(6.43) reduces to the heavy quark diffusion coefficient, Eq.(6.39); and when a = 1, Eq.(6.43)
becomes the light quark diffusion coefficient, Eq.(5.127). Hence Eq.(6.41) provides a natural interpolation
between a test string set-up where the boundary endpoint is held stationary, and a test string set-up where
the boundary endpoint is allowed to fall at the local speed of light. However, values of a ∈
/ {0, 1} correspond
to a test string whose boundary endpoint is falling at velocity smaller than the local speed of light. An
external force opposing the motion of the falling string must be present in order for this to be the case125 .
This might be realised by the introduction of a flavour D7-brane with a world-volume electric field on it
at the boundary126,127 . Considering an external force acting to retard the motion of the falling string’s
endpoint is a natural and interesting extension of this work – a possible starting point for future research.

7. Conclusions and Future Outlook

In this article we present a pedagogic account of applications of the AdS/CFT correspondence for under-
standing the dynamical behaviour of probe heavy and light quarks immersed in a thermal plasma, such as
the quark-gluon plasma created in heavy-ion experiments. As a central result, we give the first self-consistent
derivation of the light quark diffusion coefficient in an arbitrary number of dimensions, correcting previous
errors and giving some of the vaguer statements found in the literature a rigorous foundation.

To summarise, the gauge/string duality is briefly introduced in section (2). Particular attention is paid to
the throat construction of anti-de Sitter spacetime as a limit of D3-brane geometry, and the justification of
the AdS/CFT conjecture. Section (3) explored the basic theory of Brownian motion and Langevin dynamics
in the boundary theory. Specifically, particles undergoing Brownian motion in the absence of an external
force can be described by the non-retarded Langevin equation which is parametrized by two constants: (i)
the friction coefficient γ0 , and (ii) the magnitude of the random force κ0 . The second fluctuation-dissipation
theorem relates these two constants. It is possible to calculate the mean-squared displacement (Eq.(3.19)),

123 The heavy and light quark set-ups to study Brownian motion are considered in sections (4) and (5), respectively.
124 See Mathematica notebook [b] ([Link]) for a derivation of Eqs.(6.41)-(6.43) – access in appendix (Appendix G).
125 The solution where a ∈/ {0, 1} is not a physically obtainable string solution in the current set-up since the Virasoro constraints
(Eq.(4.19)) are violated and, consequently, the late time next-to-leading order behaviour in Eq.(6.42) does not smoothly
interpolate between the light and heavy quark results. This can be rectified by adding in an external field which will alter
the energy-momentum tensor (Eq.(4.8)) and, by construction, cause Tab to vanish (which ensures the Virasoro constraints are
satisfied).
126 This corresponds to an additional force acting on the external Brownian particle in the boundary theory whose motion can

now be characterised by the generalised Langevin model (described in subsection (3.2)).

127 For the case of the heavy quark, adding forced motion has been studied by de Boer et al. [42], subsection (3.2).

69
from which the time dependence of s2 (t) is apparent. At early times, the Brownian particle’s behaviour
is proportional to time and the motion is √ expected to be ballistic s(t) ∼ t; while, at late times, the Brow-
nian particle motion is diffusive, s(t) ∼ t. The cross-over time sets the scale for early and late time
behaviour and is given by trelax , which represents the time it takes for a Brownian particle which had some
initial velocity at t = 0 to thermalize in the medium. The section concludes by describing the generalised
Langevin model which adapts the friction term to depend on the past trajectory of the Brownian parti-
cle and takes an additional external force acting on the system into account. In section (6) it was seen
that the generalised Langevin model can be used in a number of applications: to model heavy quark trail-
ing strings or light quark strings whose endpoint falls with a velocity that is less than the local speed of light.

Sections (4), (5) and (6) turn towards the bulk theory. In the bulk theory, the dual description of these
probe quarks are realised as test strings in an asymptotically anti-de Sitter-Schwarzschild background. Cal-
culations are computed in AdS-Schwarzschild, and then related to quantities in the boundary theory using
the AdS/CFT dictionary. In section (4) an on-mass-shell external heavy quark is modelled as a funda-
mental open string of length `0 attached at the boundary of anti-de Sitter spacetime and hanging towards
the stretched horizon. At the semiclassical level, the Hawking radiation due to the Schwarzschild black
hole environment excites the modes on the string – resulting in the string’s boundary endpoint enduring
irregular motion. This motion can be related to the Brownian motion of the external heavy quark in the
boundary gauge theory. The main results of this section include the derivation of the leading order, static
string solution in AdS3 -Schwarzschild (Eq.(4.51)); the transverse equations of motion found by expanding
the Nambu-Goto action up to quadratic order and varying this action with respect to the transverse string
worldsheet coordinates X I (Eq.(4.67)); and from there the mean-squared displacement of the test string’s
boundary endpoint, s2 (t), in AdS3 -Schwarzschild (Eq.(4.110)). The cross-over time from early to late time
dynamics is found to be independent of the initial length of the string `0 and solely dependant on the Hawk-
ing temperature β = 1/T . As expected, in the early time limit (t  β) the motion is ballistic (Eq.(4.119));
while in the late time limit (t  β) the motion is diffusive, and the diffusion coefficient can be extracted
from s2 (t)|tβ (Eq.(4.123)). We further showed that the diffusion coefficient computed in this way and
the friction coefficient computed in a completely independent manner are exactly related by the fluctuation
dissipation theorem as required for thermalizing motion.

An off-mass-shell external light quark is modelled in section (5) as an open string, initially stretched between
the AdS boundary and just above the horizon, whose AdS boundary endpoint is released to fall at the local
speed of light. The set-up is termed the Limp Noodle configuration [52]. This initial set-up is held for
an asymptotically long amount of time, allowing for a full population of thermal modes on the initially
stretched string. Using the Bars et al. method [107, 108], the worldsheet of the Limp Noodle is partitioned
into two regions and the leading order string solution for the string once the endpoint near the boundary is
released is found in AdS3 -Schwarzschild (Eq.(5.40)). Again, the transverse fluctuations on the string excited
by the Schwarzschild black hole are considered. From the equations of motion of these fluctuations, the
string falling endpoint’s mean-squared transverse displacement s2 (t) is calculated (Eq.(5.47)). The limiting
cases of s2 (t) are then examined. For the small virtuality case (small string lengths `0 compared to the
radial position of the black-brane horizon rH ), s2small (t) is analytically evaluated (Eq.(5.58)) and the early
and late time behaviour found to be ballistic and diffusive respectively. For the arbitrary virtuality case,
s2 (t) can be analytically found in the early time limit (Eq.(5.70)); while the behaviour of arbitrary virtuality
quarks at asymptotically late times (i.e. in the near-horizon region) is found to be encoded in the small
virtuality case, i.e. s2 (t)|tβ = s2small (t) (Eq.(5.76)). Since s2small (t; d) can be solved in any d ≥ 3 dimensions,
this universality of late times motivates the generalisation of the mean-squared displacement s2 (t)|tβ to
AdSd -Schwarzschild. In terms of original advancement, subsection (5.3) presents a very important part
of this article: correcting Moerman et al. [52] by presenting the proper method in which to generalise to
s2 (t; d). Finally, the diffusion coefficient in AdSd -Schwarzschild is extricated from s2 (t; d)|tβ and given in
Eq.(5.127) by
AdSd (d − 1)2 β
DLQ (d) = √ ,
16π λ
which is – to the authors’ knowledge – the first complete, consistent derivation of the light quark diffusion
coefficient in d dimensions.

In section (6), a trailing string in AdS5 -Schwarzschild which models an infinitely massive probe quark moving
with a constant velocity v in a N = 4 SYM thermal plasma is considered. By making use of the late-time
behaviour ansatz (Eq.(6.5)), the drag force on the test string is calculated in the bulk and rewritten – via
the AdS/CFT correspondence – in terms of relevant quantities in the gauge theory (Eq.(6.25)). Because the

70
external quark is moving with a constant velocity in the boundary theory, it can be modelled by the Langevin
equation adapted such that the external force K(t) acting on the system is of equal and opposite magnitude
to the random force (K(t) = −F (t)). Hence the friction coefficient γ0 can be read off from Eq.(6.27),
and related – using the Einstein-Sutherland relation – to the diffusion coefficient in d = 5 dimensions. In
the second subsection, the drag force calculation is generalised to AdSd -Schwarzschild and the diffusion
coefficient is found to be given in Eq.(6.39) by

AdSd (d − 1)2 β
DHQ (d) = √ ,
8π λ
which agrees, for d = 3, with the diffusion coefficient found by studying the transverse fluctuations on the
heavy quark’s test string in section (4) (Eq.(4.123)).

The heavy and light quark’s diffusion constants in general d dimensions are related by a factor of a 1/2
(Eq.(6.40)). This disparity may arise through the differences in partitioning the worldsheet for the heavy
and light quark test strings. As was briefly discussed towards the end of section (6), in order to understand
AdSd AdSd
the physical origin of the 1/2 factor an interpolation between DLQ (d) and DHQ (d) can be sought. This
leads to the realisation that considering test strings with boundary endpoints falling slower than the local
speed of light will only yield physically valid solutions if an external force is introduced. Considering an
external force acting to retard the motion of the falling string’s endpoint is a natural and interesting ex-
tension of this work, and a possible starting point for future research. The publications [42, 110, 111, 112]
might provide a good commencement for this exercise.

The main contribution of this article to the field of AdS/CFT calculations is presenting a definitive, con-
solidated theoretical derivation of the light and heavy quark diffusion constants in general d dimensions.
The work acts as a springboard from which students might pursue further research avenues. Some of these
that the authors have considered – besides the addition of an external electric field – are (i) repeating
these calculations using a numerical framework in order to confirm the analytic results; (ii) examining the
fluctuation-dissipation theorem in the bulk; (iii) considering next-to-leading order transverse fluctuations
on the leading order solution (this would surely prove difficult as the expanded, near-horizon metric in
AdSd -Schwarzschild would no longer be conformal in (t, r∗ ) coordinates); and (iv) considering different test
string configurations to illuminate other aspects of the thermal plasma. With regards to the latter, one such
example is studying the quantum fluctuations in the non-transverse directions of a trailing string where
the boundary endpoint is allowed to fall. Another excellent example might be the consideration of a light
quark string setup that more closely models the phenomenologically relevant situation in which a light
quark–anti-quark pair is created highly off-shell and without a pre-population of thermal modes along the
string. Determining the fluctuations on top of the leading order classical motion of this string as it evolves
in time would require knowledge of the thermal spectrum of metric fluctuations due to the presence of the
black brane.

Acknowledgements

The authors thank Steven Gubser for the idea which inspired the Limp Noodle construction. The news of
his untimely passing is tragic; he left an indelible impact on the community. May he rest in peace.
Funding information A.K.M. appreciates the support of the Mandela Rhodes Foundation, and W.A.H.
gratefully acknowledges support from the South African National Research Foundation and the South
Africa-CERN Collaboration.

71
Appendix

Appendix A. Polyakov String Equations of Motion

Working in the static gauge, the string equations of motion are derived in this appendix by calculating the
functional derivative of the Polyakov Action with respect to the string worldsheet coordinates and setting
this variation to zero. The calculation of the string equations of motion presented here follows the layout
of appendix (A) in Moerman et al.’s exposition on light quark Brownian motion [52].

Using the definition Eq.(4.7), the Polyakov Action Eq.(4.3) is rewritten in terms of the canonical momentum
densities
Z
1
SP = d2 σ Πaµ (t, σ) ∂a X µ (t, σ) . (A.1)
2 M

Determining the functional derivative of SP with respect to X µ and setting this variation to vanish, yields

Z
1 h
i
d2 σ δΠaµ ∂a X µ + Πaµ ∂a δX µ


0 = δX SP =
2 M
Z (A.2)
1 2
h
i
δΠaµ µ
∂a Πaµ µ µ
∂a Πaµ



= d σ ∂a X + δX − δX ,
2 M

where the product rule is used in the second line.

The functional derivative of the momentum densities with respect to X µ gives two terms
1 √ h i
δX Πaµ = − ab ν ρ ν


−γ γ G ∂
µν b δX + ∂ G
ρ µν δX ∂b X , (A.3)
2πα0

since, in a curved background, the worldsheet coordinates X µ as well as
the spacetime metric Gµν (which
also depends on X µ ) are varied. Hence the first term in Eq.(A.2), δ Πaµ ∂a X µ , becomes

−1 √  −1 √ 
δX Πaµ ∂a X µ = −γ γ ab Gµν ∂b δX ν ∂a X µ + ∂ρ Gµν δX ρ −γ γ ab ∂b X ν ∂a X µ



2πα 0 2πα 0

−1 √ ab ν µ ρ
 −1 √ ab ν

Gγν ∂a X µ


= −γ γ Gµν ∂b X ∂a δX + ∂ρ Gµν δX −γ γ Gγν ∂ b X
2πα0 2πα0
= Πaµ ∂a δX µ + ∂ρ Gµν δX ρ Πaγ Gγν ∂a X µ

1 
= ∂a Πaµ δX µ − δ X µ ∂a Πaµ + 2 δX ρ Gγν ∂ρ Gµν Πaγ ∂a X µ



2
 1 γν

= ∂a Πaµ δX µ − δ X µ ∂a Πaµ + 2 δX ρ G ∂ρ Gµν + ∂µ Gρν − ∂ν Gρµ Πaγ ∂a X µ ,



2
(A.4)

where the definition of the momentum densities Eq.(4.7) is used in the third line, and the product rule is
used in the fourth line. In the final line two extra terms are added. This can be done since
1 √ h
i
Πaγ ∂a X µ ∂µ Gρν − ∂ν Gρµ = − ab ν µ


−η Gµν η ∂b X ∂a X ∂ µ Gρν − ∂ ν Gρµ = 0, (A.5)
2πα0

where the first equality holds in the conformal gauge (γ ab = η ab ) – which, as discussed in subsection (4.1),
is chosen to eliminate the Weyl invariance in the Polyakov action. To prove Eq.(A.5), notice that the
Minkowski metric can be expanded

η ab ∂a X µ ∂b X ν ∂µ Gρν − ∂ν Gρµ = ∂τ X µ ∂τ X ν ∂µ Gρν − ∂ν Gρµ + ∂σ X µ ∂σ X ν ∂µ Gρν − ∂ν Gρµ

72
and that for each set of reflective cases (µ = 1, ν = 2 and µ = 2, ν = 1), the terms generated by the
first case will cancel those generated by the second. For the identical case

(which doesn’t have a reflective
partner case), i.e. µ = ν = 1 or µ = ν = 2, the bracket ∂µ Gρν − ∂ν Gρµ vanishes.

The Christoffel symbols are now able to be defined

1 αγ
Γα


µν := G ∂µ Gνγ + ∂ν Gµγ − ∂γ Gµν . (A.6)
2

These symbols are a set of numbers which characterise a metric connection. A metric connection defines
precisely how distances are measured on a surface, by invoking the notion of consistently transporting data
in a parallel manner along a family of curves [96].

With this definition, Eq.(A.4) becomes

δX Πaµ ∂a X µ = ∂a Πaµ δX µ − δX µ ∂a Πaµ + 2 δX ρ Γγρµ Πaγ ∂a X µ .

(A.7)

Finally, inputting Eq.(A.7) into Eq.(A.2) returns

Z
1 h
i
d2 σ − 2 δX µ ∂a Πaµ + 2 δX ρ Γγρµ Πaγ ∂a X µ + 2 ∂a Πaµ δX µ


0 = δX SP =
2 M
Z h i Z
2 µ a γ a ν
d2 σ ∂a Πaµ δX µ


=− d σ δX ∂a Πµ − Γµν Πγ ∂a X +
M M
Z h i Z (A.8)
2 µ a γ a ν b
ba Πaµ µ


=− d σ δX ∂a Πµ − Γµν Πγ ∂a X + dσ δX
M ∂M
Z h i Z τf h i
σ=σf
=− d2 σ δX µ ∂a Πaµ − Γγµν Πaγ ∂a X ν + dτ δX µ Πaµ σ=0
,
M 0

where Stokes’ Theorem in d-dimensions [96] is used in the third line. In the final line the coordinates on
the worldsheet parameter space are chosen to be σ a = (t, σ)a where σ a ∈ M = [0, tf ] × [0, σf ], and one of
the integrals of the last term is simplified by using δX µ |t∈{0,tf } = 0.

Choosing the boundary conditions

σ=σf
δX µ Πaµ σ=0
= 0, (A.9)

the last term in Eq.(A.8) disappears. Hence, the string equations of motion are given by

0 = ∂a Πaµ − Γα ν a a
µν ∂a X Πα =: ∇a Πµ , (A.10)

which is the string analogue of the geodesic equation for a point particle [96].

Appendix B. The Virasoro Constraints and String Equations of Motion in Isothermal Coor-
dinates

Since the metric of any (1 + 1)-dimensional subspace can be transformed into a conformally flat metric
0
[107, 108] a new, isothermal coordinate system is introduced. These isothermal coordinates y µ (σ + , σ − ) are
0
given by Eq.(5.2), while the corresponding string mapping functions Y µ (σ + , σ − ) are defined in Eq.(5.8).
Notice that light-cone coordinates have been chosen for the parameter space. In this appendix, the Vira-
soro constraint equations and the string equations of motion, Eqs.(4.19, 4.20), are rewritten in terms of
0
Y µ (σ + , σ − ).

In the conformal gauge, the Virasoro constraint equations (Eqs.(4.19)) become

73
0 = G ηµν ∂± X µ ∂± X ν

= −G ∂± X 0 ∂± X 0 + G ∂± X 1 ∂± X 1

G h  00 0
  0 0
i G h  00 0
  0 0
i
= − ∂± Y + Y 1 ∂± Y 0 + Y 1 + ∂± Y − Y 1 ∂± Y 0 − Y 1 (B.1)
2 2
0 0
= −2 G∂± Y 0 ∂± Y 1

= ∂± Y 0 ∂± Y 1 ,

where µ, ν ∈ {0, 1} are spacetime coordinates, ∂± := ∂σ∂± , and the definition of the new mapping functions
Eq.(5.8) is used in the third line. In the final line, recognise that G is defined as a non-zero scalar function
and that the indices have been renamed (µ0 = µ). This is precisely the Virasoro constraint equations given
in subsection (5.1.1), Eq.(5.13).

The string equations of motion, Eq.(4.20), become

∂± Π± α ν ±
µ − Γµν ∂± X Πα = 0 , (B.2)

in the light-cone coordinate frame. This breaks into two equations of motion, since µ = (0, 1). For µ = 0,
the string equations of motion are
− ν −
∂+ Π+ α ν + α



0 − Γ0ν ∂+ X Πα + ∂− Π0 − Γ0ν ∂− X Πα = 0. (B.3)

The first bracket in Eq.(B.3) can be simplified

∂+ Π+ α ν +
0 − Γ0ν ∂+ X Πα
   
1 0 α ν 1 α
= ∂+ − G ∂− X − Γ0ν ∂+ X G ηαα ∂− X
2πα0 2πα0
1 
−∂+ G ∂− X 0 + Gηαα −Γα 0 α α 1 α



= 00 ∂+ X ∂− X − Γ01 ∂+ X ∂− X
2πα0
1 
−∂+ G ∂− X 0 + G Γ000 ∂+ X 0 ∂− X 0 − Γ100 ∂+ X 0 ∂− X 1 + Γ001 ∂+ X 1 ∂− X 0 − Γ101 ∂+ X 1 ∂− X 1 ,



= 0
2πα
(B.4)

where, in the first line, the canonical momentum densities Eq.(5.9) are used and the spacetime metric
simplifies due to the conformal gauge choice (i.e. Gµν (x) = G ηµν , Gµν (x) = 1/G η µν ). Further, the
Christoffel symbols, defined in Eq.(4.21), can be calculated

1 0γ 1 00 1
Γ000 = η (η0γ ∂0 G + η0γ ∂0 G − η00 ∂γ G) = η (− ∂0 G − ∂0 G + ∂0 G) = (∂0 G)
2G 2G 2G
1 1γ 1 11 1
Γ100 = η (η0γ ∂0 G + η0γ ∂0 G − η00 ∂γ G) = η (∂1 G) = (∂1 G)
2G 2G 2G
(B.5)
1 0γ 1 00 1
Γ001 = η (η1γ ∂0 G + η0γ ∂1 G − η01 ∂γ G) = η (∂1 G) = (∂1 G)
2G 2G 2G
1 1γ 1 11 1
Γ101 = η (η1γ ∂0 G + η0γ ∂1 G − η01 ∂γ G) = η (∂0 G) = (∂0 G) .
2G 2G 2G

74
Hence Eq.(B.4) becomes

 
1 0

1 0 0 0 1 1 0 1 1


−∂+ G ∂− X + ∂ 0 G ∂+ X ∂− X − ∂ 1 G ∂+ X ∂− X + ∂1 G ∂+ X ∂ − X − ∂0 G ∂+ X ∂ − X ,
2πα0 2
(B.6)

∂
where ∂µ := ∂X µ for µ = (0, 1). The second bracket in Eq.(B.3) can be similarly simplified

∂− Π− α ν −
0 − Γ0ν ∂− X Πα
   
1 0 α ν 1 α
= ∂− − G∂ + X − Γ ∂
0ν − X Gη ∂
αα + X
2πα0 2πα0
1  0

α 0 α α 1 α


= −∂ − G∂+ X + Gηαα −Γ 00 ∂ − X ∂ + X − Γ01 ∂ − X ∂ + X
2πα0
1 
−∂− G∂+ X 0 + G Γ000 ∂− X 0 ∂+ X 0 − Γ100 ∂− X 0 ∂+ X 1 + Γ001 ∂− X 1 ∂+ X 0 − Γ101 ∂− X 1 ∂+ X 1



= 0
2πα
 
1 0

1 0 0 0 1 1 0 1 1


= −∂− G∂+ X + ∂0 G ∂− X ∂+ X − ∂1 G ∂− X ∂+ X + ∂1 G ∂− X ∂+ X − ∂0 G ∂− X ∂+ X ,
2πα0 2
(B.7)

where the calculated Christoffel symbols (Eq.(B.5)) are used in the final line. Using the simplifications for
the first and second brackets (Eqs.(B.6, B.7)), the first string equation of motion Eq.(B.3) becomes

−∂+ G ∂− X 0 − ∂− G ∂+ X 0 + ∂0 G ∂+ X 0 ∂− X 0 − ∂0 G ∂+ X 1 ∂− X 1 = 0 .



(B.8)

Following a similar calculation as before, the second string equation of motion (where µ = 1) is
− ν −
∂ + Π+ α ν + α



1 − Γ1ν ∂+ X Πα + ∂− Π1 − Γ1ν ∂− X Πα = 0. (B.9)

Again, the first bracket in Eq.(B.9) can be simplified

∂+ Π+ α ν +
1 − Γ1ν ∂+ X Πα
   
1 1 α ν 1 α
= ∂+ G ∂− X − Γ1ν ∂+ X G ηαα ∂− X
2πα0 2πα0
1 
∂+ G ∂− X 1 + Gηαα −Γα 0 α α 1 α



= 0 10 ∂+ X ∂− X − Γ11 ∂+ X ∂− X
2πα
1 
∂+ G ∂− X 1 + G Γ010 ∂+ X 0 ∂− X 0 − Γ110 ∂+ X 0 ∂− X 1 + Γ011 ∂+ X 1 ∂− X 0 − Γ111 ∂+ X 1 ∂− X 1 ,



= 0
2πα
(B.10)

where, in the first line, the canonical momentum densities Eq.(5.9) are used and the spacetime metric
simplifies due to the conformal gauge choice. The Christoffel symbols now need to be calculated. However
first note that the Christoffel symbols are symmetric on their lower indexes [96], i.e. Γγαβ = Γγβα . Therefore
Γ010 = Γ001 and Γ110 = Γ101 (the latter of which have both been calculated in Eq.(B.5)). The remaining two
Christoffel symbols can be calculated

1 0γ 1 00 1
Γ011 = η (η1γ ∂1 G + η1γ ∂1 G − η11 ∂γ G) = η (− ∂0 G) = (∂0 G)
2G 2G 2G
(B.11)
1 1γ 1 11 1
Γ111 = η (η1γ ∂1 G + η1γ ∂1 G − η11 ∂γ G) = η (∂1 G + ∂1 G − ∂1 G) = (∂1 G) .
2G 2G 2G

75
Hence Eq.(B.10) becomes

 
1 1

1 0 0 0 1 1 0 1 1


∂ + G ∂− X + ∂ 1 G ∂+ X ∂− X − ∂0 G ∂+ X ∂− X + ∂0 G ∂+ X ∂ − X − ∂1 G ∂+ X ∂ − X .
2πα0 2
(B.12)

The second bracket in Eq.(B.9) can be similarly simplified

∂− Π− α ν −
1 − Γ1ν ∂− X Πα
   
1 1 α ν 1 α
= ∂− G ∂+ X − Γ ∂
1ν − X G η ∂
αα + X
2πα0 2πα0
1  1

α 0 α α 1 α


= ∂− G ∂+ X + Gηαα −Γ 10 ∂− X ∂+ X − Γ11 ∂− X ∂+ X
2πα0
1 
∂− G ∂+ X 1 + G Γ010 ∂− X 0 ∂+ X 0 − Γ110 ∂− X 0 ∂+ X 1 + Γ011 ∂− X 1 ∂+ X 0 − Γ111 ∂− X 1 ∂+ X 1



= 0
2πα
 
1 1

1 0 0 0 1 1 0 1 1


= ∂− G ∂+ X + ∂1 G ∂− X ∂+ X − ∂0 G ∂− X ∂+ X + ∂0 G ∂− X ∂+ X − ∂1 G ∂− X ∂+ X ,
2πα0 2
(B.13)

where the calculated Christoffel symbols (Eqs.(B.5, B.11)) are used in the final line. Using the simplifications
for the first and second brackets (Eqs.(B.12, B.13)), the second string equation of motion Eq.(B.9) becomes

∂+ G ∂− X 1 + ∂− G ∂+ X 1 + ∂1 G ∂+ X 0 ∂− X 0 − ∂1 G ∂+ X 1 ∂− X 1 = 0 .



(B.14)

Now, the two string equations of motion (Eqs.(B.8, B.14)) can be rewritten in terms of the new string
0
embedding functions Y µ (σ + , σ − ). Combining Eq.(B.8) and Eq.(B.14) yields

−∂+ G ∂− X 0 − X 1 − ∂− G ∂+ X 0 − X 1 + (∂0 G + ∂1 G) ∂+ X 0 ∂− X 0 − ∂+ X 1 ∂− X 1 = 0 ,




 

(B.15)

∂
where the derivatives are defined as ∂µ := ∂X µ for µ = (0, 1). By the chain rule

∂0 G = (∂00 G + ∂10 G) and ∂1 G = (∂00 G − ∂10 G) , (B.16)

0
where ∂µ0 := ∂Y∂µ0 for µ0 = (00 , 10 ). Using Eq.(B.16) and the definition of Y µ (σ + , σ − ) (Eq.(5.8)), the
equation of motion Eq.(B.15) becomes

√ h  √

0
  0
i 1  0 0
  0 0

0 = − 2 ∂+ G ∂− Y 1 + ∂− G ∂+ Y 1 ) + 2 ∂00 G ∂+ Y 0 + Y 1 ∂− Y 0 + Y 1
2

1  0 0
  0 0

− ∂+ Y 0 − Y 1 ∂− Y 0 − Y 1
2

 0
  0
 1 0 0 0 0 0 0 0 0
= ∂+ G ∂− Y 1 + ∂− G ∂+ Y 1 + ∂00 G ∂+ Y 0 ∂− Y 0 + ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0 + ∂+ Y 1 ∂− Y 1
2

00 00 00 10 10 00 10 10
− ∂+ Y ∂− Y + ∂+ Y ∂− Y + ∂+ Y ∂− Y − ∂+ Y ∂− Y

= ∂+ G∂− Y 1 + ∂− G∂+ Y 1 − (∂0 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0 ,







(B.17)

76
 where the indices have been renamed (µ0 = µ) in the last line. This is precisely one of the string equations of
motion given in subsection (5.1.1), Eq.(5.13). In order to find the other string equation of motion, subtract
Eq.(B.8) from Eq.(B.14) to yield

0 = −∂+ G ∂− X 0 + X 1 − ∂− G ∂+ X 0 + X 1
+ (∂0 G − ∂1 G) ∂+ X 0 ∂− X 0 − ∂+ X 1 ∂− X 1





 

√ h  √

00
 
00
i 1  0 0
  0 0

− 2 ∂+ G ∂− Y + ∂− G ∂+ Y ) + 2 ∂10 G ∂+ Y 0 + Y 1 ∂− Y 0 + Y 1
2

1  0 0
  0 0

− ∂+ Y 0 − Y 1 ∂− Y 0 − Y 1
2

= ∂+ G∂− Y 0 + ∂− G∂+ Y 0 − (∂1 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0 .







(B.18)

The Virasoro constraints Eq.(B.1) and the string equations of motion Eqs.(B.17, B.18) agree with the
equations given by Bars et al. in [107, 108] where u = Y 0 and v = Y 1 . Further, using the product and
chain rules, Eq.(B.17) becomes

0 = (∂+ G) ∂− Y 1 + G∂+ ∂− Y 1 + (∂− G) ∂+ Y 1 + G∂− ∂+ Y 1

− (∂0 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0






= (∂0 G) ∂+ Y 0 + (∂1 G) ∂+ Y 1 ∂− Y 1 + (∂0 G) ∂− Y 0 + (∂1 G) ∂− Y 1 ∂+ Y 1 + 2 G∂+ ∂− Y 1








− (∂0 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0






= G∂+ ∂− Y 1 + (∂1 G) ∂+ Y 1 ∂− Y 1 ,

(B.19)

and Eq.(B.18) becomes

0 = (∂+ G) ∂− Y 0 + G∂+ ∂− Y 0 + (∂− G) ∂+ Y 0 + G∂− ∂+ Y 0

− (∂1 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0






= (∂0 G) ∂+ Y 0 + (∂1 G) ∂+ Y 1 ∂− Y 0 + (∂0 G) ∂− Y 0 + (∂1 G) ∂− Y 1 ∂+ Y 0 + 2 G∂+ ∂− Y 0








− (∂1 G) ∂+ Y 0 ∂− Y 1 + ∂+ Y 1 ∂− Y 0






= G∂+ ∂− Y 0 + (∂0 G) ∂+ Y 0 ∂− Y 0 ,

(B.20)

which agrees exactly with the string equations of motion given by Moerman et al. in [52].

Appendix C. Energy of a Test String in an AdS-Schwarzschild Background

At t = 0 the test string set-up for the heavy quark agrees with the test string set-up for the light quark,
since the string is initially static in both cases – i.e. X I = 0 (where I ∈ (2, 3, ... , d − 1)) at time t = 0, with
µ
embedding functions X µ (t, σ) = (t, r(t, σ), 0) . The on-mass-shell heavy quark’s mass or the off-mass-shell
128
light quark’s mass can be calculated by finding the total energy of the relevant test string in AdSd -
Schwarzschild. In order to calculate the string’s total energy, the configuration of the string is chosen to be

128 Particles
are on the mass shell, or simply on-mass-shell, if their behaviour satisfies Einstein’s energy and momentum relation
E 2 = (pc)2 + (2mc)2 . Particles whose behaviour violates this relation are known as off-mass-shell. In the case of the light
quark it starts at t = 0 as an off-mass-shell particle (corresponding to the initial static string), radiates energy as it travels
through the thermal medium (string contracts as the boundary endpoint falls at the local speed of light), and finally stops
radiating as it becomes an on-mass-shell particle. This appendix is focused on calculating the mass (or virtuality) of the light
quark as an initially off-mass-shell particle; and the mass of an on-mass-shell heavy quark.

77
stretched between the radial positions r = 0 and r = `0 . This choice will simplify the calculation.

The AdSd -Schwarzschild metric in d dimensions is given by

ds2d := Gµν dxµ dxν

~ I2
= Gtt dt2 + Grr dr2 + GII dX
(C.1)
r2  2
~ 2 + l dr ,
 2
= 2
− h(r; d) dt2 + dX I 2
l r h(r; d)

where t ∈ [0, ∞) is the temporal coordinate, r ∈ [0, ∞) is the radial coordinate, and the transverse spatial
~ I = X 2 , X 3 , ...., X (d−1) ∈ Rd−2 . Further, l ∈ R+ is the curvature radius of


directions are denoted by X
AdSd and S d ; and the blackening factor of the Schwarzschild black hole situated at the horizon, h(r; d), is
given by Eq.(4.35). Explicitly, the target spacetime metric Gµν is given by
 2
 
− rl2 h(r; d) 0 0 0 0 0
  2  
l 1
0 0 0 0 0 
 
 r 2 h(r;d)
  2 
 r 
 0 0 l2 0 0 0 
Gµν =  , (C.2)
 
 2 .
 0 0 0 r . . 0 
 l 2 
 
 .. .. 
 0 0 0 . . 0 
  2 
r
0 0 0 0 0 l2

where µ, ν index over (t, r, 2, ... , d − 1). The induced worldsheet metric gab , defined in Eq.(4.4), takes the
explicit form
 
gtt gtσ
gab =  , (C.3)
gσt gσσ

since (t, σ) are the two parameter space coordinates in the static gauge (τ = t). In order to determine
the explicit entries of the induced metric gab , the definition Eq.(4.4) is used to calculate each component.
Specifically,
 
Gtt + GII ẊI2 GII ẊI XI0
gab =  , (C.4)
GII XI0 ẊI r02 Grr + GII XI0 2

where Ẋ ≡ ∂t X and X 0 ≡ ∂σ X; I indexes over the transverse directions XI = X 2 , X 3 , ...., X (d−1) ; and

the fact that Gµν (Eq.(C.2)) is a diagonal matrix is used.

For later use, the inverse metric g ab (where g ab ≡ (gab )−1 ) is also written out explicitly
 
ab 1 r02 Grr + GII XI0 2 −GII ẊI XI0
g =  . (C.5)
det(gab ) −GII XI0 ẊI Gtt + GII ẊI2

The determinant of the induced metric is

78
 g := det (gab ) = r02 Grr + GII XI0 2 Gtt + GII ẊI2 − G2II ẊI2 XI0 2

= r02 Grr Gtt + r02 Grr GII ẊI2 + Gtt GII XI0 2 + GII
2
XI0 2 ẊI2 − GII
2
ẊI2 XI0 2 (C.6)
02
 r 1 
= Grr Gtt r02 + GII ẊI2 + GII XI0 2 .
Gtt Grr

Now that the ground work has been laid, the total energy of the static string can be calculated
Z `0
E = dσ Πτt
0

`0
√
Z
1
= − dσ −γ γ τ b Gtν ∂b X ν
2πα0 0

`0
√
Z
1
= − dσ −g g tb Gtt ∂b X t
2πα0 0
Z `0
√ r2
 
1 tt t tσ t


= − dσ −g g ∂t X + g ∂σ X − 2 h(r; d) (C.7)
2πα0 0 l
Z `0  2
√

1 tt r
= dσ −g g h(r; d)
2πα0 0 l2
Z `0
√
r2
 
1 02 02
= dσ −g r G rr + G II X I h(r; d)
2πα0 0 l2
Z `0  2  2
√

1 02 l 1 r
= dσ −g r h(r; d) ,
2πα0 0 r2 h(r; d) l2

where the constraint equation Eq.(4.12) and the static gauge choice (τ = t) are used in the third line, and
Eq.(C.2) is used in the fourth line. The fifth line follows from remembering the string is static (XI = 0),
and Eq.(C.5) is used in the sixth line.

Using Eq.(C.6), the total energy of the static string (Eq.(C.7)) becomes

s
`0
r02
Z
1 02
 1 
E = dσ r −Grr Gtt r02 + GII ẊI2 + GII XI0 2
2πα0 0 Gtt Grr
Z `0
1 p
(C.8)
= dσ r02 −r02 Grr Gtt
2πα0 0
Z `0
1
= dσ ∂σ3 r ,
2πα0 0

where Eq.(C.6) is used in the first line, the second line follows again from XI = 0, and Eq.(C.2) is used
the final line.

From Eq.(C.8) it is apparent that the energy of the string is contingent on how the radial coordinate r
depends on the worldsheet parameter space coordinate σ. In the static gauge (which is employed in sections
(4) and (5)) the energy of the test string is easily found in AdS3 -Schwarzschild by using the identification
Eq.(4.53) and solving the integral in Eq.(C.8). Since the tortoise coordinate in anti-de Sitter-Schwarzschild
spacetime can only be inverted for d = 3, an equivalent identification to Eq.(4.53) in AdSd -Schwarzschild
can not be found. However, a different gauge choice can be made. Taking, for example, the gauge choice129

129 This gauge choice is also made in [22, 35].

79
 in section (6) where τ = t and σ = r are the worldsheet coordinates, Eq.(C.8) becomes
1  
E = `0 − 0 , (C.9)
2πα0
in AdSd -Schwarzschild. Therefore
`20
E2 ≡ m2q = , (C.10)
4π 2 α02

where the first equivalence follows from the AdS/CFT correspondence: the energy of a test string in the bulk
theory is equivalent to the mass of the given probe quark in the boundary theory. Notice that Eq.(C.10)
agrees with Eq.(4.32), which is a reflection of the fact that the spacetime in the bulk theory is curved
precisely along the additional radial direction r, and not along the temporal or transverse directions. Since
the boundary theory does not include this radial direction, from the perspective of the probe quark its mass
will be the same whether the string is in Minkowski or AdS-Schwarzschild spacetime130 . Considering that
√ l
α0 ≡ , (C.11)
λ1/4

where l ∈ R+ is the radius of curvature of AdSd and λ = gY2 M Nc is the ’t Hooft coupling (see table (1),
page 7); the quark’s mass131 becomes

λ `20
m2q = . (C.12)
4 π 2 l4

Alternatively, Eq.(C.12) can be rewritten as

√ √ √   √
λ `0 λ (rs + `0 − rs ) λ rs λ
mq = = = r̃0 − ≈ r̃0 , (C.13)
β rH β rH β rH β

where the first equality follows from using the definition of the Hawking temperature Eq.(4.36); the second
equality from Eq.(4.89); and the approximation from the definition of the stretched
√ horizon rs = (1 + )rH
(where 0 <   1). Using the definition of the Hawking temperature yields mq = λ r̃0 T . Eq.(C.13) might
be a convenient form of the quark mass mq in some cases.

Appendix D. Nambu-Goto String Equations of Motion for Transverse Fluctuations

The calculation of the string equations of motion for the transverse fluctuations presented in this appendix,
follows the layout of appendix (B) in Moerman et al.’s exposition on light quark Brownian motion [52].
Working in the static gauge, the transverse string equations of motion are derived by varying the effective
action for the transverse fluctuations with respect to the transverse string worldsheet coordinates, and set-
ting this functional variation to zero.

Using the definition Eq.(4.7), the effective transverse action Eq.(4.62) can be written in terms of the trans-
verse canonical momentum densities
Z
(2) 1
SN G = d2 σ ΠaI (t, σ) ∂a X I (t, σ) . (D.1)
2 M

(2)
Determining the functional derivative of SN G with respect to X I and setting this variation to vanish, results
in
Z
(2) 1 h
i
d2 σ δΠaI ∂a X I + ΠaI ∂a δX I


0 = δX SN G =
2 M
Z (D.2)
1 2
h
a

I a I

a

I
i
= d σ δΠ I ∂a X + ∂a Π I δX − ∂a Π I δX ,
2 M

130 In an AdS/CFT context, this implies that the energy of the string is equivalent in Minkowski or AdS-Schwarzschild spacetime.
131 Remembering this is either the initial, off-mass-shell mass for the light quark or the on-mass-shell mass for the heavy quark.

80
where the product rule is used in the second line.

The functional derivative of the momentum densities with respect to X I is given by

1 √
δX ΠaI = − −g g ab GIJ ∂b δX J ,



0 X0µ
(D.3)
2πα

where the spacetime metric Gµν does not need to be varied

since it is independent of the transverse directions
X I (Eq.(C.2)). Hence the first term in Eq.(D.2), δΠaI ∂a X I , becomes

√
 
1
δX ΠaI I ab J
∂a X µ




∂a X = − −g g GIJ X µ ∂b δX
2πα0 0

√
 
1 ab µ
∂b δX J



= − −g g GIJ X µ ∂a X
2πα 0 0 (D.4)

= ΠbJ ∂b δX J

= ∂b ΠbJ δX J − ∂b ΠbJ δX J ,

where the definition of the momentum densities Eq.(4.66) are used in the third line, and the product rule
is used in the final line. Inputting Eq.(D.4) into Eq.(D.2), yields

Z
(2) 1 h i
d2 σ ∂b ΠbJ δX J − ∂b ΠbJ δX J + ∂a ΠaI δX I − ∂a ΠaI δX I





0 = δX SN G =
2 M
Z Z
d2 σ δX I ∂a ΠaI + d2 σ ∂a ΠaI δX I



= −
M M
Z Z (D.5)
2 I a b
ba ΠaI I



= − d σ δX ∂a Π I + dσ δX
M ∂M
Z Z τf h i
2 I σ=σf
∂a ΠaI dτ ΠaI δX I


= − d σ δX + σ=0
,
M 0

where, in the second line, the indices have been renamed and the terms grouped; and Stokes’s Theorem
in d-dimensions [96] has been used in the third line. In the final line the coordinates on the worldsheet
parameter space are chosen to be σ a = (t, σ)a where σ a ∈ M = [0, tf ] × [0, σf ], and one of the integrals of
the last term is simplified by using δX I |t∈{0,tf } = 0.

Choosing the boundary conditions

σ=σf
ΠaI δX I σ=0
= 0, (D.6)

the last term in Eq.(D.5) disappears. Hence, the string equations of motion for the transverse fluctuations
are given by

0 = ∂a ΠaI = ∇a ΠaI , (D.7)

which agrees with Eq.(4.67). The second equality follows because ∇a Πaµ := ∂a Πaµ − Γα ν a
µν ∂a X Π µ in curved
α
spacetime, but the second term vanishes since the Christoffel symbols are identically zero (ΓIJ = 0) due to
symmetry between each of the transverse X I directions.

81
 Appendix E. Canonical Commutation Relations and Normalised Basis

This appendix aims to fix the normalization constant Aω , thereby completely determining the general
solution for the transverse equations of motion (Eq.(4.92)) discussed in subsection (4.3.2). In order to
determine Aω the theory is quantized: the scalar field X(t, σ) and its canonically conjugate momentum
P t (t, σ) are promoted to operators (Eqs.(4.93, 4.94) respectively), and suitable commutation relations are
imposed

h i δ(σ − σ 0 )
X̂(t, σ), nt P̂ t (t, σ 0 ) = i δ(σ, σ 0 ) = i √ ,
Σ g̃|Σ
(E.1)
h i h i
0
X̂(t, σ), X̂(t, σ ) = nt P̂ t (t, σ), nt P̂ t (t, σ 0 ) = 0,
Σ Σ

where Σ is a Cauchy hypersurface in the xµ = (t, r)µ part of spacetime that is chosen to be a constant √ time
surface132 , g̃ is the induced metric on Σ, and nµ is the future pointing normal to Σ (where nµ = δµt / −g̃tt ).
The commutation relations described in Eq.(E.1) are known as equal time commutation relations – they
encode the expectation that simultaneous measurements at different points on the string do not interfere
with each other [92].

Additionally, canonical creation and annihilation commutation relations on the Fourier coefficient operators
(âω , â†ω ) are enforced:

h i
âω , â†ω0 = 2πδ(ω − ω 0 ) ,
Σ
h i h i (E.2)
âω , âω0 = â†ω , â†ω0 = 0.
Σ Σ

Consistency between hthe commutation relations

i Eq.(E.1) and Eq.(E.2) is required133 . To this end, the
t 0
commutator bracket X̂(t, σ), nt P̂ (t, σ ) is calculated from the definitions Eqs.(4.93, 4.94), and the
Σ
result is equated to the relevant right hand side of Eq.(E.1). Specifically,

h i
X̂(t, σ), nt P̂ t (t, σ 0 ) = X̂(t, σ)|Σ nt P̂ t (t, σ 0 )|Σ − nt P̂ t (t, σ 0 )|Σ X̂(t, σ)|Σ ,
Σ
Z ∞
i r2 δtt dω dω 0
=− √ ω A ω Aω 0 ×
2πα0 l2 g̃|Σ 0 (2π)2
 0 0
fω (σ)e−iωt âω + fω∗ (σ)eiωt âω† fω (σ 0 )e−iω t âω0 − fω∗ (σ 0 )eiω t âω†0
 

(E.3)
0 0 
− fω (σ 0 )e−iω t âω0 − fω∗ (σ 0 )eiω t âω†0 fω (σ)e−iωt âω + fω∗ (σ)eiωt âω†
 

∞
i r2 1 dω dω 0
Z
=− √ ω A ω Aω 0 ×
2πα0 l2 g̃|Σ 0 2π
 0 0

− fω (σ)fω∗ (σ 0 ) ei(ω −ω)t δ(ω − ω 0 ) − fω (σ 0 )fω∗ (σ) ei(ω−ω )t δ(ω 0 − ω) ,

where the third line follows because the cross terms (âω âω0 − âω0 âω ) and (â†ω â†ω0 − â†ω0 â†ω ) vanish (using
the second commutation relation in Eq.(E.2)). The first commutation relation in Eq.(E.2) is also used in
simplification. Collecting terms, and converting to (t, r∗ ) coordinates using Eq.(4.53), yields

132 Giving initial conditions on this hypersurface determines the future (and past) evolution uniquely.
133 In order to ensure consistent quantization of the theory.

82
 Z ∞
h i i r2 10 dω
t
∴ X̂(t, σ), nt P̂ (t, σ ) = − √ ωA2ω (−fω (σ)fω∗ (σ 0 ) − fω (σ 0 )fω∗ (σ))
Σ 2πα0 l2 g̃|Σ 0 2π
Z ∞
2
rH (rs∗ + σ 0 )
 
i rH 1 dω
= 0 2
coth2
2
√ ω A2ω (fω (σ)fω∗ (σ 0 ) + fω (σ 0 )fω∗ (σ)) .
2πα l l g̃|Σ 0 2π
(E.4)

Inputting Eqs.(4.84, 4.86, 4.91), the commutator becomes134

2 Z ∞
h
t
i 1 0 i rH dω 2

−i(σ+σ 0 )ω i(σ−σ 0 )ω i(σ+σ 0 )ω i(σ 0 −σ)ω

X̂(t, σ), nt P̂ (t, σ ) = √ ωAω e + e + e + e , (E.5)
Σ g̃|Σ πα0 l2 0 2π

where the near-horizon limit (r → rH ≡ rs∗ → −∞ ≡ r̃0 → 1) has been taken after multiplication135 .
Note that in the near-horizon limit,
 
2 rH (rs∗ + σ)
coth → 1, as rs∗ → −∞ . (E.6)
l2

The commutator Eq.(E.5) can be further simplified

h i
X̂(t, σ), nt P̂ t (t, σ 0 )
Σ

2
Z ∞ Z ∞ i
1 i rH dω h 0 dω 0
i
h 0 0
=√ 0
ωA2ω e−i(σ+σ )ω + e−i(σ−σ )ω +
ωA2ω ei(σ+σ )ω + ei(σ−σ )ω
g̃|Σ πα l2
0 2π 0 2π

1 i rH2
 Z 0
dω h i Z ∞ dω h i
2 −i(σ+σ 0 )ω −i(σ−σ 0 )ω 2 i(σ+σ 0 )ω i(σ−σ 0 )ω
=√ − ωAω e +e + ωAω e +e
g̃|Σ πα0 l2 ∞ 2π 0 2π

1 i rH2
 Z 0
(−dω) h i Z ∞ dω h i
2 i(σ+σ 0 )ω i(σ−σ 0 )ω 2 i(σ+σ 0 )ω i(σ−σ 0 )ω
=√ − (−ω)(−Aω ) e +e + ωAω e +e
g̃|Σ πα0 l2 −∞ 2π 0 2π
2 Z ∞
1 i rH dω h 0 0
i
=√ 0 2
ωA2ω ei(σ+σ )ω + ei(σ−σ )ω .
g̃|Σ πα l −∞ 2π
(E.7)

In order for Eq.(E.7) to be consistent with the definition of the commutation relation Eq.(E.1), the normal-
ization constant needs to be defined as
r
l πα0 β
Aω := = √ , (E.8)
rH ω 2 πω λ1/4

where the second equality follows from using the definition of the AdS radius of curvature l (Eq.(4.65)), and
2
the relation l = βrH /2π from Eq.(4.36).

To prove that the commutation relations Eq.(E.1) now hold, input the newly defined Aω into Eq.(E.7).
Explicitly,

134 Mathematica is used to be spared from the tedious multiplication (see Mathematica Notebook [b]: [Link]).
135 Ifthe commutation relations hold in the near-horizon limit (i.e. for a specific value of (t, σ)) then they hold ∀ (t, σ). In [42] de
Boer et al. determine the normalization constant A by demanding the normalization of the modes through the Klein-Gordon
inner product. The near-horizon limit is taken during this calculation. In a later work [46], Appendix A, a similar collaboration
Atmaja et al. argue that the inner product contribution from the near-horizon region can be thought of as the overall inner
product. There is a contribution to the inner product from regions away from the horizon; however, the near-horizon region is
semi-infinite in the tortoise coordinate r∗ (r → rH ≡ rs∗ → −∞); and as such the normalization can be completely fixed by
the near-horizon regime. Therefore, the near-horizon limit can be taken here without concern.

83
 2 Z ∞
l πα0 h i(σ+σ0 )ω
 2 
h i 1 i rH dω i(σ−σ 0 )ω
i
X̂(t, σ), nt P̂ t (t, σ 0 ) =√ ω 2 e + e
Σ g̃|Σ πα0 l2 −∞ 2π rH ω
Z ∞
i dω h i(σ+σ0 )ω 0
i
=√ ω e + ei(σ−σ )ω
g̃|Σ −∞ 2π
(E.9)
i
=√ [δ(σ + σ 0 ) + δ(σ − σ 0 )]
g̃|Σ
i
=√ δ(σ − σ 0 ) ,
g̃|Σ

R∞
where the third line follows from the Dirac delta function property δ(y − x) = 2π 1
−∞
dk e−ik(y−x) ; and, in
0
the final line, the first Dirac delta function has disappeared since σ, σ ∈ [0, σf ]. Hence, the scalar field and
conjugate momentum commutation relations (Eq.(E.1)) and the Fourier coefficient creation and annihilation
commutation relations (Eq.(E.7)) are consistent; proving that the normalization constant Aω is correctly
given by Eq.(E.8).

Appendix F. Leading Order Contributions of the Near-Horizon Tortoise Coordinate

The tortoise coordinate r∗ is defined in Eq.(4.39). In the near-horizon limit (r = (1+˜

)rH ) the tortoise coor-
dinate can be expanded and truncated around ˜ = 0, in order to yield the near-horizon tortoise coordinate,
˜∗ . This coordinate is given – for general d dimensions – by

l2 (d − 4)l2 d2 − 14d + 36 l2 2 5d2 − 46d + 96 l2 3 4
˜∗ := 2 ln(˜
) + 2 ˜ + 2 (d − 1) ˜ − 2 (d − 1) ˜ + O (˜
)
rH (d − 1) 2rH (d − 1) 24rH 72rH (F.1)
2 3 4
) + a1 ˜ + a2 ˜ + a3 ˜ + O (˜
= a0 ln(˜ ) ,

where r∗ = rH ˜∗ , and a0 , a1 , a2 , a3 are constants. To leading order, the near-horizon AdSd -Schwarzschild
metric is given by
2 2
rH ~ I2 − (d − 1)rH ˜ dt2 + l2
ds2d = 2
dX 2
2 ,
d˜ (F.2)
l l (d − 1)˜


where all terms are O(˜ ). To be consistent, at leading order ˜∗ must be considered to O(˜
) – hence, it has
2
) + a1 ˜ + O (˜
the form ˜∗ = a0 ln(˜ ) .

The near-horizon tortoise coordinate is used to convert the near-horizon metric into (t, ˜∗ ) coordinates,
thereby arriving at a conformally flat description of near-horizon AdSd -Schwarzschild, from which the leading
order string solution and the transverse equations of motion can be found. Considering Eq.(5.109) it is clear
that the metric conversion factor between (t, ˜) and (t, ˜∗ ) coordinates is

2 (d−1)˜
rH ∗ 2 (d−1)
rH
e l2 =e l2
(a0 ln(˜)+a1 ˜+O(˜)2 )

2
= ec a0 ln(˜) ec a1 ˜ eO(˜)
  
2 2
= ˜c a0 1 + c a1 ˜ + O (˜ ) 1 + O (˜
) (F.3)
 
(d − 4) 2 2

= ˜ 1 + ˜ + O (˜
) 1 + O (˜
)
2
2
= ˜ + O (˜
) ,

84
 2
where the relevant leading order form of ˜∗ is used in the first equality, the constant c = rH (d − 1)/l2 is
2
defined in the second line, and in the third line the terms ec a1 ˜ and eO(˜) are series expanded around ˜ = 0.
While inputting the values for c, a0 , and a1 (from Eq.(F.1)) into the fourth line, notice that c a0 = 1 by
2
construction. All terms of O (˜ ) are disregarded, in order to be consistent with the leading order near-
horizon AdSd -Schwarzschild metric. Hence, from Eq.(F.3), the inverse near-horizon tortoise coordinate is
given by
2 (d−1)˜
rH ∗
˜ := e l2 , (F.4)

at leading order; while the near-horizon tortoise coordinate is given by

l2
˜∗ := 2 (d − 1) ln(˜
) , (F.5)
rH

at leading order. Claims that Eq.(F.5) is the leading order contribution to the near-horizon tortoise coor-
dinate have been made in [52]. This is, however, the first time to the authors’ knowledge a proof to this
effect (even a heuristic one) has been provided.

85
Appendix G. Accessing the Mathematica Code

The Mathematica code used to generate all the analytic results and plots in this article is provided as
supplementary material.

The code is organised into four comprehensively annotated Mathematica notebooks containing: (a) the
mapping of the heavy and light quark test string solutions between the worldsheet parameter space and
target spacetime ([Link]); (b) calculations pertaining to the analysis of heavy
and light quark test strings undergoing Brownian motion ([Link]); (c) the expansion of
the AdSd -Schwarzschild metric in the near-horizon region for the light quark’s test string configuration
([Link]); and (d) drag force calculations in AdS/CFT ([Link]).

For readers please always check for the latest versions of the notebooks in the GitHub repository:
[Link]

86
Bibliography

[1] A. Aronson and T. Ludlam. Hunting the Quark Gluon Plasma: Results from the First 3 Years at the
Relativistic Heavy Ion Collider (RHIC). Upton, New York: Brookhaven National Laboratory. Formal
Report: BNL-73847, (2005). Available at [Link] the [Link].

[2] U. Heinz and P. Kolb. Early Thermalization at RHIC. Nuclear Physics A, 702(1-4), (2002).
doi:10.1016/s0375-9474(02)00714-5. arXiv:hep-ph/0111075.
[3] M. Gyulassy. The QGP discovered at RHIC. In Structure and Dynamics of Elementary Matter (NATO
Advanced Study Institute), pages 159–182. Springer, (2004). arXiv:nucl-th/0403032.

[4] U. Heinz and R. Snellings. Collective Flow and Viscosity in Relativistic Heavy-Ion Collisions. An-
nual Review of Nuclear and Particle Science, 63(1), (2013). doi:10.1146/annurev-nucl-102212-170540.
arXiv:1301.2826 [nucl-th].
[5] C. Gale, S. Jeon, and B. Schenke. Hydrodynamic Modeling of Heavy-Ion Collisions. International
Journal of Modern Physics A, 28(11), (2013). doi:10.1142/s0217751x13400113. arXiv:1301.5893
[nucl-th].
[6] A. Majumder and M. van Leeuwen. The Theory and Phenomenology of Perturbative QCD based Jet
Quenching. Progress in Particle and Nuclear Physics, 66(1), (2011). doi:10.1016/[Link].2010.09.001.
arXiv:1002.2206 [hep-ph].
[7] W. Horowitz. Heavy Quark Production and Energy Loss. Nuclear Physics A, 904-905, (2013).
doi:10.1016/[Link].2013.01.061. arXiv:1210.8330 [nucl-th].
[8] M. Djordjevic, M. Djordjevic, and B. Blagojevic. RHIC and LHC Jet Suppression in Non-central
Collisions. Physics Letters B, 737, (2014). arXiv:1405.4250 [nucl-th].
√
[9] T. Hirano, P. Huovinen, and Y. Nara. Elliptic Flow in Pb+Pb Collisions at sN N =
2.76 TeV: Hybrid Model Assessment of the First Data. Physical Review C, 84(1), (2011).
doi:10.1103/physrevc.84.011901. arXiv:1012.3955 [nucl-th].
[10] C. Shen, U. Heinz, P. Huovinen, and H. Song. Radial and Elliptic Flow in Pb+Pb Collisions at Energies
Available at the CERN Large Hadron Collider from Viscous Hydrodynamics. Physical Review C, 84
(4), (2011). doi:10.1103/physrevc.84.044903. arXiv:1105.3226 [nucl-th].
√
[11] Z. Qiu, C. Shen, and U. Heinz. Hydrodynamic Elliptic and Triangular flow in Pb−Pb Collisions at s
= 2.76ATeV. Physics Letters B, 707(1), (2012). doi:10.1016/[Link].2011.12.041. arXiv:1110.3033
[nucl-th].
[12] C. Gale, S. Jeon, B. Schenke, P. Tribedy, and R. Venugopalan. Event-by-event Anisotropic Flow
in Heavy-Ion Collisions from Combined Yang-Mills and Viscous Fluid Dynamics. Physical Review
Letters, 110(1), (2013). doi:10.1103/physrevlett.110.012302. arXiv:1209.6330 [nucl-th].
[13] S. Gubser, I. Klebanov, and A. Peet. Entropy and Temperature of Black 3-Branes. Physical Review
D, 54(6), (1996). doi:10.1103/physrevd.54.3915. arXiv:hep-th/9602135.
[14] J. Casalderrey-Solana, H. Liu, D. Mateos, K. Rajagopal, and U. Wiedemann. Gauge/String Dual-
ity, Hot QCD and Heavy Ion Collisions. Cambridge University Press, (2014). arXiv:1101.0618
[hep-th].
[15] R. Morad and W. Horowitz. Strong-coupling Jet Energy Loss from AdS/CFT. Journal of High Energy
Physics, 2014(11), (2014). doi:10.1007/jhep11(2014)017. arXiv:1409.75455 [hep-th].
[16] J. Casalderrey-Solana, D. Gulhan, J. Milhano, D. Pablos, and K. Rajagopal. A Hybrid Strong/Weak
Coupling Approach to Jet Quenching. Journal of High Energy Physics, 2014(10), (2014).
doi:10.1007/jhep10(2014)019. arXiv:1405.3864 [hep-ph].
[17] W. Horowitz. Fluctuating Heavy Quark Energy Loss in a Strongly Coupled Quark-Gluon Plasma.
Physical Review D, 91(8), (2015). doi:10.1103/physrevd.91.085019. arXiv:1501.04693 [hep-ph].
[18] G. ’t Hooft. A Planar Diagram Theory for Strong Interactions. Nuclear Physics. B, 72(3):461–473,
(1974).

87
[19] I. Klebanov. TASI Lectures: Introduction to the AdS/CFT Correspondence. Strings, Branes and
Gravity, (2001). doi:10.1142/9789812799630 0007. arXiv:hep-th/0009139.
[20] M. Natsuume. AdS/CFT Duality User Guide, volume 903. Springer, (2015). arXiv:1409.3575
[hep-th].
[21] J. Maldacena. The large-N Limit of Superconformal Field Theories and Supergravity. Inter-
national Journal of Theoretical Physics, 38(4):1113–1133, (1999). doi:10.1023/a:1026654312961.
arXiv:hep-th/9711200.
[22] S. Gubser. Drag force in AdS/CFT. Physical Review D, 74(12), (2006).
doi:10.1103/physrevd.74.126005. arXiv:hep-th/0605182.
[23] D. Bak, A. Karch, and L. Yaffe. Debye Screening in Strongly Coupled N = 4 Supersymmetric Yang-
Mills Plasma. Journal of High Energy Physics, 2007(08), (2007). doi:10.1088/1126-6708/2007/08/049.
arXiv:0705.0994 [hep-th].
[24] E. Witten. Anti-de Sitter Space and Holography. Advances in Theoretical and Mathematical Physics,
2:253–291, (1998). arXiv:hep-th/9802150.
[25] S. Gubser, I. Klebanov, and A. Polyakov. Gauge Theory Correlators from Non-critical
String Theory. Physics Letters B, 428(1-2):105–114, (1998). doi:10.1016/s0370-2693(98)00377-3.
arXiv:hep-th/9802109.
[26] O. Aharony and E. Witten. Anti-de Sitter Space and the Center of the Gauge Group.
Journal of High Energy Physics, 1998(11):018–018, (1998). doi:10.1088/1126-6708/1998/11/018.
arXiv:hep-th/9807205.
[27] O. Aharony, S. Gubser, J. Maldacena, H. Ooguri, and Y. Oz. Large N Field Theories, String The-
ory and Gravity. Physics Reports, 323(3-4):183–386, (2000). doi:10.1016/s0370-1573(99)00083-6.
arXiv:hep-th/9905111.
[28] G. Policastro, D. Son, and A. Starinets. Shear Viscosity of Strongly Coupled N = 4 Supersymmet-
ric Yang-Mills Plasma. Physical Review Letters, 87(8), (2001). doi:10.1103/physrevlett.87.081601.
arXiv:hep-th/0104066.
[29] PHENIX Collaboration, S. Adler, and et al. Elliptic Flow of Identified Hadrons in Au+Au Collisions
√
at sN N = 200 GeV. Physical Review Letters, 91(18), (2003). doi:10.1103/physrevlett.91.182301.
arXiv:nucl-ex/0305013.
√
[30] STAR Collaboration, J. Adams, and et al. Azimuthal Anisotropy in Au+Au collisions at sN N = 200
GeV. Physical Review C, 72(1), (2005). doi:10.1103/physrevc.72.014904. arXiv:nucl-ex/0409033.
[31] P. Kovtun, D. Son, and A. Starinets. Holography and Hydrodynamics: Diffusion on Stretched Hori-
zons. Journal of High Energy Physics, 2003(10):064–064, (2003). doi:10.1088/1126-6708/2003/10/064.
arXiv:hep-th/0309213.
[32] P. Kovtun, D. Son, and A. Starinets. Viscosity in Strongly Interacting Quantum Field Theories
from Black Hole Physics. Physical Review Letters, 94(11), (2005). doi:10.1103/physrevlett.94.111601.
arXiv:hep-th/0405231.
[33] A. Buchel and J. Liu. Universality of the Shear Viscosity from Supergravity Duals. Physical Review
Letters, 93(9), (2004). doi:10.1103/physrevlett.93.090602. arXiv:hep-th/0311175.
[34] A. Buchel. On Universality of Stress-Energy Tensor Correlation Functions in Supergravity. Physics
Letters B, 609(3-4):392–401, (2005). doi:10.1016/[Link].2005.01.052. arXiv:hep-th/0408095.
[35] C. Herzog, A. Karch, P. Kovtun, C. Kozcaz, and L. Yaffe. Energy Loss of a Heavy Quark Moving
through N = 4 Supersymmetric Yang-Mills Plasma. Journal of High Energy Physics, 2006(07):013,
(2006). doi:10.1088/1126-6708/2006/07/013. arXiv:hep-th/0605158.
[36] H. Liu, K. Rajagopal, and U. Wiedemann. Calculating the Jet Quenching Parameter. Physical Review
Letters, 97(18), (2006). doi:10.1103/physrevlett.97.182301. arXiv:hep-ph/0605178.
[37] C. Herzog. Energy Loss of a Heavy Quark from Asymptotically AdS Geometries. Jour-
nal of High Energy Physics, 2006(09):032–032, (2006). doi:10.1088/1126-6708/2006/09/032.
arXiv:hep-th/0605191.

88
[38] J. Casalderrey-Solana and D. Teaney. Heavy Quark Diffusion in Strongly Coupled N = 4 Yang Mills.
Physical Review D, 74(8), (2006). doi:10.1103/physrevd.74.085012. arXiv:hep-ph/0605199.
[39] H. Liu, K. Rajagopal, and U. Wiedemann. Wilson Loops in Heavy Ion Collisions and their Calcu-
lation in AdS/CFT. Journal of High Energy Physics, 2007(03):066–066, (2007). doi:10.1088/1126-
6708/2007/03/066. arXiv:hep-ph/0612168.
[40] J. Casalderrey-Solana and D. Teaney. Transverse Momentum Broadening of a Fast Quark in a N = 4
Yang-Mills Plasma. Journal of High Energy Physics, 2007(04):039–039, (2007). doi:10.1088/1126-
6708/2007/04/039. arXiv:hep-th/0701123.
[41] S. Gubser. Momentum Fluctuations of Heavy Quarks in the Gauge-String Duality. Nuclear Physics
B, 790(1-2):175–199, (2008). doi:10.1016/[Link].2007.09.017. arXiv:hep-th/061214.
[42] J. de Boer, V. Hubeny, M. Rangamani, and M. Shigemori. Brownian Motion in AdS/CFT.
Journal of High Energy Physics, 2009(07):094–094, (2009). doi:10.1088/1126-6708/2009/07/094.
arXiv:0812.5112 [hep-th].
[43] D. Son and D. Teaney. Thermal Noise and Stochastic Strings in AdS/CFT. Journal of High Energy
Physics, 2009(07), (2009). doi:10.1088/1126-6708/2009/07/021. arXiv:0901.2338 [hep-th].
[44] W. Fischler, J. Pedraza, and W. Tangarife Garcia. Holographic Brownian Motion in Magnetic
Environments. Journal of High Energy Physics, 2012(12), (2012). doi:10.1007/jhep12(2012)002.
arXiv:1209.1044 [hep-th].
[45] A. Atmaja. Holographic Brownian Motion in Two Dimensional Rotating Fluid. Journal of High
Energy Physics, 2013(4), (2013). doi:10.1007/jhep04(2013)021. arXiv:1212.5319 [hep-th].
[46] A. Atmaja, J. de Boer, and M. Shigemori. Holographic Brownian Motion and Time Scales in Strongly
Coupled Plasmas. Nuclear Physics B, 880:23–75, (2014). doi:10.1016/[Link].2013.12.018.
arXiv:1002.2429 [hep-th].
[47] P. Banerjee and B. Sathiapalan. Holographic Brownian Motion in 1 + 1 Dimensions. Nuclear Physics
B, 884, (2014). doi:10.1016/[Link].2014.04.016. arXiv:1308.3352 [hep-th].
[48] S. Chakrabortty, S. Chakraborty, and N. Haque. Brownian Motion in Strongly Coupled,
Anisotropic Yang-Mills Plasma: A Holographic Approach. Physical Review D, 89(6), (2014).
doi:10.1103/physrevd.89.066013. arXiv:1311.5023 [hep-th].
[49] J. Sadeghi, F. Pourasadollah, and H. Vaez. Holographic Brownian Motion in Three-Dimensional
Gödel Black Hole. Advances in High Energy Physics, 2014, (2014). doi:10.1155/2014/762151.
arXiv:1308.2483 [hep-th].
[50] J. Sadeghi, B. Pourhassan, and F. Pourasadollah. Holographic Brownian motion in 2 + 1 Dimensional
Hairy Black Holes. The European Physical Journal C, 74(3), (2014). doi:10.1140/epjc/s10052-014-
2793-7. arXiv:1312.4906 [hep-th].
[51] W. Fischler, P. Nguyen, J. Pedraza, and W. Tangarife. Fluctuation and Dissipation in de Sitter Space.
Journal of High Energy Physics, 2014(8), (2014). doi:10.1007/jhep08(2014)028. arXiv:1404.0347
[hep-th].
[52] R. Moerman and W. Horowitz. A Semi-classical Recipe for Wobbly Limp Noodles in Partonic Soup.
arXiv preprint, (2016). arXiv:1605.09285 [hep-th].
[53] R. Brown. A Brief Account of Microscopical Observations made in the Months of June, July and
August 1827, on the Particles contained in the Pollen of Plants; and on the General Existence of Active
Molecules in Organic and Inorganic Bodies. The Philosophical Magazine, 4(21):161–173, (1828).
[54] A. Karch and E. Katz. Adding Flavor to AdS/CFT. Journal of High Energy Physics, 2002(06),
(2002). doi:10.1088/1126-6708/2002/06/043. arXiv:hep-th/0205236.
[55] A. Polyakov. String Theory and Quark Confinement. Nuclear Physics B, 68(1-3):1–8, (1998).
doi:10.1016/s0920-5632(98)00135-2. arXiv:hep-th/9711002.
[56] G. ’t Hooft. Dimensional Reduction in Quantum Gravity. Salamfestschrift: a collection of talks, 4
(A):1, (1993). arXiv:gr-qc/9310026.

89
[57] L. Susskind. The World as a Hologram. Journal of Mathematical Physics, 36(11):6377–6396, (1995).
doi:10.1063/1.531249. arXiv:hep-th/9409089.
[58] C. Thorn. Reformulating String Theory with the 1/N Expansion. In Proceedings of the first interna-
tional Sakharov conference on physics, (1992). arXiv:hep-th/9405069.
[59] G. Gibbons and K. Maeda. Black Holes and Membranes in Higher-Dimensional Theories with Dilaton
Fields. Nuclear Physics B, 298(4):741–775, (1988).
[60] G. Horowitz and A. Strominger. Black Strings and P-branes. Nuclear Physics B, 360(1):197–209,
(1991).
[61] D. Garfinkle, G. Horowitz, and A. Strominger. Charged Black Holes in String Theory. Physical Review
D, 43(10):3140, (1991).

[62] C. Ballón Bayona and N. Braga. Anti-de Sitter Boundary in Poincaré Coordinates. General Relativity
and Gravitation, 39(9), (2007). doi:10.1007/s10714-007-0446-y. arXiv:hep-th/0512182.
[63] D. Mateos. String Theory and Quantum Chromodynamics. Classical and Quantum Gravity, 24(21),
(2007). doi:10.1088/0264-9381/24/21/s01. arXiv:0709.1523 [hep-th].

[64] E. Witten. Anti-de Sitter Space, Thermal Phase Transition, and Confinement in Gauge Theories.
Adv. Theor. Math. Phys., 2(IASSNS-HEP-98-21):89–117, (1998). arXiv:hep-th/9803131.
[65] M. Douglas and D. Taylor. Branes in the Bulk of Anti-de Sitter Space, (1998). arXiv:hep-th/9807225.
[66] S. Das. Holograms of Branes in the Bulk and Acceleration terms in SYM Effective Action. Journal of
High Energy Physics, 1999(06), (1999). doi:10.1088/1126-6708/1999/06/029. arXiv:hep-th/9905037.
[67] F. Gonzalez-Rey, B. Kulik, I. Park, and M. Roček. Self-dual Effective Action of N =
4 super-Yang-Mills. Nuclear Physics B, 544(1-2), (1999). doi:10.1016/s0550-3213(99)00046-2.
arXiv:hep-th/9810152.
[68] M. Tanabashi et al. Particle Data Group Review. Physical Review D, 98:030001, (2018).

[69] L. Alday and J. Maldacena. Gluon Scattering Amplitudes at Strong Coupling. Journal of High Energy
Physics, 2007(06):064, (2007). arXiv:0705.0303 [hep-th].
[70] S. Hartnoll, C. Herzog, and G. Horowitz. Building a Holographic Superconductor. Physical Review
Letters, 101(3), (2008). doi:10.1103/physrevlett.101.031601. arXiv:0803.3295 [hep-th].

[71] S. Gubser. Breaking an Abelian Gauge Symmetry Near a Black Hole Horizon. Physical Review D, 78
(6), (2008). doi:10.1103/physrevd.78.065034. arXiv:0801.2977 [hep-th].
[72] K. Maeda, M. Natsuume, and T. Okamura. Universality Class of Holographic Superconductors.
Physical Review D, 79(12), (2009). doi:10.1103/physrevd.79.126004. arXiv:0904.1914 [hep-th].

[73] S. Hartnoll. Lectures on Holographic Methods for Condensed Matter Physics. Classical and Quantum
Gravity, 26(22):224002, (2009). arXiv:0903.3246 [hep-th].
[74] A. Einstein. On the Motion of Small Particles Suspended in Liquids at Rest Required by the Molecular-
Kinetic Theory of Heat. Annalen der physik, 17(549-560):208, (1905).
[75] M. Smoluchowski. On the Kinetic Theory of the Brownian Molecular Motion and of Suspensions.
Annalen der Physik, 21:756–780, (1906).
[76] S. Brush. A History of Random Processes: I. Brownian Movement from Brown to Perrin. Archive for
History of Exact Sciences, 5(1):1–36, (1968).
[77] L. Bachelier. Theory of Speculation. Dimson, E. and M. Mussavian (1998), A brief history of market
efficiency, European Financial Management, 4(1):91–193, (1900).
[78] N. Wiener. Differential-Space. Journal of Mathematics and Physics, 2(1-4):131–174, (1923).
[79] P. Langevin. “On the theory of Brownian Motion” (“Sur la Théorie du Mouvement Brownien,” cr
acad. sci.(paris) 146, 530–533 (1908)). American Journal of Physics, 65(11):1079–1081, (1997).

90
 [80] G. Uhlenbeck and L. Ornstein. On the Theory of the Brownian Motion. Physical Review, 36(5):823,
(1930).
[81] N. Pottier. Nonequilibrium Statistical Physics: Linear Irreversible Processes. Oxford University Press,
(2010).
[82] R. Kubo. The Fluctuation-Dissipation Theorem. Reports on progress in physics, 29(1):255, (1966).
[83] L. Ornstein. On the Brownian Motion. Acad. Amst, 26:1005, (1917).
[84] H. Mori. Transport, Collective Motion, and Brownian motion. Progress of theoretical physics, 33(3):
423–455, (1965).
[85] N. Wiener. Generalised Harmonic Analysis. Acta mathematica, 55:117–258, (1930).
[86] M. Wang and G. Uhlenbeck. On the Theory of the Brownian Motion II. Reviews of modern physics,
17(2-3):323, (1945).
[87] J. Dunkel and P. Hänggi. Relativistic Brownian Motion. Physics Reports, 471(1):1–73, (2009).
doi:10.1016/[Link].2008.12.001. arXiv:0812.1996 [hep-th].
[88] R. Kubo, M. Toda, and N. Hashitsume. Statistical Physics II: Nonequilibrium Statistical Mechanics,
volume 31. Springer Science & Business Media, (2012).
[89] L. Brink, P. Di Vecchia, and P. Howe. A Locally Supersymmetric and Reparametrization Invariant
Action for the Spinning String. Physics Letters B, 65(5):471–474, (1976).
[90] S. Deser and B. Zumino. A Complete Action for the Spinning String. Physics Letters B, 65(4):
369–373, (1976).
[91] A. Polyakov. Quantum Geometry of Bosonic Strings. In Supergravities in Diverse Dimensions:
Commentary and Reprints (In 2 Volumes), pages 1197–1200. World Scientific, (1989).
[92] B. Zwiebach. A First Course in String Theory. Cambridge University Press, (2004).
[93] J. Polchinski. String Theory: Volume 1, An Introduction To The Bosonic String. Cambridge Univer-
sity Press, (1998).
[94] Y. Nambu. Duality and Hadrodynamics, Lectures at the Copenhagen High Energy Symposium, 1970.
In Broken Symmetry: Selected Papers of Y Nambu, pages 280–301. World Scientific, (1995).
[95] T. Gotō. Relativistic Quantum Mechanics of One-dimensional Mechanical Continuum and Subsidiary
Condition of Dual Resonance Model. Progress of Theoretical Physics, 46(5):1560–1569, (1971).
[96] S. Carroll. Spacetime and Geometry: An Introduction to General Relativity. Addison Wesley, (2004).
[97] Wolfram Research, Inc. Mathematica, Version 11.3, (2018). Champaign, IL.
[98] A. Rich, P. Scheibe, and N. Abbasi. Rule-based Integration: An Extensive System of Symbolic
Integration Rules. Journal of Open Source Software, 3(32):1073, (2018). doi:10.21105/joss.01073.
Available at [Link]
[99] M. Bañados, C. Teitelboim, and J. Zanelli. Black Hole in Three-Dimensional Spacetime. Physical
Review Letters, 69(13):1849–1851, (1992). doi:10.1103/physrevlett.69.1849. arXiv:hep-th/9204099.
[100] W. Bardeen, I. Bars, A. Hanson, and R. Peccei. Study of the Longitudinal Kink Modes of the String.
Physical Review D, 13(8):2364, (1976).
[101] S. Gubser, D. Gulotta, S. Pufu, and F. Rocha. Gluon Energy Loss in the Gauge-String Dual-
ity. Journal of High Energy Physics, 2008(10):052–052, (2008). doi:10.1088/1126-6708/2008/10/052.
arXiv:0803.1470 [hep-th].
[102] P. Chesler, K. Jensen, and A. Karch. Jets in Strongly Coupled N = 4 Yang-Mills Theory. Physical
Review D, 79(2), (2009). doi:10.1103/physrevd.79.025021. arXiv:0804.3110 [hep-th].
[103] P. Chesler, K. Jensen, A. Karch, and L. Yaffe. Light Quark Energy Loss in Strongly Coupled N = 4 Su-
persymmetric Yang-Mills Plasma. Physical Review D, 79(12), (2009). doi:10.1103/physrevd.79.125015.
arXiv:0810.1985 [hep-th].

91
[104] A. Ficnar. AdS/CFT Energy Loss in Time-Dependent String Configurations. Physical Review D, 86
(4), (2012). doi:10.1103/physrevd.86.046010. arXiv:1201.1780 [hep-th].
[105] A. Ficnar, J. Noronha, and M. Gyulassy. Falling Strings and Light Quark Jet Quenching at LHC.
Nuclear Physics A, 910-911:252–255, (2013). doi:10.1016/[Link].2012.12.030. arXiv:1208.0305
[hep-th].

[106] A. Ficnar and S. Gubser. Finite Momentum at String Endpoints. Physical Review D, 89(2), (2014).
doi:10.1103/physrevd.89.026002. arXiv:1306.6648 [hep-th].
[107] I. Bars. Classical Solutions of 2D String Theory in any Curved Spacetime. In Second Paris Cosmology
Colloquium, page 399. World Scientific, (1994). arXiv:hep-th/9411217.
[108] I. Bars and J. Schulze. Folded Strings Falling into a Black Hole. Physical Review D, 51(4):1854–1868,
(1995). doi:10.1103/physrevd.51.1854. arXiv:hep-th/9405156.
[109] E. Álvarez and J. Conde. Are the Sring and Einstein Frames Equivalent? Modern Physics Letters A,
17(07):413–420, (2002). doi:10.1142/s0217732302006606. arXiv:gr-qc/0111031.
[110] T. Matsuo, D. Tomino, and W. Wen. Drag Force in SYM Plasma with B Field from
AdS/CFT. Journal of High Energy Physics, 2006(10), (2006). doi:10.1088/1126-6708/2006/10/055.
arXiv:hep-th/0607178.
[111] K. Kim, J. Shock, and J. Tarrı́o. The Open String Membrane Paradigm with External Electro-
magnetic Fields. Journal of High Energy Physics, 2011(6), (2011). doi:10.1007/jhep06(2011)017.
arXiv:1103.4581 [hep-th].

[112] M. Cederwall, A. von Gussich, A. Mikovic, B. Nilsson, and A. Westerberg. On the Dirac-Born-
Infeld Action for D-branes. Physics Letters B, 390(1-4), (1997). doi:10.1016/s0370-2693(96)01367-6.
arXiv:hep-th/9606173.

92