DESY 09-058

Resonant laser power build-up in ALPS  a light-shining-through-walls experiment 

a a c a a a a Axel Lindner , Jenny List , Tobias Meier , Niels Meyer , Dieter Notz , Javier Redondo , Andreas Ringwald , d , Benno Willkec Gnter Wiedemann

c Max-Planck-Institute

a Deutsches Elektronen-Synchrotron DESY, Notkestrae 85, D-22607 Hamburg, Germany b Laser Zentrum Hannover e.V., Hollerithallee 8, D-30419 Hannover, Germany

arXiv:0905.4159v1 [[Link]-det] 26 May 2009

for Gravitational Physics, Albert-Einstein-Institute, and Institut fr Gravitationsphysik, Leibniz Universitt, Hannover, Callinstrae 38, D-30167 Hannover, Germany d Hamburger Sternwarte, Gojenbergsweg 112, D-21029 Hamburg, Germany

Abstract

The ALPS collaboration runs a light-shining-through-walls (LSW) experiment to search for photon oscillations into weakly interacting sub-eV particles (WISPs) inside of a superconducting HERA dipole magnet at the site of DESY. In this paper we report on the rst successful integration of a large-scale optical cavity to boost the available power for WISP production in this type of experiments. The key elements are a frequency tunable narrow line-width continuous wave laser acting as the primary light source and an electronic feed-back control loop to stabilize the power build-up. We describe and characterize our apparatus and demonstrate the data analysis procedures on the basis of a brief exemplary run.
Key words:

Experimental tests, Photon regeneration, Resonators and cavities, Axions, Other gauge bosons

PACS: [Link], 14.80.-j, [Link], [Link], [Link]

1. Introduction

supersymmetry which would solve the hierarchy problem and might provide good dark matter candidates, the socalled WIMPs (for weakly interacting massive particles). Low mass particles living in hidden sectors might still have very weak interactions with the SM elds through radiative corrections involving massive mediator particles or through gravity. We name these particles WISPs (weakly

The standard model of elementary particles (SM) offers an accurate description of almost every phenomenon observed so far in particle physics. Despite this success, there are both theoretical and observational motivations to believe what the SM describes is just a small component of nature's complexity. From the theoretical side, the SM suers from naturalness, hierarchy and arbitrariness problems and, moreover, it does not describe gravity. From the observational viewpoint, cosmology and astrophysics claim that around 22% of the universe's energy density is made of a yet unidentied type of non-baryonic dark matter and

1 interacting sub-eV particles). The feebleness of

their interactions implies extremely few events in a typical experiment, making the luminosity and detector eciency two crucial requirements for WISP searches. On the other hand, having low masses, WISPs can potentially exhibit coherent interactions along macroscopic distances boosting their production probabilities. Currently, the beams of SM particles with the highest luminosities and best coherence properties are laser beams. Commercial lasers in the visible spectrum can easily reach output powers of several tens of Watts ( several kilometers. ciencies close

74%

of a yet more mysterious dark energy.

Extensions of the SM usually introduce many new particles and symmetries that provide elegant solutions to our theoretical concerns and candidates for the dark sector. Interestingly enough, these extensions often include much more particles than known today. On general grounds these particles could live in principle at any energy scale since new symmetries could protect their masses. In particular nothing prohibits that light particles beyond the SM exist as long as they have no SM charges, i.e. as long as they populate a hidden sector. Actually these hidden sectors arise quite naturally in string theory, our current best candidate for a theory of quantum gravity, and they are required to break the hypothetical

1019

photons per second) and can have coherence lengths of Sensitive detectors with quantum efare also available. Therefore, on

100%

general grounds, experimental searches for WISPs coupled to photons are the most favored.

1 Here

weak has to be understood in a broad sense, while in

WIMP, it stands for electro-weak.

Preprint submitted to Nuc. Instrum. Meth.

May 27, 2009

Let us notice that these WISPs can have a sizable impact on cosmology and astrophysics. Often not realized, they can be as good cold dark matter candidates as WIMPs despite having sub-eV masses, as it is the case for the axion [1]. Stellar evolution is often accelerated [2, 3, 4, 5], as it is by neutriby the emission of WISPs

WISP

nos, and photon-WISP interactions can modify the spectra of cosmic radiation [6, 7, 8, 9, 10, 11]. Indeed, WISPs could help in understanding a number of recent observations [11, 12, 13, 14, 15, 16, 17, 18]. Strong constraints

Figure 1: Schematic overview of a light-shining-through-walls experiment. The gray blob indicates the mixing term between photons and the weakly interacting sub-eV particle (WISP).

for

200

times through the oscillation region, thus aug-

on WISPs often arise from these eects [19], but they are certainly model dependent [20, 21, 22, 23, 24] and they rely on our very imperfect knowledge of cosmology, stellar interiors or astrophysical environments. In turn, these arguments suggest that the detection of a WISP in a laboratory experiment will deeply change our view of cosmology and astrophysics. In fact, neutrinos (the only SM WISP) were postulated and conrmed as subtle eects in laboratory experiments and nowadays they are essential ingredients in our understanding of cosmology (like in big bang nucleosynthesis or structure formation) and astrophysics (like in white dwarf cooling or supernova type-II explosions). Furthermore, WISPs have been recently shown to have interesting technological applications [25, 26]. We nd that laboratory experiments at the low energy
frontier

menting the eective laser power by a factor

100.

This

scheme suers however from two drawbacks. The rst is the necessity of a detector with a large sensitive area because the dierent beam paths must not overlap in order to avoid destructive interference. However, in general the dark noise of a detector for single photons grows with the size of its sensitive area. Thus it will increase with an increasing number of light passes in the delay line and by this deteriorating the signal to noise ratio of the detector. The second drawback is that delay lines with a much higher number of light passes than 200 are dicult to construct due to the growing complexity of the underlying geometrics. These problems can be solved by the use of optical resonators. Since they are based on the superimposition of all the dierent equivalent roundtrips within the area of only one light path, the sensitive area of the detector can be kept very small resulting in low dark noise. On the other hand, long optical cavities with equivalent round trip numbers of several thousand are used by some large-scale experiments such that they boost up the power inside to levels of several tens of kW. Good examples of these are the laser interferometric gravitational wave detectors like GEO600 [48] or LIGO [49]. Due to this power enhancement eect optical cavities could be extremely useful in a LSW experiment where they would act as ampliers for the luminosity that is available from the laser source [50, 51]. However, such a set-up was never realized so far .

[27] are therefore complementary to collider ef-

forts in the search of an accurate description of nature. A very spectacular eect of such WISPs takes place when we have a photon-WISP interaction vertex in our theory, a so-called mixing term. In this context, photons can convert into WISPs during their propagation, and the quantum amplitudes of these transitions at dierent points along the trajectory can add up coherently, enhancing the signal, or decreasing it if they add up out of phase. This phenomenon is known as quantum avor oscillations and has been observed in the context of the neutral kaon, B meson and neutrino systems. There are several direct and indirect experimental approaches for WISP searches [5, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42]. Possibly, the cleanest way to search for photon-WISP oscillations is through so-called light-shining-through-walls (LSW) experiments [43, 44, 45]. In such experiments a laser beam is shone onto a thick wall where photons are stopped but WISPs produced in oscillations can traverse. By the inverse process, WISPs can re-oscillate into photons behind the wall and may be detected in a low background environment, see Fig. 1 Since the number of photons regenerated after the wall is proportional to the ux in the oscillation region, it is crucial to get the highest possible laser powers. In this respect, it is extremely disappointing that photons in the oscillation region are lost when hitting the wall. To partially overcome this situation, in a pioneer LSW experiment [46, 47] a so-called optical delay line was set up which forced the light to travel on almost collinear paths 2

The main reason for that is the

by far increased complexity of a cavity-enhanced LSWscheme. This complexity arises on the one hand from strict constraints on the experimental ambient conditions such as high cleanliness or large vacuum chambers [48, 49, 55]. On the other hand, in some important cases, a strong magnetic eld in the background of the oscillation region is required for oscillations to happen. In this paper we report on the rst successful realization of such a cavity-enhanced LSW experiment, the Any-Light-Particle-Search (ALPS) experiment [56, 57] at DESY Hamburg. It utilizes a long optical resonator on the experiment's production side inside a superconducting HERA dipole. The resonator increases the eective

2 Cavities

with excellent characteristics have been combined with

strong magnetic elds in other types of WISP searches, like in the PVLAS [52, 53] or Q&A [54] searches of WISP-induced dichroism and birefringence.

laser power for WISP searches by more than a factor of 40. Its suitability as part of a large-scale LSW experiment is demonstrated by a rst search for WISPs at ALPS. The current limitations for its sensitivity enhancement together with other possible improvements of the experiment are identied. This should enable us to increase the sensitivity of the whole experiment signicantly in the future. This paper is organized as follows: In Sec. 2 we briey describe the WISPs we can search for in the ALPS experiment and characterize the phenomenon of oscillations. In Sec. 3 we detail the components of the ALPS experiment. The design and characterization of the optical cavity are described in Sec. 4 while in Sec. 5 we explain the process of data taking, review the performance of the optical resonator and the detector during the measurement run and present the results.

that couple to two photons depending on their intrinsic parity via interaction terms in the Lagrangian such as

L = g E B int
where eld and

1 and/or L+ = g+ (E 2 B 2 ) int 2

(4)

are dimensionful coupling constants,

the ALP

E, B

the electric and magnetic elds. In models

where parity is non-conserved [58] both couplings are allowed. Such schizons have very similar phenomenology to the ALPs studied here [59, 60]. Parity odd ALPs (0

) arise as Nambu-Goldstone bosons

[61, 62, 63] (GB) of spontaneously broken global axial symmetries, being the QCD axion [64, 65, 66, 67, 68, 69, 70, 71, 72, 73] the most famous but not unique example [74, 75, 76, 77, 78, 79, 80, 81]. Parity even ALPs (0 ) can be GBs, but also quintessence elds [82, 83, 84, 85] or elds governing the sizes of extra dimensions (moduli) or gauge couplings (dilatons) in string theories. Such particles are in principle subject to strong constraints from deviations of Newton's law [86], although some models which evade astrophysics also overcome these problems [21, 24] Further WISP candidates are hidden photons (HPs) [43].

2. Photon conversion and the WISP Zoo

The equations of motion of the photon-WISP system as a function of length can be written as

d dL

| |

1 2

2 n

m2

| |

(1)

They are gauge bosons of U(1)h symmetries of a hidden sector, which are ubiquitous in extensions of the standard model, e.g. [87, 88, 89, 90]. Gauge invariance allows HPs to have a Stckelberg mass [91, 92] and kinetic mixing with the standard photon [93, 94]

where

is the photon energy,

photon refraction index in the mass and

n = n 1 with n the medium, m is the WISP

is the quantum mechanical amplitude of the

forward transition

= |Hint |

The transition probability shows the characteristic oscillatory behavior as a function of the distance

Lmix =
with

L
(2)

1 F B , 2

(5)

4 2 M 2L P ( ) = 4 sin2 M 4
with

F (B ) the photon (HP) eld strength tensors and a dimensionless coupling ranging typically from 1023 2 to 10 [95, 96, 97, 98]. Models with kinetic mixing have
a wide range of phenomenological applications, e.g. [99, 100, 101, 102, 103, 104, 105, 106, 107, 108]. As a nal example, hidden sector particles charged under U(1)h , get a non-zero electric charge through the kinetic mixing term [93]

M 2 = ((2 2 n + m2 )2 + 4 2 )1/2 . Unfortunately, for visible light one usually has n > 0 so one cannot 2 2 match 2 n + m = 0 by choosing a suitable medium
to maximize the amplitude of the oscillations. The maximum amplitude is therefore obtained in vacuum where

n = 0 and M 2 = (m4 + 4 2 )1/2 . In coherent conditions M 2 L/4 /2 the probability takes a simpler form P ( ) = 2 L2 /(4 2 ) L
(3) where

QMCP =

eh , e

(6)

eh is the gauge coupling of U(1)h and e the electron's

charge. Other MCP model beyond this paradigm can be found in [109]. Since mini-charged particles (MCPs).

where the coherent enhancement in the interaction length is evident. Lorentz invariance forbids

transitions when the

external magnetic elds will pair produce MCPs [40, 110] and MCP loops can produce photonHP oscillations even if the HP mass is zero [111]. The Feynman diagrams giving rise to photon-WISP mixing in these dierent models are shown in Fig. 2 and the parameters characterizing the oscillation amplitudes in Tab. 1.

WISP spin is dierent from 1, and has to be explicitly broken if we want to prove oscillations into spin-0 (or of a strong magnetic eld

> 1)

particles. In the ALPS setup this is done by the inclusion

ext

orthogonal to the propa-

gation direction. In this case, photons polarized along the magnetic eld or perpendicular to it can (and will) behave dierently. They will have dierent

and

so their

WISP oscillation probability will be generically dierent. The rst WISPs to consider are axions and axion-likeparticles (ALPs). They are well motivated spin-0 particles 3

= WISP ALP

; HP(m > 0)

MCP HP(m = 0) ...

Figure 2: Feynman diagrams responsible for the mixing term between photons and dierent hypothetical weakly interacting sub-eV particles (WISPs). Photon oscillations into Axion-like particles (ALPs) and massless hidden photons (HPs) via mini-charged particle (MCP) loops require the presence of a background magnetic eld, denoted by a crossed circle. Oscillations into massive Hidden Photons occur regardless of such a background. See Table 1 for details on the photon-WISP couplings and references.

WISP Parity odd ALP (0

Needs

B ext

yes yes no yes

Parity even ALP (0 ) HP (m MCP+HP

|| g B ext
0

> 0) (for m = 0)

2 N||

m2 2

g+ B ext m2 2 2 N
||

m2 m2 m2 + m2 2 2 N||,
.

Ref. [112] [113] [43] [111]

Table 1: Parameters characterizing the photon-WISP probability of oscillations eq. (3). orientation of the photon polarization and the external eld which can be parallel the indices of refraction due to MCP loops adequate expression.

The mixing parameter

depends on the relative

and perpendicular

Note that in the MCP+HP case,

N||, = N||, (QMCP , B ext )

are generally complex and eq. (3) does not hold. See [111] for the

3. The ALPS experiment

3.1. Laser

The ALPS experiment as sketched in Fig. 3 is built up along a superconducting HERA dipole magnet which provides a eld intensity of

The beam tube inside the HERA dipole magnet is bent horizontally and leaves an aperture of only 14 mm with the vacuum tubes installed. This strongly constrains the beam quality of a suitable laser [56]. Furthermore, to eciently increase the optical power with a cavity, one needs a continuous-wave laser that emits a single longitudinal and transversal mode with a much smaller linewidth than the one of the resonator. Thus the laser source used for the ALPS experiment is a narrow-linewidth masteroscillator power amplier system (MOPA) operating at 1064 nm (Fig. 3). VIRGO. Stable narrow-linewidth emission is provided by a nonplanar ring oscillator (NPRO) [115], emitting 2 W of output power with a spectral width of 1 kHz (measured over 100 ms) and a long term frequency stability of 1 MHz/min. In four diode pumped Nd:YVO4 amplier stages the output power of this laser is increased to 35 W preserving the spectral emission characteristics and the nearly diraction limited beam quality with a fundamental transverse mode content of 95%. The MOPA is equipped with several frequency control elements, which are applied for the cavity frequency locking scheme in the ALPS experiment. A piezo-electric transducer installed on the NPRO laser crystal allows for a frequency shift of It is based on the system [114] developed for gravitational wave detectors like LIGO, GEO and

B = 5.30

T in a length of 8.8 m.

Two vacuum tubes are inserted into the magnet at each of both ends, featuring 6.3 and 7.6 m length, respectively. WISP production takes place in the rst tube; regeneration of photons in the second one. Around the rst tube we have arranged an optical resonator to increase the power available for WISP production. A light-tight plug is placed at the inside end of the second tube as our wall to absorb photons leaving the cavity towards the regeneration tube. As the light source a continuous-wave emitting laser at 1064 nm wavelength is used. For detection reasons this light is frequency doubled with a second harmonic generator before it enters the optical resonator that constitutes the production part. Regenerated photons are re-directed by an oblique mirror and focused by a lens into a few pixels of our CCD camera. The volume containing the hypothetical regenerated beam is sealed against environmental light entering into the signal region of the camera. Right at the entrance of the resonator a very small part is separated from the incident beam as a reference for the alignment of the beam axis of the resonator. This reference beam is guided along the magnet's side from the laser bench to the detection bench. After very strong attenuation the reference beam is also focused by the same lens onto a dierent position of the CCD. In the remainder of this section we present the most relevant characteristics of the dierent components of our setup. 4

100 MHz with a re-

sponse bandwidth of 100 kHz. Slow frequency drifts can be compensated by controlling the crystal temperature with a tuning coecient of -3 GHz/K. Before amplication the NPRO beam is passed through an electro-optic modulator (EOM).

AM-2

Cavity
CM

EM

Absorber

AM-1 PM L5 L4 PD3
Power meter Power meter

SM Pol-BS4 (pivoted) L1 P4 L2

AR window Vacuum tube WEBCam

AR window

Magnet 5T
Vacuum tube Attenuator

SBIG 4 Camera
WEBcam for reference Beam

PD2 L3 Pol-BS3 D1 P3

Reference Laser Beam

LASER
Dump Laserdiodes 808 nm NPRO 1064 nm 2W

screen for reference beam

Detector Bench
30 cm Nd:YVO 4 Crystals

PPKTP Oven

Laser Bench
1m

Power meter

P2 Pol-BS2 P1 Pol

Power meter

EOM

Isolator 35 W Laser

Amplifier

Figure 3:

Schematic overview of the whole experimental setup of the ALPS experiment comprising laser, second harmonic generation in

the PPKTP crystal, production cavity, magnet and detector. Magnied is the schematic representation of the ALPS master-oscillator power amplier laser system.

Outside the laser box, the amplier power can be monitored with a photo-detector (PD1) placed behind a highly reective mirror. The amplier output beam has a polarization extinction ratio of more than 20 dB. A polarizing beam splitter (Pol-BS1) is used to lter out further possible residual light in unwanted polarization states.
3.2. Second harmonic generation

are

to be

1 mm 2 mm 2 cm and its nonlinearity was measured de 7.9 1012 m/V. It is placed inside an oven

to stabilize its temperature at around 38 C in order to maintain phasematching of the infrared and green (SHG) waves. Two lenses (L1, L2, see Fig. 3) are used to focus the infrared beam into the crystal to a waist size of

135 m.

This waist size is a compromise between high conversion eciency, the risk of damaging the crystal and degradation of the green beam shape. The input polarization can be adjusted to maximum conversion eciency by a a variable attenuator consisting of another Our experiment uses a detector with a silicon CCD chip whose sensitivity is strongly peaked around the visible spectral region while it approaches zero for wavelengths above 1000 nm [116]. In the phenomenon of oscillations the regenerated light has the same characteristics as the laser beam in the WISP production vacuum tube [117]. Therefore we convert the infrared laser light from 1064 nm to green 532 nm light exploiting the nonlinear eect of second harmonic generation [118, 119] (SHG). As the nonlinear material we use PPKTP (periodically poled KTiOPO4 ) which shows high conversion eciencies due to its high intrinsic non-linearity. It is fabricated from a ux-grown KTP crystal by reorientation of its optical axis in periods of approximately

/2-plate
(P1)

(P2). The input power level of the infrared beam is set via

/2-plate

and a polarizing beam splitter (Pol-BS2). Behind the oven the converted green light is separated from the infrared by means of a dichroic mirror (D1) which is followed by a variable attenuator for the green light beam (P3, Pol-BS3) and a collimating lens (L3). Then the light passes a Faraday rotator that forms an optical diode together with the polarizing beamsplitter Pol-BS3 to protect the SHG from any back-reection from the following optical setup. With this single-pass SHG scheme a long-term stable output power of 800 mW at 532 nm is available behind D1 when the infrared beam is set to its maximum power level of 35 W.

9 m

length by strong

electrical poling leading to a much greater conversion efciency and an easier phasematching condition than for conventional SHG crystals [119]. The crystal's dimensions 5

In steady-state the transversal shape of the green beam acquires a slightly doughnut-like form, probably due to nonlinear absorption processes inside the crystal at high intensities [120]. This slight deviation of the green beam shape from a pure Gaussian lowers the coupling eciency to the optical resonator to approximately

80%

but does

not inuence the LSW experiment in any other way. To increase the SHG eciency we plan to change to a resonant second harmonic generation. an order of magnitude.
3.3. Magnet

This should in-

crease the available green light power by approximately

The production eciency of some types of WISPs depends strongly on the strength and length of the background magnetic eld (Sec. 2). The ALPS experiment at DESY uses a spare HERA dipole magnet [121]. More than 400 of these magnets are incorporated in the HERA proton storage ring, which operated from 1991 to end of June 2007. A HERA dipole magnet has an outer length of 9.8 m and is operated at cryogenic temperatures of 4.2 K. At ALPS we operate the dipole at a 5.30 T magnetic eld. Apart from a few quenches we did not encounter any difculties. The bent magnet bore is shielded by an anti-cryostat allowing to perform experiments inside the magnet at room temperature. bore. The temperature is stabilized by ushing Nitrogen at a constant rate and temperature through the As mentioned already in Sec. 3.1 the small clear aperture imposes special requirements on the laser's beam quality. The direction of the magnetic eld is vertical. Henceforth, the polarization of the laser light inside the cavity will be measured with respect to this direction.
3.4. Detector and detection bench
Figure 4: Selection of a typical one hour exposure recorded with the SBIG CCD. Tracks from radioactivity and cosmics are visible as well as warm pixels, i.e. pixels showing large pedestals or high noise rates. The signal (left) and reference (right) regions are shown enlarged. No evidence for photons from WISPs is seen while the reference beam shows up clearly and well focused.

and is probably caused by some environmental eects. To compensate for this eect, the ADUs of all pixels in an individual frame are shifted by the dierence of the average pixel value in this frame and an arbitrarily dened baseline. After this correction no hint for any additional systematic eect or unknown correlation among the pixels is found. Special care is required when mounting and xing all optical components on the detection bench because it has to be removed when the detector tube is pulled out or put into the magnet (as required by the data taking procedure, see Sec. 5). The exact position of the bench on In numerous the supporting structure (xed to the concrete of the experimental hall) is dened by special feet. tests it was proved that the focal points of reference beam and alignment beam do not change by more than one pixel on the CCD after removing and re-installing the detector bench. Re-converted photons are reected by the 45 in order to keep backgrounds as low as possible. In order to test the pointing stability a reference beam is split o the main beam in front of the cavity's input coupler (CM) and is guided from the laser bench along the side of the magnet to the detector bench. A small fraction of this reference beam is reected by the 45

The ALPS experiment uses a commercial astronomy CCD camera SBIG ST-402ME [116] as its detector for regenerated photons. It contains a Kodak KAF-0402ME silicon CCD chip with quantum eciency is of 532 nm. We operate the CCD chip at a temperature of -5 C where it shows a low dark current of 0.03 electrons per pixel and second and a readout noise of 17 electrons per pixel, independent of the exposure time. The pixel brightness is measured in ADUs (analog-to-digital units) with one ADU corresponding to 1.46 electrons. The camera allows sampling times between 0.04 s and 1 h. In order to lower the impact of the readout noise we used 1 h expositions in our search for WISPs. In dark frames (exposures with closed shutter) the values of all pixels are found to uctuate coherently. This is monitored by determining the average pixel ADU for each exposure after cutting against cosmics and other spurious signals. This average uctuates by up to 30 ADUs 6

765 510 60% for

pixels of 99

m2 .

The

main

light with a wavelength

mirror and focused by a lens on a few pixels of the CCD

window to the

detector box, attenuated and reected by two mirrors, so that it passes from behind through the main mirror and is focused on the CCD at a dierent location than the focus of the main beam. A typical CCD frame with the two regions of interest is shown in Fig. 4. In future experiments we plan to improve the signal to noise ratio by one order of magnitude by using a more sensitive camera and by reducing the signal spot size.

4. The ALPS cavity

the output coupling as small as possible in order to maximize the power build-up. But, even when choosing the best optics available and assembling the cavity under as clean ambient conditions as possible, some parasitic losses will always remain. Hence, one normally has to guess

The purpose of optical resonators (cavities) in the context of LSW experiments is to enlarge the light power in the WISP production region relative to the available primary laser power. Such an optical resonator acts therefore as an amplier for the luminosity of the whole experiment boosting its sensitivity. This is possible by coherent superposition of light elds that enter the resonator at dierent times corresponding to successive roundtrips. the regeneration tube, coupling mirror, The ratio of the laser power inside the resonator traveling towards

A in

advance during the design process of an optical resonator. The transmission coecient of the input coupler can then be chosen to maximize

PBmax .

To realize the full possible power build-up in the cavity, the alignment, shape and resonance frequency of the resonator eigenmode must be exactly matched by the incident laser beam. These parameters are dened by the alignment and radii of curvature (ROC ) of the cavity mir-

PCM ,

is called power build-up

P ,

to the power incident to the

PB .
(7)

PB = P /PCM .
spaced by a distance ecient of

rors and by the optical distance between them [123]. Single mode continuous-wave lasers emit most of their power into the fundamental transversal mode TEM00 . It is therefore most ecient to use spherical mirrors for the resonator such that the laser mode can be easily matched to the eigenmode of the cavity. However, every slight mismatch in beam shape or alignment causes some fraction of the incident power to overlap with higher order transversal eigenmodes of the cavity. In general, the higher order spatial modes have dierent resonance frequencies than the fundamental mode. Hence, this power fraction is not coherently enhanced when the fundamental mode is resonant. This eect is called mismodematching and it has to be minimized in order to maximize the intra-cavity power. In turn, such a cavity with non-degenerated resonant frequencies has the positive eect of behaving like a mode lter for the production beam, which also restricts the beam of re-converted photons to exist only in the TEM00 mode. Such a beam can then be focused to smaller spot sizes than beams consisting of multiple higher order transversal modes, which is a key issue when maximizing the signal to background ratio in our CCD camera. The full theory on modestructures of laser beams and resonators can be found in several publications, for instance [122, 123, 124]. Recall that, in addition to an optimized spatial over-

Consider a linear resonator consisting of two mirrors

fed by a laser of frequency

The

rst mirror or input coupler has a power transmission co-

TCM

and the second one of

TEM . A.

Absorption,

scattering and deection of the light during one roundtrip are combined into a parasitic loss factor quires a phase During a roundtrip between the mirrors the light ac-

= 2 (2L)/c.

Resonant enhancement

of the circulating light power is achieved when the light wave nearly reproduces itself after one roundtrip, i.e. when

= mod 2 0.

A cavity has therefore an innite num-

ber of resonances at frequencies acterized by integers

res,n = nc/(2L),

char-

n.

The frequency interval between

resonances is called free spectral range, Under the assumption that build-up by [122]

TCM ,

FSR = c/(2L). TEM , A and are all

small compared to unity one can approximate the power

PB

4 TCM (TCM + TEM + A) +

4 2

(8)

which in resonance gives

PBmax

4 TCM (TCM + TEM + A)

(9)

lap, the laser frequency has to match one resonance frequency

res .

Small uctuations of the cavity length

The errors of these approximations are negligible for all considerations that are made throughout this paper. The derivative of eq. (9) with respect to to input coupling.

induce correspondingly small relative changes of its resonance frequency given by

res /res = L/L.

To show

TCM

can be set

how even tiny length changes degrade the power build-up, we solve eq. (8) for an absolute length change

to zero to nd the maximum power build-up with respect This maximum is achieved if the socalled impedance matching condition is fullled

d that would

cause a reduction to one half of its resonant value,

TCM = TEM + A
transmission of the input coupler coupling of the cavity

(10)

PB ! 1 = PBmax 2

d=

1 (TCM + TEM + A) . 8

(11)

From this expression (and bearing in mind that we are Hence, the largest power build-up is obtained when the interested in the smallest possible values of

TCM +TEM +A

TCM

is chosen to be as

to maximize the power build-up) it becomes clear that even length uctuations on scales much smaller than the laser wavelength will cause huge changes of the power build-up of the cavity. A similar constraint for the necessary stability of the laser frequency can be deduced from equation (11) if one 7

close as possible to the sum of parasitic losses

A and output TEM

TEM . A
or

Derivation of eq.(9) with respect to that there is no maximum of

PBmax

over these parameters

clarifying that it is always best to keep parasitic losses and

substitutes

= c/

and solves for the laser frequency

at
beam radius at intensity fraction of 1/e2 /mm

3.5 3 2.5 2 1.5 1 0.5 0

which the power build-up is reduced to half its peak value. The distance between these two points on the frequency axis is called the full width at half maximum (FWHM) linewidth of the optical resonator and is given by

beam radius radial aperture causing 0.2% clipping losses position of end mirror at middle of magnet

(TCM + TEM + A) = FSR . 2

(12)

The laser frequency must be kept well within this FWHM linewidth to keep the power build-up near its peak value. The ratio

FSR/FWHM, which is a measure for the cavity's

frequency selectivity, is usually called nesse. In order to keep the power build-up stable, the dierence between the resonance frequency and the laser frequency can be minimized by a feed-back control system. This can either actuate on the cavity length or on the laser frequency. Once spatial modes and frequencies of the laser and the cavity are matched, the cavity will exhibit its full power build-up. This gain factor will remain fairly constant while scaling the incident power until it gets spoiled by thermal distortion or even by the destruction of the optics. The power inside the cavity can be easily measured by monitoring the power transmitted through the end mirror, per roundtrip due to clipping below
Figure 5: Beam radius of the fundamental eigenmode of the ALPS optical resonator together with the radius at which mode's power would be clipped.

2 4 6 8 10 12 14 distance in the direction of propagation, starting at input coupler CM /m

16

0.2 %

of the

The position of the end mirror

EM is shown as a vertical line. Clearly, the beam size is always well below the minimum aperture of our production vacuum tube (14 mm).

0.2 %.

The resonance

frequencies of all higher order transversal eigenmodes up to a mode index sum of 10 do not match the TEM00 resonance frequency within a linewidth. This simplies the readout of the length changes of the cavity, which is nec-

PEM

since, under the outlined approximations

PEM = TEM P .
4.1. Design of optical resonator and stabilization

(13)

essary for the feed-back control system explained below. The evolution of the beam radius of the chosen fundamental eigenmode is shown in Fig. 5 together with the required free circular aperture to keep losses due to clipping below the value mentioned above. Here beam radius corresponds to the lateral distance from the beam's propagation axis where the Gaussian intensity distribution of the fundamental TEM00 has fallen to a fraction of whose measured transmission is

The design of the ALPS optical resonator was constrained by several aspects. First, the site at DESY does not allow for the use of two magnets in a row so the production and regeneration regions of our LSW experiment must be located inside the same magnet. This requires that one mirror of the cavity has to be placed in the middle of the HERA magnet. Second, the bent magnet bore has an cat's pupil-shaped clear inner aperture of only 28 mm height and 14 mm width (see [56]). A linear resonator comprised by two mirrors was realized. One mirror, CM, is held in place by a rigid and manually adjustable mirror mount on an optical table in front of the magnet and the other mirror, EM, is mounted near the middle of the magnet on a self-made small and nonmagnetic mirror holder attached to the end of the WISP production vacuum tube (see [57]). The mirror holder for EM was designed such that it allows for remote alignment to compensate for alignment drifts. The distance between both mirrors is 8.6 m most of which is occupied by a vacuum tube with two anti-reectively coated windows (AR windows). Several stable resonator designs that dier in the radii of curvature of the cavity mirrors are possible [122, 123]. Our resonator uses a plano-concave design with one plane mirror (EM) and a curved one with

1/e2 .

As end mirror (EM) we chose a highly reecting mirror

TEM = 170 ppm.

Because

our resonator includes four antireective coated surfaces of the vacuum tube's windows (see Fig. 3), with rest reection coecient of input coupler.

0.3 %

each, we chose a power

transmission design value of approximately in a power transmission of

2%

for the

Accurate measurements on CM resulted

TCM 2.3 %.

Assuming the

realization of an impedance matched cavity according to equations (9) and (12) these values would result in a maximum power build-up of

PBmax 43

and a FWHM of

130 kHz.
In order to get as close as possible to this theoretical value, an optimization of the spatial overlap of laser and resonator mode has to be performed. This can be done with the two beam shaping lenses L4 and L5, the two alignment mirrors AM1 and AM2 and with the two adjustable mounts of the cavity mirrors CM and EM (see Fig. 3). If the incident laser beam as well as the cavity mode are perfect Gaussian TEM00 modes, then the combination of these actuators must be sucient to maximize the spatial overlap in all its degrees of freedom.

ROC = 15 m.

The

resulting fundamental cavity eigenmode needs a free circular aperture of only 6 mm diameter to keep power losses 8

The experimental site at DESY is exposed to strong acoustic and vibrational inuences. These ambient conditions give rise to large and fast length uctuations of our cavity as well as signicant alignment uctuations with strong impact on the power build-up. In order to compensate these uctuations we constructed an electronic feedback control loop that actuates on the laser frequency. The dierence between the laser frequency and the actual resonance frequency of the cavity is determined via a sideband modulation spectroscopy technique called Pound-Drever-Hall (PDH) readout scheme, with the help of a photodiode PD3 whose output is high-pass ltered [125, 126, 127]. The necessary phase modulation sidebands are generated by an electro-optic modulator (EOM) which is part of the laser system (see Sec. 3.1). After demodulation the resulting error signal is ltered by a controller and then amplied by a fast high voltage amplier. The resulting high voltage signal is then passed to the fast frequency actuator which is part of the laser system (see Sec. 3.1). This control loop has a bandwidth of approximately 28 kHz limited by mechanical resonances of the frequency actuator around 200 kHz. The length uctuations of the cavity are much larger than the range over which the PDH scheme would provide a valid error signal. Therefore, if the control loop is closed at an arbitrary cavity state the controller would in most cases not be able to bring cavity and laser into resonance. To solve this problem the loop is equipped with automatic lock acquisition electronics that monitors the light power that is reected back from the cavity via the DC signal of PD3. If the locked state is lost, the reected power rises above a user-dened level which is the trigger to scan the laser frequency and search for the next resonance. If this is found the system is automatically relocked. Intervention by an operator is not necessary. We record the DC signal of PD3 every 30 seconds in order to keep the information about the intracavity power for the data analysis. This signal was calibrated in test runs by measuring simultaneously the light passing through the mirror EM with known transmission

10 10
rms mismatch / ( / Hz )

2 1

(d) (a)

(b)

10 10

-1

10-2 10-3 10 10
-4 -5

(c)

(e)

10-6 100 101 102 103 104 105

(c),

that was taken from [128], with measurements

of the free-running

(d)

and stabilized

(e)

mismatch between the

laser and cavity resonance. The free-running mismatch is eectively suppressed for Fourier frequencies smaller than the bandwidth of the control loop linewidth

(b).

The power build-up is seriously degraded for

frequencies with larger rms noise than one half of the cavities FWHM

(a).
Furthermore,

prole of the cavity mode (see Sec. 3.4).

this is a rst hint suggesting that the cavity mode does not suer from large losses due to extensive clipping or heavy absorption. According to eq. (9) optical losses inside the cavity would reduce the achievable power build-up. Due to its length of 8.6 m and the way it must be mounted inside the magnet, it is not possible to assemble the cavity in a cleanroom environment. Therefore it is important to check if additional internal losses might be caused by dust or other impurities. In Sec. 4.1 we expected ity design.

PBmax 43

from our cav-

We measured this quantity by recording the

power transmitted through EM and determining the intracavity power We achieve a power build-up of

through eq. (13) in a dedicated setup.

PB = 44 2 which at rst

TEM 170 ppm. /2

glance seems to agree nicely with our expectations. However, a fast scan of the laser frequency over more than one free spectral range reveals that incident power

The polarization state of the laser light entering the cavity can be adjusted by a combination of a (P4) and a pivotable polarization beam splitter (Pol-BS4). By this combination a clean linear polarization state with an arbitrarily adjustable angle relative to the magnetic eld direction can be realized.
4.2. Characterization of the production resonator

20 % of the

PCM

is coupled to dierent higher order

transversal modes. This value could not be reduced even by careful optimization of the spatial overlap between laser beam and cavity mode. Therefore it most likely originates from distortions of the incident beam shape to some nonGaussian intensity prole by the SHG and from alignment uctuations of the long cavity [124]. In order to compare our measurement of

After installation the cavity was rst characterized at low incident powers

PCM

around 50 mW. The stabilized

PB

with the value of

PBmax

obtained

in the previous section one has to correct for the amount of

fundamental eigenmode of the ALPS cavity is observed with a CCD camera behind the mirror EM. It has a nice round shape without any observable uctuations or distortions. This is important for the detection stage of our LSW experiment because the position and intensity distribution of a possible beam of re-converted photons behind the wall is then stationary and well dened by the orientation and 9

PCM in higher order transversal modes.

This correction for

the mismodematching results in a measured power build-

PBTEM00 = 55 3
which is slightly higher than expected.

(14)

used fraction of actuator dynamic range /%

40 20

50 45 power build-up inside optical resonator /1 40 35 30 25 20 15 10 5

0 20 40 0 50

10 time /s

15

20

25

power build-up inside resonator /1

40

30

20 0

10 time /s

15

20

25

0 0

0.5

1 time /s

1.5

Figure 7:

Typical time series of the power build-up (bottom panel)

Figure 8: Typical power build-up inside the optical resonator during multiple automatic relock sequences caused by deliberately introducing a strong vibrational disturbance of the cavity length.

and the actuator signal (top panel) while the electronic frequency stabilization control loop is active.

Additionally, a value of

130 kHz

was expected.

nance frequency uctuations much greater than one half of the cavities FWHM linewidth reducing the average power build-up close to zero. But when the active stabilization system is turned on the rms frequency mismatch is largely reduced. Their remaining impact on the power build-up is virtually negligible. Our control loop is capable of stabilizing the laser to the cavity on longer timescales as well. In Fig. 7 a typical time series of remaining power build-up uctuations under lock is displayed. One can see that the laser light is kept resonant inside the cavity for tens of seconds. Less than

A direct measurement of this value resulted in

= 127

12 kHz,

in excellent agreement with the expectations.

Therefore, we have two independent ways of determining the internal losses in the cavity. Using the measured values of

PBTEM00

and

with eqs. (9) and (12) we ob-

tain for the parasitic losses introduced by the antireective coated surfaces inside our cavity values of and

0.22 0.01%

0.28 0.05%

per facet, respectively. These values are

compatible with each other. They turn out to be slightly smaller than the estimation used for the design of the resonator, which explains the fact that

PBTEM00

surpasses our

40 %

of the dynamic range of the frequency control actua-

expectations. Clearly, the windows inside the cavity dominate the overall losses. Eliminating the windows in future experiments could be the rst step towards a signicantly enlarged power build-up. To characterize the adequacy and performance of our frequency stabilization control loop we further investigate the free-running and stabilized rms uctuations of the frequency mismatch between laser and cavity. The measured spectral densities are shown in Fig. 6. Because the power build-up is seriously degraded if the frequency mismatch amounts to more than half of the cavities FWHM all measurements in Fig. 6 are normalized to its measured value of

tor is needed during this time so that slow drifts over even longer timescales can still be compensated. The remaining power build-up uctuations probably originate from changes of the alignment of the incident laser beam relative to the eigenmode of the resonator. An additional auto-alignment system may improve this situation and increase the average power build-up slightly. As far as the frequency control loop is concerned, very slow drifts of the cavity length saturate the actuator on timescales of some tens of minutes causing the stabilization to fail and hence the power build-up to be lost. In that case a fast relock was realized by means of the already mentioned automatic lock acquisition electronics and hence long measurement periods were possible. A timeseries of the power build-up during such a relock sequence is shown in Fig. 8. To stimulate a loss of the locked state of the resonator strong mechanical length uctuations were introduced into the system by deliberately hitting the optical table. While this disturbance is ringing down the stabilization fails several times because of limited dynamic range which causes the power build-up to drop to zero. This is detected by the automatic lock acquisition electronics that reacquire the locked state every time within only fractions of a second. In order to characterize the impact of alignment uctuations and to correlate the internal cavity power with

= 127 kHz.

The gure shows the squared sum of

the measured noise spectral densities of the control signal applied to the frequency actuator of the laser and of the error signal of the control loop. uation (free-running). This gives an estimate on the size of the mismatch in an uncontrolled sitAdditionally, the measured noise spectral density of the error signal of the control loop itself is shown which reects the remaining mismatch once the control loop is closed. Furthermore, a line is plotted that gives a rough estimate of the free-running laser frequency noise. One can see that the by far greatest contribution to the overall free-running frequency mismatch is due to cavity length noise. For Fourier frequencies below 100 Hz this free-running rms length noise causes reso-

10

P ! [W]

45 40 35 30 25 20 15 10 5 0 0 0.1 0.2 0.3 0.4 0.5

Prob PBTEM

00

0.1596 56.99 2.135

wall detached and the open tube re-installed to check the alignment. This third step marks the end of one data taking period. While data taking, one hour exposures are recorded by the CCD. Only frames for times where no technical problems (i.e. magnet quenches) occurred are kept for further analyses. These frames are checked for cosmics and other spurious tracks from radioactivity in a region of 25 by 25 pixels around the expected position of re-converted photons from WISPs. About 10% of all one hour frames are rejected by this requirement. Finally, re-converted photons are searched for by comparing the averaged sum of the ADU counts of all pixels

0.6

0.7

0.8

P CM - P PD3 [W]

in the signal region (see below) in data frames and dark frames. The latter ones are recorded with closed camera shutter and typically also with the magnet switched o and no or very low intensity laser light. Hence, any excess in the data frames could be a signature for WISP production. It should be stressed that this simple approach holds true as long as no hint for an excess is derived from the data. If an excess is found, more detailed studies are to be performed demonstrating the WISP origin of the recorded photons.
5.1. Signal search

P = PEM /TEM , as function of the power fed into the cavity, PCM PPD3 where PCM was kept equal to 0.8 W during the dierent measurements and PPD3 denotes
Figure 9: Power inside of the cavity, the cavity's reected power as measured by PD3. The slope gives by denition

PBTEM00 .

the power reected by CM we performed a set of calibration measurements. Their results are used to monitor the power in the cavity during data taking. For the calibration we introduced manually dierent small misalignments of the laser direction and the cavity's eigenmode and measured the power leaking through the mirror EM and the PD3 signal while the cavity was locked. The results are shown in Fig. 9. They show how the intra-cavity power

As sketched above, we used light-shining-through-themirror to align the detector optics and to dene the search region for re-converted photons on the CCD. However, a systematic dierence between the impact position of the light used for alignment and the light from photon-WISPphoton conversions arises from the non-perfect parallelism of front and back surface of the cavity end-mirror (EM). The traversing light is bent by the wedge shape of the mirror while the WISPs pass the mirror unaected. We used two methods to determine the angle of the wedge:

is degraded by small misalignments and how this is tracked by the power reected by CM. The slope of this relation gives by denition the power build-up

PBTEM00 .

The

agreement between this method and the value in eq. (14) demonstrates the overall consistency of our understanding of the cavity. Finally, let us turn to the characterization of the polarization. The laser light enters the optical cavity with a a well dened linear polarization. However, we have generally observed that this state changes inside the cavity. This is correlated with the fact that the AR windows are not perfectly perpendicular to the beam. We conrmed this by deliberately changing the angle between windows and laser light and observing the corresponding change of the polarization. The elimination of the AR windows inside the cavity should therefore solve this problem.

The mirror is rotated and the diameter of the circular path of a distant beamspot is measured. The angular distance between the light passing through the mirror and the secondary reection from the uncoated backside of the mirror is measured.

Both methods give an angle of 0.017 and back surfaces).

(0

for parallel front

With a distance of

m between

mirror and CCD and a lens in front of the CCD with a focal length of 60 mm one arrives at a dierence of
5. Data taking and analysis

8 m

between the position of the beamspots from the light used for alignment and the location of the re-converted photons from WISPs. The CCD has a pixel size of

The data taking at the ALPS experiment proceeds in three steps. First, the regeneration tube is installed without the wall. The small fraction of laser light transmitted through the central mirror (EM) attached to the end of the production tube is used to align the detector bench. In the second step, the detector tube is removed, the wall is attached, the detector tube is reinstalled and evacuated and data taking takes place. After a certain period (typically one week) the detector tube is removed again, the 11

9 m, hence the

search region is to be shifted by one pixel relative to the beamspot used for alignment (in our case upwards). The analogous eects caused by the AR windows close to the EM (inside the cavity) and at the end of the detector vacuum tube were also tested. However, the resulting angles were found to be negligible.

P ! [W]
ADU

50 45 40 35 30 25 20 15

25000 20000 15000 10000 5000 0 400 y p 398 ixe 396 l# 394 392 390 388 386

268

270

272

274

276

278

280

282
10 5 0 0 5 10 15 20 25 30 35 Measurement #

# x pixel

Figure 10: The beamspot prole of the laser light leaking through the mirror in the centre of the ALPS magnet.

Figure 11:

Monitored power in the optical resonator during the

collection of the data frames. The laser power was recorded each 30 s;

The beamspot prole of the re-converted photons is exactly the same as the beamspot used for alignment because the re-converted photons will have the same properties as the laser photons. The secondary reection mentioned above will hit the CCD 5 pixels below the primary beamspot and is, as a conservative approach, not subtracted here because of its low intensity (less than 4% of the primary spot). Therefore the signal prole used in Figure 10 shows the this analysis is a little wider than the beamspot prole of re-converted photons from WISPs. measured beamspot prole. Taking this beamspot prole and assuming a possible misalignment of

only each hundredth point is shown for simplicity. The average power relevant for the measurements (P ) is determined to be This corresponds to a power build-up of with the characterization described in Sec. 4.2.

34 4 W. 435 in excellent agreement

5.2. ALPS exemplary run

The data set used for this analysis consists of 25 CCD frames with an exposure of one hour each. These frames were selected out of 31 frames taken during one week. Three frames, where a magnet quench occurred, are not taken into account for the data analysis. Another three frames are rejected, because the visual inspection of these frames shows hints for cosmic ray activity or radioactivity in the 25 by 25 pixel region around the position for the expected signal. One hour long dark frames with the camera shutter closed were recorded to be compared with the data frames. After cutting against cosmics and radioactivity as described above 208 frames remain for the analysis. The cavity internal power during data taking is shown in Fig. 11. We achieved

pixel in each direction (corresponding

to the re-positioning accuracy mentioned above) at least 58% of the signal is contained in a region. A hypothetical signal is searched for by comparing the mean value of the sum of the ADUs in the 25 pixels of the signal region in data frames and dark frames after correcting each frame for the pedestal uctuations. One principle dierence in both frame sets is the presence of light from the reference beam in the data frames. The photon ux of the reference beam at the CCD is about 6.5 Hz (2.4

55

pixel large search

P = 34 4 W .

(15)

1018

W). By comparing dark frames and

for the production of WISPs. The uncertainty originates from a conservative estimate of the systematic uncertainty of the calibration procedure, see Fig. 9. The linear polarization state of the eigenmode of the cavity during the data taking was measured (behind the mirror EM) to be 55 degrees with respect to the orientation of the magnetic eld. We have thus 6710% of the laser photons with perpendicular polarization and 3310% with parallel polarization, see Table 2. From the alignment test before data taking the central position of the search region was determined to be at the (276,391) pixel coordinates of the CCD, while the alignment test after data taking gave the position (275,392). The dierence by

frames with the reference beam shining on the CCD we proved that the reference light does not aect the signal region. The distance between the signal region and the reference beam spot position amounts to 2 mm on the chip (230 pixels, see Fig. 4). After performing the pedestal correction, the dark frame data sets taken over the course of two years agree within statistical uncertainties. No long term drift or other changes are observed. As a nal check the dark frame data set (recorded in the course of 24 days) used for the analysis shown below is divided into a rst and second half (104 frames each). Mean and width of the distribution of the sum of the pixels in the signal region agree very well within statistics, so that no hint on systematic dierences is found.

pixels matches our experience for

the re-positioning accuracy derived from numerous tests before. The distribution of the ADU sums in the search region is presented in Fig. 12 for both data and dark frame sets. Each frame is corrected for the pedestal uctuation (see above) and gives one entry. 12

10

Data frames

Prob = 1 25 Constant 4.799 2.487e+004 1.175 67 1 Mean 2.488e+004 1.175 4.799 14 2.488e+004 Sigma 69.28 69.28 14 9.79 9.79
Entries Mean RMS Prob Constant Mean Sigma

h101

entries

ADU
24600 24800 25000 25200
Entries Mean RMS Prob Constant Mean Sigma

8 6 4 2

25400
h101

ADU

0 -2 -4 -6 395 y p 394 ixe l # 393392

Dark frames
10

Prob = 1 208 Constant 41.27 2.488e+004 3.50 66.21 1 Mean 2.487e+004 3.50 41.27 5 2.487e+004 5 Sigma 67.02 67.02 3.29 3.29

entries

391 390

389

272

273

274

275

276

277

278

# x pixel

24600

24800

25000

25200

25400

ADU

Figure 12:

Distribution of the sums of the ADUs in the 5x5 pix-

Figure 13:

The dierence between the signal (average of 25 A signal of photons

els dening the search region for photons from re-converted WISPs. Each CCD frame (25 data frames, top, and 208 dark frames, bottom) gives one entry. The boxes show the results of Gaussians tted to the distributions.

data frames) and background (average of 208 dark frames) ADUs shows no excess above noise expectations. from re-converted WISPs should be centered around the position

(275 1, 392 1).

Polarization Parallel Perpendicular Independent

Flux

(3.0 0.9) 10 Hz (6.1 0.9) 1019 Hz (9.1 0.1) 1019 Hz

19

2.614.6

Table 3: The dierence of the average of the sums of ADU values of data and dark frames in the search region is shown in the rst column. No signicant excess is observed. The result is used to determine a 95% C.L. ux limit for photons from re-converted WISPs.

Eective laser power relevant to the search for WISPs.

Parallel, perpendicular and independent denote the orientation of the laser light polarization with respect to the magnetic eld direction.

6. Conclusions

It is obvious that there is no hint for any signal in the data frames: mean and width of both distributions agree very well within statistics. Also searching for a signal at other positions (see Fig. 13) does not reveal any signicant excess. Hence we do not observe any evidence for WISP production. To derive upper limits for the WISP induced photon ux from our data we have to consider the quantum efciency of the camera (0.60.1), the electron/ADU conversion factor of the camera (1.460.05) and the fraction of the signal contained in the search region (0.750.15). This gives a ux limit of 28 mHz at 95% C.L. (see Table 3) using the method of [129]. Taking into account the eective ux of incoming photons from Table 2 one arrives at the conversion probabilities in Table 4. For determining the limits on the photon-WISP mixing strength as a function of the WISP mass we use the experimental parameters summarized in Fig. 15. The uncertainties of the magnetic eld strength,

The ALPS collaboration runs a light-shining-throughwalls (LSW) experiment to search for photon oscillations into weakly interacting sub-eV particles (WISPs) inside of a superconducting HERA dipole magnet at the site of DESY. In this paper we have described and characterized our apparatus and demonstrated the data analysis procedures. Our main result is the rst successful integration of a largescale optical resonator into a complete LSW experiment. This resonator serves as an amplier for the photon ux in the production region of the experiment and thus boosts the experiment's sensitivity. During a 31 hours lasting exemplary run, the available laser light power to search for WISPs was increased by a factor of 43. The upper limits on photon-WISP interactions derived from this brief run show that ALPS is very competitive with other state-of-the-art LSW experiments (see Fig. 14). Moreover, we have identied the main limitations of our current set-up, showing that signicant improvements, especially of the performance of the optical resonator, are possible. This opens a clear path-way for near-term future steps in increasing the sensitivity of the ALPS experiment.

B ,

(when rel-

evant, i.e. for ALPs and MCPs) and of the tube lengths,

L ,

are taken into account by a Monte Carlo calculation.

Here we conservatively use

B = 0.01

T and

L = 0.1

m.

The results are shown in Fig. 14 and in Table 5.

13

LIPSS

BMV

g+ [GeV1 ]

10

g [GeV1 ]

BFRT
OSQAR

10

BFRT
OSQAR

GammeV
10
7

ALPS

GammeV
10
7

ALPS

10

10

10

10

m+ [eV]
5

m [eV]
OSQAR

10

Coulomb
6

Sun & CAST

10

BFRT

10

ALPS
BMV GammeV LIPSS
10
6

hC

10

FI

10

BMV GammeV LIPSS

ALPS

RA

M
4

10

10

10

10

10

m [eV]

mMCP [eV]

Figure 16: Exclusion limits (95% C.L.) from the ALPS exemplary run (black). Starting from up-left corner and moving clockwise the limits Sec. 2 for details and references. As a comparison we show exclusion limits of previous LSW experiments: BFRT [27], BMV [99, 100], GammeV [101, 102], LIPSS [103, 104] and OSQAR [105] The results of a more recent OSQAR publication [106] cannot be compared due to As a comparison we show exclusion limits of previous LSW experiments: BFRT [47], BMV [130, 131], a dierent interpretation. In all cases we present the bounds with 95% C.L. In the hidden photon parameter space we have also plotted the GammeV [132, excluded from tests of the Coulomb law [107, 108], distortions of the cosmic microwave background OSQARformation of a hiddencannot be regions 133], LIPSS [134, 135] and OSQAR (preliminary) [136]. The results of a more recent and late publication [137] compared due to a dierent interpretation. In all cases hidden photons from the Sun [4]. Similar exclusionhidden photon case of ALPsspace we have photon microwave background [10], and emission of we present the bounds with 95% C.L. In the bounds for the parameter and also plotted the regions excluded from[2, 15] of the Coulomb law [138, 139], distortions of the cosmic microwave background and late formation MCPs are review for instance in tests
state for scalar and pseudoscalar ALPs, mini-charged particles (in models with a massless hidden photon) and massive hidden photons. See Sec. 2 for details and references. of a hidden photon microwave background [11], and emission of hidden photons from the Sun [4]. Similar exclusion bounds for the case of . ALPs and MCPs are review for instance in [2, 19].

state for scalar and pseudoscalar ALPs, mini-charged exemplary run (black). Starting from photon) and massive hidden photons. See Figure 14: Exclusion limits (95% C.L.) from the ALPS particles (in models with a massless hidden up-left corner and moving clockwise the limits

Polarization Parallel Perpendicular Independent

Prob.10

22

95% C.L.

99% C.L.

WISP ALP ALP

95%C.L. Limit

Mass

0.8 4.4 0.4 2.2 0.3 1.4

9.4 4.5 3.0

12 5.8 3.9

0 0+

16

HP MCP+HP

Probability of light-shining-through-walls tested with

g < 4.9 107 GeV1 g < 4.1 107 GeV1 < 2.1 106 QMCP < 3.5 106

m 0.5 meV m+ 0.5 meV m 0.7 meV mMCP 0.1 eV

the ALPS exemplary run and corresponding upper limits. Parallel, perpendicular and independent denote the orientation of the laser light polarization with respect to the magnetic eld direction.

Table 5: Upper limits for the search for WISPs in the data presented in this paper.

Acknowledgments

A. Zuber for the technical drawings. We acknowledge nancial support from the Helmholtz Association and from the Centre for Quantum Engineering and Space-Time Research (QUEST).

The ALPS collaboration thanks the MKS group at DESY for support, especially H. Brck, E. Gadwinkel, C. Hagedorn, H. Herzog, K. Prss and M. Stolper. Many thanks to U. Koetz for advice for selecting the camera and 14

CM

AR window

Vacuum tube EM Absorber AR window

Magnet

[29] R. Battesti [30] [31] [32]

Vacuum tube

arXiv:0705.0615 [hep-ex].
CAST CAST CAST
(2007) 010,

et al., Lect. Notes Phys. 741 (2008) 199237,

Collaboration, S. Andriamonje Collaboration, E. Arik

et al., JCAP 0704

4.3 m 8.8 m, 5.3 T 6.3 m Cavity 8.6 m

4.3 m 7.6 m

arXiv:0810.4482 [hep-ex]. [hep-ex].

et al., JCAP 0902 (2009) 008, et al., arXiv:0809.4581

Collaboration, G. Cantatore

[33] Y. Inoue

arXiv:0806.2230 [astro-ph]. arXiv:0804.3543 [astro-ph].

et al., Phys. Lett. B668 (2008) 9397, JCAP 0808 (2008) 026, et al., Phys. Rev. D74

Figure 15: Summary of the relevant experimental parameters for the ALPS WISP search.

[34] H. Davoudiasl and P. Huber, [35] [36]

References
[1] L. F. Abbott and P. Sikivie, 133136. [2] G. G. Raelt, [3] G. G. Raelt, [4] J. Redondo,

Phys. Lett. B120 (1983)

[37]

arXiv:astro-ph/0603108. A. C. Melissinos, arXiv:0807.1092 [hep-ph]. B. Dobrich and H. Gies, arXiv:0904.0216 [hep-ph].

Collaboration, L. D. Duy

[38] J. Jaeckel and A. Ringwald, [39] J. Jaeckel and J. Redondo,

Stars as laboratories for fundamental physics. Ann. Rev. Nucl. Part. Sci. 49 (1999) 163216,

arXiv:0707.2063 [hep-ph].

Phys. Lett. B659 (2008)

The University of Chicago Press, 1996. 664 p.

arXiv:0806.1115 [hep-ph].

Europhys. Lett. 84 (2008) 31002, Phys. Rev. Lett. 97 Europhys. Lett. 76

hep-ph/9903472.

[40] H. Gies, J. Jaeckel, and A. Ringwald, (2006) 140402, [41] H. Gies, J. Jaeckel, and A. Ringwald, (2006) 794800, [42] H. Gies and J. [43] L. B. Okun, [44] A. A. Anselm,

[hep-ph].
180184,

JCAP 0807 (2008) 008, arXiv:0801.1527 Phys. Lett. B664 (2008) Nature 306

arXiv:hep-ph/0607118.

[5] S. N. Gninenko and J. Redondo,

arXiv:0804.3736 [hep-ex].

[6] H. Georgi, P. H. Ginsparg, and S. L. Glashow, (1983) 765766. [7] A. De Angelis and R. Pain, 24912496,

Sov. Phys. JETP 56 (1982) 502. Yad. Fiz. 42 (1985) 14801483.

arXiv:hep-ph/0608238. Jaeckel, arXiv:0904.0609 [hep-ph].

arXiv:astro-ph/0205059.

Mod. Phys. Lett. A17 (2002) Phys. Rev. D72 Phys. Rev. D76

[45] K. Van Bibber, N. R. Dagdeviren, S. E. Koonin, A. Kerman, and H. N. Nelson, [46] [47]

[8] A. Mirizzi, G. G. Raelt, and P. D. Serpico, (2005) 023501, (2007) 023001,

BFRT BFRT LIGO

Collaboration, G. Ruoso

Phys. Rev. Lett. 59 (1987) 759. et al., Z. Phys. C56 (1992) et al., Phys. Rev. D47

arXiv:astro-ph/0506078.

505508. Collaboration, R. Cameron (1993) 37073725. [48] H. Luck [49]

[9] A. Mirizzi, G. G. Raelt, and P. D. Serpico, [10] A. Mirizzi, J. Redondo, and G. Sigl,

arXiv:0704.3044 [astro-ph].

arXiv:0901.0014 [hep-ph].
101 D76 D77
(2008) 131801, (2007) 121301, (2008) 123503,

JCAP 0903 (2009) 026, Phys. Rev. Lett. Phys. Rev. Phys. Rev.

et al., Class. Quant. Grav. 23 (2006) S71S78.

Collaboration,

[11] J. Jaeckel, J. Redondo, and A. Ringwald,

[50] F. Hoogeveen and T. Ziegenhagen,

Nucl. Phys. B358 (1991) Phys. Rev. Lett.

arXiv:0804.4157 [astro-ph].

[12] A. De Angelis, O. Mansutti, and M. Roncadelli,

[51] P. Sikivie, D. Tanner, and K. van Bibber, [52] [53] [54]

arXiv:0707.4312 [astro-ph]. arXiv:0711.3528 [astro-ph].

[13] J.-H. Huh, J. E. Kim, J.-C. Park, and S. C. Park,

[14] J. Isern, S. Catalan, E. Garcia-Berro, and S. Torres,

172002 (2007) . PVLAS Collaboration, E. Zavattini 96 (2006) 110406, hep-ex/0507107. PVLAS Collaboration, E. Zavattini
(2008) 032006,

et al., Phys. Rev. Lett. et al., Phys. Rev. D77

arXiv:0812.3043 [astro-ph].

[15] J. Isern, E. Garcia-Berro, S. Torres, and S. Catalan,

Q&A

arXiv:0706.3419 [hep-ex].

arXiv:0806.2807 [astro-ph]. [[Link]].

Mod. Phys. Lett. A22 (2007) 28152831,

Collaboration, S.-J. Chen, H.-H. Mei, and W.-T. Ni,

[16] C. Burrage, A.-C. Davis, and D. J. Shaw, [17] M. Fairbairn, T. Rashba, and S. Troitsky, [18] M. Ahlers, [19]

arXiv:0902.2320 arXiv:0901.4085

[55] R. Battesti, B. D. Souza, S. Batut, C. Robilliard, and Others, [56] [57]

[[Link]]. arXiv:0904.0998 [hep-ph]. J. Redondo, arXiv:0810.3200 [hep-ph]. hep-ph/0504202. hep-ph/0606163.

arXiv:0710.1703vl [[Link]] (2007) 112. ALPS Collaboration, K. Ehret et al., hep-ex/0702023.

[58] C. T. Hill and G. G. Ross, [59] Y. Liao,

[20] E. Masso and J. Redondo, [21] E. Masso and J. Redondo,

JCAP 0509 (2005) 015,

[hep-ph].

Phys. Lett. B650 (2007) 257261, arXiv:0704.1961

Nucl. Phys. B311 (1988) 253.

arXiv:0812.3495 [hep-ex].

Phys. Rev. Lett. 97 (2006) 151802, Phys. Rev. Lett. 98 (2007)

[60] J. Redondo,

Can the PVLAS particle be compatible with the astrophysical bounds? PhD thesis, IFAE (UAB), 2008.
arXiv:0807.4329 [hep-ph].

[22] R. N. Mohapatra and S. Nasri,

arXiv:hep-ph/0610068.

[61] Y. Nambu and G. Jona-Lasinio, 345358.

Phys. Rev. 122 (1961) Phys. Rev. 127

[23] J. Jaeckel, E. Masso, J. Redondo, A. Ringwald, and F. Takahashi,

hep-ph/0610203.

Phys. Rev. D75 (2007) 013004,

[62] J. Goldstone, A. Salam, and S. Weinberg, (1962) 965970. [63] S. Weinberg, [65] [66]

[24] P. Brax, C. van de Bruck, and A.-C. Davis,

99

Phys. Rev. Lett.

(2007) 121103,

[64] S. Weinberg,

[25] D. D. Stancil,

arXiv:0704.0490 [hep-ph]. [hep-ph].

Phys. Rev. D76 (2007) 111701,

Phys. Rev. Lett. 29 (1972) 16981701. Phys. Rev. Lett. 40 (1978) 223226. F. Wilczek, Phys. Rev. Lett. 40 (1978) 279282. R. D. Peccei and H. R. Quinn, Phys. Rev. Lett. 38 (1977)
14401443.

[26] J. Jaeckel, J. Redondo, and A. Ringwald, [27] S. Abel

[67] R. D. Peccei and H. R. Quinn, 17911797.

Phys. Rev. D16 (1977) Phys. Lett. B104

et al., The Physics Case for a Low Energy Frontier Prog. Part. Nucl. Phys. 62 (2009) 102134.

of Fundamental Physics. In preparation., 2009. [28] E. G. Adelberger, J. H. Gundlach, B. R. Heckel, S. Hoedl, and S. Schlamminger,

[68] M. Dine, W. Fischler, and M. Srednicki, (1981) 199. [70] J. E. Kim, [69] A. R. Zhitnitsky,

Sov. J. Nucl. Phys. 31 (1980) 260. Phys. Rev. Lett. 43 (1979) 103.

15

[71] M. A. Shifman, A. I. Vainshtein, and V. I. Zakharov,

Phys. B166 (1980) 493.

Nucl.

[111] M. Ahlers, H. Gies, J. Jaeckel, J. Redondo, and A. Ringwald, [112] G. Raelt and L. Stodolsky, [113] 015012,

[72] L. M. Krauss and F. Wilczek,

Phys. Lett. B173 (1986) Phys. Lett. B172 Phys. Phys.

Phys. Rev. D76 (2007) 115005, arXiv:0706.2836 [hep-ph]. Phys. Rev. D37 (1988) 1237. C. Biggio, E. Masso, and J. Redondo, Phys. Rev. D79 (2009)
hep-ph/0604062.
B. Willke, and D. Kracht, 459465.

[73] R. D. Peccei, T. T. Wu, and T. Yanagida, (1986) 435. [74] F. Wilczek,

Phys. Rev. Lett. 49 (1982) 15491552.

[114] M. Frede, B. Schulz, R. Wilhelm, P. Kwee, F. Seifert,

Optics Express 15 (2007) no. 2,

[75] Y. Chikashige, R. N. Mohapatra, and R. D. Peccei,

Rev. Lett. 45 (1980) 1926.

[115] T. Kane and R. Byer,

Optics Letters 10 (1985) no. 2, 6567.

[76] Y. Chikashige, R. N. Mohapatra, and R. D. Peccei,

Lett. B98 (1981) 265.

[116] [Link] [117] S. L. Adler, J. Gamboa, F. Mendez, and J. Lopez-Sarrion,

[77] E. J. Chun, J. E. Kim, and H. P. Nilles, [78] A. E. Nelson and N. Seiberg, 4662,

Nucl. Phys. B370

Annals Phys. 323 (2008) 28512872, arXiv:0801.4739

arXiv:hep-ph/9309299.

Nucl. Phys. B416 (1994) Nucl. Phys. B746

[118] G. D. Boyd and D. A. Kleinman,

Journal of Applied Physics

[79] C. Coriano', N. Irges, and E. Kiritsis, (2006) 77135, [80] C. Coriano and N. Irges,

arXiv:hep-ph/0510332.

[119] S. Spiekermann, F. Laurell, V. Pasiskevicius, H. Karlsson, and I. Freitag, [120] S. Wang, V. Pasiskevicius, and F. Laurell,

arXiv:hep-ph/0612140. arXiv:hep-th/0605206.

Phys. Lett. B651 (2007) 298305, JHEP 06 (2006) 051,

Physics 96 (2004) no. 4, 20232028. Accelerator Magnets.

Applied Physics B 79 (2004) 211219. Journal of Applied Superconducting

[81] P. Svrcek and E. Witten, [82] C. Wetterich, [84] [85]

[121] K. H. Mess, P. Schmueser, and S. Wol,

[83] P. J. E. Peebles and B. Ratra,

Nucl. Phys. B302 (1988) 668. Astrophys. J. 325 (1988) L17. J. A. Frieman, C. T. Hill, A. Stebbins, and I. Waga, Phys. Rev. Lett. 75 (1995) 20772080, arXiv:astro-ph/9505060. J. Khoury and A. Weltman, Phys. Rev. Lett. 93 (2004)
171104,

[122] A. Siegman,

Lasers.

University Science Books, 1986.

(corrected corresponding to errata list published at [123] H. Kogelnik and T. Li, 15501567.

[Link] ).

[86] A. Dupays, E. Masso, J. Redondo, and C. Rizzo, [87] [88] [89] [90]

Phys. Rev. Lett. 98 (2007) 131802, hep-ph/0610286. P. Fayet, Phys. Lett. B96 (1980) 83. P. Fayet, Nucl. Phys. B187 (1981) 184. R. Barbieri and S. Cecotti, Z. Phys. C33 (1986) 255. R. Foot, A. Y. Ignatiev, and R. R. Volkas, Phys. Lett. B503 Phys. Rev. 52 (1937) 4142. Helv. Phys. Acta 11 (1938) 225. B. Holdom, Phys. Lett. B166 (1986) 196. R. Foot and X.-G. He, Phys. Lett. B267 (1991) 509512. K. R. Dienes, C. F. Kolda, and J. March-Russell, Nucl. Phys.
B492
(1997) 104118, (2001) 355361,

astro-ph/0309300.

Applied Optics 5 (1966) no. 10,

[124] P. Kwee, F. Seifert, B. Willke, and K. Danzmann, [125]

Review of Scientic Instruments 78 (2007) no. 7, 073103. R. Pound, The Review of Scientic Instruments 17 (1946)
no. 11, 490505. A. Munley, and H. Ward, 97105.

[126] R. Drever, J. Hall, F. Kowalsky, J. Hough, G. Ford,

astro-ph/0011156.

Applied Physics B 31 (1983) no. 2,

[91] E. C. G. Stueckelberg, [92] E. C. G. Stueckelberg, [93] [94] [95]

[127] E. Black,

Am. J. Phys. 69 (2000) no. 1, 7987.

[128] M. Heurs, V. Quetschke, B. Willke, and K. Danzmann, [129] G. Feldman and R. Cousins, 38733889. [130] [131] [132]

Optics Letters 29 (2004) no. 18, 21482150. Phys. Rev. D (1998) no. 57,
BMV BMV
Collaboration, C. Robilliard Collaboration, M. Fouche (2007) 190403, (2008) 032013,

hep-ph/9610479.

[96] S. A. Abel and B. W. Schoeld, 150170,

arXiv:hep-th/0311051.

Nucl. Phys. B685 (2004) Phys.

arXiv:0707.1296 [hep-ex]. arXiv:0808.2800 [hep-ex].

et al., Phys. Rev. Lett. 99

et al., Phys. Rev. D78 Rev. Lett.

[97] S. A. Abel, J. Jaeckel, V. V. Khoze, and A. Ringwald,

Lett. B666 (2008) 6670, arXiv:hep-ph/0608248.

[98] S. A. Abel, M. D. Goodsell, J. Jaeckel, V. V. Khoze, and A. Ringwald,

[hep-ph].
115001,

JHEP 07 (2008) 124, arXiv:0803.1449

GammeV Collaboration, A. S. Chou et al., Phys. 100 (2008) 080402, arXiv:0710.3783 [hep-ex].

[133] M. Ahlers, H. Gies, J. Jaeckel, J. Redondo, and A. Ringwald, [134] [135] [136]

[99] D. Feldman, Z. Liu, and P. Nath,

arXiv:hep-ph/0702123.

Phys. Rev. D75 (2007) Phys. Lett. B662

Phys. Rev. D77 (2008) 095001, arXiv:0711.4991 [hep-ph]. LIPSS Collaboration, A. Afanasev et al., arXiv:0810.4189
[hep-ex].
LIPSS Collaboration, A. Afanasev et al., Phys. Rev. Lett. 101 (2008) 120401, arXiv:0806.2631 [hep-ex]. OSQAR Collaboration, R. Ballou et al., Summary of
OSQAR First Achievements and Main Requests for 2008. CERN-SPSC-2007-039S; PSC-M-762, December, 2007.

[100] M. Pospelov, A. Ritz, and M. B. Voloshin, (2008) 5361,

arXiv:0711.4866 [hep-ph].

[101] J. Redondo and M. Postma,

arXiv:0811.0326 [hep-ph].
003,

JCAP 0902 (2009) 005, JCAP 0901 (2009) Phys. Lett.

[102] A. Ibarra, A. Ringwald, and C. Weniger,

arXiv:0809.3196 [hep-ph].
(2009) 7176,

[137]

OSQAR

Collaboration, P. Pugnat

[103] C.-R. Chen, F. Takahashi, and T. T. Yanagida,

B671

arXiv:0809.0792 [hep-ph].

(2008) 092003, 22852287.

arXiv:0712.3362 [hep-ex].

et al., Phys. Rev. D78

[138] D. F. Bartlett and S. Loegl,

Phys. Rev. Lett. 61 (1988) Phys. Rev. Lett.

[104] N. Arkani-Hamed, D. P. Finkbeiner, T. R. Slatyer, and

[hep-ph].
063509,

N. Weiner,

Phys. Rev. D79 (2009) 015014, arXiv:0810.0713 Phys. Rev. D79 (2009)

[139] E. R. Williams, J. E. Faller, and H. A. Hill,

26

(1971) 721724.

[105] D. Feldman, Z. Liu, and P. Nath,

arXiv:0810.5762 [hep-ph].

[106] A. Ibarra, A. Ringwald, D. Tran, and C. Weniger,

arXiv:0903.3625 [hep-ph]. [hep-ph].

[107] B. Batell, M. Pospelov, and A. Ritz, [108] R. Rahman,

arXiv:0903.0363

arXiv:0901.2111 [hep-ph].

[109] B. Batell and T. Gherghetta,

arXiv:hep-ph/0512356.
D75
(2007) 035011,

Phys. Rev. D73 (2006) 045016, Phys. Rev.

[110] M. Ahlers, H. Gies, J. Jaeckel, and A. Ringwald,

arXiv:hep-ph/0612098.

16