Chapter 6

BRST Quantization
So far we have seen two methods that can be utilized to quantize the string: the covariant
approach and light-cone quantization. Each offers its advantages. Covariant quantization
makes Lorentz invariance manifest but allows for the existence of ‘ghost states’ (states
with negative norm) in the theory. In contrast, light-cone quantization is ghost-free.
However, Lorentz invariance is no longer obvious. Another trade-off is that the proof of
the number of space-time dimensions (D = 26 for the bosonic theory) is rather difficult in
covariant quantization, but its rather straightforward in light-cone quantization. Finally
identifying the physical states is easier in the light-cone approach.

Another method of quantization that in some ways is a more advanced approach is called
BRST quantization. This approach takes a middle ground between the two methods
outlined above. BRST quantization is manifestly Lorentz invariant, but includes ghost
states in the theory. Despite this, BRST quantization makes it easier to identify the
physical states of the theory and to extract the number of space-time dimensions
relatively easily.

BRST Operators and introductory Remarks

We begin by considering our old friend, Lie algebra. This is the algebra that is obeyed by
the familiar spin angular momentum operators of ordinary quantum mechanics. In
general, let some physical theory contain a gauge symmetry with operators K i . These
operators satisfy the Lie algebra:

 K i , K j   f ij k K k

k
where the f ij are called the structure constants of the theory. The structure constants
satisfy:

f ij m f mk l  f jk m f mi l  f ki m f mj l  0

The BRST quantization procedure begins with the following . We introduce two ghost
fields that are denoted by bi and c j which satisfy an anti-commutation relation given by:

 c ,b   
i
j
i
j

Futhermore  c , c    bi , b j   0 . Here the c i are ghost fields and the b j are ‘ghost
i j

momenta’.
Notice that since an anti-commutation relation is satisfied by the ghost fields, these fields
are fermionic. Now, recall that a field   z , z  has conformal dimension  h, h  provided
that it transforms under some conformal transformation z  w  z  as follows:

h h
 w   w 
  z, z        w, w 
 z   z 

The fields b and c are chosen such that they have conformal dimension 2 and -1, as we
will see below.

There are two operators that are constructed out of the ghost fields and the K i . The first
of these is the BRST operator which is given by:

1 k i j
Q  ci Ki  f ij c c bk
2

It is assumed that Q  Q† . We say that the BRST operator is nilpotent of degree two. This
means that if we square the operator we get zero:

Q2  0

Notice that (6.6) can also be expressed as  Q, Q  0 . We label the BRST operator with a
Q to imply that this is a conserved charge of the system, we often call this the BRST
charge. A second operator that is composed solely of ghost fields is called the ghost
number operator U. This is given by:

U   ci bi
i

This operator has integer eigenvalues. If the dimension of the Lie algebra is n, then the
eigenvalues of U are the integers 0,.., n . A state  has ghost number m if
U  m .

Example 6-1
Show that Q raises the ghost number by 1.

Solution
Using the anti-commutation relations for the ghost fields , notice that:
 Uc i K i   c r br c i K i   c r   ri  c i br  K i
r r

 c i K i  c i  c r br K i
r

 c i K i  c i K i  c r br  c i K i  c i K iU
r

Now consider some state  with ghost number m, that is U   m  . Then:

 UQ    U  ci Ki 
1 k i j 
f ij c c bk  
 2 

   1 
  Uc i K i  

 c b  2 f
r
r ij c c j bk  
k i

 r 

 1 
  c i K i  c i K iU  f ij k  c r br c i c j bk  
2
 Use Uc K
i
i  ci K i  c i K iU 
 r 

 1 
  c i K i  c i K iU  f ij k  c r   ri  c i br  c j bk   (Apply 6.3)
 2 r 

 1 1 
  c i K i  c i K iU  f ij k c i c j bk  fij k ci  c r br c j bk   (Use the properties of  ri )
 2 2 r 

 1 1 
  c i K i  c i K iU  f ij k c i c j bk  fij k ci  c r   rj  c j br  bk   (Apply 6.3 again)
 2 2 r 

 1 
  Q  c i K iU  fij k ci  c r   rj  c j br  bk   (Use 6.5 to write Q)
 2 r 

 1 1 
  Q  c i K iU  fij k ci c j bk  fij k ci c j  c r bk br   (Kroneker delta again)
 2 2 r 

 1 1 
  Q  c i K iU  fij k ci c j bk  fij k ci c j    rk  bk c r  br   (Yet again use 6.3)
 2 2 r 

 1 
  Q  c i K iU  f ij k c i c j  c r br  
 2 r 

  Q  QU     1  m   Q   (Apply U   m  )

Hence the BRST operator raises the ghost number by 1.

BRST-Invariant States
A state  is called BRST invariant if it is annihilated by the BRST operator :

Q   0   is BRST invariant.

BRST invariant states are the physical states of the theory. Since Q 2  0 , it follows that
any state   Q   0 is BRST invariant, since:

Q   Q2   0

Call   Q  a null state. Suppose that  is an arbitrary physical state so that

Q   0 . Notice that:

    Q     Q   0

Hence, any amplitude taken between a physical state and a null state vanishes. This
implies a result which is somewhat akin to the idea of a phase factor in ordinary quantum
i
mechanics. Two states a , b  e a represent the same physical state in quantum
mechanics since the resulting amplitudes are the same. Likewise, since any inner product
between a physical state and a null state vanishes, add a null state   Q  to a
physical state  generates a new state which is physically equivalent to  :

    Q 

because the second term on the right does not contribute to any inner product and hence
does not change the physical predictions of the theory, which are the amplitudes. Since Q
raises the ghost number of a state by 1, if the ghost number of  is m, then the ghost
number of  is m  1 . An important special case are states  with ghost number 0. If
the ghost number is 0, then:

U  0

Looking at the form of U, namely , we see that this implies that bk   0 . Furthermore,
a state annihilated by the ghost fields bk cannot be annihilated by the ghost fields c k .
Since u   c bi    i  bi c  n   bi c we see that:
i i i i

i i i
  
U    n   bi c i    n    bi c i 
 i  i

Hence if bk   0 , the c k cannot annihilate the state since this would result in a
contradiction. When bk   0 , it immediately follows that U   0 , but the above
result shows that if c   0 also, we would have U   n   0 , a contradiction.
k

The key result for states with zero ghost number is the following. Looking at Q in , it is
clear that if bk   0 then:

Q    ci Ki 
i

This implies that for a state with zero ghost number:

Ki   0

So, we can say that a state which is BRST invariant with zero ghost number is also
invariant under the symmetry described by the generators K i . Furthermore, if a state has
ghost number zero, this tells us that the state is not a ghost state, hence we avoid negative
probabilities.

BRST in String Theory-CFT

We will take a look at the BRST formalism in string theory by briefly considering two
approaches. The derivation of this approach is based on the use of path integrals, which
we are purposely avoiding due to the level of this text. So some results will simply be
stated, the reader who is interested in their derivation is encouraged to check the
references in the back of the book.

The application of BRST quantization to string theory can be done easily using conformal
field theory. The advantage of this approach is that the critical dimension D  26 arises
in a straightforward manner. We work in the conformal gauge where we take h   .
In this case the energy-momentum tensor has a holomorphic component Tzz  z  and an
anti-holomorphic component Tzz  z  where Tzz  z  was givn in (5.33) as:


Lm
Tzz  z   
m  z
m2

In Example 5-5 we worked out the OPE of Tzz  z  Tww  w  and found:
 D/2 2Tww  w   wTww  w 
Tzz  z  Tww  w    
 z  w  z  w zw
4 2

2
Where for simplicity of notation we have omitted the multiplying l s factor. The ghost
fields are introduced as functions of a complex variable z as follows. We define:

1 1
b  z  c  w  , b  z  c  w 
zw z w

Next we write down an energy-momentum tensor Tgh  z  for the ghost fields. This is
given by:

Tgh  z   2b  z   z c  z    z b  z  c  z 

The conformal dimension (6.4) of the ghost fields follows from this definition. Using the
ghost energy-momentum tensor together with the energy-momentum tensor for the string
we can arrive at the BRST current:

 1 
j  z   c  z   Tzz  z   Tgh  z    c  z  Tzz  z   c  z   z c  z  b  z 
 2 

The BRST charge is given by:

dz
 2 i j  z 
QÑ

Now, the central charge (that is, the critical space-time dimension ) comes from the
leading term in the OPE of the energy momentum tensor. Which is:

D/2
 z  w
4

The presence of this extra term is called the conformal anomaly since it prevents the
algebra from closing. So we would like to get rid of it. This is done by considering a total
energy-momentum tensor which is the sum of the string energy-momentum tensor and
the ghost energy-momentum tensor, i.e. T  Tzz  z   Tgh  z  . It can be shown that the
OPE of the ghost-energy momentum tensor is:

13 2Tgh  w   wTgh  w 

Tgh  z  Tgh  w    
 z  w  z  w zw
4 2
 D/2
Taking the leading term in this expression to be of the form  , we see that the
 z  w
4

ghost fields contribute a central charge of -26 which precisely cancels the conformal
anomaly that arises from the matter energy-momentum tensor. This result actually
follows from the nilpotency requirement (i.e. Q 2  0 ) of the BRST charge.

BRST Transformations
Next we look at BRST quantization by considering a set of BRST transformations which
are derived using a path integral approach. This is a bit beyond the level of the discussion
used in the book, so we simply state the results. Working in light-cone coordinates, we
define a ghost field c and an anti-ghost field b where c has components c  , c  and b has
components b and b . We also introduce an energy-momentum tensor for the ghost
gh
fields T with components given by:

Tgh  i  2b   c     b c  

Tgh  i  2b   c    b c  

The BRST transformations, using a small anti-commuting operator  are:

 X   i  c     c     X 

 c   i  c     c     c 

 b  i  T  Tgh 

The action for the ghost fields is:

S gh   d 2  b   c   b   c  

From which the following equations of motion follow:

 b    b    c     c   0

We can write down modal expansions of the ghost fields. These are given by:
 c    cn e  in    , c    cn e  in   
n n

b   bn e  in    , b   bn e  in   

n n

The modes satisfy the following anti-commutation relation:

 bm , cn    m n,0

with  bm , bn    cm , cn   0 . Virasoro operators are defined for the ghost fields using the
modes. Using normal ordered expansions, these are:

m    m  n  : bm  n c n :,
Lgh Lmgh    m  n  : bm  n c n :
n n

To write down the total Virasoro operator for the “real” fields + ghost fields, we form a
sum of the respective operators. That is:

m  Lm  Lm  a m ,0
Ltot gh

where the last term on the right is the normal-ordering constant for m  0 . It can be
shown that the commutation relation for the total Virasoro operator is of the form:

m , Ln    m  n  Lm  n  A  m   m  n ,0
 Ltot tot
 tot

Notice that the presence of the term A  m  on the right keeps us from obtaining a relation
that preserves the classical Virasoro algebra. As such, this term is called an anomaly. The
anomaly is determined in terms of two unknown constants which you might guess by
now are D and a. It has the form:

D 1
A  m  m  m 2  1   m  13m3   2am
12 6

To make the anomaly vanish, we take D  26, a  1 which is consistent with the other
results obtained so far in the book for bosonic string theory.

The BRST current is given by:

1 3
j  cT  : cT gh :   2 c
2 2

The BRST charge is given by the mode expansion:

 1
Q   cn L n    m  n  : cmcnb mn : c0
n 2 m,n

Using tedious algebra one can show that:

Q2 
1
2
 
 Ltot tot
 tot

m , Ln    m  n  Lm  n c m c n 
1
12
 D  26 

Hence the requirement that Q 2  0 forces us to take D  26 .

Going back to the original BRST approach outlined in (6.1)-(6.5), using the classical
algebra for the Virasoro operators:

 Lm , Ln    m  n  Lm n
We can identify the structure constants as:

f mnk   m  n   m  n , k

To see how the physical spectrum can be constructed in string theory, we consider the
open string case. The states are built up from the ghost vacuum state. Let’s call the ghost
vacuum state  . This state is annihilated by all positive ghost modes. Let n  0 , then:

bn   cn   0

The zero modes of the ghost fields are a special case. They can be used to build the
physical states of the theory. Using the anti-commutation relations (6.3), the zero modes
satisfy:

 b0 , c0   1

Using (6.3) it should also be obvious that b0  c0  0 . We also require that b0   0 for
2 2

physical states  . Now we can construct a two state system from the zero modes of the
ghost states. The basis states are denoted by  ,  . The ghost states act as:

b0   c0   0

b0    , c0   
We choose  as the ghost vacuum state. To get the total state of the system, we take the
tensor product of this state with the momentum state k to give , k . To generate a
physical state, we act on it with the BRST charge Q. It can be shown that:

Q , k   L0  1 c0 , k

The requirement that Q , k  0 gives the mass-shell condition L0  1  0 which

describes the same Tachyon state we found in chapter 4. Higher states can be generated.
We will have mode operators for each of the three fields: the X   ,  plus the two


ghost fields. To get the first excited state, we act with  1 , c1 , and b1 as follows:

     1  1c1   2b1  , k

The 1 and  2 are constants, but   is a vector with 26 components. In chapter 4, we

found that the first excited state was massless, so we expect the state to have physical
degrees of freedom in the transverse directions. That is, it should have 24 independent
components. To get rid of the extra parameters, we create a physical state with the
requirement that Q   0 . It can be shown that:

Q   2  p 2 c0   p   c1  1 p  1  , k

The requirement that Q   0 enforces constraints on the parameters. With this general
prescription, there are 26 positive norm states and 2 negative norm states.

We can eliminate the negative norm states by introducing some constraints. The first
constraint is to take p   0 and 1  0 . This rids the theory of the negative norm states.
Also note that p 2  0 , which tells us that this is a massless state. We also have two zero-
norm states:

k  1 , k and c1 , k

These states are orthogonal to the physical states. Eliminating them gives us a state with
24 degrees of freedom, as expected for a massless state in 26 space-time dimensions.

No-Ghost Theorem
The no-ghost theorem is simply a statement of the results we have seen in chapter 4 and
here, namely that if the number of space-time dimensions is given by D  26 , then
negative norm states are eliminated from the theory.
Chapter Summary
In this chapter we introduced the BRST formalism and illustrated how it can be use to
quantize strings. This is a more sophisticated approach than covariant quantization or
light-cone quantization. It takes a middle ground, preserving manifest Lorentz invariance
while living with ghost states. The approach makes the appearance of the critical D = 26
dimension simple to understand.
Chapter Quiz
1. Consider the Virasoro generators and calculate

  Li , L j  , Lk     L j , Lk  , Li     Lk , Li  , L j 
   

ANS: 0

2. Suppose that the BRST current j  z   c  z  Tzz  z   c  z   z c  z  b  z  is written as a

1
normal ordered expression. Find  Q, Q .
2

 3  D 2 3 
  2  cc  
Ñ   cc 
2
ANS:
12 3  

3. Looking at your answer to question 2, how many scalar fields does the theory contain
if we require that Q 2  0 ?

ANS: 26

4. Use L0 , k  0 to find k 2 .

1
ANS: k 
2

