You are on page 1of 8

Quantization of Nonabelian Gauge Fields and

BRST Formalism
Volker Bach
FB Mathematik; Johannes Gutenberg-Universitat;
D-55099 Mainz; Germany;
email: vbach@mathematik.uni-mainz.de
May 24, 2002
Abstract
These lecture notes are a summary of my seminar on Quantization of
Nonabelian Gauge Fields and BRST Formalism given in the Hesselberg
workshop, 24.2.1.3.02.
It is a pleasure for me to thank Rydzard Nest, Florian Scheck, and Elmar
Vogt for organizing this excellent workshop and the Hesselberg team for
providing a very pleasant environment.

Quantization + BRST, May 24, 2002

Contents
I

Lagrangians and Gauge Transformations

II Quantization and Gauge Fixing

III BRS(T) Symmetry and BRS(T) Quantization

Quantization + BRST, May 24, 2002

Lagrangians and Gauge Transformations

We start by introducing some notation. As customary in theoretical physics, we


shall always work in local coordinates using Einsteins summation convention.
Likewise, we shall not distinguish between a group or algebra and their representations. Another disclaimer concerns the mathematical contents of what is written
below. In case of doubt, all computations are formal and do not meet the usual
standard of mathematical rigor. In particular, the path-integral measure on the
space of connections is known not to exist. This does not, however, affect the
conceptual contents of these notes. Readers still feeling uneasy with this may
want to replace Minkowski space
by a finite subset of  . While this means
to sacrifice Lorentz invariance, the benefit is that there are no analytical problems
anymore.
Let be a compact semi-simple Lie group (e.g.
  or   )
and  the corresponding Lie algebra of dimension  generated by Hermitian
 satisfying the commutation relations    
matrices     

  with structure constants  .


, the basic object for constructing gauge
Denoting Minkowski space as
theories is the principal bundle   . Given a map (actually: one-form)

 , called the gauge potential, a connection on    is
defined by       . The gauge potential represents a (classical, boson) gauge field which couples through the connection  to (classical,
   of mass , which are in turn sections
fermion) matter field(s) 
on   .
A gauge transformation 
    acts on  and  as




    
 

     

(I.1)

The Langrangian density of a (boson) gauge field and a (fermion) matter


field of mass , which are minimally coupled to each other, is defined by

  


  

   
 

(I.2)

where the field strength is defined by

   


  
      
(I.3)
is abelian, then  is linear in  and   in  is quadratic in

Note that if
 . The quantization of quadratic theories is well-understood, as the corresponding field energy operator can be written as the  quantization of a one-particle
operator.
In contrast, if is nonabelian, then  is quadratic in  and   in 
is quartic in  . The corresponding quantized field is genuinely self-interacting
and its quantization is difficult.

Quantization + BRST, May 24, 2002

II Quantization and Gauge Fixing


The idea of quantization is to start with functions    of classical observables
and obtain the vacuum expectation values of the corresponding quantized
observable    (which is an operator). In particular, we can construct the vacuum expectation values of creation and annihilation operators this way. From
these, in turn, we obtain the Hilbert space of the quantum theory by using a GNS
construction.
We now quantize a pure Yang-Mills theory (without fermion fields), i.e.,

   
by using a path integral quantization. Setting   

 

(II.1)

 , the vacuum

expectation value of a gauge-invariant observable    is defined by

where



           


(II.2)

  is the space of connections,

 
 

  

(II.3)

 is a normalization factor.
is a path integral measure, and 
 do not exist, and Formula (II.2) is
Unfortunately, the measure (II.3) and 
formal in various senses. One obvious source of divergence is the redundancy
    related by a
in (II.2) coming from connections and 
gauge transformation  . This observation leads us to consider the gauge orbits

         
(II.4)
for  . Since both  and  are gauge invariant, we have      ,
     , and


 

   

 

 



     
 

(II.5)

Unfortunately, the quotient space  has a complicated topological structure,


and the formula (II.5) is not useful for concrete computations.
A different solution to this problem lies is passing to a lattice gauge theory
by replacing
by a finite lattice     
 . The integration variables in the analogue of (II.5) are then the group elements in , for each bond

Quantization + BRST, May 24, 2002

linking two neighboring sites in   . Thus, there are only finitely many integrations to perfom, and, moreover, each integration is over the Haar measure of the
compact Lie group . The difficulties arisising from (II.5) are shifted to making
sense of the limit of the integral in the limit    . Furthermore, Lorenz
invariance is lost, and the lattice imposes an artificial anisotropy.
Yet, a third approach is furnished by gauge fixing. We assume to be given a
function 
and a connection 
such that the condition    
defines a bijection between  and the level set

    .
In other words, for each  
 we require that there is a unique representant
 
,    
  such that       . The construction of  and 
is difficult and leads to the so-called Gribov problem. Locally, the bijectivity
requirement amounts to proving that the gauge orbits intersect
transversally,
i.e., that the Faddeev-Popov matrix


 

   





  

          

(II.6)



be invertible. The determinant of  is then the appropriate Jacobian for the integral

  

            

(II.7)

Here, is a function that localizes    about the constraint  . As opposed to


, the normalization factor  may turn out to be finite. (If is replaced by  ,
this contradistinction will be literally correct.) This way, we have eliminated the
gauge copies of each connection.
An (important) example is provided by the Lorentz gauge,

          

 

which yields

  











      




(II.8)

    
   
  
 

 

 










   
            


(II.9)
using an (bosonic) auxiliary field  (Nakanishi-Lautrup field). Furthermore, introducing (Lie algebra-valued) Grassmann fields, 
,  , called Faddeev-Popov

Quantization + BRST, May 24, 2002

ghosts, we may rewrite the Jacobian of the Faddeev-Popov matrix as

 






  


 

(II.10)

and inserting this expression into (II.9), we arrive at




  

where



   

   

   
 
  


 
   
 

   
  

    


   
 

and





 




 

   
    

(II.11)

(II.12)
(II.13)
(II.14)

    


(II.15)
    
 
 

(II.16)

III BRS(T) Symmetry and BRS(T) Quantization


We begin with the observation that the Faddeev-Popov ghosts are spin-0 fermion
fields which cause a violation of (the assertion of) the spin-statistics theorem. That
is, if we construct a Hilbert space  carrying the quantized fields ,  , 
, and 
from the vacuum expectation values (II.11) (essentially, by using a GNS construction), this Hilbert space  will have an indefinite scalar product. This reflects the
fact that the ghosts fields 
, and  are of no physical relevance and have been
introduced merely for mathematical convenience. One elegant way to eliminate
the ghost fields is the use of the BRST symmetry , which acts on (classical)
functionals, and the corresponding (quantum) operator, the BRST charge .
The BRST symmetry is an (odd) derivation on functions in ,  , 
, and  ,
i.e., it obeys the graded Leibniz rule          . Defining
it by

  
   


 

 

  

(III.1)
(III.2)

Quantization + BRST, May 24, 2002

one easily checks the nilpotency of ,


(III.3)
It implements gauge symmetry, i.e., 
is invariant under , but it additionally
involves the ghost and auxiliary fields and actually leaves the ghost Lagrangian
 invariant, as well.


 

  
(III.4)
The BRST charge operator is the generator of on the level of operators,



  
(III.5)
 , and   denotes the graded
where  is any of the quantized fields ,  , 
,  on 

commutator. Explicitly,


Since

 


  

, we also have



    
  
  

  



(III.6)

(III.7)

on  . We remark that, since


is symmetric, ! 
!  , Eq. (III.7)

would imply that
, if the scalar product on was positive definite.
We now define the algebras of BRST-closed operators and of BRSTexact operators on  by

 "    "    


(III.8)
 "  # "   #  
(III.9)
. Thus, we may define
and observe that    is a two-sided ideal, since
to be the quotient space
the gauge invariant observables 
  "   " 
(III.10)
 by
Analogously, we define the subspaces    

      
(III.11)
   !  !  
(III.12)
. We then define the physical Hilbert
and observe that   , due to
space 
to be the quotient space
      
(III.13)













The important facts that justify these definitions are the following:

Quantization + BRST, May 24, 2002


(b) 
(a)

(c)



is positive definite,



contains no ghost nor antighost component,

$    is left invariant under the Hamiltonian (the generator of

time translations),
(d) For any 

  and any operator " , one has




"  

 

"





(III.14)

Properties (a) and (b) insure that  is a Hilbert space appropriate for quantum
mechanics: it is positive definite and carries no unphysical (ghost) degrees of
freedom. Moreover, Properties (c) and (d) insure that an initial state chosen in
 evolves in this subspace under the physical dynamics. Thus, all transition
amplitudes are allowed to be computed with the aid of ghost fields, if this appears
to be convenient, and at the end of the day, one merely has to project all vectors
onto  .

You might also like