Sie sind auf Seite 1von 7

QFT Notes

Sarthak Duarya
a
International Centre for Theoretical Sciences,Tata Institute of Fundamental Research,Shivakote,
Bengaluru 560089, India.

E-mail: sarthak.duary@icts.res.in
Contents

1 Lecture 1 1
1.1 Path integral approach in quantum mechanics 2

1 Lecture 1

Lessons from QFT 1 One of the main ingredients of QFT is Green’s function. Because
given the Green’s function or Correlation function we can extract most of the information of
the theory from there.
For example we take a Scalar field theory with a single scalar field φ. Then the Green’s function
is vacuum expectation value of the time-ordered product of operators.Green’s function is given
by-  
hΩ|T φ̂(x~1 , t1 ) · · · φ̂(x~n , tn ) |Ωi
G(x1 , · · · , xn ) = . (1.1)
hΩ|Ωi
 
T φ̂(x~1 , t1 ) · · · φ̂(x~n , tn ) is Time ordered product which means that whatever be the order
of operators inside this product we always make sure that the one with earliest time in the
argument is to the extreme right and one with latest or largest time in the argument is to
the extreme left.
Here, |Ωi is vacuum and φ̂(~x, t) is heisenberg picture field operator which is
φ̂(~x, t) = eiĤt φ̂(~x, 0)e−iĤt .
Here, c = 1, ~ = 1 unit is used.
Now, G(x~1 , x~2 ) → contains information about the mass because taking Fourier transformation
we can write it as a function of p G(p) e and then look for poles as a function of p, p2 = −m2
is location of pole then m is the mass of the particle.
G(x~1 , · · · , xn ) → scattering amplitude via LSZ reduction formula.This is the general formal-
ism.
Now, second aspect of QFT is to calculate G(x1 , · · · , xn ) using Perturbation theory.

Ĥ = Ĥ0 + λĤint (1.2)


Here, Ĥ0 is the free hamiltonian,λ is a small parameter,Ĥint is the interaction hamiltonian.
⇒ do perturbation expansion in λ.
⇒ Gives rise to Feynman rules.
Now, we use a different approach for computing G(x1 , · · · , xn ) using Path integral.
It leads to the same Feynman rules for simple theories.
Nevertheless, there are several advantages of the Path Integral approach.

–1–
Advantages of Path Integral formalism

1. It’s manifestly Lorentz invariant. Defining a hamiltonian means singling out time coor-
dinate as special because hamiltonian gives time evolution and hamiltonian by itself is
not Lorentz invariant because momentum and energy are treated differently. Of course,
the final Feynman rules are Lorentz invariant,but at the intermediate stage we certainly
loose Lorentz invariance because we give time coordinate as special treatment. In con-
trast, we’ll see that in Path integral approach throughout we can maintain Lorentz
covariance.

2. Hamiltonian approach or Canonical approach may be complicated if interaction terms


have time derivatives. Suppose we’ve a Lagrangian of the form-
Z Z h 1 i
L = d3 xL = d3 x − ∂µ φ∂ µ φ + λφ2 ∂µ φ∂ µ φ . (1.3)
2

Notation is ηµν = (−1, 1, 1, 1). Rewriting L


Z Z h 1 i
L = d xL = d3 x − ∂µ φ∂ µ φ + λφ2 ∂µ φ∂ µ φ
3
2
Z h1 i
= d3 x φ̇2 − λφ2 φ̇2 + · · ·
2
∂L π
Now,π = = φ̇ − 2λφ2 φ̇ ⇒ φ̇ = 1−2λφ
∂ φ̇ 2.
R 3 R 3 h 1 π2 i
Now H(π, φ) = d x[φ̇π−L] = d x 2 1−2λφ2 +· · · where · · · refers to space derivative
terms. Expanding λ we get infinite number of terms, to a given order in λ we’ve
finite number of terms,as we go higher and higher powers in λ we get more and more
interaction terms.

3. Much easier to implement for non-abelian gauge theories.

1.1 Path integral approach in quantum mechanics


p̂2
Ĥ = + V (q̂). (1.4)
2m
2p
Classical Hamiltonian is 2m + V (q).
1
Classical Lagrangian is 2 mq̇ 2 − V (q).
Action Principle
R
Equation of motion follow from extremizing the action S = dtL.
Imagine,tinitial = t0 , tfinal = t00 . qinitial = q 0 , qfinal = q 00 . Varying this path till S is extremised
which means that for classical path δS = 0 to first order for any variation of the path.
Define,
q̂(t) = eiĤt q̂(0)e−iĤt = eiĤt q̂e−iĤt
p̂(t) = eiĤt p̂(0)e−iĤt = eiĤt p̂e−iĤt .

–2–
Now,
q̂ |qi = q |qi , hq|pi0 = eiqp ,
p̂ |pi0 = p |pi0 ,0 hp|qi = e−iqp .
Define,
|q, ti ≡ eiĤt |qi .
Now,

q̂(t) |q, ti = eiĤt q̂e−iĤt eiĤt |qi

e−i
= eiĤt q̂ Ĥt eiĤt |qi
 

= eiĤt q̂ |qi
= q |q, ti .

Define, for t00 > t0

K(q 0 , t0 ; q 00 , t00 ) ≡ q 00 , t00 q 0 , t0



00 0
= q 00 e−iĤ(t −t ) q 0 .

Define
t00 − t0
∆= .
N
where N is large integer.

K(q 0 , t0 ; q 00 , t00 ) = q 00 e−iN ∆Ĥ q 0



= q 00 |e−i∆Ĥ e−i∆{zĤ
· · · e−i∆Ĥ} q 0

N-factors
R
Inserting I = dqi |qi i hqi |, where i = N − 1, N − 2, · · · 1.
Therefore,

K(q 0 , t0 ; q 00 , t00 ) = q 00 e−iN ∆Ĥ q 0



= q 00 |e−i∆Ĥ e−i∆{zĤ
· · · e−i∆Ĥ} q 0

N-factors
Z NY−1
= dq1 . . . dqN −1 hqi+1 |e−i∆Ĥ |qi i
i=0

where qN ≡ q 00 , q0 ≡ q 0 .

Z N
Y −1
0 0 00 00
K(q , t ; q , t ) = dq1 . . . dqN −1 hqi+1 |e−i∆Ĥ |qi i . (1.5)
i=0

–3–
Now we want to calculate hqi+1 |e−i∆Ĥ |qi i.
 
p̂ 2
−i∆ V (q̂)+ 2m
hqi+1 |e−i∆Ĥ |qi i = hqi+1 |e |qi i (1.6)
1
Using, BCH lemma eA eB = eA+B+ 2 [A,B]+··· ,we get
 
p̂ 2
−i∆ V (q̂)+ 2m p̂2 2 ×(··· )
hqi+1 |e |qi i = hqi+1 |e−i∆V (q̂) e−i∆ 2m e∆ |qi i
p̂2
= hqi+1 |e−i∆V (q̂) e−i∆ 2m |qi i
p̂2
= e−i∆V (qi+1 ) hqi+1 |e−i∆ 2m |qi i
Z
Inserting I = dp |pi0 0 hp|
Z
p2
=e −i∆V (qi+1 )
dpe−i∆ 2m hqi+1 |pi0 0 hp|qi i
| {z } | {z }
=eipqi+1 =e−ipqi
Z
p2
= e−i∆V (qi+1 ) dpe −i∆ 2m
eip(qi+1 −qi )

Completing the squares


h i2
i∆ m im
Z
− p− (qi+1 −qi ) + 2∆ (qi+1 −qi )2
= e−i∆V (qi+1 ) dpe 2m ∆

Performing gaussian integral


im 2
= Ae−i∆V (qi+1 )+ 2∆ (qi+1 −qi )

From eq.(1.5) we get,


Z N −1
Y im 2
K(q 0 , t0 ; q 00 , t00 ) = dq1 . . . dqN −1 Ae−i∆V (qi+1 )+ 2∆ (qi+1 −qi )
i=0
N −1 N −1
" #
∆m X  qi+1 − qi 2
Z X
= dq1 . . . dqN −1 AN exp − i∆ V (qi+1 ) + i
2 ∆
i=0 i=0
Z
= dq1 . . . dqN −1 AN eiS

R R 1 2 
Here, S = Ldt = 2 mq̇ − V (q) dt.
Let’s check this.
Z t00 N
X −1
− V (q)dt = −∆ V (qi+1 ).
t0 i=0

–4–
Z t00 N −1
1 2 1 X 2
mq̇ dt = m∆ q̇|t0 +i∆
t0 2 2
i=0
N −1 h
1 X qi+1 − qi i2
= m∆ .
2 ∆
i=0

Therefore, h i2
PN −1 PN −1 qi+1 −qi
−i∆ V (qi+1 )+i 2i m∆
eiS = e
i=0 i=0 ∆
.
Formally,
Z
00 00
K(q 0 , t0 ; q 00 , t00 ) = N
|{z} [Dq]qq0 ,t,t0 eiS . (1.7)
Normalization constant
| {z }
Integral over paths

Integral over Paths: First we break a path into many continuous segments. Path is a
collection of q− values at regular intervals. We represent S as a sum which is function of the
q−values at those segments. Once we discretize S, it depends on the q-values at these points
q 0 , q1 , q2 , · · · , q 00 . Integrate over path precisely means integrate over each of the qi ’s. As we
vary a path each of these qi ’s change.
Let’s consider

00 00  
q , t T q̂(t1 ) · · · q̂(tn ) q 0 , t0 .

(1.8)

Here, t00 > t1 , t2 , · · · , tn > t0 .


Let’s suppose the correct ordering is t̃n > t̃n−1 > · · · > t̃1 where {t̃1 , · · · , t̃n } : rearrangement
of {t1 , · · · , tn }.

00 00  
q , t T q̂(t1 ) · · · q̂(tn ) q 0 , t0 = q 00 , t00 q̂(t̃n )q̂(t̃n−1 ) · · · q̂(t̃1 ) q 0 , t0


(1.9)

Now,
q̂(t) = eiĤt q̂(0)e−iĤt = eiĤt q̂e−iĤt ,
0 0
q , t ≡ eiHt ˆ 0 0
00 00 00
q , q , t = q 00 e−iHt .

Using these relations eq.(1.9) can be written as

00 00  
q , t T q̂(t1 ) · · · q̂(tn ) q 0 , t0 = q 00 , t00 q̂(t̃n )q̂(t̃n−1 ) · · · q̂(t̃1 ) q 0 , t0

00 0
= q 00 e−iĤ(t −t̃n ) q̂e−iĤ(t̃n −t̃n−1 ) q̂ · · · e−iĤ(t̃2 −t̃1 ) q̂e−iĤ(t̃1 −t ) q 0 .

Suppose, t̃i − t0 = ki ∆, here ki are integer, ∆ to be sufficiently small. Now,


0
e−iĤ(t̃1 −t ) = e−iĤk1 ∆ = |e−iĤ∆ · · · e{z
−iĤ∆ −iĤ∆
e } (1.10)
k1 times

e−iĤ(t̃2 −t̃1 ) = e−iĤ(k2 −k1 )∆ = |e−iĤ∆ · · · e{z


−iĤ∆ −iĤ∆
e } (1.11)
k2 −k1 times

–5–
Using eq.(1.10) and eq.(1.11) we can simplify

00 −iĤ(t00 −t̃n ) −iĤ(t̃n −t̃n−1 ) 0
q e q̂e q̂ · · · e−iĤ(t̃2 −t̃1 ) q̂e−iĤ(t̃1 −t ) q 0
00 0
Now, hq 00 |e−iĤ(t −t̃n ) q̂e−iĤ(t̃n −t̃n−1 ) q̂ · · · e−iĤ(t̃2 −t̃1 ) q̂e−iĤ(t̃1 −t ) |q 0 i
= hq 00 |· · · q̂ |qk2 i hqk2 | · · · e−iĤ∆ q̂ |qk1 i hqk1 | e−iĤ∆ · · · |q2 i hq2 | e−iĤ∆ |k1 i hq1 | e−iĤ∆ |q 0 i
| {z } | {z }
qk2 |qk2 i =qk1 |qk1 i
= dq1 · · · dqN −1 eiS qkn qkn−1 · · · qk1
R

= [Dq]eiS q(t1 ) · · · q(tn ).


R

Here, qk = q(t̃k ), t0 + k∆ = t̃k .


Path integral always automatically computes matrix element of time ordered product given
by
Z

00 00  
0 0
q , t T q̂(t1 ) · · · q̂(tn ) q , t = [Dq]eiS q(t1 ) · · · q(tn ).

(1.12)

–6–

Das könnte Ihnen auch gefallen