Sie sind auf Seite 1von 24

Perturbative String: Internal structure of String-like Object(s)

Trung Van Phan

Physics Department, Massachusetts Institute of Technology,


77 Massachusetts Avenue, Cambridge MA02139
Abstract
Point-like object to Quantum Mechanics Theory is similar to String-like object to String The-
ory. Indeed, we start generalize our Physics Theory out from no internal structure object(s) to
simple/complex internal structure one(s).
1
I. ACTION OF A CLASSICAL STRING
Let
a
( R
1,1
) is the parametrization of the world-sheet - the area caused by the
string motion in the target space-time R
1,d1
(one can write the position of an element of
the string as X = X()). As a matter of convenient, since the string is a spatial-limited
object, choose [0, l]. The Nambu-Goto action for classical string, with G = det (G
ab
) =
det (g

a
X

b
X

) (G
ab
is called the induce metric), is described by:
S
NG
=
1
2

dA =
1
2

d
2

G (1)
The prefactor
1
2

is the constant assign weight for each unit of the world-sheet and the
sign comes from the chosen metric signature (time: +, space: ). To simplify the
action by taking the dynamical G
ab
out of the square root, one can introduce an auxiliary
eld (world-sheet metric) h
ab
= h
ab
() and dene the Polyakov action, with h = det (h
ab
):
S
P
=
1
4

d
2

hh
ab
G
ab
=
1
4

d
2

hh
ab
g

a
X

b
X

=
1
4

d
2

h(X)
2
(2)
Since the theory of interest is a local one as the equation of motion should not receive any
non-local boundary contribution, we must suppress all the boundary term of the dynamical
elds. A string, is an object with spatial limitation, gives the non-local contribution to the
2
variation of S
p
at its spatial end when I change the dynamical elds by X:
S
p
=
1
2

_
d
0

hX

1
X

1
=l

1
=0
(3)
To make this contribution goes away, we need the string to be a close string or an open
string with
1
X

= 0 (free points), or in general n


a

a
X

at the end points with n


a
.
The world-sheet stress-energy tensor is dened as T
ab
= 4
S

hh
ab
(S is a general world-
sheet action that have the same local symmetry of interest with S
P
). With Polyakov action,
one gets S when varying h
ab
(

h =
1
2

h
h =
1
2

hh
ab
h
ab
):
(
_

d
2
)
1
S
P
=
1
4

hh
ab
G
ab
+

hh
ab
G
ab
) =
1
4

h(
1
2
h
ab
h
cd
G
cd
G
ab
)h
ab
T
ab
=
1

(
1
2
h
ab
h
cd
G
cd
G
ab
) =
1

(
1
2
h
ab
h
cd

c
X

d
X

a
X

b
X

) (4)
The classical equation of motion gives T
ab
= 0 and from that one can integrate out the
auxiliary eld h:
G
ab
=
1
2
h
ab
h
cd
G
cd
G =
1
4
(h
cd
G
cd
)
2
h

h =
2

G
h
cd
G
cd
(5)
Plug that in the Polyakov action, we get back to the Nambu-Goto action!
The above action is similar to d (the number of space-time dimensions) matter elds
couple to 2-dimension gravity. In detail, one can think of it as a theory of General Relativity
in (1 + 1) (spatial-limited from 0 to l) dimension (correspond to ) with gravity couple to
d scalar eld (correspond to X, note that the word scalar here means X

is transformed
as a scalar under any transformation; the internal transformation of X is represented by
a matrix that correspond to g

).
The Polyakov action has 1 global symmetry (space-time Poincare symmetry) and 2 local
symmetry (world-sheet dieomorphism symmetry and conformal symmetry). We can also
nd the correspond conserved current for each of these symmetries:
1/ Poincare global transformation acts on the dynamical elds X and leave g

...X

...X

unchange. We get the conserved momentum for shifting X:


P
a

a
X
L =
1
2

hh
ab
g

b
X

(6)
and conserved angular momentum for Lorentzian rotating X around axis n:
J
a
n
= P(n X) = n(XP)
J
a

= X

P
a

P
a

=
1
2

hh
ab
(X

b
X

b
X

) (7)
3
They just come out straight from classical Physics! The conservation laws for these
currents indicate that
a
P
a

=
a
J
a

= 0
2/ Dieomorphism local transformation reparametrizes as the volume element d
2

h
as well as h
ab

a
...
b
... stay the same. Its interesting to see the small variation in innitestimal
language:

a
=
a
(); X =
c

c
X(); h
ab
=
c

c
h
ab
+ (
a

c
)h
cb
+ (
b

c
)h
ca
=
a

b
+
b

a
(8)

h =
1
2

hh
ab
h
ab
=
1
2

hh
ab
(
a

b
+
b

a
) =

h
a

a
(9)
The innitesimal change in any general S with the dieomorphism local symmetry can
be written as (I will need S = 0 and use that
a

h = 0):
S =
1
4
_

d
2

hT
ab
h
a
b =
1
2
_

d
2

hT
ab

b
=
1
2
_

d
2

a
(

hT
ab
)
b
=
1
2
_

d
2

h
a
T
ab

b
= 0 (10)
I use integration by part and kill the boundary term (which = 0), then nally arrive
at
a
T
ab
= 0 with the world-sheet stress-energy tensor as a conserved current. One can
also show that directly for the Polyakov world-sheet stress energy tensor with brute-force
calculation. Start from the equation of motion for eld X:
(
X

a

a
X
)L =
1
2

a
(

hh
ab

b
X

) =
1
2

h
2
X

= 0 (11)
So we get
2
X

=
a

a
X

=
a

a
X

= 0. Use this:

a
T
ab
=
1

(
1
2

b
(h
cd

c
X

d
X

) (
a

a
X

)
b
X

a
X

(
a

b
X

))
=
1

(
1
2

b
(h
cd

c
X

d
X

)
a
X

b
X

)
=
1

(
1
2

b
h
cd

c
X

d
X

+ h
cd

c
X

d
X

a
X

h
ac

b
X

)
=
1

(
1
2

b
h
cd

c
X

d
X

+ h
ac

d
cb

a
X

d
X

) = 0 (12)
In the last step, I use the symmetry argument to kill the whole thing (
a
bc
=
1
2
h
ad
(
b
h
cd
+

c
h
db

d
h
bc
))
3/ Weyl local transformation change the auxiliary world-sheet metric eld h
ab
e
2
h
ab
( = ()) and it makes

hh
ab
[(e
2
)
d
W
2

h][(e
2
)
1
h
ab
] =

hh
ab
invariant (d
W
is the world-sheet dimension, which is equal to 2). One can derive the trace of the world-
sheet stress-energy tensor T
a
a
must be vanished as a consequence of general local conformal
4
symmetry (the innitestimal variations are h
ab
= ()h
ab
and

h = ()

h since
h
ab
h
ab
= d
W
= 2). Use S for a general local Weyl symmetric action:
S =
1
4
_

d
2

hT
ab
h
a
b =
1
4
_

d
2

hT
ab
h
ab

=
1
4
_

d
2

hT
a
a
= 0 T
a
a
= 0 (13)
One can verify this for the Polyakov world-sheet stress energy tensor:
h
ab
T
ab
=
1

(
1
2
h
ab
h
ab
h
cd
G
cd
h
ab
G
ab
) =
1

(
d
W
2
1)h
ab
G
ab
= 0 (14)
Combining dieomorphism and Weyl symmetry, one arrives at conformal symmetry.
Theres a very interesting mistake that people might make when they study about the
theory: since the theory have a denite length scale (to be more precise, length square scale
for

), how can the theory be conformal invariance? The answer is straight-forward: this
denite scale is for the target space, and the conformal symmetry in the theory is for the
world-sheet.
Before I move on to the next section, consider a 2-dimensional Einstein-Hilbert action
that also preserves these 3 symmetries (R is the world-sheet Ricci scalar):
S
2
=

2
4
_

d
2

hR =
2
(()
1
2
_

dSK) (15)
Mathematically, the action S
2
only depend on the topological properties of the world-
sheet , with () is the Euler-characteristic of and K = t
a
n
b

a
t
b
is the extrinsic
5
curvature of with t n and the sign was depend on the type of boundary (timelike: +,
spacelike: ). 2-dimension gravity is dynamically trivial as S
2
gives no dynamics: indeed,
in 2-dimension, I can write the curvature as R
abcd
= r
ab

cd
(purely symmetrical argument:
R
abcd
= R
cdab
= R
bacd
= R
abdc
with a, b, c, d = 0, 1, and I only have one parameter r())
and get R = 2r. This means R
abcd
=
1
2
R
ab

cd
trivially and so does R
ab
= h
cd
R
abcd
=
1
2
Rh
ab
- the equation of motion for Einstein-Hilbert action (vacuum T
ab
= 0) in 2-dimension is
satised naturally. So h
ab
is non-dynamical and even can be set to h
ab
=
ab
(at gauge
choice) locally - completely x the metric by gauging away all parameters (symmetric h
ab
has
1
2
d
W
(d
W
+1) = 3 degrees of freedom, and by gauging out d
W
= 2 dieomorphism parameters
and 1 conformal parameter, we left with none degree of freedom that cannot by gauge
away!). Actually, look at the innitestimal transformation (under both dieomorphism
a
,
and conformal transformation):
(
_

d
2
)
1
S
2
=

h2
2
=
a
(2

h
a
) (16)
One can see that S
2
is invariant if has no boundary or the term under total derivative
in the above expression at boundary S disappear. The reason why I care about possible
addition term(s) for the Lagrangian is because the theory of quantum (modern Physics) is
symmetry-base. The degree of freedom in our accessible energy scale is the result of the
underlying elds and the symmetry between them. By looking at the symmetry between
the observable degree of freedom, one can make a guess for the symmetry at deeper level,
and therefore possible interactions between degrees of freedom that also obey the symmetry
may enter the picture.
II. CLASSICAL SOLUTION
After xing the metric, therere still (some) residual gauge(s) freedom (thats generated
by conformal Killing vectors). To see that, lets see how the metric transform innitestimally
by dieomorphism and conformal:
h
ab
=
a

b
+
b

a
+ h
ab
= (
a

b
+
b

a
h
ab

c
) + ( +
c

c
)h
ab
(17)
That means the innitestimal transformation vector
a
must satisfy
a

b
+
b

a
= h
ab

c
(condition for dieomorphism generators) if the new metric after the variation is Weylly the
6
same as the old one. Weylly the same means one can go from the new metric back to
the old one by performing a Weyl transformation. These vector (that generate the Weylly
family of metric) are called the conformal Killing vector. Start from h
ab
=
_
_
1 0
0 1
_
_
(
is a prefactor), with the change along the conformal Killing vector, one can use the Weyl
generator to get that metric back. Use
0
= ,
1
= for at metric and with the constrains
that the (Weylly the same with at metric) metric diagonal must be opposite and the other
2 terms must be vanished, one arrives at

= 0. One can rewrite


it as (

)(

) = (

)(

) = 0 and change to the light-cone coordinate

= to end up with a nice and short

= 0

). With the conformal


Killing vectors, the families of correspond conserved currents J
a

= T
ab

b
with
a
J
a

= 0.
The proof is straight-forward:

a
J
a

= T
ab

b
=
1
2
T
ab
(
a

b
+
b

a
) =
1
2
T
ab
h
ab

c
=
1
2
T
a
a

c
= 0 (18)
Dene the associate conserved charge of the string for the above conserved current (time
independent):
L

=
l
2
4
_
l
0
dT

)
+
(

) =
l
2
4
_
l
0
dT

()
+
() (19)
Lets call it the Virasoro charge. This charge is very important in 2-dimension conformal
eld theory in general.
Now, after choosing a nice gauge to work with, one can write the Polyakov action as
follow (in the world-sheet at gauge and in the world-sheet light-cone gauge):
S
P
=
1
4

dd((

X)
2
(

X)
2
) =
1
4

d
2

+
X

(20)
The equation of motion for the string classically obey
2
X

= 0 for at gauge and

= 0 for light-cone gauge. After xing the gauge, the energy-momentum equation
of motion T
ab
= 0 will just become a constrain. For at gauge: T

= T

=

XX

= 0 =
T

= T

=
1
2
(

X
2
+ X
2
) and for light-cone gauge: T
++
= (
+
X)
2
= T

= (

X)
2
= 0 =
T
+
= T
+
. Note that the world sheet metric in light-cone gauge is
_
h
++
h
+
h
+
h

_
=
1
2
(
0 1
1 0
).
The fastest way to nd the solution for the motion of the string is look at the world-
sheet light-cone gauge and nd that one can split X into left (holomorphic) and right
(antiholomorphic) mover: X() = X
L
(
+
) + X
L
(

). Let the string at spatial parameter


7
run from 0 to l (l is the string length, and we want to make it a constant). After brute-
force calculations, we get the mode expansion for (x and p are the center of mass position
and momentum):
1/ Close string solution:
X

L
=
1
2
x

+
1
2
(
2

l
)p

+
+ i
_

n=0
1
n

n
e

2
l
in
+
(21)
X

R
=
1
2
x

+
1
2
(
2

l
)p

+ i
_

n=0
1
n

n
e

2
l
in

(22)
with the derivatives with

0
=

0
=
_

2
p

+
X

L
=
2
l
_

m=

m
e

2
l
im
+
;

L
=
2
l
_

m=

m
e

2
l
im

(23)
Note that for x and p inside X, I actually use x = x(
+
= 0) and p = p(
+
= 0). For
other X solution from now on (both scalar and operator form), Ill stick with that subtle
remind. Reality implies (

n
)

n
and (

n
)

n
.
Classical quantities associate with the closed string are:
- Halmintonian:
H =
1
2

_
l
0
d(

X

L) =

l

m=
(
m

m
+
m

m
) (24)
- Angular momentum (the orbital angular momentum L and the intrinsic angular mo-
mentum E +

E - one can think of it as spin):
J

=
_
l
0
d(X

) = (x

)
i

m=
1
m
(

m
) i

m=
1
m
(

m
) (25)
= L

+ E

+

E

(26)
- Energy-stress tensor (must be vanished):
T
++
= (
2

l
)

k=

m=

km

m
e

2
l
im
+
= 0 (27)
T

= (
2

l
)

k=

m=

km

m
e

2
l
im

= 0 (28)
8
- Families of Virasoro charge (write

with the basis

(k)
= e
2
l
ik

- each k gives us a
distinct charge):

L
k
=
l
2
4
_
l
0
dT
++
()
+
(k)
() =
1
2

m=

km

m
=

L

k
(29)
L
k
=
l
2
4
_
l
0
dT

()

(k)
() =
1
2

m=

km

m
= L
k

(30)
Note that one can write the Halmintonian as H =
2
l
(L
0
+

L
0
). So the information about
the Halmintion of a closed string is included in the families of conserved Virasoro charge.
One can also write the energy-stress tensor as:
T
++
= 2(
2

l
)

k=

L
k
e

2
l
im
+
; T

= 2(
2

l
)

k=
L
k
e

2
l
im
+
(31)
So classically, all Virasoro charge vanish as the energy-stress tensor equal to 0. Consider
the mass-shell condition p
2
+M
2
=
1

(
2
0
+
2
0
) +M
2
= 0 and with H = 0, one get the mass
of the closed string depend on the amplitude of the oscillation modes as follow:
M
2
=
1

n=0
(
n

n
+
n

n
) =
2

n=1
(
n

n
+
n

n
) (32)
Theres a subtle here: actually M =
4

n=1

n
=
4

n=1

n

n
), but I want to
represent it in a symmetric form, so that I have the above expression instead.
2/ Open string solution
X

= x

+ (
2

l
)p

+ i

n=0
1
n

n
e

l
in
cos (
m
l
) (33)
and its derivatives with

0
=

:
2

m=

m
e

l
im

(34)
Reality implies (

n
)

n
.
Classical quantities associate with the open string are:
- Halmintonian:
H =
1
2

_
l
0
d(

X

L) =

2l

m=

m
(35)
9
- Angular momentum (the orbital angular momentum L and the intrinsic angular mo-
mentum E):
J

=
_
l
0
d(X

) (36)
= (x

) i

m=
1
m
(

m
) = L

+ E

(37)
- Families of Virasoro charge (write

with the basis

(k)
= e
2
l
ik

- each k gives us a
distinct charge). You dene this charge it dierently from the closed string case:
L
k
=
l
2
4
_
l
0
d(T
++
()
+
(k)
() + T

()

(k)
()) =
1
2

m=

km

m
= L
k

(38)
So classically, all Virasoro charge vanish as the energy-stress tensor equal to 0. Note that
one can write the Halmintonian as H =

l
L
0
. So the information about the Halmintion of an
open string is included in the families of conserved Virasoro charge. Since the stress-energy
tensor vanish, H also must be gone. Use the mass-shell condition (with the same trick as in
the case of the closed string), one gets the mass of the open string:
M
2
=
1

n=1

n
(39)
III. OPERATOR-BASED QUANTIZATION
A. Old Covariance Quantization
Now lets try to quantize the theory with the same method as in Quantum Mechanics rst
(for convenient, note that the Dirac delta function can be written in the Fourier expansion
modes as () =
1

m=
e
2im
)! In Quantum Mechanics of point-like object, we assign
the canonical equation [X

(), P

)] = ig

), so in String Theory, we also assign


that relation for every element of the string (P

=
S
p
X

), and nally arrive at the center


of mass operators [x

, p

] = ig

and [

m
,

n
] = [

m
,

n
] = mg

m+n,0
. Using raising-
lowering operator notation (m, n > 0) [a
i
m
, a
j
n

] = g
ij

i,j
and [a
0
m
, a
0
n

] =
m,n
(sign is
ipped, compare to normal relation), one gets

ma

m
=

m
and

ma

m
. The
way to construct states from raising-lowering operators is simple - just like quantum simple
harmonic oscillation state, with the start from the vacuum state |0 (since I want to dress
10
the oscillation over the center of mass motion, the vacuum state is actually the oscillating
vacuum state with some give momentum, so I should represent that state with more detail:
|0; p). So the general form of a particular 1-particle state is (instead of working with the
normal raising-lowering operators, in String Theory I prefer to work with

m
, with the
oscillation mode raising

m
and lowering

m
as m > 0):
|
1
=

1
...
k
k

j=1

j
m
j
|0; p (40)
For 1-particle state, (
j
, m
j
) in general doesnt need to be distinct. One always need to
include the polarization tensor

1
...
k
to indicate the polarize structure of the particle. So
now, after I know how to construct a state, how can I know if its a physical meaningful state
or not? The question here is due to the fact that we I quantize the theory in this world-sheet
light-cone gauge I get some constrain for the conserved Virasoro charge, and I havent use
that information yet. Now its the time to guess and write down the operator representation
for Virasoro charges by asking what we expect to see after acting these operator on the
physical state. Consider the treatment for open string with only one Virasoro charge (the
case of a closed string is similar with one more copy of this charge family). Classically for all
m one gets L
m
= 0 so in quantum version one expects to impose the constrain | L
m
| = 0
for physical state |. Since classically L
m
= (L
m
)

, so for m > 0 (for m = 0 we should treat


it dierently, since unlike L
m
of positive m that are complex, L
0
must be real) by require
that the operator analog must be L
m
= L

m
, the quantum constrain for the expectation of
Virasoro operator is satised when L
m
| = 0 for all m positive. For m = 0, there comes
the problem: in the scenario of raising-lowering operators as building block of the theory,
I always want to organize operators in the right order (just a convention with no actual
deep meaning), so let dene each operator with its right order (right order means raising to
the left and lowering to the right). For m > 0 the order is right automatically, since one
can move the operators pass through each other with ease since the m-index of the raising-
lowering operators are distinct. For m = 0, theres an ambiguity for it since we dont know
how L
0
actually organize its operators, so one can change/shift the operator L
0
to L
0
A,
with the new 0-index Virasoro operator is normal ordering and A is pop-out naturally. So, in
mathematical language, (: O : is the raising-lowering operator right ordering of the original
operator O) L
m
=: L
m
: for m > 0 and redene L
0
: L
0
: then make (L
0
A) | = 0
instead of L
0
| (originally, L
0
=: L
0
: A). Lets take a look at the structure of Virasoro
11
operators (L
n
=
1
2

m=
:
nm

m
:)- the Virasoro algebra:
[

k
, L
n
] =
1
2

m=
[

k
, :
nm

m
:] =
1
2

m=
(k
k+m,0

nm
+ k
m,0

n+km
) = n

n+k
(41)
With that elementary commutation, one can perform the brute-force calculations for
Virasoro operators commutation rule (during the process, the space-time dimension d pop-
out from the commutator [

m
,

nm
] = md
n,0
):
[L
m
, L
n
] = (mn)L
m+n
+
d
12
m(m
2
1)
m+n,0
(42)
This is the Virasoro commutation relation for the conformal eld theory with action
S
P
. In general conformal eld theory, its dierent. The theory now have physical states
(with Virasoro L
m
constrain) and fundamental building block operator X (no quadratic
or higher terms in the classical version, so we dont need to worry about ordering problem
when we change and quantize it to operator) for the Hamiltonian and all other dynamical
operators that inuence the Physics of the theory! The next things I want to play with is
the physical constrain for the physical states: their norm must be non-negative.
A spurious state in Mathematical description is a state that satisfy the mass-shell con-
dition (L
0
A) | and orthogonal to all physical states (if its also satisfy L
m>0
| = 0
then it is a spurious physical state - so its null, since its orthogonal with itself). Sometime
one can dene an eective spurious state without the mass-shell condition, and we get the
form of that state |

= L
n
| with arbitrary state |. Let try to write down a general
spurious state | =

n>0
L
n
|
n
with some basis |
n
, one can see that automatically it
is orthogonal to all physical state |): | L
n
|
n
= (| L

m
) |
n
= 0. Now, lets look at
the spurious basis in more detail (use the commutation relation for Virasoro operators to
get [L
0
, L
n
] = nL
n
):
(L
0
A) | =

n>0
L
0
L
n
|
n
A| =

n>0
([L
0
, L
n
] + L
n
L
0
) |
n
A

n>0
L
n
|
n
= 0
=

n>0
L
n
(L
0
A + n) |
n
= 0 (L
0
A + n) |
n
= 0 (43)
A worth-to-mention property of the Virasoro operators that we use in the above equation
is L
n
(n > 0) will rise the eigenvalue of an eigenstate of L
0
by n. Consider a state |k with
L
0
|k = k |k, so L
0
(L
n
|k) = ([L
0
, L
n
] + L
n
L
0
) |k = (n + k)(L
n
|k).
Since L
n
for n > 2 can be represented as iterated commutators of L
1
and L
2
, so any
spurious state can be written as | = L
1
|
1
+ L
2
|
2
with (L
0
a + 1) |
1
= 0 and
12
(L
0
a + 2) |
2
= 0. In this theory, one can get negative norm states because of time-like
oscillation mode (for example: |a
0
m

|0 |
2
= 0| a
0
m
a
0
m

|0 = 0| [a
0
m
, a
0
m

] |0 = 0| 0 < 0),
since the signature of time is opposite with space, and I dont want to get these states in
the physical picture.
Lets start with the Mathematical derivation for how can I nd the value for A and
quantize the number of dimensions of space-time so that I can rule all the negative state out
of the theory. At the boundary of physical positive and physical negative state (with A and
d as the parameters, for dierent algebraic structure of L), there must be a boundary which
is made of null physical states. Look at the boundary through the 1st channel physical
spurious state | = L
1
|
1
. For m = 0 we get exactly what we expect from from the
mass-shell condition (L
0
+ 1 A) |
1
= 0 and for m = 1 one arrives at:
L
1
| = L
1
L
1
|
1
= ([L
1
, L
1
] + L
1
L
1
) |
1

= 2L
0
|
1
= 2(A 1) |
1
= 0 A = 1 (44)
The only reason why I choose to look at that special form of physical spurious state is
simply because its the fastest way to x A. For m > 1, the condition L
m>0
| = 0 will only
put further constrains for us the get | exactly physical. One can, for example, simply use
the condition L
m>0
|
1
= 0 (we have the freedom to choose |
1
):
L
m
| = L
m
L
1
|
1
= ([L
m
, L
1
] + L
1
L
m
) |
1

= ((m + 1)L
m1
+ L
1
L
m
) |
1
= 0 L
m>0
|
1
= 0 (45)
Now lets take a look at the 2nd channel, the fastest way to get d is by considering a
13
state of the form | = (L
2
+
3
2
L
1
L
1
) |
2
. Then by and by choosing L
m>0
|
2
= 0 the
term L
1
| vanish (mass-shell condition at A = 1 is (L
0
+ 1) |
2
= 0):
L
1
| = L
1
(L
2
+
3
2
L
1
L
1
) |
2

= ([L
1
, (L
2
+
3
2
L
1
L
1
)] + (L
2
+
3
2
L
1
L
1
)L
1
) |
2

= [L
1
, (L
2
+
3
2
L
1
L
1
)] |
2
L
m>0
|
2
= 0
= (6L
1
+ 6L
1
L
0
) |
2
= 0 (L
0
+ 1) |
2
= 0 (46)
I also get L
m>2
| disappear:
L
m
| = L
m
(L
2
+
3
2
L
1
L
1
) |
2

= ([L
m
, (L
2
+
3
2
L
1
L
1
)] + (L
2
+
3
2
L
1
L
1
)L
m
) |
2

= [L
m
, (L
2
+
3
2
L
1
L
1
)] |
2
L
m>0
|
2
= 0
= ((
5
2
m + 2)L
m2
+ 3L
1
L
m1
) |
2
= 0 L
m>0
|
2
= 0 (47)
At m = 2, fortunately d pop-out and I can nd the critical value for it:
L
2
| = L
2
(L
2
+
3
2
L
1
L
1
) |
2

= ([L
2
, (L
2
+
3
2
L
1
L
1
)] + (L
2
+
3
2
L
1
L
1
)L
2
) |
2

= [L
2
, (L
2
+
3
2
L
1
L
1
)] |
2
L
m>0
|
2
= 0
= (13L
0
+
d
2
+ 9L
1
L
1
) |
2
L
m>0
|
2
= 0
= (13 +
d
2
) |
2
(L
0
+ 1) |
2
= 0
= 0 d = 26 (48)
So, at critical value of A = 1 and d = 26, the algebraic structure of L gives us physical
null state. These values give us the boundary for the ghost-free (no negative norm physical
state) spectrum.
One might prefer a more-Physics approach. First, what kind of states I want to have in
the physical picture? Obviously, the fundamental 1-particle states must be included! Since
one can treat the closed string in the same manner with the open string (the only dierent is
for closed string I have one more set of oscillators to write), choose open string to work with.
Lets start with vacuum oscillation mode of a string with center of mass momentum p, the
14
representation state is |0; p. Since theres nothing in the oscillation space, I automatically
get L
m>0
|0; p = 0. For L
0
operators, I can use it to nd the mass of the string, as p
2
= M
2
(mass-shell condition):
(L
0
A) |0; p = (

p
2
A) |0; p = 0 M
2
= p
2
=
A

(49)
For the lowest 1-oscillator mode

1
|0; p, the L
0
condition gives M
2
=
1A

and the
L
1
gives the polarization condition: L
1

m
|0; p =

|0; p = 0

= 0. With
A < 1, M
2
> 0 and one can go to the rest frame and see the polarization in time-direction
vanish
0
= 0, so all the oscillation with time-like modes are disappear and the observable
spectrum in the theory is just like d 1 space-like oscillations, which make this state gets
positive norm. With A > 1, M
2
< 0 and one can go to the frame with momentum goes
along the time-direction and the observable spectrum in the theory is just like time-like
oscillations, which is bad with this state has negative norm. With A = 0, M
2
= 0 and a
degree of freedom for the polarization can be eliminated as I get the observable spectrum
of d 2 space-like oscillations - d 2 states with positive norm. The theory for open string
with A = 1 is really interesting since it gives a massless vector particle (which is similar to a
photon in the real world). Although its a bit unnatural, lets use A = 1. Therere a lots of
unexplained reasons for why I want the critical values are A = 1 but for now let stick with
this and provide better explanations in the next operator-base quantization (the light-cone
gauge quantization) and the later part of the e-book. Indeed, at A < 1, I have d 1 states
and this is more than what I expect from the light-cone quantization - the number of state
shouldnt depend on what kind of gauge I choose, so A = 1 must be the right answer. For
further detail, A = 1 is also required by the world sheet symmetries since only in this case
the vertex operator (for string interaction) is conformally invariant.
The next 1-particle state I want to look at is the next lowest level (in term of mass). The
general representation for that state is |, ; p = (

1
+

2
) |0; p, and I will need
to nd what constrain I should put in it so that Ill get a physical state. The norm of this
state is straight-forward: , ; p| , ; p = 2(

). Now lets nd the conditions


for physical state, start with the mass-shell condition (A = 1) L
0
|, ; p = |, ; p:
L
0
|, ; p = (

p
2
+
1

1
+
2

2
)(

1
+

2
) |0; p
=

p
2
(

1
+

2
) |0; p + 2(

1
+

2
) |0; p
= (

k
2
+ 2) |, ; p = |, ; p M
2
= p
2
=
1

> 0 (50)
15
So this particle is massive, and one can try to go to its rest frame to simplify the calcu-
lation (p
0
=
1/2
and p
i
= 0). Look at the condition for m = 1 (Virasoro index):
L
1
|, ; p = (

p
1
+
1

2
)(

1
+

2
) |0; p
= (2

1
+ 2

1
) |0; p = 0

1
+

1
= 0

2
0
(51)
For m = 2 (Virasoro index):
L
2
|, ; p = (

p
2
+
1
2

1
)(

1
+

2
) |0; p
= (

+ 2

) |0; p = 0

+ 2

= 0
ii
= 5
00
(52)
With higher value of m (m > 2), the condition for L
m
satisfy automatically. Now consider
an eective spurious state given by:
|, ; q = (L
1

1
+ L
2
) |0; q
= ((

(q

+ q

) +
1
2
g

1
+ (

+ q

2
) |0; q (53)
If these state are physical, I need to set:

(q

+ q

) +
1
2
g

+ q

(54)
to get the conditions for eective spurious physical states (no constrain for
i
):
1
2
(d 8) = 3

2
0
; 3

2 =
0
(55)
I want to get them out of the picture. Indeed, the physical states that I want to include
in the theory from this quantization are only physical states with positive norm. The reason
for that is null states will not make any contribution to the spectrum of the theory, since
it orthogonal to all physical states include itself, so it is essentially zero. One can just
rule them out. Theoretically, I get
d(d+3)
2
possible states just from the form of |, ; p. The
condition for physical state removes d +1 degree of freedom, and also the condition for null
states (in general) removes d 1 (for no constrain in
i
) with 1 (for constrain =
0
= 0)
more degree of freedom. At d = 26 the null states constrain make and independence,
thus removes 2 instead of 1. So the number of physical states which are not null is
d(d1)2
2
16
- the same as with light-cone gauge quantization. For d < 26, I has
d(d1)
2
and this is
obviously more than in the light-cone quantization. Since the number of state must be
gauge dependence, it means the correct choice is d = 26 dimensions.
With results I nd in this mass level of the open string, let me show that negative norm
physical state appear in d > 26 with the state has
ij
=
ij
(so the condition for physical
state implies that
00
=
d1
5
and
0
=

2(d1)
5
):
, ; p| , ; p = 2(

) =
2(d 1)(26 d)
25

2
< 0 (56)
B. Light-cone Gauge Quantization
A big problems for the old quantization method is not all the gauge symmetry freedoms
are xed - choosing a at gauge is not enough to eliminate all possible symmetries. Theres
no treatment for stoping the conformal Killing generators - this old covariance approach just
simply neglect it. One might ask for another quantization approach, with all the freedoms of
gauge are eliminated. And thats exactly what Im going to do: let me introduce the method
of light-cone gauge quantization. Unlike the above old covariance method (all components
are covariance, since they are all written with well-dene Lorentzian structure with index ),
the prize to pay when one perform the light-cone quantization is the Lorentzian symmetry
is broken with the new choice of dimensions as the components are now not all covariance
(this choice of gauge picks out a particular time direction and a particular spatial direction
and hybrid them, so it means that any calculations that we do involving these hybrid
dimensions will not be manifestly Lorentz invariant). One can think of it as I choose to
break down the Lorentzian symmetry to x the gauge, then by considering the struture of
the generators (after quantize) that classically generate Lorentz transformation, the only
way one can reconstruct the correct algebraic structure of Lorentz generators is xing A = 1
and d = 26. Of course a classical symmetry might not always a symmetry for the quantum
theory as theres always a possibility for anomaly to enter the picture - after quantizing the
theory in the light-cone gauge there might be some anomaly in the theory. Ill give that
some treatment and show that at these critical values the anomaly is disappear - make the
theory consistence with the choice of gauge.
Dene a light-cone coordinate in space-time with x

=
1

2
(x
0
+ x
1
) and the rest of the
dimensions are labeled by i = 2, 3, ..., d 1. Then the metric is changed a bit g

=
17
1, g

= 0 and theres no communication between these light-cone dimensions and the


other spatial dimensions. The residual gauge symmetry is actually the reparametrization
of the world-sheet light-cone components

. This change =
1
2
(
+
+

) and
=
1
2
(
+

) to =
1
2
(
+
+

) and =
1
2
(
+

). So always a solution for the


free massless wave equation of the form
+

= 0. In general the classical consequence


is not survive after the quantization, so its not a good idea to impose the equation of
motion to x the gauge before quantizing the theory. But lets use this trick rst to get a
quick derivation - as X also satisfy the
+

= 0, I can identify proportionally with


a component of X

and x the gauge. Naturally, because of the Lorentzain struture of


, one might want to assign it with X
0
, but unfortunately, it will not lead to the same
simplication that Ill show you right now. Light-cone gauge mean the choice of picking
X
+
= x
+
+(
2

l
)p
+
. This classically correspond to set all the coecient
+
n
to 0 for n = 0
- thats somehow expected, since I can interpret it as using the inite dimensional conformal
symmetry generator to gauge away an innite number of oscillator degrees of freedom.
And thats how the problem of anomaly comes in the picture. When one try to quantize the
theory of string, there are 2 important things should be consider through the whole process of
quantization for meaningful physical states: unitary and Lorentz invariance. Unitary means
there are no negative norm states and there might be some null state with no contribution
to the dynamical picture of the theory (so they will be neglected). Lorentz invariance, with
the old covariance quantization method, is always preserve step by step, and the price for
that is there are negative norm states (which means the theory need further treatment). In
the light-cone gauge quantization, the degrees of freedom that possibly give the negative
norm states (Ill show that later) are killed in the beginning, but that can only be the right
method if after the quantization, theres no conformal anomaly (which result in the wrong
commutation rule of Lorentz operator). In general, conformal anomaly is a problem with
the theory (and indeed, there are theories of string in non-critical dimension, but Ill not
talk about them now) since it pop-out from the parametrization of the world-sheet and it
isnt something physical - its not a condition for the theory to be consistence. Indeed,
conformal invariance of the world-sheet parametrization of theory is merely for light-cone
gauge quantization (to be more precise, for the process of take out the
+
oscillation degrees
of freedom) to be a consistence quantization method. And actually with the massless vector
object can be intepreted as the photon comes from that light-cone gauge quantization with
18
the critical dimension d = 26, it seems like a good choice to build a theory that describe the
universe. If the conformal symmetry is broken, one can not use the gauge to rule out the
degrees of freedom, and it also means that the theory might have UV divergence behavior.
The Virasoro constrain for T
ab
= 0 leads to:
(

X X

)
2
= 0

X

=
l
2

(

X
i
X
i
)
2
2p
+
(57)
Classically, one can derives the following solution in the light-cone gauge:
X

= x

+ (
2

l
)p

+ i

n=0
1
n

n
e

l
in
cos (
n
l
) (58)
and the relation between oscillation modes (n = 0):

n
=
1

1
2p
+

m=

i
nm

i
m
(59)
At n = 0, one gets
i
0
=

p
i
and

0
=

, so one can represent p

as:
p

=
1
2

1
2p
+

m=

i
m

i
m
(60)
The mass-shell condition gives:
M
2
= p
2
= p
+
p

p
i
p
i
=
1

n>0

i
n

i
n
(61)
To start quantizing the theory, lets determine the dynamical variable: x
+
is not a
dynamical variable since I can toss it away by reparametrize (change the gauge to
X
+
= (
2

l
)p
+
), p

also depend on and p


+
so I can just not count it. Im left with
the following dynamical variables: x
i
, p
i
, x

, p
+
(center of mass) and (oscillation). A
family of oscillation mode is missing (compare to the old quantization method), and thats
the family which gives negative norm state. One should also take a look at the action:
S
P
=
1
4

d d (X
i
)
2
+
_
d p
+
x

[X
i
, P
j
] = i
ij
; [x

, p
+
] = i (62)
The interpretation for [x

, p
+
] = i is actually a modication for [x
0
, p
0
] = i. The
result for [X
i
, P
j
] = i
ij
is the relation between oscillation mode [
i
m
,
j
n
] = m
m+n,0

ij
,
and unlike with the old covariance quantization, I dont have any diculty with negative
norm state since the relation of the ip sign form [
0
m
,
0
n
] = m
m+n,0
disappear! In this
19
quantization, I also run into the order problem for the term

0
, and I also need an ordering
constant A to x it:

0
A

0
=
1

1
2p
+

m=

i
m

i
m
A (63)
Let rst check the general expression for an innitestimal Lorentzian transformation

on the coordinate in the theory with a reparametrization


a
that is compatible with the
at (world-sheet) gauge condition - along the conformal Killing vector: X

a
X

. The reason I include the reparametrization term is because want to preserve the
gauge condition X
+
= x
+
+ (
2

l
)p
+
, which transform innitestimally as X
+
=
+

(
2

l
)p

. Matching these equations gives:

0
= (
l
2

p
+
(x

+ (
2

l
)p

);
1
=
_

0
d

0
(

, )
Thats how the new action of Lorentz transformation change after taking into account the
noncovariant gauge xing feature. Since I have to include the complicate
a
term in the new
Lorentz Conformal gauge-xing transformation, there are nonliniear (quadratic) dynamical
terms in the innitestimal change. In quantum theory, bilinear terms raise issues with
the ordering, so it might make some anomaly and mess up with the Lorentzian algebra.
The generators of Lorentz transformation for open string (no ordering problem with this
operator) have the folowing algebra:
[J

, J

] = ig

+ ig

ig

ig

(64)
Quantizing the system in the light-cone gauge has the advantage that all physically
relevant states are easy to identify, but the price to pay is that Lorentz invariance is not
easy to establish, since the
0
coordinate was identied with the hybrid X
+
. Since the
theory must be Lorentz invariance, the string states that one can nd from this quantization
should be a representation of the Lorentz group (I can also generalize this argument to
Poincare group, but the interesting feature of this quantization is only appear in Lorentz
group). It means even with the light-cone index, the Lorentzian operators need to satisfy
Lorentz algebra. Check the algebra with the light-cone index operators, one nd that theres
a possible anomaly terms arise from the commutation with the following form (fortunately,
I dont run into any anomaly problems with other form):
[J
i
, J
j
] =
i
p
+
2

m>0
(
26 d
12
m +
1
m
(
d 26
12
+ 2(1 A)))(
i
m

j
m

j
m

i
m
) (65)
20
It must equal to 0 in order to satisfy the Lorentzian algebra. Requiring the consistency
in the theory by eliminate the anomaly, one arrives at the critical value A = 1 and d = 26
Another way to x the gauge in the light-cone coordinate without imposing the equation
of motion before choosing how to x the gauge and quantize can be done if I start everything
over again. Consider no at world-sheet gauge xing at all and all local gauge freedoms are
still in the theory. Let start by xing X
+
=
0
naturally (the choice of this xing is not
because of the relation between the extra leftover gauge freedom and the equation of motion
in the at world-sheet gauge xing - its natural) and also impose
1
h
11
= 0 with h = 1
- believe it or not, all the local gauge freedom are swept away with these condition! After
quantization, also by eliminate the anomaly, I also get the constrain A = 1 and d = 26.
Lets see how it works, start with X
+
=
0
. That means h
11

hd
1
= fd
1
= dl must
be unchanged under
1
reparametrization (as the choice for
0
is xed with the background
parameter X). The innitestimal length dl is invariant, and one can try to x d
1
propor-
tional with dl (think about the leftover dieomorphism for reparametrize the world-sheet
in the spatial dimension
1

1
(
0
,
1
), by choosing the proportional relation,

only
depend on
0
) at any give world-sheet time slide
0
, as f =
dl
d
1
= f(
0
). The constant
of proportionality is determined by require that the coordinate of end points remain at 0
and l. The only gauge freedom left after xing dieomorphism is scaling of
1
. Using Weyl
transformation to x h = 1 as the only possible
0
dependence term in f is h
11
, I arrives
at
1
h
11
= 0 (as the obvious result of xing d dl). All gauge freedoms are eliminated!
To see how it disappear, note that for the target space x, I x these with the light-cone
coordinate. If I can show that all the terms of the world-sheet metric h
ab
is a constant or
equal to a quantity that made of target parameter (x, X, p or P), it means that actually
theres no freedoms in the local world-sheet left! From the condition of determinant h, one
of the 3 world-sheet metric freedom is eliminated, and one can assume the remain freedom
with that conditions are h
11
(
0
) (
1
h
11
= 0) and h
10
(
0
,
1
). Use the inverse of world-sheet
metric:
_
_
h
00
h
01
h
10
h
11
_
_
=
_
_
h
11
(
0
) h
10
(
0
,
1
)
h
10
(
0
,
1
)
1h
2
10
(
0
,
1
)
h
11
(
0
)
_
_
(66)
21
The Polyakov Lagrangian (Y

= X

(
0
), x

(
0
) =
1
l
_
l
0
d
1
X

):
L
P
=
1
4

_
l
0
d
1
(h
11
(2
0
x

(
0
X
i
)
2
)
2h
10
(
1
Y

0
X
i

1
X
i
) +
1 h
2
10
h
11

1
X
i

1
X
i
) (67)
Y

is non-dynamical since theres no time derivatives term of it, so it acts as a Lagrangian


multiplier and put the world-sheet spatial derivative of
1
h
10
vanish (
1
2
h
10
= 0). The open
string boundary condition at the 2 ends give h
10

0
X

h
00

1
X

= 0 and with = +, one


gets h
10
= 0 at end points. Since
1
2
h
10
= 0 means h
10
can depend linearly on
1
at most, so
h
10
= 0 everywhere. For h
11
, it actually just the momentum conjugate p

=
L
x

=
l
2

h
11
(up to a prefactor)! So no gauge freedom is left!
The Hamiltonian can be written in this gauge as:
H =
l
2

1
2p
+
_
l
0
d
1
(2

P
i
2
+
1
2

(
1
X
i
)
2
)
S
P
=
l
2

1
2p
+
_

d
2
(2

P
i
2
+
1
2

(
1
X
i
)
2
)
_
d
0
p

=
1
4

d
2
(X
i
)
2
+
_
d
0
p
+
x

(68)
From here, one can do the quantization exactly identical as the earlier light-cone quanti-
zation that Ive shown!
In the end of this section, lets perform a trick to get the string vacuum energy (Casimir
energy). This trick also provides a deeper interpretation of the ordering constant A. This
constant is dened as the ambiguity ordering in promoting the classical generator to a
quantum operator. If the real order of the operator is indeed as symmetric as (choose
light-cone gauge quantization, since it xes all the local gauge freedom and only quantize
the right degree of freedom in the theory):
L
0
=
1
2

m=

m
=: L
0
: +
d 2
2

m>0
m =: L
0
: +A A =
d 2
2

m>0
m (69)
The constant A also correspond to the extra energy E inside the Halmintonian, since
H =

l
L
0
. Thats a divergence vacuum energy E =

l
d2
2

m>0
m, and the computation
of that energy is standard in QFT and proceeds via regularization and renormalisation.
Introduce a cut-o momentum (so l will be dimensionless) to regularize the divergent
22
expression in such a way that the divergence only appears when I let :
E

l
d 2
2

m>0
me

m
l
=

l
d 2
2
(l
2

1
12
+ O(
1
l
)) (70)
One can regularize the theory by kill o the divergence part
2
with the cosmological
term in the action
2
_

d
2

h (since they are all scale with l) and then neglect the
small correction O(
1
l
). Notice that including the cosmological term in the action will break
down the conformal symmetry, so by only eliminate the divergence term the theory can
satisfy conformal symmetry. The Casimir energy is the leftover energy after regularization
E =

l
d2
24
, and this energy appear in the theory because of the spatial limitation of the
string - if one let l , the Casimir energy will be gone. At d = 26, one gets A = 1, and
the trick really has done a good job!
IV. PATH INTEGRAL QUANTIZATION
path integral quantization
BRST quantization
No-gohst theorem
V. PARTICLES SPECTRUM
VI. GENERALIZATION TO SUPERSTRING
VII. EXERCISE
For any part of this chapter that I show only the calculations for the case of open string,
do the calculations for the case of closed string. Its a ridiculously simple exercise, but by
actually write down the calculations, it might help you to understand every step in this
chapter better.
23
VIII. REFERENCES

Electronic address: phanvltt@mit.edu


24

Das könnte Ihnen auch gefallen