9.3 Quantization of Fields in Curved Space 213 which might arise from an interaction such as L = 1 2 m 2 P A i A i - 4 p ⇡A i @ i φ. (9.30) The sign of the mass term is wrong, however. 9.3 Quantization of Fields in Curved Space Some good references for these ideas are: Einstein’s Gravity in a Nutshell by Zee Quantum Fields in Curved Space by Birrell and Davies, Conformal Field Theory by Di Francesco, Mathieu, and S´ en´ echal Let’s focus on scalar fields for simplicity. We usually expand a scalar field in flat space as Fourier might have (1.56) φ(x)= Z d 3 p p (2⇡) 3 2p 0 h a(p) e ip·x + a † (p) e -ip·x i . (9.31) The field φ obeys the Klein-Gordon equation (r 2 - @ 2 0 - m 2 ) φ(x) ⌘ (2 - m 2 ) φ(x)=0 (9.32) because the flat-space modes, which have p 2 = -m 2 , f p (x)= e ip·x (9.33) do (r 2 - @ 2 0 - m 2 ) f p (x) ⌘ (2 - m 2 ) f p (x)=0. (9.34) In terms of these functions f p (x), the field is φ(x)= Z d 3 p p (2⇡) 3 2p 0 h a(p) f p (x)+ a † (p) f ⇤ p (x) i . (9.35) In terms of a discrete set of modes f n , it is φ(x)= X n h a n (p) f n (x)+ a † n f ⇤ n (x) i . (9.36) The action for a scalar field in a space described by the metric g ik is S = - 1 2 Z p gd 4 x h g ik φ ,i φ ,k + ( m 2 + ⇠ R ) φ 2 i (9.37) in which R is the scalar curvature, ⇠ is a constant, commas denote derivatives as in φ ,k = @ k φ, g ij is the inverse of the metric g ij , and g is the absolute
13
Embed
9.3 Quantization of Fields in Curved Space 213quantum.phys.unm.edu/523-18/HawkingRadiation.pdfEinstein’s Gravity in a Nutshell by Zee Quantum Fields in Curved Space by Birrell and
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
9.3 Quantization of Fields in Curved Space 213
which might arise from an interaction such as
L = 12m
2P AiA
i � 4p⇡Ai@
i�. (9.30)
The sign of the mass term is wrong, however.
9.3 Quantization of Fields in Curved Space
Some good references for these ideas are:
Einstein’s Gravity in a Nutshell by ZeeQuantum Fields in Curved Space by Birrell and Davies,Conformal Field Theory by Di Francesco, Mathieu, and Senechal
Let’s focus on scalar fields for simplicity. We usually expand a scalar fieldin flat space as Fourier might have (1.56)
�(x) =
Zd3pp
(2⇡)32p0
ha(p) eip·x + a†(p) e�ip·x
i. (9.31)
The field � obeys the Klein-Gordon equation
(r2 � @20 �m2)�(x) ⌘ (2�m2)�(x) = 0 (9.32)
because the flat-space modes, which have p2 = �m2,
fp(x) = eip·x (9.33)
do
(r2 � @20 �m2) fp(x) ⌘ (2�m2) fp(x) = 0. (9.34)
In terms of these functions fp(x), the field is
�(x) =
Zd3pp
(2⇡)32p0
ha(p) fp(x) + a†(p) f⇤
p (x)i. (9.35)
In terms of a discrete set of modes fn, it is
�(x) =X
n
han(p) fn(x) + a†n f
⇤n(x)
i. (9.36)
The action for a scalar field in a space described by the metric gik is
S = � 1
2
Z pg d4x
hgik�,i�,k +
�m2 + ⇠R
��2
i(9.37)
in which R is the scalar curvature, ⇠ is a constant, commas denote derivativesas in �,k = @k�, gij is the inverse of the metric gij , and g is the absolute
214 E↵ective field theories and gravity
value of the determinant of the metric gij . If the spacetime metric is gij , theninstead of (9.32), the field � obeys the covariant Klein-Gordon equation
@ihp
g gij @j �(x)i��m2 + ⇠R
�pg �(x) = 0. (9.38)
However, to simplify what follows, we will now set ⇠ = 0.To quantize in the new coordinates or in the gravitational field of the
metric gij , we need solutions f 0n(x) of the curved-space equation (9.38)
@ihp
g gij @j f0n(x)
i�m2pg f 0
n(x) = 0 (9.39)
which we label with primes to distinguish them from the flat-space solutions(9.33). We use these solutions to expand the field in terms of curved-spaceannihilation and creation operators, which we also label with primes
�(x) =X
n
ha0n(p) f
0n(x) + a0†n f 0⇤
n (x)i. (9.40)
The flat-space modes obey the orthonormality relations
(fp, fq) = i
Zd3x
hf⇤p (x) @tfq(x)�
�@tf
⇤p (x)
�fq(x)
i
= i
Zd3x
he�ipx (�iq0)eiqx � ip0e�ipx eiqx
i
= 2p0(2⇡)3�3(p� q),
(9.41)
(f⇤p , f
⇤q ) = � 2p0(2⇡)3�3(p� q) and (fp, f
⇤q ) = 0. (9.42)
In terms of discrete modes, the flat-space scalar product is
(fn, fm) = i
Zd3x
hf⇤n(x) @tfm(x)�
�@tf
⇤n(x)
�fm(x)
i= �nm (9.43)
and its orthonormality relations are
(fn, fm) = �nm, (f⇤n, f
⇤m) = � �nm and (fn, f
⇤m) = 0. (9.44)
The scalar product (9.43) is a special case of more general curved-spacescalar product
(f, g) = i
Z
S
pgS d3S va
hf⇤(x) @ag(x)�
�@af
⇤(x)�g(x)
i(9.45)
in which the integral is over a spacelike surface S with a future-pointingtimelike vector va, and gS is the absolute value of the spatial part of themetric gik. This more general scalar product (9.45) is hermitian
(f, g)⇤ = (g, f). (9.46)
9.3 Quantization of Fields in Curved Space 215
It also satisfies the rule
(f, g)⇤ = � (f⇤, g⇤). (9.47)
One may use Gauss’s theorem to show (Hawking and Ellis, 1973, section2.8) that this inner product is independent of S and v. The curved-spacemodes fn(x) obey orthonormality relations
(f 0n, f
0m) = �nm, (f 0⇤
n , f 0⇤m) = � �nm, and (f 0
n, f0⇤m) = 0 (9.48)
like those (9.44) of the flat-space modes.The flat-space modes fn(x) = eipnx naturally describe particles of mo-
mentum pn in flat Minkowski space. In curved space, however, there are ingeneral no equally natural modes. We can consider other complete sets ofmodes f 00
n(x) that are solutions of the curved-space Klein-Gordon equation(9.39) and obey the orthonormality relations (9.48). We can use any of thesecomplete sets of mode functions to expand a scalar field �(x)
�(x) =X
n
anfn(x) + a†nf⇤n(x)
�(x) =X
n
a0nf0n(x) + a0†nf
0⇤n (x)
�(x) =X
n
a00nf00n(x) + a00†n f 00⇤
n (x).
(9.49)
The curved-space orthonormality relations (9.48) imply that
(f`,�) =X
n
an(f`, fn) + a†n(f`, f⇤n) = a`
(f 0`,�) =
X
n
a0n(f0`, f 0
n) + a0†n (f0`, f 0⇤
n ) = a0`
(f 00`,�) =
X
n
a00n(f00`, f 00
n) + a00†n (f 00`, f 00⇤
n ) = a00`
(9.50)
and that
(f⇤`,�) =
X
n
an(f⇤`, fn) + a†n(f
⇤`, f⇤
n) = �a†`
(f 0⇤`,�) =
X
n
a0n(f0⇤`, f 0
n) + a0†n (f0⇤`, f 0⇤
n ) = �a0†`
(f 00⇤`,�) =
X
n
a00n(f00⇤`, f 00
n) + a00†n (f 00⇤`, f 00⇤
n ) = �a00†`.
(9.51)
The completeness of the mode functions lets us expand them in terms of
216 E↵ective field theories and gravity
each other. Suppressing the spacetime argument x, we have
f 0j =
X
i
�↵jifi + �ji f
⇤i
�. (9.52)
The curved-space orthonormality relations (9.48) let us identify these Bo-goliubov coe�cients
(f`, f0j) =
X
i
h↵ji(f`, fi) + �ji (f`, f
⇤i )i= ↵j`
(f⇤`, f 0
j) =X
i
h↵ji(f
⇤`, fi) + �ji (f
⇤`, f⇤
i )i= ��j`.
(9.53)
To find the inverse relations, we use the completeness of the mode functionsf 0ito expand the fj ’s
fj =X
i
�cjif
0i + dji f
0⇤i
�(9.54)
and then use the orthonormality relations (9.48) to form the inner products
(f 0`, fj) =
X
i
hcji(f
0`, f 0
i) + dji (f0`, f 0⇤
i )i= cj`
(f 0⇤`, fj) =
X
i
hcji(f
0⇤`, f 0
i) + dji (f0⇤`, f 0⇤
i )i= �dj`.
(9.55)
The hermiticity (9.46) of the scalar product tells us that the cj`’s are relatedto the ↵’s
cj` = (fj , f0`)⇤ = ↵⇤
`j. (9.56)
The hermiticity (9.46) of the scalar product and the rule (9.47) show that
dj` = � (f 0⇤`, fj) = � (fj , f
0⇤`)⇤ = (f⇤
j , f0`) = ��`j . (9.57)
So the inverse relation (9.54) is
fj =X
i
�↵⇤ijf
0i � �ij f
0⇤i
�. (9.58)
The formulas (9.52 & 9.58) that relate the mode functions of di↵erent metricsare known as Bogoliubov transformations.The vacuum state for a given metric is the state that is mapped to zero
by all the annihilation operators. Our formulas (9.51) for the annihilationand creation operators
a` = (f`,�) and a0`= (f 0
`,�)
a†`= � (f⇤
`,�) and a0†
`= � (f 0⇤
`,�)
(9.59)
9.3 Quantization of Fields in Curved Space 217
let us express the annihilation and creation operators for one metric in termsof those for a di↵erent metric. Thus, using our formula (9.52) for the f 0’s interms of the f ’s, we find
a0j = (f 0j ,�) =
X
i
h↵ji (fi,�) + �ji (f
⇤i ,�)
i=
X
i
⇣↵ji ai � �ji a
†i
⌘. (9.60)
Our formula (9.58) for the f ’s in terms of the f 0’s gives us the inverse relation
aj = (fj ,�) =X
i
h↵⇤ij (f
0i ,�)� �ij (f
0⇤i ,�)
i=
X
i
⇣↵⇤ij a
0i + �ij a
0†i
⌘. (9.61)
The annihilation operators a0jdefine the vacuum state |0i0 by the rules
a0j |0i0 = 0 (9.62)
for all modes j. Thus our formula (9.61) for aj says that
aj |0i0 =X
i
⇣↵⇤ij a
0i + �ij a
0†i
⌘|0i0 =
X
i
�ij a0†i|0i0 (9.63)
The adjoint of this equation is
8h0|a†j= 8h0|
X
i
a0i�⇤ij . (9.64)
Thus the mean value of the number operator a†jaj in the |0i0 vacuum is
8h0|a†jaj |0i0 = 8h0|
X
i
a0i�⇤ij
X
k
�kj a0†k|0i0. (9.65)
The commutation relations
[a0i, a0†k] = �ik (9.66)
and the definition a0j|0i0 = 0 (9.62) of the vacuum |0i0 imply that the average
number (9.65) of particles of mode j in the (normalized) vacuum |0i0 is8h0|a†
jaj |0i0 = 8h0|
X
ik
�⇤ij�kj �ik|0i0 =
X
i
�⇤ij�ij . (9.67)
It follows that the vacuum of one metric contains particles of the othermetric unless the Bogoliubov matrix
�j` = � (f⇤`, f 0
j) (9.68)
vanishes. The value of �j` in the Minkowski-space scalar product (9.41) is
�j` = � (f⇤`, f 0
j) = �i
Zd3x
hf`(x) @tf
0j(x)�
�@tf`(x)
�f 0j(x)
i. (9.69)
218 E↵ective field theories and gravity
This integral for �j` is nonzero, for example, when the functions f` and f 0j
have di↵erent frequencies but are not spatially orthogonal.An example due to Rindler. Let us consider the two metrics
Now u, r, y, z are independent coordinates, so this equation of motion is
� r�1@2uf + @r(r@rf) + r@2
yf + r@2zf = m2rf (9.75)
or
� r�2@2uf + r�1@rf + @2
rf + @2yf + @2
zf = m2f. (9.76)
Let’s make f a plane wave in the y and z directions
f(u, r, y, z) = f(u, r)eiypy+izPz . (9.77)
If we also set
M2 = m2 + P 2y + P 2
z , (9.78)
then we must solve
� r�2@2uf + r�1@rf + @2
rf = M2f (9.79)
9.4 Accelerated coordinate systems 219
where f = f(u, r). We now make the Daniel-Middleton transformation,separating the dependence of f upon r and u
f(u, r) = a(u)b(r). (9.80)
Our di↵erential equation (9.79) reduces to
� r�2ab+ r�1ab0 + ab00 = M2ab (9.81)
in which dots denote @u and primes @r. Dividing by a, we get
� a
a
b
r2+
b0
r+ b00 �M2b = 0. (9.82)
As a(u), we choose
a(u) = ei!u anda
a= � !2. (9.83)
Then our equation (9.82) for b(r) is
b00 +b0
r�⇣M2 � !2
r2
⌘b = 0 (9.84)
or equivalently
r2b00 + rb0 �⇣M2r2 � !2
⌘b = 0. (9.85)
Yi’s solution is a modified Bessel function I⌫(z) for imaginary ⌫.Bogoliubov’s � matrix is
�j` = � (f⇤`, f 0
j). (9.86)
The scalar product (9.45) uses the metric of one, not both, of the solu-tions. If we choose the flat-space metric, then we need to write the solutionf 0j(u, r, y, z) in terms of the flat-space coordinates t, x, y, z. The mean occu-
pation number (9.67) is the sumX
j
8h0|a†jaj |0i0 =
X
ij
|�ij |2 =X
ij
|(f⇤j , f
0i)|2. (9.87)
9.4 Accelerated coordinate systems
We recall the Lorentz transformations
t0 = �(t� vx), x0 = �(x� vt), y0 = y, and z0 = z
t = �(t0 + vx0), x = �(x0 + vt0), y = y0, and z = z0(9.88)
220 E↵ective field theories and gravity
in which � = 1/p1� v2. The velocities (in the x direction) are
u =dx
dtand u0 =
dx0
dt0=
�(dx� vdt)
�(dt� vdx)=
(u� v)
(1� vu). (9.89)
So the accelerations (in the x direction) are
a =du
dt
a0 =du0
dt0= d
h (u� v)
(1� vu)
i��(dt� vdx)
=h du
(1� vu)+
(u� v)vdu
(1� vu)2
i��(dt� vdx)
=hdu(1� vu)
(1� vu)2+
(u� v)vdu
(1� vu)2
i��(dt� vdx)
=du(1� v2)
�(1� vu)2(dt� vdx)
=(1� v2)a
�(1� vu)2(1� vu)
=(1� v2)a
�(1� vu)3=
a
�3(1� vu)3.
(9.90)
Now we let the acceleration a0 be a constant. That is, the acceleration inthe (instantaneous) rest frame of the frame moving instantaneously at u = vin the laboratory frame is a constant, a0 = ↵. In this case, since u = v, weget an equation (Rindler, 2006, sec. 3.7)
a0 = ↵ =a
�3(1� vu)3=
a
�3(1� v2)3=
a
(1� v2)3/2
=1
(1� v2)3/2du
dt=
1
(1� u2)3/2du
dt=
d
dt
✓up
1� u2
◆ (9.91)
that we can integrateup
1� u2= ↵ (t� t0). (9.92)
Squaring and solving for u, we find
u =dx
dt=
↵(t� t0)p1 + ↵2(t� t0)2
(9.93)
which we can integrate to
x = ↵�1Z
dt↵2(t� t0)p
1 + ↵2(t� t0)2=
p1 + ↵2(t� t0)2 � 1
↵+ x0. (9.94)
9.5 Scalar field in an accelerating frame 221
The proper time of the accelerating frame is ⌧ = t0. An interval dt0 ofproper time is one at which dx0 = 0. So dt0 =
p1� v2dt or dt⌧ =
p1� u2dt
where u(t) is the velocity (9.93). Integrating the equation
dt0 = d⌧ =p
1� u2dt =
s
1� ↵2(t� t0)2
1 + ↵2(t� t0)2dt =
dtp1 + ↵2(t� t0)2
,
(9.95)we get
↵(t0 � t00) = ↵(⌧ � ⌧0) =
Z↵ dtp
1 + ↵2(t� t0)2= arcsinh(↵(t� t0)). (9.96)
So
↵(t� t0) = sinh(↵(⌧ � ⌧0)). (9.97)
The formula (9.94) for x now gives
↵(x� x0) = cosh(↵(⌧ � ⌧0)). (9.98)
9.5 Scalar field in an accelerating frame
For simplicity, we’ll work with a real scalar field (1.54)
�(x) =
Zd3pp
(2⇡)32p0
ha(p) eip·x + a†(p) e�ip·x
i. (9.99)
If the field is quantized in a box of volume V , then the expansion of the fieldis
�(x) =X
k
1p2!kV
ha(k) eik·x + a†(k) e�ik·x
i. (9.100)
For the scalar field (9.99), the zero-temperature correlation function is the
222 E↵ective field theories and gravity
mean value in the vacuum
h0|�(x, t)�(x0, t0)|0i = h0|Z
d3pp(2⇡)32p0
ha(p) eip·x + a†(p) e�ip·x
i
⇥Z
d3p0p(2⇡)32p00
ha(p0) eip
0·x0+ a†(p0) e�ip
0·x0i|0i
= h0|Z
d3pd3p0
(2⇡)3p
2p02p00a(p)a†(p0) eip·x�ip
0·x0 |0i
=
Zd3pd3p0
(2⇡)3p2p02p00
�3(p� p0)eip·x�ip0·x0
=
Zd3p
(2⇡)32p0eip·(x�x
0). (9.101)
This integral is simpler for massless fields. Setting p = |p| and r = |x� x0|,we add a small imaginary part to the exponential and find
h0|�(x, t)�(x0, t0)|0i =Z
d3p
(2⇡)32peipr cos ✓�ip(t�t
0�i✏)
=
Zpdp dcos ✓
(2⇡)22eipr cos ✓�ip(t�t
0�i✏)
=
Zpdp
(2⇡)22
✓eipr � e�ipr
ipr
◆e�ip(t�t
0�i✏)
=
Z 1
0
dp
(2⇡)22ir
�eipr � e�ipr
�e�ip(t�t
0�i✏)
=1
(2⇡)22ir
✓� 1
i(r � (t� t0))� 1
i(r + (t� t0))
◆
=1
(2⇡)22r
✓1
r � (t� t0)+
1
r + (t� t0)
◆
=1
(2⇡)2[(x� x0)2 � (t� t0)2]. (9.102)
This is the zero-temperature correlation function.In the instantaneous rest frame of an accelerated observer moving in the
x direction with coordinates (9.97 & 9.98), this two-point function is
as well as cosh2 ↵⌧ � sinh2 ↵⌧ = 1 were used.Now let’s compute the same correlation function at a finite inverse tem-
perature � = 1/(kBT )
h�(0, ⌧)�(0, ⌧ 0)i� = Tr⇥�(0, 0)�(0, t)e��H
⇤.Tr
�e��H
�. (9.106)
The mean value of the number operator for a given momentum k is
ha†(k) a(k)i� = nk = Tr⇥a†(k) a(k)e��H
⇤.Tr
�e��H
�(9.107)
in which
H0 =X
k
!k
�a†(k) a(k) + 1
2
�. (9.108)
The trace over all momenta with k0 6= k is unity, and we are left with
ha† ai� = nk = Tr⇥a† ae��!ka
†a⇤.
Tr�e��!ka
†a�
= � 1
!kTr�e��!ka
†a� @
@�Tr
�e��!ka
†a� (9.109)
in which a†a ⌘ a†(k) a(k) and the 1/2 terms have cancelled. The trace is
Tr�e��!ka
†a�=
X
n
e��n!k =1
1� e��!k. (9.110)
So we have
ha† ai� = ��1� e��!k
�
!k
(�!ke��!k)�1� e��!k
�2 =1
e�!k � 1. (9.111)
In the trace (9.106), only terms that don’t change the number of quantain each mode contribute. So the mean value of the correlation function atinverse temperature � = 1/(kBT ) is
h�(0, ⌧)�(0, ⌧ 0)i� =X
k
1
2kV
⇢⇣e~!k/kBT � 1
⌘�1ei!k(⌧�⌧
0)
+
⇣e~!k/kBT � 1
⌘�1+ 1
�e�i!k(⌧�⌧
0)
�.
(9.112)
224 E↵ective field theories and gravity
In the continuum limit, this ⌧, ⌧ 0 correlation function is two integrals. Formassless particles, the simpler one requires some regularization because itinvolves zero-point energies
A� =X
k
1
2kVe�ik(⌧�⌧
0) =
Zd3k
(2⇡)32ke�ik(⌧�⌧
0) =
Zk2dk
(2⇡)2ke�ik(⌧�⌧
0)
=
Z 1
0
kdk
(2⇡)2e�ik(⌧�⌧
0).
(9.113)
We send ⌧ � ⌧ 0 ! ⌧ � ⌧ 0 + i✏ and find
A� =
Z 1
0
kdk
(2⇡)2e�ik(⌧�⌧
0+i✏) = id
d⌧
Z 1
0
dk
(2⇡)2e�ik(⌧�⌧
0+i✏)
= id
d⌧
Z 1
0
dk
(2⇡)2e�ik(⌧�⌧
0+i✏) =1
(2⇡)2d
d⌧
1
(⌧ � ⌧ 0)
= � 1
(2⇡)21
(⌧ � ⌧ 0)2
(9.114)
as ✏ ! 0.The second integral is
B� =
Z 1
0
kdk
(2⇡)2
⇣e�k � 1
⌘�12 cos(k(⌧ � ⌧ 0)) (9.115)
which Mathematica says is
B� =1
(2⇡)2(⌧ � ⌧ 0)2� csch2(⇡(⌧ � ⌧ 0)/�)
4�2. (9.116)
Thus the finite-temperature correlation function is
h�(0, ⌧)�(0, ⌧ 0)i� = A� +B� = � 1
4�2 sinh2(⇡(⌧ � ⌧ 0)/�). (9.117)
Equating this formula to the zero–temperature correlation function (9.103)in the accelerating frame, we find
1
4�2 sinh2(⇡(⌧ � ⌧ 0)/�)=
↵2
(4⇡)2 sinh2[↵(⌧ � ⌧ 0)/2](9.118)
which says redundantly
kBT =↵
2⇡and ⇡kBT =
↵
2. (9.119)
So a detector accelerating uniformly with acceleration ↵ in the vacuum feels
9.6 Maximally symmetric spaces 225
a nonzero temperature
T =~↵
2⇡ckB. (9.120)
This result (Davies, 1975) is equivalent to the finding (Hawking, 1974) thata gravitational field of local acceleration g makes empty space radiate at atemperature
T =~g
2⇡ckB. (9.121)
Black holes are not black.
9.6 Maximally symmetric spaces
The spheres S2 and S3 and the hyperboloids H2 and H3 are maximally sym-metric spaces. A transformation x ! x0 is an isometry if g0
ik(x0) = gik(x0)
in which case the distances gik(x)dxidxk = g0ik(x0)dx0idx0k = gik(x0)dx0idx0k
are the same. To see what this symmetry condition means, we consider theinfinitesimal transformation x0` = x` + ✏y`(x) under which to lowest ordergik(x0) = gik(x)+ gik,`✏y` and dx0i = dxi+ ✏yi,jdx
j . The symmetry conditionrequires
gik(x)dxidxk = (gik(x) + gik,`✏y
`)(dxi + ✏yi,jdxj)(dxk + ✏yk,mdxm) (9.122)
or
0 = gik,` y` + gim ym,k + gjk y
j
,i. (9.123)
The vector field yi(x) must satisfy this condition if x0i = xi+ ✏yi(x) is to bea symmetry of the metric gik(x). Since the covariant derivative of the metrictensor vanishes, gik;` = 0, we may write the condition on the symmetryvector y`(x) as