Simplified derivation of the Hawking–Unruh temperature
for an accelerated observer in vacuum
Paul M. Alsinga)
Department of Physics and Astronomy, University of New Mexico, Albuquerque, New Mexico 87131
Peter W. Milonnib)
Theoretical Division (T-DOT), Los Alamos National Laboratory, Los Alamos, New Mexico 87545
共Received 27 January 2004; accepted 23 April 2004兲
A detector undergoing uniform acceleration a in a vacuum field responds as though it were
immersed in thermal radiation of temperature T⫽បa/2␲ kc. An intuitive derivation of this result is
given for a scalar field in one spatial dimension. The approach is extended to the case where the field
detected by the accelerated observer is a spin 1/2 Dirac field. © 2004 American Association of Physics
Teachers.
关DOI: 10.1119/1.1761064兴
I. INTRODUCTION
Hawking1 predicted that a black hole should radiate with a
temperature T⫽បg/2␲ kc, where g is the gravitational accel-
eration at the surface of the black hole, k is Boltzmann’s
constant, and c is the speed of light 共the Hawking effect兲. The
radiation results from the effect of the strong gravitation on
the vacuum field. Shortly thereafter it was shown separately
by Davies and Unruh that a uniformly accelerated detector
moving through the usual flat space–time vacuum of con-
vential quantum field theory responds as though it were in a
thermal field of temperature2– 6
បa
T⫽ , 共1兲
2 ␲ kc
where a is the acceleration in the instantaneous rest frame of
the detector 共the Davies–Unruh effect兲. We will refer to Eq.
共1兲 as the Hawking–Unruh temperature because its applica-
bility to flat space–time accelerated detectors or to stationary
detectors situated outside the horizon of a black hole7 de-
pends only on the interpretation of the source of the accel-
eration a.
The results of Hawking, Davies, and Unruh suggest pro-
found consequences for the merger of quantum field theory
and general relativity and sparked intense debates over unre-
solved questions that are still being actively investigated. If
black holes are not really ‘‘black,’’ are naked singularities the
ultimate fate of black holes, or will the long-sought fusion of
quantum mechanics and general relativity into a coherent
theory of quantum gravity prevent such occurrences? If a
quantum mechanical pure state is dropped into a black hole
and pure thermal 共uncorrelated兲 radiation results, how does
one explain the apparent nonunitary evolution of the pure
state to a mixed state?
The intriguing consequence of quantum field theory for
accelerated detectors indicated by Eq. 共1兲 has not been de-
rived in a physically intuitive way. Numerous explicit and
detailed calculations have appeared in the scientific literature
over the last 30 years for a wide variety of space–times.
However, even for the simplest calculation involving a scalar
field 共bosons兲, the intricacies of field theory techniques,
coupled with a forest of special function properties, make
most derivations intractable for the nonspecialist. An inves-
tigation of the flat space–time Davies–Unruh effect for
1524 Am. J. Phys. 72 共12兲, December 2004 http://aapt.org/ajp
Dirac particles of spin 1/2 共fermions兲 brings in a whole host
of new machinery, the least of which is the formulation of
the Dirac equation in curvilinear coordinates 共that is, in
curved space–time兲. This formulation quickly goes beyond
the expertise of most nonexperts. However, in both the boson
and fermion cases, the beautiful and important results of
Hawking, Davies, and Unruh can be stated quite simply: for
a scalar field 共bosons兲 a detector carried by an accelerated
observer detects a Bose–Einstein 共BE兲 number distribution
of particles at temperature T given by Eq. 共1兲, while for a
spin 1/2 field 共fermions兲 a detector carried by an accelerated
observer detects a Fermi–Dirac number distribution at the
same temperature.
The purpose of this paper is to present a simplified deri-
vation of Eq. 共1兲 that is suitable for advanced undergraduate
or beginning graduate students and elucidates the essential
underlying physics of the flat space–time Davies–Unruh ef-
fect. Once the simplest features of a quantized vacuum field
are accepted, Eq. 共1兲 emerges as a consequence of time-
dependent Doppler shifts in the field detected by the accel-
erated observer.
In Sec. II the essential features of uniform acceleration are
reviewed, and in Sec. III we use these results to obtain Eq.
共1兲 in an almost trivial way based on the Doppler effect. In
Sec. IV this simple approach to the derivation of the tem-
perature in Eq. 共1兲 is developed in more detail. In Sec. V we
extend the previous calculations for scalar fields to spin 1/2
Dirac fields. We close with a brief summary and discussion.
II. UNIFORM ACCELERATION
We will refer to an observer moving with constant velocity
in flat space–time as a Minkowski observer and refer to a
Rindler observer8 as one who travels with uniform accelera-
tion in the positive z direction with respect to the former.
Uniform acceleration is defined as a constant acceleration a
in an instantaneous inertial frame in which the 共Rindler兲 ob-
server is at rest. The acceleration d v /dt of the Rindler ob-
server as measured in the lab frame 共that is, by the
Minkowski observer兲 is given in terms of a by a Lorentz
transformation formula which relates the acceleration in the
two frames9
冉 冊
dv v2 3/2
⫽a 1⫺ 2 . 共2兲
dt c
© 2004 American Association of Physics Teachers 1524
If we integrate Eq. 共2兲 and take v ⫽0 at t⫽0, we have
v (t)⫽at/ 冑1⫹a 2 t 2 /c 2 . The relation dt⫽d ␶ / 冑1⫺ v 2 /c 2 be- 冕⫺⬁
⬁
d ␶ e i⍀ ␶ e i( ␻ K c/a)e
a ␶ /c
⫽
c
a
冕0
⬁
dy y (i⍀c/a⫺1) e i( ␻ K c/a)y
tween the lab time, t, and the proper time, ␶, for the accel-
erated observer gives t( ␶ )⫽(c/a)sinh(a␶/c) if we take t( ␶
⫽0)⫽0. The velocity v of the accelerated observer as de-
c
⫽ ⌫
a
i⍀c
a 冉 冊
冉 冊
tected from the lab frame can be expressed in terms of the ⫺i⍀c/a
proper time ␶ as ␻ Kc
⫻ e ⫺ ␲ ⍀c/2a , 共8兲

冉 冊
a
a␶
v共 ␶ 兲 ⫽c tanh . 共3兲 where ⌫ is the gamma function.10 Then, because
c 兩 ⌫(i⍀c/a) 兩 2 ⫽ ␲ / 关 (⍀c/a)sinh(␲⍀c/a)兴,11 we obtain
A straightforward integration of Eq. 共3兲 using v (t( ␶ ))
⫽dz/dt⫽dz( ␶ )/d ␶ •d ␶ /dt yields the well-known hyperbolic 冏冕 ⬁

⫺⬁
d ␶ e i⍀ ␶ e i( ␻ K c/a)e
a ␶ /c
冏 2
⫽
2␲c
⍀a e
1
2 ␲ ⍀c/a
⫺1
. 共9兲
orbit of the accelerated, Rindler observer in the z direction:8
The time-dependent Doppler shift detected by the acceler-
c
t 共 ␶ 兲 ⫽ sinh
a
a␶
c
, 冉 冊 z共 ␶ 兲⫽
c
a
2
cosh
c冉 冊
a␶
. 共4兲
ated observer therefore leads to the Planck factor (e ប⍀/kT
⫺1) ⫺1 indicative of a Bose–Einstein distribution for scalar
共boson兲 particles with T⬅បa/2␲ kc, which is just Eq. 共1兲.
We will consider a⬎0, that is, the observer accelerates in the We obtain the same result for a wave propagating in the ⫹z
direction.
positive z direction.
Note that the time-dependent phase can also be obtained
directly by considering the standard nonaccelerated
Minkowski plane wave exp关i␸⫾兴⬅exp关i(Kz⫾␻Kt)兴 and sub-
III. INDICATION OF THERMAL EFFECT OF stituting Eq. 共4兲: ␸ ⫾ ( ␶ )⫽Kz( ␶ )⫾ ␻ K t( ␶ )⫽( ␻ K c/a)
ACCELERATION ⫻exp(⫾a␶/c) with K⫽ ␻ K /c. 12,13

Consider a plane wave field of frequency ␻ K and wave

vector K parallel or anti-parallel to the z direction along
which the observer is accelerated. In the instantaneous rest
frame of the observer, the frequency ␻ K⬘ of this field is given
by the Lorentz transformation
IV. A MORE FORMAL DERIVATION
The derivation of the temperature in Eq. 共1兲 leaves much
to be desired. We have restricted ourselves to a single field
frequency ␻ K , whereas a quantum field in vacuum has com-

␻ K⬘ 共 ␶ 兲 ⫽
␻ K ⫺K v共 ␶ 兲
⫽
冋
␻ K 1⫺tanh 冉 冊册a␶
c
ponents at all frequencies. Moreover, we have noted the ap-
pearance of the Planck factor, but have not actually com-
pared our result to that appropriate to an observer at rest in a

冑 冉 冊
冑1⫺ v 2 共 ␶ 兲 /c 2 thermal field 共that is, a field in which the average number of
a␶ particles is given by a BE distribution for bosons or a Fermi–
1⫺tanh2
c Dirac distribution for fermions at a fixed temperature T).
To rectify these deficiencies, let us consider a massless
⫽ ␻ K e ⫺a ␶ /c 共5兲 scalar field in one spatial dimension (z), quantized in a box
of volume V: 14
for K⫽⫹ ␻ K /c, that is, for plane wave propagation along
the z direction of the observer’s acceleration. For propaga-
tion in the ⫺z direction, ␾ˆ ⫽ 兺
K
冉 2 ␲ បc 2
␻ KV 冊 1/2
关 â K e ⫺i ␻ K t ⫹â K† e i ␻ K t 兴 . 共10兲

␻ K⬘ 共 ␶ 兲 ⫽ ␻ K e a ␶ /c ,
where K⫽⫺ ␻ K /c. Note that for small values of a ␶ , ␻ K⬘
⬵ ␻ K (1⫿a ␶ /c), the familiar Doppler shift. Equations 共5兲
and 共6兲 involve time-dependent Doppler shifts detected by
the accelerated observer.
Because of these Doppler shifts, the accelerated observer
sees waves with a time-dependent phase ␸ ( ␶ )
⬅ 兰 ␶ ␻ K⬘ ( ␶ ⬘ )d ␶ ⬘ ⫽( ␻ K c/a)exp(a␶/c). We suppose therefore
that, for a wave propagating in the ⫺z direction, for which
兰 ␶ ␻ K⬘ ( ␶ ⬘ )d ␶ ⬘ ⫽( ␻ K c/a)exp(a␶/c), the observer sees a fre-
quency spectrum S(⍀) proportional to
共6兲 Here K⫽⫾ ␻ K /c, and â K and â K† are, respectively, the an-
†
nihilation and creation operators for mode K ( 关 â K ,a K ⬘ 兴
⫽ ␦ KK ⬘ , 关 â K ,a K ⬘ 兴 ⫽0). We use ˆ to denote quantum me-
chanical operators. The expectation value 具 (d ␾ ˆ /dt) 2 典 /4␲ c 2
⫺1
of the energy density of this field is V 兺 K ប ␻ K 关 具 â K† â K 典
⫹1/2兴 . For simplicity, we consider the field at a particular
point in space 共say, z⫽0), because spatial variations of the
field will be of no consequence for our purposes.
For a 共bosonic兲 thermal state the number operator â K† â K
has the expectation value (e ប ␻ K /kT ⫺1) ⫺1 . Consider the
Fourier transform operator
1
冕 ⬁

冏冕 冏 ĝ 共 ⍀ 兲 ⫽ dt ␾
⬁ 2 ˆ e i⍀t
i⍀ ␶ i( ␻ K c/a)e a ␶ /c 2␲ ⫺⬁
d␶ e e . 共7兲

冉 冊
⫺⬁
2 ␲ បc 2 1/2

If we change variables to y⫽e a ␶ /c , we have

⫽ 兺K ␻ KV
â K ␦ 共 ␻ K ⫺⍀ 兲 共 ⍀⬎0 兲 . 共11兲

1525 Am. J. Phys., Vol. 72, No. 12, December 2004 Paul M. Alsing and Peter W. Milonni 1525
The expectation value 具 ĝ † (⍀)ĝ(⍀ ⬘ ) 典 in thermal equilibrium which is identical to the thermal result Eq. 共13兲 if we define
is therefore the temperature by Eq. 共1兲.

冉 冊
Note that the expectation value 具 â K a K ⬘ 典 ⫽ ␦ KK ⬘ involves
†
2 ␲ បc 2
具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典 ⫽ 兺 具 â K† â K 典 ␦ 共 ⍀⫺⍀ ⬘ 兲 the creation and annihilation operators of the accelerated ob-
K ␻ KV server and is taken with respect to the accelerated observer’s
vacuum, which is different from the vacuum detected by the
⫻ ␦ 共 ␻ K ⫺⍀ 兲 nonaccelerated observer. This point is discussed more fully

⫽ 兺K 冉 2 ␲ បc 2
冊
1
␻ K V e ប ␻ K /kT ⫺1
in Sec. VI.

V. FERMI–DIRAC STATISTICS FOR DIRAC

⫻ ␦ 共 ⍀⫺⍀ ⬘ 兲 ␦ 共 ␻ K ⫺⍀ 兲 . 共12兲 PARTICLES
We go to the limit where the volume of the quantization box We have considered a scalar field and derived the Planck
becomes very large, V→⬁, so that we can replace the sum factor (e ប⍀/kT ⫺1) ⫺1 indicative of Bose–Einstein 共BE兲 sta-
over K by an integral: 兺 K →(V/2␲ ) 兰 dK. 15 Thus tistics. We began with the standard plane-wave solutions of
the form exp关i(Kz⫾␻Kt)兴 for the nonaccelerated Minkowski
具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典 ⫽បc 2 冕 ⬁

⫺⬁
dK
1 1
␻ K e ប⍀/kT ⫺1
observer and considered the time-dependent Doppler shifts
as detected by the accelerated observer. For spin 1/2 Dirac
particles we would expect an analogous derivation to obtain
⫻ ␦ 共 兩 K 兩 c⫺⍀ 兲 ␦ 共 ⍀⫺⍀ ⬘ 兲 (e ប⍀/kT ⫹1) ⫺1 indicative of Fermi–Dirac 共FD兲 statistics.
2បc/⍀ Mathematically, the essential point involves the replace-
⫽ ␦ 共 ⍀⫺⍀ ⬘ 兲 . 共13兲 ment i⍀c/a→i⍀c/a⫹1/2 in the integrals in Eqs. 共7兲–共9兲,16
e ប⍀/kT ⫺1 and the relation 兩 ⌫(i⍀c/a⫹1/2) 兩 2 ⫽ ␲ /cosh(␲⍀c/a).11
Now consider an observer in uniform acceleration in the Physically, this replacment arises from the additional spinor
quantized vacuum field, that is, the particle free vacuum ap- nature of the Dirac wave function over that of the scalar
propriate for the accelerated Rindler observer. This observer plane wave. For a scalar field, only the phase had to be
sees each field frequency Doppler-shifted according to Eqs. instantaneously Lorentz-transformed to the comoving frame
共5兲 and 共6兲, and so for him/her the operator ĝ(⍀) has the of the accelerated observer. For non-zero spin, the spinor
form structure of the particles must also be transformed,17 or
Fermi–Walker transported18 along a particle’s trajectory to

ĝ 共 ⍀ 兲 ⫽
1
2␲
冕 ⫺⬁
⬁
d ␶ e i⍀ ␶ 兺K 冉 2 ␲ បc 2
␻ KV 冊 1/2 ensure that it does not ‘‘rotate’’ as it travels along the accel-
erated trajectory. Ensuring this ‘‘nonrotating’’ condition in
the observer’s instantaneous rest frame leads to a time-
R␶ d ␶ ␻ ⬘ ( ␶ ) dependent Lorentz transformation of the Dirac bispinor of
⫻ 关 â K e ⫺i ⬘ K ⬘ ⫹â K† e i ⬘ K ⬘
R␶ d ␶ ␻ ( ␶ )
兴
form19 Ŝ( ␶ )⫽exp(␥0 ␥3 a␶/2c)⫽cosh(a␶/2c)
冉 冊
the
冕 2 ␲ បc
⬁ 2 1/2
1 ⫹ ␥ ␥ sinh(a␶/2c), where the 4⫻4 constant Dirac matrices
0 3
⫽
2␲ ⫺⬁
d ␶ e i⍀ ␶ 兺K ␻ KV are given by

⫻ 关 â K e i( ⑀ K ␻ K c/a)e
⫺ ⑀ K a ␶ /c
⫹â K† e ⫺i( ⑀ K ␻ K c/a)e
⫺ ⑀ K ␻ K ␶ /c
兴,
共14兲
␥ 0⫽ 冉 1
0
0
⫺1
冊 , ␥ 3⫽ 冉 0
⫺␴z
␴z
0
冊 , 共17兲

and ␴ z ⫽diagonal(1,⫺1) is the usual 2⫻2 Pauli spin matrix

where ⑀ K ⫽ 兩 K 兩 /K. Because â K 兩 vacuum典 ⫽0, only the â K†
in the z direction. If Ŝ( ␶ ) acts on a spin up state 兩 ↑ 典
terms in Eq. 共14兲 contribute to the vacuum expectation value
具 ĝ † (⍀)ĝ(⍀ ⬘ ) 典 . If we perform the integrals over ␶ as before ⫽ 关 1,0,1,0 兴 T , 20 Ŝ( ␶ ) 兩 ↑ 典 ⫽exp(a␶/2c) 兩 ↑ 典 . Thus, for the spin
up Dirac particle we should replace the plane wave scalar
and use 具 â K a K ⬘ 典 ⫽ ␦ KK ⬘ , we obtain
†
wave function exp关i␸(␶)兴 used in Eq. 共7兲 by

具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典 ⫽ 冉 冊冉
c
2␲a
2
2 ␲ បc 2
V
⌫
i⍀c
a 冊冏 冉 冊冏 2 exp(a␶/2c)exp关i␸(␶)兴.21 This replacement leads to i⍀c/a
→i⍀c/a⫹1/2 in Eq. 共8兲, and therefore the result

⫻e ⫺ ␲ c⍀/a
兺K
1 ␻ Kc
␻K a 冉 冊 i ⑀ K (⍀⫺⍀ ⬘ )c/a
,
冏冕 ⬁

⫺⬁
d ␶ e i⍀ ␶ e a ␶ /2c e i( ␻ K c/a)e 冏
a ␶ /c
2
⫽
2␲c
␻ Ka e
1
2 ␲ ⍀c/a
⫹1
.
共18兲
共15兲 If we compare Eq. 共18兲 with Eq. 共9兲, we note the crucial
where we have used the fact that the sum over K vanishes change of sign in the denominator from ⫺1 for BE statistics
unless ⍀⫽⍀ ⬘ . We show in the Appendix that the sum over to ⫹1 for FD statistics. We also note that the prefactor in Eq.
K is (2Va/c 2 ) ␦ (⍀⫺⍀ ⬘ ), so that 共9兲 involves the dimensionless frequency ⍀c/a while in Eq.
共18兲 the prefactor involves the factor ␻ K c/a 共the argument of

具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典 ⫽
បc 2
␲a
⌫ 冏 冉 冊冏
i⍀c
a
2
e ⫺ ␲ c⍀/a ␦ 共 ⍀⫺⍀ ⬘ 兲
the exponential in the distribution function is still ប⍀/kT
with the same Hawking–Unruh temperature T⫽បa/2␲ kc).
This difference in the prefactor is no cause for concern, be-
2បc/⍀ cause in fact a single frequency ␻ K detectable by a
⫽ 2 ␲ ⍀c/a␦ 共 ⍀⫺⍀ ⬘ 兲 , 共16兲
e ⫺1 Minkowski observer is actually spread over a continuous

1526 Am. J. Phys., Vol. 72, No. 12, December 2004 Paul M. Alsing and Peter W. Milonni 1526
range of frequencies ⍀ detectable by the accelerated Rindler
observer, with a peak centered at ⍀⫽ ␻ K . 22 This fact allows
us to replace ␻ K by ⍀ in the final result. 关This frequency
replacement is explicitly evidenced by the delta function
␦ ( ␻ K ⫺⍀) in Eqs. 共11兲–共16兲 in the comparison of the ther-
mal and accelerated correlation functions.兴
For the spin up Dirac particle, the more formal field-
theoretic derivation of Sec. IV proceeds in exactly the same
fashion, with the modification of the accelerated wave func-
tion from exp关i␸(␶)兴→exp(a␶/2c)exp关i␸(␶)兴 and the use of
anti-commutators 兵 â K ,a K ⬘ 其 ⫽ ␦ KK ⬘ for the quantum me-
† fluctuations in the vacuum state are crucial for the thermal
chanical creation and annihilation operators instead of the
commutators appropriate for scalar BE particles. For the cor-
relation function we find
2បc/ ␻ K
具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典 ⫽ ␦ 共 ⍀⫺⍀ ⬘ 兲 ,
e 2 ␲ ⍀c/a ⫹1
the FD analogue of Eq. 共16兲.
VI. SUMMARY AND DISCUSSION
In the usual derivation of the Hawking-Unruh temperature
Eq. 共1兲,2– 6 one solves the wave 共or Dirac兲 equation for the
field mode functions in the Rindler coordinates Eq. 共4兲, and
then quantizes them. Because the hyperbolic orbit of the ac-
celerated observer Eq. 共4兲 is confined to the region of
Minkowski space–time z⬎0, z⬎ 兩 t 兩 bounded by the asymp-
totes t⫽⫾z called the right Rindler wedge 关with mirror or-
bits confined to the left Rindler wedge defined by z⬍0, 兩 z 兩
⬎ 兩 t 兩 obtained from defining the accelerated observer’s coor-
dinates as t( ␶ )⫽⫺(c/a)sinh(a␶/c) and z( ␶ )⫽⫺(c 2 /
a)cosh(a␶/c)], it turns out that the vacuum detected by the
accelerated observer in say, the right Rindler wedge is dif-
ferent than the usual Minkowski vacuum 共defined for all z
and t) detected by the unaccelerated observer. The inequiva-
lence of these vacua 共and hence the Minkowski versus Rin-
dler quantization procedures23兲 is due to the fact that the
right and left Rindler wedges are causally disconnected from
each other. Readers can easily convince themselves of the
causal disconnectedness of the right and left Rindler wedges
by drawing a Minkowski diagram in (z,t) coordinates and
observing that light rays at ⫾45° emanating from one wedge
do not penetrate the other wedge. Hence the Minkowski
vacuum that the accelerated observer moves through appears
to her/him as an excited state containing particles, and not as
the particle free vacuum appropriate for the right Rindler
wedge. The Bose–Einstein distribution with the Hawking–
Unruh temperature T for scalar fields 共Fermi–Dirac for Dirac
fields兲 is usually derived by considering the expectation
value of the number operator â R† â R for the accelerated ob-
server 共in the right Rindler wedge兲 in the unaccelerated
Minkowski vacuum 兩 0 M 典 , that is, 具 0 M 兩 â R† â R 兩 0 M 典
⬃ 关 exp(ប⍀/kT)⫾1兴⫺1 共with the upper sign for scalar fields
and lower sign for Dirac fields兲. The proportionality of the
particle number spectrum detected by the accelerated Rindler
observer moving through the Minkowski vacuum to a ther-
mal spectrum is referred to as the thermalization theorem by
Takagi.5
In this work we have taken a slightly different
viewpoint.24 For a scalar field, we first consider an unaccel-
erated Minkowski observer in a thermal state and find that
the expectation value of a field correlation function is pro-
1527 Am. J. Phys., Vol. 72, No. 12, December 2004
portional to the Bose–Einstein distribution. We then consider
the calculation of this correlation function again, but this
time for an accelerated observer in his particle free Rindler
vacuum25 state 兩 0 R 典 , such that for a single mode,
具 0 R 兩 â R â R† 兩 0 R 典 ⫽1. The new feature is that from his local sta-
tionary perspective, the accelerated observer detects all
Minkowski frequencies 共arising from the the usual plane
waves associated with Minkowski states兲 as time-dependent
Doppler shifted frequencies.
The derivation presented here shows why quantum field
effect of acceleration: 具 ĝ † (⍀)ĝ(⍀ ⬘ ) 典 is nonvanishing be-
cause the vacuum expectation 具 â K â K† 典 ⫽0. But there is more
to it than that, because 具 â K â K† 典 also is nonvanishing for an
observer with a⫽0. For such an observer, however,
共19兲
冕 ⫺⬁
⬁
d ␶ e i⍀ ␶ e i ⬘ K ⬘
R␶ d ␶ ␻ ( ␶ )
⫽ 冕⫺⬁
⬁
d ␶ e i(⍀⫹ ␻ K ) ␶
⫽2 ␲ ␦ 共 ⍀⫹ ␻ K 兲
⫽0 共20兲
for scalar particles, because both ⍀ and ␻ K are positive. In
other words, the thermal effect of acceleration in our model
arises because of the nontrivial nature of the quantum
vacuum and the time-dependent Doppler shifts detected by
the accelerated observer. For Dirac particles, the essential
new feature is the additional spinor structure of the wave
function over that of the scalar plane wave. To keep the spin
nonrotating in the comoving frame of the accelerated ob-
server, the Dirac bispinor must be Fermi–Walker transported
along the accelerated trajectory, resulting in an additional
time-dependent Lorentz transformation. Formally, this trans-
formation induces a shifting of i⍀c/a→i⍀c/a⫹1/2 in the
calculation of relevant gamma function-like integrals, lead-
ing to the FD Planck factor.
In the following we briefly discuss the relationship of our
correlation function to those used in the usual literature on
this subject and point out a not widely appreciated subtlety
relating details of the spatial Rindler mode functions 共which
we have ignored in our model兲 to the statistics of the noise
spectrum detected by the accelerated observer.
In our model, we have not motivated the use of the corre-
lation function 具 ĝ † (⍀)ĝ(⍀ ⬘ ) 典 aside from the fact that we
could calculate it for a nonaccelerated observer in a thermal
field and for a uniformly accelerated observer in vacuum and
compare the results. It is easy to show that a harmonic oscil-
lator with frequency ␻ 0 and dissipation coefficient ␥, lin-
early coupled to the field Eq. 共10兲, reaches a steady-state
energy expectation value
具E典⬀ 冕 冕⬁
0
d⍀
0
⬁
d⍀ ⬘
具 ĝ † 共 ⍀ 兲 ĝ 共 ⍀ ⬘ 兲 典
共 ⍀⫺ ␻ 0 ⫺i ␥ 兲共 ⍀ ⬘ ⫺ ␻ 0 ⫹i ␥ 兲
, 共21兲
which offers some motivation for considering
具 ĝ † (⍀)ĝ(⍀ ⬘ ) 典 . In fact, it can be shown that 具 E 典
⫽ 关 e ប ␻ 0 /kT ⫺1 兴 ⫺1 , which shows again that our accelerated
observer acquires the characteristics appropriate to being in a
thermal field at the temperature T⫽បa/2␲ kc.
In an extensive review of the Davies–Unruh effect,
Takagi5 utilizes the quantum two-point correlation 共Wight-
Paul M. Alsing and Peter W. Milonni 1527
man兲 function g W ( ␶ , ␶ ⬘ )⬅ 具 0 M 兩 ␾ ˆ † ( ␶ ⬘ ) 兩 0 M 典 to deter-
ˆ (␶)␾ able by utilizing the expression a⫽GM /r 2 for the gravita-
mine the power spectrum of the vacuum noise detected by tional acceleration of a test mass at a distance r from a mass
the accelerated observer for a scalar field, M , and determine T at the surface of the earth, the Sun, and
a Schwarzschild black hole.
S 共 ⍀ 兲 ⬅lim
s↓0
冕
⬁
⬁
e ⫺i⍀ ␶ ⫺s 兩 ␶ 兩 g W 共 ␶ 兲 , 共22兲
ACKNOWLEDGMENTS
which is very much in the spirit of our calculation in Sec. IV. Thanks go to H. Fearn, D. McMahon, G. J. Milburn E.
Here the field operator ␾ ˆ ( ␶ ) is expanded in terms of the Mottola, M. O. Scully, and M. Wolinsky for brief discussions
Rindler mode functions and involves the creation and anni- relating to acceleration in vacuum.
hilation operators for both the right and left Rindler wedges.
Takagi shows the remarkable, although not widely known
result, that for a scalar field in a Rindler space–time of di- APPENDIX: MODE SUM CALCULATION IN EQ.
mension n, S n (⍀)⬃ f n (⍀)/ 关 exp(ប⍀/kT)⫺(⫺1)n兴. For even- „15…
dimensional space–times 共for example, n⫽2 as considered
in this work or the usual n⫽4) S n (⍀) is proportional to the If we convert the sum over K to an integral, we have, for
Bose–Einstein distribution function 关 exp(ប⍀/kT)⫺1兴⫺1, and K⬎0,
essentially reproduces our Eq. 共16兲 关up to powers of ⍀c/a,
contained in the function f n (⍀)]. However, for odd n,
S n (⍀) is proportional to the Fermi–Dirac distribution
兺
K⬎0 ␻ K
1
冉 冊
␻ Kc
a
i ⑀ K (⍀⫺⍀ ⬘ )c/a
⫽
V
2␲
冕0
⬁
dK
1
␻K
关 exp(ប⍀/kT)⫹1兴⫺1. For Dirac particles the opposite is true,
namely for even space–time dimensions S n (⍀) is propor-
tional to the FD distribution and for odd space–time dimen-
⫻ 冉 冊
␻ Kc
a
i(⍀⫺⍀ ⬘ )c/a
. 共A1兲

sions S n (⍀) is proportional to the BE distribution. This cu- We let x⫽log(␻Kc/a) and write Eq. 共A1兲 as

冕
rious fact arises from the dependence of S n (⍀) on two V ⬁ Va
factors in its calculation. The first is the previously men- dx e ⫺ix(⍀⫺⍀ ⬘ )c/a ⫽ ␦ 共 ⍀⫺⍀ ⬘ 兲 . 共A2兲
tioned thermalization theorem, that is, the number spectrum 2␲c ⫺⬁ c2
of accelerated 共Rindler兲 particles in the usual nonaccelerated The same result is obtained for the sum over K⬍0, so that
Minkowski vacuum is proportional to the BE distribution
the sum over all K is (2Va/c 2 ) ␦ (⍀⫺⍀ ⬘ ).
function for scalar fields and is proportional to the FD dis-
tribution function for Dirac fields. The second factor that a兲
Electronic mail: alsing@hpc.unm.edu
switches the form of S n (⍀) from BE to FD depends on the b兲
Electronic mail: pwm@lanl.gov
detailed form of the Rindler mode functions.5 Although the 1
S. W. Hawking, ‘‘Black hole explosions,’’ Nature 共London兲 248, 30–31
trajectory of the accelerated observer takes place in 1⫹1 共1974兲; ‘‘Particle creation by black holes,’’ Commun. Math. Phys. 43,
199–220 共1975兲.
dimensions 关the (z,t) plane兴, the quantum field exists in the 2
W. G. Unruh, ‘‘Notes on black hole evaporation,’’ Phys. Rev. D 14, 870–
full n-dimensional space–time, and thus S n (⍀) ultimately 892 共1976兲.
depends on the form of the mode functions in the full space– 3
P. C. W. Davies, ‘‘Scalar production in Schwarzschild and Rindler met-
time. In space–times of even dimensions the number spec- rics,’’ J. Phys. A 8, 609– 616 共1975兲.
4
trum of Rindler particles in the Minkowski vacuum and the The literature on this subject is vast. For some articles directly relevant for
noise spectrum of the vacuum fluctuations 共that is, the re- the work presented here see, for instance, P. Candelas and D. W. Sciama,
‘‘Irreversible thermodynamics of black holes,’’ Phys. Rev. Lett. 38, 1372–
sponse of the accelerated particle detector兲 both depend on 1375 共1977兲; T. H. Boyer, ‘‘Thermal effects of acceleration through ran-
the same distribution function, and these two effects often dom classical radiation,’’ Phys. Rev. D 21, 2137–2148 共1980兲 and ‘‘Ther-
are incorrectly equated. mal effects of acceleration for a classical dipole oscillator in classical
In our simplified derivation we have bypassed this techni- electromagnetic zero-point radiation,’’ 29, 1089–1095 共1984兲; D. W.
cality by performing our calculations in 1⫹1 dimensions Sciama, P. Candelas, and D. Deutsch, ‘‘Quantum field theory, horizons,
and thermodynamics,’’ Adv. Phys. 30, 327–366 共1981兲.
共that is, n⫽2). We have concentrated on the power spectrum 5
An extensive review is given by S. Takagi, ‘‘Vacuum noise and stress
of vacuum fluctuations as seen by a particle detector carried induced by uniform acceleration,’’ Prog. Theor. Phys. 88, 1–142 共1986兲.
by the accelerated observer. We have shown that in 1⫹1 See Chap. 2 for a review of the Davies–Unruh effect.
dimensions the spectrum of fluctuations is proportional to the 6
N. D. Birrell and P. C. W. Davies, Quantum Fields in Curved Space 共Cam-
Bose–Einstein distribution function for scalar fields and to bridge U.P., New York, 1982兲.
7
Note that in order for a detector to remain at a fixed location outside the
the Fermi–Dirac distribution for spin 1/2 fields, with the
horizon of a black hole, it must undergo constant acceleration just to
Hawking–Unruh temperature defined by Eq. 共1兲. The depen- remain in place.
dence of the noise spectrum on these distribution functions is 8
W. Rindler, ‘‘Kruskal space and the uniformly accelerated observer,’’ Am.
ultimately traced back to the time-dependent Doppler shifts J. Phys. 34, 1174 –1178 共1966兲.
as detected by the accelerated observer as he/she moves 9
P. W. Milonni, The Quantum Vacuum 共Academic, New York, 1994兲, pp.
through the usual nonaccelerated Minkowski vacuum. 60– 64.
10
We hope that the calculations exhibited here are suffi- I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products
共Academic, New York, 1980兲. We have used integrals 3.761.4 and 3.761.9,
ciently straightforward to give an intuitive understanding of 兰 ⬁0 x ␮ ⫺1 sin(ax)dx⫽关⌫(␮)/a␮兴sin(␮␲/2) and 兰 ⬁0 x ␮ ⫺1 cos(ax)dx
the essential physical origin of the Hawking–Unruh tempera- ⫽关⌫(␮)/a␮兴cos(␮␲/2) respectively, in the combination of the second plus i
ture experienced by a uniformly accelerated observer. times the first. Taken together, both integrals have a domain of definition
Suggested Problem. Discuss when the Hawking–Unruh a⬎0, 0⬍Re(␮)⬍1. In Eq. 共8兲 we have ␮ ⫽i⍀c/a with Re(␮)⫽0. The
temperature from Eq. 共1兲 would become physically detect- integrals can be regularized and thus remain valid in the limit Re(␮)→0 as

1528 Am. J. Phys., Vol. 72, No. 12, December 2004 Paul M. Alsing and Peter W. Milonni 1528
 can be seen by adding a small imaginary part ⫺i ⑀ a/c, ⑀ ⬎0 to the fre-
quency ⍀ so that ␮ → ␮ ⬘ ⬅ ␮ ⫹ ⑀ and 0⬍Re(␮⬘)⫽⑀⬍1 is strictly in the
domain of definition of the integrals. In the limit of ⑀ →0 we have 关 ⌫( ␮
⫹ ⑀ )/a ␮ ⫹ ⑀ 兴 e i( ␮ ⫹ ⑀ ) ␲ /2→ 关 ⌫( ␮ )/a ␮ 兴 e i ␮ ␲ /2 and thus obtain the same values
for the integrals as if we had just set Re(␮⬘)⫽0 initially. The equivalence
of these two approaches to evaluate the integrals occurs because the stan-
dard integral form of the gamma function ⌫(z)⫽ 兰 ⬁0 dt e ⫺z t z⫺1 , Re(z)⬎0
can be analytically continued in the complex plane and in fact remains
well defined, in particular, for Re(z)→0, Im(z)⫽0, which is the case in Eq.
共8兲. See for for example, J. T. Cushing, Applied Analytical Mathematics
for Physical Scientists 共Wiley, New York, 1975兲, p. 343.
11
Reference 10, Sec. 8.332.
12
Related, although much more involved derivations of the Davies–Unruh
effect based on a similar substitution can be found in L. Pringle, ‘‘Rindler
observers, correlated states, boundary conditions, and the meaning of the
thermal spectrum,’’ Phys. Rev. D 39, 2178 –2186 共1989兲; U. H. Gerlach,
‘‘Minkowski Bessel modes,’’ ibid. 38, 514 –521 共1988兲, gr-qc/9910097
and ‘‘Quantum states of a field partitioned by an accelerated frame,’’ 40,
1037–1047 共1989兲. We can justify this substitution from the form ␸
⫽ 兰 k ␮ (x)dx ␮ of the phase of a quantum mechanical particle in curved
space–time. See L. Stodolsky ‘‘Matter and light wave interferometry in
gravitational fields,’’ Gen. Relativ. Gravit. 11, 391– 405 共1979兲 and P. M.
Alsing, J. C. Evans, and K. K. Nandi, ‘‘The phase of a quantum mechani-
cal particle in curved spacetime,’’ ibid. 33, 1459–1487 共2001兲, gr-qc/
0010065.
13
A similar derivation in terms of Doppler shifts appears in the appendix of
H. Kolbenstvedt, ‘‘The principle of equivalence and quantum detectors,’’
Eur. J. Phys. 12, 119–121 共1991兲. T. Padmanabhan and coauthors also
have derived Eq. 共9兲 by similarly considering the power spectrum of Dop-
pler shifted plane waves as detected by the accelerated observer. See K.
Srinivasan, L. Sriramkumar, and T. Padmanabhan, ‘‘Plane waves viewed
from an accelerated frame: Quantum physics in a classical setting,’’ Phys.
Rev. D 56, 6692– 6694 共1997兲; T. Padmanabhan, ‘‘Gravity and the ther-
modynamics of horizons,’’ gr-qc/0311036.
14
When we later convert a sum over K to an integral, we obtain the Lorentz-
invariant measure dK/ ␻ K as a consequence of the 1/冑␻ K in Eq. 共10兲. The
use of this invariant measure eliminates the need to explicitly transform
the frequency term 1/冑␻ K in Eq. 共14兲, for instance.
15
Because we use a one-dimensional model, the factor V/2␲ appears instead
of the more familiar V/(2 ␲ ) 3 . In other works, our volume V here is really
just a length.
16
If we use the same gamma function integrals as in Ref. 10, we will have
for the Dirac case ␮ ⫽i⍀c/a⫹1/2, with Re(␮)⫽1/2 clearly in their do-
main of definition 0⬍Re(␮)⬍1.
17
S. Weinberg, Gravitation and Cosmology 共Wiley, New York, 1972兲, pp.
365–370.
18
C. W. Misner, K. S. Thorne, and J. A. Wheeler, Gravitation 共Freeman, San
Francisco, 1973兲, Chap. 6, pp. 163–176.
19
Reference 5, p. 101, and P. Candelas and D. Deutsch, ‘‘Fermion fields in
accelerated states,’’ Proc. R. Soc. London, Ser. A 362, 251–262 共1978兲.
1529 Am. J. Phys., Vol. 72, No. 12, December 2004
This result can be understood as follows. If the observer was traveling at
constant velocity v in the positive z direction, we would Lorentz transform
the spinor in the usual special relativistic way via the operator Ŝ( v )
⫽exp(␥0␥3␰/2) where tanh ␰⫽v/c. See J. D. Bjorkin and S. D. Drell, Rela-
tivistic Quantum Mechanics 共McGraw–Hill, New York, 1964兲, pp. 28 –30.
From Eq. 共3兲 we have v /c⫽tanh(a␶/c) so that ␰ ⫽a ␶ /c, yielding the
spinor Lorentz transformation to the instantaneous rest frame of the accel-
erated observer.
20
For simplicity, we have chosen the spin up wave function as an eigenstate
of Ŝ( ␶ ) with eigenvalue exp(a␶/2c). The exact spatial dependence of the
accelerated 共Rindler兲 spin up wave function is more complicated than this
simple form, although both have the same zero bispinor components. See
W. Greiner, B. Müller, and J. Rafelski, Quantum Electrodynamics in
Strong Fields 共Springer, New York, 1985兲, Chap. 21.3, pp. 563–567; M.
Soffel, B. Müller, and W. Greiner, ‘‘Dirac particles in Rindler spacetime,’’
Phys. Rev. D 22, 1935–1937 共1980兲.
21
The spinor Lorentz transformation Ŝ( ␶ ) does not mix spin components.
Thus, for example, a spin up Minkowski state remains a spin up acceler-
ated 共Rindler兲 state. We can therefore drop the constant spinor 兩↑典 from our
calculations and retain the essential, new time-dependent modification
exp(a␶/2c) to the plane wave for our Dirac ‘‘wave function.’’
22
Reference 5, Sec. 2, in particular Eqs. 共2.7.4兲 and 共2.8.8兲.
23
S. A. Fulling, ‘‘Nonuniqueness of canonical field quantization in Riemann-
ian spacetime,’’ Phys. Rev. D 7, 2850–2862 共1973兲 and Aspects of Quan-
tum Field Theory in Curved Space-Time 共Cambridge U.P., New York,
1989兲.
24
This viewpoint is also taken in a different derivation of the Unruh–Davies
effect in Ref. 9.
25
The unaccelerated Minkowski vacuum 兩 0 M 典 is unitarily related to the Rin-
dler vacuum 兩 0 R 典 丢 兩 0 L 典 via 兩 0 M 典 ⫽Ŝ(r) 兩 0 R 典 丢 兩 0 L 典 , where Ŝ(r) is the
squeezing operator Ŝ(r)⫽exp关r(âRâL⫺â†RâL† )兴. The subscripts R and L de-
note the right (z⬎0,z⬎ 兩 t 兩 ) and left (z⬍0,兩 z 兩 ⬎ 兩 t 兩 ) Rindler wedges, re-
spectively, which are regions of Minkowski space–time bounded by the
asymptotes t⫽⫾z. 兩 0 R 典 is the Fock state of zero particles in the right
Rindler wedge and 兩 0 L 典 is the Fock state of zero particles in the left
Rindler wedge. Note that the orbit of the accelerated Rindler observer
given by Eq. 共4兲 is confined to the right Rindler wedge. Because the right
and left Rindler wedges of Minkowski space–time are causally discon-
nected from each other, the creation and annihilation operators â †R , â R and
â L† , â L live in the right and left wedges, respectively, and mutually com-
mute with each other, that is, 关 â R ,â L† 兴 ⫽0, etc. Because, physical states
that live in the right wedge have zero support in the left wedge 共and vice
versa兲, they are described by functions solely of the operators â †R , â R
appropriate for the right wedge, that is, 兩 ⌿ R 典 ⫽ f (â R ,â †R ) 兩 0 R 典 丢 兩 0 L 典
⫽ 兩 ␺ R 典 丢 兩 0 L 典 . It is in this sense that we can speak of 兩 0 R 典 as the vacuum
for the right Rindler wedge, and similarly 兩 0 L 典 as the vacuum for the left
Rindler wedge.
Paul M. Alsing and Peter W. Milonni 1529