Professional Documents
Culture Documents
Compton Scattering On Blackbody Photons: Lowell S. Brown Ronald S. Steinke
Compton Scattering On Blackbody Photons: Lowell S. Brown Ronald S. Steinke
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
Compton scattering on blackbody photons
Lowell S. Brown and Ronald S. Steinke
Department of Physics, University of Washington, Seattle, Washington 98195
~Received 14 June 1996; accepted 8 October 1996!
We examine Compton scattering of electrons on blackbody photons in the case where the electrons
are highly relativistic, but the center of mass energy is small in comparison with the electron mass.
We derive the partial lifetime of electrons in the large electron positron ~LEP! accelerator due to this
form of scattering in the vacuum beam pipe and compare it with previous results. © 1997 American
Association of Physics Teachers.
304 Am. J. Phys. 65 ~4!, April 1997 © 1997 American Association of Physics Teachers 304
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
In Sec. II we use simple relativistic techniques to compute In the nonrelativistic limit, s may be replaced with the
the total cross section for the scattering of an ultrarelativistic Thomson cross section, s T 58 p r 20 /3, where r 0 5e 2 /4p m is
electron on the Planck distribution of blackbody photons. the classical electron radius. Since this is independent of k,
Section III describes the more detailed calculation needed for the scattering rate may be rewritten as
the cross section in which the electron loses an energy
greater than DE. If the energy loss DE in an electron– Ḡ0 5 s T j̄ 0 , ~2.6!
photon collision is too large, the electron’s motion falls out- where
side of the acceptance parameters of the machine. At LEP
this happens when DE/E is greater than about 1%. As we
shall see, this means that even if the beam were in a perfect
vacuum, it would decay with a half life of about two days.
j̄ m 5 E ~ d 3 k̄ ! k̄ m
~ 2 p ! 3 k̄ 0
f̄ ~ k̄! ~2.7!
The vacuum in the LEP accelerator is so good that the beam is the photon number flux four vector. Since p̄ m /m5(1,0) in
scattering of the blackbody photons is, in fact, the primary the electron’s rest frame, we may write this leading approxi-
mechanism for beam loss when the machine is run with a mation, denoted with a 0 subscript, as
S D
single beam. Scattering on the residual gas in the beam pipe
gives a considerably longer half life of about 6 days.8 When dn p̄ m
52 s T j̄ m , ~2.8!
the machine is run in the usual mode with two beams for dt 0
m
e 1 e 2 collision experiments, beam–beam collisions reduce
the beam half life to about 14 h.8 with the minus sign arising from our Lorentz metric conven-
tion in which the metric has the signature (2,1,1,1). The
result ~2.8! is now in an invariant form which holds in any
II. TOTAL SCATTERING RATE frame. With a thermal photon distribution in the lab frame,
In the general case of an electron scattering off some pho- 2
f ~ k !5 ,
v /T ~2.9!
ton distribution, the scattering rate Ḡ in the electron’s rest e 21
frame may be computed using the formula
where v 5k 0 is the photon energy, the photon number dis-
E ~ d k̄ !
3 tribution is isotropic, and so only the number density com-
Ḡ5 f̄ ~ k̄! s ~ k̄! , ~2.1! ponent j 0 is nonvanishing. Thus, in the lab frame,
~ 2p !3
m p0
where f̄ (k̄) is the photon phase-space density @with the nor- G 05 s T j 0 5 s T j 0. ~2.10!
malization defined such that j̄ 0 in Eq. ~2.7! is the photon E m
number density# as a function of the photon momentum and The lab photon number density obtained from integrating
s (k̄) is the scattering cross section, which is similarly a ~2.7! with the distribution ~2.9! is the familiar result
function of the photon momentum. Here all quantities are 2z~ 3 ! 3
evaluated in the electron’s rest frame as indicated by the j 05 T , ~2.11!
overbar. This scattering rate may be viewed as a time deriva- p2
tive in which z (3)51.202 . . . is the Riemann zeta function.
dn Thus, G 0 is given by
Ḡ5 , ~2.2!
dt 2z~ 3 ! 3
G 05 T sT . ~2.12!
where t is the time in the electron’s rest frame. Since num- p2
bers are Lorentz invariant and t may be defined to be the The first-order relativistic correction to this result is ob-
invariant proper time of the electron, the rate dn/d t is, in tained with the use of the corrected cross section
S D
fact, a Lorentz invariant. Thus, if the integral on the right-
hand side of the rate formula ~2.1! is written in a Lorentz 2 pk
invariant manner, we can immediately evaluate the rate in s 5 s T 11 . ~2.13!
m2
the laboratory frame. In the lab frame, the electron moves
with four momenta Note that the product kp5k m p m of the two four-momenta is
negative with our metric. Because s is no longer indepen-
dz m
p m 5m , ~2.3! dent of k, the corresponding form of Eq. ~2.10! is
dt
where z m ( t ) is the world line of this particle, its space-time
position as a function of proper time, and m is the electron
G 15
m
s
E T
E ~ d 3 k ! 2kp
~ 2 p !3 k 0m
S
2 pk
f ~ k ! 11 2
m D
mass. This gives the familiar time-dilation formula
dt dz p E 0 0
52
1
E S m
2
s T j m p m1 2 T mn p m p n , D ~2.14!
5 5 5 , ~2.4!
dt dt m m where
where E is the electron’s total energy. Thus the scattering
rate G in the lab frame may be easily evaluated using T mn5 E ~ d 3k ! k mk n
~ 2p !3 k0
f ~k! ~2.15!
305 Am. J. Phys., Vol. 65, No. 4, April 1997 L. S. Brown and R. S. Steinke 305
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
diagonal components, and it is also traceless because the 2
photon is massless, k m k m 50. Therefore, in the lab frame, f ~ k !5 2bmk
. ~3.2!
m 21
S D S D
e
1 4 m 2 00
T m n p m p n 5 E 2 1 u p2 u T 005 E 2 12 T . From the definition of b m , 2 b m b m 51/T 2 and
3 3 4E 2
2 b m p m 5E/T, because multiplication by b m selects out the
~2.16!
time component in the lab frame. In the electron rest frame,
2 2
The m /E term is very small, and it may be neglected. b m therefore takes on the value
S D
Integrating over the photon distribution in Eq. ~2.15! gives
the well-known blackbody energy density E p
b m5 ,2 , ~3.3!
Tm Tm
6z~ 4 ! 4
T 005 T , ~2.17!
p2 where E and p are taken in the lab frame. We have not yet
taken into account the lower bound on the electron energy
where z (4)5 p 4 /9051.082•••. This yields the corrected loss in the lab frame. Because multiplication by b m selects
scattering rate the time component in the lab frame, this limit may be insti-
G 15
2z~ 3 ! 3
p 2 T s T 124s F
z~ 4 !
z~ 3 !
. G ~2.18!
tuted by the inclusion of an ‘‘energy loss’’ step function in
the integrand of Eq. ~3.1!,
plitude u T u 2 . The total electron scattering rate as observed in We do the k8 integral in spherical coordinates and take the
the lab frame reads z axis to be parallel to k, with u the angle between these two
G5
1
2E
E ~ d 3k ! 1
~ 2p !3 2v
f ~k! E ~ d 3k 8 ! 1
~ 2p !3 2v8
vectors. The angle u is the photon scattering angle in the
electron rest frame, and
2kk 8 5 vv 8 ~ 12cos u ! . ~3.7!
3 E ~ d3p8! 1
~ 2 p ! 3 2E 8
~ 2 p ! 4 d ~ 4 ! ~ k 8 1p 8 2k2 p ! u T u 2 , The d function can now be solved for v 8 to yield
~3.1! d ~ 22m v 12m v 8 22kk 8 !
where p and p 8 are the initial and final electron four mo-
menta, k and k 8 the initial and final photon four momenta,
with E5p 0 , E 8 5 p 8 0 , v 5k 0 , and v 8 5k 8 0 the time com-
5
1
2 @ m1 v ~ 12cos u !#
d v 82
mv
S
m1 v ~ 12cos u !
, D
ponents of these four vectors. Except for the initial factor of ~3.8!
1/2E which is the lab energy of the initial electron and which which requires that v 8 , v and thus makes the u ( v 2 v 8 )
converts the invariant proper time into the lab time, the right- step function redundant. The scattered photon energy v 8
hand side of this expression is a Lorentz invariant. The prob- given by the d function is, of course, just the Compton en-
lem proves to be greatly simplified if the integrals are evalu- ergy. To deal with the ‘‘energy loss’’ step function, we note
ated in the rest frame of the electron, because Compton that the Lorentz transformation from the lab frame to the
scattering of a photon on an electron at rest has a very simple initial electron rest frame turns the lab frame isotropic black-
nonrelativistic limit. This complicates the initial photon dis- body photon distribution into a very narrow pencil in the
tribution, but, if we introduce a four-vector b m , whose time electron rest frame in which we are now working. Thus the
component in the lab frame is one over the temperature of initial photon distribution is sharply peaked about the aver-
the photon distribution and whose spatial components are age value
zero in the lab frame, the distribution in an arbitrary frame
still has the simple form ~ k m/ v !5 b m/ b 0. ~3.9!
306 Am. J. Phys., Vol. 65, No. 4, April 1997 L. S. Brown and R. S. Steinke 306
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
Hence, we can approximate vb2 vm
2k b . v ~ b 0 2 u bu ! 5 . . ~3.20!
b0 b 1 u bu 2TE
0
2 b k 8 . ~ 2kk 8 ! 5 b 0 v 8 ~ 12cos u ! . ~3.10!
v Thus, with the neglect of order m 2 /E 2 corrections, and re-
To verify this and assess the order of accuracy, we define the membering that u bu .E/mT,
E E
average more precisely by 1 mT 2 v E/mT
* ~ d k ! / ~ 2 p ! ~ 1/2 v ! f ~ 2k b ! X
3 3 d cos x 5 d ~ 2k b ! . ~3.21!
21 vE v m/2TE
^X&5 . ~3.11!
* ~ d 3 k ! / ~ 2 p ! 3 ~ 1/2 v ! f ~ 2k b !
We shall do the 2k b integral last, due to its dependence on
Then, by virtue of the relativistic invariance of this defini- the initial photon distribution. In order to interchange the
tion, order of the v and 2k b integrations, we note that the lower
^ k m k n & 5 ~ b m b n 2 41 b 2 g m n ! A ~ b 2 ! , ~3.12! limit on 2k b ,
l
since b is the only four vector available and k k m 50. Re- m vm
2k b . , ~3.22!
membering that v 5k 0 , this presents the squared fluctuation 2TE
about the average as gives the upper bound on v ,
KS v
DS v
k m2 b m 0 k n2 b n 0
b b DL v,
2TE ~ 2k b !
m
5s ~ 2k b ! m. ~3.23!
^ v 2&
The upper limit on 2k b ,
S bn
b
bm
b
b mb n
5B g m n 2 0 g m 0 2 0 g n 0 1 0 2 ,
~b !
D ~3.13!
2k b ,
2vE
mT
, ~3.24!
where
gives the lower bound on v ,
2b2 m2
B5 2. . ~3.14! ~ 2k b ! mT m2
4~ b ! 1b
0 2
4E 2 v. 5s ~ 2k b ! m . ~3.25!
2E 4E 2
Thus the deviations away from our approximation may be
neglected because they involve the very small quantity In view of the extreme smallness of m 2 /E 2 , we may replace
m 2 /E 2 . Using this approximation for 2k 8 b simplifies the this lower limit by v 50. Hence, switching the order of in-
‘‘energy loss’’ step function to tegration gives
S
u v 82
DE
b T ~ 12cos u !
0 5 u v 82D S
DE m
E 12cos u
, D E vE
0
`
d
2 v E/mT
v m/2TE
d ~ 2k b ! 5 E0
`
d ~ 2k b ! E
0
s ~ 2k b ! m
dv.
~3.15! ~3.26!
where the 2 b k in the original step function has been ne- We perform this reversal of integrals, do the two trivial azi-
glected because it is much less than DE/T. Inserting the muthal integrals, and do the v 8 integrals using the d function
value of v 8 given by the energy-conserving d function ~3.8! to obtain
E
into the step function gives m 2T `
S D
G ~ DE ! 5 d ~ 2k b ! f ~ 2k b !
mv DE m 16E 2 ~ 2 p ! 3 0
u 2
m1 v ~ 12cos u ! E 12cos u
E 1
E s ~ 2k b ! m
5u v2 S DE m
E 8 12cos u
, D ~3.16!
3
21
d cos u
v
0
dv
where on the right-hand side we have solved for v and de- 3 uTu2
@ m1 v ~ 12cos u !# 2
fined
E 8 5E2DE,
which is the maximum final electron energy in the lab frame.
~3.17! 3u v2 S DE m
E 8 12cos u
. D ~3.27!
We perform the k integral in spherical coordinates, with To work out the integrals which appear here, it is conve-
the polar angle x taken to be the angle between k and b, so nient to first introduce the appropriate, dimensionless vari-
that ables,
2k b 5 v ~ b 0 2 u bu cos x ! . ~3.18! v
x52k b , z512cos u , n5 , ~3.28!
We rewrite the angular integral for k in terms of an integra- s ~ 2k b ! m
tion over k b by noting the limits and define
2vE DE
2k b , v ~ b 0 1 u bu ! .2 v b 0 5 , ~3.19! u5 . ~3.29!
mT 2sE 8
and With this new notation, we have
307 Am. J. Phys., Vol. 65, No. 4, April 1997 L. S. Brown and R. S. Steinke 307
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
G ~ DE ! 5
T3
4m 2 ~ 2 p ! 3
E
0
`
dxx 2 f ~ x ! E
0
2
dz
3 E 0
1
dn
n
~ 11s n xz ! 2 uTu u n2
2
2u
xz
. S D ~3.30!
G ~ DE ! 5
T3
4m ~ 2 p ! 3
2 E
u
`
dxx 2 f ~ x ! E
2u/x
2
dz
E
Fig. 1. The dimensionless F 0 (u) defined in Eq. ~3.35! as a function of the
1 n dimensionless variable u defined in Eq. ~3.29!.
3 dn uTu2. ~3.31!
2u/xz ~ 11s n xz ! 2
This will be evaluated in the nonrelativistic limit, keeping
first-order corrections in s. The exact squared amplitude dif-
fers from its nonrelativistic limit
^ E2E 8 & 5 E 0
`
du
G ~ DE ! dDE
G1 du
, ~3.38!
F
G ~ DE ! 5G 0 F 0 ~ u ! 24s
z~ 4 !
F ~u! ,
z~ 3 ! 1 G ~3.34! 2
z~ 4 !
z~ 3 !
F 0~ u ! 1
z~ 4 !
G
F ~ u ! 1O ~ s 2 ! .
z~ 3 ! 1 J
where G 0 is the approximate total scattering rate from Eq. ~3.40!
~2.12!. The straightforward integrations give Inserting the expressions for the F’s from Eq. ~3.35! and Eq.
F 0~ u ! 5
1
2z~ 3 !
E ` dx
e x 21
x 2
S
23ux13u 2
ln
x
u
1
2u 3
x
, SD D ~3.36! and interchanging the order of the x and u integrals,
the integrals may be evaluated analytically, and we find that
H F G J
u
~3.35! z~ 4 ! z ~ 4 ! 63 z ~ 5 !
^ E2E 8 & 52sE 11s 4 2 1O ~ s 2 !
and 10 z~ 3 ! z~ 3 ! 5 z~ 4 !
F 1~ u ! 5
1
6z~ 4 !
E `
u
dx
e 21
x
9
S
x 3 2 u 2 x2u 3
2
51.80sE @ 128.5s # .
To check that no mistakes have been made in our calcu-
~3.41!
SD D
lation of F 0 (u) and F 1 (u) given in Eq. ~3.35! and Eq. ~3.36!,
x 9u 4 we have independently evaluated the average energy loss
16u 3 ln 1 . ~3.36!
u 2x ^ E2E 8 & starting from Eq. ~3.6! and only make the small s
It can be seen that in the DE→0 limit, F 0 (0)5F 1 (0)51, so approximation toward the end of the calculation. We find the
same result with this different method.
G(DE) reduces to the result ~2.18! for G 1 .
The integrals in the definitions ~3.35! and ~3.36! of the
At this stage, one must resort to numerical calculations to
functions F 0 (u) and F 1 (u) have been calculated numeri-
evaluate the integrals. However, analytic calculations of the
cally, and the results are displayed in Figs. 1 and 2. As a
energy weighted moments of the distribution can still be per-
check on this numerical result, we have used it to evaluate
formed. The simplest of these is the average energy loss
the integrals in Eq. ~3.40! numerically, and the results agree
observed in the lab frame, ^ E2E 8 & . Because G(DE) de-
with the analytic expression ~3.41! to within 0.2%.
scribes the rate due to all scattering events where We may compare our calculations with those of
E2E 8 .DE, this average value may be computed by Domenico6 who employed a Monte Carlo method. He used
^ E2E 8 & 52 E 0
`
dDEDE
d
dDE S
G ~ DE !
G1
, D ~3.37!
the values E546.1 GeV and T5291 K ~which we have pre-
viously employed! that give s50.0089. He also took
DE50.012 E which places u50.69. Numerical integration
where G 1 is the total scattering rate including the first cor- gives F 0 (0.69)50.44 and F 1 (0.69)50.83, and from these
rection in s given by Eq. ~2.18!. Changing variables to u and values we calculate a mean beam lifetime of 64 h to zeroth
integrating by parts gives order in the nonrelativistic limit, and of 68 h when the first-
308 Am. J. Phys., Vol. 65, No. 4, April 1997 L. S. Brown and R. S. Steinke 308
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46
Note added in proof: We should stress that our results do
agree to within 0.6% with those obtained by V. Telnov1 us-
ing Monte Carlo methods.
ACKNOWLEDGMENTS
We would like to thank J. Rothberg for making us aware
of the electron scattering on the photons in the LEP beam
pipe. We would also like to thank H. Burkhardt for inform-
ing us of relevant literature and C. Woll for checking many
of our calculations. This work was supported, in part, by the
U.S. Department of Energy under Grant. No. DE-FG03-96
ER40956.
1
V. I. Telnov, ‘‘Scattering of electrons on thermal radiation photons in
electron–positron storage rings,’’ Nucl. Instrum. Methods A 260, 304–
Fig. 2. The dimensionless F 1 (u) defined in Eq. ~3.36! as a function of u. 308 ~1987!.
2
B. Dehning, A. C. Melissinos, F. Perrone, C. Rizzo, and G. von Holtey,
‘‘Scattering of high energy electrons off thermal photons,’’ Phys. Lett. B
order relativistic corrections are included. This is to be com- 249, 145–148 ~1990!.
3
C. Bini, G. De Zorzi, G. Diambrini-Palazzi, G. Di Cosimo, A.
pared with Domenico’s value of 90 h for the same input
Di Domenico, P. Gauzzi, and D. Zanello, ‘‘Scattering of thermal photons
parameters. We do not understand the reason for this discrep- by a 46 GeV positron beam at LEP,’’ Phys. Lett. B 262, 135–138 ~1991!.
ancy. 4
C. Bini, G. De Zorzi, G. Diambrini-Palazzi, G. Di Cosimo, A.
We may also compare our results with those of Di Domenico, P. Gauzzi, and D. Zanello, ‘‘Fast measurement of luminos-
Burkhardt,7 who also used a Monte Carlo method. Burkhardt ity at LEP by detecting the single bremsstrahlung photons,’’ Nucl. In-
uses parameters that are slightly different than those used by strum. Methods A 306, 467–473 ~1991!.
5
G. R. Blumenthal and R. J. Gould, ‘‘Bremsstrahlung, synchrotron radia-
Domenico, namely, E545.6 GeV and T5295 K. These pa-
tion, and Compton scattering of high-energy electrons traversing dilute
rameters yield the same s50.0089. However, since the over- gases,’’ Rev. Mod. Phys. 42, 237–270 ~1970!.
all rate scales as T 3 , which is 4% larger with Burkhardt’s 6
A. D. Domenico, ‘‘Inverse Compton scattering of thermal radiation at LEP
temperature, our value of the beam lifetime for and LEP-200,’’ Part. Accel. 39, 137–146 ~1992!.
7
H. Burkhardt, ‘‘Monte Carlo Simulation of Beam Particles and Thermal
DE50.012 E with his parameters is reduced from 68 to
Photons,’’ CERN/SL Note 93-73 ~OP!, 1993 ~Internal Note, unpublished!.
65 h. Burkhardt finds 83 h. ~To compare with Domenico, we 8
See, for example, H. Burkhardt and R. Kleiss, ‘‘Beam lifetimes in LEP,’’
note that modifying his result of 90 h by the 4% change in Proceedings of the Fourth European Particle Acceleration Conference
T 3 produces 86 h.! Burkhardt and Kleiss8 also state that the EPAC, London 1994, edited by V. Suller and Ch. Petit-Jean-Genaz ~World
average fractional energy loss ^ E2E 8 & /E is 1.1% for this Scientific, Singapore, 1994!, Vol. II, pp. 1353–1355.
9
See, for example, L. S. Brown, Quantum Field Theory ~Cambridge U.P.,
value of s, but our result ~3.41! gives the larger value of Cambridge, 1992!, Sec. 3.4.
1.5% corresponding to our shorter beam lifetime. Again, we 10
The result for F 0 (u) given in Eq. ~3.35! can be obtained from an integra-
can only state that we do not understand the reason for these tion of Eq. ~2.42! of Blumenthal and Gould ~Ref. 5!, but the result for
discrepancies. F 1 (u) in Eq. ~3.36! appears to be new.
Edward M. Purcell, ‘‘Research in nuclear magnetism’’ ~Nobel Lecture, December 11, 1952, reprinted in Nobel Lectures,
Physics, Vol. 3, 1942–1962, Elsevier Amsterdam, 1964!.
309 Am. J. Phys., Vol. 65, No. 4, April 1997 L. S. Brown and R. S. Steinke 309
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
107.222.41.79 On: Wed, 12 Mar 2014 21:59:46