You are on page 1of 6

PHYSICAL REVIE% B VOLUME 25, NUMBER 1 1 JANUARY 1982

Fermi density effect on the stopping power of metallic aluminum

Mitio Inokuti and David Y. Smith


Argonne National Laboratory, Argonne, Illinois 60439
(Received 2 July 1981)

The density-effect correction 8 to the Bethe stopping-power formula for fast charged particles
is evaluated for metallic aluminum from the dielectric-response function e(E). The latter has
been accurately determined over the entire range of excitation energy E by Shiles, Sasaki,
Inokuti, and Smith through comprehensive analysis of all pertinent experimental data, The
resulting values of 8 (which is a function of the particle speed Pc) should be the most reliable
to date. The present result agrees well with that of Sternheimer, who used a simpler and less
rigorous procedure, and thus corroborates the general view that 5 is insensitive to fine details of
the behavior of ~(E). We also present general remarks on the evaluation of 5 and on the ana-
lytic continuation of e(E) as a function of the complex energy E.

I. INTRODUCTION given by

At higher and higher speeds of an incident charged


particle, the stopping power receives progressively
lnI = —E~ ' roo

dE E Im — l
e(E)
lnE
more important contributions from glancing collisions
at larger and larger impact parameters (more precise- where
ly, at smaller and smaller momentum transfers).
'z
E~ =t(4rrNZe /m)' 2
Unless the medium is an extremely dilute gas, the
relevant impact parameters for relativistic particles is the plasma energy corresponding to the total elec-
are so large that there are many atoms or molecules tron density NZ of the medium. For a material with
in the medium between the incident particle at its density p g/cm', atomic weight A, and atomic
closest approach and the particular atom or molecule number Z, we may write
that is excited. As Fermi showed, the dielectric po-
larization of the medium in effect screens the Ep = 28. 816(pZ/A )
charged-particle interactions and thus makes the stop-
ping power less than the value given by the Bethe where E~ is expressed in units of eV.
formula. Fano has given the most general expression
For a particle of charge ze and high speed u =Pc,
the stopping power S of a medium with atomic
number Z and atomic number density N may be ex-
5= — Ep
2
faoo
dE EIm- 1
e(E)
ln 1+ L
E
i j
pressed:as

S = (4mz'e'/m u') (I —P')


Ep
x NZ[ln(2m''/I) —ln(l —Pz) —p2 —25] . (I)
where L is a real-valued function of P, defined as
The reduction represented by 8 is called the Fermi the positive root of the equation
density correction. If one uses the (complex) dielec-
tric response function I —P e(iL) =0 (8)
e(E) = e)(E) +i e2(E)
In practice, Eq. (7) enables one to calculate 8 as a
to characterize the medium, the probability of energy function of given value of L and P2, while Eq. (8)
transfer E from fast charged partcles is proportional enables one to calculate P' as a function of given L.
to The two functional relations thus provide a
Im[ —I/e(E) ] = e2(E)/[e|(E) + e2(E) ] . (3) parametric representation of the desired relation
between 5 and P .
Then, the mean excitation energy I in Eq. (I) is It is the purpose of the present article to report the

25 61 1982 The American Physical Society


62 MITIO INOKUTI AND DAVID Y. SMITH 25

evaluation of 8 by use of a comprehensive set of II. THEORY


«(E) data in Eqs. (7) and (8). The data set stems
from critical examination5 of all pertinent experimen- A. Relation between L and P2
tal results on aluminum, followed by dispersion-
relation analyses, and perhaps represents the best As%ick"' first pointed out, it is best to treat Eq.
knowledge of «(E) currently available for any materi- (8) by expressing «(iL) in terms of an integral in-
al. Figure 1 presents an overview of the data. The volving «(E) for real values of E In .general, there
motivation for the present work is to provide a defin-
itive evaluation of 8 for one substance and to com- tation. %e follow Fano '
are various possibilities for such an integral represen-
as well as Landau and
pare the findings with earlier results based on more Lifshitz, t4 and cast Eq. (8) in the form
approximate procedures, ' which have been applied
2
dE E Im- E'+L')
to many materials. 1 1
«(iL) m "& «(E)

Io' (9)

IO
The integra1 is a monotonically decreasing function of
L 2, because Im [ — 1/«(E) ] & 0 for any real E & 0.
The quantity Im[ —1/«(E) 1 is peaked over a limited
IO
range of E, and is small for E 0 and for E
This is exemplified by the data for metallic aluminum
IO shown in Fig. 1. Therefore, the integral is suitable
for numerical evaluation. (See Sec. IIC for addition-
IO al remarks. )
Figure 2 shows the behavior of 1 —1/«(iL) as a
)
function of J. 0, for metallic aluminum. For an
electric conductor in general the behavior will be simi-
lar to what we discuss below.
IO
Recall that, for small E,
IO~2
«(E) = I +4«rilro/E (10)
where 0- is the static conductivity. ' Therefore,
IO

1 —1/«(iL) 1 —L/(4rrf cr)


IO

IO OS

IO t t t
06
IO IO IO IO IO IO lQ

PHOTON ENERGY (eV) OA

FIG. 1. Dielectric-response function of metallic alumi-


num. The real part ei(E) is negative with large absolute
values at low E, vanishes at E~ -15.0 eV (i.e., the plasma
OD
energy for conduction electrons), and is positive and ap- 0
proaches unity at higher E. The imaginary part, ~2(E), is L (eV)
largest at low E and decreases with E except at thresholds
for newer modes of excitation (first at the beginning of in- FIG. 2. Quantity 1 —1/~(iL) as a function of L for metal-
terband excitation of valence electrons, next at the L-shell lic aluminum. The horizontal axis shows L in eV. The
threshold, and finally at the K-shell threshold, on going to function begins with unity at L 0, and decreases monoton-
higher and higher E). The quantity Imf — 1/e(E) j which ically with L The initial decrease at low L is practically pro-
governs the energy transfer from fast charged particles, portional to L . The function takes the value 0.5 at about
shows a prominent maximum near E~, and is small at both L -E~ =15 eV, as one expects from the sharp peaking of
lower and higher E. Im[ —1/e(E)] at E E~.
25 FERMI DENSITY EFFECT ON THE STOPPING POWER OF. . . 63

for small L. As a consequence, one has Then, one has, on expanding in powers of L,
—I/«(iL) =1 —GL/(E")'
2 ~ dE E ) Im— 1
1

«(E)
(12) —[1 —( G/E') '] (L/E') '+ . (16)
Here the L term dominates over the L term un-
by taking the limit L 0 in Eq. (9). In other words, less L & G. The constant G is an index for damping
1 —1/«(iL) starts with unity at L =0, and monotoni- of the conduction-electron motion, and is much
cally decreases toward zero at L ~.
Thus, the smaller than the constant E', which represents the
unique root L =L (p2) of Eq. (8) exists for any p' plasma energy. Indeed, the «(E) data for metallic
(0 & p' & 1); it starts with L =0 at p'=0, and aluminum are closely approximated by Eq. (14) with
monotonically increases without bound with increas- G =0.0614 eV and E'=12.5 eV; thus, L' dominates
ing p2. over the L term unless L (0.
06 eV. The same con-
From Eqs. (9) and (ll), one has clusion may be reached by examining the contribu-
tions to the integral in Eq. (9) from low Eva-lues.
For an electric insulator, «(E) is analytic at E =0
(13)
and has a value «(0) & 1. Consequently, there is a
root L(pt) of Eq. (9) only if 1/«(0) & p2 & 1. The
for small p'. However, the domain of validity of limit L 0 in Eq. (9) gives
Eqs. (11) and (13) is extremely limited. In practice,
1 —1/«(iL) is almost proportional to L2 for moderate
L, and thus L' (not L) is almost proportional to p' at
2
dEE 'Im-
i 1 =1— 1 (17)
«(E) «(0)
moderate p'. To see this circumstance, we may use
the Drude model (applicable to a conductor and at in place of Eq. (12). The function L (p') starts with
low E) zero at p'=1/«(0), and grows monotonically and
without bound with increasing p'. The initial rise of
L is proportional to p' =—p' —I/«(0) for extremely
«(E) = I —(E") /(E y/GE) (14) low p'. However, for moderate p', L' (not L) be-
comes proportional to p'2.
where E' and G are constants with the dimension of Finally, the behavior of L (p') at extreme relativis-
energy and are related to each other through tic speeds (i.e., for p' 1) may be discussed in the
same way for both insulators and conductors. %'e
(15) may rewrite the integral in Eq. (9) as

—J dE E Im—
1
«(E)
E2+ L 2) = L 2
„dE E Im ——«(E) —L 22 — ~
dE E Im—
f
E'/(E'+L. ') .
!
(18)
I

The first term on the right-hand side is equal to to the result" that the effective oscillator strength
(Er/L)' by virtue of the weil-known sum rule. 4'"
The second term is certainly bounded by s2Es'/L4, "
density per unit range of E, which is proportional to
E Im[ —1/«(E) j, behaves as E at high E. Thus,
"at
and may be approximated by the same quantity at the integrand of Eq. (19) behaves as E high E
large L, where s2 is a quantity with the dimension of and the integral is convergent.
energy squared, defined by Using Eqs. (18) and (19), we may rewrite Eq. (9)
in the form
s2E'= — dE E'Im- 1
«(E)
(19) (Er/L ) 2 —E~~s2/L~ —1 —P2
= (21)

In view of the limiting behavior' and hence we obtain the asymptotic behavior for ex-
treme relativistic particles:
«(E) =1 —Er2/E2 as E (20)
(L/Es) =
—(1 —p ) ' —s2/Er as p 1 . (22)
it might appear as though s2 were divergent. Howev-
er, Eq. (20) neglects absorption at high E. More
realistically the energy absorption from a fast charged B. 8 as a function of L and P2
particle or from a photon at high E occurs because of
the excitation of tightly bound electrons in inner Let us consider Eq. (7), first for small values of L
atomic shells. Consideration of this absorption leads and p'. We may set ln(l + L'/E') = —L'/E' in the in-
64 MITIO INOKUTI AND DAVID Y. SMITH

gion, the other factor in the integrand E Im[ —


(
tegrand except for the region 0 & E L. In this re-
1/
(27) to evaluate 8 for liquid water from a semiempiri-
cal set of e(E) data. For a conductor, Eq. (27) is im-
a(E) ) is small [-E'/4ntoac. cording to Eq. (10)] practical because each side of this equation increases
and the contribution to the integral is negligible. without bound as p2 0 and because the accurate
Thus, we obtain with the use of Eq. (12) evaluation of the integral requires highly detailed
knowledge of e2(E) at low E.
8=(LIE )'p' (23) The extensive series of work by Sternheimer
for a conductor. Similarly, we obtain with the use of and by Sternheimer and Peierls' is in effect based on
Eq. (17) Eq. (27), and on various approximations to e(E)
starting with atomic oscillator-strength spectra. Most
8 = (L/E ) [P —I/e(0) ) —
= (L/E ) P' (24) of those approximations were necessary in the period
of the work, are sensible on physical grounds, and
for an insuiator. Recalling the low-p' behavior of L also remain appropriate for the calculation of 5 for
discussed in Sec. II A, one now sees that 5 is ex- many media, because 5 should be insensitive to fine
tremely small at low speeds (8 ~ p4 for a conductor details of the behavior of e(E), as pointed out by
and 8 ~ p'4 for an insulator). Fano4 (on p. 22 of his article).
At extreme relativistic speeds, 5 increases so much
as to modify the relativistic rise of the Bethe stopping
power appreciably. To see this quantitatively, we III. NUMERICAL WORK
may insert Eq. (22) into Eq. (7) to obtain the asymp-
totic behavior The major part of the e(E) data' consists of values
of e~(E) and e2(E) at 506 points of E ranging from
8 =in(Z, /f)'-ln(I - p') -i+(i - p')s, /E,' . (25) 0.04 eV to 10 keV. Whenever the knowledge of
e(E) outside this range is required, analytic extrapo-
lations may be used. For E (0.04 eV, Eq. (14) is
The first three terms are the same as Eq. (50) of
Fano. 4 )
adequate. For E 10 keV, Eq. (20) is useful for
c](E), and e2(E) may be approximated by a power
law as discussed in connection with Eq. (20).
C. Additional remarks on the relation The plasma energy E~ for use in Eqs. (4) and (7)
between L and P2 has the value 32.86 eV, according to Eq. (6) with the
known density p =2.6989 g/cm' for metallic alumi-
As we noted in the beginning of Sec. II A, various num, This E~ value must be sharply distinguished
integral representations of e(iL) are possible. Those from the energy E,~ = 15.0 eV, at which the plasma
integral representations are all derivable from the oscillation of conduction electrons occurs in metallic
Cauchy theorem by considering different integrands aluminum. At E =E,~, e~ vanishes, and
as discussed by Morse and Feshbach. ' For instance, Im[ —I/e(E) ) exhibits a sharp maximum as seen in
Fig. 1. The E,~ value signifies that the number of
conduction electrons participating in plasma oscilla-
— 1 t- 1— 1
E'+L') tions is 13(15.0/32. 86) =2.7 per atom. By contrast,
1 =L dERe
e(iL) 0 s(E) the E~ value corresponds to the plasma energy if all
the 13 electrons in aluminum were to participate.
=1—2 The e(E) data give I =165.7 eV (as stated in Ref.
(26)
5), and give 1.001 for the left-hand side of Eq. (12),
in excellent confirmation of the dispersion-relation
is equivalent to Eq. (9). However, the above integral analysis.
is less suitable for numerical evaluation except for For a set of chosen values of L ranging from 0.5
large L & 0, because the integrand changes sign as E eV to 100 keV, the integral in Eq. (9) was evaluated.
runs from zero to infinity. Equating the integral value to 1 —p2 for each L, we
Another alternative is to rewrite Eq. (8) as established numerically the functional relation
between L and p . Then, for each value of p, the

1 —6(/L) = —4om
dE Ee, (E)/(E'+L') = P ' —1
root L (p') of Eq. (9) was readily determined by in-
terpolation.
Finally, for each value of the chosen L, 5 was
(27) evaluated by the use of Eq. (7). Interpolation of the
results enabled us to determine 5 for any value of p'.
The integrand here, as in Eq. (9), is always positive. All the numerical results were found to be consistent
For an insulator, Eq. (27) is just as good a starting with the limiting behavior of 8, both for low p' and
point for numerical work; indeed, Ashley" used Eq. for high p', as stated in Secs. II A and IIB.
25 FERMI DENSITY EFFECT ON THE STOPPING POWER OF. . . 65

IV. RESULTS 14% at p'=0. 5 (i.e. , where the scale changes in Fig.
4) and becomes larger at lower p. Yet, this
The results of numerical work are given in Table I discrepancy is unimportant for the evaluation of the
and in Fig. 3. Figure 4 shows a comparison of the stopping power because the density correction 5 is ex-
present result with that of Sternheimer. ' The tremely small compared to the total stopping power.
discrepancy is largest at low speeds; it amounts to At higher p' the discrepancy diminishes. This is

TABLE I. Numerical results for the density correction 8 for metallic aluminum. The first
column shows values of L in units of ev chosen at the outset. The second column shows the
values of 1 —1/~(iL) evaluated by the use of Eq. (9). The third column shows the corresponding
values of P =1/~(iL). The fourth column shows Sternheimer's variable X =log~p(p/Mc)
= 2 log]of18 /(1 —P ) ], where p is the momentum and M is the rest mass of the incident particle.
Finally, the fifth column shows the values of 5 evaluated according to Eq. (7). The 5 value for an
arbitrary X value can be readily obtained by interpolation. The entry after each D signifies a
decimal exponent.

1 —1/. (IL} X

2.5D —01 9.9971D —01 2.9147D —04 —1.7676D +00 1.7692D —07
5.0D —01 9.9828D —01 1.7229D —03 —1.3815D +00 6.6998D —07
7.5D —01 9.9691D —01 3.0880D —03 —1.2545D +00 1.4567 D —06
1.0D +00 9.9559D —01 4.4097 D —03 —1.1768D +00 2.5247D —06
2.0D +00 9.9057D —01 9.4307 D —03 —1.0107D +00 9.4636D —06
5.0D +00 9.767 4D —01 2.3264D —02 —8.1155D —01 5.3976D —05
1.0D +01 9.5557D —01 4.4433 D —02 —6.6628D —01 2.0031D —04
2.0D +01 9.1710D —01 8.2899D —02 —5.2193D —01 7.3129D —04
3.5D +01 8.6602D —01 1.3398D —01 —4.0524D —01 2.0242D —03
5.0D +01 8.2099D —01 1.7901D —01 —3.3073D —01 3.7903D —03
8.0D +01 7.4471 D —01 2.5529D —01 —2.3247 D —01 8.3445D —03
1.0D +02 7.0180D —01 2.9820D —01 —1.8585D —01 1.1912D —02
1.5D +02 6.1460D —01 3.8540D —01 —1.0134D —01 2. 1910D —02
2.0D +02 5.4780D —01 4.5220D —01 —4.1642D —02 3.2673D —02
3.0D +02 4.5193D —01 5.4807D —01 4. 1886D —02 5.4562D —02
4.0D +02 3.8628D —01 6.1372D —01 1.0054D —01 7, 5669D —02
5.0D +02 3.3841D —01 6.6159D —01 1.4558D —01 9, 5511D —02
7.0D + 02 2.7312D —01 7.2688D —01 2. 1255D —01 1.3131D —01
1.0D +03 2. 1435D —01 7, 8565 D —01 2.8205D —01 1.7683 D —01
1.5D +03 1.6079D —01 8.3921D —01 3.5881D —01 2.3748 D —01
2.QD +03 1.3061D —01 8.6939D —01 4. 1162D —01 2.8582D —01
3.0D +03 9.7338D —02 9.0266D —01 4.8362D —01 3.6105D —01
4.0D +03 7.9097D —02 9.2090D —01 5.3303D —01 4. 1948D —01
5.0D +03 6.7404D —02 9.3260D —01 5.7050D —01 4.6785 D —01
7.0D +03 5.3034D —02 9.4697 D —01 6.2589D —01 5.4642D —01
1.0D +04 4. 1157D —02 9.5884D —01 6.8365D —01 6.3824D —01
2.0D +04 2.4987 D —02 9.7501D —01 7.9565D —01 8.4852D —01
5.0D +04 1.247 4D —02 9.8753D —01 9.4926D —01 1.2105D +00
1.0D +05 7.1094D —03 9.9289D —01 1.0725D + 00 1.5586D +00
2.0D +05 3.9266D —03 9.9607D —01 1.2021D + 00 1.9708D + 00
1.0D +06 9.0596D —04 9.9909D —01 1.5212D +00 3.1273D +00
2.0D +06 4.7156D —04 9.9953D —01 1.6631D +00 3.6867D + 00
5.0D +06 1.9711D —04 9.9980D —01 1.8526D +00 4.4662D + 00
1.0D +07 1.0126D —04 9.9990D —01 1.9973D +00 5.0827D +00
2.0D +07 5.1708D —05 9.9995D —01 2. 1432D + 00 5.7200D + 00
8.0D +07 1.3257D —05 9.9999D —01 2.4388D + 00 7.0424D + 00
4.0D +08 2.6797 D —06 1.0000D + 00 2.7860D +00 8.6229D + 00
2.0D +09 5.3734D —07 1.0000D +00 3.1349D + 00 1.0221D + 01
1.0D +10 1.0753D —07 1.0000D + 00 3.4842D +00 1 1822D +01
~

5.0D +10 2. 1508D —08 1.0000D + 00 3.8337D +00 1.3425D +01


66 MITIO INOKUTI AND DAVID Y. SMITH

V
1
~
$ V
I W IIJO

Ill
IS
ION

N
40
40
RO

00
RO
P
-RO
' I'

I t
oo MA)
«Q -1 0 1 I 3 4 -I 3
X X
FIG. 3. Density-effect correction 8 plotted against the FIG. 4. Comparison of the present result with earlier
momentum of an incident charged particle. The horizontal results. The vertical axis represents the deviation of
axis shows Sternheimer's variable X =log&0(p/Mc) Sternheimer's results from the present result expressed in
= 21 log)0[p /(1 —p )], ~here p is the momentum and M is percent. The horizontal axis represents the same variable X
the rest mass of the incident particle. as in Fig. 3. The solid line refers to the result of the
Sternheimer 1956 paper (Ref. 7). The broken line refers to
the result of Berger (personal communication), who used
the same method as that of Ref. 7 and current data for
unsurprising because the asymptotic value of 8 is, in inner-shell binding energies. Note a change in the vertical
effect, determined solely by the mean excitation en- scale for the region X & 0 (i.e., p2 (
0.5); what is plotted
ergy I, for which Sternheimer' used 12.0 Ry =164
1
there represents && of the percentage deviation.
eV, which differs little from 165.7 eV of Ref. 5 [see
Eq. (25)l.
vide support to the series of studies on other materi-
als by Ster nheimer.
V. CONCLUSIONS

The present result, based on the most reliable e(E) ACKNOWLEDGMENTS


data for metallic aluminum, closely agrees with ear-
lier results of Sternheimer based on simpler pro- The present work was performed under the
cedures. This confirms the earlier view of Fano" that auspices of the U. S. DOE. We also thank J. C. Ash-
the Fermi density correction is no more sensitive to ley, M. J. Berger, and U. Pano for valuable advice
details of the dielectric response function than the and information that they generously gave us in the
mean excitation energy. The present results also pro- course of the present work.

H. Bethe, Ann. Phys. (Leipzig) 5, 325 (1930). ~&G. C. W'ick, Ric. Sci. 12, &58 (1941).
H. Bethe, in Handbuch der Physik, edited by H. Geiger and ' G. C. Wick, Nuovo Cimento (9) 1, 302 (1943).
K. Scheel (Springer, Berlin, 1933), Vol. 24/1, p. 273. ' U. Fano, Phys. Rev. 103, 1202 (1956).
3E. Fermi, Phys. Rev. 57, 485 (1940). L. D. Landau and E. M. Lifshitz, Electrodynamics of Con-
4U. Fano, Annu. Rev. Nucl. Sci. 13, 1 (1963). tinuous Media, translated by J. B. Sykes and J. S. Bell (Per-
5E. Shiles, T. Sasaki, M. Inokuti, and D. Y. Smith, Phys. gamon, London, 1960). See Chaps. IX and XII, in partic-
Rev. B 22, 1612 (1980). ular.
R. M. Sternheimer, Phys. Rev. 88, 851 (1952). ' A. R. P. Rau and U. Fano, Phys. Rev. 162, 68 (1967).
~R. M. Sternheimer, Phys. Rev. 103, 511 (1956). '6P, M. Morse and H. Feshbach, Methods of Theoretical
R. M. Sternheimer, Phys. Rev. 145, 247 (1966). Physics (McGraw-Hill, New York, 1953), Vol. I, pp.
9R. M. Sternheimer, Phys. Rev. 164, 349 (1967). 370-372.
' R. M. Sternheimer and R. F. Peierls, Phys. Rev. B 3, 3681 J. C. Ashley, Radiat. Res. (in press).
(1971).

You might also like