You are on page 1of 7

PHYSICAL REVIE% B VOLUME 25, NUMBER 5 1 MARCH 1982

Spin dependence of the electron inelastic mean free path


and the elastic scattering cross section —
a high-energy
atomic approximation

J. A. D. Matthew'
Surface Science Diuision, National Bureau of Standards, Washington, D C. 2. 0234
(Received 20 August 1981)

The spin dependences of the inelastic scattering cross section (inverse mean free path)
and the elastic scattering cross section are calculated for polarized electrons scattered
from oriented atoms in the Born-Ockhur approximation with a view to understanding
spin-dependent scattering in ferromagnets. In the medium-to-high-energy range() 100
eV) the elastic scattering for parallel spins is greater than for antiparallel spins, while the
inelastic cross section for parallel spins is less than for antiparallel. Elastic spin depen-
dence appears to be greater than inelastic, and the exchange effects fall off rapidly with
increasing energy. The relation of this atomistic scattering approach to solid-state models
rs d&scussed.

INTRODUCTION where n, and n, are the number of majority and


minority spins per atom (or strictly per Wigner-
The mean free path (MFP) of electrons above Seitz cell), respectively. Rendell and Penn treat
the Fermi level in a ferromagnet has spin depen- the spin-dependent MFP in a statistical approxima-
dence because electrons with spin parallel to the tion introduced by Lindhard et al. and used by
majority-spin density undergo an enhanced ex- Tung et al. to calculate the electron MFP for Al,
change interaction. The effect may be important Si, Ni, Cu, Ag, and Au, However, Rendell and
in interpreting spin-polarized low-energy electron Penn have extended the previous analysis to in-
diffraction (LEED), ' spin-polarized scattering from clude exchange effects, which had previously been
amorphous materials, and spin-polarized photo- studied just above the Fermi level by Ritchie and
emission. The magnitude of the spin dependence Ashley. ' These models treat the conduction-band
has created some controversy with Bringer et al. electrons, including the d electrons, as an electron
suggesting that the inelastic exchange effects may gas in a kind of local density approximation in
even dominate elastic exchange, while Rendell and contrast to the atomistic arguments of Bringer
Penn estimate that inelastic exchange effects are et al.
much smaller and of opposite sign to those predict- It is of interest, therefore, to investigate a more
ed by Bringer et al. at low energy. In addition detailed atomistic model to give insight into the re-
Helman and Baltensperger have considered spin lation between these two treatments, and to under-
dependence of plasmon scattering in ferromagnetic stand how the low-energy argument of Bringer ex-
materials. tends to higher (but not necessarily very high) ener-
Bringer et al. assume that electrons interact gies. Here we treat the exchange scattering in the
inelastically only with electrons of opposite spin at Born-Ockhur approximation, a form of the Born-
low energy due to cancellation of the direct Oppenheimer scattering approximation. The
Coulomb interaction by the exchange term, so that model gives results which are somewhat similar to
the component of the mean free path due to one those of Bringer et al. at low incident energies,
electron excitation of the conduction electrons will where, however, the scattering approximation em-
for spins parallel (A, , ') and antiparallel (A&, ') to ployed is not strictly valid, and where solid-state
the majority-spin direction obey the relation effects are likely to be important. At higher ener-
gies the effect is much reduced in magnitude and
agrees in sign as well as approximate magnitude
with the estimates of Rendell and Penn for Fe and

3326
25 SPIN DEPENDENCE OF THE ELECTRON INELASTIC MEAN .. 3327

Co, but the variation of the exchange contribution K


goj(~)=
J fpj(~»
J
to the inelastic cross section with energy varies k2
rather differently. It is also possible to make an
estimate of the exchange contribution to the elastic
where K = k 0 — kz is the scattering vector. Inelas-
scattering cross section within the same approxi-
tic scattering is dominantly in the forward direc-
mation. In the high-energy regime it is shown that
tion at high incident energy, and so for scattering
the elastic scattering is greater for electrons with
spins parallel to the majority spins of the sample
angles 8-
0, E = kp — kj, and
than for antiparallel spins in direct contrast to the kp —kJ.
inelastic case. Estimates of the relative magnitudes go) = fo, (6)
kp
of the elastic and inelastic effects are made.
Following Kessler" the differential scattering
BORN-OCKHUR MODEL cross section for an electron to be excited to state j
with parallel spin will be of the form
The Born-Oppenheimer' approximation is a
do
modification of the conventional Born scattering = Ifpj goj I— (7a)
approximation which implies overall antisymmetry .tt
on the many-electron wave function of the system while the sum of all antiparallel scattering process-
(atom plus scattered electron) with respect to coor- es gives
dinate interchange. Let us first consider an elec-
tron of energy E and wave vector k p incident on a
= fo, '+ go,
'. (7b)
single electron atom with initial wave function go I I I I

(energy Eo) excited to a final state l j (energy E~)


with the scattered electron having final wave vec- At the threshold of an inelastic process kj 0, and -
tor k J Depending on whether the spatial part of
~
so fptjt 0. This is exactly the result used by
the wave function is symmetric or antisymmetric, Bringer et a/. However, above threshold gpj de-
the scattering amplitude for 0~ j
transitions is of creases rapidly with increasing incident energy, and
the form " then to a good approximation
2
f+pj fpj+gpj
gpj
1 EO
— fpj . (g)
22E
where f pj is the first Born scattering amplitude
and gpj is the exchange given in atomic units by Let us now attempt to adapt this well-known re-
sult to the problem of scattering from ferromagnet-
g„= —(2tr)- ' II r &
r &2
1(p( r2)'t( j( r t ) ic crystals or amorphous ferromagnetic materials.
Consider a pseudoatom in which there are n,
—k valence electrons per atom in the majority-spin
Xe
i( ko r &
' ~
r2)
dr)dr2 direction and n, in the minority direction (for Ni
n, +n, = 10, n, n, =0.6— ). We treat each of the
electrons as equivalent and scattering independent-
Ockhur obtained a simple method of evaluating ly, and neglect contributions to the mean free path
gpj at reasonably high impact energy by Fourier from inner-core electrons and plasmons.
expanding r ' & & Now, summing over all electrons and over all
possible inelastic transitions, and using the fact
— iq (r& —r2)
r)p e q dq (4) that the inverse mean free path is approximately
2 8.2
proportional to the differential inelastic scattering
Setting q + k p — p it is possible to neglect p in re- cross section at small angles, because cross sections
lation to kp to first order, and gpj may be related are largest there, the asymmetry parameter A is
to the simple Born amplitude by given by

(nt —nt)(& '&— Ifpj '& — gpj '&)


A = ~tt ~tt Ifoj goj I & I & I I

~tt +~tt (nt+nt)(& Ifpj gpj I &+& —


Ifpj I
&+& Igo I
3328 J. A. D. MATTHE% 2S

where (foj) and (goj) are the direct and exchange scattering amplitudes at zero scattering angle averaged
over all possible orientations of the valence-electron wave functions, and A,«' (A, , ') is the inverse mean free
path for electrons parallel (antiparallel) to the majority-spin direction.
Applying the Born-Ockhur approximation we now obtain

—2(n„n,—
)g
J . 2E. ( I f I
)2
i
2 2 ~
4
AEJ.
(n„+n, ) g 1— +1+ (fl, )

In the regime where (bE~/2E) && 1, i.e. , the exci- compare the relative importance of elastic and in-
tation energies are significantly less than the in- elastic effects in the high-energy limit. Firstly, it
cident energy, the expression reduces to is important to note that the elastic scattering cross
section for parallel spins cr", , is greater than for
, n)y— (If
(n— antiparallel spins 0",, in contrast to the result for
)
2E inelastic scattering, where A t t' g A, «'. There is a
I

A== simple physical reason for this. In the inelastic


(ng+ng) X & I fo, I
&'
case the excitations arise purely from repulsive
J
which in turn simplifies to electron-electron interaction in the high-energy
limit: The presence of parallel spin exchange
(n, n, ) —
(bE)
2
reduces close encounters between electrons (i.e. , the
(12)
(n, +n, ) 2E exchange potential is attractive), and the magnitude
of the driving potential of the excitation is re-
where (bE) is the mean excitation energy ob- duced. If the spin of the incident electron is paral-
tained by weighting all possible transition energies lel to the majority-spin direction in the system it
by the relevant cross sections. The result may be experiences more exchange, sees a repulsive
compared with the low-energy estimate of Bringer electron-electron interaction of smaller magnitude,
et al. and of this model at the threshold of an in- and therefore a smaller inelastic scattering cross
elastic channel section. In the elastic scattering case the electron
scatters both from the nuclear potential and the
(13) repulsive potential due to the electrons in the sys-
tem. There is a net attractive potential (the nu-
An estimate of the appropriate (b,E) may be ob- cleus exerts a stronger influence) at least for high-
tained from the peak values of electron loss from energy scattering; with incident spin parallel to the
valence electrons, as observed in Im( — 1/e), which majority-spin direction there is a greater net ex-
for transition metals is typically of order 20 — 25
"
eV. This should at least give a rough magnitude
change interaction so the magnitude of the scatter-
ing potential is now increased, and o.t, & 0.'t&.
of the effect. The results of the model for Ni, Co, These arguments are only valid in the Born limit
and Fe are given in Table I for incident energies in where the cross sections are directly related to the
the 100 — 500-eV range, and are compared with the Fourier transform of a potential, so at low energies
values of Bringer et al. In this model the sign of the sign of the effects may change.
3 is always negative, i.e., At t' & A, , ' as in the Let us consider the quantity
Bringer model, but by 100 eV incident energy 3 is
much smaller in magnitude than Bringer's predic- tt «
(14)
tion, but agrees in approximate magnitude with I A~ I
'+ If~i I'
Rendell and Penn; the exchange effects do, howev-
er, decrease in magnitude much more rapidly than where f&& (f, &)
is the elastic scattering amplitude
for incident spins parallel (antiparallel) to the
in their model. The relationship between these
various approaches will be discussed later in the majority-spin direction.
paper. For closed shells Bonham and Fink' have
It is useful to apply the same approximation to shown that within the Born-Ockhur approximation
exchange effects in elastic scattering in order to the elastic scattering intensity f
is given by
I I
SPIN DEPENDENCE OF THE ELECTRON INELASTIC MEAN. .. 3329

TABLE I. Calculation of 3 =(A I I' —k, , ')/(A, «'+1, »') for incident energies in the range
100 —500 eV for Ni, Co, and Fe.

Co Fe

fl I —ng 0.6 1.5 2.3


n, +n, 10 9 8
(SZ)' (eV) 23 25 23

Co Fe

100 —7.9X10-4 —2.6X10 ' —3.8X10-'


150 —3.5X 10-4 —1.2X 10 —1.7X 10
200 —2X10-4 —6.5X 10-' —9.5 X 10
300 —8.8X 10-' —2.8X10-' —4.2X10-4
500 —3.2X 10-' —1X 10-4 —1.5X10-'
Bringer eI; al. ' —6.0X 10-' —1.7X10-'
'Reference 19.
Reference 12.
'Reference 3.

)
f (
= [Z —F„(K)] 300, and 500 eV for the case of Fe [results for Co
and Ni will approximately scale with (n, n, )]. —
The values of S are positive as explained and rath-
+ f Z F„(K)]F„(—
K)
E er larger than the corresponding results for inelas-
ko tic scattering; although the approximation is cer-
tainly not valid for backscattering, the predictions
are of the same sign and the same magnitude range
0
as those obtained from phase shift analysis at high
where F„(K) is the x-ray atomic scattering factor. scattering angle assuming different exchange densi-
Modifying this result to the case of unbalanced ties for up and down incident spina above 100 eV
spins we obtain, to lowest order, (Davis et al. ' ). At zero scattering angle (forward
K (n,
ko iZ F„(K)
n, )F3g(K)—
where F3d(E) is the atomic scattering factor of a
I— scattering) the effect falls off as F. ', but at higher
scattering angles the effect falls off rather more
quickly with incident energy.

single 3d electron (which is assumed responsible


for the unbalanced exchange scattering for atoms TABLE II. Calculation of S = ( ) — f„( f„()/
in the first transition series). S may alternatively ( f» + f
I i~
I
)

) for Fe in the incident energy range


100—
I I

be written in the form 500 eV. I'3d(E) from ~ejss (Ref. 20); p„(I„) from

5= 2(n, n,
.2
)F3d(K)— Ref. 21.

kof.)
where f,
) is the conventional high-energy electron
(eV)
scattering amplitude. The formula is valid when
the exchange scattering amplitude for an electron
100 5.2X1-' 5.2X 10
is not too large in relation to the direct amplitude,
150 03AX-' 3AX 10-'
which through Eq. (5) implies reasonably high en- 200 2.6X10-' 2.2X 10-'
ergies and reasonably small scattering angles. 17.2 X 10-'
300 13AX 10-'
Table II gives estimates of S for scattering angles 500 10.0X 10-' 5AX10-'
8=0', 30' with incident energies 100, 150, 200,
3330 J. A. D. MATTHEW

DISCUSSION we expect that X & 0 in the high-energy limit as is


observed. Although it is not possible to get reli-
A primary objective of this work is to determine able elastic backscattering cross sections in the
whether elastic or inelastic effects are primarily re- Born-Ockhur approximation, the calculations
sponsible for the spin dependence of the intensities presented here do suggest that elastic spin depen-
of electrons elastically scattered from solids. '
In dence is more prominent than inelastic. Since both
crystals the phenomenon is further complicated by effects vary linearly with n, n„— at fixed incident
diffraction, ' but in amorphous materials a simpler energy X will be proportional to the surface mag-
approach is possible. If one adopts a layer scatter- netization whichever mechanism dominates the
ing model, e.g. , along the lines of Matthew and spin dependence.
Underhill' and if the mean free path A, measured A second objective of this paper is to get insight
in layers is reasonably large (& 3), the incoherent into the relation between different approaches to
backscattered elastic signal per incident electron calculating the spin dependence of inelastic scatter-
will be approximately of the form ing. The approach adopted here is entirely con-
sistent with that of Bringer et al. , but it em-
I= 2 phasizes that the relatively large effects of the type
predicted in that model occur only for primary en-
for a one-dimensional model, where r is the back- ergies very close to important inelastic thresholds.
scattering probability per layer (proportional to the This condition may at least be partially satisfied in
elastic backscattering probability), and A, is pro- low-energy photoemission and in electron scatter-
portional to the inelastic forward scattering proba- ing experiments with primary energies below -40
bility. The effective depth of probing is A, /2 layers eV, but the effect falls off rapidly as the incident
in a simple one-dimensional model, because on the energy increases. Although the Born-Ockhur ap-
average the electrons go in, are backscattered, and proximation is unlikely to give very accurate cross
go out with a total path in the solid of A, . In sections as low as 100 eV the ratio of the exchange
three-dimensional situations I
will still be propor- amplitude to the direct amplitude for atomic
tional to r and A, , but the proportionality constant scattering should be reasonably well represented
may be somewhat different. Now putting in spin above about 100 eV.
dependence in an oriented system As already noted, the sign of the spin depen-
dence agrees with that predicted by Rendell and
Itt oc Penn at high energies, and the magnitude is in
Att reasonable accord with their predictions for Co and
Fe around 100 eV, an interesting result considering
It) o: that the two approaches are wholly dissimilar.
Att
However, there are important differences between
where the two equations have the same propor- the energy dependence predicted by the two calcu-
tionality constant. The calculations performed lations. Rendell and Penn's results for A decrease
here show that in the high-energy limit relatively slowly with increasing energy, while the
atomic model predicts an E behavior. Rendell
and Penn's model is of the local density type and
and treats all conduction and valence electrons as an
—j —1 inhomogeneous electron gas. It differs from an
atomistic calculation in the following ways:
Now (1) the exchange interactions have density depen-
dences akin to those of a free-electron gas,
(20) (2) solid-state screening is implicit in the calcula-
I t t+I tg 2r 2X tion, and
where br=r„—r„, AA, =Att— —A.—. . and
1 1
r, k
—1 (3) constraints are included which account for
are spin-averaged values. Although, in the high- the fact that the continuum states are already filled
energy limit, the cross sections for elastic and in- up to a wave vector kz.
elastic scattering have opposite spin dependence, It is interesting to probe how these effects might
their contribution to the observed spin dependence modify the atomistic result. Screening is expected
of the intensities are in the same direction, and so to reduce both direct and exchange amplitudes, but
SPIN DEPENDENCE OF THE ELECTRON INELASTIC MEAN. .. 3331

will this occur to the same extent in both~ Screen- to plasmons are involved in these transitions, "but
ing may crudely be introduced into the Born- there will be interchange of oscillator strength be-
Ockhur approximation by replacing the bare tween the two kinds of process, so that the atomis-
Coulomb potential between electrons with a tic assumptions %111 glvc R rclcvRnt first-order cstl-

f ~~8,
screened Coulomb potential' of the form
where $1 18 RQ lnvcrsc scI'ccnlng dis-
tance. Following through the theory q in Eq.
mate. Implicitly the theory has assumed localized
1RthcI' than ltlncrant magnetism, which may Qot bc
unreasonable in the magnetic glasses.
It is also interesting to compare the results of
(4) will be replaced by (q +y f) ', and a screened
exchange amplitude goj will be of the form this one-electron model with the spin-dependent
plasmon scattering model of Helman and Balten-
E
— (21) spcI'gcI'. This thcoly RssUIDcs R IQagnctlzcd ffcc-
goj 2 zfoj &

ko+y) clectron gas, and so completely ignores the hybrid


character of electron excitations in magnetic solids.
where foj is the "bare-atom" direct amplitude of
Eq. (5). Screening parameters are at maximum -1 A is predicted to be negative at all energies, and
the effect increases with increasing wave vector of
a.u. ', so that above 100 CV, y~ will be very small,
and screening relatively unimportant. In the direct the plasmon up to a cutoff wave vector k, . In-
will like- tegrated over all scattering angles A varies linearly
process the screened direct amplitude fo with (n, n, )—
wise be of the form /(n, +n, ) as in other models, and
decreases with incident energy more slowly than
—,+f2
foj- E E ufo,
the atomistic model, with the exchange contribu-
tion to A, being energy independent.

whcrc 'Q ls a scI'ccQlng parameter. Now Rt small


angles K-(ko — kj ), and so screerung may now be
more appreciable because E « ko for small
scattering wave vectors. This suggests that the ex-
change amplitude may be less affected by screening The spin dependence of the elastic and inelastic
than the direct amplitude, implying that A will in scattering cross sections for polarized electrons
reality fall off much less quickly in a solid than scattered from atoms with oriented spins has been
the Z predicted in the atomic theory. The calculated within the Bom-Ockhur approximation,
IDuch weakcl cncrgy dcpcndcncc ln thc locRl dcnsl- which is likely to be reasonably valid above 100 CV
ty model may be related to this. OQ the other incident energy. The results are used to give in-
hand, thc local dcnslty approxlIDRtlon ls thought to sight into the scattering of polarized electrons from
exaggerate exchange effects in LEED at higher en- ferromagnets.
ergies' and the correct energy dependence may lie (1) At high energies the elastic scattering cross
somewhere between the two models. Such simple section for spins parallel to the majority-spin direc-
screening arguments give Qo lnslght into the rever- tion is greater than for antiparallel spins.
sal of sign found by Rendell and Penn at low en- (2) The inelastic scattering cross sections for
ergy (indeed for Ni A does not become negative till parallel spins is less than that for antiparallel spins.
above 200 eV). In their model the sign reversal (3) The magnitude of the spin dependence of the
follows directly from the properties of BA& '/Bn, elastic scattering cross sections is greater than for
the derivative of the exchange component of the inelastic scattering, at least for small scattering an-
inverse mean free path with electron density, and gles.
at low energies this is controlled by the variation (4) The spin dependence of the inverse mean free
of k~ with electron density. This quantity is some- path is much smaller than the predictions of
what artificial in the model in that it is defined by Bringer et al. , but agrees in sign and approximate
lumping the d and s electron densities together in magnitude with Rcndcll Rnd Penn s values foI' Co
the gas, and the detailed behavior may well be and Fe. The effect falls off with incident energy
scnsltlvc to such model assumptions. ln thc atomic IDodcl IDolc quickly than ln thc local
The atoIDic model may also be criticized in as- density model.
suming that the (hE) measured from the peak in (5) The atomistic model directly utilizes empiri-
the observed Im( — I /e) corresponds to an effective cal information about the excitations that occur in
energy of one-electron excitations. There is no the material, but does not take account of screen-
doubt that interband onc-electron processes coupled ing in the solid. Simple arguments suggest that
3332 J. A. D. MATTHEW

screening will reduce the direct scattering ampli- ACKNOWLEDGMENTS


tude rather more than the exchange amplitude with
the result that their ratio falls off more slowly with The author wishes to thank R. J. Celotta and D.
energy in solids than in atoms. Feder' has recent- T. Pierce for drawing attention to the problem, and
ly reviewed spin polarized scattering from magnet- acknowledges helpful discussions with R. W. Ren-
ic surfaces, and the results presented here give dell and D. R. Penn on their work. In addition, S.
some further insight into the scattering mechan- M. Girvin made useful comments on the
isms involved. manuscnpt.

*Permanent address: Physics Department, University of "J.Kessler, Polarized Electrons (Springer, Berlin, 1976),
York, Heslington, York Y01 5DD. Chap. 4.
'R. J. Celotta, D. T. Pierce, G. C. Wang, G. D. Bader, ~~C. Wehenkel and B. Gauthe, Phys. Status Solidi-8 64,
and G. P. Felder, Phys. Rev. Lett. 43, 728 (1979). 515 (1974).
R. J. Celotta, D. T. Pierce, H. C. Siegman, and J. ' R. A. Bonham and M. Fink, High Energy Electron
Unguris, Appl. Phys. Lett. 38, 577 {1981). Scattering (Van Nostrand-Reinhold, New York, 1974).
A. Bringer, M. Campagna, R. Feder, W. Gudat, E. ~4H. L. Davis et al. (unpublished).
Kisker, and E. Kuhlmann, Phys. Rev. Lett. 42, 1705 '~J. A. D. Matthew and P. R. Underhill, J. Electron.
(1979). Spectrosc. Relat. Phenom 14, 371 (1978).
4R. W. Rendell and D. R. Penn, Phys. Rev. Lett. 45, ' J. M. Ziman, Principles of the Theory of Solids (Cam-
2057 {1980). bridge University, Cambridge, England, 1969), p. 126.
5J. S. Helman and W. Baltensperger, Phys. Rev. B 22, '7M. R. Welton-Cook and M. Prutton, J. Phys. C 13,
1300 (1980). 3993 (1980).
J. Lindhard, M. Scharff, and H. E. Schiftt, K. Dan. ~8R. Feder, J. Phys. C 14, 2049 {1981).
Vindensk. Selsk. Mat. Fys. Medd. 33, (14), 39 {1963). ' R. S. Tebble and D. J. Craik, Magnetic Materials (Wi-
7J. Tung, J. C. Ashley, and R. H. Ritchie, Surf. Sci. 81, ley, New York, 1969), and references contained
427 (1979). therein.
R. H. Ritchie and J. C. Ashley, J. Phys. Chem. Solids R. J. Weiss, X ray Determina-tion of Electron Distribu
26, 1689 (1965). tions (North-Holland, Amsterdam, 1966), p. 71.
9V. J. Ockhur, Zh. Eksp. Teor. Fiz. 45, 734 (1963) [Sov. 'International Tables for X-ray Crystallography, edited

Phys. JETP 18, 503 (1964)]. by C. H. MacGillavry and G. R. Rieck {Kynoch,
' J. R. Oppenheimer, Phys. Rev. 32, 361 {1928). Birmingham, 1968), Vol. 3, p. 201.

You might also like