PRB 2000
PRB 2000
PRB 2000
net/publication/235578234
CITATIONS READS
19 212
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Maurizio Dapor on 27 June 2014.
Antonio Miotello
Istituto Nazionale per la Fisica della Materia and Dipartimento di Fisica dell’Università di Trento, I-38050 Povo, Trento, Italy
Davide Zari
Istituto Trentino di Cultura and Dipartimento di Fisica dell’Università di Trento, I-38050 Povo, Trento, Italy
共Received 8 June 1999; revised manuscript received 1 October 1999兲
We have analyzed the problem of the secondary electrons emitted from a solid irradiated by a positron or
electron particle beam. The relevant concepts that form the basis for a theoretical understanding of the sec-
ondary electron emission problem are reanalyzed by looking at Wolff’s theory. However, to attain analytical
results, such a theory must include some simplification, which may be valid in only a limited number of
situations. To overcome such a limitation we have elaborated a Monte Carlo procedure for the calculation of
the secondary electrons emitted from a solid, irradiated with a positron beam. The choice of a primary positron
beam is justified because the experiments involving electron emission are not contaminated by the reemitted
primary electrons. The calculations were performed in the positron primary energy range between 50 eV and
2 keV, and for different incidence angles with respect to the surface of a copper sample. Many numerical
results are reported, namely: i兲 the elastic mean-free paths of positrons in copper, ii兲 the mean number of
electrons emitted per positron, iii兲 the penetration depth of the positrons, and iv兲 the depths from which the
secondary electrons are emitted. Finally, the numerical results concerning the secondary electron energy
distribution are compared with the experimental data recently presented by Overton and Coleman showing a
general good agreement.
I. INTRODUCTION emerge from the surface. If the target is not a thin film
共namely, if there are not transmitted particles兲 the remaining
The problem of secondary electron emission from solids primary particles are trapped into it. For bulk targets the ratio
irradiated by a particle beam is relevant, mainly in connec- between the number of bakscattered and total
tion with the analytical techniques that utilize secondary (backscattered⫹trapped) particles is generally called the
electrons to investigate chemical and compositional proper- backscattering coefficient.
ties of solids in the near surface layers: Auger electron spec- If the primary particles are electrons, then the spectrum of
troscopy and X-photoelectron spectroscopy. In general, the the secondary electrons is clearly contaminated by the con-
energy spectra of the emitted electrons are quite complicated tribution of the backscattered primary electrons. On the other
because many features appear in such a spectra in connection hand, as recently noted by Overton and Coleman,4 the prob-
with the different collisional processes involved before low- lem of distinguishing between true secondary and backscat-
energy secondary electron emission. As a consequence, a tered electrons is absent, if the secondary electron emission
better understanding of the collisional events occurring in the is stimulated by positron beams. In the quoted paper,4 the
surface layers before emission, should permit a more general authors performed an interesting experimental study of the
understanding of the surface physics including, for example, spectra of fast secondary electrons not contaminated by the
plasmon excitation. Recently, the collisional processes of backscattered electrons. Specifically, secondary electrons
positron beams impinging on solids received great attention produced by positron beams, impinging on copper target at
because of the possibility to realize, by positron annihilation glancing and 35° incidence angle and for primary energies in
spectroscopy, nondestructive investigations of point and ex- the range from 50 eV to 2 keV, have been analyzed.4
tended defects of surfaces, interfaces, and of bulk materials. Experimental results have been compared to the
Review papers on the subject have been prepared by Dupas- Sickafus5–7 empirical law:
quier and Zecca,1 Schultz and Lynn,2 and Asoka-Kumar
et al.3 j 共 E 兲 ⫽AE ⫺m , 共1兲
When a particle beam, with particle energy exceeding
some threshold value, impinges on a solid target, it stimu-
lates the emission of secondary electrons through collisions where j(E) represents the measured secondary electron en-
with target atoms. On the other hand, a fraction of the par- ergy distribution, E being the energy of the emitted second-
ticles of the primary beam is also ejected from the surface ary electrons, while A and m are constants that depend on the
because, after a number of elastic and/or inelastic collisions solid and on the energy of the impinging positrons.
with the target atoms, some of those particles come back and The comparison shows excellent linear correlation when
the data are plotted as log10关 j(E) 兴 vs log10(E) and the values Here N is the number of electrons between r and r⫹dr, ⍀
of m are reported as a function of the energy of the colliding and ⍀⫹d⍀, E and E⫹dE. ⍀ is a unit vector in the velocity
positrons. direction, S is a source term that represents the secondary
In the present paper, we try to explain the experimental electrons produced by the primary beam, and F is the prob-
results4 by using a Monte Carlo code in which, essentially, ability that, given an electron at ⍀⬘ with energy E ⬘ , another
the positron-solid interaction is included as described by the electron will be found after scattering at ⍀ with energy E. ⌰
relevant cross sections and stopping-powers of the collisional is the angle between the directions ⍀ and ⍀⬘ . P l are the
processes 共both elastic and inelastic兲 and the corresponding Legendres’s polynomials.
mean-free paths. It is evident that the set of equations 共2兲 is quite compli-
However, before to illustrate the numerical procedure we cated to be utilized and an approximate approach is needed.
have adopted, it is important to gain some insights into the In particular, Wolff8 observed that the high-l harmonics be-
relation 共1兲 by looking at the previously proposed analytical come important only for high energies. For low energies 0
models. In this way, we will understand the theoretical overwhelms the other harmonics because it grows rapidly as
framework that forms the basis to approach the problem of E→0. In this case
the secondary electron emission as well as the limits of the
analytical procedures, which, on the contrary, may be over- j 共 E 兲⫽ 0共 E 兲. 共7兲
come by numerical methods, mainly based on a Monte Carlo By utilizing the Goldberger’s evaluation of the probability of
simulation. Let us now reanalyze Wolff’s theory8 to under- scattering from E ⬘ to E 共for S-wave scattering from a degen-
line the basic concepts necessary to analyze the secondary erate Fermi gas兲9 and taking into account the exclusion prin-
electron emission problem. ciple, Wolff showed that8
The secondary electrons production occurs in two steps. where x(E) decreases as energy increases from E f to ⬃4E f
The first step is the secondary electrons production due to the and equals 2 for energies higher than ⬃4E f (E f is the Fermi
impacts between the primaries particles and the electrons energy兲. For E→2E f , x(E)→2.3. E o is the primary
bound in the solid. The second process is the so-called cas- particle-beam energy. The proportionality constant, in Eq.
cade: the secondary electrons travel in the material produc- 共8兲, that we will indicate as P(E), is a geometrical factor
ing other secondaries before being trapped in the solid or representing the probability that an electron reaching the sur-
before emerging from the surface. The equations that govern face have a large enough normal velocity component to es-
the cascade process, as deduced from the Boltzmann trans- cape. Assuming a spherical symmetric distribution and indi-
port equation, for the steady state, when the primary particles cating with the work function, P(E) can be calculated
collide normally with the target surface in the z direction, are through
the following:8
l ⫽ 共 E 兲 冋
l l⫺1
⫹
l⫹1 l⫹1
册 P 共 E 兲 ⫽1⫺ 冑 ⫹E f
E
. 共9兲
2l⫹1 z 2l⫹1 z
The Goldberger’s evaluation of the reduction of the total
⫹ 冕E
⬁
dE ⬘ F l 共 E,E ⬘ 兲 l 共 z,E ⬘ 兲 ⫹S l 共 z,E 兲 . 共2兲
scattering cross-section due to the exclusion principle entails
the following equation for the mean-free path 共valid for E
⬎2E f )
In this set of integro-differential equations,
1
l ⫽ v N l / 共 E 兲 , 共3兲 共 E 兲⫽ , 共10兲
n c inel 共 E 兲共 1⫺7E f /5E 兲
(E) is the electron mean-free path, v is the electron veloc- where n c is the number density of conduction band electrons,
ity, E is the electron energy, and N l , F l , and S l are the and inel (E) is the electron-electron inelastic scattering
coefficients of the following expansions in spherical harmon- cross section. By utilizing Eqs. 共7兲–共10兲, Wolff’s formula
ics: follows:
冉 冊
⬁
1 1⫺ 冑共 ⫹E f 兲 /E x(E)
N 共 z,cos ,E 兲 ⫽
4 兺 共 2l⫹1 兲 N l共 z,E 兲 P l共 cos 兲 ,
l⫽0
共4兲 j共 E 兲⫽
Eo
n c inel 共 E 兲共 1⫺7E f /5E 兲 E
. 共11兲
⬁
1
S 共 z,cos ,E 兲 ⫽
4 兺 共 2l⫹1 兲 S l共 z,E 兲 P l共 cos 兲 ,
l⫽0
共5兲 III. APPROXIMATE ANALYTICAL RELATIONS FOR
PRIMARY ELECTRON ENERGY EXCEEDING 100 eV
sumed to be spherically symmetric. Since the shapes of the function approach for inelastic scattering and on the Relativ-
secondary electron spectra, on the other hand, are substan- istic Partial Wave Expansion Method 共RPWEM兲 for elastic
tially determined by electrons with energies lower than scattering.
40 eV, the spectra are quite insensitive to the choice of W.8
Sickafus5 empirically tried to extend the energy range of
IV. THE MONTE CARLO SCHEME
Wolff’s cascade theory. He assumed that EⰇE f and ob-
served that, in such a case, P(E)⬃1, (1⫺7E f /5E)⬃1, and The Monte Carlo method is a very powerfull and reliable
x(E)⬃2. As a consequence procedure to evaluate important quantities related to the
electron-solid and positron-solid interaction, such as back-
E xo scattering coefficients and implantation profiles,14,16,17,15 sec-
j共 E 兲⬀ . 共12兲 ondary electron emission,18,19 and spectra of backscattered
inel 共 E 兲 E x and of secondary electrons.20,21 In the present calculation, the
Assuming that the inelastic cross section has the following elastic scattering of both positrons and secondary electrons is
form:10 computed step by step along the path of the particle by cubic
spline interpolation of extensive tabulations of the differen-
log E tial elastic scattering cross section obtained by the code we
inel ⬀ , 共13兲 described elsewhere.22,23 The use of cubic spline interpola-
E tion rather than direct calculation is due to the fact that the
then direct computation of the differential scattering cross sec-
tions requires a large amount of computational time. Any
way, the adopted Monte Carlo scheme is very accurate in the
E xo
j 共 E 兲 ⫽C , 共14兲 determination of the scattering angle, , after every elastic
E x⫺1 log E collision 共errors being within 1–5 %兲.23
Concerning the inelastic scattering, we have used the
where C is a constant. As a consequence, Ashley treatment11,13 for the calculation of the stopping
power, dE/ds, and inelastic mean-free path, inel . Our cal-
log j 共 E 兲 ⫽x log E o ⫺ 共 x⫺1 兲 log E⫺log log E⫹log C. culations are in excellent agreement with those of
共15兲 Ashley.11,13 In order to expedite the computer simulation we
Since the function log log E is almost constant, on a given also have used a cubic spline interpolation of previously
range of E, we conclude that, for any given primary energy computed and tabulated data for inelastic scattering, as for
E o , log j(E) as a function of log E is given by the elastic case. Before each step of the particle trajectory, a
random number, uniformly distributed in the range 关0,1兴, is
log j 共 E 兲 ⫽k⫺m log E. 共16兲 generated and compared to the probability of inelastic scat-
tering. If the random number is lower than that probability,
where m⬃1 and k is a constant. then the collision will be inelastic. The probability of inelas-
In other words, a plot of log j(E) versus log E presents a tic scattering p inel is given by
linear trend 共the so-called Sickafus region of the spectrum兲
⫺1 ⫺1
and the value of m should be approximately 1. p inel ⫽ inel / 共 inel ⫹ ⫺1
el 兲 . 共17兲
On the other hand, by utilizing the dielectric function
approach11–13 to calculate the inelastic cross section, and by In this equation, el and inel are, respectively, the elastic
empirically introducing the results directly into the Wolff and the inelastic mean-free paths:
formula 关Eq. 共11兲兴, we obtained, for primary energies lower
than 1000 eV, spectra with values of m higher than ⬃2. In 1
particular, for copper, we found that m⫽2.24⫾0.04 and m el ⫽ , 共18兲
⫽2.01⫾0.04 when E o is equal to 500 and 1000 eV, respec- N el
tively. Note that the interval regions where the slopes of the
log-log plots have been calculated were quite restricted, be- 1
ing 50–155 eV for E o ⫽500 eV, and 50–270 eV for E o inel ⫽ , 共19兲
N inel
⫽1000 eV, while the corresponding linear correlations co-
efficients were, respectively, 0.997 and 0.995. where el is the total elastic scattering cross section, inel is
Wolff’s theory give results that are surprisingly good, if the total inelastic scattering cross section, and N is the num-
compared with many experimental data, even if these data ber of atoms per unit of volume in the target.
are outside the claimed limits of validity of the theory. It is The step-length, ⌬s, is calculated as
reasonable to conclude that Wolff’s theory could be ex-
tended outside the ⬃100 eV because the shapes of the sec-
⌬s⫽⫺ tot •ln共 r ⌬s 兲 , 共20兲
ondary electrons spectra are dominated by the low-energy
electrons.
where r ⌬s is a random number uniformly distributed in the
However, in order to avoid empirical procedures to ob-
range 关0,1兴 and tot is defined by
tain m, and to include also the elastic scattering collisions,
we decided to simulate the positron and the electron trajec- ⫺1 ⫺1
tories with a Monte Carlo procedure based on the dielectric tot ⫽ inel ⫹ ⫺1
el . 共21兲
5982 MAURIZIO DAPOR, ANTONIO MIOTELLO, AND DAVIDE ZARI PRB 61
If the collision is elastic, then we assume that the incident The azimuthal angle after each collision 共elastic or in-
particle changes its direction in agreement with the differen- elastic兲 takes on any value in the range 关0,2 兴 selected by a
tial elastic scattering cross section obtained by the random number, r , uniformly distributed in that range.
RPWEM.22,23 Both the and angles are calculated relative to the last
The polar scattering angle, , after an elastic collision is direction in which the electron was moving before the im-
selected via a random number, r , uniformly distributed in pact. The direction z1 along which the electron is moving
the range 关0,1兴. The choice is such that the integrated prob- after the last deflection, relative to the z direction, is given by
ability for scattering in the angular range from 0 to equals
r : cos z1 ⫽cos z cos ⫹sin z sin cos , 共26兲
2 冕 d
0 d⍀
sin d
where z is the angle relative to the z direction before the
impact.
The motion ⌬z along the z direction is then calculated
r ⫽ 共22兲
冕
d
.
as
2 sin d
0 d⍀
⌬z⫽⌬s•cos z1 . 共27兲
If the collision is inelastic, the energy loss can be evaluated
by utilizing a random number r ⌬E , uniformly distributed in The new angle z1 then becomes the incident angle z for
the range 关0,1兴 so that the next path length.
The initial energy of each secondary electron, produced
冕 0
⌬E
⫺1
共 d inel /dT 兲 dT by an inelastic collision, is calculated by
r ⌬E ⫽ 共23兲
冕 E⫺E f
, E s ⫽⌬E⫹E f , 共28兲
⫺1
共 d inel /dT 兲 dT
0 where ⌬E is now the total energy lost by the particle 共posi-
⫺1
tron兲 as calculated from the last inelastic collision, E f is the
where (d inel /dT) represents the probability for energy loss Fermi energy 关here included because ⌬E is calculated from
T per unit distance travelled by an electron of energy E, i.e., the dielectric excitation function 共Ref. 21兲兴. The secondary
the inverse mean-free path differential in energy transfer. electron trajectory is described exactly as the trajectory of
This process of evaluation of the energy loss is quite time the positrons of the primary beam, by including, of course,
consuming. Then, in order to expedite the calculations, we the appropriate cross sections. Note that the whole secondary
introduced the following approximation. Along each step ⌬s electron cascade is followed in the numerical simulation.
of the particle trajectory we approximated the energy losses Both positrons and secondary electrons are followed until
assuming that the particle loses an amount of energy, which their energy becomes lower than 16 eV 共which for electrons
is evaluated by multiplying the stopping power, dE/ds, by corresponds to the Fermi energy plus the work function of
the step-length ⌬s, namely: the copper target, in the present case兲 with respect to the
vacuum level. We assume that the secondary electrons are
dE emitted with an angular distribution having spherical sym-
⌬E⫽ ⌬s. 共24兲
ds metry, as suggested by Shimizu and Ze-Jun.21
The surface energy barrier clearly influences the energy
Actually the use of the stopping power within the continuous distribution of the ejected low-energy electrons. In particular,
energy loss approximation for the evaluation of the energy it should be noted that an electron cannot escape from the
loss is a rough simplification: positrons and electrons can surface into vacuum with an angle higher than
lose, indeed, large amounts of their energy in single colli-
冑
sions. So, in order to check the validity of the present nu-
共E f⫹兲
merical procedure in relation to the whole energy loss of the c ⫽cos⫺1 , 共29兲
positrons penetrating into a solid, we compared our calcu- 共 E⫹E f ⫹ 兲
lated maximum range with the data of Mills and Wilson where E f is the Fermi energy and is the work function of
concerning the transmission of 1–6 keV positrons through the irradiated material 共copper in the present case兲. Our nu-
thin metal films.24 These authors evaluated with great accu- merical simulation may use both the classical and quantum-
racy the maximum range of keV-positrons in Al and in Cu: mechanical formula for the transmission coefficient. The
for 3100 eV positrons in Cu they found an experimental data reported here used the quantum-mechanical formula.21
range equal to 69.1 nm while with our numerical code we The number of positron trajectories simulated for each en-
obtained, in the same conditions adopted in the ergy spectrum ranges from 104 to 106 and the overall cas-
experiments,24 a range of 70.8⫾0.7 nm. The comparison cade of the secondary electrons is followed in the computa-
with other energies may be found in Table IV that we will tion.
comment below.
The polar scattering angle, , after each inelastic collision
is calculated as V. RESULTS AND DISCUSSION
冑
The results we are presenting have been obtained by the
⌬E Monte Carlo procedure just described. The inelastic and elas-
sin ⫽ . 共25兲
E tic mean-free paths and the stopping powers we utilized are
PRB 61 MONTE CARLO SIMULATION OF POSITRON- . . . 5983
TABLE I. Inelastic mean-free path and stopping power of elec- TABLE III. Mean number, , of secondary electrons emitted
trons 共Ashley et al., 1976兲 and positrons 共Ashley, 1990兲 in Cu. per incident positron 共of energy E o ).
⫺ ⫺ ⫹ ⫹
E⫺E f inel
e
关 ⫺(dE/ds) 兴 e inel
e
关 ⫺(dE/ds) 兴 e Eo
共eV兲 (Å) (eV/Å) (Å) (eV/Å) 共eV兲
Eo m E min E max r
共eV兲 共eV兲 共eV兲
1
A. Dupasquier and A. Zecca, Riv. Nuovo Cimento 8, 1 共1985兲. 10
C.J. Powell, J. Vac. Sci. Technol. 13, 219 共1976兲.
2
P.J. Schultz and K.G. Lynn, Rev. Mod. Phys. 60, 701 共1988兲. 11
J.C. Ashley, C.J. Tung, R.H. Ritchie, and V.E. Anderson, IEEE
3
P. Asoka-Kumar, K.G. Lynn, and D.O. Welch, J. Appl. Phys. 76, Trans. Nucl. Sci. NS23, 1833 共1976兲.
4935 共1994兲. 12
J.C. Ashley, J. Electron Spectrosc. Relat. Phenom. 46, 199
4
N. Overton and P.G. Coleman, Phys. Rev. Lett. 79, 305 共1997兲. 共1988兲.
5
E.N. Sickafus, Phys. Rev. B 16, 1436 共1977兲. 13
J.C. Ashley, J. Electron Spectrosc. Relat. Phenom. 50, 323
6
E.N. Sickafus, Phys. Rev. B 16, 1448 共1977兲. 共1990兲.
7
E.N. Sickafus and C. Kukla, Phys. Rev. B 19, 4056 共1979兲. 14
M. Dapor, Phys. Rev. B 46, 618 共1992兲.
8
P.A. Wolff, Phys. Rev. 95, 56 共1954兲. 15
M. Dapor and A. Miotello, Eur. Phys. J.: Appl. Phys. 5, 143
9
M.L. Goldberger, Phys. Rev. 74, 1269 共1948兲. 共1999兲.
5986 MAURIZIO DAPOR, ANTONIO MIOTELLO, AND DAVIDE ZARI PRB 61
16
J. Barò, J. Sempau, J.M. Fernàndez Varea, and F. Salvat, Nucl. 13, 149 共1988兲.
Instrum. Methods Phys. Res. B 100, 31 共1995兲. 21
R. Shimizu and Ding Ze-Jun, Rep. Prog. Phys. 55, 487 共1992兲.
17
A. Miotello and M. Dapor, Phys. Rev. B 56, 2241 共1997兲. 22
M. Dapor, J. Appl. Phys. 79, 8406 共1996兲.
18 23
A. Dubus, J.C. Dehaes, J.P. Ganachaud, A. Hafni, and M. Cailler, M. Dapor and A. Miotello, At. Data Nucl. Data Tables 69, 1
Phys. Rev. B 47, 11 056 共1993兲. 共1998兲.
19
K. Ohya, Jpn. J. Appl. Phys., Part 1 33, 4735 共1994兲. 24
A.P. Mills and R.J. Wilson, Phys. Rev. A 26, 490 共1982兲.
20
S. Ichimura, R. Shimizu, and Ding Ze-Jun, Surf. Interface Anal. 25
N.F. Mott, Proc. R. Soc. London, Ser. A 126, 259 共1930兲.