ISSN 1063 7842, Technical Physics, 2009, Vol. 54, No. 11, pp. 15661570. Pleiades Publishing, Ltd.

., 2009. Original Russian Text Yu. Kudriavtsev, R. Asomoza, 2009, published in Zhurnal Tekhnichesko Fiziki, 2009, Vol. 79, No. 11, pp. 69.

ATOMS, SPECTRA, RADIATION

Refined Statistical Model of Secondary Ion Formation

Yu. Kudriavtsev and R. Asomoza
Department of Solid State Electronics, Research and Advanced Education Center, National Polytechnical Institute, 07360 Mexico, Mexico e mail: yuriyk@cinvestav.mx
Received August 18, 2008; in final form, February 12, 2009

AbstractA model of formation of secondary ion during ion beam sputtering of a target is considered. The model is based on the so called statistical model of formation of secondary ions at a certain critical distance from the surface, which was proposed earlier. The concept of dynamic temperature introduced in earlier pub lications for a cascade of collisions initiated by a primary ion, as well as a new interpretation of the interaction of the formed ion with the surface charge of opposite polarity, enabled us to derive an analytic expression for the ion formation probability. Comparison of the results of calculation with experimental data shows good agreement testifying the correctness of the proposed model. PACS numbers: 34.50.Bw, 79.20.Ap, 79.20.Rf, 82.80.Ms DOI: 10.1134/S1063784209110024

INTRODUCTION The interaction of accelerated ions with the surface of a solid leads, among other effects, to sputtering of surface particles. This phenomenon is of considerable practical importance since the study of the mass spectra of sputtered particles permits elemental analysis of materials with a high sensitivity (up to 1012 atoms/cm3). In view of technical problems of postionization of sec ondary atoms, such analysis is carried out for second ary ions (secondary ion mass spectrometry, SIMS). Since the 1960s, when the first publication in this field appeared, until now, SIMS specialists have been inter ested in determining the ionization probability () of a sputtered particle. Unfortunately, even the best of the large number of secondary ionization models make it possible to obtain only order of magnitude estimates of the value of , which is naturally insufficient for practical spectrometry. For this reason, the SIMS method remains qualitative; quantitative analysis can be carried out only for pure materials (e.g., semicon ductors) using specially prepared (implanted) stan dards. Among other models proposed in this field, the Dobretsov statistical model (SM) of ionization [1] was actively discussed in 19601979. Dobretsov consid ered ionization processes in the quasi adiabatic approximation. The motion of a recoil atom is too slow in comparison with the rate of electron transi tions; for this reason, the electronic subsystem of a solid and the recoil atom is rapidly rearranged and the ion formation process takes place for a system in dynamic equilibrium. The electrons of the recoil atom and electrons of the solid form a single multielectron system obeying the FermiDirac distribution law.

The situation radically changes when the recoil atom or, to be more precise, quasi atom (since the charge state of the recoil particle is not defined) attains a certain critical distance xc from the surface, at which the probability of electron exchange between the energy levels of the atom and the surface becomes neg ligibly low. Thus, the charge of the recoil atom is defined by the probability of filling of the electron energy levels of this atom precisely at the critical dis tance. The ratio of the probability of filling of states of the recoil particle to the probability of vacancy forma tion immediately before reaching xc defines the ion ization probability. On the basis of the FermiDirac statistics, Dobretsov derived the following relations for positive and negative ions, respectively, in the case of a metal surface [1]: Px =
+

I * + exp x , kT c 0

Px =

* Ax . (1) kT c 0

Here, is the work function of the surface under investigation, is the statistical weight of the charged (+ or ) or neutral (0) state of a particle, I* and A* are the ionization potential and the electron affinity of the elements being analyzed at the critical distance, k is the Boltzmann constant, and Tc is the temperature of the emitting surface. Dobretsov proposed the follow ing relations for the ionization potential and electron affinity, which take into account the interaction of the formed ion with its image at the critical distance: I* = I e 4x c
2

and

A* = A +

e , 4x c

(2)

where e is the electron charge and I and A are the ion ization potential and the electron affinity of the given

1566

REFINED STATISTICAL MODEL OF SECONDARY ION FORMATION

1567

elements at an infinitely long distance from the sur face. No estimates were obtained by Dobretsov or his successors for the temperature of the bombarded sur face in spite of the critical nature of this parameter for the verification of the model in the whole. Thus, tem perature was used as a certain fitting parameter for comparing the model with experimental data. Such a comparison led to a temperature estimate of 2000 8000 K, which has not been confirmed by direct mea surements of the temperature of the target. 1. THEORETICAL ANALYSIS We analyzed the SM critically and proposed a number of important modifications. The relations derived here were used for estimating the ionization probability for various elements sputtered from silicon by oxygen and cesium ions and compared with litera ture data. 1.1. Temperature of the Cascade First, we considered the possibility of a local increase in temperature in the region of collision cas cade initiated by a primary ion. The size of the cascade is limited to approximately 50 atomic spacings for considered primary energies. In this region, the energy of the primary ion after a series of quasi elastic colli sions with atoms of the solid is transferred to these atoms during approximately 1012 s. It is important to note that, in spite of the short lifetime of the cascade, a certain dynamically equilibrium energy distribution of recoil atoms sets in it. In addition, intense electron transitions occur in the collision cascade region, which is manifested, in particular, in secondary elec tron emission and electromagnetic radiation. Obvi ously, part of the energy of the primary ion is trans ferred to the electron subsystem of the solid primarily in the collision cascade region; in other words, taking into account the above arguments, we can assume that a certain dynamically equilibrium distribution over energy levels exists for the electronic subsystem also. Nonequilibrium electrons and holes are responsible for the ionization of recoil particles. In our previous publication, we proposed a defini tion of dynamic temperature characterizing very fast processes including the formation of collision cas cades in solids. The available models of ion sputtering predict that recoil atoms in a collision cascade have a strictly definite energy distribution described by a potential function ~E2 [2]. It was demonstrated exper imentally that the energy distribution of atoms indeed decreases in proportion to the square of energy [3]. This fact led us to the conclusion that a certain dynamic distribution of atoms, characterized by dynamic temperature Td, exists during a time period less than 1012 s in the cascade region. Like the con ventional temperature for other polyatomic quasi
TECHNICAL PHYSICS Vol. 54 No. 11 2009

isolated systems, the dynamic temperature was defined as the mean kinetic energy of the system per atom. From the energy distribution of the form E2, we derived the following relation for the dynamic temper ature [4]: (3) kT d E c kT ln [ E 0 / ( kT ) ], where T is the temperature of the target, E0 is the energy of primary ions, and is the kinematic factor defined as (4) = ( 4M 1 M 2 )/ ( M 1 + M 2 ) , here, M1 is the mass of the ion and M2 is the mass of the target atom. An analogous definition of temperature can also be used for other quasi isolated short lived systems of particles, for which an energy distribution can be determined. It is important to note that the measur ing time for dynamic temperature is limited by the lifetime of such an equilibrium state with a certain energy distribution of particles; moreover, it must be much shorter than the lifetime of this dynamically equilibrium state. Since the time of emission of a sec ondary atom is 10131014 s depending on the mass and energy of the atom (which is shorter than the life time of the cascade by a factor of 10100), the dynamic temperature of the cascade can be mea sured with the help of sputtered atoms (from their energy distribution). This feature forms the main dif ference between the dynamic temperature and the thermodynamic temperature (e.g., the temperature of the sample). Thus, for the temperature of the emitting surface, we used in Eq. (1) dynamic temperature (3) as the quantity characterizing the energy of the cascade. 1.2. Position of the Ionization Level and Electron Affinity at the Critical Distance Quantum mechanical calculations of the position of ionization levels and electron affinity in the emitted atom at the critical distance from the surface are a complex problem that has not been solved as yet. In our opinion, required information can be obtained from analysis of chemically and physically adsorbed particles at the surface (e.g., using X ray photoelec tron spectroscopy, XPS). This method makes it possi ble to determine experimentally the shifts of valence energy levels for adsorbed particles. On the basis of a large body of experimental data, the following semi empirical relation was obtained for such a chemical shift [5]: (5) E ch e + q q 1 . 4d 2d r Here, q is the charge of the adatom, e is the electron charge, d is the effective dipole length, and r is the characteristic radius of valence orbitals for the ada tom. In the first approximation, we can set r d.
2 2

1568

KUDRIAVTSEV, ASOMOZA
E max

We consider a recoil particle at a distance consider ably longer than the equilibrium distance between a chemisorbed atom and the nearest atom of the solid (to be more precise, between the nuclei of these two atoms). The degree of overlapping of electron clouds of such a recoil particle and the nearest surface atom (and, hence, the local redistribution of charges deter mining the value of q) exponentially decreases with increasing distance between these atoms. Conse quently, we can assume that the value of q at distance xc is much smaller than e, and relation (5) is simplified: E ch e . 4x c
2

W =

Eq

f ( )d
, (7)

E max

f ( )d
0

(6)

Thus, we have obtained the same relation for the ionization potential and electron affinity at the critical distance as that proposed by Dobretsov (Eqs. (2)). 1.3. Statistical Weight of States Expression (1) for the ionization probability con tains the ratio of statistical weights of the charged and neutral states. The statistical weight of a state can be calculated as the possible number of combinations of the given state. For example, in the case of alkali met als, the outer shell contains one electron. Taking into account possible orientations of the electron spin, we can speak of two possible statistical states of a neutral atom. For a positive alkali metal ion, the allowed energy levels in the preceding shells are filled com pletely; i.e., the ion state can be preset in a unique way. Consequently, the ratio of statistical weights for such elements is 0.5. Detailed analysis of the electronic structure for other elements from the Periodic Table shows that the value of 0.5 is valid for most elements. The exception is Be, Mg, Ca, Zn, Sr, Pd, Cd, Ba, and Hg. For these elements, the ratio of statistical weights is 2. Precisely the same ratio is observed for negative ions of the same elements: 2 for the elements listed above and 0.5 for all remaining elements. 1.4. Secondary Ion Emission Probability The Dobretsov formula was initially derived for ionization of fast particles during their scattering from an incandescent surface. It was assumed that all parti cles (charged or neutral) leave the surface due to their own kinetic energy. In the case of sputtered particles, it was shown [6] that a part of secondary ions formed with an energy lower that the energy of Coulomb attraction between this ion and an opposite charge formed at the surface as a result of the electron transi tion cannot leave the surface. The ratio of the proba bility of emission of sputtered atoms and ions, which must appear as an additional coefficient in the Dobretsov equation, can be found from energy distri bution f () of sputtered atoms:

where Eq is the energy of Coulomb interaction between two charges at the critical distance of ioniza tion of sputtered atoms and Emax is the maximal energy of these atoms. In comparison with experimental data, the maximal energy corresponds to the transmission band of the energy analyzer of the mass spectrometer (130 eV for magnetic and 1525 eV for quadrupole mass spectrometers). In the case of magnetic mass spectrometer, estimates based on formula (7) give val ues of W in the range of 0.80.9; consequently, this coefficient can be omitted in the first approximation. Taking into account the above arguments, we can write the final expression for the ionization probability in the form I e E+ v 4x c + P x = 0.5*Wexp kT ln ( E 0 /kT ) for positive ions and
2 2

(8a)

E A e c 4x c (8b) P x = 0.5*Wexp kT ln ( E 0 /kT ) for negative ions. The asterisk in these equations indi cates the existence of exceptional elements for which the ratio of statistical weights is 2; Ec and Ev are the Fermi levels (Ef ) for metals (Ec = Ev = Ef ), the bot tom of the conduction band, and the top of the valence band with allowance for the shift caused by ion bom bardment for negative and positive ions, respectively, in the case of semiconductors [7]. 2. COMPARISON OF THE MODEL WITH EXPERIMENT The expressions derived above were used for esti mating the ionization probability of various elements sputtered from silicon by oxygen and cesium ions. The calculated values were compared with experimental data. For this purpose, we calculated the so called rel ative sensitivity factors (RSF) defined as I (9) EEF = C X M , IX where CX is the concentration of elements X we are interested in and Ii is the intensity of ions of this ele ment (X) and ions of the matrix (M) (silicon in our case). The intensity of secondary ions of this species is
TECHNICAL PHYSICS Vol. 54 No. 11 2009

REFINED STATISTICAL MODEL OF SECONDARY ION FORMATION (a) Calculated RSF, atoms/cm3 Calculated RSF, atoms/cm3 1026 (b)

1569

1024

1022

1024

1021

1023

1025

1023

1025

1027

Experimental RSF, atoms/cm3

Experimental RSF, atoms/cm3

Fig. 1. Comparison of calculated and experimental values of RSF for (a) positive and (b) negative ions of various elements sput tered by (a) oxygen and (b) cesium.

proportional to the ionization probability and concen tration of this element in the target: (10) I X = j 0 Y tot C X P X Tr X . Here, j0 is the current of primary ions, Ytot is the sput tering yield, and Tr is the transmission coefficient of the instrument for the given element. The ratio of these three factors in formula (9) was set at unity, while the concentration of all elements was assumed to be equal to the atomic density of silicon. After trivial transformations, we derived the following relations for RSF: IM IX + 22 RSF X = 5 10 exp kT ln ( E 0 /kT )
2 2 e eX M 4x c 4x c exp , kT ln ( E 0 /kT )

(11a)

AM AX 22 RSF X = 5 10 exp kT ln ( E 0 /kT )

2 2 (11b) e e X M 4x c 4x c exp , kT ln ( E 0 /kT ) where indices X and M indicate the values for the ele ments of interest and the matrix, respectively. The calculated and experimental [8] values of RSF for positive (Fig. 1a) and negative (Fig. 1b) ions of dif ferent elements sputtered by oxygen (Fig. 1a) and

cesium (Fig. 1b) from silicon are compared in the Fig. 1. The figure shows the experimental data for all ele ments represented in [8, Table E9, Appendix E]. The straight lines demonstrate equal values of RSF. In the case of positive ions, the values calculated for halogens considerably differ from experimental data, which is probably due to an additional mechanism of ioniza tion existing precisely for these elements. In the case of negative ions, the probability of ionization of elements with an electron affinity larger than 1.6 eV is unity [7]; for this reason, we excluded these elements from com parison. For the dynamic temperature, we used the values obtained in [4] for analogous regimes: 0.657 eV in the case of sputtering by oxygen and 0.187 eV for sputtering by cesium. Experimental values in [8] are averaged values of RSF for a large number of measurements performed on different instruments at different laboratories; for this reason, the experimental error attains 50%. Tak ing this into account, as well as the fact that the value of transmission factor of the instrument may differ considerably for different elements (which was disre garded in computational formulas (11)), the results represented in the figure lead to the conclusion about good agreement between the calculated and experi mental results. 3. CONCLUSIONS Thus, the formation of secondary ions during ion beam sputtering is described theoretically as a quasi adiabatic process; in this case, the statistics of electron distribution for an emitted atom obeys the Fermi Dirac law up to the critical ionization distance, and the temperature in the region of the collisions cascade of the primary ion with the surface is defined by the mean kinetic energy of recoil atoms. The expressions obtained for the ionization probability are in good agreement with experimental data.

TECHNICAL PHYSICS

Vol. 54

No. 11

2009

1570

KUDRIAVTSEV, ASOMOZA

ACKNOWLEDGMENTS This study was supported by CONACYT (Mexico), grant no. 48454. REFERENCES
1. L. N. Dobretsov, Tr. Fiz. Tekh. Inst. Akad. Nauk Uzb. SSR 3, 39 (1950). 2. P. Sigmund, in Sputtering by Particle Bombardment I, Ed. by R. Behrisch (Springer, Berlin, 1981), p. 35 3. M. W. Thompson, Philos. Mag. 18, 377 (1968).

4. Yu. Kudriavtsev and R. Asomoza, Nucl. Instrum. Methods Phys. Res. B 266, 3540 (2008). 5. R. Gomer, Solid State Phys. 30, 93 (1975). 6. Y. Kudriavtsev, A. Villegas, S. Gallardo, G. Ramirez, and R. Asomoza, Appl. Surf. Sci. 254, 3801 (2008). 7. Y. Kudriavtsev, A. Villegas, S. Gallardo, and R. Aso moza, Appl. Surf. Sci. 254, 2059 (2008). 8. R. G. Wilson, F. A. Stevie, and C. W. Magee, Secondary Ion Mass Spectrometry: A Practical Handbook for Depth Profiling and Bulk Impurity Analysis (Wiley, New York, 1989).

Translated by N. Wadhwa

TECHNICAL PHYSICS

Vol. 54

No. 11

2009