Characterization of Semiconductors by Mossbauer ... - Springer Link

__________________ 10 Characterization of Semiconductors by Mossbauer Spectroscopy Guido Langouche

1. Introduction In 1962, shortly after the discovery of the Mossbauer effect in 1958, two papers were published i ,2 on the characterization of impurity atoms in semiconductors by Mossbauer spectroscopy. Both dealt with 57Fe in silicon and germanium. In the 25 years that have passed since, more than 500 Mossbauer papers have been published on semiconductors. In Mossbauer spectroscopy experiments, the nucleus of the Mossbauer isotope acts as an extremely sensitive microscopic probe inside the semiconductor. From the study of the radiation emitted or absorbed without recoil by this nucleus, the symmetry and density of the surrounding electron configuration and the vibrational properties of the solid in which it is imbedded can be obtained. Extremely valuable and fairly unique information can be extracted in this way on the structural and electronic properties of the individual atom and the immediate surrounding lattice. In Mossbauer spectroscopy experiments, the position of the dopant atom in the lattice, the formation of complexes with defects or other impurity atoms, the charge state of the Mossbauer atom, and the changes in this charge state when the Fermi level is shifted can be studied. Also dynamic processes can be investigated, which are happening within the time frame of the lifetime of the Mossbauer level, which is typically 10-7 to 10-8 s. The method is especially attractive for the study of dopant atoms. In particular, in view of the fact that only the Mossbauer nuclei contribute to the signal, and, on the other hand, that all of them do contribute, Mossbauer spectroscopy compares favorably with some of the more common techniques Guido Langouche • Leuven, Belgium.

Instituut voor Kern- en Stralingsfysika, University of Leuven, B-3030

445 G. J. Long et al. (eds.), Mössbauer Spectroscopy Applied to Inorganic Chemistry © Springer Science+Business Media New York 1989

446

G. Langouche

in semiconductor research, which are often based on the study of properties of the bulk of the material or are restricted to atoms that are electrically active. The limits of the method should also be realized, however. The number of available isotopes offering sufficient spectroscopic resolution is rather limited, and these are not always the most common dopant atoms. Also the interpretation of the hyperfine spectra can sometimes be ambiguous. By far the largest part of the Mossbauer literature on semiconductors deals with the characterization of impurity atoms in group IV semiconductors and in the group III-V, II-VI, and IV-VI binary compound semiconductors. In this review, we will limit ourselves to this class of technologically important materials. We will therefore not discuss a number of special topics such as the chalcogenide glasses and their crystalline counterparts, semiconducting oxides, magnetic semiconductors, or ternary semiconductors. Reviews on Mossbauer spectroscopy of semiconducting glasses can be found in the recent literature. 3

2. The Use of Mossbauer Isotopes in Semiconductor Studies Before looking in detail into the scientific results obtained with Mossbauer spectroscopy, a few points should be mentioned concerning the application of the Mossbauer technique in this field. 2.1. Mossbauer Transitions

Although about one hundred Mossbauer transitions are known, few of them are used extensively. The ones encountered in the semiconductor literature are shown in Table 1. Mossbauer isotopes as constituent or dopant atoms have been studied in so-called "source experiments" as well as in "absorber experiments." In the first case, a radioactive atom which decays via a Mossbauer nuclear transition is introduced into the semiconductor and is studied in combination with a single-line absorber. In the second case, the Mossbauer isotope is introduced into the semiconductor in its stable ground state and studied in combination with an unsplit radioactive source. Although the same Mossbauer transition is studied, the Mossbauer spectra are often very different. In source experiments, the decay from the radioactive parent to the excited Mossbauer state happens in an extremely short time interval, usually a matter of nanoseconds, before the Mossbauer transition takes place in the actual Mossbauer experiment. The source preparation, most often by alloying, diffusion, or ion implantation, therefore takes place with the parent atom and is thus dominated by the physical and chemical properties of this element. As many parent atoms decay by f3-decay or electron capture, the parent atom is often a neighboring chemical element. Therefore, the lattice site occupied by the parent atom is

Characterization of Semiconductors

447

TABLE 1. Different Elements and Corresponding Mossbauer Transitions Used in Semiconductor Studies Element As Au Cd Co Cs Eu Fe Oa Oe I In Kr Pt Rb Sb Sm Sn Te Xe

Source experiment

Absorber experiment 197Au

119Cd ~ 119Sn 57 Co ~ 57Fe 133CS 151Eu,153Eu 57Fe 730a ~ 730e 1251 ~ 125Te 1191n ~ 119Sn 83Kr

119Te ~ 119Xe ~

197Pt ~ 197 Au 83Rb ~ 83Kr 119Sb ~ 119Sn, 125Sb ~ 125Te 153Sm ~ 153Eu 119mSn ~ 119Sn 119Sn, 121Te ~ l21Sb, 125mTe ~ 125Te, 129mTe ~ 1291 119Sn, l21Xe ~ l21Sb, 129mXe ~ 129Xe, 133Xe ~ 133CS

governed by the diffusion and implantation behavior of this element and does not change on the time scale of the nuclear decay, unless high recoil energies are given to the decaying atom. The electronic configuration of the atom, on the other hand, has rearranged itself under the influence of the newly formed nucleus in its center, on a time scale much faster than the lifetime of the excited Mossbauer state, so that isomer shifts and electric field gradients in source and absorber experiments can readily be compared. Thus, the measured isomer shift and the hyperfine splitting in a source experiment contain mixed information. They reflect the microscopic lattice configuration that existed around the parent atom, as well as the electronic configuration of the Mossbauer atom. The possible existence of "after-decay effects" in source experiments, in terms of unusual charge states or large recoil energies, has been discussed by Seregin et al. 4 for several Mossbauer transitions of interest in this work. Source experiments have two distinct advantages with respect to absorber experiments. First, the concentration of impurity atoms needed to obtain a measurable effect is typically several orders of magnitude lower than in absorber experiments. Second, the ability to study the lattice site of the parent atom allows many more elements to be studied than is possible in absorber experiments, given the limited number of Mossbauer isotopes. This advantage is exploited extensively in 119Sn work, where the Mossbauer resonance can be fed by many different parent atoms, so that cadmium, indium, tin, antimony,

G. Langouche

448

tellurium, and xenon can be studied using the same resonance. The relevant decay schemes are shown in Figure 1. In the literature on Mossbauer spectroscopy on semiconductors, 1l9Sn is the Mossbauer resonance for which most data are available and also for which the Mossbauer data are best understood. Most 1l9Sn spectra are studied in the conversion electron Mossbauer spectroscopy (CEMS) geometry and generally consist of a number of overlapping resonances, some of them broadened by a quadrupole interaction. Nevertheless, 1l9Sn Mossbauer spectra showing not much more than an asymmetric broad resonance are nowadays analyzed with great confidence into a number of components with well-defined isomer shifts and linewidths. This procedure is based on a vast amount of systematic studies, using the different parent isotopes, which has allowed the assignment of these components to particular lattice configurations around the Sn atom. 2.2. Doping the Semiconductor with Mossbauer Isotopes

In only a few cases, the Mossbauer dopant atoms have been introduced from the melt. A more common way of doping the semiconducting material with Mossbauer isotopes is by diffusion at elevated temperatures and subsequent slow cooling or rapid quenching. Depending on the concentration, solubility limit, and quenching speed, various configurations result: dilute atoms on regular or defect sites, clusters, large precipitates, and even amorphized lattices are obtained. In semiconductor technology, ion implantation has become the most widely used technique in the controlled doping of surface layers. It has been

,~~5.8m

I!'.Ei 54 X•

.~7-i;:- 19.1m

If. E7

1112~4.7d

~~I

;;1~T• ~ 52

EC

o FIGURE 1. Different parent isotopes feeding the 119Sn resonance. (From reference 5.)


449

applied also in many Mossbauer studies, both with stable Mossbauer isotopes and radioactive parent atoms. In general, only parent isotopes with lifetimes shorter than one year are used in ion implantation processes. It is known that ion implantation of medium and heavy elements into semiconductors at room temperature leads to amorphization of the lattice. In the single-track amorphization model, this already happens for individually implanted atoms. The lattice position of the primary implanted atom with respect to this amorphized region is hard to predict, as it is governed by statistical laws. There is a large probability for the atom to be caught in an amorphous network. It might also find itself outside this amorphized track or in a recrystallized part of it. For increasing implantation doses, these tracks gradually overlap. For typical room temperature implantation energies around 100 keV and for medium-mass atoms, the total amorphization dose is _10 14 atoms/cm2. Implantation at elevated temperatures is known to prevent amorphization of the lattice but might promote dopant diffusion and precipitate formation. The studies reported in this review involve implantation doses in the range between 10 10 and 1018 atoms/ cm2 at different temperatures, so that various lattice configurations of the implanted atoms have to be considered.

3. Group IV Semiconductors More than 95% of the Mossbauer literature on the doping of semiconductors deals with only four Mossbauer resonances: the 5sp elements 119Sn, with its many different parent isotopes, 125Te, and 1291 and the 3d transition element 57Fe. We will discuss this literature according to the parent element because, as explained in Section 2.1, this element determines the lattice site occupied by the Mossbauer atom.

3.1. Tin 3.1.1. Alloys The fact that a-Sn itself is a group IV semiconductor suggests, of course, that Sn impurities will be easily incorporated into Si or Ge. This is indeed the case for Ge, as the room temperature solubility of Sn in Ge is extremely high (almost 1021 atoms/cm3 ). The room temperature solubility of Sn in Si is substantially lower (5 x 1019 atoms/cm 3 ). The diffusivity of Sn is very low, however, so that supersaturated solutions can be obtained by various techniques. The properties of Sn in such solutions were the subject of several Mossbauer studies. In early experiments,6,7 alloys such as Sn x Ge 1- x were studied. The 119Sn spectra indicate the presence of two components. One of them has an isomer shift close to that of metallic Sn, indicating the presence of f3-Sn precipitates.

450

G. Langouche

The other line has an isomer shift close to that of a-Sn and was assigned to a solid solution of Sn in Ge. 7 Bakhchieva et al. 8 doped Si and Ge crystals from the melt with Sn concentrations in the range from 6 x 10 18 to 6 x 1020 atoms/ cm3. For concentrations lower than 2 x 10 19 atoms/ cm\ single-line 119Sn spectra (8 = 1.80 mm/ s) were obtained with the isomer shift close to that of a-Sn. In this work, we will cite all 119Sn isomer shifts as absorber isomer shifts with respect to the BaSn03 standard, unless otherwise specified. This resonance was assigned to substitutional Sn atoms. For higher concentrations, the spectra were analyzed in terms of the substitutional single line and an extra quadrupole doublet (8 = 2.15 mm/s, a = 0.70 mm/s), which was assigned to small Sn-Sn complexes. For the highest concentrations (> 5 x 1020 atoms/ cm3) above the solubility limit, an additional single line (8 = 2.60 mm/s) was observed and assigned to f3-Sn precipitates. 3.1.2. Diffusion

Diffusion of radioactive 119mSn in silicon was reported in 1975 independently by two groups.9-11 A single line with isomer shift 8 = 1.85(5) mm/s was assigned to substitutional Sn in Si. This value of the isomer shift is close to that of a-Sn. Seregin et al. 11 used both p- and n-doped Si and observed no difference in isomer shift. This lack of Fermi-level dependence is consistent with the lack of electrical activity of the isoelectronic substitutional impurity. Bonchev et a/. 12 later reported the formation of Sn2Si upon diffusion of Sn in Si. Upon increasing the concentration of diffused Sn in Si above -2 x 10 19 atoms/cm3, an additional quadrupole doublet (8 = 2.15 mm/s, .:l = 0.70mm/s) was observed by Seregin et al.13 and attributed to Sn complexes. For concentrations above 1020 atoms/ cm 3, the single line of metallic tin (8 = 2.60 mm/ s) appeared. Bakhchieva et al. 8 doped Si and Ge samples with 119mSn by diffusion with surface concentrations of up to 7 x 10 19 atoms/ cm 3 in Si and 2 x 1020 atoms/ cm3 in Ge. The spectra consisted of single lines with the 1l9Sn isomer shift similar to that of a-Sn and were therefore assigned to substitutional Sn. No difference was observed between p- and n-doped samples. 3.1.3. Ion Implantation

In 1974 Weyer et al. reported l4 the first Mossbauer study on ion-implanted 119mSn in Si. The implantation dose was varied between 10 13 atoms/ cm2 and 10 17 atoms/ cm2, and the implantation temperature was also varied from room temperature to 40~oC. The implantation energy was of the order of 80 keV. At high implantation temperatures, a single-line resonance was observed with 8 = 1.86 (6) mm/s. Channeling measurements confirmed 100% substitutionality. 119mSn was also implanted in the other group IV semiconductors/ 4 and


451

Antoncik l5 - 17 analyzed the host dependence of the measured isomer shifts in terms of the electronic configuration of the Sn atom. The interpretation of 119Sn isomer shift data was refined in further theoretical treatments lS - 22 by the same author. The correlation between the isomer shift and the host lattice constant was analyzed in terms of an increased dehybridization of the Sp3 configuration. Room temperature implantation was found to lead to a heavily damaged host/ 3,24 as evidenced by the loss of channeling behavior of the host. The Mossbauer spectra were hardly affected, however, indicating that the local environment around the implanted Sn atom remained perfectly tetrahedral. 25 High-dose (>10 16 atoms/cm2) implantation slightly increased the isomer shift of the substitutionalline26 due to amorphization of the lattice (discussed below). Annealing of these supersaturated samples resulted in the precipitation of metallic Sn. A two-step annealing process, including a rapid thermal annealing step, was proposed by Scherer et al. 27 for high-dose (10 16 atoms/ cm2) 119Sn_ implanted Si. By this process a complete regrowth of the amorphized layer could be reached, as well as a complete substitutional solid solution of 119Sn, for a concentration two orders of magnitude higher than the solid solubility limit. This was evidenced by channeling and Mossbauer data. Rapid thermal annealing was also used by Weyer et s in order to obtain substitutional supersaturated solutions of Sn in Si, up to implantation doses of 5 x lOiS atoms/cm 2.

ae

3.1.4. Defects

Several methods have been used to generate defects in the semiconductor host, in order to be able to study impurity-defect complexes. Matsui et al.,9 whose report on the diffusion of 119Sn was mentioned in Section 3.1.2 irradiated their sample with neutrons. The appearance of the resulting spectrum depended on the doping of the sample: a quadrupole doublet [8 = 1.85 (5) mm/s, tl = 1.08 (10) mm/s] resulted in the case of n-type silicon, but a single line [8 = 1.50 (5) mm/s] was obtained for p-type silicon. They were interpreted as Fermi-level-dependent Sn-vacancy pairs. Their assignment is not consistent with the extensive work, cited below, on ionimplanted samples. A similar quadrupole splitting but a quite different isomer shift (5 = 2.12 mm/s, tl = 0.98 mm/s) were measureds,29 after neutron transmutation doping of 119Sn-doped Si and were also assigned to a Sn-vacancy complex. Quenching from high temperatures also results in a side resonance, II a single line with 5 = 2.44 (5) mm/s, attributed by Seregin et al. to interstitial Sn. All these tentative assignments are not consistent with the extensive work done on ion-implanted samples from different parent isotopes. Postirradiation of an ion-implanted sample, showing only substitutional Sn, with a-particles,30

G. Langouche

452

in order to produce point defects in Si, gave different results, depending on the irradiation temperature, as shown in Figure 2. No change occurs after room temperature irradiation, but a-irradiation at liquid nitrogen temperature resulted in the appearance of substantial amounts of a side resonance (8 = 2.3 mm/s, .:1 = 0.35 mm/s). From a comparison with data from other parents, this resonance was identified as a defect line, associated with a Sn-vacancy pair. The S2p2 configuration is proposed for such an atom, with a dangling bond towards a neighboring vacancy, which leads to an increase in the isomer shift. Also its Debye temperature was found to be 20% smaller than for the substitutional line: aD = 230 (10) K for substitutional Sn versus aD = 188 (20) K for the Sn-vacancy pair. The Debye temperatures of Sn in Si have been discussed in terms of the Mannheim modee 1 and using an adiabatic bond-charge model. 32 In room temperature ion implantation of 119mSn into Si and Ge, no defect lines were reported, except for a slight change in isomer shift due to the a

.......

~

z z

~

u a::

~

~

::I

b

0 ~

>-

~

iii

100

z

I!! 3!:

-4

-3

-2

-I

0

VELOCITY (mm/s)

2

FIGURE 2. 119Sn Mossbauer spectrum of 119mSn implanted at room temperature in Si at a dose of 1014 atoms/cm 2 : (a) after a-irradiation at room temperature; (b) after a-irradiation at 77 K. (From reference 30.)


453

amorphization of the lattice. 25 Implantation of 119mSn into a-Sn at room temperature, however, resulted33 in three extra resonances in the Mossbauer spectrum. A small fraction of the 119Sn spectrum belongs to a resonance with isomer shift 8 = 3.5 mm/s, which was assigned to Sn atoms in an interstitial lattice position. Two more spectral components were related to oxygen. A quadrupole-split component [8 = 0.0 (1) mm/s, tl = 0.5 (1) mm/s1 was assigned to Sn02 due to implantation in the surface oxide. This component was especially intense in polycrystalline a-Sn samples and, accompanied by a new single line [8 = 3.95 (5) mm/s1 which was assigned to a complex defect structure involving oxygen atoms that were recoil implanted from the surface oxide layer into the bulk.

3.1.5. Amorphous Hosts The different Mossbauer studies on 119Sn in amorphous silicon all agree on the experimental data; a broadened resonance; with a position close to that of 119Sn in c-Si, is measured. This resonance, however, is analyzed and interpreted in different ways by different authors. Amorphous silicon was prepared by Andreev et al. 34,35 by thermal evaporation of 119Sn-doped Si. A single broadened line was measured, with an isomer shift [8 = 1.94 (2) mm/s1 close to that of substitutional Sn in crystalline Si, but with a severe line broadening [f = 1.45 (3) mm/s instead of f = 0.80 (3) mm/s1. The authors state that this broadening can be analyzed both with a quadrupole interaction (tl = 0.55 mm/s) and with a Gaussian distribution (with width f = 0.40 mm/s) of the isomer shifts. It was therefore concluded that the Sn atoms have the same tetrahedral lattice configuration as in c-Si, with the broadening due to distortion of the tetrahedral angles and to the stabilization of some Sn atoms with dangling bonds in unusual charge states. Saturation of these dangling bonds with hydrogen was thoughe 6 to be responsible for a shift of the Mossbauerresonance to lower velocities (8 = 1.65 mm/s). A similar small increase [0.15 (3) mm/s1 in isomer shift was observed after ion implantation of 119mSn in a_Si/ 7 ,38 with a line broadening of only 20%, and with the same Debye temperature as in c-Si. The line broadening is thougHt to be due to a sum of lines with different isomer shifts. The fact that the Debye temperature is the same in c-Si and a-Si is consistent with the same tetrahedral lattice configuration around the Sn impurity, which results in very similar vibrational properties around Sn in c-Si and a-Si on a microscopic scale. Williamson and co_workers 39,4o studied the doping of amorphous hydrogenated Si with Sn. In contrast with the previous studies, only a small difference in isomer shift «0.05 mm/s) was derived between substitutional Sn in c-Si and a-Si: H. Rather than attributing the observed line broadening to a distribution in isomer shifts only, a distribution of electric field gradients [average tl = 0.46 (5) mm/ s1 together with an accompanying distribution of isomer shifts

454

G. Langouche

was invoked. No nearest-neighbor bonding between Sn and H was observed. It is possible to incorporate a large concentration of Sn in a-Si: H, in sharp contrast with the very low solubility of Sn in c-Si «0.1 at. %). From the Mossbauer data it can be concluded, however, that some of the Sn is found in precipitates of metallic Sn, Sn02, and SnO. Nasredinov et al. 41 interpreted the observed line broadening as a bandstructure effect. Sn in a- Si is believed to produce an acceptor band of localized states near the valence band, to which the Fermi level is pinned. Instead of being an isoelectronic neutral impurity as in c-Si, some Sn atoms form charged Sn centers and produce experimentally observed changes in the electronic and optical properties of a-Si. These charged Sn centers have different isomer shifts and hence contribute to the observed line broadening. Amorphous Si 1 - x Snx alloys were studied42,43 by 119Sn Mossbauer spectroscopy. It was found that homogeneous amorphous alloys of Si and Sn can be formed up to a Sn concentration of 50% without precipitation of metallic Sn. It is remarkable that for all concentrations a single Lorentzian line is observed with an average linewidth of r = 1.16 (2) mmls which can almost completely be accounted for by the thickness of the sample. The isomer shift changes linearly as a function of composition, from 8 = 1.89 mmls for Sn in pure a-Si to 8 = 2.05 mmls for Sn in amorphous a-Sn. This behavior was ascribed to a systematic decrease in the number of p-like electrons in covalent bonds, upon an increase in tin concentration. 15 ,24 From this behavior it is concluded that Sn atoms are substituted for Si atoms on a random continuous network, where, however, each tin atom tends to be selectively surrounded by silicon atoms only, in almost perfect tetrahedral units. As almost natural linewidths were observed in this study on amorphous Si doped with Sn, it is tempting to conclude that the line broadening observed in other studies is-due to isomer shift distributions connected with Sn concentration gradients in the sample.

3.1.6. Laser and Ion Beam Irradiation In 1980 two groups reported laser annealing and laser implantation experiments on Si samples doped with Sn. Altudov et al. 44 implanted 119Sn in Si at a dose of 1015 atoms/cm 2 and irradiated this sample with a 20-ns ruby laser pulse of 1.2 J/cm 2 • A substitutional supersaturated solution of Sn in single-crystal Si was obtained. Laser implantation of a Si sample covered with a O.I-#Lm Sn film with a 1.5-J/cm2 light pulse of the ruby laser also gave rise to substitutional Sn in single-crystal Si. Damgaard et al. 45 ,46 reported similar laser implantation experiments with a neodymium laser and formed supersaturated solutions. The number of incorporated Sn atoms for different film thicknesses was studied.


455

Prasad et al. 47 studied the process of ion-beam mixing in the Sn-Si system by 119Sn Mossbauer spectroscopy and Rutherford backscattering. A 30- to 45-nm-thick metallic film of Sn was deposited on Si and bombarded with Ar ions of different energies and doses. In the example shown in Figure 3, about 40% of the Sn contributed to the formation of a tin silicide containing 20 to 50% Sn. The rest of the Sn is detected as metallic tin. 3.2. Antimony As can be seen in Table 1, antimony can be studied from two parent states, 119Sb and 125Sb, as well as in a I21Sb absorber experiment. All three have been used to study Sb impurities in group IV semiconductors. 3.2.1. Alloys I21Sb-doped Si single crystals, from a commercial supplier, were studied by Teague et al. 48 in a concentration range of 3 x 10 18 atoms/ cm 2 to 3 x 10 19 atoms/ cm 2 • Up to 10 19 atoms/ cm 2 , a single line was observed with a linewidth of 2.5 mm/ s and an isomer shift varying from 1.32 mm/ s to 0.68 mm/ s, with respect to InSb. The authors believe the Sb atoms to be incorporated substitutionally and invoke a lattice expansion to account for this concentration dependence. 3.2.2. Implantation The first Mossbauer study on ion-implanted 119Sb in Si was reported by Weyer et al. in 1974/ 4 where 80-keV 119Sb atoms were implanted at 400°C, at a dose of 5 x 10 14 atoms/ cm 2 • A single line was observed in the 119Sn Mossbauer spectrum, with an isomer shift of 1.9 (1) mm/ s, and therefore 119Sb was believed to be 100% substitutional. More extensive studies were reported later/ 5,49,5o and four independent lines could be recognized 51 in the 119Sn spectrum oflow-dose room temperature 119Sb-implanted Si. These lines were characterized by the measured isomer shifts, linewidths, quadrupole splittings, and Debye-Waller factors. A large fraction of the Sb atoms occupy substitutional lattice sites [8 = 1.83 (6) mm/ s, 0 D = 220 (20) K]. A minor fraction is found in interstitial lattice sites [8 = 3.3 (1) mm/s, 0 D = 240 (30) K]. The two other lines are assigned to Sbvacancy complexes: one line [8 = 2.6 (1) mm/s, Il = 0.3 mm/s, 0 D = 160 (20) K] to Sb atoms with a dangling bond towards an adjacent vacancy, and another line, with the same isomer shift, but with no quadrupole interaction and a much higher Debye temperature [0 D = 250 (40) K], to Sb atoms in an extended vacancy complex. The former complex anneals out between 700 and 900 K. The latter one is only formed after high-temperature treatment (> 1200°C) of the sample and is probably associated with vacancy loops. The

G. Langouche

456

119Sn ON Si

1·24

AS EVAPORATED

1·20

a

1·16 1·12 1·08 1·04

----

1·00

EAr.: 130 KeV DOSE: 2. 1016/cm2

1·24 ~

I-

1·20

b

(f)

Z

w

1·16

Z

1·12

I-

EAr. : 130 KeV

1·24

DOSE: 5. 1016/cm2

1·20

c

1·16 1·12 1·08 1·04 1·00 +5·0

+4·0

+3·0

+ 2·0

+1·0

0·0

VELOCITY (mm/s) FIGURE 3. 1I9Sn Mossbauer spectra of an ion-beam-mixed Sn-Si layer, starting from a 45-nmthick 1I9Sn layer evaporated on [100] Si. (From reference 47.)


457

interstitial fraction increases with annealing temperature, probably due to the trapping of the Sb atoms in agglomerates with Si interstitials. A similar study was reported 52 for 1I9Sb implantation in Ge. Here only three sites were recognized: the dominating substitutional one (8 = 1.96 mm/s, 0 D =189K), the interstitial one (8=3.40mm/s, 0 D =149K), which increases strongly for annealing temperatures above 500°C, and antimonyvacancy complexes (8 = 2.55 mm/s, r = 1.24 mm/s, 0 D = 220 K), which start to anneal out above 300°C. In 1I9Sb-implanted a_Sn,33 four sites could be recognized: substitutional, interstitial, vacancy-associated Sb, and, as also observed in the 119m Sn implantations in a-Sn, an oxygen-related defect site. Postirradiation of 119Sb_doped Si, at liquid nitrogen temperature, with a-particles results in an increase of the number of Sb-vacancy complexes. 3o Rapid thermal annealing with an incoherent light source was used 53 ,54 to obtain concentrations exceeding solid solubility. The equilibrium solubility of Sb in Si is comparable to the solubility of Sn in Si. It was found that in 1I9Sb-doped Si samples, implanted to 5 x 10 15 atoms/ cm2 and subjected to rapid thermal annealing at 700°C, 95% of the Sb atoms are found in substitutional sites. This corresponds to a metastable, electrically active, substitutional solution with a concentration of 5 x 1020 Sb atoms/cm3. The remaining 5% is found in Sb-vacancy complexes. Further annealing at temperatures 900°C, Sb precipitates are formed. In the present high-concentration study, the velocity region between 2 and 3 mm/ s is analyzed in a different way than in the previously mentioned study51 where two defect sites, with the same isomer shift, due to two different Sb-vacancy complexes, were proposed. Two resonances are now found in this velocity region. One line [13 = 2.32 mm/ s, r = 1.10 (10) mm/s, 0 D = 183 (10) K] is associated with Sb atoms with a dangling bond pointing to an adjacent vacancy. The other line [8 = 2.74 (5) mm/s, r = 1.18 (10) mm/s, aD = 237 (10) K] is interpreted as stemming from Sb in antimony precipitates. It is found that the Sb going out of solution is predominantly found in Sb-vacancy complexes for low doses and low annealing temperatures and in Sb precipitates for high doses and high annealing temperatures. An interesting observation arises when the results from these experiments are compared to those from channeling and electrical activation measurements. Complete agreement is found between the electrical and Mossbauer data. The channeling data, however, seriously overestimate the substitutional fraction after highertemperature annealing. It is not clear to which of the two defect structures, as measured by Mossbauer spectroscopy, the 25 to 30% overestimate in substitutional fraction, as measured by channeling, belongs. Scherer et al. 27 comment on the previous data and propose the Sb-vacancy complex as being the fraction seen as nonsubstitutional by Mossbauer spectroscopy, but as substitutional by channeling. In their 119Sn-implanted and

458

G. Langouche

annealed Si samples no precipitate component was seen by Mossbauer spectroscopy. A similar overestimate in substitutional fraction was observed by channeling measurements as compared to the Mossbauer data. A comparison has been made 28 between the annealing behavior of highconcentration Sn and Sb implanted in silicon. While in the Sn-implanted samples the Sn atoms are found invariably on substitutional lattice sites (at least up to 1050°C annealing temperature), considerable fractions of Sb leave the substitutional site after 700°C annealing. It is concluded that the coulomb interaction between positively charged Sb donors and negatively charged vacancies is the natural driving force for the reaction in which Sb is going out of solution and becomes electrically inactive by the formation of Sb-vacancy complexes. Hence, the Sb migration is mediated by vacancies. The third Mossbauer transition that can be used in Sb studies is the 125Sb ~ 125Te transition. The 125Te Mossbauer transition, however, has a very poor resolution, due to the very large intrinsic linewidth. On the other hand, due to the large quadrupole moment of the excited state, it is much more sensitive to the presence of quadrupole interactions. When 125Sb (half-life, 2.7 years) was implanted (10 14 atoms/cm 2) in group IV semiconductors,s5,56 only one somewhat broadened single-line component [8 = 0.20 (6) mm/s, r = 8.2 mm/s] was observed in the 125Te Mossbauer spectrum and was assigned to substitutional Sb. The 125Te isomer shifts will be quoted in this work as absorber isomer shifts with respect to ZnTe, unless otherwise specified. The linewidth in this experiment is about 25% larger than expected, so that clearly other sites with slightly different isomer shifts must be involved. The Mossbauer resonance, on the other hand, is not quadrupole split, as observed, for example, for the implanted 1251 and 125mTe parent leading to the same Mossbauer transition. For these parents an off-substitutional site was proposed in asimplanted samples, which is clearly not the case for the Sb parent.

3.2.3. Laser Annealing Laser annealing 56 of the 125Sb-implanted Si sample with a Q-switched ruby laser with an energy density of 1.8 J / cm2 changed the parameters of the Mossbauer resonance to 8 = 0.41 (9) mm/s and r = 6.5 (3) mm/s. This is the expected linewidth for a single-line resonance for an atom on a defect-free site. It is therefore Concluded that these parameters correspond to purely substitutional Sb in Si.

3.3. Indium Radioactive 1191n ions have a hal"r-life of only 2.1 min and were obtained by Weyer et al. 57 ,58 from the ISOLDE facility at CERN and implanted in various targets.


459

When implanted into silicon58 with an implantation dose of < atoms/ cm2, five independent lines were recognized in the 119Sn Mossbauer spectra. Four of them were discussed before: the substitutional In [8 = 1.80 (3) mm/s, 0 v = 250 (15) K], the In atom with a dangling bond towards an adjacent Si monovacancy [8 = 2.35 (8) mm/s, 0 v = 198 (20) K], the interstitial In [8 = 3.46 (15) mm/s, 0 v = 160 (25) K], and In associated with oxygen [8 = 0.0 (2) mm/s, 0 v = 334 (5) K]. The fifth line [8 = 1.35 (7) mm/s, 0 v = 178 (20) K], according to Nylandsted-Larsen et al.,s9 belongs to Sn atoms in interstitial sites in the center of a divacancy with hybridized d 2sp3 bonds attached to six nearest neighbors. The interstitial line has a lower Debye temperature than after the previously mentioned Sb implantations. This might suggest that here isolated interstitial impurities are obtained, in contrast to Sb (and Te) implantations, where the interstitials probably are in extended agglomerates. Also the vacancy trapping process is different for Sb and In impurities, leading to different properties of the vacancy-associated site (8 = 2.35 m/s), especially in its annealing behavior. This can be attributed to a different impurity-vacancy coulomb attraction. After implantations into Ge, the same lines were recognized,57 except for the oxide line, which is missing. A comparison between the data from the different parents was made in a number of review papers. 52 ,60,61 10 12

3.4. Cadmium 119Cd is also produced at ISOLDE (CERN), has a half-life of only 2.7 min, and decays to 119ln, which has a comparable half-life before decaying to 119Sn. It is therefore not surprising that in the reported experiment60 the results obtained with the implanted Cd parent (implantation dose < 10 12 atoms/ cm 2) are very close to the results obtained with the In parent.

3.5. Tellurium Extensive data have been reported on the behavior of Te in group IV semiconductors. This is partly due to the fact that four different parent isotopes as well as one absorber isotope can be used (see Table 1). Another reason is that some surprising new facts were observed such as a Fermi-level-dependent isomer shift and a dynamic behavior of Te in a Jahn-Teller-distorted site. We will discuss the data from the 119Te isotope and from the 125mTe and 129mTe parents separately because some disagreement exists between the results of these studies.

O. Langouche

460

3.5.1. 119Te Parent Because of the radioactive chain 119Te (4.7 days) ~ 119Sb (38 h) ~ Jl9Sn, one can study the annealing properties of similar defect configurations associated with both the Sb parent and the Te parent. When the implanted 119Te is in radioactive equilibrium with its decay products (>60 h after the implantation), two types of annealings can be performed. If the annealing time and the measuring time following the annealing are short compared to the 119Sb half-life, the measured radiation comes from decaying atoms that have been annealed as Sb atoms since most of the decaying Sb atoms were formed before the cool-down. On the other hand, if the measurement is performed after a long time compared to the Sb half-life, then the radiation stems from atoms that have been annealed as Te atoms, since most of the annealed Sb atoms have now decayed. After a number of partial initial reports,14,25,49,50 a very extensive study was published by Nylandsted-Larsen et al. 59 In this study five different lines were recognized in the 119Sn spectrum. They are given in Table 2, together with their isomer shifts, their Debye-Waller factors, and their populations for an implantation at 5 x 10 14 atoms/cm 2 at room temperature and at 723 K. Figure 4 shows the measured Mossbauer spectrum for a 723 K implantation temperature. From the table it is clear that the fraction of substitutional Te atoms increases with increasing implantation temperature. Annealing up to 1200 K did not change the Te-atom location substantially. Samples prepared under the same conditions were also investigated by 125Te and 1291 Mossbauer spectroscopy and will be discussed below. Petersen et al. 33 studied the somewhat special case of 119mTe implantation in a-Sn. Five sites are populated: substitutional, interstitial, the Snmonovacancy pair, a Sn-oxygen complex, and a line which has the same isomer shift as 119Sn in Sn02 and, therefore, is assigned to the oxidized surface l~yer.

Kemerink et al. 62 studied the effects oflaser annealing on 119Te-implanted Si with doses between 1013 and 10 14 atoms/cm 2. A single-line resonance was TABLE 2. Line Assignments and Site Populations (P) after a 119Te Implantation (5 x 10 14 atoms/cm2 ) in Si at Room Temperature (RT) and at 723 K U

a b

eD

8b (mm/s)

P(RT)

P(700K)

(K)

(%)

(%)

Assignment

0.92 1.84 2.61 3.30 4.41

230 250 165 250 175

1 27 60 3

6 68 12 2

8

11

Sn + divacancy Substitutional Sn + monovacancy Interstitial Oxygen complex

From reference 59. Isomer shift with respect to BaSn03'


461

8.0 75

::; w z z c

70 65

::J:

u

a:: w

6.0

en

5.5

no

...z

.'

:;)

0

u

6.0

...iii z ...w!

SO

"5i ~

4.0 3.0 2.0 1.0 3.0

20 VELOCITY (mm/sl

FIGURE 4. 119Sn Miissbauer spectra, measured at room temperature, of 1I9mTe and 1I9mSn implanted in Si at 723 K at a dose of 5 x 10 14 atoms/cm 2 • (From reference 59.)

observed, with the same isomer shift in n-Si and p-Si, and with a slightly broadened linewidth [f = 1.12 (3) mm/s]. This excludes defect association to Te in laser-annealed Si and gives evidence again that Sn occurs in only one charge state. 3.5.2. 125Te

The 125Te as well the 129 1 Mossbauer spectrum of Te-implanted Si is dominated by two resonances. For 125Te Mossbauer spectroscopy, this can give rise to an interpretation problem, as a doublet spectrum from a ~ ~ ! nuclear transition can be interpreted as two single-line resonances (two regular sites) or as one quadrupole doublet (one noncubic site). For 1291 Mossbauer spectroscopy, with its ~ ~ ~ transition, a quadrupole interaction gives rise, in general, to a 12-line spectrum, so that, in principle, no ambiguity can exist. However, the different spectrum components are grouped more or less in two groups with some weak side resonances. A distribution in electric field gradients andlor isomer shifts can smear out these side resonances, in such a way that a two-line spectrum results, which can be misinterpreted as a spectrum with two independent single lines. Such a misinterpretation existed in the analysis of both the 129 1 and the 125Te Mossbauer spectra after Te implantations in group IV semiconductors, published before 1979,55,63-68 until a laser annealing experiment69 led to a reinterpretation of the spectra.

G. Langouche

462

The 125Te doublet observed55 .65 after implantation of 125mTe in Si and Ge was originally analyzed as two single-line resonances, following the analysis of the 129mTe spectra.63 This, however, led immediately to problems, which were pointed out by the authors. 65 The isomer shifts of these two single lines were completely outside the range of previously observed values. The corresponding very large and very small electron density at these two sites (substitutional and interstitial?) was very puzzling. A second puzzling result55 was that the 125Te Mossbauer spectrum resulting from l2SSb implantation, which was believed to give rise to the population of a substitutional site only, indeed gave rise to a single-line spectrum, as discussed above, but its isomer shift did not correspond to either of the two observed resonances after 125mTe implants. De bruyn et aCo.7l showed that laser annealing changed the doublet spectrum into a single line [8 = 0.42 (5) mm/s], as shown in Figure 5. Since the isomer shift was the same as observed after l2SSb implantations, Te was concluded to be substitutional after laser annealing, in agreement with channeling data. Thermal annealing, on the other hand, did not alter7l the doublet nature of the spectrum. It was concluded that Te atoms in as-implanted and annealed Si are residing on a unique site, characterized by a large quadrupole interaction, which has, therefore, lower than cubic symmetry. The observed quadrupole splitting and isomer shift of the doublet [8 = 0.24 (4) mm/ s, a = 4.55 (7) mm/s] were shown to be consistent7l with the values observed after 129mTe implants. It is somewhat puzzling that the 119Te experiments are interpreted as giving evidence for the existence of five different sites (Table 2) with strong differences in site population as a function of implantation temperature. The 125mTe experiments on the same system, however, are analyzed in terms of a single defect site. Also the 129mTe implants can be analyzed in terms of a single 181fcm2

-15

-10

10

15

VELOCITY (mmlsl

FIGURE 5. 125Te Mossbauer spectra of 125mTe implanted at room temperature in Si at a dose of 7 x 1014 atoms/ cm 2: as-implanted (bottom) and after laser annealing (top). (From reference 71.)


463

defect site. This discrepancy may partly be due to the fact, which has already been mentioned, that 119Sn Mossbauer spectroscopy is not very sensitive to quadrupole interactions, while 125Te and 1291 Mossbauer spectroscopy is very sensitive to quadrupole interactions. On the other hand, due to its inherently large linewidth, 125Te Mossbauer spectroscopy is unable to resolve components with slightly different isomer shifts. Some agreement between the 119Te and the 125Te analysis can be reached by assuming that in the 119Sn analysis the contribution from 119Te atoms on defect sites was underestimated, while, on the other hand, it is perfectly possible to include into the 125Te analysis a substantial (10 to 20%) single-line contribution without seriously affecting the quality of the fit. Another possibility is that the substitutional site for the Te atoms in as-implanted and thermally annealed samples is distorted. Kemerink et al.72 observed such a distorted substitutional site, due to a Jahn-Teller-like distortion of the site of substitutional iodine atoms in n-Si after laser annealing. Such a distortion would hardly be detected in the 119Sn work, but it will be very obvious in 125Te and 1291 spectroscopy. A direct proofforthe quadrupole doublet character of the 125Te Mossbauer spectrum was recently obtained in a low-temperature nuclear orientation experiment. 73 At sufficiently low temperatures, the hyperfine levels of the parent state have unequal Boltzmann populations, which are transferred to the excited Mossbauer state. This gave rise to a large asymmetry in the doublet, offering direct evidence for its quadrupole-split character. Dezsi et al. 74 •75 found a striking similarity between the Mossbauer parameters of 125Te in as-implanted Ge and the parameters of 125Te in amorphous germanium telluride. They suggested therefore that in the ion implantation process single-track amorphization occurs, giving rise to a local symmetry around the Te atom as in a-Ge x Tel-x' This atomic configuration around the

Te atom appears to be so stable that it persists even after thermal annealing and is only broken up by laser annealing. The microscopic nature of this atomic configuration is not known for sure, but a threefold coordination, due to a dangling bond towards a missing nearest neighbor, is proposed for Te in Ge as well as in Si. 76 Van Rossum et al. 77 calculated the electric field gradient that would result from such a microscopic model, using an extended Huckel approach. They found that the displacement from the substitutional site towards the neighboring vacancy, which is needed to account for this electric field gradient, was giving rise to interatomic Ge-Te and Si-Te distances that were exactly equal to the sum of the covalent radii. Kemerink et al. 62 studied the Fermi-level dependence of 125mTe_implanted Si after laser annealing. Depending on the doping of the samples, two different charge states of the Te atoms were identified. In n-Si a single line [5 = 0.35 (5) mm/s with respect to ZnTe, @D = 207 (3) K] was attributed to Teo. In p-Si different parameters were measured [5 = +0.07 mm/s, @D = 232 (3)] and assigned to Te 2+.

G. Langouche

464 3.5.3.

129[

We mentioned in the preceding section that the absence of side wings in the 1291 Mossbauer spectrum after implantation of 129mTe into group IV semiconductors led to a two-site (substitutional and interstitial) model for the lattice location of implanted Te. 55 ,63-68 De bruyn et al. 56 ,69-71 showed that laser annealing of a 129mTe sample with a 1.8-JI cm2 pulse from a ruby laser results in dominant single-line resonance, which on the basis of channeling data has to be attributed to substitutional Te. The previous data were reinterpreted and the two resonances in the as-implanted and thermally annealed 129 1 Mossbauer spectrum were considered to be a quadrupole multiplet [5 = 1.42 (7) mmls (absorber isomer shift with respect to Cu 129I), eQVzzl h = +560 (20) MHz, T/ = 0]. The side resonances in the spectrum were smeared out due to some distribution in the hyperfine interaction parameters. The existence of a quadrupole interaction in asimplanted samples was confirmed by a perturbed angular correlation experiment. 78 One thus has to conclude that immediately after implantation most of the Te atoms occupy positions with lower than cubic symmetry. This is consistent with the 125Te data. The discrepancy with the 119Sn data was discussed above. As for 125Te, Dezsi and co-workers 79-81 compared the 1291 data from as-implanted 129mTe in Ge with the Mossbauer data from 129mTe in amorphous GeTe. Also here a similar chemical bonding seems to occur for the Te implanted in Ge [5 = 1.02 (5) mmls, eQVzzl h = +480 (5) MHz, T/ = 0,8] and for part of the Te atoms in a-GeTe, which have presumably a threefold coordination. The other part of the Te atoms implanted in a-GeTe might be twofold coordinated [5 = 0.87 (4) mmls, eQVzzlh = -704 (12) MHz, T/ = 0]. At a low implantation dose of 5 x 10 12 atoms/ cm2, Dezsi 80 observed a substantial (-30%) extra single-line component in the 1291 spectrum of asimplanted 129mTe and assigned it to substitutional Te. This is seen as support for a model which states that below the full amorphization limit of Si, part of the implanted Te atoms have a chance to escape their own amorphized implantation track and land substitutionally in a crystalline or recrystallized phase. A similar behavior was observed in low-dose 57CO implantations in Si, as discussed in Section 3.9.2. 82 Kemerink and co_workers 62 ,83-85 made a careful study of the influence of the position of the Fermi level on the 1291 spectrum in the decay of laserannealed Si containing substitutional Te. They showed the existence of three doping-dependent charge states of substitutional iodine in Si. The relevant spectra are shown in Figure 6. Heavily p-doped Si shows a single-line component (SI) with 5 = 0.96 (4) mm/s with respect to Cu1291 and 0 D = 193 (3) K. It is attributed to 12+. For Si that is more or less compensated, a single-line component (S2) is found with 5 = 2.39 (4) mm/s and 0 D = 170 (3) K. It is attributed to 1+. For the n-doped Si a more complex situation occurs. A

465


a

100

0.99

p - type

098

097

z 100

Q