Hyperfine Interactions 45 (1989) 199-216

ION IMPLANTATION IN SEMICONDUCTORS STUDIED BY MOSSBAUER SPECTROSCOPY

G. L A N G O U C H E Instituut voor Kern- en Stralingsfysika, University of Leuven, B-3030 Leuven, Belgium

The application of MSssbauer spectroscopy as an extremely sensitive characterization technique for ion-implanted semiconductors, is illustrated. Factors influencing the final landing site of implanted ions are first reviewed, as well as ion beam induced material modifications. Recent applications of MSssbauer spectroscopy in this field are then discussed including the study of supersaturated solutions of Sb and Sn in Si, the formation of epitaxial and buried silicides and the search for the DX-center in GaAs. 1

INTRODUCTION

Ion implantation has nowadays become the dominant technique in the industrial well-controlled doping of shallow layers in semiconductors. This is the major application of ion implanters which use a typical energy in the 50 keV to 500 keV range, a beam diameter of a few mm, usually a sweeping facility to cover an implantation area of a few cm~, and a typical maximum current of the order of a few mA. Dedicated ion accelerators, however, will produce e.g. an oxygen beam current of 100 mA for buried oxide formation, a 10 nm beam spot for microlithography, or several MeV particle energy for the.implantation of buried layers. Semiconductors are of course not the only target material for ion implanters. In principle any ion can be implanted in any target. Insulators like oxides, plastic foils and even frozen noble gases have been implanted for research purposes. Numerous ion implantation studies on the production and mobility of defects in metals have been reported. Ion implantation in metals and alloys in order to improve the wear resistance has also become a commercially viable technique, typical examples being heavy duty ball bearings and surgical implants. M6ssbauer spectroscopy has fairly unique properties as a characterization method for ion implanted materials. The nucleus of the implanted MSssbauer isotope acts as an extremely sensitive microscopic probe inside the solid in which it is embedded. Valuable information can often be extracted in this way on the structural and electronic properties of the individual atom and the immediate surrounding lattice. The position of the implanted atom in the lattice can be studied, the formation of complexes with defects or other impurity atoms, the charge state of the MSssbauer atom, and also dynamic processes can be studied, which are happening within the time window of the lifetime of the M6ssbauer level, which is typically 10 -r to 10 -s s. In particular the fact that only the M6ssbauer nuclei contribute to the signal, and on the other hand that all of them do contribute, compares favorably with some of the more common characterization techniques, which are often based on the study of the bulk of the material, require crystallinity, or are restricted to atoms that are electrically active. A number of reviews on MSssbauer spectroscopy of ion implanted materials has been given in recent years [1-4]. In this review paper, which is meant for a conference on industrial applications of MSssbauer spectroscopy, we will limit ourselves to a discussion of a number of recent M6ssbauer studies on the ion implantation in semiconductors which are clearly 9 J.C. Baltzer A.G., Scientific Publishing Company

199

200

G. Langouche, Ion implantation in semiconductors

interesting for the microelectronics industry. Supersaturated solutions of Sb in Si are studied in order to obtain semiconductors with high carrier concentration. Epitaxial and buried silicide layers are used as surface contacts or as conducting layers in three dimensional arrays respectively. The DX-center in GaAs and AlxGa,-xAs is heavily investigated as it inhibits high donor activation. Before discussing these systems, we want to make some general remarks on the landing site of the implanted ion and on the effects of ion beam irradiation of materials. 2

THE LANDING SITE OF THE IMPLANTED ION

A detailed microscopic description of the individual collision cascade is still not fully developed. Early models [5] use a simple binary collision approach to describe the collision cascade. Soon it was realized, however, that during the collision process, atoms in the cascade receive amounts of energy which, when converted to a temperature scale, exceed the melting point of the material. The concept of "thermal spike" was therefore introduced, and a controversy started on whether or not the temperature concept can be used in such a non--equilibrium situation. Molecular dynamics computer simulations are presently used in attempts to visualize the physical evolution of an implantation cascade, starting from first principles. The crucial point in such calculations is the choice of an interaction potential which accurately describes the interatomic forces. Typical results of such calculations [6] are collective phenomena like correlated sputtering of clusters of atoms, collective movements of lattice planes forming real shock waves in the lattice, and focused collisions along high index crystallographic axes. The final lattice site of the primary implanted atom and its possible association with defects will determine the electronic properties of the semiconductor. Few experimental techniques can deal with the determination of individual lattice sites. When single crystals are available, the channeling technique can be used to control the possible (partial) blocking of open crystallographic channels by impurity atoms. Shifts from equilibrium position as small as 0.01 nm can be determined. Hyperfine interaction techniques, e.g. MSssbauer spectroscopy and perturbed angular correlations (PAC), probe the immediate microscopic surrounding of the implanted radioactive atortas. Extended X-ray absorption (EXAFS) has recently also been applied to implanted systems. From such studies it is found that a number of parameters have their influence on the final lattice site of the primary atom [7]. We will review these shortly. In earlier attempts to predict ion implantation sites, the probability for a final replacement collision [5] is calculated, and from this the substitutional fraction of implanted atoms can be estimated. From the lack of agreement between such calculations and experiment, it is concluded that simple ballistic effects are not sufficient to explain substitutional fractions of implanted atoms. The chemical activity of each element and its affinity to make well specified chemical bonds, often with geometrical restrictions, plays an important role during the final quenching of the energy spike. Chemical effects as well as size effects are included in models that try to predict the final site selection using conventional alloying considerations, like the Hume-Rothery rules [8], and the modified Hume-Rothery rules formulated by Sood [9]. Differences in electronegativity and atomic size are plotted in Darken-Gurry plots to predict regions of substitutional incorporation of impurity atoms. A strong improvement in such predictions is obtained by using Miedema parameters [10], somewhat empirical thermochemical parameters, originally used to predict the heat of formation of alloys. When plotting the elements in a two-dimensional plot, according to the Miedema parameters, chemical potential and the electron density at the boundary of the Wigner-Seitz cell, separate regions can be outlined corresponding to different types of implantation lattice sites, as is beautifidly illustrated for elements in a Be host [11]. Sing and Zunger [12] introduced a new approach to predict final lattice sites for impurity atoms, solely based on size effects. Orbital radii are introduced, defined as the effective core size of atoms as sampled by s, p and d valence electrons respectively. They are derived entirely non-empirically, in contrast to both the Hume-Rothery and Miedema approaches, using only first-principles quantummechanical calculations of a non-local

G. Langouche, Ion implantation in semiconductors

3.0

I

I

!

201

9

ARb Cs

2.0 5 d

9 Hf

Au 9

AZr

9 ,, 9

1.o _ Te

9 No

Cd

//r~B~':~ 9

p Sn.yao 9

Xe"L"~:'/_I. "Go 0.0 Se'~-~'A Bi As

ALi

0.25

0.0

I

0.5

0.75

1.0

R, n- (a.u.) Figure 1

Implantation sites in Si plotted according to the Singh-Zunger orbital radii. The experimentally determined implantation sites are indicated by * (substitutional) and 9 (non-substitutional). [from Reference 12]

screened pseudopotential for each element. Fig. 1 shows the Singh-Zunger predictions for 26 implants in Si, and Table 1 shows a table of merit made up by Singh and Zunger [12] comparing the success of the different models just discussed. Table 1

Table of merit comparing different model implantation sites in Si. [from Reference 12]

Theory Hume-Rothery Modified Hume-Rothery Miedema Orbital radii

predictions

for

substitutional (%) non-substitutional (%) 22 100 83 100

50 29 75 100

Depending on specific parameters such as the implantation dose, the implantation energy, target temperature, the time delay between implantation and measurement, and even del)ending on the measurement technique, differences call Occur. It is important, therefore, to obtain data where the direct landing site ill an undamaged crystal is probed. Site changes due to annealing effects should be avoided, which is often hard as radiation enhanced diffusion can alter the landing site during the sample preparation procedure. Also conditions should be such that amorphization of the lattice is prevented. Both conditions are often hard to obey simultaneously as the probability for the former effect increases when rising the temperature, while the probability for amorphization of the lattice increases when lowering the temperature.

G. Langouche, Ion implantation in semiconductors

202

In a study of implantation sites in metals, where amorphization is not likely to occur, impressive results were obtained by O. Meyer [13] by lowering the temperature of the implantation and the characterization - by channeling - to temperatures where vacancies are not yet mobile. A systematic behavior was observed when plotting the measured substitutional fraction as a function of the heat of solution, derived from the Miedema parameters. For higher values of the heat of solution, the impurity-host system becomes less miscible, so that correlated vacancy trapping already occurs during the implantation process of single ions, even at low temperatures. Kalbitzer [14] tried to obtain accurate data points on the implantation sites of different elements in Si. Implantations were performed at room temperature, in order to prevent lattice amorphization. By choosing a low implantation energy (2 - 3 keV) and by selecting the mass of the impurity atom considerably heavier than the target atom mass, the separation between the implanted impurity depth profile and the damage depth profile could be kept as large as possible. Figure 2 shows the 20 data points obtained by Kalbitzer et al, ordered according to different columns in the periodic table of elements. Substitutionality is only observed for group III, IV and V elements, typical donor and acceptor atoms, so that several site assignments used in figure 1 have to be reconsidered. The dominant role of the chemical character of the implanted element in the semiconductor is clearly illustrated.

LATTICE LOCATION OF IMPURITIES IN SILICON SIDE M A I N GROUPS

o~

v

Z

0 I00 ~,- . . . . .

~:~

g tb so

X

Cs Au, Ag

Figure 2

~./

/ ",,

Ba Cd

9

o

9 INTERSTITIAL (NON-TETRAHEDRAL)

SUBSTITUTIONAL

~,

9

TE TRAHEDRAL

)i

,' .

r

o

TI, In Y

. ,'

.

...... .

k---k*

....

Sn Pb Zr

-,~i

,'

....

Bi, Sb Ta

',,

Te W

/',

I Re

Xe PI

Fractional occupation mumbers for different implanted impurities in Si on three different kinds of lattice sites. [from Reference 14]

G. Langouche, Ion implantation in semiconductors

The dominant role of chemical effects and size effects in the process of lattice site selection has been clearly demonstrated in the previous discussion. It is becoming more and more clear in recent experiments that in semiconductor hosts the lattice configuration of the impurity atom can also be governed by pure electronic effects. It is well known that depending on the position of the Fermi level in the bandgap of the material the charge state of deep level impurities is changing. This has been demonstrated also by isomer shift changes in IVI6ssbauer spectroscopy: for Fe in GaAs [15], for Sn in PbSe [16] and CdTe [17], for Te in Si [18] and for I in Si [19]. All of these atoms were found to occupy substitutional lattice sites, except for I 0 in the last reference. It was found that iodine atoms, when in the neutral charge state, occupy all off-substitutional lattice site due to a Jahn-Tellcr like distortion. The electronic structure around the impurity atom is clearly the driving force for this off-substitutional site occupation. The DX-center is.probably another example of such a behavior and will be discussed in section 6. 3

ION BEAM MODIFICATION OF MATERIALS

3. I

Structural modifications

The structural modifications of the ion bombarded lattice range from simple point defect generation, over the creation of large defect clusters such as loops, stacking faults, bubbles, up to the complete amorphization of the implanted lattice. Amorphization by ion implantation is well known to happen for non-metallic targets, and is easily observed in channeling and X-ray diffraction experiments. The mechanisms playing a role in this amorphization process are still under investigation. Transmission electron microscope pictures of individual collision cascades [20] suggest that individual implantation tracks created by medium heavy and heavy atoms are already amorphous. Ultra high resolution electron microscope pictures of individual tracks are also interpreted as being amorphized regions in the Si host. The low implantation dose MSssbauer data on 57Co implanted in Si can be interpreted [21] in favor of such an impact amorphization model. A large quadrupole doublet, due to some so far unspecified defect structure around the Co atom, dominates the MSssbauer spectra of '~7Co in ion beam amorphized Si and evaporated amorphous Si films, and is therefore associated with the amorphous character of the host. A substantial population of this typical defect configuration remains present even at the lowest doses, and can therefore be an indication of impact amorphization. Electron and positron channeling studies [22], on the other hand, from ion implanted radioactive electron or positron emitters, indicate that in as-implanted samples the channeling ability of the host is not lost. This is interpreted as an indication for a large degree of crystalline order around the implanted probe atom, so that in these cases track amorphization seems unlikely. Further experiments are needed to clarify this controversy. 3. o

Topological modificalion

Tile topology of tile surface of ion bombarded targets can be strongly modified due to anisotropic sputtering of surface atoms. Surface structures, often with conical shapes, have been observed covering the sputtered surfaces {23]. The origin is to be found in the fact that the sputtering speed depends strongly on the angle between the sputtering ion beam direction and the crystallographic axes of the target, and on the impact angle with respect to the surface normal. Secondary processes also play a role such as the presence of impurity atoms on the surface, the redeposition of sputtered atoms, surface diffusion, surface stress and radiation damage. 3.3

Electronic modifications

As already mentioned tile doping of semiconductors with donor or acceptor atoms is by far the most important technological application of ion implantation, with obvious advantages with respect to classical diffusion techniques. Such are the complete control over the impurity

203

G. Langouche, Ion implantation in semiconductors

204

concentration as a function of depth, the lack of heating of the target, the full reproducibility of the process, and the absence of problems with the solubility or the diffusion coefficient of the impurity atom in the target. Supcrsaturated solutions can be obtained by ion implantation, sometimes by orders of magnitude in excess of the equilibrium solid solubility. Such an example, studied by MSssbauer spectroscopy, will be discussed in section 4. A recent development in the technological application of ion implantation in semiconductors is the production of insulating or metallic buried layers, by high energy (often MeV) ion implantation of oxygen or metal ions respectively. This offers the possibility to build three-dimensional networks, rather than two-dimensional ones. The formation of such metallic silicide layers extending to the surface offers the possibility to fabricate good ohmic contacts. A MSssbauer study on buried and epitaxial silicides will be discussed in section 5. Very impressive are the changes in electrical conductivity in insulators, such as polymeric films, induced by ion implantation. A change in conductivity over fifteen orders of magnitude (Figure 3) was observed [24] as a function of Ar implantation dose from 1014 to 10t7 atoms/cm2. 10 6 10l. i

cE 10 2 o I >

10-2

"~G 10 "t' "1 "~E 106 0 O 10"8

10-10 I

I

1

I

I

I

I

io~3 iov- lO15 ~ lO17 lO~ 1o19 implantation dose (ionslcm 2 ) Figure 3

3.4

Conductivity changes in a polymeric fihn as a function of 2 MeV Ar implantation dose. [from Reference 24]

Compositional modifications

Direct compositional modifications include of course the inclusion of foreign atoms in tile target material. In principle any kind of ion can be introduced in this way into any kind of solid target. By indirect compositional changes we mean the change in the original composition of a multi-element target, as a consequence of the ion implantation process, in which the implanted species does not play a role. Ion beam mixing of multilayered structures is the typical example. This technique is sketched in Fig. 4. A chemically inert gas beam (like xenon) is used as mediator of the mixing process. Ion beam mixing can give rise as well to a homogeneous mixing on an atomic scale as to the formation of stable binary phases. This process is studied as well in fundamental as in applicd research. Micro-surface alloys are formed that cannot be made by other techniques. The reason is that the ion beam mixing process is characterized by ultra-short cooling times: the time needcd for a collision cascade to cool down, which is of the order of 1 to 10 picoseconds. In this way cooling rates of the order of 101~ to 1014 K/s can be reached, which is orders of magnitude faster than 107 K/s reached e.g. in splat cooling.

G. Langouche, Ion implantation in semiconductors

Xe

>

ABAB AB Figure 4

AB

xy

Schematic view of the ion beam mixing process. A multilayer system of alternating layers of two elements is bombarded with inert gas ions. An alloy is formed with a composition that is determined by the original layer thicknesses.

The mechanisms playing a role in the ion beam mixing process offer an interesting subject for fundamental research. Ballistic effects certainly play a role in the intermixing of cascades. Radiation enhanced diffusion is very effective in the transport of vast amounts of mobile point defects, dragging along target atoms. Chemical effects arc also clearly present in experiments carried out at temperatures were diffusion is severely limited. M6ssbauer studies on the ion.beam mixing of Fe layers on Si are reported in section 5. 3.5

Mechanical properties

The improved wear and corrosion resistance of materials after ion implantation has drawn considerable attention. Ion implanted orthopedic hip and knee joints, made of Ti-A1-V alloys show a 100 times longer lifetime after ion implantation with I0 t~ to 10tr nitrogen atoms/cm2 [23]. The understanding of the different meci~anisms involved in this process is far from complete. M5ssbauer studies [25] havc been reported that try to analyze the effects of nitrogen implantation on steels. 4

SUPERSATURATED SOLUTIONS OF Sn and Sb IN Si

The fact that a,--Sn itself is a group-IV semiconductor "suggests of course that Sn impurities will be easily incorporated in Si or Ge. This is indeed the case for Ge, as the thermal equilibrium solubility limit for Sn in Ge is extremely high (almost 1021 atoms/cm3). The solubility limit of Sn in Si is substantially lower (5 x 10 t9 atoms/cm.~). The diffusivity of Sn is very low, however, so that supersaturated solutions can be obtained by various techniques. The same holds for Sb in Si, which has very similar solubility and diffusion parameters as Sn in Si. The properties of Sn and Sb in such supersaturated solutions were the object of M(issbauer studies. Before reporting two such recent studies we want to mention that Sn in Si (and other semiconductors) has been the object of a large number of very extensive MSssbauer studies, especially by the Aarhus-CERN_group. These studies have been reviewed recently [26] and concern topics like Sn--doped Si crystals doped from the melt, by diffusion, by ion implantation and by laser implantation. Point defects and amorphized hosts were generated by various techniques, and the Sn resonance was studied in the decay of parent isotopes belonging to Cd, In, Sb, Sn, Te and Xe elements. At least six different Sn-defect configurations were recognized in these studie~s.

205

G. Langouche, Ion implantation itz semiconductors

206

Weyer [27-29] studied the behavior of supersaturated solutions of both Sn and Sb in Si. Sn is an isoelectronic impurity, while Sb acts as a donor in Si. Supersaturated solutions were formed by ion implantation at concentrations above the solubility limit, followed by rapid thermal annealing by an incoherent light source in a matter of seconds. Despite the similarity of the maximum solubility and diffusion-constant values of Sb and Sn in Si, the two species were found to behave differently. Figure 5 shows the 119Sn spectra after implantation of 2 x l0 ts atoms/cm 2 and 5 x l0 ts atoms/era 2 respectively, after incoherent light annealing (ILA) at three different temperatures. Although the implantation doses are only slightly more than a factor two different, the spectra are very different in shape. The 2 x 10 L5 atoms/era 2 implantation dose corresponds to a Sb concentration of about 5 x 1020 atoms/cmL This corresponds to the maximum that can be incorporated electrically active on undisturbed substitutional sites, after ILA to 700"C. The upper left corresponding spectrum indeed shows an almost pure single line, corresponding to the substitutional line (6 = 1.79(3) ram/s, @ = 234(5) K). This concentration is about an order of magnitude higher than the maximum thermal equilibrium solid solubility. Higher implantation doses and annealing to higher temperatures results in the population of a (vacancy-associated Sb) defect site (~ = 2.32(5) ram/s, (9 = 183(10) K) and of Sb-metal precipitates (6 = 2.74(5) ram/s, @ = 237(10) K) line (Fig. G). l

,

,

I

I

1

i

i

i

l

1

I

I

L

I

mo

I

I

I

I

1 I

e:f$

I

I

81~

S~

uz

375

~ 62~

25O

lSC

SO(

115

I

I

i

1

I

I

I

I

I

I

I

I

-IL2

-~le

I

I

I

71

6+

S|

I 2S 500 )I$

I.(

31

,,+!,

,

.o',

-,',

-,:

-,',

.;o

vEtOCltV t ~ z ~ )

Figure 5

.,',

2+

.:,

I ~'05

-Or6

-~

-lt~

-;'~

--'l

--s

u0Sn MSssbauer spectra of Si implanted to a dose of (left) 2 x 10 t5 tt0Sb/cm2 and (right) 5 x 10 I~ ttQSb/cm2 after 15 s ILA at the indicated temperatures. [from Reference 29]

G. Langouche, Ion in+plantation in semiconductors

1.0

+

,

1.0

~

ist/~,

i

i

5 x |OlSSblcm /

~ ? x 10~5Sblcm2 .

0.8

207

O.8

"~ ~g

Su

z06

PreClp.

Subs1, ~ A

0s

,,,c,~~....--:

0+4 o. 0.2

"

~ o~ 0.2

9

O@ e l l I

I

700

I

0

J

900

7 0

1100

1100

900

ANNEAL TEMPERATURE {*C)

ANNEAL TEMPERATURE ( ' C )

Population of the different sites as a function of annealing temperature, as derived from M6ssbauer spectroscopy. [from Reference 28]

Figure 6

+

i

i

5 x 10"sSDIcm2

2 x 10155D Irm Z u o

o uJ

~3 o Chonne(Ing

~2 3 gll~tncal

0

7 0

-]

~

'

I

900

ANNEAL TEMPERATURE (~

Figure 7

1100

0

I

i

I

a

700 900 ANNEAL TEMPERATURE ( ' C )

1100

Substitutional doses from channeling (*), M6ssbauer (m) and electrical (4) measurements versus annealing temperature. [from Reference 28]

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 tern eratures W h e n comnared to channeling data and electricalmeasurements (Fig. 7) the close similarity between the results from Mossbauer experiments and the electrical data ~s striking, while the channeling measurements clearly overestimate the substitutional fraction by 25 to 30 %. An interesting question therefore arises. What kind of Sb configuration is seen as substitutional by the channeling measurements, but as non-substitutional by the electricalas well as the MSssbauer experiments. Both the vacancy-associated Sb and the Sb metal precipitates could, in principle, result in only small displacements from exact substitutional positions. Another question raised by the authors concerns the source of the vacancies generated at low annealing temperatures and the numbcr of vacancies involved in the vacancy-associated Sb defect.

G. Langouche, Ion implantation in semiconductors

208

200C tltO0 II00

riO0 r

0

I

I

I

I

I

I

I

I

tS~O 'i'~ ,, I 0 0 K

~'5(

I

I

l

I

I

I

I

T N , 192 K IS0~ I10( g0( 60(

~o

o' . h

.b

-,: -i, 30 -h -:,

.,,

VELOCll I' Imm/ll4.(,

Figure 8

tt~ M6ssbauer spectra from IO ta HgSb/cm2 implanted in 5 x 10 t.~ Sn/cm 2 preimplanted Si samples. T~, and T,, are annealing and measurement temperatures. [from Reference 27]

Figure 8 shows a quite different behavior for 5 x 10 t5 atoms/cm 2 Sn-preimplanted Si samples. The substitutional line dominates at all annealing temperatures. No precipitate line is observed in the spectrum. Two defect configurations present are thc vacancy associated line and what is called a "surface" line, a high isomer shift (/f = 3.25(10) mm/s, O = 183 K) line attributed to interstitials in the surface layer, which can be removed by stripping. No substantial precipitation is seen for Sn. As Sb precipitates upon annealing and Sn does not, Weyer concludes [27] that the Coulomb attraction between tile positively charged Sb-donors and the negatively-charged vacancies must be the natural driving force for this reaction. The Sb migration is thus mediated by vacancies. Scherer et al. [30] made a careful channeling and M6ssbauer study of a still higher implantation dose (10t~ atoms/em2 at 250 keV implantation energy) of ttgSn in Si. After a two-step annealing process (1 h at 600"C followed by I s rapid thermal anneal at 1200"C) only substitutional Sn is seen by MSssbauer spectroscopy. Complete epitaxial regrowth of the amorphized layer is reached, as well as a complete substitutional solid solution of tt0Sn to a concentration two orders of magnitude above the solid solubility limit. Without this two-step process a bigger substitutional fraction is seen in the channeling measurements than in the MOssbauer data. The excess Sn atoms are found in vacancy-complexes. This led the authors to suggest that also in the Sb-experiments of Weyer et al. Sb-vacancy complexes, rather than precipitates, might be responsible for the overestimation in the channeling substitutional fraction.

G. Langouche, Ion implantation in semiconductors

5

EPITAXIAL AND BURIED SILICIDES

5.1

gpi~azial and buried OoSi2

209

In recent years Co-silicides have been studied very hard because of their possible application as low resistivity contacts to Si, and for millimeter wave applications. These studies consisted mainly in X - r a y diffraction, Rutherford backscattering and electrical measurements, all giving only macroscopic information on the properties of the formed suicides. A M6ssbauer study was performed by Vantomme [31] in order to obtain a more local microscopic structural information. Co-Si surface layers were prepared by implanting radioactive .~rCo into a 80 nm thick evaporated Co layer on top of a Si single crystal. This system was subsequently thermally annealed and the formation of the different cobalt silicides could be clearly followed as a flmction of temperature, as they are characterized by well-defined and different M6ssbauer spectra. Simultaneous Rutherford Backscattering Spectroscopy measurements show that the silicides are formed as parallel layers on top of the Si substrate. The final silicides formed, after annealing above 600"C is CoSi~. The relative areas of the various spectrum components in the formation of Co-silicide surface layers on Si as a function of annealing temperature, as deduced from the M6ssbauer spectra, are shown in Figure 9.

0

<

100-

la

.

'13

.N

o,,o o

f

............

80-

0

E 0 Z

600

L0-

20"

O0 4, 0

O-

,'

Co2Si CoSi CoSi2

6 Figure 9

160

260

36o

660

760

Annealing Temperature {~ The normalized area of the various comj~onents in the formation of Co--sUicide surface layers on Si as a function of annealing temperature, as deduced from the M6ssbauer spectra.

G. Langouche, Ion imphmlation in semiconductors

210

A problem still under investigation is the anoma.lous spectrum of ,~7Co in CoSi2. Although CoSi~ is known to be cubic, and therefore is expected to give rise to a single line MOssbauer spectrum, the spectrum ahvays contains a broad side resonance (Fig. lfl), i n d i c a t i n g a n o n - c u b i c s u r r o u n d i n g for a. large part of tlle Co-al.oms. The contribution of this extra, component was found to depend on the orientation of the crystal with respect to the gamma observation direction, ltowevcr, not ill such a way that a well-defined quadrupole doublet with a well-defined orientation could be recognized. Local off-stoichiometry in a non-random way might explain such a behavior, but awaits further investigation.

.8 .oo

g o.95"~ 0.90

0.85-

0.80-

-is

-6s

ob

0:s Velocity (ram/s)

Figure 10

,~q~e MSssbauer spectrum of ,~TCo doped epitaxial CoSi2 after annealing above 55111"C.

Buried silicide layers were also studied. They were obtained by iml)lanting 160 keV r,.~ alternated with small fractions of r,TCo, up to a dose of 2 x 10 ~7 atoms/era 2. During the implantation tile temperature was kept at 280"C. After thermal annealing above 600"C the (;tnonlalous) spectrum of CoSi2 appeared.

5.2

Ion beam mixing of Fc-Si

Ogale [32] deposited a 5 nm layer of SWe on a <111> Si crystal, followed by a 25 nm overlayer of natural iron, containing only 2 % of r,q"e. Ill this way tile ,~TFe MSssbauer spectrum should be extremely sensitive to intcrl'ace phenonmna. First tire sample was studied without ion-beam mixing l,rea.tment. After deposition magnetic subspectra were observed (Fig. II), including the Sllecl.rum of o.-Fe, but superimposed on that a doublet (6 = (I.45 mm/s with respect to r A = U.99 ram/s) appears. The authors point out that the parameters of this spectrum are different from tile ones of STFe in FeSi (6 = 0.28 ram/s, A = 0.49 ram/s), although Cohen [33], in a similar evaporation experiment reported the formation of FeSi at tire interface. Tile hypcrfine interaction parameters of tim doublet measured by Ogale are strikingly similar to the doublet obtained after ion implantation of STFe in Si. Maybe the evaporation system did contain an accelerating voltage.

211

G. Lun~ouche, Ion imlganlulion in .semk'onductor,

The sa.nlple was then bonlbarded wil, h 1 to 3 x 10 i6 a t o m s / t i n 2 of ll10 keV" Xe a.toms. Ac(:ording 1,() the MOssbaner dal, a., Ve.~Si i-x alloys ,3.re [ornied ;il, I.he intt~rface, showing ;1 broad distril)ution in lll;I, gllOI, iC hyperfine splil, l,ings in the II.10 to 3fl0 kOe l'allgO. Upon anliea.lillg l.u ,15il ~ C the forrn;il, ion of ~-[;'e, I"oSi ~ild I"e:lSi is observed while after anlloa.ling al, 71)11~ C, only t i - l " e and l?e3Si reniain.

:

(a)

155 9

.:

L '.,

I~ ~.

.

:.

: .

";

"?.

9

:

!

.

.'

.:.:. ~.-...:' ,:.:.:..:..

o

~

,

:'.

;"~<,:~.!' .,

z

~q 112

g {b) m

L-, ?,

(.. .,.'=; '. "<:~.:~:'~'~:'~(i:'

167

']'* L

.L.:.~.~4(,:..:.: .

-I

-~

-I

-I

I

o

9 ":~":~":":':';./.:.:t i

VELOCIIY (mm/sr

l"iguro 1 l

133},

l/.oonl-temperature r'71:e ctmversion citer:trOll Mi)ssl)aner spr of (a) as-deposited and (b) icJn-bc'am nlixtxI Ve-Si coinlxlSites. Ifrtml ll,efer(,nce 32]

All as-inll)lantt'.d sa.niple was allneai(x[ Fur coitlpa.rlSOll. In conl, laSI, a~ain I,o (".oll('n who observed only bulk l:eSi a.ftor 651):C ,'lllll0al, Ogale observed iJnly ( l - F e after 7(ill ~ C

{I n heal. lil a nlore recent s t u d y

I3,1], [lie difl'eroliCOS b e t w e e n I,hl:~ ioil ht!anl n i i x i n g

ol" ]:o w i l h

crystalline a.lld ~llllOll)hothS Si woro invostig;ttod, with illixing dlisc,,4 rl.lHll 5 x ll)l:i to 5 x i(il,~ x e al,onls/cni 7. The c-,qi results a.ro COliSis{(,nt wil, h I,he i.Iievious study, except for the qnadrlipOle dtnll)iel, hi, I,he inl,erl'ace, which tills I,inle h~ts l, hp corrt.x:l, tln.r;ilnel.ers LiP I"oSi. ["tip tile a-Si sample I,his I"oSi douljlel. ('.(niiponont is liol, observed ;i.1. low inixing do,sos, but instead ;i sin;ill cuiitl-it)iltion ()[ a singh! line (~ = II.i/9 nlnl/s) appears, ~l.lltl is al3,rilnil.od I.o Po atoms ;it regula.r lal, l,ieo sites in Si, prol);ll)ly geneiatc'd lly rucoil inil)ianl.nl.icln. At high nii• doses lljl Fc:a-Si, (~l~e ;ind I"cSi is l'lnlllOd, while in the l"c:c-Si c;i8o o:--];'e ;131d l~eSi7 are [orlned. "l'he,~e differt!nt iun b('.il.ln nlixillg resull, s for tile ;L-Si a.nd c-Si sul)sl, ra.tes deiilonstia.te thai. the iOil be;tin inixitig pro(:e,<;s ill this case illVOiVOS inore th;in just hallistic niixilig, but I,hn.I difft'rent conClibutiollS o1" i'adi~l.l.iun cn]ianc(xI diffusion play a role ill ('-~i and ili a-Si.

G. Langouche, lo#7 implantation in semiconductors

212

6

SEARCH FOIl. T I l E DX--CENTEll.

The term DX-center was introduced by Lang [35, 36] with D standing for donor and X relating to an unknown associated defect. It was used to accot, nt for a defect center that could explain the persistent-ifllotoconductivity trapping center in Te-dol)ed AlxGai-xAs. It was found that for x > 0.22 the conductivity ill n-type Al.~Gai.xAs is controlled by a deep donor level. When the position of this level is followed as a function of x, for x < 0.22 this level is progressively deeper inside the conduction band. It can be pushed down out of the conduction band by applying external pressure exceeding 2 GPa. It is reached by the Fermi-level in bulk GaAs for doping levels exceeding 1019 atonls/cm2. It is therefore thought that the presence of this DX-level can explain thc observed lack of electrical activation for high donor doping levels in GaAs. 123

I

l x:~

I

I

V'-

0.995

0.993

u~ Z

~r

o.99~{

J

""-.~

J

~t %-

J

",

w 0 997 L

~:'I' F""o.43""'" ~" ' - ' ~ 0.993

0 996 ~

t.,:':r

."

'~u

,!'""'," "3

Figure 12

J

~"

-*..

J

,

-

0 3 V E L O C I T Y (ram/s)

tt+Sn MSssbauer spectra of u0Sn-doped AlxGat.xAs. Three resonances are used to fit the spectra, ll.esonance 1 is assigned to substitutional Sn, while resonances 2 and 3 arc associated with the DX-center. [from Reference 41]

G. Langouche, Ion i/nplantation in semiconductors

213

One of the simplest models to account for the lack of electrical activation of n-doped GaAs fur high donor concentrations is a configuration in which the dopant atom is not substitutional, but associated with a vacancy ill the nearest neighbor shell. For Te in GaAs, which is residing on the As site when it is active as a donor, this means all associated Ga vacancy in the first neighbor shell. This donor-accgptor p.air would provide self-compensation. Surprising results were reported from EXAFS-investigations on S and Te donors in GaAs, however. Part of the S-atoms [37] and all of the Te atoms [38] were proposed to have an unperturbed, but slightly contracted (for S) or expanded (for Te) first neighbor shell, with an As vacancy in the second neighbor shell. The analysis of the experimental data was questioned by Morgan [39], who suggested that a DX-center due to a "displaced donor", without any associated defects, equally well fits the experimental EXAFS data. Several microscopic "displaced 'donor" models are being proposed nowadays for the DX-center. They all involve donor atoms that are driven away from the substitutional lattice site, by a driving force that finds its origin it) the electronic level structure of the impurity-host system. Lang [35] originally proposed a large-lattice-relaxation model, but other models are regularly proposed nowadays, all having in common that no external defect is present. This DX-center probably belongs to a new large class of defect centers in I I I - V and I]-VI compound semiconductors. A fundamental understanding of this deep level is needed to evaluate the usefulness of these compound semiconductors in various device St rut..~,ure8. M6ssbauer st)ectroscopy is a very attractive toot to study defect (:enters of this kind. Its extreme sensitivity to perturbations in the immediate neighborhood of the impurity atom might allow to obtain information on this particular defect. A survey of all M6ssbauer studies on Te in GaAs has recently been published [40]. We will therefore not repeat this discussion but focus shortly on two relevant experiments. Gibart [41] reported a clear correlation between the presence of a spectrum component (i~round 3 ram/s) in the u0Sn M6ssba~l,er spectra of tt0Su-doped AlxGal-xAs as a. function of x (Fig.- 12). A sharp increase in the presence of this component is noticed (Fig. 13) for x > 0.2, where the DX-center should appear, lte therefore identifies this spectrum component with tile Sn DX-center, and offers a broad range of possibilities for its interl)retation. i

i

0

Z

60

re

50

0 ICJ

40

g7 2ofl.Iiii!iiiiiiii?ii!i iiiiiiiiiiiiii i;iiiiiii;ii!i!iii!i!iiiiii:iiiiiiiiiii:i iiiiiiii!i :i! iiiiiiii!i ;':',ili :iiii:i!iiiii 1i iiiiiiiiiiii?iiiiiiiiiiiiiiiiii!iiiiiii:iiii ! !iiiiiiiiiii!;iiiiiii i .ii:iii:!i 0

Vigure 13

Summed

0.2 0.4 0.6 0.8 AI C O N C E N T R A T I O N (x)

resonance

fractious

of

littcs

1

2

mtd

3

versus

AI

concentrations using a 2 site hypothesis (solid circles) or a quadrupole doublet hypothesis (open circles). Tile region above the shaded area is attributed to DX-centers. [from Reference 4/]

214

G. La,.~o,chd. Ion imlflU, lulio, in .senlico,duclo;:s

129,,,Te was also iml)lauted in, GaAs a..d the 1291 M6ssl.)auer sl~ecl,rum was studied as a function of concentration. The electrical acl,ivil,y ~ff the impla, nl,ed Te atoms was also veri[ied. Some of l,he specl,ra are shown in Fig. I,I. I']ectrical a(:l,ival,i(m measure.mevH,s 1,12, ,13] showecl I,hnl., while I~;r tllc~ Iowesl, dose (2 x Ill,:* al.imls/cm~) ;HnloSl, full elecl.rical acl, ivatiun could be r(~;l(:hed, only negligible a('tivaLion was ul)l, ained I'~)r the IlJlr, al,oms/em~ impla.nl,ed sample. Comparison with the M6ssbaucr data shows I.hat the disal)pearance or the electrical activation is correlated with the transitic)n from a si.gle line resuna.c:e (with absorber isomer shift with respect to Cul ~; = 11.71)(5) ram/s), inclical,ing a. regular lattice site, to a quadrupule split l\lSssbauer spectrum, indicating a sl.rong deviation [rum a. central l)~Jsition in an unperturbed ('orlfigural, ion. This quadrul)ole dottblel, (b = ().9[)(5) ram/s, eQV~.~./h = 29()(111) Mllz and 7/_~ II) is therefore tentatively associated with a l)X-ccnl, cr.

99

IUO

99'

~

b

O 98

99 ~

98-

-~,

-2

0

2

VELOCITY (ram/a) Vigure 1,1

1291 Miisshauer spectra or inq)lanted 12!luI'l'e ill (;aAs a.her rapid d~ermal a.nealing for doses or (a) :~ x II~:~ (h) IOJ4 (c) ll~.~ and (d) llllt~ al,(m,s/clvl2. A (llll ahsorber was used.

G. Langouche, Ion implantation in sendconductors

The tentative character of this assignment is because there might be other possibilities to account for tile observed correlation: tile formation of precipitates in this supersaturated solution is one of them, although the measured Mfssbauer parameters do not correspond to one of the known Te-Ga or Te-As compounds. Also the donor character of iodine awaits further confirmation. Concerning a possible interpretation of the observed quadrupole interaction and isomer shift, a straightforward model is difficult to put forward. Both an associated vacancy in the first neighbor shell, as a displaced donor seem to be able to account for the larger quadrupole splitting. A second-neighbor defect-association is expected to produce a much smaller electric field gradient than the observed one. We can state that Mfssbauer data on Sn, Te and I in GaAs all show that for high concentrations of dopant atoms, a defect configuration is observed, while for low concentrations they all occupy substil;utional lattice sites. Further tests are necessary to verify is the observed defect has also other typical DX-center properties. CONCLUSION By reviewing a number of recent Mfssbauer studies on ion implanted semiconductors, we have illustrated the ability of Mfssbauer spectroscopy to characterize microscopically matet;als which are important for semiconductor technology9 Especially the problem of the electronic and atomic configuration around impurity atoms in semiconductors remains a very interesting ono for fundamental physics as well from an applied physics point of view. The microscopic M6ssbauer probes inside these defect centers can reveal substantial information which is essential for further progress in this field. B.EFERENCES

Ii I

[iJ

[ 8] I;!t

Jl5J []16 [ii] [20] [21] [22]

L. Niesen, tlyp. Int. 13 (1983) 65. I. Ddzsi, Hyp. Int. 24/26 (1985) 1051. G. Weyer, Hyp. Int. 27 (1986) 249. H. de ~htaard, Hyp. lnt. 40 (1988) 31. D.K. Brice, Inst. Phys. Conf. Set. 28 (1976) 334. E. Verbiest, Ph.D. Thesis, University of Leuven, 1983, unpublished. G. Langouche and R. Coussement, Applications of the Mfssbauer Effect, (Gordon and Breach, New York, 1985) Volume l, 395. W. Iiume-Rothery, R.E. Smalhnan and C.W. llaworth, StT~tclure of Metals and Alloys, (Institute of Metals, London, 1969). D.K. Sood, Phys. Lett. 68A (1978)469. A.R. Miedema, F.R. de Boer and P.F. de Chatcl, J. Phys. F. 3 (1973) 1558. E.N. Kaufmann, Ft. Vianden,J.R. Chelikowsky and J.C. Philil)s, Phys. Rev. Lett. 39 (1977) 1671. O. Meyer and h. Turos, Nucl. Instr. Meth. B 19/21J (1987) 136. II.P. Frerichs and S. Kalbitzer, Rad. Eff. 83 (198,1) 135. P.P. Setegin, F.S. Nasredinov and A.S. Bakhtiyarov, Phys. Stat. Sol. b 91 (1979)35. F.S. Nasredinov, ].V. Prokofieva, P.P. Seregin, S.V. Zarubo, A.V. Ermolacv and A.N. Kurmantaev, Phys. Star. Sol. b 130 (1985) 727. II. Grann, F.T. Pedersen and G. Weyer, llyp. h,t. 29 (L986)1237. G.J. Kemerink, II. de Waard, L. Niesen and D.O. Boerma, llyp. Int. 1,1 (1983) 37. G.J. Kemerink, 1I. de Waard, L. Niesen and D.O. Bocrma, llvp. Int. 14 (1983) 53. L.M. tlowe, M.H. Rainville, H.K. llaugen and D.A. Thoml)s(~n, Nucl. Instr., lvieth. 170 (1980) 419. G. bangouche, M. de Potter, I. Ddzsi and M. Van Rossum, Rad. Eft. Lett. 67 (1982) 101 II. IIofs~ss, G. Lindner, S. Winter, B. Besold, E. Recknagel, G. Weyer and J.W. Pctcrsen, Mat. Sc. Forum 10/12 (1986) ]183.

215

216

[23] [24] [~1

[27] [28] [29] [31J] [31t 32 [33] {34] 35] 361 :~7] 38

f

39] 40] 41] ,12J

[43]

G. Langouche, Ion implantation in semiconductors

R. Kelly, Ion bombardment modification of surfaces, (Elsfwier, Amsterdam, 1984) Volume 1. M.S. Dresselhaus, B. Wasserman and G.E. Wnek, Mat. I-tes. Soc. Syrup. Proc. 27 (1984) 413. G. Wagner, R. Leutenecker, T. Louis and U. Gonser, Hyp. Int. ,12 (1988) 1017. G. Langouche, M'6ssbauer spectroscopy applied to Inorganic Chemistry., Volume 3, Eds. G. Long and F. Grandjean, (Plenum Press, New York, 1989), Chapter 10, to be published G. Weyer, A. Nylandsted Larsen, I,'.T. Pedersen, It. Galloni and R. Rizzoli, Mat. Sc. Forum 10/12 (1986) 1135. A. Nylandsted Larsen, F.T. Pedersen, G. Weyer, R. Galloni and R. Rizzoli, Energy beam - solid interactions and transient therTnal processzng, Eds. V.T. Nguyen and A.G. Cullis (Les Editions de Physique, Les Ulis, 1985)p. 319. A. Nylandsted Larsen, F.T. Pedersen, G. Weyer, I/.. Galloni and R. Rizzoli, J. Appl. Phys. 59 (1986) 1998. E. Scherer, J.P. de Souza, C.M Hazenack, and I.J.R. Ba.umvol: Semicond. Sci. Technol. I (1986) 241. A. Vantomme, I. Ddzsi and G. Langouche, Hyp. Int. ,11 (1988) 725. S.R. Ogale, R. Joshee, V.P. Godbole, S.M. Kanetkar and V.G. t~hide, J. Appl. Phys.

57 (198 1291 .

R.L. Cohen, I,.C. Fcldman, K.W. We,st and I'.,I. Silvcrman, IVlat. Ires. S(~c. Syrup. Proc. 3 (1981) 357. D.IVl. Phase, V.P. Godbole, V.N. Kulkarni, S.V fHmisas..SI~. Ogale and V.G. Bhide, Nucl. Instr. Meth. B 19/20 (1987) 737. D.V. I,ang and R.A. Logan, Phys. Rev. Lett. 39 (1977) 635. I).V. Lang, R.A. Logan and M. ,laros, Phys. Itev 1~ 19 (1979) 11115. F. Sette, S.,I. Pearton, ,1.M. Poate and .I.E. Rowe, Phys. [{.t,v. [,ctl.. 56 (1986) 2637. G.N. Greaves, P.J. Ilalfpenny, G.M. Lamhlc and K.I. lloberts, ,lourn. (le Phys. (Colloque) 47 (1986) 901. T.N IVlorgan, Phys. Rev. Lett. 58 (1987) 1280. G. Langouche, Proceedings of the third. Seeheim l,Vorkshop on illi')ssbauer Spectroscopy, (to be published in llyp. Int.) I'. Gibart, I).L. Williamson, B. El ,la.ni a.n(I P. Basnmji, Phys. I{.cv. B 38 (1988) 1885. G. I,angouche,ll. I~emehnansl .I. Odeurs, G. Ilorghs, M. de I'ot'tet, W. I)eracdt and M. Van Rossum, Proceedings of tin: 15th InteTTtatzonal Confcrc~ce on Defects i.n Semiconductors, Budapest 1988 (to be puhlished in Materials Science Forum). G., Langouche, D. Schroycn, It. Bemelrnans, M. Van Rq~ssum, W. Deraedt and M. de Potter, Mat. Res. Soc. Syrup. Proc. 111,1(1988) 527.