Infrared microcharacterization of grain boundaries in polycrystalline silicon

Infrared microcharacterization of grain boundaries in polycrystalline silicon

Solid State Communications, Vol. Printed in Great Britain. 69, No. 5, pp.457-460, 1989. INFRARED MICROCHARACTERIZATION OF GRAIN BOUNDARIES IN POLYC...

376KB Sizes 0 Downloads 28 Views

Solid State Communications, Vol. Printed in Great Britain.

69, No. 5, pp.457-460,

1989.

INFRARED MICROCHARACTERIZATION OF GRAIN BOUNDARIES IN POLYCRYSTALLINE Dipartimento

0038-1098189 $3.00 + .OO Pergamon Press plc

SILICON

A. Borghesi, M. Geddo, G. Guizzetti di Fisica ‘A. Volta’, Universita’ di Pavia, I-27100 Pavia, Italy

S. Pizzini, D. Narducci, A. Sandrinelli Dipartimento di Chimica-Fisica ed Elettrochimica, Universita’ di Milano, I-20133 Milano, Italy G. Zachmann Bruker Analytische Messtechnik Gmbh, Karlsruhe, Federal Republic of Germany (Received 24 June 1988 by E. Tosatti) Fourier transform infrared analysis of boron-doped polycrystalline silicon, in the 4000-600 cm-1 wavenumber range, is performed by scanning inter- and intra-grain regions of the samples with a 20 pm size spot. The results provide the first direct. optical evidence of free carrier excess at grain boundaries (GB), attributed to impurity segregation. Moreover, oxygen absorption band behaviour shows that infrared active species of oxygen segregate at the surface of the GB only, in agreement with results from SIMS and EBIC measurements on the same samples.

IT

erentially segregate at GB, give rise to a variety of recombination states, depending on the nature and type of bond exchanged with non-metallic impurities (boron, oxygen, carbon) and unsaturated silicon atoms. In order to provide a direct correlation between GB electrical activity in poly-Si and the microscopic structure of the center(s) responsible for it, we used IR microanalysis, which appears a most powerful technique. The samples analyzed belonged to a polycrystalline silicon ingot already subjected to a complete structural, chemical and electrical characterizations, as well as to local secondary ions mass spectrometry (SIMS) and electron beam induced current fEBIC1 analvsis 112.14.15.171. The ingot was grown bi the hirect&nal \olidi&ation technique in SiaNI coated quartz crucible from electronic grade silicon intentionally doped with boron. The samples to be measured were cut to a thickness of 1.5 mm&from 10 x 10 cm2 polycrystalline silicon blocks using an I.D. diamond saw and then mechanicallv and chemically polished to a final flatness of 3 %. -Microstructural features of the samples were evidenced by conventional Sirtl etch procedures. The average diameter of the grains was of the order of 1 mm. The average resistivity of the samples was 3.8 ncm (corresponding to a boron concentration of 3.6~10~~ at/cm3) and the average diffusion length was 190 pm, as determined by the surface photovoltage method. Oxygen and carbon content were 12.5 and 7.1 ppmw, respectively. We demonstrated, by the use of EBIC technique 112,141. that. after ohosohorous diffusion. for about 15 minutk$, at ‘1150”6, all_GB were slighly’ recombining, except those in twin relationship, and independent of the relative orientation of the grains. The same heat treatment in an arson atmosohere resulted in annreciable oxygen (and carbon) segregation, which was monitored bv SIMS in the lateral scannine mode. So. we have stu”died with IR microanalysis boyh as grown and heat treated samples, cut from the same Si-poly slice. Optical measurements were performed using a Fourier-transform infrared soectrometer (Bruker mod. 113~) equipped with an IR microscope. ‘The spectroscopic features of the bulk and of the GB regions, in

IS WELL

ESTABLISHED [l-7] that grain in polycrystalline silicon (poly-Si) could act as e cient carrier recombination surfaces. with a significant reduction of carrier mobility in the surrounding space charge region [S-lo]. A variety of potential barrier heigths exists in the space charge regions albeit only the tallest determine the temperature deoendence of the bulk samnle resistivitv 191. While this picture is supported by experimental evidence showing that the GB’ s are often quite inhomogeneous recombination surfaces [ll-131, many of the aspects related to GB’ s electrical activity have not been fully addressed so far. For example, it is not clear whether in the polySi (n-type and p-type) the high density of states found close to the gap center has an intrinsic or extrinsic orinin. Furthermore. the nearlv indenendence of the recombination efficiency with Respect-to the relative orientation of the neiehbourine grains 1141and. therefore. to the structure of-any GBI Kas not’ yet been fully ex: plained. The hypothesis that oxygen contamination of GB results in localized GB states 191is substantiated bv experimental evidence; namely, be strong enhancemgnt of the electrical activity of GB observed after heat treatments of the samples is accompanied by oxygen segregation at GB [4]. Neverthless, a direct corrispondence between the oxygen segregation and the GB activity has not been found, nor have the physical consequences of oxygen segregation at GB been completely understood. Most impurities, intentionally or unintentionally added to nolv-Si. segregate within the GB region 141 as oxygen‘does, withpo&ible interaction among them: selves and with unsaturated silicon bonds, giving rise to various GB states. Recent literature added novel complexity to the problem, as the contemporary presence of carbon [15] and of deep level impurities [16] has been shown to influence final oxygen segregation at GB. Oxygen segregates separately from carbon (15 when carbon is present alone, while all impurities, mc \uding oxygen, segregate together within the GB region [IS] in the contemporary presence of carbon and deep level impurities, like titanium. Metallic impurities, which pref457

458

GRAIN BOUNDARIES IN POLYCRYSTALLINE SILICON

the 4000-600 cm-’ wavenumber range, were obtained by spotwise scanning the samples along a line crossing a specific GB with a 20 pm size light spot. The distance between each spot was kept constant and equal to 20 pm. The absorbance spectra of the bulk and the GB region for both a untreated and a heat-treated sample are reported in Fig. 1 and Fig. 2, respectively. Within the experimental uncertainty, we did not observe, in the examined spectral range, any dependence of the reflectance (R) on the individual sample nor on the spot position. The uncertainty in R is less than 2 %, to be compared with transmission variations up to 30 %. Moreover the behaviour of the absorbance spectra versus the spot position, which are almost coincident at high frequencies and separate with decreasing frequencies, could not be attributed to light scattering. Therefore any contribution of light scattered at GB is negligible in our present experimental conditions. A preliminary analysis of the spectra of Fig. 1 and Fig. 2 shows that: a) The absorbance band centered at 1107 cm-1 does not vary substantially when the light spot is scanned from the bulk to the GB region. This band is well known ia]to correspond to transitions involving localized vib ration mode of interstitial oxygen in silicon (the antisymmetric stretching mode ~3 of the S&O group). HOWever the absolute value of the peak is about 10 % higher for the heat-treated sample. b) No excess absorption at 1225 cm-’ and at 810 cm-‘, pertaining to the localized vibration modes of oxygen in crystalline and amorphous silica [i8,19], was observed in the heat treated sample. c) Both samples present an absorbance background whose spectral behaviour is typical of free carrier absorbance, which increases as the light spot is moved from the bulk to the GB. These features are more clearly observed in Fig. 3 which displays, for a heat-treated sample, the relative absorbance spectrum, i.e. the difference between the GB absorbance and bulk absorbance. It can be observed that all structures, in particular the absorption at 1107 cm-‘, in the spectrum of Fig. 2 disappear. The experimental curve in Fig. 3 is well fitted by a free carrier absorption, calculated using a Drude-like model for the dielectric function [ZO] B= E~+ ig with:

Vol. 69, No. 5

1.0 si

UNTREATED

0.75

&

io.5:_

a

b)

0.25f

0.0

IL_____ 1800

1600

1400

1200

WAVENUMBERS

1000

800

(cm’)

Fig. 1. Absorbance spectra for a untreated polycrystalline silicon sample: a) near the grain boundary; b) on the bulk.

1.0

si

z f

E

0.75

-

0.5

-

HEAT

TREATED

0

ti

a

b)

0.25 .

0.0

I 1800

1400

1600

1000

1200

WAVENUMBERS

800

(ci-n-‘)

Fig. 2. Absorbance spectra for a heat-treated polycrystalline silicon sample: a) near the grain boundary; b) on the bulk.

W*%

f2 =

w(l +cIw)

where 7 is the mean relaxation time for electrons (holes), Q_, is the high-frequency dielectric constant and up = JT is the plasma frequency, with N the free carrier concentration, e the electron charge and m* the optical effective mass for the electrons (holes). The best fit procedure between the experimental and calculated absorbance has been performed with the MINUIT program of the CERN library. This program, based on a least square algorithm, calls in series three minimizing subroutines: SEEK, a Montecarlo searching; SIMPLEX, which uses a simplex method by Nelder and Mead; MIGRAD, based on a variable metric method by Davidon. If we assume that m* has about the same value of 0.3 mo (m, is the free electron mass) as in bulk silicon, and observe that e, is essentially unaffected by low free carrier concentrations in this spectral region i.e.: coo fl 11.8), we obtain the following best fit values Ior the free carrier concentration and the relaxation time: N &.5x1~9~ cm-3 and + =1.45x10-‘~ s. The corresponding effective mobility p = er/m*, for carriers

jl’

1

__..__.” ....

.._......_...

3500

3000

. .. . .

2500

/

_.z / ,./ f . ..’

..W-1500

2000

WAVENUMBERS

1000

(d)

Fig. 3. Relative absorbance spectrum obtained from curves a) and b) in Fig. 2 full line); absorbance -alculated using a Drude-like mo 6 el (dotted line).

Vol. 69, No. 5

GRAIN BOUNDARIES IN POLYCRYSTALLINE SILiCON

moving in the potential well at the GB, is approximately ten times lower than the bulk mobility, in agreement with the data reported in refs. [21] and [22 . It is interesting to note that the dependence of the Iree carrier absorbance (at a fixed wavenumber n = 800 cm-l on the snot position is peaked in correspondence to t he position of the GB.-This dependence may be fitted by a aaussian curve with a half-width g EJQzrrrnfor both the intreated and the heat-treated sample (Fig. 4a and 4b). The only difference is a slightly higher absorption profile in the heat-treated sample relative to the untreated one. The curves in Fig. 4 also represent the free carrier concentration profile. In fact, the absorption coefficient a = E~w/~c,with n and w fixed, is proportional to wpa (see Eq. 1) and consequently to N. This is the first direct optical evidence that impurities electrically active segregate at GB. Apparently, the substantial independence d the free car&r concentration profile on thermal treatments proves that it is the previous hystory of the sample, and not the subsequent heat treatment, the main responsible of this segregation at GB. A similar result was obtained with conductance measurements along iron-doped silicon GB 1221,which gave a nearly temperature independent sheet conductivity, parallel,to the GB. The observed independence of the oxygen absorbance peak at 1107 cm-l from the spot position and the absence of absorutions at 1225 cm-’ and 810 cm-l are a good evidence that any IR active species of oxygen seereeate in detectable excess at GB. This conclusion ag;e, with previous SIMS depth profile measurements [15] showing that oxygen segregates at GB only in a thin subsurface layer (M 20 nm thick): therefore the resulting optical density is too small to be detected by IR absorption. Considering the very limited extension of the GB interfaces in coarsely grained Si-poly, their surfaces, as in single crystal silicon, are the most active sink for impurities, as well as the most efficient recombination site. While this last conclusion needs a confirmation by depth sensitive recombination experiments, our results provide a very significant support to previous conclusions about the role of GB in Si-poly, where intragrain defects dominate the recombination properties [13], and add an important contribution to our knowledge about the chemical configuration of GB.

459

SI

UNTREATED

(a) I

0.15-

5

O.lO-

i z

0.05 -

$ :

0.00,

cC

0.25,

20

40

60

60

a

120

140

160

160 I

sl

2

d

100

HEAT

TREATED

(b)

0.20

I I

0

.

A . r’

20

40

/

60 SPOT

\

60

100

POSITION

\

120

I

140

160

160

(ym)

Fig. 4. Relative absorbance maps at c = 800cm-~ over silicon grain boundaries of the same samples in Fig. 1 and Fig. 2: a) untreated sample; b) heat-treated sample. Note added in proof - SIMS measurements carried out after the submission of this paper, on samples coming from the same polycristalline silicon block, uantitatively confirm the segregation of boron at GB 7 231. Acknowledgements - This work was financed by Progetto Finalizzato ‘Materiali e dispositivi per elettronica a stato solido’ and by Gruppo Nazionale di Struttura della Materia de1 Consiglio Nazionale delle Ricerche.

The technical assistance of Mr. M. Moscardini is gratefully acknowledged.

REF ‘ERENCES 1. L.L.Kazmerski and P.E.Russel, J. Phys.

(Colloq.)

Anul.Phvs.Lett. 38,174 (1981). 3. C.H.Seager, J:Appl.Puys. 52,396O (ISSl).’ 4. L.L.Kazmerski, P.E.Russel, P.J.Ireland, J.R.Dick, C.R.Herrington, R.J.Matson and K.M.Jones, Proceed. 16th IEEE Photovoltaic Specialists Conf. (IEEE, New York,1982), p.622. 5. C.Donolato, J.Appl.Phys. 54,1314 (1983). 6. J.Marek, J.Appl.Phys. 55,318 (1984 . 7. S.Martinuzzi, Rev.Phys.Appl. 22,63 $ (1987). 8. J.Y.Seto,J.Appl.Phys. 46,5247 (1975). 9. C.H.Seager and T.G.Castner, J.Appl.Phys. 48,3879 (1978). 10. G.Baccarani, B.Ricci and G.Spadini, J.Appl.Phys. 49,5565 (1978). 11. F.Battistella, A.Rocher and A.George, Mater.Res. Soc.Symp. vol.59 Materials Res.Soc. (1986) p.347.

12. S.Pizzini, A.Sandrinelli, M.Beghi, D.Narducci and P.L.Fabbri, Rev.Phys.Appl. 22,631 (1987 . 13. S.Pizzini, D.Narducci and M.Rodot, k ev.Phys. ;4ppl. 23,188 (l-988). . S.Plzzuu, A.Sandrinelli, M.Beghi,D.Narducci! F.Allegretti, S.Torchio, G.Fabbri, G.Ottaviani, A.Fusl and F.Demartin, J.Electrochem.Soc. 135, I55 (1988). 15. S.Pizzini, P.Cagnoni, A.Sandrinelli, M.Anderle and Proceed. 2th GADEST (Garzau, DDR) 1987, p.268. S.Pizzini, L.Bigoni M.Beghi and C.Chemelli, :%ectrochem.Soc. 133 2i63 (1986) 18. B.Pajot, H.J.Stein, B.Cales and C.Naud, J.Electrochem.Soc. 132,3034 (1985). 19. F.Shimura, Y.Ohnishi and H.Tsuya, Appl.Phys. Lett. 38,867 (1981).

460

GRAIN BOUNDARIES IN POLYCRYSTALLINB SILICON

20. See, for example, F. Wooten in Ontical urouerties of solids (Academic Press, 1972). 21. S.M.Sze, Phvsics of Semiconductor Devics (Wiley, New York, 1969). 22. B.K.Miremadi and S.R.Morrison, J.Appl.Phys. 55, 3658 (1984).

Vol. 69, No. 5

23. S.Pizzini, F.Borsani, A.Sandrinelli, D.Narducci, M.Anderle and R.Canteri, Proceed. Symposium on Polycrystalline Semiconductors, Springer Verlag (1988) in press.