Defect creation and hydrogen evolution in amorphous Si:H

Journal of Non-Crystalline Sollds 35 & 36 (1980) 285-290 ©North-Holland Publishing Company

DEFECT CREATION AND HYDROGEN EVOLUTION IN AMORPHOUS Si:H = D. K. Biegelsen, R. A. Street, C. C. Tsai, and J. C. Knights Xerox Palo Alto Research Center

Palo Alto, C.alifornia 94304

The origin and nature of electrically active defects created by annealing are studied by luminescence, electron spin resonance, infrared absorption and hydrogen manometry. The initial defect density decreases similarly for all samples in annealing to -250°C. The temperature dependence of the subsequent increase in defect density is dominated by the effects of diffusion and sample microstructure and depends only indirectly on hydrogen bonding configurations. The peak in the rate of hydrogen evolution near 420°C is .:=ssigned to immediate release from isolated -Sill and/or = Sill 2 sites; this peak is found to shift to > 700°C because of diffusive delays in thick, homogeneous samples. Peaks at temperatures below 420°C are ascribed to hydrogen bonding sites on or near microstructural surfac,~s.

INTRODUCTION Previously published studies 1"4 of hydrogen evolution from plasma deposited a.Si'H have reported broad peaks in the rate of effusion and have attributed these peaks to release from sites having differing bonding configurations, e.g., -Sill, = Sill 2, -Sill 3 or (Sill2) n. Subsequent to these studies the existence of microstructure in a.Si'H (i.e., columnar growth having a chara~-teristic diameter ~100.200A) has been observed and found to depend on the deposition conditions. 5 One might suppose that this inhomogenous morphology should affect the specific hydrogen incorporation and evolution. Furthermore, evidence from SIMS measurements indicates that hydrogen diffusion is activated with an energy -1.5 eV,s although further studies 7 have shown diffusion coefficients at 235°C roughly 103 larger than values extrapolated from data of Ref. 6. In this article we present the results of parallel annealing measurements of luminescence, ESR, infrared absorption and hydrogen evolution on a set of samples spanning the range of deposition parameters and consequent range of as.deposited internal structures. We show that evolution is in fact dominated by the effects of diffusion and microstructural inhomogeneities in the samples. We show also that the electrically active centers depend only on the amount of hydrogen evolved and are characteristically the same as defects found in as.deposited (unannealed) films. 285

286

D.K. Biegelsen et al. / Hgdrogen Evolution in Amorphous Si:H

RESULTS

Eight to ten samples of undoped a.Si:H were simultaneously plasma deposited from SiH4/Ar mixtures onto roughened Coming 7059 glass substrates. The deposition parameters are given in Table I. Also included are the dominant hydrogen bonding configurations as determined by infrared absorption (for details see Ref. 8.) Deposition Conditions

H Environments

100%; 1W; RT; Anode

= (SiH2)n, -Sill

&

100%; 1W; RT; Cathode

=(SiH?)., _=Sill

O

100%; 1W; 230°C; Anode

_=Sill

0

5%; 20W; 150°C; Anode

(Sill2) n, = Sill 2

•

[]

5%; 25W; RT; Anode

(Sill2) n, = Sill 2

*

II

5%; 25W; RT; Cathode

=Sill 2, -Sill, (Sill2) n

Symbol

Microstructure

Table I. Sample Properties. (Hydrogen environments are listed in order of relative abundance.) The individual samples of each batch were isochronady annealed at the various temperatures, T A, in an initial pressure of 100t= dry N2 for ten minutes. Evolved hydrogen was monitored manometrically. A single sample of each batch was annealed at constant heating rate of 20°C/min to obtain continuous differential gas evolution curves. Total pressures of the order of one torr were generated in a constant volume of ~50cc. After annealing, the ESR and luminescence spectra of each sample were measured. (For details, see ref. 8) infrared measurements were run on a sample codeposited on a roughened crystalline silicon substrate and sequentially annealed. Figure 1A shows the variation of the ESR spin density, Ns, with annealing temperature, T A. The initial spin density in all samples studied is seen to decrease with an approximate activation energy of 0.5 eV reaching a minimum between 200 and 300°C. With further increase in T A, Ns increases in significantly different ways for samples with different deposition conditions. For example, only the anod e samples deposited at 5% concentration, 25 W power show a large increase around 300°C. Similarly, other sample dependent structure can be seen around 400°C and above 500°C. In Figure 1B are shown the luminescence efficiency data on the same samples as in Fig. 1A. The similarity in structure in (the reciprocal of) the luminescence and the spin density is striking. Figure 1C shows the annealing dependence of the energy of the luminescence peak. Again the gross features above ~250°C correlate, maximum shifts occurring along with maximum changes in IL and Ns. From the detailed similarity in the changes with annealing temperature, we are led to the conclusion that irrespective of the microscopic origin of defects at low and high temperatures, an increase in spin density leads to a decrease in intensity and

287

D.K. Biegelsen et al. / Hgdrogen Evolution in Amorphous Si:H

energy of the luminescence. In contrast, Figure 1D shows that the variation of luminescence linewidth (FWHH) with annealing temperature is different for the low and high temperature regimes. The width generally decreases from -0.45 eV to -0.3 eV by -300°C (i.e. where N s is a minimum) and then stays approximately constant. I

II

I

•

(a)

10 TM

I

I

Oo

I

O

I

(b)

m

¢

10-1

,/ iI

~10

TM

E i¢1 ,z, _1 10-2 n-

1017

IIX I

~¢

I

I

I

I

I

--

--

,

,

,

i

._

J

I

I

I

I

-

I

I

,.et

(d) -0.5 1

1.5

--

iiI

(3 O

[] _

--

1.0 0.9

0

I

I

I

I

I

I

100

200

300

400

500

600

0

I

I

100

200

T A (°C)

Fig. 1

o~

o I,

I

I

400

500

600

o 300

700

T A (°C)

Annealing of samples listed in Table 1.

We turn now to the hydrogen evolution during annealing. Figure 2A shows the increase in hydrogen evolved (normalized to sample volume) as a function of temperature. Figure 2R contains the derivative spectra of representative curves (plotted logarithmically) to show more clearly the structure. It can be see" that for a given sample the peaks correspond very closely to the regions of rapid change (Figure 1) in the luminescence and spin density. The abrupt changes above 650°C are due to H release on sample crystallization. (cf. reference 4.)

288

D.K. Biegg!sen et al. / Bgdrogen Evolution in Amorphous Si:H

101 300 (a)

I

I

i

!

I

I

I

I

I

I

I

(bl

100 -

M

! 150

100

J 2oo

I soo

I 4o0

I ~00

I 600

T,, (°C)

Fig. 2

Hydrogen evolution.

I

70o

i

800

I 1o-31

1100

I

200

I

300

I

400

I

500

I

600

l

700

800

Measurements of the changes in Si-H infrared absorption bands versus TA allow us to ascertain the microscopic origin of the evolved hydrogen. For example, in 100%, 1W, 30°C, anode samples the 320°C evolution peak correlates with the decrease of IR modes assigned to (Sill2) n whilst the - S i l l and - S i l l 2 bands decrease over a broad temperature range extending to greater than 600°C. A set of 5%; 230°C; anode samples deposited over a large range of powers (all of which have observable columnar morphology) all show evolution peaks only at 330°C and 430°C. The 330°C peak again correlates with IR modes assigned to (Sill2) n, whereas now the - S i l l strength decreases at 430°C. Thus, the 430°C and ~600°C peaks are not uniquely associated with either --Sill or = Sill 2. Moreover, the association of the peak near 320°C with (Sill2) n does not hold in one of the samples studied here. Cathode samples prepared at 5%; 25W; RT show almost no (Sill2) n in the infrared. The - S i l l 2 modes remain constant to above 400°C, whereas the - S i l l mode drops off near 350°C. The derivative of the evolution curve in Fig. 2A shows a peak near 3500C corresponding to the - S i l l decrease. Diffusion, for hydrogen release from bulk sites, plays an important role in determining the temperature of the peak near 600 ° (as opposed to the 320 ° and 420 ° peaks). The evolution for (100%; 1W; 2300C; anode) samples having apparently homogeneous morphology was measured for two samples of different thickness. The relatively lower H release at low temperatures and the greater release upon crystallization in the thick sample is consistent with a diffusive delay. Another set of samples, varying in thickness from 0.1 I* to 21/= was deposited under nominally identical conditions (5%, 1W, 230°C, anode). For samples thicker than ~11,, disintegration and a sudden increase in H-evolution from (Sill2) n occur near 350°C. This leads us to conclude that intercolumnar regions are filled with polysilane-like material. Also peaks near 450°C and 600°C are observed to coexist. Below ~1:=, no disintegration is evident and the "600°C" peak shifts down toward 4200C. All these spectra can be explained by assuming that the 420oC peak corresponds to Sill/Sill2 bonded on or very near to a surface, whereas the peak structure occurring for varying thickness from ~450°C to above 650°C (limited by crystallization) correspond to bulk Sill/Sill 2. The peak shift is determined by diffusion.

289

D.K. Siegelsen et al. / Hgdrogen Evolution in Amorphous Si:H

DISCUSSION The correlation between luminescence intensity and Ns (Fig. 1) is made more quantitative in Fig. 3. The relation is universal in the sense that the paramagnetic

. • . , a O . M IK •

I

. :'%, 10-1 II A

m

.i

10-2

Fig. 3

Luminescence intensity versus spin density.

10-3 1O'S

101;

101e

lO~Se

Ns

non-radiative recombination centers are the same independent of their origin for annealing or hydrogen release from various hydrogen configurations. The two theoretical curves represent the parametric bounds of the data 8 found for asdeposited samples. (Note that this plot is shifted to higher Ns by a factor of 2.1 due to a correction overlooked in the original paper. This shift does not change the conclusions.) Although we have found the quenching and shifting of th6 luminescence band to be similar at low and high temperatures and to be related simply to the spin density, the origin of the shitt in iuminescence energy and change in line width is in fact different for the two regimes. In the low temperature regime (<250°C), the shift to higher energies and narrowing with increasing anneal temperature are quantitatively consistent and accounted for by a reduction in the disorder potential in the band tails. The shift to lower energies at high temperatures with no consequent increase in width is in agreement with the decrease in bandgap when hydrogen is evolved. In summarizing the results of the local origins of evolved hydrogen we find: (1) that in samples containing polysilane the 320°C peak corresponds to hydrogen release from polysilane.like regions but in high power, high dilution cathode samples containing almost no (Sill2) n, nevertheless a peak near 350°C occum; (2), that the 420°C peak corresponds to hydrogen released from -Sill and/or = Sill 2 sites both in samples having strong columnar microstructure and in very thin samples -- therefore, most likely, near-surface-bonded; (3), that the peak near 6000C is related to diffusively delayed evolution of hydrogen from bulk, isolated -Sill and/or = Sill2; and (4), that -Sill 3 sites which occur in samples grown at low substrate temperatures, release hydrogen at < 2500C. The latter observation is made in the IR spectra, but is seen in evolution only as a low temperature tail on the polysilane peak. We find that we can account for all structure reported in the literature 1-4 in this unified way.

290

D.K. Biegelsen et al. / Hydrogen Evolution in Amorphous Si:H

The binding energy of the relaxed Si.H bond is stronger than the Si.Si bond and therefore, in the absence of a compensating energy gain mechanism, one would not expect to observe hydrogen evolution at such low temperatures. A plausible release mechanism is the bond number conserving reaction involving reconstruction in more or less weak bonds and molecule formation: Si-H + Si-H --, 2 Si.H --* Si.Si + H2. The ,evolution rate limiting process is either Si-H bond switching to a suitable second Si-H i~ite (e.g., on a surface) or local release of H2 but slow H2 diffusion to a surface. It is probable that any microstructure in the sample would enhance the likelihood of H2 release either by increasing the Si-H/Si.H pairing crossection by reducing the effective dimensionality of the space for easy diffusive motion, or by reducing the energy of the release arising from steric constraints. The presence of microstructure in samples would thus lower the temperature of maximum evolution rate from the 420°C for homogeneous regions. The fact that samples having columnar morphology always have a 320°C peak associated with release from (Sill2) n and show sample disintegration seems to imply that polysilane exists predominantly in the intercolumnar regions. The occurrence of a peak near 350°C in (the 5%; 25W; RT: cathode) samples for which no (Sill2) n or columnar morphology is observed, may then imply a different microstructural growth pattern. For example, small yoids which contain little or no (Sill2) n could also enhance the probability of two --Sill units becoming neighbors with consequent H2 release and bond reconstruction. CONCLUSIONS We conclude from the above results the following picture. First, annealing initially reduces the defect density. The mechanism, whether defect motion or hydrogen diffusion, has not been ascertained, but hydrogen release below 150°C2 tends to support the latter. The subsequent increase of defect density is dependent both on the nature of hydrogen bonding and the microstructure and thickness of the samples. Nevertheless, the defects annealed out below 250°C and created above, act as nonradiative recombination centers in the same way as found in as.deposited samples-independently of the microscopic origin of the hydrogen evolved. REFERENCES *Work was supported in part by DOE contract 03-79-ET-23033. !. M.H. Brodsky, M. A. Frisch, J. F. Ziegler, and W. A. Lanford, Appl. Phys. Lett., 30, 561 (1977). 2. K.J. Matysik, C. J. Mogab, and B. G. Bagley, J. Vac. Sci. Tech. 15, 302 (1978). 3. J . A . McMillan and E. M. Peterson, (to be published in J. Appl. Phys.). 4. C.C. Tsai, H. Fritzsche, M. H. Tanielian, P. J. Gaczi, P. D. Persans and M. A. Vesaghi, Proc. 7th Int. Conf. Amorph. Liq. Semicond., edit~.d by W. E. Spear, (Univ. of Edinburgh, 1977), p. 339 and C. C. Tsai and H. Fritzsche, Solar Energy Mat., 1, 29 (1979). 5. J.C. Knights, and R. Lujan, (to 'be published). 6. D.E. Carlson and C. W. Magee, Appl. Phys. Lett. 33, 81 (1978). 7. D.E. Carlson, Publication SAN-2219-1, Dept. of Energy, San Francisco (1979). 8. R.A. Street, J. C. Knights and D. K. Biegelsen, Phys. Rev. B 1__8,1880 (1978). 9. J.C. Knights, G. Lucovsky and R. J. Nemanich, Phil. Mag. B 37, 467 (1978); M. H. Brodsky, M. Cardona, and J. J. Cuomo, Phys. Rev. B 16, 3556 (1977).