LUMINESCENCE IN a-Si:H

HAL Id: jpa-00220917

https://hal.archives-ouvertes.fr/jpa-00220917

Submitted on 1 Jan 1981

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

LUMINESCENCE IN a-Si:H

R. Street

To cite this version:

R. Street. LUMINESCENCE IN a-Si:H. Journal de Physique Colloques, 1981, 42 (C4), pp.C4-283- C4-291. �10.1051/jphyscol:1981460�. �jpa-00220917�

LUMINESCENCE IN a-Si:H

R.A. S t r e e t

Xerox Palo Alto Research Center, Palo A Z h , California 94304, U.S.A.

Abstract. - Dangling bonds, as observed in ESR, act as both radiative and non-radiative recombination centers in a-Si:H. The luminescence properties of a-Si:H are therefore very sensitive to the presence of dangling bonds.

Non-radiative recombination is shown to occur by electron tunnelling at low temperature and by diffusion and capture at high temperature. The evidence for a weak radiative transition of electrons trapped at dangling bonds is discussed. The h~minescence properties of doped and compensated a-Si:H are shown to be dominated by the changes in dangling bond density.

Introduction. - The inclusion of a tew percent of hydrogen during the deposition of amorphous silicon leads to a material with good semiconducting properties. The effect of hydrogen is to reduce the density of states in the gap, which in turn allows effective doping and results in long carrier lifetimes. Another consequence of the reduced gap state density is that efficient luminescence is observed at low temperature (1). The luminescence provides a convenient means of studying recombination processes in a-Si:H and has been widely investigated in recent years.

Hydrogen, although effective, does not completely eliminate the localized states in the gap, and the residual defects play an important role in the recombination. The purpose of this paper is to describe the various mechanisms of recombination through defect states which are found from luminescence experiments. The measure we use for determining the density of defect states is electron spin resonance (ESR). All forms of amorphous silicon have a characteristic ESR resonance at g=2.0055. This signal has been identified as a silicon dangling bond, much of the evidence coming from studies of dangling bonds at the crystalline Si/SiO, interface (2). In a- Si:H the dangling bond density N, can be controllably varied from about 3 ~ 1 0 ~ ~ c m - ~ up to 1019cm-3 by a suitable choice of deposition conditions (3). In this way, the defect density can be chosen as a variable in order to study its effect on the recombination. The question of whether diamagnetic defects also exist and are important in the recombination is discussed in 05.

Fig. 1 summarizes the recombination mechanisms that have been deduced from luminescence studies of a-Si:H. The luminescence spectrum is dominated by a broad band centered at 1.3- 1.4eV. which has high quantum efficiency at low temperature and low spin density. A second transition at 0.8-0.9eV is reIated to defect states and is discussed in 43.

A characteristic feature of amorphous semiconductors is the presence of localized states at the

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1981460

JOURNAL DE PHYSIQUE

EXCITATION OF ELECTRON-HOLE PAIRS

RADIATIVE PROCESSES

I

(LOW T, LOW Ns) (HIGH T, HIGH N,) (HIGH E X )

(LOW T, HlGH NS)

THERMAL AUGER

IONIZATION (HIGH lo) (HIGH T)

CAPTURE BY DEFECTS (HIGH N,)

Fig. 1 Schematic diagram of the recombination mechanisms in a-Si:H showing the experimental conditions when each dominates. (Temperature T, Spin density Ns, Excitation intensity I,, Excitation energy Ex)

conduction and valence band edges. An important consequence is that at si~fficiently low temperature, carriers do not diffuse after trapping in the band tails. Instead recombination, both radiative and non-radiative, involve tunnelling. Thus the 1.3-1.4eV "band edge" transition occurs by radiative tunnelling of electrons and holes from the band tail states (4) with a lifetime

rR given by

where 7, w 10-8 sec. R is the electron-hole separation, and R, is the effective Bohr radius, found to be about 10

k

(4). Thus, the recombination is characterized by a wide distribution of lifetimes corresponding to the distribution in the values of R with an average lifetime of approximately

lov3

sec (4). Diffusion is found to be significant above 30-50K and is observed as a change in the decay properties (5). The equivalent transition to a non-radiative process controlled by diffusion is discussed below.

Four non-radiative mechanisms have so far been identified, and Fig. 1 shows the experimental conditions under which each one prevails. Two of the mechanisms involve defect states and are discussed in 92. In addition, surface recombination is observed (6). although its effects are much weaker than is generally the case in crystalline semiconductors, because the carrier mobility and diffusion length are both much smaller in an amorphous semiconductor.

Finally, an Auger recombination mechanism has been identified at low temperatures from the dependence of the luminescence intensity on excitation intensity (7). The long radiative lifetime leads to a large density of photo-excited carriers, which favors the Auger process.

discharge deposited a-Si:H samples. The solid line shows a theoretical fit as described in the text.

I I I

lot6 loq7 1018

ESR SPIN DENSITY ( ~ r n - ~ )

Defect-Related Non-Radiative Transitions,

-

Fig. 2 shows the dependence of the band edge luminescence intensity at low temperature on the ESR spin density for a large number of glow discharge deposited a-Si:H samples (3). When the spin density exceeds about 1017cm-3 the luminescence is strongly quenched. These data establish that dangling bonds are non-radiative centers and also explain why no luminescence is observed in unhydrogenated material which typically has a spin density of about 1019cm-3. Just as for the radiative process, the data in Fig. 2 must be explained by tunnelling because diffusion in band tail states will be insignificant. An electron will tunnel non-radiatively to a defect at a distance R, with rate p given by

p = V, exp (-2RD/R,) (2)

where v,

-

1012sec-1. Non-radiative recombination of this type will therefore occur if this rate exceeds the radiative rate Thus, for a non-radiative transition R, must be less than Rc where

R, = % R O L n ( ~ ~ R ) (3)

A random distribution of defects leads to a luminescence efficiency Y, given by (3)

The solid lines in Fig. 2 show the fit to this expression. When a mean value for ^7, of sec and R,=10

k

is assumed, then from Eq.3, R,=100

k

in good agreement with observation. The tunnelling model therefore provides a qualitative and quailtitative fit to the data. However, it is interesting to note that although the model assumes a random distribution of defects, in fact material with a high defect density !as a pronounced columnar structure (8). P~~esumably the colums are sufficiently small (-100 A) that surface defects are effectively distributed at random in the bulk.

An even more direct link between the band edge luminescence and the dangling bond defects is found in spin denendent Ii~minescence (SDPL) experiments. In this experiment, the

JOURNAL DE PHYSIQUE

MAGNETIC FIELD H

-

Fig. 3 SDPL resonance compared to the ESR resonance. The dashed line corrects the ESR for a weak substrate signal.

INCIDENT LASER POWER ( m ~ / c m ~ )

Fig. 4 The relative luminescence efficiency as a function of excitation intensitjl in samples of different spin density.

change in luminescence intensity is observed as electrons are excited into resonance by absorption of microwave power in a magnetic field. The luminescence intensity is modified if either the radiative or non-radiative process depends on spin selection rules and therefore on the spin orientation. The SDPL experiment is quite complex in all its details since both radiative and non-radiative processes are spin dependent and several aspects of it have not been resolved (9). However, above about 30K an SDPL resonance is observed as in Fig. 3, which is clearly identifiable as the dangling bond signal. In resonance, the band edge luminescence intensity is reduced, as expected for the non-radiative tunnelling process (10). In addition, the magnitude of the SDPL increases with the spin density, again in agreement with expectations. (Street, unpublished)

At temperatures above about 50K, the onset of carrier diffusion modifies the non-radiative recombination. The luminescence intensity decreases with increasing temperature because the electron moves away from the hole, and this process replaces tunnelling to the defects as the primary non-radiative mechanism. However, to complete this non-radiative process, the free carriers must recombine. Some evidence of how this occurs is shown in Fig. 4 in which the luminescence efficiency is plotted as a function of excitation intensity in samples of different spin density (7). At the measurement temperature of 170K, thermal quenching has reduced the luminescence intensity by 1-2 orders of magnitude. However, as the excitation intensity G is increased, there is an enhancement of the luminescence efficiency Y,(G), particularly in samples of low spin density. The explanation of the effect is that the free carriers can either recombine non-radiatively at defects or radiatively with free holes (10). A simple model assuming capture at dangling bonds predicts

the value of G for which the Si:H showing the band edge and second term in Eq 5 is unity. defect transitions.

where Yo is the efficiency in the low intensity limit, and y and are parameters related to the capture rates. Eq. 5 describes the intensity dependence qualitatively, although there are substantial deviations at high G, probably indicatiijg that the recombination model assumed is over-simplified. Furthermore, a fit to the data at low G has the predicted dependence on Nc2 as shown in Fig. 5. These results show that dangling bonds provide the major non-radiative recombination process at both high and low temperatures for the band edge luminescence transition.

A Radiative Transition at Defect States. - A luminescence transition at 0.8-0.9eV is oRen observed in a-Si:H (11). This lransition is typically found in samples in which the band edge transition is strongly quenched, for example by doping or particle bombardment. Typical spectra for doped a-Si:H are shown in Fig. 6. The energy and properties of the lower energy luminescence is the same in undoped and both n-type and p-type material, which indicates that the recombination center is a native defect rather than an impurity.

Some direct evidence that the center is in fact a dangling bond is shown in Fig. 7 (12). Here, the temperature dependence of the defect band is compared to that of the band edge transition.

Also shown is the identical temperature dependence of light induced ESR (LESK) in the same doped samples. In doped samples, the Fermi energy moves to the band tails so that the electron occupancy of the dangling bands changes and these states are diamagnetic in equilibrium. Spin resonance is only observed when carriers are optically excited out of the dangling bonds. The resulting LESR spin density depends on the recombination time ^T such that

JOURNAL DE PHYSIQUE

The luminescence intensity Y, of the defect band also depends on the recombination rates for both radiative (P,) and non-radiative (P,J processes,

7 = (P,

+

P,J-l (7)

It is usually the case that the radiative rate P, is not strongly temperature dependent, so that the temperature dependence of YL is the same as that of ^{7 .} Thus, the similar temperature dependence of Ns and Y, indicates that the same recombination mechanism occurs in both experiments. It is therefore deduced that the luminescence transition occurs at the dangling bonds (12).

Fig. 8 shows a model for the radiative transitions (11). The defect transition is assumed to occur after an electron has been captured from the conduction band by a dangling bond. From the luminescence energy the dangling bond is estimated to be a electron trap of depth about 0.5eV. However, various aspects of the defect transition remain unclear and further work is required to confirm the model shown in Fig. 8. For example, the efficiency of the transition is evidently very small for reasons which are not fully understood. The transition is expected to be by tunnelling as for the band edge transition, but again there is very limited evidence on this point, and the value for the effective Bohr radius is unknown. Finally, we point out that none of the mechanisms shown in Fig. 1 can account for the thermal quenching of the defect transition.

;

^SAMPLES

/

CONDUCTION BAND

TAIL

-

- - -

STATES

-

^/ - - -i-- Ns

yo.-

LUMINESCENCE LUMINESCENCE

^EZ

i

^-

SELF-TRAPPED

HOLESTATES ,

- ^- I _-

^-

Nh VALENCE BAND

I

( 1 l l l l l I I

4 8 8 10 12 Fig. 8 Schematic model for the

TEMPERATURE laoo/r (K-') luminescence transitions.

Fig. 7 Temperature dependence of the defect luminescence compared to that of LESR.

/

^BANDEDGE

TRANSITION

/

^x DEFECTPEAK

LUMINESCENCE 10 K

in doped a-Si:H.

I I I I I I I L

lo4 104 iv UD 106 ICP 10

@He] NOMINAL DOPING [pH,]

Defectsin Doped and Compensated a-Si:H. - The preceeding discussion emphasizes the

- .

information that is obtained about the recombination mechanisms in undoped a-Si:H in which the defect density is known. This information allows the luminescence to be used as a probe of defects in related material. As an example of how well this potential can be realized, the luminescence of doped and compensated a-Si:H is described (13). Fig. 9 shows that doping both p-type and n-type strongly quenchcs the band-edge luminescence. At doping levels above about the defect transition starts to dominate the luminescence. Typical spectra are shown in Fig. 6. Both of these features of the luminescence indicate an increase in defect density with doping. Further evidence for this behavior is found from the temperature dependence, from intensity measurements as shown in Fig. 4, and also from decay measurements (13, 14). In each case, consistent estimates of the defect density result. Confirmation of these results comes from LESR data in Fig. 9 which shows a large increase in dangling bond density with doping (13).

However, the LESR data represents a lower limit on the true defect density because it is a non- equilibrium measurement, and the observed signal depends on the illumination intensity. An alternative means of estimating the density is to assume the relation between luminescence intensity and defect density shown in Fig. 2 holds for doped material. It is unclear whether the recombination mechanism is sufficiently similar in the two cases to allow this approach.

However, recent optical measurements of the defect density yield the same densities, suggesting that the method is sound (Jackson, private communication).

C4-290 JOURNAL DE PHYSIQUE

Fig. 10 Luminescence intensity and dangling bond LESR spin density in 10-3[PHJ doped a-Si:H as a fhnction of boron compensation.

c

B

5 0

Z

V)

NOMINAL DOPING [B2Hel

I T I I

DOPED AND COMPENSATED aSi:H

- 103 [pH,] -

LUMINESCENCE 10 K X LESR 30 K

Fig. 10 shows the luminescence properties of compensated a-Si:H (13). As the compensation level increases, the luminescence intensity passes through a maximum value which is about an order of magnitude larger than in the singly doped n-type material. These data suggest that the defect density decreases with compensation. Again, the results are confirmed by the LEBR measurements also shown in Fig. 10, which find a minimum in the dangling bond density at compensation. These results lead us to deduce that the defect density in doped and compensated a-Si:H is largely determined by the position of the Fermi energy. The proposed explanation of these results in terms of an autocompensation mechanism is described in more detail elsewhere (13, 15).

>.lo - -

t

K

Conclusions.

-

The experiments described in this paper are intended to demonstrate the strong link between luminescence and ESR measurements from which a great deal of information about the recombination mechanisms can be obtained. The two experiments together provide a powerful probe of a-Si:H and related materials. It is worth commenting that ESR measures only paramagnetic centers and so its connection to the luminescence data is not immediately obvious.

For example, band tail states are not observed by ESR in equilibrium. However, these states are the radiative centers and their role in the luminescence process is relatively well established.

Rather, the fact that in all the cases discussed, the non-radiative recombination (and the defect luminescence) can be linked to the dangling bond resonance, provides clear evidence that these are the major class of defect states. Most of the experiments involve material with dangling bond densities above 1016cm-3, and so other defects present at lower densities cannot be excluded, although there is no evidence for a change in recombination properties in samples of lower spin density. One exception to this conclusion is in ion- or electron-bombarded a-Si:H in which there is clear evidence for diamagnetic defects (16). However, even in doped and compensated a-Si:H, the recombination appears to be dominated by dangling bonds.

The second point emphasized in this paper is the way in which carrier diffusion in band tail localized states determines the recombination mechanisms. There is a transition temperature of about 50K (but the transition is extcnded over a fairly wide temperalure range), which separates the low temperature tunnelling processes from the high temperature diffusion and capture mechanisms. This transition is observed in both the radiative and non-radiative processes. It

Acknowledgments. - Much of the experimental data presented here was collected in collaboration with D. K. Biegelsen and J. C. Knights, and their participation is gratefully acknowledged.

References

[ 11 ENGEMANN, D., and FISCHER, R., Proc. 5th Int. Conf. on Amorphous and Liquid Semicondi~ctors (edited by J. Stuke and W. Brenig) (1973) 949.

[2] BIEGELSEN, D. K., Solar Cells 2 (1980) 421.

[ 31 STREET, R. A., KNIGHTS, J. C., and BIEGELSEN, D. K., Phys. Rev B 18 (1978) 1880.

[4] TSANG, C. and STREET, R. A., Phys. Rev. B 19 (1979) 3027.

[ 51 NOOLANDI, J., HONG, K. M. and STREET, R. A., Solid State Commun. 34 (1980) 45.

[6] REHM, W., FISCHER, R., and BEICHLER, J., Appl. Phys. Lett., in press.

[7] STREET, R. A., Phys. Rev B 23 (1980) 861.

[ 81 BIEGELSEN, D. K., KNIGHTS, J. C., STREET, R. A., TSANG, C., and WHITE, R. M., Phil. Mag. B 37 (1978) 677.

191 KAPLAN, D., SOLOMON, I. and MOTT. N. F., J. de Physique Lettres, 39 (1978) 4.

[lo] KNIGHTS, J. C. and LUJAN, R., Appl. Phys. Lett. 35 (1979) 244.

[ll] STREET, R. A., Phys. Rev. B 21 (1980) 5775.

[12] STREET, R. A. and BIEGELSEN, D. K., Solid State Commun. 33 (1980) 1159.

[13] STREET, R. A., BIEGELSEN, D. K. and KNIGHTS, J. C., Phys. Rev (1981) in press. _{- - - - - - - -} [14] TSANG, C., and STREET, R. A., Phil. Mag. B 37 (1978) 601.

[15] BIEGELSEN, D. K., STREET, R. A. and KNIGHTS, J. C., Proc. Conf. on Tetrahedrally Bonded Amorphous Semiconductors, in press.

1161 VOGET-GROTE, U., KUMMERLE, W., FISCHER, R. and STUKE, J. Phil. Mag. 3 41 (1980) 127.