Memory effects in highly resistive p–i–n heterojunctions for optical applications

Memory effects in highly resistive p–i–n heterojunctions for optical applications

Thin Solid Films 403 – 404 (2002) 363–367 Memory effects in highly resistive p–i–n heterojunctions for optical applications R. Schwarza,b, P. Louroa,...

705KB Sizes 0 Downloads 5 Views

Thin Solid Films 403 – 404 (2002) 363–367

Memory effects in highly resistive p–i–n heterojunctions for optical applications R. Schwarza,b, P. Louroa,*, Yu. Vygranenkoa, M. Fernandesa, M. Vieiraa, M. Schubertc a

Electronics and Communications Department, ISEL, P-1949-014 Lisboa, Portugal b ´ Departamento de Fısica, IST, P-1096 Lisboa, Portugal c ¨ Physikalische Elektronik, IPE, Universitat ¨ Stuttgart, D-70569 Stuttgart, Germany Institut fur

Abstract Large area p–i–n diode structures based on amorphous hydrogenated silicon can be used as single element image sensors where the information is read out by a scanning laser beam. A high sensitivity is reached with silicon–carbon alloy contact layers. The higher defect density in the large band gap material is usually a problem for efficient carrier collection in solar cell applications. When used as an image sensor, however, the charge stored in deep defects represents an easy way to realize short-term image storage. In the case of a p-(Si:H)yi-(Si:H)yn-(Six C1yx :H) sensor structure we have measured a memory effect of approximately 1% after several minutes of image projection. Metastable sensor degradation is observed in accordance with the Staebler–Wronski effect. Fast degradation of sensor performance — corresponding to 90% erasable image storage capability — was studied in an unalloyed structure using a Nd:YAG laser system. The response can be modeled by a stretched exponential decay with parameters depending on the laser pulse energy. 䊚 2002 Elsevier Science B.V. All rights reserved. Keywords: Diode structures; Amorphous hydrogenated silicon; Scanning laser beam

1. Introduction Recently, detailed characterization was done on thin film image sensors composed of a p–i–n structure based on hydrogenated amorphous silicon (a-Si:H) and deposited onto ZnO-coated glass substrates. Particularly low minimal image intensities were achieved in structures where the slightly doped contact layers consisted of hydrogenated amorphous silicon–carbon films (aSixC1yx:H). Such laser scanned photodiode (LSP) image sensors were characterized with respect to optical and electrical parameters of the individual layers w1x, I–V and C–V characteristics under illumination of the devices w2x, and spectral response under various bias conditions w3 x . It is to be expected that the well-known degradation of optoelectronic properties of a-Si:H under illumination, known as the Staebler–Wronski effect (SWE) w4x, will also affect the image sensor response. In fact, the higher * Corresponding author.

the material quality, the larger this effect should be. As an example, the efficiency of an a-Si:H solar cell decreased to 50% of its initial level after light soaking under 500 mWycm2 intensity during 1000 h w5x. However, the degradation should be reversible either by long-term relaxation or by thermal annealing. This metastable effect can be exploited in a positive way for image storage. In an earlier work Paasche et al. measured a charge storage time of several days in a homogeneous SiC thin film w6x. The details of sensor degradation, image storage, and sensor recovery will actually be much more complex than what is known from single layer behavior. In a heterostructure image sensor we also have to deal with interface quality, band offsets, potential drops across the wide-band gap contact layers, and screening of the electric field due to charges accumulated in deep defects. This work describes the small reduction in sensor response of a p-(Si:H)yi-(Si:H)yn-(Six C1yx :H) sensor structure after illumination with a square-shaped white light image of some 0.5 mWycm2 intensity. The stored

0040-6090/02/$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 0 - 6 0 9 0 Ž 0 1 . 0 1 5 1 5 - 2

R. Schwarz et al. / Thin Solid Films 403 – 404 (2002) 363–367

364 Table 1 Main characteristics of the sensors Sensor code 噛M008821 噛M006301

RS (Vycm2) 3

1.2=10 4.0=105

Voc (V) 0.84 0.70

image has a level of approximately 1% after several minutes of illumination. To study the kinetics of sensor degradation we have used short pulses from a Nd:YAG laser with a pulse energy of several milliJoules and modeled the decay with a stretched exponential as is done in the case of the SWE effect. As an example, the sensor response of an all a-Si:H based structure decreased by 50% after only 10 shots of 5-ns duration at a power level of 6 MWycm2. We expect that the sensor recovery should also be controlled by the kinetics know from the metastable SWE effect, unless permanent damage is done to the device because of excessive heating during exposure. 2. Experimental details Two types of large area image sensors are used in this study. The first sensor (a-Si-sensor) has the following layer series: glassyZnO:Alyp-(a-Si:H)yi-(a-Si:H)y n-(a-Si:H)yAl, where all layers of the p–i–n diode were deposited by plasma enhanced chemical vapor deposition at 13.56 MHz w7x. In the second sensor (a-SiCsensor) the n-type contact layer with a bandgap of 1.82 eV was replaced by an alloy film of a-SixC1yx:H with a bandgap of 2.10 eV. The sensor area was 4=4 cm2 in both cases.

Jsc (mAycm2) 2.12 0.20

J0 (mAycm2) y12

6.0=10 1.0=10y10

n 1.8 2.9

Key cell characteristics such as: open-circuit voltage Voc under AM1 illumination, the short-circuit current JSC, the ideality factor of the dark characteristic n, and the reverse bias saturation current JO are listed in Table 1. For image read-out a HeNe laser beam with an intensity of 25 mWycm2 or lower was scanned across the sensor. The a.c. component of the photocurrent after chopping the laser beam with 83 Hz was fed into a lock-in amplifier and recorded by a PC. The image intensity was approximately 500 mWycm2. Fast degradation was achieved with the green line at 532 nm of a frequency-doubled Nd:YAG laser that was projected onto the image sensor. The laser spot size was approximately 3 mm in diameter and the energy was 0.8–3.2 mJ per pulse corresponding to a momentariness power density of 6 MWycm2. The maximum fluence reached was 1020 photonsycm2. The low repetition rate of 10 Hz assures that the sample temperature was kept at room temperature level. 3. Memory effect Fig. 1 shows the recorded image of a white light square projected onto the a-SiC sensor 噛M006301. The a.c. signal, Ia.c., of the detector response is reduced in

Fig. 1. Square image projected onto the heterostructure sensor 噛M006301.

R. Schwarz et al. / Thin Solid Films 403 – 404 (2002) 363–367

365

Fig. 3. Laser spot image remaining in the a-Si:H sensor (噛M002281) after high power pulses from the Nd:YAG laser.

can be reversed by annealing, as is known from light soaking experiments in solar cells. 4. Pulsed laser degradation

Fig. 2. (a) Line scans before (upper curve) and after illumination (lower curve) of the heterostructure sensor 噛M006301 with a square image. (b) Difference signal showing the memory effect.

the illuminated regions. As has been reported earlier w2x the lowest image intensity level that can easily be detected is approximately 20 mWycm2. For intensities larger than approximately 10 mWycm2, the a.c. signal saturates at its lowest level. In Fig. 2a we present a line scan without projected image where some spatial inhomogeneity is seen. After exposure during 5 min we have recorded the sensor response in the dark, I2, and compared it with the response before exposure, I1. The difference signal (memory signal), IM, defined as:

To further investigate the relation between sensor response degradation and the memory effect we have used rapid light soaking with strong light pulses from the Nd:YAG laser. Fig. 3 shows a 2-D plot of the response of the all a-Si:H sensor 噛M008821 in a region where two spots where illuminated, spot A with 0.8-mJ pulses and spot B, approximately 3 mm apart, at an energy of 3.2 mJ per pulse. The line scan of Fig. 4 shows the sensor signal after a fluence of approximately 1020 phycm2 for both spots. The final signal level is similar, however, the decay kinetics are quite different for the two different laser energies used, as has been described in the literature w8 x . The signal decay during laser exposure is plotted on a double-logarithmic plot in Fig. 6. We have also plotted the fit to a stretched exponential type of decay as given by:

IMsyŽI1yI2.yI1 is shown in Fig. 2b. Two effects can be detected, first, the difference signal corresponds to approximately 0.6% of the initially projected image. Secondly, there is a small overall sensor signal reduction of approximately 0.5%. Both effects

Fig. 4. Line scan across the exposed sensor area of Fig. 3.

R. Schwarz et al. / Thin Solid Films 403 – 404 (2002) 363–367

366

Table 2 Decay parameters from stretched exponential fit

Spot A Spot B

Pulse energy

t0 (s)

b (y)

F0 (phycm2)

0.8 mJ 3.2 mJ

5400 40

0.45 0.29

8=1018 1017

Characteristic time t0, exponent b, characteristic flux F0 corresponding to t0).

IŽt.sIminqŽImaxyImin.UexpŽyŽtyt.b.

(2)

where the characteristic time t and the exponent b depend on the laser pulse energy as given in Table 2. Already a factor of 4 of increased power is enough to accelerate the degradation 100 times. 5. Discussion

Fig. 5. Band diagram obtained by numerical simulation of the two sensor structures under AM1 illumination.

We will now look at the correlation between the three effects we observed, image storage (memory effect), metastability (Staebler–Wronski effect), and degradation.

monoatomic silicon layers we would expect the strongest memory effect in the SiC heterostructures. This will be the object of a further study.

5.1. Memory effect

5.2. Degradation kinetics

The basic hypothesis we propose to explain the observed memory effect is based on charge storage in deep defects of the central a-Si:H and, to some extent, the thin a-SiC:H layers. It is, for example, known that in single silicon-carbon alloy films charge storage times of several days are achieved at room temperature since the reemission rate of trapped carriers is low w6x. Other high-defect density regions can contribute to the memory effect in a p–i–n structure, like the pyi interface or the iyn interface at the front and back of the detector, respectively. Finally, in a sensor structure the additional stored charges will affect the potential distribution and can, in particular, screen the electric field and thus reduce the sensor signal in the illuminated region. Typical potential profiles for the two structures investigated here are shown in Fig. 5 for AM1 illumination and zero applied bias using the AMPS numerical simulation program w9x. Details of the changes of the I–V characteristic (the so-called S-shape) and the concomitant change of internal fields was studied in 1993 with respect to solar cell degradation of a p–i–n heterojunction including an aSiC:H window layer w10x. There it was shown that stored charges can even lead to field reversal and therefore low carrier collection at intermediate forward bias or under short-circuit conditions. The weak solar cell performance is an advantage for sensor applications. The image sensor is actually based on the signal reduction by illumination. Since high band gap alloy layers contain traps in larger numbers and at lower energies compared to the

Table 2 summarizes the parameters found in the fast degradation experiment. The number of pulses (x-axis of Fig. 6) can be recalculated into fluence by taking the laser pulse energy, the repetition rate of 10 Hz, and the exposed area of approximately 0.1 cm2 into account. The decay kinetics is compatible with reports on the SWE in amorphous silicon solar cells. As an example, the shape of the Nd:YAG laser spot is stored in the homojunction sensor already after two pulses of 5 ns each. The stored image level reaches approximately 90% after a 1-min exposure. This point

Fig. 6. Pulsed laser degradation described by stretched-exponential fits wEq. (2)x.

R. Schwarz et al. / Thin Solid Films 403 – 404 (2002) 363–367

of view means looking at fast degradation in a positive way as an image memory effect. 5.3. Metastability All the light-induced changes described so far, should be reversible since only metastable defects were supposedly created by prolonged white light or pulsed laser exposure. Only dopand diffusion or laser induced phase transitions could lead to permanent degradation. In that sense, the two terms degradation and memory effect are equivalent. 6. Conclusions We studied the possibility of using a large-area aSi:H based image sensors for image storage. The usually unwanted memory effect is explained by the well known reversible degradation of a-Si:H layers and p-i–n structures after prolonged illumination. The stored image has a level of below 1% under low-level white light illumination and reaches 90% after high power pulsed laser irradiation during 1 min. The sensor response can be recovered by thermal annealing as known from the Staebler–Wronski effect in solar cells.

367

Acknowledgements We would like to thank J. Gloeckner for sample deposition and helpful discussions. This work was financially supported by PRAXISyPyEEIy12183y1998. References w1x M. Vieira, M. Fernandes, J. Martins, P. Louro, A. Macarico, R. ¸ Schwarz, M. Schubert, Mater. Res. Soc. Symp. Proc., San Francisco, USA, 2000 (in press). w2x M. Vieira, M. Fernandes, P. Louro, A. Macarico, R. Schwarz, ¸ M. Schubert, IEEE Sensor J., June, 2001. w3x M. Fernandes, Yu. Vygranenko, N. Leite, M. Vieira, Presented at the European MRS Meeting, Symposium F, Strasbourg, France, June 2001. w4x D.L. Staebler, C.R. Wronski, Appl. Phys. Lett. 31 (1977) 292. w5x A. Terakawa, M. Shima, M. Isomura, M. Tanaka, S. Kiyama, S. Tsuda, H. Matsunami, J. Non-Cryst. Solids 227-230 (1998) 1267. w6x S. Paasche, G.H. Bauer, J. Non-Cryst. Solids 77&78 (1985) 1433. w7x C. Koch, M. Ito, M. Schubert, J.H. Werner, Mater. Res. Soc. Symp. Proc. 575 (1999) 749. w8x M. Stutzmann, J. Nunnenkamp, M.S. Brandt, A. Asano, M.C. Rossi, J. Non-Cryst. Solids 137&138 (1991) 231. w9x Web site: http:y ywww.psu.eduydeptyamps w10x W. Kopetzky, R. Schwarz, Appl. Phys. Lett. 62 (1993) 2959.