- Email: [email protected]
High field behaviour of ZnO—II
J. Phy.
Chm~. Solids
Pergamon
HIGH
Press 1970. Vol. 3 I. pp. 2391-2395.
FIELD
INVESTICJATION
Laboratories
Printed in Great
BEHAVIOUR OF THE
OF ZnO-II
PHOTOCURRENTS
H. KIESS RCA. Ltd.. Zurich.
(Heceiwd
Britain.
29 December
Switzerland 1969)
AbstractPhotocurrents in ZnO have been investigated under high field conditions. An anomalous increase of photocurrent with voltage was observed at wavelengths between A = 3750A and A = 3900 A. It is shown that this increase in Dhotocurrent is due to a photoinduced increase in the tunnelling rate of charge carriers from surface states.
IN THE past photocurrents in the system ZnO-electrolyte[ l-41, have been investigated by many people mostly with the aim of obtaining information about chemical transfer reactions. Williams and Willis[S] were the first to report about the transport mechanism of photoexcited charge carriers in ZnO using an electrolyte as blocking contact and as a means of achieving high electric fields in the ZnO surface layer adjacent to the electrolyte. They found a strong increase in photocurrent at high voltages and concluded that impact ionization across the band gap causes the photocurrent to rise steeply. It is the intention to show here that an anomalous increase in photocurrent is not necessarily due to impact ionization and that some care must be taken in the interpretation of experimental results. For the ZnO single crystals used in the work presented here, it is shown that another mechanism is responsible for the strong increase in photocurrent. The ZnO single crystals were obtained from the 3 M company and came from the same batches as used in a former investigation[6] on the high field behaviour of ZnO. The crystals were prepared as in the above work. For the experiments the crystals were mounted in a way which exposed one face to the electrolyte which formed the blocking contact with the ZnO. The other face was provided with an ohmic In-contact. The crystals were illu-
minated through the electrolyte for photocurrent measurements. A Sylvania iodine quartz lamp (700 W) or an Osram high pressure mercury lamp HBO 200 was used with a conventional monochromator or band pass filters for varying the wavelength of the incident light. The illumination of the crystal at different wavelengths was usually reduced to equal incident photon flux by varying the lamp voltage. The calibration was made with an Eppley thermo-pile and a Keithley 148 nanovoltmeter. The photocurrents were recorded with a Moseley 135 x-y recorder. In most cases an acetate buffered 2 M KCI solution with p,, - 4.5 was used as electrolyte. Variation of the electrolyte solutions and of the p,-value did not have a significant influence on the results. Figure 1 shows the photocurrent as a function of the positively biased crystal. The applied voltage is recorded on the horizontal axis and the current on the vertical axis. This recorder trace was obtained using chopped light, which allows the separation of the photocurrent from the dark current provided that the chopping period is longer than the decay or rise time of the photocurrent. Using white light the photocurrent is nearly constant up to 130 V and rises sharply at higher voltages. Significantly the sharp rise of the photocurrent is connected with a strong increase in ‘dark current’. Using monochromatic light of wave-
2392
H. KIESS
Voltage
,V
Fig. 1. Photocurrent excited by chopped light as a function of reverse bias. The lower recorder trace was gamed with light of wavelength A = 3650 A, the upper recorder trace with white light. The scales for the current and the voltage are identical for both recorder traces. The photocurrent at A = 3650 8, shows a pronounced saturation; the dark current increases from 10e6A at 150 V to 1.7. 10m5A at 200 V. With white light a strong increase in photocurrent with voltage is observed. At 150 V the dark current is also about 10e6.4, but at 180V an apparent ‘dark current’ of 8 x 10m5A is observed, which is about one order of magnitude higher than the dark current measured in the low recorder trace.
length A = 3650 A, there are two striking differences: firstly, no increase in photocurrent is observed and secondly, the dark current remains below the value of the apparent ‘dark current’ which is found in the off periods under illumination with white light. The spectral dependence of the photocurrentvoltage characteristic is shown in Fig. 2, both plotted logarithmically. The photocurrents at different wavelengths were measured with equal incident photon flux, the light intensities being about two orders of magnitude smaller than those of Fig. 1. The photocurrents represent the steady state values. For short wavelengths (A < 3700 A) the photocurrent is independent of voltage apart from
the small increase at voltages below about 2 V). At longer wavelengths it increases with voltage and especially at A = 3800 A and 3850 A exceeds the saturated photocurrent at shorter wavelengths. As the quantum efficiency defined as transported charge carriers per absorbed photon, was found to be unity at short wavelengths (A < 3750 A), it must be greater than one at the long wavelengths between 3750-39OOA. When the photocurrents were measured in the same way as in Fig. 1, using chopped light, then an apparent ‘dark current’ was also observed at wavelengths greater than 375OA. The threshold voltage for the increase in both photocurrent and apparent ‘dark current’
HIGH FIELD
I@ -
BEHAVIOUR
OF ZnO-II
2393
Voltage( V ) I 10-l
I I,.,.
I . . ..I loo
10’
I I....1
4 102
Fig. 2. Photocurrent as a function of reverse bias at different wavelengths. The incident light intensities are reduced to equal photon flux.
is shifted to lower voltage for lower light intensities. At high voltages the rise and decay time of the photocurrent on switching on and off the light is different for short and long wavelengths. At low voltages rise and decay times of the photocurrent for all wavelengths are below the response time of the recorder. At high voltages and at wavelengths greater than 3750 A, however, the photocurrent rises rapidly only to a certain value and then increases with a time constant of about I-IO set to higher values. In Fig. 3 the fast component and the steady state value of the photocurrent are plotted as a function of voltage. The fast component of the photocurrent saturates at a value which corresponds to a quantum efficiency of one, whereas the slow component represents an additional current which is not observed under illumination at A < 3750 A. It is obvious that the results of our measurements exclude the possibility of impact ionization, because no current increase beyond a quantum efficiency of one is observed at sufficiently short wavelengths. The increase
JPCS Vol. 1 I No. I I
B
with voltage of the fast component of the photocurrent at longer wavelengths as shown in Fig. 3 must be due to the diffusion of electron hole pairs into the depletion layer, to an increase in depletion layer width and to a shift of the absorption edge by the FranzKeldysh-effect. We will not be concerned with these physical phenomena in this article but restrict our interest to the increase in quantum efficiency due to the slow process. It is known from studies[6] of the dark l-V-characteristics and from the dependence of the capacitance as a function of positive d.c.-bias, that the sharp rise in dark current is due to tunnelling of charge carriers from surface states. Furthermore it has been established that at low voltages a Schottky barrier exists with a spatially homogeneous concentration of deep and shallow ionized donors of about 8 X 10”cm-“. At high voltages (> 10 V), however, the depletion layer consists of a layer with deep and shallow ionized donors right at the surface, adjacent to a layer with shallow ionized donors with a concentration of about 2 X lOI cme3. The region with the low concentration of ionized donors is
2394
H.
KIESS
Fig. 3. Photocurrent as a function of voltage at A = 3800 A. The (A) refer to the steady state value of the photocurrent and the (0) to its fast component. The horizontal line represents the saturated photocurrent
formed by capture of tunnel-injected electrons by the deep donor centres, leaving only the shallow centres ionized. Assume now that high voltage (> 10 V) is applied and that the crystal is illuminated with light of such a wavelength that it penetrates the layer with low ionized donor concentration. Then the following situation arises. Trapping of holes in the region with low donor concentration increases the space charge density there and hence increases the field at the surface. This gives rise to a higher tunnelling rate of charge carriers from the surface states. Therefore a light induced tunnelling current flows through the depletion layer in addition to the recombination free, saturated photocurrent. The magnitude of this tunnelling current depends only on the degree to which the space charge density is increased by the capture of holes. In these circumstances the observed quantum efficiency will be greater than unity. If the penetration depth of the
at ,\ = 3700 A.
light is smaller than or equal to the width of the region with high space charge density. the trapping of holes may have an insignificant effect on the space charge density or no trapping at all may be possible, if for example the trapping centres were identical with the deep donors in our ZnO crystals. This mechanism explains the main features of the spectral dependence of the photocurrent and the increase of quantum efficiency above one. It also explains why the increase of photocurrent at low light levels starts at lower voltages, since then the saturated photocurrent is so small that it becomes comparable with the tunnelling current. The question arises whether the trapping of holes can be detected. Supposing the voltage is kept constant and the space charge density is increased by hole trapping. then the total charge stored in the depletion layer also increases. Therefore the charges measured in the dark and under illumination should be different. This was
HIGH
FIELD
BEHAVIOUR
OF ZnO-II
2395
Time ( rd. units 1 I
.
.
..I
I
*
aa.1
I
.
.
..I
Fig. 4. Transient current on switching off the voltage. The lower curve is observed when the crystal is in the dark or illuminated with light of a wavelength shorter than 3750 A. The upper curve represents the transient current at A = 3800 A.
actually found. The result is shown in Fig. 4, where the transient current on switching off the voltage is shown. The curve in the dark and under illumination with short wavelength light (A 5 3700 A) are identical. However, a higher discharge current is observed under illumination with light of wavelength between 3750-4000,& The plateau, which shows up in the millisecond region, is not understood up to now, but it is probably connected with the regeneration of the surface states.
would like to thank W. J. Merz and J. Sandercock for suggestions to improve this manuscript.
Acknowledgements-l
REFERENCES 1. GERISCHER
H., J. elecrrochem.
Sot.
113, 1174
( 1966). 2. MORRISON S. R. and FREUND T.,J. them. Phys. 47, I543 ( 1967). 3. GOMES W. P., FREUNDT. and MORRISON S. R.. J. electrochem. Sot. 115.8 I8 ( 1968). 4. HAUFFE K., Rec. pure opp/. Chem. 18.79 (1968). 5. WILLIAMS R. and WILLIS A.. J. appl. Phys. 39. 373 I (I 968). 6. KIESS H.,./. Phys. Chem. Solids31.2379 (1970).
Chm~. Solids
Pergamon
HIGH
Press 1970. Vol. 3 I. pp. 2391-2395.
FIELD
INVESTICJATION
Laboratories
Printed in Great
BEHAVIOUR OF THE
OF ZnO-II
PHOTOCURRENTS
H. KIESS RCA. Ltd.. Zurich.
(Heceiwd
Britain.
29 December
Switzerland 1969)
AbstractPhotocurrents in ZnO have been investigated under high field conditions. An anomalous increase of photocurrent with voltage was observed at wavelengths between A = 3750A and A = 3900 A. It is shown that this increase in Dhotocurrent is due to a photoinduced increase in the tunnelling rate of charge carriers from surface states.
IN THE past photocurrents in the system ZnO-electrolyte[ l-41, have been investigated by many people mostly with the aim of obtaining information about chemical transfer reactions. Williams and Willis[S] were the first to report about the transport mechanism of photoexcited charge carriers in ZnO using an electrolyte as blocking contact and as a means of achieving high electric fields in the ZnO surface layer adjacent to the electrolyte. They found a strong increase in photocurrent at high voltages and concluded that impact ionization across the band gap causes the photocurrent to rise steeply. It is the intention to show here that an anomalous increase in photocurrent is not necessarily due to impact ionization and that some care must be taken in the interpretation of experimental results. For the ZnO single crystals used in the work presented here, it is shown that another mechanism is responsible for the strong increase in photocurrent. The ZnO single crystals were obtained from the 3 M company and came from the same batches as used in a former investigation[6] on the high field behaviour of ZnO. The crystals were prepared as in the above work. For the experiments the crystals were mounted in a way which exposed one face to the electrolyte which formed the blocking contact with the ZnO. The other face was provided with an ohmic In-contact. The crystals were illu-
minated through the electrolyte for photocurrent measurements. A Sylvania iodine quartz lamp (700 W) or an Osram high pressure mercury lamp HBO 200 was used with a conventional monochromator or band pass filters for varying the wavelength of the incident light. The illumination of the crystal at different wavelengths was usually reduced to equal incident photon flux by varying the lamp voltage. The calibration was made with an Eppley thermo-pile and a Keithley 148 nanovoltmeter. The photocurrents were recorded with a Moseley 135 x-y recorder. In most cases an acetate buffered 2 M KCI solution with p,, - 4.5 was used as electrolyte. Variation of the electrolyte solutions and of the p,-value did not have a significant influence on the results. Figure 1 shows the photocurrent as a function of the positively biased crystal. The applied voltage is recorded on the horizontal axis and the current on the vertical axis. This recorder trace was obtained using chopped light, which allows the separation of the photocurrent from the dark current provided that the chopping period is longer than the decay or rise time of the photocurrent. Using white light the photocurrent is nearly constant up to 130 V and rises sharply at higher voltages. Significantly the sharp rise of the photocurrent is connected with a strong increase in ‘dark current’. Using monochromatic light of wave-
2392
H. KIESS
Voltage
,V
Fig. 1. Photocurrent excited by chopped light as a function of reverse bias. The lower recorder trace was gamed with light of wavelength A = 3650 A, the upper recorder trace with white light. The scales for the current and the voltage are identical for both recorder traces. The photocurrent at A = 3650 8, shows a pronounced saturation; the dark current increases from 10e6A at 150 V to 1.7. 10m5A at 200 V. With white light a strong increase in photocurrent with voltage is observed. At 150 V the dark current is also about 10e6.4, but at 180V an apparent ‘dark current’ of 8 x 10m5A is observed, which is about one order of magnitude higher than the dark current measured in the low recorder trace.
length A = 3650 A, there are two striking differences: firstly, no increase in photocurrent is observed and secondly, the dark current remains below the value of the apparent ‘dark current’ which is found in the off periods under illumination with white light. The spectral dependence of the photocurrentvoltage characteristic is shown in Fig. 2, both plotted logarithmically. The photocurrents at different wavelengths were measured with equal incident photon flux, the light intensities being about two orders of magnitude smaller than those of Fig. 1. The photocurrents represent the steady state values. For short wavelengths (A < 3700 A) the photocurrent is independent of voltage apart from
the small increase at voltages below about 2 V). At longer wavelengths it increases with voltage and especially at A = 3800 A and 3850 A exceeds the saturated photocurrent at shorter wavelengths. As the quantum efficiency defined as transported charge carriers per absorbed photon, was found to be unity at short wavelengths (A < 3750 A), it must be greater than one at the long wavelengths between 3750-39OOA. When the photocurrents were measured in the same way as in Fig. 1, using chopped light, then an apparent ‘dark current’ was also observed at wavelengths greater than 375OA. The threshold voltage for the increase in both photocurrent and apparent ‘dark current’
HIGH FIELD
I@ -
BEHAVIOUR
OF ZnO-II
2393
Voltage( V ) I 10-l
I I,.,.
I . . ..I loo
10’
I I....1
4 102
Fig. 2. Photocurrent as a function of reverse bias at different wavelengths. The incident light intensities are reduced to equal photon flux.
is shifted to lower voltage for lower light intensities. At high voltages the rise and decay time of the photocurrent on switching on and off the light is different for short and long wavelengths. At low voltages rise and decay times of the photocurrent for all wavelengths are below the response time of the recorder. At high voltages and at wavelengths greater than 3750 A, however, the photocurrent rises rapidly only to a certain value and then increases with a time constant of about I-IO set to higher values. In Fig. 3 the fast component and the steady state value of the photocurrent are plotted as a function of voltage. The fast component of the photocurrent saturates at a value which corresponds to a quantum efficiency of one, whereas the slow component represents an additional current which is not observed under illumination at A < 3750 A. It is obvious that the results of our measurements exclude the possibility of impact ionization, because no current increase beyond a quantum efficiency of one is observed at sufficiently short wavelengths. The increase
JPCS Vol. 1 I No. I I
B
with voltage of the fast component of the photocurrent at longer wavelengths as shown in Fig. 3 must be due to the diffusion of electron hole pairs into the depletion layer, to an increase in depletion layer width and to a shift of the absorption edge by the FranzKeldysh-effect. We will not be concerned with these physical phenomena in this article but restrict our interest to the increase in quantum efficiency due to the slow process. It is known from studies[6] of the dark l-V-characteristics and from the dependence of the capacitance as a function of positive d.c.-bias, that the sharp rise in dark current is due to tunnelling of charge carriers from surface states. Furthermore it has been established that at low voltages a Schottky barrier exists with a spatially homogeneous concentration of deep and shallow ionized donors of about 8 X 10”cm-“. At high voltages (> 10 V), however, the depletion layer consists of a layer with deep and shallow ionized donors right at the surface, adjacent to a layer with shallow ionized donors with a concentration of about 2 X lOI cme3. The region with the low concentration of ionized donors is
2394
H.
KIESS
Fig. 3. Photocurrent as a function of voltage at A = 3800 A. The (A) refer to the steady state value of the photocurrent and the (0) to its fast component. The horizontal line represents the saturated photocurrent
formed by capture of tunnel-injected electrons by the deep donor centres, leaving only the shallow centres ionized. Assume now that high voltage (> 10 V) is applied and that the crystal is illuminated with light of such a wavelength that it penetrates the layer with low ionized donor concentration. Then the following situation arises. Trapping of holes in the region with low donor concentration increases the space charge density there and hence increases the field at the surface. This gives rise to a higher tunnelling rate of charge carriers from the surface states. Therefore a light induced tunnelling current flows through the depletion layer in addition to the recombination free, saturated photocurrent. The magnitude of this tunnelling current depends only on the degree to which the space charge density is increased by the capture of holes. In these circumstances the observed quantum efficiency will be greater than unity. If the penetration depth of the
at ,\ = 3700 A.
light is smaller than or equal to the width of the region with high space charge density. the trapping of holes may have an insignificant effect on the space charge density or no trapping at all may be possible, if for example the trapping centres were identical with the deep donors in our ZnO crystals. This mechanism explains the main features of the spectral dependence of the photocurrent and the increase of quantum efficiency above one. It also explains why the increase of photocurrent at low light levels starts at lower voltages, since then the saturated photocurrent is so small that it becomes comparable with the tunnelling current. The question arises whether the trapping of holes can be detected. Supposing the voltage is kept constant and the space charge density is increased by hole trapping. then the total charge stored in the depletion layer also increases. Therefore the charges measured in the dark and under illumination should be different. This was
HIGH
FIELD
BEHAVIOUR
OF ZnO-II
2395
Time ( rd. units 1 I
.
.
..I
I
*
aa.1
I
.
.
..I
Fig. 4. Transient current on switching off the voltage. The lower curve is observed when the crystal is in the dark or illuminated with light of a wavelength shorter than 3750 A. The upper curve represents the transient current at A = 3800 A.
actually found. The result is shown in Fig. 4, where the transient current on switching off the voltage is shown. The curve in the dark and under illumination with short wavelength light (A 5 3700 A) are identical. However, a higher discharge current is observed under illumination with light of wavelength between 3750-4000,& The plateau, which shows up in the millisecond region, is not understood up to now, but it is probably connected with the regeneration of the surface states.
would like to thank W. J. Merz and J. Sandercock for suggestions to improve this manuscript.
Acknowledgements-l
REFERENCES 1. GERISCHER
H., J. elecrrochem.
Sot.
113, 1174
( 1966). 2. MORRISON S. R. and FREUND T.,J. them. Phys. 47, I543 ( 1967). 3. GOMES W. P., FREUNDT. and MORRISON S. R.. J. electrochem. Sot. 115.8 I8 ( 1968). 4. HAUFFE K., Rec. pure opp/. Chem. 18.79 (1968). 5. WILLIAMS R. and WILLIS A.. J. appl. Phys. 39. 373 I (I 968). 6. KIESS H.,./. Phys. Chem. Solids31.2379 (1970).