Photoconductor gain mechanisms in GaN ultraviolet ...

Photoconductor gain mechanisms in GaN ultraviolet detectors E. Munõz,a) E. Monroy, J. A. Garrido, I. Izpura, F. J. Sańchez, M. A. Sańchez-Garcıá, and E. Calleja Depto. de Ingenierıá Electrońica, ETSI Telecommunicacioń, Ciudad Universitaria, 28040-Madrid, Spain

B. Beaumont and P. Gibart CRHEA-CNRS, Parc Sophia Antipolis, rue Bernard Gregory, 06560 Valbonne, France

~Received 3 April 1997; accepted for publication 13 June 1997! GaN photoconductive detectors have been fabricated on sapphire substrates by metal organic vapor phase epitaxy and gas-source molecular beam epitaxy on Si ~111! substrates. The photodetectors showed high photoconductor gains, a very nonlinear response with illuminating power, and an intrinsic nonexponential photoconductance recovery process. A novel photoconductor gain mechanism is proposed to explain such results, based on a modulation of the conductive volume of the layer. © 1997 American Institute of Physics. @S0003-6951~97!01933-5#

GaN and AlGaN alloys are very promising semiconductor materials for ultraviolet ~UV! photodetectors. These wide band gap semiconductors ~3.4–6.2 eV!, match the UV photon energy region, withstand high temperatures, and can be obtained through the technology already being developed for GaN-based light emitting diode manufacturing. Because of their simplicity, photoconductive GaN and AlGaN UV detectors have attracted significant interest early on. Properties of undoped n-type and high resistivity GaN photoconductive UV detectors have been reported.1–4 Besides, photoconductivity has been one of the characterization tools used to elucidate the properties and origin of the shallow and deep levels that are present in current GaN layers.5,6 A simple modeling of n-type photoconductive devices is based on considering a gain mechanism ~number of electrons circulating through the sample per absorbed photons per unit time! equal to the ratio between the hole ~minority! lifetime and the electron transit time ( t h /t t ). In this model,7 the photoconductivity decay and frequency response would be determined by t h , and linear detectors could be obtained for both static ~dc! and dynamic ~ac! photon fluxes. Present GaN technology is far from producing low defect-density layers. Significant concentrations of shallow and deep levels are found in both metal-organic vapor phase epitaxy ~MOVPE! and gas-source molecular beam epitaxy ~GSMBE! layers. In nonintentionally doped layers, shallow donors render the material n type, and a yellow luminescence band ('2.3 eV) is present in most of the cases. The origin and gap position of such levels are still under debate.8,9 Current GaN epitaxial layers have a large concentration of dislocations (109 – 1010 cm22), resulting from the low-angle grains ~columns! present in these epitaxial layers.10 The data reported on GaN photoconductive UV detectors vary significantly, and, because of the above material problems of GaN layers, it is sometimes difficult to distinguish between intrinsic and sample-dependent photoconductive behavior. However, a number of consistent findings have been reported, some of them showing that GaN photoconductive devices do not behave as the above simple model a!

Electronic mail: [email protected]

870

Appl. Phys. Lett. 71 (7), 18 August 1997

indicates: ~i! the photocurrent response starts for incident photons with energy below the gap, and the ac gain raises to the 103 – 104 range for photons above the gap; ~ii! the photocurrent response is highly ~sub! nonlinear, i.e., depends on the illumination irradiance ~P!, and a gain dependence with ' P 20.5 has been reported; ~iii! in these undoped GaN layers, very slow ~tens of minutes!, nonexponential conductance decays are produced after shutting off the light source; ~iv! although detailed measurements were not reported, available data indicate that the photoconductor gain depended significantly on temperature. Some of these characteristics were claimed to be due to the presence of shallow and deep traps. The lack of linearity is a very relevant issue when considering applications of these devices. In this work, GaN photoconductive detectors are characterized and a novel gain mechanism, that is dominant in these detectors, is described. It is proposed that the photoconductive gain mechanism in GaN devices is due to the modulation of the layer’s electrical conduction volume. This spatial modulation is caused by the photovoltaic response of the space charge regions present at the GaN surfaces and charge arrays, which change their widths as carriers are photogenerated. Wurtzite GaN samples grown by MOVPE on sapphire substrates and by GSMBE on Si ~111! substrates, have been used in this study.11,12 The GaN layers were n-type, nominally undoped, with a thickness from 0.5 to 7 mm, and AlN buffer layers were always used. In the MOVPE layers, the electron concentration, as obtained by Hall measurements, was in the 1016 – 1018 cm23 range, and electron mobility at room temperature ranged from 50 to 120 cm2/V s. The GSMBE GaN showed typical electron mobilities of 30 cm2/V s. GaN photodetector bars with Ti/Al or In ohmic contacts, separated up to 1 mm, were typically used. Spectral responsivity studies were performed by using a 600 W globar ~quartz tungsten! lamp and a 900 W Xe lamp. To study the dependence of the photoconductor gain with optical power, an He–Cd laser ~325 nm! was used as excitation source. When performing dc gain measurements, very special care was taken in determining the equilibrium ~dark! conductance by waiting for very long periods of time ~hours!. For ac gain measurements, a MKII frequency programmable chopper

0003-6951/97/71(7)/870/3/$10.00

Copyright ©2001. All Rights Reserved.

© 1997 American Institute of Physics

FIG. 1. dc photoconductor gain vs optical power density, for different temperatures, measured in GaN photodetectors. A dependence with P 2k (k ;0.9) is found over five decades, obtaining that k decreases as the temperature is increased. The insert shows the responsivity with the incident photon wavelength.

and an EG&G 5208 lock-in amplifier were used. While performing the photoconductance measurements, the bias electric fields applied to the samples were below 103 V/cm. A typical spectral dc responsivity in MOVPE samples at 300 K is presented in the insert of Fig. 1. A responsivity above 103 A/W is reached for photons above the gap. A study of the dc gain with incident optical power and temperature was made (l5325 nm). Results are shown in Fig. 1. For the powers used in these experiments, higher than 1022 W/m2, the photoconductivity gain decreases as P 2k , with k'0.9 over five decades. In GSMBE GaN layers on Si, a similar behavior is observed. The photoconductor gain decreases with temperature. The ac photoconductor gain was also studied. As a general trend, the same P 2k dependence is followed, with k decreasing as the frequency of the light chopper is increased. As an example, the exponent k decreases from 0.65 at 66 Hz to a value of 0.55 by chopping at 135 Hz. These results are consistent with the data reported by Kung2 and by Binet.4 The photoconductance transient recovery has been recorded for very extended time periods ~Fig. 2!, after turning

FIG. 2. Photoresponse decay of a GaN detector after a He–Cd laser light pulse (l5325 nm) of 4.6 mW/m2. The nonexponential law ~dotted line! is fitted quite well by the simulation results ~solid line! based on the spatial conduction modulation. The semilog plot of the inset shows the very nonexponential initial temporal response. Appl. Phys. Lett., Vol. 71, No. 7, 18 August 1997

off an excitation of 4.6 W/m2. A nonexponential conductance decay is always found, indicating that we face an intrinsic nonexponential process. As described below, the changes in spatial conduction modulation by the photovoltaic effect fit such a result. Our experimental results are in agreement with the photoconductivity studies performed by Kung5 and by Qiu.6 The high sensitivity of the GaN photoconductors, their very nonlinear response with illuminating power, and the intrinsic nonexponential photoconductance recovery process, are all explained with the modulation of the conductive volume of the layer. In high band gap materials, significant band bendings are expected at pinned surfaces, interfaces, grain boundaries, and charged dislocations, that are most likely to be present in current undoped GaN layers. By illumination above 2.5 eV9 or above the gap, electrons are photoionized, swept through the space charge regions, thus being spatially separated from the localized capture centers. This process changes the potential barrier heights, f b , and the space charge region widths, thus modulating the effective conductive volume of the layer and the sample conductance. Let us point out that power densities of 100 W/m2 correspond to UV photon fluxes in the 1016 photons/cm2 s range, and the intrinsic GaN recombination times obtained in photoluminescence measurements, below the 10 ns range, would only generate 1012 cm23 free carriers in our samples. Furthermore, because of this spatial separation of the photogenerated carriers, once in the dark, the extra free electrons will slowly be recaptured basically through thermoionic processes, and this capture process modulates again the potential barrier, and thus the conductance recovery dynamics. A kind of feedback effect is active resulting in a very strong nonexponential regime. Under illumination above the gap of the n-type samples, holes will be photogenerated and will be captured by surface states and negative charge arrays. This will contribute markedly to the f b change and will result in a very significant sample conductance modulation. Holes will slowly recombine either with free electrons ~spatially separated! or via the trapped electrons at the interfaces. The proposed model has been tested in detail by computer simulation. As an example, for the simplest case of a GaN layer (N d 51017 cm23) with a surface band bending of f b 50.7 eV, that corresponds to a charge density of 9.1011 electrons cm22, the thermoionic currents, photovoltage, space charge region width, and spatial modulation effects were determined. The simulated photoconductance decay is shown in Fig. 2, fitting very well the experimental curve. For this structure, the calculated dependence of gain with irradiance and temperature are shown in Fig. 3 for the dc case, indicating good agreement with the results shown in Fig. 1. Thus, the spatial modulation mechanism is able to describe the basic behavior of GaN photoconductors. The predicted flat gain region in Fig. 3, at very low irradiances and high temperatures, has not been found in our present experiments. For the ac gain, the above dependence of the exponent k with chopper frequency has also been confirmed by our model. More details will be presented in another publication. In thin photoconductive GaAs photodetectors @~HEMT! channels#, a similar behavior for the dc and ac gains was found,13 and the role of surface band bendings in


Munõz et al.

871

This work was partially supported by Plan Futures 1996–9, Contract No. TECN-96-1054, and by CICYT TIC95-0770. We would like to thank UAM Professors F. Jaque and I. Aguirre, and C. Co´rdoba, for their UV calibration assistance. Discussions with Professor R. Beresford, under BBV support, are acknowledged. 1

FIG. 3. Computer simulations of the dc photoconductor gain vs irradiance and temperature, as obtained from the proposed spatial conduction modulation mechanism. The model predicts the experimental behavior shown in Fig. 1.

this dependence was invoked. The importance of these surface photovoltaic effects in photoconductivity spectroscopy of thin layers, has recently been reported by Izpura et al.14 In summary, the very high gains obtained in GaN photoconductive detectors are due to a modulation mechanism of the conductive volume of the layer. As carriers are photogenerated, they are spatially separated by potential barriers generated by surface and bulk dislocation-related band bendings. Carrier recombination and capture are controlled by such potential barriers, and an intrinsic nonexponential recovery process is present.

872

M. Asif Khan, J. N. Kuznia, D. T. Olson, J. M. Van Hove, and M. Blasingame, Appl. Phys. Lett. 60, 2917 ~1992!. 2 K. S. Stevens, M. Kinniburgh, and R. Beresford, Appl. Phys. Lett. 66, 3518 ~1995!. 3 D. Walker, X. Zhang, P. Kung, A. Saxler, S. Javadpou, J. Xu, and M. Razeghi, Appl. Phys. Lett. 68, 2100 ~1996!. 4 F. Binet, J. Y. Duboz, E. Rosencher, F. Scholz, and V. Ha¨rle, Appl. Phys. Lett. 69, 1202 ~1996!. 5 P. Kung, X. Zhang, D. Walker, A. Saxler, J. Piotrowski, A. Rogalski, and M. Razeghi, Appl. Phys. Lett. 67, 3792 ~1995!. 6 C. H. Qiu, W. Melton, M. W. Leksono, J. I. Pankove, B. P. Keller, and S. P. DenBaars, Appl. Phys. Lett. 69, 1282 ~1996!. 7 S. M. Sze, Physics of Semiconductor Devices, 2nd ed. ~Wiley, New York, 1981!. 8 J. Neugebauer and C. Van de Walle, Appl. Phys. Lett. 69, 503 ~1996!. 9 õz, I. E. Calleja, F. J. Sańchez, D. Basak, M. A. Sańchez-Garcıá, E. Mun Izpura, F. Calle, J. M. G. Tijero, B. Beaumont, P. Lorenzini, and P. Gibart, Phys. Rev. B 55, 4689 ~1997!. 10 S. D. Lester, F. A. Ponce, M. G. Craford, and D. A. Steigerwald, Appl. Phys. Lett. 66, 1249 ~1995!. 11 P. Vennegues, B. Beaumont, and P. Gibart, European MRS Conference, June 1996 ~unpublished!. 12 M. A. Sańchez-Garcıá, E. Calleja, E. Monroy, F. J. Sańchez, F. Calle, E. õz, and R. Beresford, J. Cryst. Growth ~to be published!. Mun 13 J. P. Vilcot, D. Decoster, L. Raczy, and M. Constant, Electron. Lett. 20, 274 ~1984!. 14 ã, F. J. Sańchez, and E. Mun õz, Semicond. Sci. I. Izpura, J. F. Valtuen Technol. 12, 678 ~1997!.

Appl. Phys. Lett., Vol. 71, No. 7, 18 August 1997


Munõz et al.