You are on page 1of 4

JOURNAL OF APPLIED PHYSICS

VOLUME 87, NUMBER 4

15 FEBRUARY 2000

Heavy doping effects in Mg-doped GaN


Peter Kozodoy,a) Huili Xing, Steven P. DenBaars, and Umesh K. Mishra
Department of Electrical and Computer Engineering, University of California, Santa Barbara, California 93106

A. Saxler, R. Perrin, S. Elhamri,b) and W. C. Mitchel


Air Force Research Laboratory, Materials and Manufacturing Directorate, AFRL/MLPO, Wright-Patterson AFB, Ohio 45433-7707

Received 23 August 1999; accepted for publication 2 November 1999 The electrical properties of p-type Mg-doped GaN are investigated through variable-temperature Hall effect measurements. Samples with a range of Mg-doping concentrations were prepared by metalorganic chemical vapor phase deposition. A number of phenomena are observed as the dopant density is increased to the high values typically used in device applications: the effective acceptor energy depth decreases from 190 to 112 meV, impurity conduction at low temperature becomes more prominent, the compensation ratio increases, and the valence band mobility drops sharply. The measured doping efciency drops in samples with Mg concentration above 2 1020 cm 3 . 2000 American Institute of Physics. S0021-89790004304-8

INTRODUCTION

Mg-doped p-type GaN is of critical importance for a host of nitride-based devices including light emitting diodes, lasers, photodetectors, and bipolar transistors.1 Because of the deep nature of the Mg acceptor very high doping levels approximately 1020 cm 3 ) are frequently used in device structures. Heavy doping effects may be important at these high doping levels, leading to such phenomena as valence band-tail states and impurity band formation.2 Temperature-dependent Hall effect measurements have been performed on Mg-doped GaN by a number of groups, yielding a range of activation energies for the Mg dopant between 125 and 215 meV.38 This wide spread in experimental data may be at least partially explained by a variation in doping level, and consequently in the importance of heavy doping effects, in the various samples. In this work we present a careful examination of the electrical properties of Mg-doped GaN as a function of dopant concentration. A wide range of dopant densities are studied in order to investigate the evolution of high-doping effects both below and above the optimal dopant concentration. Temperaturedependent Hall effect measurements are employed to analyze the electrical properties of each sample.
EXPERIMENT

The GaN samples were grown by metalorganic chemical vapor deposition on c-plane sapphire substrates. Trimethylgallium TMGa, biscyclopentadienyl-magnesium (Cp2 Mg), and ammonia (NH3 ) were used as precursors. The layer structure used was that of a p n junction: a base layer of approximately 3 m of n-type GaN was rst deposited, followed by a top layer of Mg-doped GaN 0.5 or 1.0 m thick.
a

Electronic mail: kozodoy@engineering.ucsb.edu Permanent address: Department of Physics, University of Dayton, Dayton, OH 45469. 1832

The Mg-doping level was varied by changing the ow of Cp2 Mg during growth from 24 to 428 nmol/min. In all samples the TMGa molar ow was 19.3 mol/min. Secondary ion mass spectroscopy SIMS measurements were performed on the samples in order to measure the chemical Mg concentration, which scaled roughly with Cp2 Mg ow and varied between 2 1019 and 8 1020 cm 3 . The samples with the highest magnesium concentration are overdoped, i.e., the magnesium concentration is so high that the material qualities have begun to degrade. In the most heavily doped sample sample F the Mg concentration is estimated to be about 2% of the Ga concentration, and a severely degraded surface morphology consisting of densely packed hexagonal pyramids was observed. The samples were prepared for Hall effect measurements using lithographically dened van der Pauw structures. Variable-temperature Hall effect measurements were performed using magnetic elds between 1 and 2 T. During the measurements the applied voltage was kept below 3 V to minimize any leakage through the underlying n-type layer. The hole concentration p was obtained from the Hall constants R H using p r H / qR H with the Hall scattering factor r H assumed to be of value unity. In Fig. 1 the calculated hole concentration is plotted as a function of the inverse temperature for each of the samples. An activation energy dependence is clearly evident in the high-temperature regime for all of the samples. In all but the most lightly doped sample the measured hole concentration increases again at low temperature, indicating the onset of hopping or impurityband conduction. In the most heavily doped samples this impurity conduction is clearly an important component of the total room temperature conductivity. The acceptor activation energy ( E A ), acceptor concentration ( N A ), and compensating donor concentration ( N D ) were extracted by tting the high-temperature data to the formula
2000 American Institute of Physics

0021-8979/2000/87(4)/1832/4/$17.00

Downloaded 23 Aug 2006 to 129.74.159.226. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

J. Appl. Phys., Vol. 87, No. 4, 15 February 2000

Kozodoy et al.

1833

p pND NV EA exp , N A N D p g kT

where k is the Boltzmann constant, T is the temperature, g is acceptor degeneracy which is assumed to be equal to four, and N V is the effective valence band density of states

* kT 3/2/ h 3 , N V 2 2 m h

* where h is the Planck constant. The hole effective mass m h in GaN is not well known. Early reports suggested values * 0.8 m 0 , 9 while more recent results have yielded around m h larger values of 1.1m 0 10 and 2.2 m 0 . 11 The higher value of * 2.2 m 0 was used in these calculations because it yields mh an improved agreement between the extracted acceptor concentration and the SIMS measurements of Mg concentration. We note that the model used to t the data is quite simple; Eq. 1 assumes a single acceptor level and Eq. 2 assumes a parabolic valence band. The formation of a broad impurity band, or of valence band-tail states, is therefore beyond the scope of this model. If these effects play an important role in the heavily doped samples, then the tting procedure employed may provide results that are somewhat inaccurate.
RESULTS AND DISCUSSION

FIG. 2. Concentration data extracted from the Hall effect and SIMS measurements are presented in the top plot. In the lower plot, the solid triangles represent the measured activation energy and the open triangles represent that predicted by Eq. 3.

Figure 2 and Table I summarize the parameters that provide the best t between Eq. 1 and the temperaturedependent Hall effect measurements. The measured activation energy for the Mg acceptor decreases as the doping is

FIG. 1. Hole concentration measured as a function of temperature on Mgdoped GaN samples. For clarity of presentation the data have been divided between two separate plots; note that the scale differs on the two plots. The solid lines represent ts to Eq. 1.

increased, going from 190 meV in sample A to 112 meV in sample E. The material quality begins to degrade before the Mott transition can be reached and the activation energy rises again in the severely overdoped sample. The extracted values of N A and N D are also plotted in Fig. 2. While the activation energy can be extracted quite accurately from the least-squares t to Eq. 1, there is some uncertainty in the acceptor and compensating donor concentrations. This is due not only to the assumption of the hole mass, but also to the nature of the data t especially in the more heavily doped samples where there is little indication of saturation in the hole concentration measured at high temperature. Nonetheless, the extracted concentrations match closely with expected values. The acceptor concentration rises as the dopant density is increased, in good agreement with the Mg concentration as measured by SIMS. The highest acceptor concentration around 2 1020 cm 3 ) was obtained in sample D. When the Mg concentration is increased beyond this point the measured acceptor concentration is actually reduced, which may indicate that much of the Mg is not incorporating in the desired substitutional site. We also note that the calculated compensation level ( N D / N A ) is observed to rise dramatically as the doping level is increased. The measured decrease in acceptor activation energy at high doping levels explains the variation in published results for the depth of the Mg acceptor. This phenomenon is well known from conventional semiconductors2,12 and may have various causes including: the formation of a broad Mg acceptor band extending toward the valence band edge, the creation of valence band-tail states extending into the forbidden gap, screening of the acceptor potential by free carriers, and binding energy reduction through a Coulomb interaction

Downloaded 23 Aug 2006 to 129.74.159.226. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

1834

J. Appl. Phys., Vol. 87, No. 4, 15 February 2000

Kozodoy et al.

TABLE I. Summary of tting results from Hall effect measurements. The Mg concentration measured by SIMS is also listed. Cp2 Mg ow nmol/min E A meV N A (cm 3 ) N D (cm 3 ) Mg (cm 3 ) 24.4 42.3 76.5 135 228 428 190 174 152 118 112 165 1.8 1019 4.6 1019 1.4 1020 2.2 1020 8.6 1019 7.6 1018 1.1 1018 3.0 1018 1.2 1019 4.1 1019 2.7 1019 5.0 1018 1.6 1019 4 1019 8 1019 2 1020 3 1020 8 1020

Sample A B C D E F

between valence-band holes and ionized acceptors. tz et al.7 have suggested an important role for the last Go of these effects in Mg-doped GaN. In this case the acceptor activation energy may be written as
E A N A E A ,0 f

q2 N 1/3, 4 A

where E A ,0 is the inherent acceptor energy observed at very (1/3) ) is the average dislow doping concentrations, ( N A tance between ionized acceptors, q is the electronic charge, is the dielectric constant assumed to be 9.5 0 ), and f is a geometric factor of value (2/3)(4 /3) 1/3. 7 We note that this formulation neglects the repulsive potential of the ionized compensating donors, an assumption which is somewhat justied because the free holes are likely to concentrate around the ionized acceptors and avoid the donors.12 tz suggests that the acceptor energy in Eq. 1 be Go re-evaluated at each temperature depending on the ionized acceptor concentration. However, in this case we are justied in using a simpler approach in which a single temperatureindependent activation energy is used for each sample. This is because the ionized acceptor concentration is determined mainly by the number of compensating donors, not the number of free holes, across the great majority of the temperature range used for data tting. We therefore use Eq. 1 to t the Hall effect data and then compare the measured acceptor N D . Assumenergy to that predicted by Eq. 3 using N A ing an inherent activation energy of E A ,0 220 meV, a very

close agreement is obtained between the measured activation energy and that predicted by Eq. 3, as shown in Fig. 2. In fact, the t is surprisingly good considering the uncertainty in the values of N D .) The compensation is believed to be due to a native donor and/or a Mg-related state. Several authors have suggested that nitrogen vacancies play a role in the compensation of p-type GaN.1315 We have shown in other work that the compensation level in Mg-doped GaN may be adjusted by properly tailoring the growth conditions,14 indicating that this is a parameter which will be dependent, at least partially, on the choice of growth conditions and the details of reactor design. Figure 3 presents the hole mobility measured as a function of temperature. The highest mobility recorded is 62 cm2 /V s for the lightly doped sample at T 146 K. As the doping level is increased the peak mobility begins to drop rapidly due to the high concentration of ionized species resulting from the higher doping level and compensation ratio. In the most heavily doped sample the hole mobility is observed to increase again slightly, most likely a consequence of the reduced ionized dopant density in this sample. In most samples the measured mobility exhibits a precipitous drop at low temperature; this is due to the onset of hopping conduction, which is characterized by a very low mobility. The factors determining the exact temperature dependence of the valence-band mobility are not fully understood at this pointa more complete analysis of the mobility measurements is currently underway.
CONCLUSIONS

In conclusion, we have examined the electrical properties of Mg-doped GaN as a function of doping level. As the doping level is increased a number of heavy doping effects are observed, including an increase in both impurity conduction and degree of compensation. The increased compensation appears to drive other effects such as a pronounced reduction in acceptor activation energy and lower hole mobility values. The measured acceptor concentration increases with Mg concentration up to the optimum doping level of 2 1020 cm 3 . Beyond this point the doping efciency drops and at very high doping levels severe morphological changes are observed.
ACKNOWLEDGMENTS

This work was supported by the Air Force Ofce of Scientic Research through a contract monitored by Dr. Gerald Witt and by the Ofce of Naval Research through a contract monitored by Dr. John Zolper. The authors are grateful to J. Antoszewski, J. M. Dell, and L. Faraone at the University of Western Australia for preliminary measurements and useful discussions and to S. Davidson of Wright-Patterson AFB for technical assistance.
O. Ambacher, J. Phys. D 31, 2653 1998. E. F. Schubert, Doping in III V Semiconductors Cambridge University Press, Cambridge, 1993. 3 T. Tanaka, A. Watanabe, H. Amano, Y. Kobayashi, I. Akasaki, S. Yamazaki, and M. Koike, Appl. Phys. Lett. 65, 593 1994.
1 2

FIG. 3. Mobility measured as a function of temperature on Mg-doped GaN samples.

Downloaded 23 Aug 2006 to 129.74.159.226. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

J. Appl. Phys., Vol. 87, No. 4, 15 February 2000


4

Kozodoy et al.
10 11

1835

tz, N. M. Johnson, J. Walker, D. P. Bour, and R. A. Street, Appl. W. Go Phys. Lett. 68, 667 1996. 5 H. Nakayama, P. Hacke, M. R. H. Khan, T. Detchprohm, T. D. Kazumasa, K. Hiramatsu, and N. Sawaki, Jpn. J. Appl. Phys., Part 2 35, L282 1996. 6 C. J. Eiting, P. A. Grudowski, and R. D. Dupuis, Electron. Lett. 33, 1987 1997. 7 tz, R. S. Kern, C. H. Chen, H. Liu, D. A. Steigerwald, and R. M. W. Go Fletcher, Mater. Sci. Eng., B 59, 211 1999. 8 W. Kim, A. E. Botchkarev, A. Salvador, G. Popovidi, H. Tang, and H. Morkoc, J. Appl. Phys. 82, 219 1997. 9 J. J. Pankove, S. Bloom, and G. Harbecke, RCA Rev. 36, 163 1975.

M. Suzuki and T. Uenoyama, Jpn. J. Appl. Phys., Part 1 34, 3442 1995. J. S. Im, A. Moritz, F. Steuber, V. Harle, F. Scholz, and A. Hangleiter, Appl. Phys. Lett. 70, 631 1997. 12 P. P. Debye and E. M. Conwell, Phys. Rev. 93, 693 1954. 13 U. Kaufmann, M. Kunzer, M. Maier, H. Obloh, A. Ramakrishnan, B. Santic, and P. Schlotter, Appl. Phys. Lett. 72, 1326 1998. 14 P. Kozodoy, S. Keller, S. P. DenBaars, and U. K. Mishra, J. Cryst. Growth 195, 265 1998. 15 C. G. Van de Walle, C. Stamp, and J. Neugebauer, J. Cryst. Growth 189190, 505 1998.

Downloaded 23 Aug 2006 to 129.74.159.226. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

You might also like