The Roles of Threading Dislocations on Electrical Properties of AlGaN/GaN Heterostructure Grown by MBE

(1)

The Roles of Threading Dislocations on Electrical Properties

of AlGaN/GaN Heterostructure Grown by MBE

Yuen-Yee Wong,aEdward Yi Chang,a,b,zTsung-Hsi Yang,b,cJet-Rung Chang,b Jui-Tai Ku,dMantu K. Hudait,eWu-Ching Chou,dMicheal Chen,fand Kung-Liang Lina

a

Department of Materials Science and Engineering, bDepartment of Electronics Engineering, cMicroelectronics and Information System Research Center, and dDepartment of Electrophysics, National Chiao Tung

University, Hsinchu 30010, Taiwan e

Department of Electrical and Computer Engineering, Virginia Tech, Blacksburg, Virginia 24061, USA f

ULVAC Taiwan Incorporated, Hsinchu 30078, Taiwan

The role played by different types of threading dislocations共TDs兲 on the electrical properties of AlGaN/GaN heterostructure grown by plasma-assisted molecular beam epitaxy共MBE兲 was investigated. Samples with different defect structures and densities were prepared and measurements were taken from the same sample to study the correlative behavior of various TDs. From the Hall measurement, the electron mobility in two-dimensional electron gas channel was mainly controlled by the edge dislocation, which has a dominant amount in the material. The edge TDs acted as Coulomb scattering centers inside the channel and reduces the carrier mobility and increased its resistance. Screw TDs played a much significant role than edge TDs in determining the reverse-bias leakage current of Schottky barrier diodes. Leakage current is affected slightly by the reduction of free carrier density in the channel for samples with a higher edge TD density, but screw TD, which acted as the current leakage path, was more deleterious to the reverse-bias leakage current of AlGaN/GaN structure.

Manuscript submitted January 19, 2010; revised manuscript received March 9, 2010. Published May 17, 2010.

GaN high electron mobility transistor共HEMT兲 devices are ex-cellent candidates for high power and high frequency applications.1,2 GaN HEMTs are usually achieved by using either AlGaN or AlInN as the barrier layer on top of a GaN buffer layer. Due to the strong polarization effect and large amount of surface states, high electron density can be induced at the AlGaN/GaN or AlInN/GaN interfaces, as two-dimensional electron gas 共2DEG兲 without any intentional doping in the materials.3,4 However, GaN and its heterostructure materials are usually epitaxial grown on a foreign substrate, such as sapphire, silicon, or silicon carbide because a large-size commercial grade native substrate is still not available at a low cost. A high density of threading dislocation共TD兲 is therefore generated in the materials as a result of large lattice and thermal expansion coeffi-cient mismatches between the GaN films and these substrates. TDs are known to induce electron scattering in the GaN material, thus reducing the carrier mobility,5-7and increasing the current leakage path in GaN-based devices.8,9High electron mobility in the 2DEG and low reverse-bias leakage current at the gate are the two most important parameters for GaN-based HEMTs to achieve a high power performance at high frequency operations. In a lateral current flow electron device, the former is controlled by factors affecting electron moving transversely in the channel, while the later is deter-mined by factors influencing current flow vertically across the bar-rier material. Three types of TDs are normally associated with the hexagonal material of GaN, namely, screw type, edge type, and mixed type. It is believed that each of the TD types may play dif-ferent roles in the GaN HEMT. However, most of the previous re-sults on the dislocation in GaN materials were obtained on the GaN film,5,10,11 but did not consider the HEMT devices, while others were measuring the performance of GaN HEMT as a function of the total dislocation density,6-8 but did not take into account the role played by each type of dislocation. In fact, the structural and elec-trical properties of dislocation are not only depended sensitively on the dislocation type but also on the growth method, growth condi-tions as well as the presence of dopants and impurities. For example, theoretical calculation have shown that both the screw and edge TDs are electrically inactive with a bandgap free from deep levels in the absence of impurities.12 However, the characterizations on real samples revealed that most of the TDs in GaN are in fact electrically active. The dislocation core can be filled up with excess Ga

atoms11,13or dopants and impurities 共including Mg, Si, and O兲,13 and therefore, changing the core structure and creating extra energy states in the material bandgap. In this paper, we have investigated the effect of screw and edge TDs on the electrical properties, namely, the electron mobility and reverse-bias leakage current of AlGaN/GaN heterostructure grown by plasma-assisted molecular beam epitaxy 共PAMBE兲. Both carrier mobility and reverse-bias leakage current were obtained from the same sample and therefore their correlative behavior can be studied.

Experimental

The AlGaN/GaN heterostructures were grown by PAMBE 关UL-VAC Molecular Beam Epitaxy共MBE兲 System兴 on sapphire 共0001兲 substrates. The epi-ready substrates were first thermally cleaned in the growth chamber at 820°C until the reflection high energy elec-tron deflection共RHEED兲 pattern became streaky. Nitridation of the substrate was then performed at 600°C to form a starting layer for the deposition of AlN buffer and followed by 1.5 ␮m GaN film. In this study, a series of AlGaN/GaN heterostructures with different amounts of screw and edge TDs were achieved by modifying the growth condition of AlN and the initial growth condition of GaN layers. In all cases, the final surfaces of GaN layers were smooth as indicated by the streaky RHEED patterns. Finally, AlGaN layers with constant Al composition of 25% with two different thicknesses 共20 and 25 nm兲 were grown on the GaN layers. Atomic force mi-croscope共AFM, Digital Instrument, D3100兲 was used to investigate the as-grown sample surfaces. Defect structure of the AlGaN/GaN heterostructures was determined by high resolution X-ray diffrac-tometry共HRXRD, Bede D1 system兲 scanning on the rocking curves of different diffraction planes. The aluminum composition and thickness of each AlGaN barrier layers were also determined by HRXRD, using ␻-2␪ and X-ray reflectivity scan modes, respec-tively. After the structural characterization, each of the samples was divided into two parts and prepared for Hall measurement and Schottky barrier diodes共SBDs兲 fabrication. From the Hall measure-ment using the van der Pauw contact configuration, electron mobil-ity, sheet resistance, and carrier density of 2DEG were obtained. SBDs fabricated on the AlGaN/GaN samples were used to measure the reverse-bias current. Before the fabrication, any metal droplet on the wafer surfaces was removed using diluted hydrochloric acid. The circular SBDs consist of a Schottky contact with diameter of 130 ␮m at the center of diode, and an ohmic contact formed as the outer ring. Using standard photolithography and lift-off methods, the z

E-mail: edc@mail.nctu.edu.tw

Journal of The Electrochemical Society, 157共7兲 H746-H749 共2010兲 0013-4651/2010/157共7兲/H746/4/$28.00 © The Electrochemical Society

H746

) unless CC License in place (see abstract).

ecsdl.org/site/terms_use

address. Redistribution subject to ECS terms of use (see

140.113.38.11

(2)

ohmic contacts of SBD were first fabricated by depositing Ti共200 Å兲/Al 共1200 Å兲/Ni 共250 Å兲/Au 共1000 Å兲 and followed by rapid thermal annealing at 800°C for 1 min. Then, the Schottky contact was formed by depositing Ni共200 Å兲/Au 共3000 Å兲.

Results and Discussion

In a previous study,14we have shown that the defect structure of GaN can be controlled by the growth condition of AlN buffer layer. Generally, if the starting surface of the AlN layer was relatively rough, the edge TDs in GaN were grown in the inclined direction and their self-interaction was enhanced. Therefore, the amount of edge TD density could be reduced through the recombination and annihilation processes. The formation of screw TD in GaN was sup-pressed on a smooth buffer surface. In this work, samples A, B, C, and D were prepared by using different AlN growth temperatures with constant thickness of 15 nm. By changing the growth tempera-ture of AlN buffer from 550 to 800°C, GaN layers with different dislocation densities, as indicated by the full width at half-maximum 共fwhm兲 values of the rocking curves, were prepared as shown in TableI. To further reduce the edge TD density, a second buffer layer of GaN grown under Ga-lean condition was also deposited on the AlN for samples M and N. The GaN materials grown under this condition would have surface with inclined facets which could fur-ther enhance the bending of edge TD propagation direction and reduce its density.15,16On top of all these buffers, initial GaN layer grown under Ga-rich growth condition was first applied to recover the rough surface of these buffer layers. Once the GaN surface be-came smooth 共within tens of nanometers兲, as indicated by the streaky RHEED patterns, the effective Ga/N ratio was set to slightly larger than unity for the rest of the growth to ensure a smooth surface of GaN layer was maintained. AFM images show the typical surface morphologies of GaN共Fig.1a兲 and AlGaN 共Fig.1b兲 grown

in this study. The surface of AlGaN decorated by small “mounds” is believed to be the result of kinetic roughening in the MBE growth of the AlGaN material.17The root-mean-square roughnesses were 0.44 and 0.65 nm, respectively, on a 4 ␮m2_{area. The RHEED patterns} of these surfaces are also shown in the insert of each diagram.

The dislocation densities of the GaN samples were determined using HRXRD. X-ray rocking curves共␻-scan兲 of both symmetric 共0002兲 and asymmetric 共10–12兲 planes were scanned for these samples. Due to the specific defect structure of screw TD 共with Burgers vector b =具0001典兲, edge TD 共b = 1/3具11–20典兲, and mixed TD 共b = 1/3具11–23典兲, the 共0002兲 rocking curve is broadened by screw- and mixed-type TDs, while the 共10–12兲 rocking curve is broadened by all TDs.18,19The contribution of both screw and edge TDs to the rocking width can be distinguished using the following equation

␤共10–12兲2 =␤s2+␤e2 关1兴 where␤_{共10–12兲}is the fwhm of共10–12兲 plane and ␤sand␤eare the contributions of screw and edge TDs, respectively, to the fwhm of 共10–12兲 plane.18

To simplify the calculation of the dislocation den-sities, the mixed TDs have been divided into their screw- and

edge-type components. So the “screw TD” in this study is referred to the contribution from pure screw and the screw component of mixed TDs, while the “edge TD” is consisted of pure edge and the edge component of mixed TDs. Furthermore, it was also demonstrated by Gay et al.20that the dislocation density is related to the XRD result through the equation

D_dis⬃ ␤2/9b2 关2兴

where D_dis is the TD density in the material, ␤ is the fwhm of a given XRD peak, and b is the length of Burgers vector of their corresponding dislocation. From Eq.1 and 2, both the screw and edge TD densities in GaN can be calculated.14

The effects of dislocation in the GaN film were compared with the electron mobility obtained from the Hall measurements. Figure2

shows the relationship between the electron mobility and edge TD density in the AlGaN/GaN samples. The mobility is decreased monotonically with the increase in the edge TD density in the samples. For sample D, with an edge TD density of more than 1.7⫻ 1010 _cm2_{, the mobility cannot be measured due to the} over-ranged sheet resistance 共⬎106 _ohm_{/square兲 and has been set as} zero. The reduction in the electron mobility in the AlGaN/GaN samples can be explained by considering the edge-type TD as the scattering centers for electrons moving in the 2DEG. The dangling bonds along the edge dislocation line can act as an electron acceptor5,6and filled up with electrons from the channel. Therefore,

Table I. fwhm of GaN(0002) and (10–12) planes prepared on buffers with different growth conditions.

Sample

AlN buffer growth temperature

共°C兲

fwhm of rocking curve共arc sec兲 _AlGaN thickness 共nm兲共0002兲 plane 共10–12兲 plane A 740 71 2404 20 B 700 85 2276 20 C 550 264 1890 20 D 800 75 2584 20 M 800 100 1460 25 N 750 233 1221 25

Figure 1.共Color online兲 AFM images of the typical as-grown 共a兲 GaN and

共b兲 AlGaN surfaces. Inserts show their corresponding RHEED patterns.

H747

Journal of The Electrochemical Society, 157共7兲 H746-H749 共2010兲 H747

140.113.38.11

(3)

these negatively charged edge dislocations scattered and reduced the carrier mobility. This is supported by the reduction of sheet carrier density with the increase in the edge TD density as shown in Fig.3. All the AlGaN layers were grown under the same growth condition 共despite two different thicknesses, these thicknesses are well above the critical thickness 共⬃35 Å兲 required to induce carriers into 2DEG 3from the surface state兲 and the carrier concentrations in-duced in the 2DEG are supposed to be similar for all the samples. It is the occurrence of trap sites in the dislocations that has reduced the availability of electron as free carriers in the 2DEG. As a result, the measured sheet resistances also increase accordingly. For a better illustration of the effect of the edge TD, the sheet resistance and free carrier concentration for sample D are set as 106 _{⍀/䊐 and zero,} respectively.

In AlGaN/GaN structure, electrons are confined in the 2DEG formed at the close vicinity of the heterostructure interface. These carriers are moving parallel to the AlGaN/GaN interface and there-fore their mobility can be affected by a variety of factors, such as phonon scattering, interface roughness, alloy scattering from the Al-GaN barrier, and scattering from the charged defects共dislocations, dipoles, residual impurities兲.21 Because the growth condition of AlGaN/GaN is similar for all the samples in our experiment, these

factors are the invariance properties except the TD density. The electron mobility decreases proportionally with the increase in the edge TD density, at least in the density range in this study.

There is no clear relationship between the electron mobility and screw TD density共insert of Fig.2兲. This is understandable because

the edge TD densities in all the samples are at least 2 orders of magnitude larger than the screw TD densities. For GaN materials grown on sapphire by common methods such as MBE and metallor-ganic chemical vapor deposition, the dislocation is always domi-nated by edge TD. The effect of the screw dislocation on electron mobility could have been overshadowed by the presence of the large number of edge dislocations. However, it is still reasonable to sug-gest that the screw TD does not play a significant role in controlling the electron mobility in most of the common cases.

Another important parameter of HEMT device is the reverse-bias leakage current at the gate contact. This current can be originated from both the surface and bulk material of the AlGaN/GaN struc-ture. In this work, only factors controlling the bulk component are investigated and the surface leakage currents are assumed similar for all the samples because they were grown under the same condition. Bulk leakage mechanism play the dominant role in the reverse-bias current22and it is believed that the TD is the major factors control-ling this property, as suggested by the literatures.9,23The leakage currents of the AlGaN/GaN structures were determined from the fabricated SBDs. For useful comparison, only samples A, B, C, M, and N, with measurable sheet resistances and current densities, have been fabricated into SBDs. Leakage current in SBDs is flowed ver-tically across the AlGaN layer as in contrast to the Hall measure-ment with electrons moving transversely in the 2DEG. The quality of AlGaN layer is, therefore, crucial under this circumstance. In the present study, crystal qualities of the AlGaN layers are too thin to be determined using HRXRD. Because they were all grown continu-ously on GaN without applying a growth stop, the dislocations in AlGaN are presumed to be similar to or imitate most of those from the GaN layers. The strain induced from the lattice mismatches be-tween these two materials was believed only to bend TDs slightly in the growth direction, but did not obstruct dislocation from propagat-ing to the sample surface. The leakage currents measured at reverse-bias for the AlGaN/GaN samples are shown in Fig.4. For samples A, B, and C, all with AlGaN thickness of 20 nm, the amount of leakage current increases with increasing screw TD density. Insert of Fig.4shows the values of leakage current measured at the reverse-bias of 5 V for these samples. For MBE grown GaN materials, the

Figure 2. Hall electron mobility of AlGaN/GaN samples as a function of

edge TD density. Insert shows the effect of screw TD density on electron mobility. Open symbols in the diagrams represent the values for sample D.

Figure 3. 共Color online兲 Sheet resistance and carrier concentration of the

AlGaN/GaN samples as a function of edge TD density. Open symbols rep-resent the values for sample D.

Figure 4. The reverse bias currents of AlGaN/GaN SBD samples. Insert

shows the reverse bias current for samples A, B, and C at⫺5 V as a function of screw TD density.

H748 Journal of The Electrochemical Society, 157共7兲 H746-H749 共2010兲

H748

140.113.38.11

(4)

screw dislocations acted as a leakage path for reverse-bias current as suggested by Hsu et al.9Their study also revealed that the screw dislocations, which are likely to be associated with excess Ga under Ga-rich growth condition, have a relaxed core and become electri-cally active.11 For samples M and N, with AlGaN thickness of 25 nm, their leakage also increased with screw TD density, but with a much slower rate. This has suggested that the longer leakage paths in the thicker AlGaN layers have increased the resistance for the leakage current.

The influence of the edge TD density on the leakage currents is not following a clear trend. At first thought, the availability of the free carrier density in the 2DEG共decrease with edge TD density兲 seems to limit the amount of leakage current as shown in samples A, B, and C 共with thinner AlGaN layer兲. But it is not the case for samples M and N共with thicker AlGaN layer兲. Although the edge TD densities are much higher than the screw TD densities, the role of the latter is dominant in this case. If only samples A, B, and C are compared, the slope of the insert of Fig.4is much steeper than the one in Fig.5. This has suggested that the screw dislocation plays a more important role than the edge dislocation in affecting the reverse-bias leakage current of the AlGaN/GaN structure.

Conclusion

In this study, the roles of TD on the electrical properties of AlGaN/GaN structure have been investigated. These properties, namely, electron mobility in the 2DEG and reverse-bias leakage current at the gate contact, were determined from the Hall measure-ment and SBDs, respectively. GaN materials with different defect structures and densities, which have been estimated using HRXRD, were first achieved by controlling the growth conditions of both AlN and GaN buffer layers. As the dominant type of dislocation in GaN, the edge TD affects the electron mobility in the 2DEG. They tended to trap carriers in the channel and in turn, acted as the scattering center to reduce the electron mobility. As a result, the availability of the free carrier density was reduced, while the channel resistance was increased with the increase in the edge TD density. The screw TD seemed to not affect the electron mobility significantly, but

served as the major player in controlling the reverse-bias leakage current. Each of the screw TD may provide a conducting path in the AlGaN layer for the leakage current and therefore a thicker barrier layer could be used to suppress the leakage current. This is because the current resistance was increased due to the longer leakage paths in the thicker AlGaN layer. The availability of free carrier density in the 2DEG, which is a function of edge TD density, also seems to affect the leakage current for sample with thinner AlGaN layer, but the effect is relatively insignificant as compared to the effect of the screw TD density.

Acknowledgment

This work was financially supported by the Ministry of Eco-nomic Affairs and the National Science Council of Taiwan under research grant no. 95-EC-17-A-05-S1-020, no. NSC95-2752-E-009-001-PAE, and no. NSC96-2221-E-009-236. The authors thank UL-VAC Taiwan Inc., Micheal Chen, and Stanley Wu for the MBE maintenance support.

National Chiao Tung University assisted in meeting the publication costs of this article.

References

1. N. Onojima, N. Hirose, T. Mimura, and T. Matsui, Jpn. J. Appl. Phys., 48, 094502 共2009兲.

2. A. Chini, R. Coffie, G. Meneghesso, E. Zanoni, D. Buttari, S. Heikman, S. Keller, and U. K. Mishra, Electron. Lett., 39, 625共2003兲.

3. J. P. Ibbetson, P. T. Fini, K. D. Ness, S. P. DenBaars, J. S. Speck, and U. K. Mishra, Appl. Phys. Lett., 77, 250共2000兲.

4. M. Miyoshi, Y. Kuraoka, M. Tanaka, and T. Egawa, Appl. Phys. Express, 1, 081102共2008兲.

5. J. H. You, J.-Q. Lu, and H. T. Johnson, J. Appl. Phys., 99, 033706共2006兲. 6. N. G. Weimann, L. F. Eastman, D. Doppalapudi, H. M. Ng, and T. D. Moustakas,

J. Appl. Phys., 83, 3656共1998兲.

7. C. Skierbiszewski, K. Dybko, W. Knap, M. Siekacz, W. Krupczynski, G. Nowak, M. Bockowski, J. Lusakowski, Z. R. Wasilewski, D. Maude, et al., Appl. Phys. Lett., 86, 102106共2005兲.

8. K. H. Lee, S. J. Chang, P. C. Chang, Y. C. Wang, and C. H. Kuo, J. Electrochem. Soc., 155, H716共2008兲.

9. J. W. P. Hsu, M. J. Manfra, D. V. Lang, S. Richter, S. N. G. Chu, A. M. Sergent, R. N. Kleiman, L. N. Pfeiffer, and R. J. Molnar, Appl. Phys. Lett., 78, 1685共2001兲. 10. P. Ebert, L. Ivanova, S. Borisova, H. Eisele, A. Laubsch, and M. Dahne, Appl.

Phys. Lett., 94, 062104共2009兲.

11. J. W. P. Hsu, M. J. Manfra, S. N. G. Chu, C. H. Chen, L. N. Pfeiffer, and R. J. Molnar, Appl. Phys. Lett., 78, 3980共2001兲.

12. J. Elsner, R. Jones, P. K. Sitch, V. D. Porezag, M. Elstner, T. Frauenheim, M. I. Heggie, S. Oberg, and P. R. Briddon, Phys. Rev. Lett., 79, 3672共1997兲. 13. D. Cherns and M. Hawkridge, J. Mater. Sci., 41, 2685共2006兲.

14. Y.-Y. Wong, E. Y. Chang, T.-H. Yang, J.-R. Chang, Y.-C. Chen, J.-T. Ku, C.-T. Lee, and C.-W. Chang, J. Cryst. Growth, 311, 1487共2009兲.

15. C. D. Lee, A. Sagar, R. M. Feenstra, C. K. Inoki, T. S. Kuan, W. L. Sarney, and L. Salamanca-Riba, Appl. Phys. Lett., 79, 3428共2001兲.

16. P. Waltereit, C. Poblenz, S. Rajan, F. Wu, U. K. Mishra, and J. S. Speck, Jpn. J. Appl. Phys., Part 2, 43, L1520共2004兲.

17. S. Vézian, F. Natali, F. Semond, and J. Massies, Phys. Rev. B, 69, 125329共2004兲. 18. J. C. Zhang, D. G. Zhao, J. F. Wang, Y. T. Wang, J. Chen, J. P. Liu, and H. Yang,

J. Cryst. Growth, 268, 24共2004兲.

19. B. Heying, X. H. Wu, S. Keller, Y. Li, D. Kapolnek, B. P. Keller, S. P. DenBaars, and J. S. Speck, Appl. Phys. Lett., 68, 643共1996兲.

20. P. Gay, P. B. Hirsch, and A. Kelly, Acta Metall., 1, 315共1953兲.

21. D. Jena, in Polarization Effects in Semiconductors, from Ab Initio Theory to Device Applications, C. Wood and D. Jena, Editors, p. 161, Springer, New York共2007兲. 22. W. S. Tan, P. A. Houston, M. J. Uren, R. S. Balmer, J. L. Glasper, and T. Martin,

in 2nd EMRS DTC Technical Conference, United Kingdom Ministry of Defense, p. A27共2005兲.

23. E. J. Miller, E. T. Yu, P. Waltereit, and J. S. Speck, Appl. Phys. Lett., 84, 535 共2004兲.

Figure 5. The reverse bias currents of AlGaN/GaN SBD samples as a

func-tion of sheet carrier density in the 2DEG.

H749

Journal of The Electrochemical Society, 157共7兲 H746-H749 共2010兲 H749

140.113.38.11