Rectifying characteristics of WSi0.8-GaN Schottky barrier diodes with a GaN cap layer grown at low temperature

(1)

Rectifying characteristics of WSi

_0.8

– GaN Schottky barrier diodes with a GaN cap layer grown at low temperature

J. K. Sheu^a兲

Institute of Electro-optical Science and Engineering, National Cheng Kung University, Tainan 70101, Taiwan, Republic of China

M. L. Lee and W. C. Lai

Institute of Microelectronics and Department of Electrical Engineering, National Cheng Kung University, Tainan 70101, Taiwan, Republic of China

H. C. Tseng and G. C. Chi

Department of Physics, National Central University, Chung-Li 320, Taiwan, Republic of China 共Received 20 January 2005; accepted 29 June 2005; published online 12 August 2005兲

Undoped GaN/low-temperature 共LT兲 GaN/WSi0.8 and undoped GaN / WSi_0.8 Schottky barrier contacts were prepared. Introducing the LT GaN on top of the conventional structures markedly reduced the leakage current and increased the barrier height. The measured barrier heights of the LT GaN-caped samples and the conventional samples were around 1 and 0.55 eV, respectively. The thermal stability of the Schottky barrier contacts was also studied and the barrier height was shown to be very stable even when the annealing temperature was increased to 950 ° C for 1 h. © 2005 American Institute of Physics.关DOI: 10.1063/1.2006220兴

Gallium nitride共GaN兲 and its related alloys have great potential for use in field-effect transistors 共FETs兲 at high power, high temperature, and high frequency because the band gap is wide. However, the large gate leakage current, particularly at high temperature or high bias, limits device performance. Restated, a thermally stable Schottky contact is required. The leakage current in a Schottky barrier diode 共SBD兲 depends strongly on the barrier height. A high Schottky barrier height 共SBH兲 is required at the metal/

semiconductor interface to reduce the leakage current in the SBDs. Various metals^1–4 and transparent conducting oxide films^5–8 have been deposited on GaN to produce high- performance SBDs. W-based Schottky metals have been used on GaN to yield highly thermally stable Schottky contacts.⁷ Schottky contacts formed by W-based metals on GaN had a lower barrier height and a higher reverse leakage current than those formed by other metals,⁸limiting the im- portance of the W-based GaN Schottky contacts in high- performance GaN FETs. The barrier height and/or reverse leakage current of SBDs also depend strongly on the properties of the surface layer. The gate leakage current in GaAs FETs has been shown to be reducible using a low- temperature-grown共LTG兲 GaAs layer.^9,10A similar concept has been applied to GaN-based MSM PDs.¹¹In this study, a LTG GaN layer was deposited on top of an undoped i-GaN layer to achieve a surface state that differs from those of conventional GaN films. Nitride-based SBDs were fabricated using WSi_0.8 as the Schottky metal.

The samples used in this study were all grown on c-face 共0001兲 sapphire 共Al2O₃兲 substrates by organometallic vapor- phase epitaxy 共OMVPE兲. After the sapphire substrate was annealed at 1100 ° C in H₂ ambient to remove surface con- tamination, a 30-nm-thick GaN nucleation layer was depos-

ited onto the sapphire substrate at 550 ° C. The temperature was then raised to 1060 ° C to grow a 3-␮m-thick Si-doped n⁺-GaN 共n=3⫻10¹⁸cm⁻³兲 layer. On top of this n⁺-GaN layer was grown a 2-␮m-thick undoped GaN 共n=3

⫻10¹⁶cm⁻³兲, followed by a 30-nm-thick 共LTG兲 GaN as the cap layer grown at low temperature共samples A兲. The growth temperature of the LTG GaN cap layer was 550 ° C. Notably, this LTG GaN cap layer shows a semi-insulating property with a sheet resistivity larger than 10⁹⍀/䊐. Samples B without the LTG GaN cap layer were also prepared for com- parison. In this study, WSi_0.8– GaN SBDs were fabricated by dry etching to expose the n⁺-GaN, and then Ti/ Pt/ Au 共20/20/300 nm兲 was deposited on the exposed n⁺-GaN to serve as an Ohmic contact. Finally, WSi_0.8 共100 nm兲 was deposited on LTG GaN as a Schottky metal. The WSi_0.8films were deposited on GaN samples using a W / Si cosputtering process with an Ar discharge. The diameter of the fabricated circular devices was 500␮m. The room-temperature 共RT兲 current-voltage 共I-V兲 characteristics of the fabricated SBDs were then measured using a HP4156 semiconductor param- eter analyzer.

Figure 1 shows the typical I-V characteristics of the WSi_0.8/ GaN SBDs obtained at RT. Under a reverse bias, the current was around 1⫻10⁻¹¹A for samples A. Samples B exhibited a large reverse current of around 1⫻10⁻⁶– 1

⫻10⁻⁵ A. The low reverse current in samples A was attributable to the fact that the highly resistive LTG GaN cap layer between WSi_0.8 and i-GaN resulted in a thicker and/or a higher potential barrier than the conventional SBDs共samples B兲. Under forward bias, the typical I-V curves of the samples are similar to the reverse I-V characteristics. The forward current of samples A is far less than that of samples B. The lower forward current of samples A was attributable to the high-resistivity LTG GaN cap layer and the fact that injected carriers could be compensated for by the poor-crystal-

a兲Electronic mail: [email protected]

JOURNAL OF APPLIED PHYSICS 98, 036106共2005兲

Downloaded 08 Dec 2008 to 140.116.81.121. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp

(2)

quality-related trap levels within the band gap of LTG GaN,¹² as in the case of annealed LTG GaAs with high resistivity.¹³Therefore, the annealed LT GaAs is suitable for use as a buffer layer in metal-semiconductor FETs 共MESFETs兲 or as a gate insulator layer in metal-insulator- semiconductor FETs共MISFETs兲. In contrast, the preliminary findings herein reveal that no significant difference in resistivity exists between the as-grown LTG GaN and the annealed LTG GaN films.

In this study, the effective barrier heights of samples A and B can be tentatively derived from the forward I-V char- acteristics and Nord method.⁶Table I shows the average effective barrier heights of samples A and B. Notably, the values in Table I were obtained from at least 15 samples A and 15 samples B. The effective Schottky barrier height of samples B is around 0.55 eV, which value is less than the ideal value of about 0.7 eV, if the work function of WSi_0.8 and n-type GaN are about 4.8 and 4.1 eV, respectively.^7,8 The barrier height of the contact to n-type GaN without the cap layer共samples B兲 may be lower because of surface defects in the GaN films. That the ideal barrier height can be determined using simple Schottky-Mott theory is well known. Without any charged states on the semiconductor surface, the barrier height can be given by

␾bn=␾m−␹s, 共1兲

where ␾m is the work function of the metal and ␹s is the electron affinity of the semiconductor. However, in most practical metal-semiconductor contacts, the ideal barrier height is never reached because surface states are present.

The sum of charges throughout the junction of a

metal/ n-semiconductor contact should be zero. Restated, the relationship between the negative charge Qmon the surface of the metal, the positive charge Qd associated with uncom- pensated donors, and the surface states in the metal/

semiconductor interface 共i.e., surface charge兲 Qsshould be

Qm+ Qd+ Qs= 0. 共2兲

If the surface states contain a net positive charge, Qd should be smaller than that in the ideal case. Under such conditions, the width of the depletion region in the semiconductor is reduced, reducing the bending of the bands. If the density of surface states is not very high, then the corresponding barrier height can be expressed as

␾bn= q共Vbi+ V_n兲, 共3兲 where V_n is the energy difference between the Fermi level and the bottom of the conduction band and V_biis the amount of band bending in the semiconductor. Hierro et al. indicated that charged states are present near the dislocations, and are responsible for the locally high reverse leakage current in GaN SBDs and the consequent reduction in the Schottky barrier height.¹⁴Therefore, the low barrier height of samples B is attributable to the positively charged surface states associated with the GaN surface pits, corresponding to the ter- mination of thread dislocations. However, the effective barrier height of samples A is 1.02 eV, as shown in Table I. The higher barrier height is attributable to the fact that the surface states 共pits兲 are passivated 共covered兲 by the LTG GaN cap layer, reducing the leakage current and therefore increasing the barrier height.¹⁵ However, this value greatly exceeds the ideal value of around 0.7 eV. Therefore, a barrier height that exceeds the theoretical value has other causes. The Schottky barrier heights of various metals onto n-type wurtzite GaN have been systematically studied and the experimental data revealed that the barrier height varies with the work function of the metal.¹⁶The preliminary results herein reveal that the Schottky barrier heights of the Ni/ Au共100/100 nm兲 bilayer metal deposited on n-type GaN with and without LTG GaN cap layer, which were around 1.1 and 0.9 eV, respectively, followed similar trends to those of the aforementioned WSi_0.8/ GaN Schottky barrier diodes. In the latter case, the measured barrier height is close to the reported values.¹⁷ A reasonable growth temperature of high-quality GaN is typi- cally around 1000 ° C. In this case, the LTG GaN cap layers were grown at 550 ° C. Therefore, its crystalline quality was very inferior to that of GaN grown at high temperature. As stated above, the LTG GaN layers are highly resistive and behave like insulators. Therefore, defects in the surface of the GaN grown at high temperature with the LTG GaN cap layer including surface pits can be passivated. However, the surface of the LTG GaN layers should have a very high density of surface states. Accordingly, Fermi-level pinning, caused by the high density of the surface states, may cause the measured barrier height to be around the difference be- tween the energy-band gap Egand the neutral level␾0which is␾bn⬇Eg/ e −␾0. Restated, in this case, the Schottky barrier height seems to be fixed at around 1 eV, corresponding to a

␾0value of about 2.4 eV above the valence band, indepen- dent of the difference between the work functions of the

FIG. 1. Typical I-V characteristics of the WSi_0.8/ GaN SBDs taken at room temperature.

TABLE I. The typical Schottky barrier heights, which are determined by I-V characteristics, of samples A and B as a function of annealing tempera- ture. Notably, the data in the table are average values from at lest 15 diodes;

the standard deviation of the barrier height is 0.01.

Sample A

␾e共eV兲

Sample B

␾e共eV兲

As-deposited 1.02 0.55

550 ° C, 1 h 1.03 0.52

750 ° C, 1 h 0.99 0.53

950 ° C, 1 h 0.98 0.65

036106-2 Sheu et al. J. Appl. Phys. 98, 036106共2005兲

(3)

metals and the n-GaN. For clarity, more metals were depos- ited on n-type GaN with a LTG GaN cap layer to measure the effective Schottky barrier heights. These experiments are under way and the results will be published elsewhere.

As stated above, W-based Schottky contacts on n-type GaN are very thermally stable. In this study, WSi_0.8Schottky contacts were deposited on GaN films 共samples A and B兲 followed by high-temperature thermal annealing, to study the thermal stability. The samples with the WSi_0.8Schottky contacts on GaN were first annealed at temperatures from 750 to 950 ° C for 1 h in N₂ ambient, and Ti/ Pt/ Au 共20/20/300 nm兲 was deposited on exposed n⁺-GaN as an Ohmic contact. Table I presents the typical Schottky barrier heights, which were determined by the I-V characteristics, of samples A and B as a function of annealing temperature. The barrier heights of samples A decline slightly with an increase of annealing temperatures. As speculated above, Fermi-level pinning may occur at the LT GaN interface, even when these samples were annealed at high temperature. Therefore, the barrier height of samples A was still around 1 eV. In addition to the 950 ° C annealed samples, samples B also showed a stable Schottky barrier height of around 0.5 eV. As is gener- ally known, interfacial reactions between the metal and the semiconductor can substantially affect the electrical characteristics of a metal-semiconductor contact. In this study, the barrier height of the WSi_0.8Schottky contact decreases only slightly as the annealing temperature increases, as presented in Table I. This fact reveals that the WSi_0.8/ GaN 共or LT GaN兲 Schottky contacts are highly thermally stable. Upon annealing at 950 ° C, the barrier height of samples B markedly increased, rather than decreasing further. X-ray diffrac- tion revealed that the as-deposited WSi_0.8 films exhibited amorphouslike properties. However, when the WSi_0.8 films were annealed up to 950 ° C, some characteristic peaks of WSi₂compound were detected. Restated, when the annealed temperatures were increased to 950 ° C, the WSi_0.8 film gradually became crystalline and tended locally to form WSi₂ grains. This annealing-induced recrystallization may have increased the work function of WSixfilms by the for- mation of WSi₂ grains. Therefore, the 950 ° C annealed samples B showed an increase in barrier height rather than a further decrease. Upon annealing at 950 ° C, samples A exhibited only a slight decrease in the measured barrier height, because of Fermi-level pinning caused by the LT GaN cap layer.

In summary, WSi_0.8 contacts were formed on n-type GaN with a LTG GaN cap layer to characterize the Schottky

barrier diodes. The higher Schottky barrier height of the samples with the LTG GaN cap layer had two possible causes: the high-resistivity LTG GaN may passivate the surface defects 共pits兲 caused by dislocations or may cause Fermi-level pinning at the metal-semiconductor interface.

Furthermore, thermally stable WSi_0.8contacts on n-type GaN were demonstrated. This finding is consistent with other re- ports. The slight difference among the barrier heights of samples A with various annealing temperatures could again be caused by Fermi-level pinning. However, this tentative conclusion can only be verified by conducting further experiments.

ACKNOWLEDGMENTS

The authors would like to acknowledge the Center for Micro/Nano Technology Research, National Cheng Kung University, Tainan, Taiwan for equipment access and techni- cal support and the financial support from the National Sci- ence Council of Taiwan for their research grants of NSC93- 2112-M-008-034 and NSC93-2215-E-006-036.

1X. A. Cao, S. J. Pearton, G. Dang, A. P. Zhang, F. Ren, and J. M. Van Hove, Appl. Phys. Lett. 75, 4130共1999兲.

2Q. Z. Liu, L. S. Yu, S. S. Lau, J. M. Redwing, N. R. Perkins, and T. F.

Kuech, Appl. Phys. Lett. 70, 1275共1997兲.

3L. Wang, M. I. Nathan, T. H. Lim, M. A. Khan, and Q. Chen, Appl. Phys.

Lett. 68, 1267共1996兲.

4C. T. Lee, Y. J. Lin, and D. S. Liu, Appl. Phys. Lett. 79, 2573共2001兲.

5J. K. Kim, H. W. Jang, C. M. Jeon, and J.-L. Lee, Appl. Phys. Lett. 81, 4655共2002兲.

6J. K. Sheu, Y. K. Su, G. C. Chi, M. J. Jou, and C. M. Chang, Appl. Phys.

Lett. 72, 3317共1998兲.

7J. Kim, F. Ren, A. G. Baca, and S. J. Pearton, Appl. Phys. Lett. 82, 3263 共2003兲.

8X. A. Cao, S. J. Pearton, F. Ren, and J. R. Lothian, Appl. Phys. Lett. 73, 942共1998兲.

9S. Lodha, D. B. Janes, and N. P. Chen, Appl. Phys. Lett. 80, 4452共2002兲.

10L. W. Yin, Y. Hwang, J. H. Lee, R. M. Kolabas, R. J. Trew, and U. K.

Mishra, IEEE Electron Device Lett. 11, 561共1990兲.

11J. K. Sheu et al., J. Electron. Mater. 32, 400共2003兲.

12M. Lampert and P. Mark, Current Injection in Solids共Academica, New York, 1970兲.

13A. Osinsky, S. Gangopadhyay, R. Gaska, B. Williams, M. A. Khan, D.

Kuksenkov, and H. Temkin, Appl. Phys. Lett. 71, 2334共1997兲.

14A. Hierro, D. Kwon, S. A. Ringela, M. Hansen, J. S. Speck, U. K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 76, 3064共2000兲.

15J. K. Sheu, M. L. Lee, and W. C. Lai, Appl. Phys. Lett. 86, 052103 共2005兲.

16A. C. Schmitz, A. T. Pong, M. A. Khan, Q. Chen, J. W. Yang, and I.

Adesida, J. Electron. Mater. 27, 255共1998兲, and references therein.

17Q. Z. Liu, L. S. Yu, F. Deng, S. S. Lau, and J. M. Redwing, J. Appl. Phys.

84, 881共1998兲, and references therein.

036106-3 Sheu et al. J. Appl. Phys. 98, 036106共2005兲