Vacuum 144 (2017) 256e260

Contents lists available at ScienceDirect

Vacuum journal homepage: www.elsevier.com/locate/vacuum

Short communication

Tuning electrical properties of Au/n-InP junctions by inserting atomic layer deposited Al2O3 layer Hogyoung Kim a, b, *, Dong Ha Kim c, Sungyeon Ryu c, Byung Joon Choi c, ** a

Department of Visual Optics, Seoul National University of Science and Technology (Seoultech), Seoul 01811, South Korea Convergence Institute of Biomedical Engineering & Biomaterials, Seoul National University of Science and Technology (Seoultech), Seoul 01811, South Korea c Department of Materials Science and Engineering, Seoul National University of Science and Technology (Seoultech), Seoul 01811, South Korea b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 March 2017 Received in revised form 30 June 2017 Accepted 4 August 2017 Available online 5 August 2017

In order to modulate the electrical properties of Au/n-InP contacts, the insertion of an Al2O3 layer deposited by atomic layer deposition (ALD) was employed. All the samples with an Al2O3 layer showed the increased barrier height compared to the sample without Al2O3 layer. The barrier height was also found to be almost constant regardless of the Al2O3 thickness. Analysis on the reverse bias current revealed that the dominant current transport is Poole-Frenkel emission for all the samples with an Al2O3 layer, indicating that the defects are generated in the Al2O3 layer. Our results suggest that the metal induced gap states (MIGS) or interface dipole effects did not play major role in the Au/Al2O3/InP junction properties. Rather, the removal of native oxide and the termination of dangling bonds by ALD process determined the modulation of barrier heights. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Al2O3 layer Barrier height Interface dipole

InP and related semiconductors have received much attention as potential materials in electronic and optoelectronic devices such as field effect transistors and laser diodes due to its high electron mobility and saturation velocity [1]. In these devices, Schottky contacts with low leakage current and ohmic contacts with low contact resistance are pivotal to obtain high device performance. However, it is still a challenge to obtain good ohmic contact due to the Fermi level pinning (FLP) at 0.4e0.5 eV below the conduction band in n-InP [2]. Strong FLP effect also hinders to tune the Schottky barrier heights by metal work functions, which was explained by metal-induced gap states (MIGS) [3] or interfacial dipole effect [4]. As an approach to solve the FLP issue, Connelly et al. have reduced the barrier height of metal/Si contact by using a thin Si3N4 through the creation of a dielectric dipole [5]. Similar methods have been made by growing high quality high-k gate dielectrics on semiconductors using atomic layer deposition (ALD) [6e8]. The interfacial dielectric layer acting as a dangling bond terminator at the semiconductor surface reduces the surface charge trap density

* Corresponding author. Department of Visual Optics, Seoul National University of Science and Technology (Seoultech), Seoul 01811, South Korea. ** Corresponding author. E-mail addresses: [email protected] (H. Kim), [email protected] (B.J. Choi). http://dx.doi.org/10.1016/j.vacuum.2017.08.004 0042-207X/© 2017 Elsevier Ltd. All rights reserved.

which originally pinned the Fermi level [9]. Among dielectric materials, aluminum oxide (Al2O3) is a very stable and robust material, and as an alternate gate dielectric Al2O3 has many favorable properties, including a high band gap, thermodynamic stability on Si up to high temperatures [10]. Previously metal/Al2O3/InP structure was mainly employed as metal-oxidesemiconductor (MOS) capacitors [11,12]. As an effort to obtain the optimal thickness for low resistance ohmic contact, Zheng et al. varied the thickness of Al2O3 layer in metal/n-InP contacts and explained the reduction of barrier heights through the formation of interface dipole [13]. The direction of interface dipole was determined from the difference of areal density of oxygen atoms at the interface according to Kita's model [14]. As a barrier height modulator for metal/InP contacts, however, metal/Al2O3/InP structure has not been studied well. In this work, we investigated the thickness effect of Al2O3 layer grown by ALD on the electrical properties of Au/n-InP Schottky junctions and the detailed interface formation was discussed. A single side polished undoped (unintentionally n-type doped) InP (100) wafer (thickness: 350 mm, carrier concentration: 5
1015 cm3) was used as a starting material. The wafer was cut into small pieces (about 5
10 mm2) and some of these pieces were cleaned in acetone and methanol and the native oxide was removed in a HCl:H2O (1:1) solution and were loaded into an ALD

H. Kim et al. / Vacuum 144 (2017) 256e260

chamber. Before Al2O3 deposition, the deposition temperature ramped up to 250  C for 5 min in an N2 ambient and the samples were subjected to H2O prepulse treatment for 3 min. Then four samples with different Al2O3 thicknesses were prepared by varying the number of ALD cycles. Trimethylaluminum (TMA) and deionized water (H2O) were used as the precursors with a purging gas of nitrogen (N2). A reference sample without Al2O3 layer was also prepared to observe the different characteristics of metal/n-InP contacts. For Schottky contacts, 50 nm thick Au metal with a diameter of 500 mm was deposited by using e-beam evaporation through a shadow mask on the surface of both samples. For ohmic contacts, In metal was rubbed over the entire back surface of the samples. Using spectroscopic ellipsometer, the thicknesses of Al2O3 were measured to be 1.3, 2.7, 4.9 and 8.3 nm. Currentevoltage (IeV) measurements were carried out with a Keithley 238 current source at room temperature. X-ray photoelectron spectroscopy (XPS) measurements were carried out using a monochromatic Al Κa xray source to observe the formation mechanism near the InP surface. Fig. 1(a) and (b) are cross-sectional scanning transmission electron microscope (STEM) images of Au/Al2O3/InP junctions with Al2O3 thicknesses of 2.7 and 8.3 nm, respectively, showing the high quality contacts of metal and the uniform Al2O3 films. STEM measurements also show that the thicknesses of Al2O3 layers are very similar to the values from spectroscopic ellipsometer. Fig. 1(c) shows the depth profiles from energy dispersive X-ray spectroscopy (EDS) measurements for each element around the Al2O3 layer region for the sample with an 8.3 nm thick Al2O3 layer. It is seen

257

that Au atoms are not present in the InP layer. The diffusion of many oxygen atoms into Au metal are also observed (atomic percent ratio of [O]/[Al] in Al2O3 layer is less than 1), which leave behind many oxygen vacancies. It should be noted that from EDS data we found that the atomic percent ratio of [O]/[Al] for the sample with a 2.7 nm thick Al2O3 layer is larger than unity, forming less oxygen vacancies. Fig. 2 (a) shows the IeV curves for all the samples. Compared to the sample without Al2O3 layer, the current values decreased with Al2O3 layers, revealing rectifying characteristics. Fig. 2 (a) also shows that the IeV curves for the samples with an Al2O3 layer are different from the sample without Al2O3 layer. A shoulder appeared at low voltages (below 0.2 V; denoted as an arrow), associated with interfacial defect states forming low barrier regions [15]. Based on the thermionic emission (TE) model, the electrical parameters such as barrier height, ideality factor and series resistance were determined from IeV data, which are shown in Fig. 2 (b) and (c). With a 1.3 nm thick Al2O3 layer, the barrier height increased from 0.47 to 0.62 eV. Then the barrier height remained almost constant with increasing the Al2O3 thickness. The ideality factor was about 2.0 with increasing the thickness up to 4.9 nm and then increased to 2.9 at the Al2O3 thickness of 8.3 nm. As shown in Fig. 2(c), the series resistance at the Al2O3 thickness of 1.3 nm was very similar to that without Al2O3 layer. However, the series resistance increased significantly for the thicknesses of 2.7 and 4.9 nm. Interestingly, the series resistance decreased when the Al2O3 thickness became 8.3 nm. With considering either a Schottky effect or a PooleeFrenkel

Fig. 1. Cross-sectional scanning transmission electron microscope (STEM) images across the Al2O3 layer for the samples with (a) 2.7 and (b) 8.3 nm thick Al2O3 layer and (c) atomic percent vs. depth profile obtained from EDS line scan for the sample with an 8.3 nm thick Al2O3 layer.

258

H. Kim et al. / Vacuum 144 (2017) 256e260

Fig. 2. (a) Semilogarithmic currentevoltage (IeV) characteristics and (b) Ideality factor and barrier height and (c) series resistance vs. Al2O3 thickness.

effect, reverse bias IeVpffiffiffiffiffiffiffiffi characteristics are described with a ffi dependence of expðb=kT V=dÞ [16], where b is the emission coefficient, and d is the film thickness. From the linear fitting to the plots of ln(I) vs. V1/2 (see Fig. 3(a)), emission coefficients can be calculated from the slopes, which are shown in Fig. 3(b) bSC and bPF are the theoretical Schottky and PooleeFrenkel emission coefficients, respectively, given as bPF ¼ bSC ¼ ðq3 =pεS ε0 Þ1=2 , which are determined to be bPF ¼ 2.15
105 Vm1/2 V1/2 and bSC ¼ 1.08
105 Vm1/2 V1/2 (using εs ¼ 12.4 for InP). The experimental values were closer to the Schottky emission for the sample without Al2O3 layer. In contrast, the PooleeFrenkel emission became dominant for all the samples with an Al2O3 layer. Schottky emission refers to direct thermionic emission from the metal, whereas Poole-Frenkel (PF) emission refers to electric fieldenhanced thermal emission from a trap state into a continuum of

Fig. 3. (a) Plots of ln(IR) vs. V1/2 and (b) experimental emission coefficients. bSC and bPF indicate the theoretical Schottky and PooleeFrenkel emission coefficients, respectively.

electronic states; usually, the conduction band in an insulator [17]. When the deposited insulators contain a high density of defects, these defects cause energy states close to the band edges and restrict the current flow by capture and emission processes, thereby becoming the dominant PF emission mechanism. The electronic states of the Al2O3 layer play an important role in the reverse current. Hence, strong PF emission observed for the sample with a 4.7 nm thick Al2O3 layer indicates the presence of large electron traps. After depositing 10 nm thick metal layers on both InP and Al2O3 (2.7 nm)/InP samples, we performed XPS measurements. Fig. 4 (a) and (b) shows the atomic percent vs. etching depth for both samples. For Au/InP junction, oxygen atoms are present near the Au/InP interface. As shown in Fig. 4 (c), O 1s core levels exhibit the emission peak at 530e531 eV, associated with InOx, In2O3 or InPO4 [18,19]. Although the exact origin is unclear at present moment, it can be suggested that the oxide layer is present at Au/InP interface. Meanwhile, oxygen atoms are dominant near the Au/InP interface for Au/Al2O3/InP junction (Fig. 4 (b)). As shown in Fig. 4 (d), this is related with the Al2O3 layer (emission peak at ~ 531.9 eV [20]). The In 3d5/2 core levels in Fig. 4 (e) and (f) show that unlike Au/InP junction, Au/Al2O3/InP junction has the emission band related with In(OH)3 [21]. The supply of H2O precursor during the ALD process (including both H2O prepulse treatment and Al2O3 deposition) might contribute to the In(OH)3 formation, which eventually reduced the density of dangling bonds. Modulation of barrier heights with an inserted interfacial layer has been explained by several mechanisms. If the MIGS effect is dominant, the inserted interfacial layer will block the electron wave function from metal to semiconductor and therefore decrease the number of MIGS. Because the extrinsic MIGS pinning effect is limited with increasing the thickness of dielectric layer [22], the correspondent variation of barrier heights is expected. As shown in Fig. 2(a), however, the barrier heights are almost constant regardless of the Al2O3 thickness, indicating that the MIGS effect is not the dominant factor. When the native oxide-In2O3 is present on n-InP

H. Kim et al. / Vacuum 144 (2017) 256e260

259

Fig. 4. (a) and (b) XPS depth profiles, (c) and (d) O 1s core levels, and (e) and (f) In 3d5/2 core levels for Au/InP and Au/Al2O3(2.7 nm)/InP junctions.

surface, the reduction of barrier height in metal/n-InP junctions were attributed to the interface dipole effect due to the difference of areal density of oxygen atoms at Al2O3/In2O3 [13]. Dong et al. investigated the dependence of “self-cleaning” effect of the substrate oxides on InP substrate temperature during HfO2 deposition and found that the removal of In-oxide is more efficient at higher ALD temperatures [23]. The removal of surface oxide from GaAs substrate during the deposition of Al2O3 and HfO2 by ALD has also been reported [24]. Chauhan et al. observed that the barrier height for the Pt/Al2O3/n-InGaAs diode prepared with sulfur passivation is higher than that for the sample with native oxide (i.e., untreated sample) [25]. Because the barrier height increased with an Al2O3 layer in our work, it can be expected that a minimal native oxide is remained at the Al2O3/InP interface after ALD process. Meanwhile, Wang et al. used the dipole formation at the Al2O3/InGaAs interface to explain the increase of barrier height with a 2 nm thick Al2O3 layer [6]. If the Al2O3/InP interface dipole is a dominant mechanism to determine the barrier height, the variation of barrier height should be observed with different Al2O3 thicknesses (in other words, different dipole magnitudes), which is contradictory to the results in Fig. 2(a). This phenomenon is more likely that the Fermi level has been repinned close to midgap rather than become unpinned [6,25], which resulted from donor-type interface states after inserting the Al2O3 layer or the redistribution of interface states [6]. In order to confirm our results, we deposited 0.7 nm thick Al2O3 on InP and fabricated Au/Al2O3/InP diodes. Further, we performed temperature dependent IeV measurements for all the samples. The temperature dependent barrier heights in Fig. 5 show that even with a 0.7 nm thick Al2O3 layer, the barrier height increased compared to the sample without Al2O3 layer. This assures again that the Fermi level was repinned instead of being unpinned. Jackson et al. characterized the interface properties of ALDgrown Al2O3/GaN MIS capacitors and found that the interface

Fig. 5. Temperature dependent barrier heights for each sample.

state density decreases with increasing the Al2O3 thickness, i.e., the interface quality is improved with increased film thickness [26]. They suggested that the increased time at 300  C during the deposition of thicker layers may have led to the reduction in the interface state density through an in-situ annealing process. Wang et al also observed a similar trend for ALD Al2O3 on Si, suggesting that the longer deposition times for thicker layers could result in an annealing effect for thicker films [27]. These imply that the interface state density at Al2O3/InP interface can be reduced for the thicker Al2O3 layer due to the annealing effect. Meanwhile, oxygen plasma treatment on bulk ZnO produced the higher series resistance compared to the untreated ZnO, which was associated with the induced defects involving a split-oxygen interstitial complex acting as double donors and/or oxygen containing adsorbates

260

H. Kim et al. / Vacuum 144 (2017) 256e260

acting as trapping sites [28]. Both strong PF emission and large series resistance observed for the sample with a 4.7 nm thick Al2O3 layer, thus, can be due to the higher density of electron traps present at the Al2O3/InP interface. Based on the results so far, we propose the following mechanisms. For the sample without Al2O3 layer, the native oxide on InP surface and the surface states affect the current transport, producing the relatively low barrier height. For the samples with an Al2O3 layer, the native oxide was removed effectively and the dangling bonds were terminated, which in turn enhanced the barrier height. When the Al2O3 thicknesses are 1.3 nm, electrons can tunnel through the Al2O3 layer easily, showing low series resistance. At the Al2O3 thickness of 2.7 nm, the series resistance increases due to the increase of tunneling resistance. At the Al2O3 thickness of 4.9 nm, along with the tunneling through the Al2O3 layer, electron transport through the defects near the Al2O3/InP interface also play an important role in the total current. The presence of these electron traps may increase the series resistance and produce strong PF emission. With further increasing the Al2O3 thickness up to 8.3 nm, the interface defect density will decrease due to the in-situ annealing effect, reducing the series resistance. In conclusion, we have shown that the electrical properties of Au/n-InP contacts can be modulated by inserting an Al2O3 layer deposited by ALD. With an Al2O3 layer, the barrier height increased compared to the sample without Al2O3 layer, which was associated with the removal of native oxide and the termination of dangling bonds. The dominant current transport under reverse bias conditions was Poole-Frenkel emission for all the samples with an Al2O3 layer, indicating that the defects were generated in the Al2O3 layer. Regarding the series resistance, the tunneling resistance is the main determinant for thin Al2O3 layers (1.3 and 2.7 nm) whereas electron transport via localized defects in the Al2O3 layer becomes important for thick Al2O3 layers (4.9 and 8.3 nm). Acknowledgments This research was supported by Nano Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future

Planning (2009-0082580). This work was also supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2014R1A1A2054597). References [1] M. Xu, J. Gu, C. Wang, D. Zhernokletov, R. Wallace, P. Ye, J. Appl. Phys. 113 (2013) 013711. [2] L. Brillson, C. Brucker, A. Katnani, N. Stoffel, R. Daniels, G. Margaritondo, J. Vac. Sci. Technol. 21 (1982) 564. [3] V. Heine, Phys. Rev. 138 (1965) A1689. [4] R. Tung, Phys. Rev. B 64 (2001) 205310. [5] D. Connelly, C. Faulkner, P. Clifton, D. Grupp, Appl. Phys. Lett. 88 (2006) 012105. [6] R. Wang, M. Xu, P. Ye, R. Huang, J. Vac. Sci. Technol. B 29 (2011) 041206. [7] J. Lin, A. Roy, A. Nainani, Y. Sun, K. Saraswat, Appl. Phys. Lett. 98 (2011) 092113. [8] G. Choi, H. Yoon, S. Jung, Y. Jeon, J. Lee, W. Bahng, K. Park, Appl. Phys. Lett. 107 (2015) 252101. [9] Y. Zhou, M. Ogawa, X. Han, K. Wang, Appl. Phys. Lett. 93 (2008) 202105. [10] G. Wilk, R. Wallace, J. Anthony, J. Appl. Phys. 89 (2001) 5243. [11] Y. Wu, Y. Xuan, T. Shen, P. Ye, Z. Cheng, A. Lochtefeld, Appl. Phys. Lett. 91 (2007) 022108. [12] S. Wang, M. Cao, B. Sun, H. Li, H. Liu, Appl. Phys. Exp. 8 (2015) 091201. [13] S. Zheng, W. Yang, Q. Sun, L. Chen, P. Zhou, P. Wang, D. Zhang, F. Xiao, Appl. Phys. Lett. 103 (2013) 261602. [14] K. Kita, A. Toriumi, Appl. Phys. Lett. 94 (2009) 132902. [15] K. Suzue, S. Mohammad, Z. Fan, W. Kim, O. Aktas, A. Botchkarev, H. Morkoc, J. Appl. Phys. 80 (1996) 4467. [16] J. Sioomns, J. Phys. D. Appl. Phys. 4 (1971) 613. [17] H. Zhang, E. Miller, E. Yu, J. Appl. Phys. 99 (2006) 023703. [18] A. Lin, N. Armstrong, T. Kuwana, Anal. Chem. 49 (1977) 1228. [19] C. Sundararaman, H. Lafontaine, S. Poulin, A. Mouton, J. Currie, J. Vac. Sci. Technol. B 9 (1991) 1433. [20] B. Tan, K. Klabunde, P. Sherwood, J. Am. Chem. Soc. 113 (1991) 855. [21] M. Faur, D. Jayne, M. Goradia, C. Goradia, Surf. Interface Anal. 15 (1990) 641. [22] W. Monch, J. Appl. Phys. 111 (2012) 073706. [23] H. Dong, K. Santosh, X. Qin, B. Brennan, S. McDonnell, D. Zhernokletov, C. Hinkle, J. Kim, K. Cho, R. Wallace, J. Appl. Phys. 114 (2013) 154105. [24] C. Hinkle, A. Sonnet, F. Vogel, S. McDonnell, G. Hughes, M. Milojevic, B. Lee, F. Aguirre-Tostado, K. Choi, H. Kim, J. Kim, R. Wallace, Appl. Phys. Lett. 92 (2008) 071901. [25] L. Chauhan, S. Gupta, P. Jaiswal, N. Bhat, S. Shivashankar, G. Hughes, Thin Solid Films 589 (2015) 264. [26] C. Jackson, A. Arehart, E. Cinkilic, B. McSkimming, J. Speck, S. Ringe, J. Appl. Phys. 113 (2013) 204505. [27] J. Wang, S. Mottaghian, M. Baroughi, IEEE Trans. Electron. Dev. 59 (2012) 342. [28] H. Kim, H. Kim, D. Kim, Vacuum 101 (2014) 92.