Enhanced peak-to-valley current ratio in InGaAs/ InAlAs trench-type quantum-wire negative differential resistance field-effect transistors

Transcription

1 JOURNAL OF APPLIED PHYSICS 97, Enhanced peak-to-valley current ratio in InGaAs/ InAlAs trench-type quantum-wire negative differential resistance field-effect transistors Takeyoshi Sugaya, a Kee-Youn Jang, Cheol-Koo Hahn, Mutsuo Ogura, and Kazuhiro Komori National Institute of Advanced Industrial Science and Technology (AIST), and Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology (JST), Umezono, Tsukuba, Ibaraki, , Japan Akito Shinoda and Kenji Yonei Shibaura Institute of Technology, , Shibaura, Minato-ku, Tokyo , Japan Received 16 July 2004; accepted 2 December 2004; published online 19 January 2005 Trench-type narrow InGaAs quantum-wire field-effect transistors QWR-FETs have been fabricated on 311 A InP V-groove substrates by hydrogen-assisted molecular-beam epitaxy. Enhanced negative differential resistance NDR effects with a peak-to-valley ratio PVR as high as 13.3 have been observed at an onset voltage of 0.16 V in the QWR-FETs at 24 K. The PVR increased with reductions in the InGaAs epitaxial layer thickness, which caused an enhanced mobility difference between the QWR and side quantum wells QWs. This forms a velocity modulation transistor based on the real-space transfer of electrons from the high mobility QWR to the low mobility side QWs. The NDR effects were observed up to 230 K as the gate length was decreased to 50 nm. A unique feature of the QWR-FET is that NDR effects are controllable with the gate bias in a three-terminal configuration American Institute of Physics. DOI: / I. INTRODUCTION Negative differential resistance NDR effects in semiconductor heterostructures have attracted much interest for high-frequency oscillators, high-speed logic, and memory devices. Resonant tunneling diodes have been the most established NDR devices. 1 3 However, they are two-terminal devices and combination either with passive elements, fieldeffect transistors, or heterobipolar transistors have been necessary. 4 6 In contrast, a three-terminal NDR device such as a real-space transfer RST transistor effectively simplifies the device structure and reduces the circuit complexity In the conventional RST device, an onset voltage of the NDR V NDR is large, and the peak-to-valley current ratio PVR is poor because a significant portion of electrons remains in a high electron mobility channel. Recently, we have fabricated a trench-type InGaAs/ InAlAs quantum wire QWR structure on an InP 311 A V-groove substrate, which was grown by a selective epitaxial growth using molecular-beam epitaxy MBE with atomic hydrogen and a dimer arsenic source. 11,12 A field-effect transistor FET with a trench-type QWR channel has pronounced NDR effects with the highest PVR of 6.2 and the V NDR as low as 0.1 V at 40 K. 13 In this work, we realized more ideal NDR effects by reducing the InGaAs epitaxial layer thickness, which caused an enhanced mobility difference between the QWR and side quantum wells QWs. We also confirmed the gate length and temperature dependence of the NDR effects. The InGaAs/ InAlAs QWR heterostructures were grown by MBE on a 311 A semi-insulating InP V-groove patterned substrate. V grooves were prepared by photolithography and wet etching HCl:H 3 PO 4 :H 2 O 2, 50:10:1 by volume. The grooves were formed along the 011 direction and their sidewalls comprised of 100 and 011 planes. The substrate was loaded into a MBE chamber and cleaned with atomic hydrogen at 500 C for 2 min, after which a 400-nm In 0.52 Al 0.48 As barrier layer was then grown under an As 2 source at 510 C with atomic hydrogen. The resulting trench formed on this layer consisted of 111 A and 331 B facets, comprised of the oblique slopes of the two trapezoids, and grown selectively on the 100 and 011 sidewalls of the original substrate, as shown in Fig. 1. The trench angle between these new facets was approximately 22. Such a nara Electronic mail: t.sugaya@aist.go.jp II. EXPERIMENT FIG. 1. A cross-sectional view of the trench-type InGaAs/ InAlAs QWR structures observed with transmission electron microscopy TEM. Natural superlattices with a period of 2 nm are formed at the bottom of the InAlAs trench structure /2005/97 3 /034507/5/$ , American Institute of Physics

2 Sugaya et al. J. Appl. Phys. 97, FIG. 3. a Schematic illustration of cross-sectional QWR-FET, b SEM plan-view image of the trench-type QWR-FET. c The detailed layer structure of the FET. illustration and b a plan-view scanning electron microscopy SEM image of the 100-nm gate-length trench-type InGaAs QWR-FET we used in our experiments. We have fabricated InGaAs/ InAlAs QWR-FETs with various InGaAs QWR layer thicknesses 7 20 nm. FIG. 2. Fabrication process flow for QWR-FET and a detailed grown layer structure is shown. row facet angle has not been reported previously with MBE growth. Natural superlattices with a period of 2 nm are formed at the bottom of the InAlAs trench structure. The formation of these superlattices provides an atomically flat surface for the trench-type QWR. A 10-nm In 0.53 Ga 0.47 As QWR layer was grown on the InAlAs trench under an As 4 source with atomic hydrogen. This corresponds to the dark region in the transmission electron micrograph image in Fig. 1. The resulting QWR had a cross-sectional depth of 10 nm and a width of 25 nm. The 111 A and 331 B QWs are also grown on the sidewalls. To fabricate the QWR-FET, a 10 -nm spacer of undoped InAlAs was grown on top of the InGaAs, followed by a Si -doped layer. The detailed layer structure is shown in Fig. 2 b, in which the additional -doped layers facilitate the formation of nonalloyed ohmic contacts to the QWR. The fabrication process flow for the QWR-FET is shown in Fig. 2. After etching away the sidewalls to leave just a single wire Fig. 2 a, the etched surface was passivated by depositing a 100-nm SiO 2 film, into which windows were opened and Ti/ Au evaporated to form nonalloyed source and drain contacts with a lateral separation of 4 m Fig. 2 c. As a final processing step, a Ti/Pt/Au Schottky gate with a length of nm was deposited between the source and drain, using recess etching and lift-off Fig. 2 d. Figure 3 includes a a schematic III. NEGATIVE DIFFERENTIAL RESISTANCE IN INGAAS QWR-FET Figure 4 shows the static NDR characteristics of QWR- FETs with a 100-nm gate length and QWR thicknesses of a 20, b 10, and c 7 nm. The evolution of the NDR effects is clearly observed with increasing gate voltage V G. The PVR of the NDR effects increases systematically as the QWR thickness decreases, while the V NDR remains almost the same for all the samples. The maximum PVR is as high as 13.3 at V G =4.5 V in the 7-nm-thick QWR-FET. Such a high PVR has not been previously reported for a QWR-FET device. Figure 5 a shows the NDR effects with a 1000-nm gate length, in which the V NDR is about 0.4 V. Figure 5 b plots the gate-length dependence of V NDR and the PVR of a trenchtype QWR-FET from 50 to 1000 nm. As the gate length is prolonged, the V NDR also increases from 0.1 to 0.4 V, while the PVR decreases from 8 to 2. These results mean that the NDR is improved effectively by reducing the gate lengths. Figure 6 shows the NDR effects of the 50-nm-gate In- GaAs QWR-FET V G =4.0 V in the K temperature range. The QWR thickness is 10 nm. The solid lines represent the NDR effects at 40 and 230 K. At 40 K, the V NDR is clearly seen at a drain-source voltage as low as 0.1 V, and the PVR for the QWR-FET is 6.2. The NDR characteristic was clear until 230 K with the PVR and V NDR of 1.4 and 1.7 V. Above 270 K, the NDR effects disappeared. IV. MECHANISMS FOR NDR EFFECTS The NDR mechanism in the trench-type QWR is interpreted as a subband transition from the high mobility fundamental level, predominantly in the QWR, to the low mobility higher subband levels in the side QWs. 13 There are two other possible mechanisms for the NDR, which are the Gunn effect and a RST of the carriers to the InAlAs space layer. The Gunn effect is explained by the intervalley transition of the electrons. The energy separation, E L, between the valley

3 Sugaya et al. J. Appl. Phys. 97, FIG. 5. a Typical NDR effects with 1000-nm gate length and b the gate-length dependence of the NDR effects in the trench-type QWR-FET. FIG. 4. NDR characteristics of the 100-nm-gate trench-type QWR-FETs at 24 K with a 20-nm QWR thickness V NDR =0.22 V, PVR=2 at V G =4.5 V, b 10-nm QWR thickness V NDR =0.2 V, PVR=4.3 at V G =4.5 V., and c 7-nm QWR thickness V NDR =0.16 V, PVR=13.3 at V G =4.5 V. trench-type QWR-FET was undetectable around the onset voltage. 13 Therefore, the V NDR of 0.1 V cannot be explained by an overflow of electrons from the InGaAs QWR layer into the InAlAs space layer. The subband transition from the high mobility fundamental level to the low mobility higher subband levels in the side QWs is another type of RST because the QWR and side QWs are structurally connected and embedded inside the barrier layer. We calculated the electron wave function distribution at each subband energy level based on the actual geometry of the QWR taken from the cross-sectional trans- and the lowest satellite valley L valley of In 0.53 Ga 0.47 As, is about 0.55 ev, namely, larger than the V NDR of our device; therefore, the observed NDR cannot be explained by the Gunn effect. For the RST mechanism, the electrons are transferred from the InGaAs QWR layer to the InAlAs space layer beneath the Schottky gate. 7 The conduction-band offset between the InGaAs QWR and the InAlAs space layer is 0.46 ev. The channel electrons have to obtain at least ev before they can escape into the InAlAs space layer, because the fundamental subband level in the trench-type QWR is located at ev above the InGaAs conduction band. Therefore, a V NDR of 0.1 V is too small for the above mechanism. With respect to the NDR effects of the ridgetype QWR-FETs, the V NDR was about 0.3 V and the gate leakage current was detected after the NDR had been observed. 14 In contrast, the gate leakage current of the FIG. 6. Temperature dependence of the NDR effects for the InGaAs QWR- FET with 50-nm gate length and V G fixed at 4.0 V.

4 Sugaya et al. J. Appl. Phys. 97, FIG. 8. DOS of the accelerated electrons is inversely proportional to the square root of the electron kinetic energy along the QWR. Therefore, the DOS of the fundamental subband level E 1 becomes relatively smaller than that of the higher subband levels, and it forces hot electrons to transfer to the higher subband levels. FIG. 7. Color Wave functions at each subband energy are calculated by the actual geometry of the QWR. a Cross-sectional TEM image of trench-type QWR, b fundamental subband energy level E 1 =0.122 ev, c second energy level E 2 =0.169 ev, and d third energy level E 3 =0.206 ev. Electrons at the fundamental level are located mainly at the bottom of the trench-type QWR, while the probability of electrons at the higher subband levels increases at the side QW. mission electron microscopy TEM image by numerically solving the two-dimensional 2D single band Schrödinger equation using the finite element method, as shown in Fig. 7. Effective-mass ratios of m * /m o =0.041 and m * /m o =0.075 were employed for In 0.53 Ga 0.47 As and In 0.52 Al 0.48 As, respectively, at 10 K. 15 The electrons at the fundamental level are located mainly at the bottom of the trench-type QWR, as shown in Fig. 7 b. While the electron wave function extends towards the side QWs located on both sides of the QWR as the order of the subband levels increases, as shown in Figs. 7 c and 7 d. These tendencies mean that electrons move towards the side QW when the source-drain field accelerates the electrons. The relative energy difference between QWR and QWs reduces at the positive gate bias because the side QWs are closer to the gate electrode than the trench QWR in our device configuration. So, positive gate bias is favorable to reduce the V NDR. The calculated energy separations between the fundamental, second, and third subband levels were 47 and 84 mev, respectively. A V NDR of 0.1 V is consistent with the subband energy separations of the trench QWR because some parts of the drain voltage are dissipated at the contact and source-drain series resistances. The effect of interface roughness is stronger at higher subband levels because the electron wave function penetrates deeper into the thin side QWs as the eigenenergy of the transverse mode increases. 16 Therefore, the electron mobility is smaller at the higher subband levels than that at the fundamental level. As a result, negative resistance is produced when channel electrons are transferred from the high mobility fundamental subband level, predominantly in the QWR, to the low mobility higher subband levels that predominate in the side QWs at a sourcedrain voltage larger than V NDR. In Fig. 4, the PVRs of the 10- and 20-nm-thick FETs are 4.3 and 2, respectively. The large PVR with the thin InGaAs layer indicates that the effect of interface roughness is dominant with the thin side QWs. This corresponds to the observation that electronic states are isolated in the monoatomic indentation in thin QWs. 17 The above results confirm that the mechanism for the NDR effects is the RST from the QWR to the thin side QWs, because the Gunn effect and the RST to the InAlAs barrier layer cannot explain the enhanced PVR with a decrease in the QWR thickness. For the RST to the side QWs, it is necessary that the carriers have to be heated more effectively to transfer to the high-energy subband. As shown in Fig. 5, the short gatelength device is effective because of the less scattering probability, leading to intersubband transfer, and NDR. 14 The RST mentioned above is categorized as a velocity modulation transistor 18,19 with the expectation that NDR based on the subband transition will be fast compared with the carrier overflow in conventional RST devices. As shown in Fig. 8, the saw toothlike density of states DOS of the QWR also favors a pronounced PVR, because the DOS of the accelerated electrons decreases in inversely proportional to the square root of the electron energy and it forces the accelerated electrons to transfer from the fundamental level to the higher subband levels. This effect is intrinsic to the one-dimensional 1D electrons. In the 2D electrons with steplike DOS, in contrast, almost the same amount of hot electrons remains at the high mobility fundamental subband level, and it causes a small PVR ratio. V. CONCLUSIONS In conclusion, we fabricated the trench-type InGaAs QWR-FETs on 311 A InP nonplanar substrates. Enhanced NDR effects with the PVR as high as 13.3 at an onset voltage of 0.16 V were observed in the QWR-FETs at 24 K. The NDR mechanism in the trench-type QWR is based on the RST of carriers from the high mobility QWR layer to the low mobility side QWs. The PVR increased as the InGaAs epitaxial layer thickness decreased, which caused an enhanced mobility difference between the QWR and the side QWs. The saw toothlike DOS of the 1D electrons also favors the pronounced PVR. The RST transistors with such a low V NDR and large PVR have been realized at high temperature.

5 Sugaya et al. J. Appl. Phys. 97, Because the RST transistors by the trench QWRs are simple in the device structure and controllable with the gate bias in a three-terminal configuration, they are favorable for high speed and low-power modules with reduced circuit complexity. 1 L. L. Chang, L. Esaki, and R. Tsu, Appl. Phys. Lett. 24, T. C. L. G. Sollner, W. D. Goodhue, P. E. Tannenwald, C. D. Parker, and D. D. Peck, Appl. Phys. Lett. 43, D. G. Austing, T. Honda, Y. Tokura, and S. Tarucha, Jpn. J. Appl. Phys., Part 1 34, B. F. Aull, K. B. Nichols, P. A. Maki, S. C. Palmateer, E. R. Brown, and T. A. Lind, Appl. Phys. Lett. 63, N. El-Zein, G. Maracas, V. Nair, G. Kramer, and H. Gorokin, Proceedings of the IEEE International Symposium on Compound Semiconductors, San Diego, California, 8 11 September 1997 unpublished, p N. Yokoyama, K. Imamura, S. Muto, S. Hiyamizu, and H. Nishi, Jpn. J. Appl. Phys., Part 2 24, L A. Kastalsky and S. Luryi, IEEE Electron Device Lett. 4, I. C. Kizilyalli and K. Hess, J. Appl. Phys. 65, P. M. Mensz, P. A. Garbinski, A. Y. Cho, D. L. Sivco, and S. Luryi, Appl. Phys. Lett. 57, C. L. Wu, W. C. Hsu, M. S. Tsai, and H. M. Shieh, Appl. Phys. Lett. 66, T. Sugaya, K. Matsumoto, K. Yonei, T. Sekiguchi, M. Ogura, and Y. Sugiyama, Appl. Phys. Lett. 78, T. Sugaya, K. Matsumoto, K. Yonei, M. Ogura, Y. Sugiyama, and K. Y. Jang, Appl. Phys. Lett. 78, K.-Y. Jang, T. Sugaya, C.-K. Hahn, A. Shinoda, K. Yonei, M. Ogura, and K. Komori, Appl. Phys. Lett. 83, S. J. Kim, T. Sugaya, M. Ogura, and Y. Sugiyama, Jpn. J. Appl. Phys., Part 1 39, R. J. Nicholas, J. C. Portal, C. Houlbert, P. Perrir, and T. P. Pearsall, Appl. Phys. Lett. 34, X. L. Wang, M. Ogura, H. Matsuhata, and A. Hamoudi, Appl. Phys. Lett. 71, R. Grousson, V. Voliotis, N. Grandjean, J. Massies, M. Leroux, and C. Deparis, Phys. Rev. B 55, R. Dingle, H. L. Stoermer, A. C. Gossard, and W. Wiegmann, Appl. Phys. Lett. 33, H. Sakaki, Jpn. J. Appl. Phys., Part 2 21, L