Characterization of the InGaAs/InAlAs HEMT Transit Output Response by Using an Electro-Optical Sampling Technique

Transcription

1 Journal of the Korean Physical Society, Vol. 47, No. 3, September 2005, pp Characterization of the InGaAs/InAlAs HEMT Transit Output Response by Using an Electro-Optical Sampling Technique Seong-Jin Kim Photonic Devices Research Laboratory, Itswell Co., Ltd, 9-4BL, Ochang Scientific Industrial Complex, Chungbuk T. Itatani, T. Sugaya, M. Ogura and Y. Sugiyama National Institute of Advanced Industrial Science and Technology, Umezono, Tsukuba, Ibaraki , Japan (Received 9 May 2005) In order to characterize the transit output response of a 0.2-µm dual-gate InGaAs/InAlAs fieldeffect transistor (HEMT), we monolithically integrate a 0.2-µm interdigitated InAlAs-based photoconductive switch (PCS), a 0.2-µm dual-gate InGaAs/InAlAs HEMT, and coplanar waveguide transmission lines on an InP substrate. The impulse response of the 0.2-µm interdigitated InAlAsbased PCS is estimated to be as fast as 550 fs at full-width at half maximum (FWHM), corresponding to a 3-dB bandwidth of approximately 800 GHz. The transit output response of the 0.2-µm dual-gate InGaAs/InAlAs HEMT is estimated to be 7.3 ps at FWHM, indicating the cutoff frequency (f t) of approximately 136 GHz. This demonstrates that the electro-optical sampling technique is useful for accurately estimating the transit output response of ultrafast devices. PACS numbers: 85 Keywords: InGaAs/InAlAs HEMT, Monolithic integration circuit, Ultrafast device, Photoconductive switch, EO sampling, Optoelectronics I. INTRODUCTION In recent years, the tremendous progress in semi conductor-device speed has led to a strong demand for a new technique capable of accurately characterizing device performance over the range of device operating speeds. Performances of field-effect transistors (FETs) and resonant tunneling diodes are already over 350 GHz [1 3], and their performances are improving rapidly. However, above 100 GHz, the current gain cutoff frequency (f t ) and the maximum power gain (f max ) of FETs have been usually estimated by extrapolating the data measured at lower frequencies [4 6]; thus, the extrapolation rules can adversely affect the accuracy of device performance characterizations at higher frequency regions. A conventional electrical sampling technique using an oscilloscope or a network analyzer limits accurate device performance characterization due to its low bandwidth, error accumulation due to time-based drift and Jitter, and reflection due to impedance mismatch between the device under test (DUT) and the microwave probe. However, an electro-optical (EO) sampling method can kseongjin@itswell.com frequencies proved to be the performance limitation due to the poor material properties. Therefore, other ma overcome some of the difficulties concerning the abovementioned limitations of the conventional characterization techniques because the measurement bandwidth can be increased into the terahertz (THz) regions [7] and Jitter is inherently absent. The most direct way for generating a short-duration electrical pulse is by using a combination of PCS and an ultrafast pulsed laser. This implemental method is simpler than that of conventional electrical sampling. Typical EO sampling system uses a pulsed laser, a photoconductive switch (PCS) for electrical pulse generation, the DUT, and an EO crystal to optically probe the electrical field. The analysis of a high-electron-mobility field-effect transistor (HEMT) using the EO sampling technique was already demonstrated [8]. However, the PCS and the DUT were separately mounted on a chip, and they were interconnected by a wire-bonding technique. This could give rise to a reflection due to an impedance discontinuity at the wire bonding point; thus, the device performance might be degraded. The study of ultrafast PCS has concentrated on GaAs grown at low temperatures [7]. Although GaAs-based FETs have been widely used in high-speed optoelectronic devices and circuits, their operating speeds and

2 Characterization of the InGaAs/InAlAs HEMT Transit Output Seong-Jin Kim et al terials with characteristics of high mobility and high peak velocity are required in order to solve this problem. An InGaAs/InAlAs FET formed on an InP substrate is the most promising candidate for high-speed analog and digital applications because it possesses a high conduction-band discontinuity, a high mobility, and high peak velocity. Therefore, studies of InP-based devices and monolithic circuits are important for future high-speed telecommunications. In this paper, we report the transit output response of a 0.2-µm dual-gate InGaAs/InAlAs HEMT estimated by using an EO sampling technique. A 0.2-µm dual-gate InGaAs/InAlAs HEMT, an ultrafast InAlAs-based PCS, and coplanar waveguide transmission lines were monolithically integrated on an InP substrate. The impulse response of the InAlAs-based PCS and the transit output response of the InGaAs/InAlAs HEMT were 550 fs and 7.3 ps, respectively, at full-width at half maximum (HWHM). II. EXPERIMENT 1. Fabrication of a Monolithically Integrated Circuit An epitaxial structure for InGaAs/InAlAs HEMT and an InAlAs-based PCS was grown on a (100)-oriented InP substrate by using solid-source molecular beam epitaxy (MBE) at a growth temperature of 460 C. Prior to the epitaxial growth, the InP wafer was baked at 550 C for 5 min under As 4 -stabilized and atomic-hydrogen irradiation conditions to remove surface contamination. The epitaxial structure consisted of a 400-nm i-inalas buffer, a 10-nm i-ingaas channel, a 10-nm i-inalas spacer, a 15-nm i-inalas barrier, and a non-alloyed ohmic contact layer, as shown in Fig. 1. A Si δ-doped layer was inserted between the i-inalas barrier layer and the i- InAlAs spacer layer in order to supply electrons into the Fig. 1. Epitaxial structure of the InGaAs/InAlAs HEMT grown on a InP substrate. Fig. 2. SEM image of the monolithically integrated device with a 0.2-µm dual-gate InAlAs/InGaAs HEMT, a 0.2- µm interdigitated InAlAs-based PCS, and coplanar waveguide transmission lines. The inset is a high-resolution SEM image of the 0.2-µm interdigitated InAlAs-based PCS. InGaAs channel. The non-alloyed ohmic contact layer consisted of 1-nm n + -InAlAs, 5-nm n + -InGaAs, 1-nm n + -InGaAs, and 1-nm n + -InAs layers. A Si δ-doped layer was inserted between the 5-nm n + -InGaAs layer and the 1-nm n + -InGaAs layer in order to improve the ohmic contact property. Fig. 2 shows a scanning electron microscope (SEM) image of the monolithically integrated device. The inset shows a high-resolution SEM image of the InAlAsbased PCS having 0.2-µm line & 0.2-µm space interdigitated metal-semiconductor-metal (MSM) diode. A device set consisting of a 0.2-µm interdigitated InAlAsbased PCS, a 0.2-µm dual-gate InGaAs/InAlAs HEMT, and coplanar waveguide transmission lines was monolithically integrated on the InP substrate by using a combination of conventional photolithography, e-beam lithography, metal deposition, and liftoff. The 0.2-µm interdigitated InAlAs-based PCS and the 0.2-µm dual-gate InGaAs/InAlAs HEMT were simultaneously defined by using e-beam lithography and self-aligned recess-etching techniques. A polymethylmetacrylate (PMMA) was used to serve as the e-beam lithography resist and was exposed using a high-resolution e-beam system (JBX- 6000SA) with a beam energy of 50 kev and a beam diameter of 3 nm. The PMMA exposed with the e-beam system was developed in a methylisobutylketon (MIBK) solution. The self-aligned recess etching for the gate formation was performed with a citric acid : H 2 O 2 solution (1 : 1), which has a high selectivity between the InGaAs layer and the InAlAs layer. Ti (100 nm)/pt (10 nm)/au (30 nm) metals were used to serve as the MSM-PCS and the gate electrodes. AuGe (100 nm)/ni (30 nm)/au (200 nm) metals were employed to serve as the non-alloyed ohmic contact. The electric pad and the coplanar waveguide transmission lines were formed with Ti (100 nm)/au (200 nm) metals. The width and the space of the coplanar waveguide transmission line were each 5 µm. Individual devices were electrically isolated by wet etching the epitaxial layers into the i-inalas buffer layer. The wet etching was carried out using a solution of H 3 PO 4 : H 2 O 2 : H 2 O (3 : 1 : 50) at room

3 -522- Journal of the Korean Physical Society, Vol. 47, No. 3, September 2005 temperature. A 100-nm SiO 2 film was deposited on the entire wafer by using thermal-chemical vapor deposition (CVD) for device passivation. 2. Experimental setup for Electro-optical Sampling Fig. 3 shows a schematic diagram of the EO sampling system with an HR-coated LiNbO 3 EO crystal placed on the DUT circuit. The inset is a schematic diagram of the monolithically integrated device, where A and B are sampling points for measuring the PCS impulse response and the HEMT transit output response, respectively. An ultrafast colliding pulsed mode-locked laser (IMRA) [8] was employed to provide two optical pulses of 1.55 m and 0.78 µm in wavelength. The IMRA had a 10-mW optical output power, a 180-fs pulse width, and a 50-MHz repetition rate. Optical beams of 0.78 µm and 1.55 µm were used as an excitation source for the PCS and as a sampling source of the electrical field, respectively. The pulse beam was split into two beams with a ratio of 9 to 1 by using a beam splitter. The stronger branch beam excited the PCS so as to generate a Gaussianshaped short-electrical signal propagating along a coplanar waveguide transmission line. The other linearly polarized branch beam probed the electrical field induced by the electric pulse penetrating into the EO crystal. The probing beam reflected by the EO crystal was separated into horizontal and vertical polarizations by using a polarizing beam splitter and the two polarizations were detected by using two photodiodes connected to a differential amplifier. In order to achieve a linear EO response and maximum sensitivity at the same time, we used a polarization compensator to optically bias the modulator at its quarter-wave point. III. RESULTS AND DISCUSSION The DC current-voltage (I V ) characteristics of all devices were tested on a probe station by using a HP- 4156A semiconductor parameter analyzer. Fig. 4 shows the I DS -V GS characteristics under illumination condition (solid line) and dark condition (dashed line), where the InAlAs-based PCS was exited with a 750-nm CW semiconductor laser. The I DS -V GS characteristics are more distinct for the illumination condition than for the dark condition. The drain current (I DS ) under illumination normally increased as the gate bias was increased. This shows that a monolithically integrated device consisting of a 0.2-µm InAlAs-based PCS, a 0.2-µm dual-gate InGaAs/InAlAs HEMT, and coplanar waveguide transmission lines had been successfully interconnected. The pinch-off characteristic is less for the dark condition than for the illumination condition because there is no gate bias on the gate electrode under the dark condition. The impulse response of the PCS was measured with a HR-coated LiNbO 3 -tip probe placed 10 µm away from the PCS edge, corresponding to position A shown in Fig. 3. The PCS was biased by as much as 10 V because the response time was strongly affected by the strength of the electrical field. The photon-generated carriers at low electrical field may move with small drift velocities, which results in a long transit time. The impulse response of a line & space interdigitated PCS is limited by the electron-hole pair recombination time and by the capacitance (RC) time constant [9]. The finger spacing determines the intrinsic response time of the PCS. The ratio of the finger width to finger spacing determines the PCS capacitance, which affects the external response time. The capacitance of the PCS can be decreased by reducing the ratio of the finger width to the finger pitch, resulting in a smaller RC-time constant; thus, it is desirable to minimize both the finger spacing Fig. 3. Schematic diagram of the EO sampling measurement system. Fig. 4. DC I DS-V GS characteristics of the 0.2-µm dualgate InGaAs/InAlAs HEMT under illumination conditions (solid line) and dark conditions (dashed line) for the 0.2-µm interdigitated InAlAs-based PCS.

4 Characterization of the InGaAs/InAlAs HEMT Transit Output Seong-Jin Kim et al measured for V DS ranging from 0 V to 0.6 V and V GS = 1.6 V. The spectrum shape of the HEMT transit output response was almost symmetric. The transit output response of the 0.2-µm dual-gate InGaAs/InAlAs HEMT was 7.3 ps at FWHM, which implies a cutoff frequency (f t ) of approximately 136 GHz. This result is similar to that of the trend previously reported (f t = 125 GHz) for InGaAs/InAlAs HEMTs [11]. However, the transit output response of the 0.2-µm dual-gate InGaAs/InAlAs HEMT should be compared with the data measured by using electrical sampling in order to probe the accuracy of the EO sampling technique. Fig. 5. Impulse response, estimated by using the EO sampling technique, of the 0.2-µm interdigitated InAlAs-based PCS. IV. CONCLUSION We estimated the transit output response of a 0.2-µm dual-gate InAlAs/InGaAs HEMT by using an EO sampling technique. In order to prevent reflection and deterioration of the pulse waveform induced from the bonding wires, we monolithically interconnected the 0.2-µm dual-gate InAlAs/InGaAs HEMT and the 0.2-µm interdigitated InAlAs-based PCS by using coplanar waveguide transmission lines. The impulse response of the InAlAs-based PCS was estimated to be 550 fs at FWHM, corresponding to a 3-dB bandwidth of approximately 800 GHz. The transit output response of the 0.2-µm gate In- AlAs/InGaAs HEMT was 7.3 ps at FWHM, corresponding to a cutoff frequency (f t ) of approximately 136 GHz. Fig. 6. Transit output response, estimated by using the EO sampling technique, of the 0.2-µm dual-gate InGaAs/InAlAs HEMT. and width to obtain an ultrafast impulse response. Fig. 5 shows the impulse response spectrum of the InAlAs-based PCS, as measured by the EO sampling technique. The impulse response of the InAlAs-based PCS was estimated to be 550 fs at FWHM. This indicates that the 3-dB bandwidth of the InAlAs-based PCS, calculated by using a 0.441/FWHM, is approximately 800 GHz. This impulse response is faster than that of a previously reported InGaAs-based PCS [10]. The obtained impulse response of the PCS was enough to characterize the transit output response of the 0.2-µm dual-gate InGaAs/InAlAs HEMT formed on the InP substrate. Fig. 6 shows the transit output response spectrum of the 0.2-µm dual-gate InGaAs/InAlAs HEMT, as measured by using the EO sampling technique. The transit output response of the 0.2-µm dual-gate InGaAs/InAlAs HEMT was sampled using the HR-coated LiNbO 3 -tip probe placed on the drain transmission line 10 µm away from the gate edge, corresponding to position B shown in Fig. 3. The transient output signal of the HEMT was ACKNOWLEDGMENTS The authors would like to thank Drs. T. Sakamoto and M. Komuro for useful discussions on and encouragement width this study. They also thank Drs. H. Hiroshima and S. Haraichi for the valuable suggestions about the e-beam lithography, and Mr. K. Hikosaka for the PCbased instrumentation. REFERENCES [1] L. D. Nguyen, A. S. Brown, M. A. Thompson and L. M. Jelloian, IEEE Trans. Electron Devices 39, 2007 (1992). [2] E. R. Brown, J. R. Söderstöm, C. D. Parker, L. J. Mahoney, K. M. Molvar and T. C. Mcgill, Appl. Phys. Lett. 58, 2291 (1991). [3] N. Shimizu, T. Nagatsuma, M. Shinagawa and T. Waho, IEEE Electron Device Lett. 16, 262 (1995). [4] J. Y. Shim, J. H. Lee, H. S. Yoon, K. H. Lee, D. J. Kim and J. H. Lee, J. Korean Phys. Soc. 41, 528 (2002). [5] S. C. Kim, B. O. Lim, T. J. Baek, B. S. Ko, D. H. Shin and J. K. Rhee, J. Korean Phys. Soc. 45, 1089 (2004). [6] S. J. Kim, J. Y. Shim, J. H. Lee, H. S. Yoon, K. H. Lee, D. J. Kim and J. H. Lee, J. Korean Phys. Soc. 42, 276 (2003).

5 -524- Journal of the Korean Physical Society, Vol. 47, No. 3, September 2005 [7] S. Gupta, M. Y. Frankel, J. A. Valdmanis, J. F. Whitaker, G. A. Mourou, F. W. Smith and A. R. Calawa, Appl. Phys. Lett. 59, 3276 (1991). [8] J. A. Valdmanis and R. L. Fork, IEEE J. Quantum Electronics 22, 112 (1986). [9] S. Y. Chou and M. Y. Liu, IEEE J. Quantum Electronics 28, 2358 (1992). [10] Y. G. Wey, K. Giboney, J. Bowers, M. Rodwell, P. Silvestre, P. Thiagarajan and G. Robinson, IEEE J. Lightwave Technology 13, 1490 (1995). [11] A. Mahajan, P. Fay, M. Arafa, G. Cueva and I. Adesida, IEEE International Electron Dev. Meeting Technical Digest (New York, 1996), p. 51.