GaAs HEMT structure for on-chip RF power detection

Superlattices and Microstructures 47 (2010) 274–287

Contents lists available at ScienceDirect

Superlattices and Microstructures journal homepage: www.elsevier.com/locate/superlattices

Design, fabrication and characterization of a Schottky diode on an AlGaAs/GaAs HEMT structure for on-chip RF power detection Farahiyah Mustafa a , Norfarariyanti Parimon a , Abdul Manaf Hashim a,∗ , Shaharin Fadzli Abd Rahman a , Abdul Rahim Abdul Rahman a , Mohd Nizam Osman b a

Material Innovations and Nanoelectronics Research Group, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia b

Telekom Research & Development, TM Innovation Centre, 63000 Cyberjaya, Selangor, Malaysia

article

info

Article history: Received 6 July 2009 Received in revised form 7 October 2009 Accepted 27 October 2009 Available online 17 November 2009 Keywords: Schottky diode Dipole antenna AlGaAs/GaAs RF power detector HEMT

∗

abstract A Schottky diode was designed and fabricated on an n-AlGaAs/GaAs high electron mobility transistor (HEMT) structure for RF power detection. The processing steps used in the fabrication were the conventional steps used in standard GaAs processing. Current–voltage measurements showed that the devices had rectifying properties with a barrier height of 0.5289–0.5468 eV. The fabricated Schottky diodes detected RF signals well and their cut-off frequencies up to 20 GHz were estimated in direct injection experiments. To achieve a high cut-off frequency, a smaller Schottky contact area is required. The feasibility of direct integration with the planar dipole antenna via a coplanar waveguide transmission line without insertion of matching circuits was discussed. A higher cut-off frequency can also be achieved by reducing the length of the coplanar waveguide transmission line. These preliminary results represent a breakthrough as regards direct on-chip integration technology, towards the realization of a ubiquitous network society. © 2009 Elsevier Ltd. All rights reserved.

Corresponding author. Tel.: +60 7 553 5688; fax: +60 7 556 6272. E-mail address: [email protected] (A.M. Hashim).

0749-6036/$ – see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.spmi.2009.10.011

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

275

1. Introduction The explosive growth of internet and wireless technologies starting in the late 21st century has opened up prospects towards an advanced ubiquitous network society; nanoelectronic devices are the most promising candidates for yielding such technologies. Therefore, nanoelectronic systems are increasingly vulnerable to malfunction due to incident electromagnetic (EM) radiation, particularly since many integrated circuits operate at lower and lower voltages. Damaging RF radiation can be produced intentionally, such as by high power microwave generators [1], or accidentally, such as by ambient sources like lightning. Then, it becomes a great interest to know how, and at what level, microwaves penetrate equipment shielding and reach the vulnerable chips. This motivated our group to work on on-chip RF detectors, both for measuring power at the chip level and for developing strategies to mitigate its effects. Knowing the RF power levels in various chips and locations within chips is likely to be more useful than the ‘‘digital’’ information that a given external RF power level made the circuits fail. III–V materials are the most promising materials for high frequency device use because of the high electron mobility and other unique features such as the formation of two-dimensional electron gas (2DEG) layer [2,3]. The result of this is that the devices can be switched more quickly because there are lesser effects of collisions. Recently, the concept of the intelligent quantum (IQ) chip introduced by Hasegawa et al. [4] applied this compound semiconductor material as a base material for on-chip integration. Ideally, the chip power detectors would have a wide dynamic range, and be fabricated on this compound semiconductor. Schottky diodes are known as fast rectifying devices and can be used as RF detectors [5]. In special molecular beam epitaxy (MBE) grown geometries, RF detection up to 100 GHz has been reported [6]. However, in foundry fabricated Si-based diodes, detection of only up to 600 MHz has been reported [7]. Recently, a CMOS fabricated Schottky diode detecting RF signals up to 10 GHz in direct injection experiments and in the range of 9.5–19.5 GHz in microwave irradiation experiments has also been reported [8]. However, to our knowledge there has been no report on the design and fabrication of an n-AlGaAs/GaAs HEMT Schottky diode for RF power detector use. Recently our group have developed, besides RF power detectors, some other new functional devices such as THz wave detectors and plasma wave THz amplifiers utilizing the same AlGaAs/GaAs HEMT structure [9–11]. In this paper, the design and fabrication of a Schottky diode on an n-AlGaAs/GaAs high electron mobility transistor (HEMT) structure for RF power detection was reported. We also discussed the feasibility of direct integration with a planar dipole antenna via a coplanar waveguide transmission line without insertion of a matching circuit for real practical application. These preliminary results represent a breakthrough for direct on-chip integration technology, towards the realization of a ubiquitous network society. 2. Design and fabrication of the Schottky diode We have chosen to fabricate a Schottky diode on the AlGaAs/GaAs HEMT structure because of the higher electron mobility that can be provided by the 2DEG layer. In addition, among novel modern structures, AlGaAs/GaAs heterostructures have emerged as the most popular material for confining electrons. As a semiconductor material for Schottky diode use, GaAs has been considered a most promising material because of its stability, capability of making a good Schottky contact and well-developed fabrication process technology. This material structure is also suitable for the development of the so-called IQ chip which has been considered as the most promising chip structure for future ubiquitous network society use [4]. The sample is an AlGaAs/GaAs modulation-doped heterostructure grown by molecular beam epitaxy. The interface of the n-doped AlGaAs layer and undoped GaAs layer defines a 2DEG system where electron motion perpendicular to the layer is frozen out, thus producing highly mobile electrons. The thickness of the main layers, from bottom to top, are as follows: 625 µm semi-insulated high dielectric constant GaAs substrate; 500 nm GaAs buffer layer; 10 nm AlGaAs buffer layer; 20 nm undoped GaAs layer; 10 nm AlGaAs spacer layer; ndoped AlGaAs (Si δ doping) barrier layer; 10 nm GaAs undoped cap layer. The devices were designed and fabricated using photolithography and a standard lift-off technique. The carrier mobility and the

276

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

Fig. 1. (a) Schematic structure of the Schottky diode and (b) the fabricated Schottky diode (top view).

carrier sheet density obtained from Hall measurements at room temperature were 6040 cm2 /V s and 8.34 × 1011 cm−2 , respectively. The Schottky electrode is formed by Ni/Au and the ohmic electrode is formed by alloyed Ge/ Au/Ni/Au. As shown in Fig. 1(a), our devices are facilitated with a coplanar waveguide (CPW) structure on both sides of the Schottky and ohmic contacts which possess ground–signal–ground (GSG) pad structures. The dimensions of the gap, a, and width, b, for a CPW calculated on the basis of the Wheeler equation [12] were chosen to be 60 µm and 90 µm, respectively, in order to produce the characteristic impedance, Z0 , of 50 . This CPW structure is similar to the proposed antenna structure described in Section 3, where it can make direct integration without insertion of a matching circuit possible. This CPW structure permits direct injection of the RF signal through a Cascade ground–signal–ground Infinity-150 microprober. In this preliminary study, the Schottky contact area, A, is 20 µm × 20 µm, the lengths of the CPW, L, are 20 µm and 100 µm and the distances, d, between Schottky–ohmic contacts is 40 µm. Hereafter, the device with the CPW length, L, of 20 µm is named Schottky diode A and the device with the CPW length, L, of 100 µm is named Schottky diode B. It is discussed in Section 3 that, to achieve a high cut-off frequency, the rectifying metal–semiconductor contact area needs to be reduced. However, too small a contact area limits the maximum power that the device can detect before the diode burns out. Therefore, the area of the diode, A, is the main design parameter since most of the other parameters such as the work functions of the metal and the semiconductor are determined by the process. A photo of the fabricated Schottky diode is shown in Fig. 1(b).

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

277

Fig. 2. DC I–V characteristics of the fabricated Schottky diode.

3. Measured results for the Schottky diode and discussion 3.1. Current–voltage (I–V ) measurement After fabricating the Schottky diode, the DC I–V characteristics were measured using a Keithley Semiconductor Characterization System model 4200 and a Micromanipulator Probe Station. As shown in Fig. 2, the DC I–V curve of a fabricated Schottky diode A and Schottky diode B shows a diode I–V curve with a 1.176 k and 909.1  series resistance, respectively, defined as the slope between 2 and 3 V. The threshold voltages, VTH , for both devices are estimated to be 1.1 V, as shown in Fig. 2 (inset). Measurements of the reverse saturation currents of the devices are used to calculate the Schottky barrier heights (SBHs) from the Richardson–Dushman equation for the thermionic current [5]:

φB = Vt · ln



A · A∗ · T 2 IS



.

(1)

In Eq. (1), φB is the barrier height in volts, IS is the reverse saturation current, Vt is the thermal voltage, A∗ is the effective Richardson constant, A is the area of the metal–semiconductor contact, and T is the absolute temperature. The reverse leakage current for device A was 1.99 nA and the SBH was calculated to be 0.5468 eV, while the reverse leakage current for device B was 3.97 nA and the barrier height was calculated to be 0.5289 eV. These SBH values are almost three times lower than the ideal calculated value which is 1.443 eV due to the small contact area. The reduction of the Schottky barrier height may be due to the fabrication process, i.e. the annealing process, which can result in a decrease in barrier height as suggested by Zhang et al. in Ref. [13]. They have reported Schottky contacts of different metals to n-type AlGaAs/GaAs structures and proposed a model which involves the quality of the contact and defect formation at the semiconductor surface due to interdiffusion and/or penetration of metal into the semiconductor. This model can qualitatively explain the difference in barrier heights and degradation of the barrier due to a certain process. In addition, this was also reported in Ref. [7], where the work functions of the metal and the semiconductor are determined by the process. The actual nature of the metal–semiconductor contact is not controllable and in fact may vary substantially from one process to another. The lowering of the Schottky barrier height is also

278

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

Fig. 3. Generated input voltages as a function of injection powers.

due to the small contact area, as this parameter is included in Eq. (1). To improve the RF response, the barrier height should be reduced, because a smaller barrier height gives better RF rectification due to the lower turn-on voltage [8]. 3.2. RF detection measurement Firstly, we investigated the voltages that can be generated by the signal generator in order to confirm the level of voltage at each input power level. In this measurement, an oscilloscope is connected directly to the signal generator. This is only performed at low frequencies of 1 and 10 MHz in order to confirm the relationship. It is confirmed that due to equipment characteristics/capability, the generated voltages cannot be increased when exceeding a certain input power. Fig. 3 shows the generated half-peak voltage, Vin(peak) , of the signal generator as a function of the input power, Pin . As shown in Fig. 3, we need to apply more than 5 dBm of input power in order to turn the diode on since the turn-on voltage of diode is about 1.1 V. The RF power detecting characteristics of the Schottky diodes, which had a 20 µm × 20 µm contact area, were measured by directly injecting RF power through the GSG CPW structure using a Cascade Infinity-150 microprober. We assembled a simple measurement setup as shown in Fig. 1(a). In this setup, an oscilloscope (model: Tektronics TDS 3054C) with an internal input resistance, Rosc , of 1 M and an internal capacitance, Cosc , of 10 pF is connected at the output side. Fig. 4(a) shows the average rectified voltages, Vout(peak) , as a function of the input voltages. It can be seen that the rectified output voltages are only obtainable when the input voltages slightly exceed the turn-on voltage of the diodes. An example of a measured rectified output voltage waveform is shown in Fig. 4(b). It is clearly seen that a good DC output voltage waveform is produced and good rectification is obtained. We also added a resistor, Radd , of 1 k and a capacitor, Cadd , of 2.2 nF in parallel with the oscilloscope in order to lower the total load to 1 k. Fig. 5(a) and (b) show the rectified output voltages as a function of the input voltages and an example of measured waveforms, respectively. Quadratic rises are observed in both Figs. 4(a) and 5(a). It can be seen that the diode is no longer in the square law operation region and moves into the linear region where the output voltage is proportional to the input voltage [14]. Here, also a good DC output voltage waveform is produced and good rectification is obtained. Since stable DC output

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

279

a

b

Fig. 4. (a) Rectified output voltages as a function of input voltages and (b) an example of measured rectified output voltage waveforms (Rosc = 1 M, Cosc = 10 pF).

voltages are produced at the output side, the DC current can be directly derived from the measured DC voltages and applied load. Fig. 6 shows the rectified output voltages as a function of frequency at input power levels of 5 dBm, 10 dBm, 15 dBm and 20 dBm for device B. The cut-off frequency can be seen in the test frequency range of 10 MHz to 30 GHz where it increases with the level of input power. Fig. 7 shows the cut-off frequencies as a function of input power for device A and device B. Here, the cut-off frequency is the frequency when the DC output voltage is zero. As shown in Fig. 7, the cut-off frequency increases with the input power. The cut-off frequencies of device A are higher than those

280

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

a

b

Fig. 5. (a) Rectified output voltages as a function of input voltages and (b) an example of measured rectified output voltage waveforms (Radd = 1 k, Cadd = 2.2 nF).

of device B at each test input power levels. It can be simply stated that a shorter CPW is required for creating a high cut-off frequency. It is noticed here that a different length of CPW yields a different fundamental resonant frequency, based on our separate pre-simulation results on the CPW analysis using a commercial Electromagnetic Sonnet Suites simulator (full wave EM simulation software). We can assume that a Schottky diode obeys the same expressions as a pn junction diode. The junction capacitance, Cj , is proportional to the contact area, A, and the series resistance including the

√

contact resistance is proportional to 1/ A. Since the cut-off frequency is proportional to 1/RS Cj , we

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

281

Fig. 6. Rectified output voltages as a function of the frequencies at different input power levels.

Fig. 7. Cut-off frequencies as a function of input power.

√

can conclude that the cut-off frequency is proportional to 1/ A [8]. Therefore, to achieve a higher cutoff frequency, the Schottky contact area also needs to be reduced. From Fig. 7, cut-off frequencies of around 20 GHz for device A and around 3 GHz for device B at an input power of 20 dBm are observed.

282

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

Fig. 8. Schematic of direct integration between the Schottky diode and dipole antenna.

From the cut-off frequency equation fC =

1 2π RS Cj

(2)

the junction capacitance, Cj , was calculated to be 6.77 fF for device A and 58.35 fF for device B (with R = 1.176 k and fc = 20 GHz for device A, R = 909.1  and fc = 3 GHz for device B) which is larger than 2.01 fF (the theoretical value). It is not unusual to find a discrepancy between capacitances obtained from an experiment and from theoretical considerations for two reasons. First, no capacitor can provide the ideal geometry of a parallel plate. The other reason is the parasitic capacitances and possibly also a parasitic resistance blow up effect as proposed by Greenberg et al. and Arora et al. [15–17]. 4. Possible direct integration with a dipole antenna to form a RF power detector Fig. 8 shows the possible direct connection between the Schottky diode and antenna. On the basis of the design and the Schottky diode characteristics obtained, presented in Sections 2 and 3, it is expected that direct integration with the dipole antenna via a CPW transmission line can be achieved without any matching circuit. For this purpose the behaviour of the antenna and the Schottky diode have to be modelled at and around the operating frequency. Since all the components are designed on the same substrate, a planar fabrication technique can guarantee excellent mechanical tolerances as well as tuning-free design. The measured results show the usefulness of the proposed antenna configuration and the effectiveness of uniplanar technology in terms of both performance and cost. Therefore, this integration can contribute to novel RF power detector. In this section, we present the RF characteristics of a planar dipole antenna facilitated with a CPW structure on a semi-insulated high dielectric constant substrate of GaAs, for which εr = 12.9 in the millimeter-wave region. 4.1. The coplanar waveguide and antenna design As the popularity of the CPW transmission line has increased significantly in recent years, antenna elements that are suitable for CPW feed configurations have also become important. In the light of this, a design guideline for a CPW-fed antenna is presented herein. Fig. 9(a) shows the proposed transition from the CPW to the dipole antenna configuration within the same dielectric substrate. The fabricated dipole antenna is shown in Fig. 9(b). The dimensions of the gap, a, and width, b, for the CPW calculated on the basis of Wheeler’s equation [12] were also chosen to be 60 µm and 90 µm, respectively, for the characteristic impedance, Z0 , of 50 . The thickness, h, is 625 µm. This CPW structure is also suited to the dimension of the Cascade GSG microprober used in the measurement process. The length of the CPW was chosen to be 120 µm. It is also noticed here that a longer CPW length may affect

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

283

Fig. 9. (a) Schematic dipole antenna structure and (b) fabricated dipole antenna.

the resonant frequency of the dipole antenna, according to our separate pre-simulation results from the integrated CPW and dipole antenna analysis using the commercial Electromagnetic Sonnet Suites simulator. Therefore, a shorter CPW length is chosen in order to allow us to omit a matching circuit during the direct integration. Besides that, it can also reduce the specific area, making it a low cost antenna. The metals of the antenna are Cr/Au with a thickness of 10/60 nm. Basically, Cr, Ni and Au are used to make links between the antenna and Schottky contact because most Schottky contacts for GaAs-based devices are made using Cr/Au or Ni/Au combinations. Thus, this can remove a step in the fabrication when the integration of the antenna and Schottky diode is performed. In this study, an S-parameter reflection measurement on the fabricated dipole antenna in the microwave region at room temperature was performed to characterize the return loss of the CPW facilitated planar dipole antenna. An HP8722S Network Analyzer facilitated with a Cascade GSG microprober Infinity-150 is used in this measurement. 4.2. Simulation and experimental results for the dipole antenna In our previous study, the dipole antenna structure was investigated by varying the length, L, and width, W , of the antenna [18] and those designed antennas are working in the superhigh frequency (SHF) band. An example of typical return loss characteristics for a fabricated antenna is shown in Fig. 10(a) and (b). The length and width of the antenna structure as shown in Fig. 10(a) are 6 mm and 100 µm, respectively. From the graph, there is almost 2% difference in frequency bandwidth at −10 dB between the measured and simulated responses. Here, it can be seen that a high return loss magnitude down to −25 dB at 9 GHz was obtained experimentally. As seen in Fig. 10(b), for the antenna length of 3 mm and width of 60 µm, there is almost 4% difference in frequency bandwidth at −10 dB between the measured and simulated responses. Here, it can also be clearly seen that very high return loss magnitude down to −46 dB at 18 GHz was obtained experimentally. The positive magnitude of the return loss characteristics of the fundamental (first-harmonic) resonant frequency as a function of the antenna width for various lengths is shown in Fig. 11. The high return loss obtained makes the antenna more meaningful and also improves the direct integration with other microwave devices like Schottky diodes. Consequently, it will reduce the quantity of reflected signal.

284

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

a

b

Fig. 10. Measured and simulated return loss for the CPW-fed dipole antenna. (a) L = 6 mm, W = 100 µm. (b) L = 3 mm, W = 90 µm.

Fig. 12(a) and (b) show the characteristics of the antenna as a function of the resonant frequency for various dimensions of length and width, respectively. As expected, it can be seen in Fig. 12(a) that the fundamental resonant frequency shifts to higher frequency when the length of the antenna decreases. Interestingly, as shown in Fig. 12(b), the resonant frequencies of the antenna are almost unchanged with the variation of the antenna’s width. In a nutshell, the changing width of the antenna does not influence the position of the resonant frequency much. From Fig. 12(b), it was seen that

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

285

Fig. 11. Magnitude of the return loss at the fundamental resonant frequency as a function of the antenna width for various lengths (the magnitude was changed to be positive).

the width dependence only has an effect on the magnitude of the return loss, as expected. To get a higher magnitude of return loss, a greater width of the dipole antenna should be adopted. With this information, we can design a dipole antenna with the desired resonant frequency by selecting the length and width of the antenna to match with the Schottky diode. The preliminary results presented in this paper represent a breakthrough for the direct integration for real application. However, we expect the output voltage to be smaller due to propagation loss and coupling loss, causing transmission, and hence rectification by the Schottky diode. There are two possible improvements that can be made in future in order to further improve our present devices. First, the cap layer (undoped GaAs) needs to be removed so that the ohmic resistance can be reduced by forming an ohmic contact directly to the n-AlGaAs layer. Another method is to redesign the metal contact of the input node in such a way that any possible channel is eliminated. 5. Conclusion In this paper, a preliminary investigation of the design, fabrication and RF characterization of a Schottky diode was performed and the feasibility for direct integration with the proposed planar dipole antenna was discussed. The cut-off frequencies of the fabricated Schottky diodes have been shown to be adequate for potentially damaging RF signals, since electromagnetic signals in the approximate frequency range of 200 MHz to 5 GHz are known to cause electronic damage in many systems. These preliminary results represent a breakthrough as regards direct on-chip integration technology, towards the realization of a ubiquitous network society. Acknowledgements The authors wish to extend their thanks for the support provided by the Ibnu Sina Institute, Universiti Teknologi Malaysia, Malaysia, and Nano-Optoelectronics Laboratory, Universiti Sains Malaysia, Malaysia. This work was supported by the Ministry of Science, Technology and Innovation under Science-Fund Grant 03-01-06-SF0277, Malaysia government. We wish to thank our colleagues

286

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

a

b

Fig. 12. Fundamental resonant frequency for various (a) lengths and (b) widths.

for useful discussions, particularly Assoc. Prof. Dr. Azlan Abdul Aziz, Assoc. Prof. Dr. Roslan Mat Hashim at Universiti Sains Malaysia, Malaysia, and Assoc. Prof. Dr. Zulkafli Othman at Universiti Teknologi Malaysia, Malaysia. References [1] [2] [3] [4]

M. Abrams, IEEE Spectr. 40 (2003) 24–30. T. Mimura, S. Hiyamizu, T. Fujii, K. Nanbu, Japan. J. Appl. Phys. 19 (1980) 225. M.S. Shur, W.C. Peatman, H. Park, W. Grimm, M. Hurt, Solid State Electron. 38 (1995) 1727. http://www.rciqe.hokudai.ac.jp 2005.

F. Mustafa et al. / Superlattices and Microstructures 47 (2010) 274–287

287

[5] B.L. Sharma, Metal–Semiconductor Schottky Barrier Junctions and their Applications, Plenum Press, New York, London, 1984, p. 20. [6] J.F. Luy, K.M. Strohm, J. Buechier, P. Russer, IEE Proc. 139 (1992) 209–216. [7] V. Milanovic, M. Gaitan, J.C. Marshall, M.E. Zaghloul, IEEE Trans. Electron Devices 43 (1996) 2210–2214. [8] W. Jeon, T.M. Firestone, J.C. Rodgers, J. Melngailis, Solid State Electron. 48 (2004) 2089–2093. [9] A.M. Hashim, S. Kasai, T. Hashizume, H. Hasegawa, Microelectron. J. 38 (2007) 1263–1267. [10] A.M. Hashim, S. Kasai, T. Hashizume, H. Hasegawa, Microelectron. J. 38 (2007) 1268–1272. [11] A.M. Hashim, S. Kasai, H. Hasegawa, Superlattices Microstruct. 44 (2008) 754–760. [12] C.P. Wen, IEEE Trans. Microw. Theory Tech. MTT-17 (1969) 1087–1088. [13] D.H. Zhang, Mater. Sci. Eng. B 60 (1999) 189–193. [14] Application Notes AN 64-1C, Hewlett Packard Co., 1996. [15] D.R. Greenberg, J.A. Alamo, IEEE Trans. Electron Devices 41 (1994) 1334–1339. [16] V.K. Arora, M.B. Das, Semicond. Sci. Technol. 5 (1990) 967–973. [17] V.K. Arora, M.B. Das, Electron. Lett. 25 (1989) 820. [18] F. Mustafa, N.I. Rusli, A.M. Hashim, 2nd Asia Modelling & Simulation AMS2008, 2008, pp. 994–996.