2.1

Silicon-based high-speed flexible MOSFETs

Si is the most widely used conventional material for electronics, as it is abundant and has descent intrinsic properties required for high-performance electronics. Although it is less utilized for high-frequency applications, Si can be tweaked to boost intrinsic properties to be useful as a high-frequency device [6]. For the Si based devices, further improvement of device speed implies significant technical and economic advantages. While the mobility of bulk Si can be routinely enhanced using strain techniques, implementing these techniques into transferrable single-crystalline Si nanomembranes (NM) for fast flexible electronics has been challenging and not demonstrated. Here we demonstrate the combination of self-sustained strain sharing with strained-NM compatible doping techniques to create strained, print-transferrable Si

NMs [7]. The combination led us to demonstrate a new speed record of Si based flexible electronics without using aggressively scaled critical device dimensions. Figure 1(a) illustrates the strain sharing techniques. Creation of self-sustained strain in Si NMs started from growing 80 nm of undoped Si_0.795Ge_0.205 epitaxial layer on silicon-on-insulator (SOI) wafer with a thin-downed 48 nm top Si layer. A nearly symmetrical trilayer structure (Si/SiGe/Si) was formed by growing 46 nm of undoped Si on top of the SiGe. Due to the lattice mismatch between Si and SiGe, the SiGe layer in the trilayer structure was compressively strained to the Si in-plane lattice constant (mismatch strain, ε_m = -0.77%). The trilayer NM was released by selective removal of the SiO₂ (BOX) layer of the SOI. During the release, strain sharing occured between the SiGe and Si layers. To demonstrate the effectiveness of the self-sustained straining approach, Figure 1(b)(i) shows an X-ray diffraction (XRD) off-axis reciprocal-space map (RSM) around the (044) reflection for the as-grown trilayer structure. The RSM indicates that the SiGe layer was strained to the Si lattice constant; the Si and SiGe peaks lied along the same vertical line. This result confirms that there was no plastic relaxation of the mismatch strain in the alloy before release of the trilayer. To realize high device speed, the NMs must be properly doped. We designed an alternative approach by applying the ion implantation and anneal processes to an unstrained Si layer on SOI as shown in Figure 2(a) and (b). The doping and thinning down procedures have some effect on the crystallinity of the as-grown Si/SiGe/Si trilayers by a general broadening of the diffraction peaks in Figure 2(c).

Figure 1.

(a) Process flow to implement the strain sharing using Si/SiGe/Si epitaxial trilayer structure. (b) X-ray diffraction before and after release of the trilayer NM. i, On-axis line scan around the (004) reflection before and after release of the trilayer NM. ii, Off-axis RSM around the (044) reflection for the as-grown trilayer structure. Adapted from [7].

Figure 2.

(a) Doping of strained trilayer structure. (b) Image of doped strained NM after transferring to a plastic substrate. (c) XRD of as-grown ion-implanted trilayer NM. i. Off-axis RSM around the (044) reflection. ii. On-axis line scans around the (004) reflection comparing the doped and undoped as-grown trilayer NMs. Adapted from [7].

Figure 3(a) shows the transfer curve and transconductance (g_m) versus gate voltage (V_g) of a strained-channel TFT along with the results measured from the unstrained-channel TFT for comparison. The highest g_m for the strained TFTs was 386 μS. Comparing to the unstrained reference TFTs with identical dimensions, the 47.3% enhancement of the peak g_m values in the strained TFT was observed and is mainly ascribed to mobility enhancement, which is due to the introduction of the tensile strain in the Si channel of the trilayer. The g_m and mobility enhancement ratios were consistent with the expected value. Figure 3(b) plots the g_m versus drain current. Figure 3(c) and (d) show the measured current gain (H₂₁) and power gain (G_max) of the strained TFT, indicating that the cut-off frequency (f_T) is 5.1 GHz and the maximum oscillation frequency (f_max) is 15.1 GHz.

Figure 3.

(a) Transfer curves and transconductance (g_m) curves of unstrained and strained devices (V_ds = 50mV). (b) g_m is plotted as a function of drain current. Point A indicates the peak g_m where peak f_T/f_max were measured. Point B is where 3.5GHz f_max can be obtained. (c) Current gain (H₂₁) and power gain (G_max) as a function of frequency of unstrained and strained devices (V_g=4V, V_ds=5V). (d) f_T and f_max of strained devices as a function of drain bias under fixed gate bias (V_g=4V). Adapted from [7].

2.2

GaAs-based high-speed flexible HBTs

GaAs based HBTs and Schottky diodes have been fabricated for flexible RF applications on flexible cellulose nanofiber (CNF) substrate [3]. The fabrication of GaAs HBT started with conventional non-self-aligned GaAs HBT fabrication process on GaAs wafer with AlGaAs sacrificial layer. A thick layer of photoresist (~ 7 μm) was used as anchor for holding and protecting GaAs HBTs during undercut process, in which the device was undercut etched in diluted hydrofluoric (HF) solution (1:100) for 3 hours. After the fully undercut, the GaAs HBT was transferred onto a temporary substrate with a thin layer of polyimide (PI) and poly(methyl methacrylate) (PMMA) on a Si substrate. Then, another layer of PI was spin casted on the GaAs HBT on temporary substrate as encapsulating layer. After via opening, ground-signal-ground RF pads were deposited on the device for characterization later. The PMMA sacrificial layer of PI encapsulated device was undercut etched in hot acetone. Following fully dissolving PMMA, the device was picked up by PDMS elastomer stamp and transferred onto the CNF substrate which had a layer of SU-8 as adhesive. Figure 4(a) shows GaAs HBT on CNF substrate on a tree leaf. DC characterization of the GaAs HBT on CNF substrate shows highest gain (β) of 14.49 when V_BE was biased at 1.86 V, as shown in Figure 4(b). Scattering (S-) parameters measurement of GaAs HBT was conducted with bias of V_C = 2 V and I_B = 2 mA to characterize RF performance of GaAs HBT. After de-embedding the parasitic effects using short and open pattern on same substrate, the current gain (H₂₁) and power gain (G_MAX) of the GaAs HBT are demonstrated in Figure 4(c). The GaAs HBT can achieve cut-off frequency (f_T) of 37.5 GHz and maximum oscillation frequency (f_max) of 6.9 GHz. The relatively low f_max is caused by the non-self-aligned structure. The 2 μm emitter to base spacing resulted from high base resistance, which diminished the f_max. This GaAs HBT is capable of working in GSM and Wi-Fi channel, which is in the range of 800 to 2500 MHz. Schottky diodes based on GaAs were also fabricated in similar way as the GaAs HBT. The current-voltage measurement in Figure 4(d) shows good ideality factor of 1.058 and low turn-on voltage of 0.7 V. S-parameters of GaAs Schottky diode demonstrated in Figure 4(e) and (f) show the RF performance of the device at forward current bias of 10 mA and reverse voltage bias of -0.5 V, respectively. The insertion loss (S₂₁) was only -1 dB at 20 GHz under forward bias and -2 dB at 4.3 GHz, which presents potential for microwave switching applications. A full bridge rectifier using GaAs Schottky diodes and metal-insulator-metal (MIM) capacitor was fabricated on CNF substrate, as shown in Figure 4(g). The rectification behavior was measured using an input RF signal at 5.8 GHz. With increasing RF input power, the DC output power could achieve 2.43 mW output power with input power of 21 dBm as shown in Figure 3(h). Among the compound semiconductors, GaAs is the most widely used material in electronics as more than 85% of today’s consumer electronics with wireless communication capabilities, such as cell phones and mobile tablets, employ GaAs-based microwave devices for their superior high-frequency operation and power handling capabilities. The thin film GaAs devices obtained using this method would not only save the expensive material, but also allow mobile devices in flexible formats.

Figure 4.

(a) Photograph of an array of HBTs on CNF substrate on a tree leaf. (b) Gummel plot and gain (β) of GaAs HBT (inset, optical image of a GaAs HBT). (c) Current gain and power gain as a function of frequency. (d) Current-Voltage curve of GaAs Schottky diode (inset, optical image of a GaAs Schottky diode). (e) S-parameters (S₁₁ and S₂₁) of GaAs Schottky diode under forward current bias of 10 mA. (f) S-parameters (S₁₁ and S₂₁) of GaAs Schottky diode under reverse voltage bias of -0.5 V. (g) Optical image of a full bridge rectifier on CNF substrate. (h) DC output power of the rectifier as a function of RF input power. Adapted from [9].

2.3

GaN-based high-speed flexible HEMTs

While Si- or GaAs-based transistors may be the suitable choice for high speed flexible applications, GaN-based high electron mobility transistors (HEMTs) have achieved one-order magnitude higher power density and power conversion efficiency compared to the competing technologies. Considering the low thermal dissipation coefficient of the general flexible materials, GaN based transistors with best conversion efficiency are the best candidate for the high power flexible amplifiers. To effectively embed the thin film GaN HEMTs on the flexible substrate, as shown in Figure 5a, we have transferred the thin film AlGaN/GaN HEMTs on a plastic substrate by selectively removing the growth substrate through XeF₂ gas phase etching. Figure 5b shows the large scale, highly transparent, 8×8 mm GaN NM transferred on a plastic substrate. The pre-fabricated GaN HEMTs on the transferred GaN membrane, as presented in Figure 5c, demonstrated good DC properties, as shown in Figure 5c and Figure 5d, and measured with high frequency response (f_T/f_max = 50/115 GHz). Large scale flexible GaN nanomembrane features good thermal properties on the flexible film [8], and the same technique can be utilized to achieve the stretchable GaN HEMTs and hence the stretchable power amplifiers. In Figure 6a, the stretchable power transistors are combined with the pre-defined circular-shaped GaN NM and low-loss stretchable coplanar transmission line. The embedded GaN HEMTs transferred on Dupont plastic tape demonstrated high stretchability under applied strain. The stretchable/flexible GaN HEMTs demonstrated here can form the route to design capabilities of next generation high performance flexible electronics for a variety of applications.

Figure 5.

High speed transparent and flexible AlGaN/GaN HEMTs on PET film. (a) Schematic of the flexible AlGaN HEMTs transferred on PET film. (b) Optical image of transferred 8mm×8mm GaN nanomembrane on PET film. (c) AlGaN/GaN HEMTs with gate length 400nm. DC characteristics of (a) Id-Vg, (b) Id-Vd, and (c) RF characteristics including h₂₁ and MAG. Adapted from [8].

Figure 6.

High Speed stretchable AlGaN/GaN HEMTs with coplanar transmission. (a) Schematic of the stretchable AlGaN HEMTs (b) Stretchable AlGaN/GaN under 0% and 20% strain.

3.1

Flexible capacitors, inductors and filters

RF circuits use spiral inductors and metal-insulator-metal (MIM) capacitors as matching components between RF blocks. Like conventional CMOS and MMIC technology, the passive components such as spiral inductors and MIM capacitors need to have their own library to adapt with various dielectric constants of flexible substrates. In conjunction with flexible active devices, many authors researched both robust and mechanically bendable, high frequency passive components such as inductors and capacitors. Numerous types of filters with various combinations of flexible inductors and capacitors were also demonstrated for a more complex matching. The first inductors and capacitors demonstrated on flexible substrates could achieve high-frequency under bending conditions [9]. However, the self-resonance frequency (f_res) of these inductors and capacitors were limited by 9.1 GHz and 13.5 GHz, respectively. Optimizing the materials and fabrication processes, enhanced versions of such passive devices were demonstrated with capabilities reaching X-band (8 to 12 GHz) operation, as shown in Figure 7 [10]. Finally, passive components that could reach Ku-band (12 to 18 GHz) operation, including filters were demonstrated, as presented in Figure 8(a)-(d) with optimized designd and fabrication processes [11]. The self-resonance frequency of inductors and capacitors optimized here showed 19.1 GHz and 25.4 GHz, respectively. It also combined its high f_res lumped flexible inductors and capacitors to make band pass filters in Figure 8(e) and notch filters in Figure 8(f). These filters are represented by optical image in Figure 8(g). The device is ultrathin that it can be laminated on a fingernail as demonstrated in Figure 8(h). The flexible series and parallel resonance lumped circuits can be implemented with high speed active components for microwave integrated circuit.

Figure 7.

Measured 4.5 turns spiral inductors with 4.5 turns, 20 μm line width, and 4 μm line space. (a) L values and (b) Q values as a function of frequency under flat and various bending states. Measured (c) capacitance values and (d) Q values of an 88 × 88 μm² MIM capacitor as a function of frequency under flat and bending conditions. Adapted from [10].

Figure 8.

(a) Inductance against frequency and (b) Q factor of a 2.5 turn spiral inductor. (c) Capacitance against frequency and (d) Q factor of a 40 µm × 40 µm large MIM capacitor. S-parameters plotted against frequency for (e) a 4.8 GHz band pass filter and (f) 4.5 GHz notch filter. (g) Optical microscopy image of the flexible low pass filter. (h) Ultrathin passive devices coated on a fingernail. Adapted from [11].

3.2

Stretchable transmission lines

Current technologies for stretchable electrodes have been utilizing different types of serpentine designs to enhance intimate fabrication of wearable electronics onto skin [12,13]. However, studies on those stretchable interconnects were limited to designs primarily for electrical operation at direct current (DC) signals or alternating current (AC) signals at very low frequency that is typically less than 1 GHz. As the working frequency of current wireless communication system increases to higher frequency (RF) level (i.e. multi-gigahertz), it is necessary to develop essential elements including active and passive elements of microwave circuit as a form of soft, flexible, or stretchable component which is still in compliance with conventional microwave circuit technology. Therefore, underlying studies on passive elements such as stretchable RF transmission lines are mandatory and innovative design consideration becomes crucial.

The stretchable RF transmission line can have unique twisted-pair design integrated into serpentine microstructure [14]. It can minimize electromagnetic interference from surrounding area, such that the transmission line is minimally affected by the environment. The new design enables the RF electrodes to be utilized in thin-film bioelectronics, such as the epidermal electronic system (EES) [15]. Two fundamental RF characteristics of the scattering (S-) parameters determine the performance of the transmission lines: insertion loss representing S21 is the loss of power that results from an inserted device; return loss representing S11 is the loss of power in the signal returned or reflected. Figure 9(a) illustrates a unique design-concept of stretchable RF transmission line that is able to work at microwave frequencies with significantly low level of loss by integrating twisted-pair geometry into the serpentine shaped structures. Originally the design is inspired by the twisted-pair cabling from the telephone cables where the two conducting wires are twisted together. The balanced cable is well-known to minimize electromagnetic interference (EMI) from external surroundings. Figure 9(b) and (c) show the experimental results of the fabricated twisted-pair-based stretchable transmission line. For each line, the width was fixed to 25 µm and the thickness, via-hole size, and dielectric spacer thickness were optimized to be 1 µm, 150 µm², and 5 µm, respectively. Also, optical microscopy images of the transmission lines at different elongation are shown in Figure 10(b). The twisted-pair-based transmission lines presented remarkable performance with low insertion loss (about -1.1 dB at 40 GHz) and high return loss up to 40 GHz, which are attributed to the excellent confinement of the electromagnetic fields in the structure as well. To examine the effects during elongation, the transmission line was measured under different elongations (0%, 20%, 25%, and 35%) as shown in Figure 10(c). Almost negligible increases in insertion loss were observed with negligible performance change in return loss characteristics were observed as well.

Figure 9.

(a) The concept of twisted-pair-based stretchable transmission line. (b) Optical microscopy image of the stretchable transmission line with 0%, 20%, 25%, and 35% elongation (c) Measured S₂₁ and S₁₁ of the stretchable transmission line with two turns of serpentine at 0%, 20%, 25%, and 35% elongation plotted against frequency. (d) Optical microscopy image of the twisted-pair-based stretchable low-pass filter (e) S-parameters of the stretchable low-pass filter at 0%, 20%, 25%, and 35% elongation plotted against frequency. Adapted from [14].

In addition, the RF microwave filters are used to attenuate or transmit signals at certain frequency bands in microwave circuit. As an example of the application of the new design of RF transmission line, the low-pass filter presented in Figure 10(d) and the measured parameters are presented in Figure 10(e). The filter performed a wide band low-pass characteristic where the 3 dB cut-off frequency was 9.9 GHz, with a relatively flat band and low insertion loss between 11.6 GHz (-3.5 dB) and 15 GHz (-3.7 dB). The performance of the filter is slightly changed under stretched (0%, 20%, 25%, and 35% elongations) conditions. However, the capability to generate stretchable filters working at high frequency indicate potential of the new design of transmission line for stretchable microwave integrated circuits. Furthermore, the twisted-pair geometry can minimize interference with external noise, which would allow its operation on human skin. The design and fabrication represented in this work for stretchable transmission lines that operate at microwave frequencies are applicable in EES demanding wireless communication capabilities. Furthermore, stretchable microwave filters that is based on the twisted-pair structure promises potential of the twisted-pair-based transmission lines as passive components. This type of new transmission line is also suitable for high-speed digital circuits which requires minimized EMI from external environment. Integration of stretchable transmission lines with active devices, such as transistors and diodes to construct microwave integrated circuits remains one of the future challenges.