Large Scale Fabrication of Graphene Based Nano‐Electromechanical (NEM) Contact Switches with Sub‐0.5 Volt Actuation

Nano‐electromechanical (NEM) contact switches are extensively studied to suppress the limitations of conventional complementary metal‐oxide‐semiconductor (CMOS) transistors. The attributes of NEM contact switches includes reduced power consumption, reduced off‐state leakage current, and increased on‐state current and sub‐thermal switching. However, unacceptably high pull‐in voltage and low contact lifetime posed a significant challenge for the use of NEM contact switches in energy efficient CMOS applications. Here, graphene‐based electro‐statically actuated NEM contact switches with ultra‐low pull‐in voltage and significant improvement in the contact lifetime are demonstrated. This is achieved by using the graphene on gold electrode as a contact material. The graphene NEM contact switches with graphene as a contact material exhibit an ultra‐low pull‐in voltage of < 0.5 V and high contact lifetime of more than 1.5× 106 cycles. The switches also show an excellent switching performance with high on/off ratio of ≈108, an extremely low off‐state current of ≈100 fA, and small hysteresis window of < 0.1 V.


Introduction
The complementary metal-oxide-semiconductor (CMOS) -based silicon industry is continually demanding energy efficient performance and low power utilization in CMOS devices for its new era of applications, such as the Internet of Things (IoT) and smart dust. [1,2] As a result, the Nano-electromechanical (NEM) contact switches emerge as one of the attractive alternatives to CMOS devices that suffer from their inherent limitations such as high leakage current and limited subthreshold swing of 60 mV dec -1 DOI: 10.1002/apxr.202200050 at room temperature. [3,4] Owing to its mechanical degree of freedom in switching mechanism, NEM contact switches can attain the ideal switching characteristics in terms of steep sub-threshold swing, quasi-zero off-state leakage current, thermally independent switching and ultralow power consumption. [5][6][7][8][9] Furthermore, they exhibit minimal degradation in the performance even in the harsh environments, such as high radiation fields, high temperatures, and in the external electric fields. [10][11][12] Because of its excellent device performance, NEM contact switches can be suitable for new era of CMOS applications. However, NEM contact switches generally suffer from two main parameters, high pull-in voltage due to the limitations in the scalability of conventional 3D bulk materials and lack of switching reliability due to contact surface degradation caused by repeatable physical contact in mechanical switching. [13,14] The contact degradation leads to several forms of damages including wear, fraction, creep, electromigration, delamination, and permanent stiction. [15][16][17] Lately, a vast range of new methods and materials in the structural designs of NEM contact switches have been implemented to improve their pull-in voltage and contact lifetime. [18][19][20] For example, mechanical switches based on the silicon carbide and ruthenium has shown improved switching cycles. [21,22] On the other hand, newly emerging nano materials, including carbon nano tubes (CNT), [23] nano wires, [24] transition-metal dichalcogenides, [25] and graphene [26] have been studied to improve the performance of the NEM contact switches. Among the various nano materials proposed, graphene, an atomic layer of carbon atoms tightly arranged into a 2D honeycomb lattice, promise an encouraging role in the field of NEM contact switches. [27] Attributes of the graphene such as excellent electrical and mechanical property with high Young's modulus and tensile strength, exceptionally high electrical and thermal conductivities makes it as a suitable material for ideal NEM contact switch. [28][29][30] In addition, various carbon-based nano materials are also employed as the contact material to improve the contact lifetime of the NEM contact switch. For instance, the suspended CNT beam with diamond-like carbon as contact, [31] vertically aligned CNTs with CNT as contact layer, [32] multilayer graphene with multilayer graphene as contact, [33] nano crystalline graphene (NCG) with NCG as contact material [34] and graphite-to-graphite contact, [35] have been studied earlier for electromechanical contact devices, inertial switches [32] and charge transfer based gas sensor applications. [36] However, this new generation materials based NEM contact switches were also failed to offer a high contact lifetime due to contact degradation. Furthermore, the graphene based NEM contact switches failed to offer a long contact lifetime-an essential parameter for reliable NEM contact switches-due to the irreversible stiction. Jian et al. reported that graphene-based two terminal NEM contact switch fabricated with gold (Au) as the contact material and achieved a low pull-in voltage of < 2 V. [37] Nevertheless, the device performance was limited to few switching cycles due to the irreversible stiction between the graphene beam and Au electrode. Wenzhen et al. There are studies oreported on the graphene-based NEM contact switches with chromium oxide (Cr 2 O 3 ) as a contact material with surface trenches; however, device failed to offer low pull-in voltage as well as contact lifetime due to the atomically uncontrolled nature of the contact surface. [38,39] Graphene was used as the contact material in graphene-based NEM contact switch to avoid the permanent stiction. Van et al. reported a three terminal graphene NEM contact switch with graphene as the contact material but the switch failed within few cycles due to poor geometrical design of the device. [40] Moreover, the switch was reported with high pullin voltage of > 7 V. Graphene was successfully utilized as a contact material in nickel based micro-electromechanical switches with long contact lifetime but failed to scale down the pull-in voltage. [41] As a result of the inability to reduce the thickness of the suspended Ni beam, the reported pull-in voltage is > 70 V.
In this paper, we report the experimental demonstration of a graphene-based two terminal NEM contact switches with ultralow pull-in voltage suitable for new era of energy efficient lowpower CMOS applications. The NEM contact switches were fabricated using the chemical vapor deposition (CVD) grown graphene as suspended beam as well as the contact material and referred as G-G NEM contact switches. The irreversible stiction was suppressed by utilizing the weak van der Waals (vdW) interaction between the suspended graphene beam and the contact graphene layer. The two terminal G-G NEM contact switches exhibit superior switching characteristics with a ultra-low pull-in voltage of ≈0.45 V, sub threshold slope as small as < 10 mV dec -1 , stable pull-out at 0.4 V, high on/off ratio of ≈ 10 8 and a small hysteresis window of < 0.1 V. Figure 1a shows the schematic illustration of the G-G NEM contact switches. The suspended graphene beam was used as the active element of the NEM contact switch and graphene covered top actuation electrode was used as the fixed element. The extented zoom-in view in Figure 1a illustrates the graphene covered top actuation electrode. The extensive fabrication process is given in the Supporting Information Sections S1 and S2. Figure 1b demonstrates the large scale array of the fabricated devices on Si/SiO 2 substrate with dimensions (Length ×Width) of 10×10 mm. Here, the CVD grown graphene was also used as contact material. By exploiting the weak van der Waals interac-tion between the graphene layers, long endurance highly reliable NEM contact switches were achieved. Figure 1c shows the optical image of a device after transferred the contact graphene layer. The contact graphene layer was transferred on top of the fabricated sample using wet chemical based transfer method. (Refer Supporting Information Section S3 for more information on graphene transfer method).The Raman spectroscopy with Nddoped Y-Al-garnet Laser as the excitation source ( = 532 nm) was employed to verify the existence of the graphene layer after the transfer and strain analysis of the suspended graphene beam. (See Supporting Information Section S4 for more on Raman analysis of the graphene on Si/SiO 2 substrate). The Raman spectroscopy of the graphene layer on Si/SiO 2 substrate relative to the suspended graphene beam is shown in Figure 1d. The D peak observed at 1350 cm −1 indicates the existence of the lattice defects in the CVD grown graphene. [42] The G and 2D peaks were observed at 1587.5 cm −1 and 2681.4 cm −1 , respectively for the graphene on the substrate. Whereas the G and 2D peaks were observed at 1582.6 cm −1 and 2676.8 cm −1 , respectively for the suspended graphene beam of length of 1 μm and width of 0.5 μm. Both the G and 2D peak were red shifted by ≈ 5 cm −1 and ≈ 5 cm −1 , respectively, for the suspended graphene beam, which indicated the increased strain in the suspended graphene beam. [43] The calculated strain and pre-tension was about ≈0.3% and ≈9 N m −1 , respectively. It is worth to note that, the single layer graphene had weak built strain after transferring the graphene on SiO 2 surface. (See Supporting Information Section S5 for more on strain analysis of suspended graphene beam). The G-G NEM contact switches were fabricated with three different air gap thickness of 30 nm, 60 nm, and 90 nm, respectively. This is achieved by varying the thickness of the sacrificial SiO 2 thickness. (See Supporting Information Section S6 for more on AFM analysis of the thickness of sacrificial SiO 2 layer). Figure 1e shows the AFM image of the device taken after the sacrificial SiO 2 deposition of 30 nm. The line profiles obtained for the various thicknesses from AFM results are furnished in Figure 1f. Figure 1g shows the tilted view of the SEM image obtained for a device S1 after the electrical measurement.

Electrical Measurement Configuration of G-G NEM Contact Switches
All the electrical measurements were carried out using semiconductor device analyzer (Keithley-4200 SCS) with measure resolution of ≈10 aA. To avoid the moisture related failures of the switches, the measurement chamber was vacuumed to ≈10 −4 Pa. To confirm the conductivity of the suspended graphene beam, the drain current (Id) across the graphene beam was measured as a function of the applied drain voltage (Vd). To investigate the mechanical switching characteristics of the device, the two-point probe method was used. Subsequently, the measurements were carried out in the following order: 1) high-resolution sweeps by ramping the applied voltage (Va) with step size of 7 mV and monitoring the switching current (Is). 2) low-resolution fast-cycling measurements. During the low-resolution measurements, the applied bias voltage (Va) of 0.5 V was applied and the current was continually monitored with a compliance limit of 5 μA for on-state and voltage of 10 mV was applied to measure the current at off-state. All the measurements were carried out in vacuum at room temperature. In total, seven devices of G-G NEM contact switches were presented and all devices shown qualitatively similar switching behaviors except the failure modes. Table 1 lists the designed geometrical dimensions as well as the observed electrical characteristics of the devices. The electrical measurement results obtained from the device S1 are presented, unless otherwise noted.

G-G NEM Contact Switch Fabrication
First, the CVD grown graphene on Si/SiO 2 surface was used as the substrate (See Supporting Information Sections S1 and S2 for the detailed fabrication processes). Then graphene ribbon was fabricated by using the positive resist poly (methylmethacrylate) (PMMA), and then the pattern was defined by electron beam lithography (EBL) exposure. The unwanted graphene was removed using oxygen plasma based dry etching. The anchor metal electrode was fabricated using the positive bilayer resist of PMMA/methylmethacrylate (MMA) and the metal electrode (Cr:Au :: 5:85 nm) was deposited by using electron beam (EB) evaporation. The sacrificial SiO 2 was fabricated using PMMA/MMA bi-layer resist followed by EBL pattern and SiO 2 was deposited using EB evaporation. After fabricating the sacrificial SiO 2 layer, another layer of CVD graphene was transferred onto the sample using a wet chemical based transfer method (see Supporting Information Section S3 for more details on graphene transfer process). The transferred graphene layer was patterned using the PMMA, and the unwanted graphene was removed using oxygen plasma etching. The top actuation electrode was fabricated using the PMMA/MMA bi-layer resist, the metal electrode (Cr/Au: 5/85 nm) was deposited by using electron beam (EB) evaporation. Unwanted graphene remaining on the sample was removed by using oxygen plasma etching. The transferred CVD graphene layer bonded with Cr metal layer. The bonding between out-of-plane dangling C atoms in the graphene layer and metal atoms led to a strong Cr-C bond. The transferred graphene layer acted as an anti-stiction coating to reduce the mechanical failures related to the surface adhesion and other reliability issues. Finally, the sacrificial layer (SiO 2 ) between the bottom GNR and the top actuation electrode was etched in buffered hydrofluoric (BHF) acid (1:5) and dried in super-critical point dryer. This process led to suspension of GNR from bottom SiO 2 and air-gap between the suspended graphene beam and the actuation electrode.

Mechanical Performance Calculation
The mechanical restoring force of a doubly clamped graphene beam can be calculated theoretically using the following method. Suspended graphene can exhibit both linear (Hookean) and cubic (non-Hookean) restoring force behaviour. The Hookean limit applied for suspended graphene beam with small deflections and high pre-tension, while the non-Hookean limit applied for large deflections and small pre-tension. [57] The cross-over from linear regime to the cubic regime was decided by the absolute pre-tension and deflection of an individual suspended graphene beam. Here, the non-Hookean form was adopted to calculate the mechanical restoring force of the suspended graphene beam, detailed mechanical calculation methods are given in the ref. [58].
The van der Waals force between the two graphene layers can be calculated as where K is the spring constant of the doubly clamped beam, Θ is the dimensionless constant, is the Poisson ratio of graphene, A b is the area of the suspended beam, E is the Young's modulus of the graphene, t, L, and W are the thickness of the single layer graphene of 0.35 nm, length, and width of the suspended graphene beam, respectively. A c is the contact area of the switch (0.25 μm 2 ) and A H is the Hamaker's constant [59] of graphene taken to be 4.7 x 10 −19 J, Z is the interlayer distance between two graphene layers is taken to be 0.335 nm. [60] The mechanical restoring force of doubly clamped graphene beam calculated for 30 nm air gap (d) was ≈2.7 nN, whereas vdW force between graphene-graphene interfaces was ≈28 fN.

Finite Element Method (FEM) Simulations
In FEM simulations, COMSOL Multiphysics (5.6, COMSOL Inc., Burlington, MA, USA) package with MEMS module was used. MEMS module solves the electrostatic field with mechanical forces. In addition, MEMS module also allowed to couple these fields and simultaneously map the beam deflection, electric field distribution, charge density around the beam, and stress profile of the NEM contact switches in 3D mode. The electrostatic field in the air and in the graphene beam was governed by Poisson's equation -∇ · (ϵ∇V) = 0. The bias potential is applied to the top actuation electrode and the suspended graphene beam was grounded. All the other boundaries in the model are electrically isolated. In this simulation, a 3D analysis was done by solving coupled electromechanical equations. The NEM contact switch model was made up of seven layers. All the seven layers were encapsulated with an additional layer that was modeled as air. The material properties were imported from COMSOL Multiphysics materials library except for graphene, which was cited from the literature. In order to numerically solve the coupled model by FEM, the whole device structure was meshed with a user defined mesh. The different element size was used for each layer. The average element quality of the mesh was ≈0.7, where 1 is represented as an ideal and 0 is represented as degenerated mesh elements. Figure 2 shows the switching characteristics of the G-G NEM contact switch S1. The current-voltage (I-V) responses measured between the anchor to anchor electrodes and anchor to top actuation electrode before mechanical switching operation is shown in Figure 2a. It can be seen from Figure 2a, the suspended beam and the top actuation electrode were isolated electrically. The two point probe based switching measurement was conducted between the suspended graphene beam and graphene covered top actuation electrode as shown in Figure 1a. The measured switching current (Is) with respect to the applied voltage (Va) with current compliance of 5 μA in vacuum is shown in Figure 2b. During the forward sweep of the applied voltage Va, the contact between the graphene beam and the contact graphene (Gr/Au) occurred at 0.45 V (pull-in voltage) and the measured switching current is increased abruptly to the compliance value of 5 μA from very low off-state current floor of < 100 fA. At pull-in the suspended graphene beam made a contact with contact graphene/Au top actuation electrode. The region of the contact area increases as the applied voltage is increased, which in turn reduces the contact resistance. During the reverse sweep, the switching current instantly falls in to off-state current floor (completely turned off) at the applied voltage of 0.4 V (pull-out voltage). It is worth to mention that, at pull-out, on-state to off-state transition occurred in a single sweep voltage step of 7 mV. The measured hysteresis window of the switch is < 0.1 V. For the first 30 switching cycles, the deflection in the pull-in voltage is significantly very minimum of < 10 mV as illustrated in Figure 2c. To verify the steep and stable switching behavior, the switching current (Is) is measured for voltage steps with different values. The switching characteristics for applied voltage pulses as a function of time are shown in Figure 2d. It is observed that the current (Is) is only reached the compliance value for voltages higher than the pull-in voltage, for any values of Va less than pull-in voltage the switching current (Is) is remains in the off-state. This measurement is useful for detecting the deviation in pull-in and pull-out voltages and reveals the stability of a switch. The linear fit of switching slopes of the pull-in and pull-out transition is illustrated in Figure 2e,f, respectively. The steep switching slope (SS) for the first switching cycle from the Is-Va curves were determined as <10 mV dec -1 for the pull-in (SS pull-in) and < 10 mV dec -1 for pull-out (SS pull-out). The respective on/off ratio of the switching current is ≈10 9 orders of magnitude, which is higher than those of previously reported studies on graphene-based NEM contact switches. [34,37,38,40] (See Supporting Information S7 for more information on electrical failure analysis of G-G NEM contact switch).

Electrical Characterization of G-G NEM Contact Switches
The effect of the graphene layer at the contact interface is well evident from the switching characteristics of G-G NEM contact switch. Especially, the sharp transition in the pull-out curve is direct evidence of the reduced contact adhesion between suspended graphene beam and top actuation electrode. During the mechanical switching operations (on/off transitions) the contact graphene plays an important role. The contact graphene significantly reduces the physico-chemical degradation at the contacting interface owing to its stable mechanical and chemical properties of the graphene. [44] The exceptionally small variation in the pull-out voltage (≈10 mV) is obvious evidence for the wellcontrolled stiction at the graphene-to-graphene contact interface. Unlike the GNEM switches reported earlier, which failed to perform mechanically, the G-G NEM contact switches demonstrated with stable and sharp pull-in and pull-out switching slopes over numerous switching cycles. The result indicates that the G-G NEM contact switches were stiction free, and stiction was surpassed by the weak van der Waals interaction at the graphene-to-graphene contact. The switching behavior of GNEM switches with gold (Au/Cr 2 O 3 ) as contact material is given in Supporting Information Section S8 for comparison. However, the control device (Au/Cr 2 O 3 ) was failed within few switching cycles. The failure is mainly due to chemical bond formation between the carbon and gold atoms. [45]

Long-Term Contact Lifetime Measurements
The low resolution hot-switching experiment is carried out to verify the contact liftime of G-G NEM contact switch as well as the current carrying capabilities of graphene as a contact material. The deformability and low adhesion of the double-clamped graphene beam on the contact graphene were also investigated during the long-term contact switching operation. In this measurement, the switch was placed under continuous mechanical stress especially on the contact region of switch. These rapid cycling measurements also divulged the failure mechanism associated to the mechanical wear of the contact surfaces. Barring mechanical wear, other contact failure mechanisms such as surface oxidation or contamination can also occur during storage. [46] The contact reliability of G-G NEM contact switch was obtained as an outcome of the hot-switching measurements. Figure 3 illustrates the long-term contact lifetime measurements showing on-state and off-state switching current for each cycle. The electrical measurements were performed by applying static voltage between the suspended graphene beam and actuation electrode, defined as hot switching condition, with bias voltage of 0.5 V for on-state with current compliance limit of 5 μA and 10 mV for off-state. The switch continued to cycle with a stable on-state current of 5 μA for more than 1.5 million switching cycles and did not suffer mechanical failure. It can be seen from Figure 3a that the on-state current of 5 μA was reached for all switching cycles (> 1.5 million) without any significant reduction in on-state current. On the other hand, the off-state current was gradually increased from ≈10 fA to ≈1 pA as switching cycles increases. Figure 3b illustrates the long-term endurance of the switch S2 with ≈5 millions of switching cycles. It is well evident that each switch exhibits different contact life time over its switching course and it varies for each device. The common traits observed among all the measured devices was that on-state current remains stable throughout the long-term measurement and the off-state current gradually increases over the time of switching operation. Additionally, more than 10 2 -10 4 orders of magnitude change in the off-state current was observed in all the measured devices. The change in the off-state current is illustrated in three different regions in Figure 3b. In all the measured switches, the electrical migration or welding caused by the Joule heating was not observed. The increase in off-state current is attributed to the mechanical degradation of the SiO 2 layer due to the multithousand hot-switching cycles. The repeated high-speed mechanical switching cycles will lead to the enhancement of physical stress in the clamping electrode regions, which is already partially hang due to the over-etching of the SiO 2 layer by buffered hydrofluoric acid. [47] In addition, examining the SEM images of the NEM contact switches after electrical measurement, we found various degrees of structural changes in the suspended graphene beam that is randomly varies for each device. Figure 3c illustrates . Long-term hot switching charateristics of G-G NEM contact switches. a) On-state and off-state switching current (Is) of switch S1 for > 1.5 million switching cycles. b) On-state and off-state switching current (Is) of switch S2 for > 5 million switching cycles. Switching current compliance limit was set to 5 μA for (a) and (b). c) Failure of the switch S1 during the inter-spread high resolution switching measurement after 1.5 million switching cycles. d) Failure of the switch S2 during the hot-switching measurement itself. The +5m annotation on the x-axis of (d) indicates the switching cycles measured after the 5 millions of switching. e) The Radar plot of various types of failures observed in the G-G NEM contact switches for 30, 60, and 90 nm. f) Relationship between the pull-in voltage and failure by stiction plotted against airgap thickness. the failure of the switch S1 while conducting the inter-spread high-resolution switching measurement. In the hot-switching measurements, switching transition time is ≈50 ms and contact time is also ≈50 ms. Whereas in the high-resolution switching measurement, the applied voltage was swept at the rate of 0.05 V s −1 . In addition, the suspended beam remains in contact for ≈2.5 s for 0.05 V hysteresis window, which is lowest measured hysteresis. It is worth to note that the inter-spread high-resolution switching was conducted after a million of hot-switching cycles, at the point the suspended graphene is started to damage (which is confirmed from the increase in leakage current as shown in Figure 3b). The increased contact time and area make the suspended beam vulnerable to being stuck at the contact, which subsequently leads to the device failure. Figure 3d illustrates the failure of the switch S2 during the hot-switching measurements itself. After more than five millions of switching cycles the switch continues to work at stable off-current for few hundreds of switching cycles. Then, off-current linearly increases from the off-state current floor of ≈1 nA to compliance limit of 5 μA with in 100 switching cycles and got stuck with the on-state current  The radar plot in Figure 3e illustrates the various types of failures observed in G-G NEM contact switches. A total of 192 devices were considered for this analysis with 64 devices each for 30, 60, and 90 nm air gap thicknesses, respectively. A significant portion of all the fabricated devices were failed prior to the electrical measurement. The pre-electrical failures including writing errors in electron beam lithography (EBL) process, overlapping of electrodes due to the misalignment of exposed patterns in EBL and the broken electrodes/contact pads during the fabrication process. (Refer Supporting Information Section S10 for more on the pre-electrical measurement failures in the G-G NEM contact switches). Stiction, rolled-up graphene beam, and broken graphene beam are three different types of failures observed postelectrically. Stiction constitutes more than one third (≈35%) of failures in all fabricated devices. This is attributed to the degradation of the contact graphene during the long-term continuous hot-switching measurements. It is also worth to mention that the failure by stiction is very high (≈50%) for the devices fabricated with small air gap thickness (30 nm). Figure 3f illustrates the rate of failure by stiction over different air gap thickness. It is well evident from the figure that the small air gap thickness significantly reduces the pull-in voltage, but the rate of failure by stiction is increased with the decrease in the air gap thickness. Even though the air gap thickness was set well above the active range of van der Waals force, [48] increased rate of stiction at 30 nm was attributed to buckling of graphene beam. In addition, the broken graphene beam during the electrical measurement causes another 25% of the failures observed. It can be important to note that, the broken graphene beam is either caused during the fabrication process itself or during the electrical measurement.

Various Types of Failures in G-G NEM Contact Switches
To understand the failure modes in the fabricated switches, we examine the SEM image of devices after the electrical measurement. We consider six different switches and their geometrical dimensions as well as the electrical performance are furnished in Table 1. Figure 4 illustrates the SEM images of G-G NEM contact switches with various types of failures. Figure 4a-f corresponds to the switches of S2-S7 in Table 1, respectively. Stiction is one of the common failure modes in the G-G NEM contact switches; the suspended graphene beam is stuck on to the fixed top actuation electrode and keeps the switch closed even after the applied electrical voltage is completely removed. As mentioned earlier, the adhesive forces at the contact interface are greatly reduced owing to the van der Waals interaction between the graphene-tographene contact. However, we found that three different types of failure modes in the fabricated G-G NEM contact switches. Figure 4a shows the SEM micrograph of a failed device, the graphene beam is stuck on to the actuation electrode. The geometrical dimensions of the device obviously remain undamaged after the electrical measurement. However, in a closer look one can observe that the suspended graphene bem has many defect sites. As we have used CVD grown graphene, it is assumed that the defects were present in the graphene beam even before the electrical measurement. [49] More likely, the continuous mechan-www.advancedsciencenews.com www.advphysicsres.com ical switching may increase the defects in graphene beam and subsequently reduces mechanical stability of the switch. We attribute a similar scenario at the contact graphene for the device failure. The failure of the switch with graphene as contact layer in vacuum was owing to irreversible stiction after extinction of contact graphene at the contact site of the top actuation eletrode. The single layer CVD grown graphene was used as the contact layer, the CVD graphene has point and line defects. [50] The continuous mechanical switching with very high speed at such a defect site may eventually lead to more defects and subsequently extinction of graphene layer at the contact. [51] After the extinction of graphene at the contact, the suspended graphene directly makes contact with metal electrode. This may lead to permanent stiction of suspended graphene beam on to the metal electrode. Figure 4b represents a similar failure of the device with geometrical damage and changes were observed in the graphene beam. The failure of the device is due to stiction of the graphene beam which might have been caused by extinction of graphene at the contact site. Figure 4c represents a device failure with significant observable changes in the geometrical dimensions. The edges of the suspended graphene beam start to roll-up and remain unaffected at the anchor electrodes. The red dotted lines in Figure 4c represent the originally designed width of the graphene beam; the arrows indicate the structural damage caused by rolling-up of graphene beam. Figure 4d represents a device failure with completely rolled-up graphene beam. Figure 4e illustrates a device failure with completely rolled-up graphene beam and almost broken at one end of the anchor. In addition, long suspended side of the graphene beam with respect to actuation electrode rolledup more than the shorter side. Owing to reduced restoring force at the anchors of long suspended side of the graphene beam, it deteriorates more and the damages are significant which is consistant with the classical theory of micro beams. [52] Another common failure mode is, broken suspended graphene beam as shown in Figure 4f. Breaking mainly occurred at anchoring ends of the graphene beam owing to under etching of the SiO 2 at the anchoring electrodes. Under etching is caused by the BHF based isotropic wet chemical etching method. [47] It is also worth to note that, most of the fractures in the graphene beam were observed at the long suspended side. Asymmetry in the suspended graphene beam with respect to top actuation electrode is unintentional and it was caused by the misalignment in electron beam exposure during the fabrication process. Owing to the repeated mechanical switching, the mechanical stability of graphene beam is reduced and the edges started to roll-up. The rolling leads to the reduction in contact area of the switch. The increase in contact resistance can be attributed to reduced width of the suspended graphene beam. The relationship between asymmetric electric field and mechanical stability will be further investigated in the following section.

Variations in the Pull-In Voltage of G-G NEM Contact Switches with Identical Device Dimensions
One of the very important aspect of transistors in the CMOS electronics is that it's stable threshold or on-state voltage. [53] The on-state voltage or pull-in voltage in the NEM contact switch is highly depends on the geometrical dimensions of the switch. So it is essential to discuss the fabrication degree of merits that influences the pull-in voltage in the G-G NEM contact switch. We have observed variations in pull-in voltage of the G-G NEM contact switches with identical device dimensions. To understand the reason behind this phenomenon, we have fabricated G-G NEM contact switches with systematic intentional shift of the the top actuation electrode with respect to the suspended graphene beam. (See Supporting Information section S11 for more on pullin voltage variation). The SEM micrographs of the devices are shown in Figure 5. The SEM images were taken after the electrical measurement. All the switches are fabricated with identical geometrical dimensions of length of 1 μm, width of 0.5 μm and air gap thickness of 60 nm. From the SEM micrographs, it is evident that the top actuation electrodes of the switches were shifted toward the right anchor electrode with respect to center of the graphene beam. The narrowing of the gap between top actuation electrode and anchor electrode is clearly visible in the Figure 5a-e. The systematic shift of the top actuation electrode induces the asymmetry in the device structure. This asymmetricity induces the asymmetric electric field between actuation electrode and suspended graphene beam. The electric field strength at the center of the graphene is linearly decreases with respect to the shift of the actuation electrode.
The FEM based simulation was carried out to understand the effect of asymmetric electric filed on pull-in voltage. The detailed information on FEM simulation is given in the FEM simulation section. The Figure 5f demonstrates the measured pull-in voltage experimentally and obtained pull-in voltage along with electric field strength in FEM simulation. When the shift is increased, the pull-in voltages are measured to be 0.5, 0.8, 1.5, 2.4, and 3 V for shift in position x = 0, 50, 100, 150, and 200 nm, respectively. The electric field strength is reduced from 0.22 to 0.13 V nm -1 for x = 0 and x = 200 nm, respectively. It is well evident from Figure 5f that the electric field is reduced at the large shifted actuation electrode and needs more voltage to induce a pull-in. In other words, the switches with large shifted top actuation electrode need more voltage to deform the suspended graphene beam for same airgap thickness compared to switches with no shift in the top actuation electrode.

Finite Element Method Simulation of G-G NEM Contact Switches
In order to understand the manifestations of the mechanical failure modes owing to the asymmetric electric field generated by the misalignment of top actuation electrode, a 3D FEM based simulation of the G-G NEM contact switch was conducted by replicating the experimental device structure. The electrical and mechanical characteristics of the G-G NEM contact switch were simulated using the FEM based tool COMSOL (5.6, COMSOL Inc., Burlington, MA, USA). [54] (See Experimental Section for more details on FEM modeling). Furthermore, to reduce computational complexity the dimensions of device S2 is adopted for FEM simulations. Figure 6a shows initial structure of the simulated G-G NEM contact switch in COMSOL simulation environment with meshing nodes. To understand the mechanical properties owing to the asymmetric electric field, two different structures were studied. The schematic diagrams of the structures are illustrated in Figure 6b. In the first case, the center of the top actuation electrode is aligned well with the center of the graphene beam, referred as the C-C (center-to-center) aligned structure. In the second case, center of top actuation electrode is shifted with respect to center of the graphene beam. This was done purposefully to replicate the experimental device with misalignment, and referred as the C-E (center-to-edge) misaligned structure. This shift creates asymmetric mechanical behaviour of graphene beam with respect to center aligned device structure. The graphene beam was modeled as a linear isotropic material with a thickness of 0.35 nm. The initial air gap thickness between the suspended graphene beam and the actuation electrode was set to 90 nm, with a length and width of the beam of 2 and 0.5 μm, respectively. The model was meshed with a global tetrahedral mesh with minimum element size of 0.35 nm was adopted to refine the mechanical deformation of the suspended graphene beam precisely. In addition, to obtain the accurate stress gradients along with the deflected graphene beam, the graphene beam has meshed with swift mesh. (Refer Supporting Information S12 for more information on FEM analysis of the NEM contact switch).
For graphene the mechanical properties such as density of 2.2 g cm −3 , Young's modulus of 1 TPa, and Poisson's ratio of 0.17 were used. [55] In this FEM model, the out of plane mechanical properties of the graphene beam is considered only within the limits of linear elastic theory. [29] First, the pull-in was confirmed by beam displacement. Figure 6c illustrates the beam displacement of 90 nm for 1.3 V, which is the complete air-gap closing distance by graphene beam. The pull-in voltage obtained in FEM simulation closely resembled to the experimental result. Figure 6d illustrates the beam displacement in 3D at pull-in voltage for C-C aligned structure. The potential distribution of the switch across the center of the graphene beam is illustrated in Figure 6e and arrows represents the electric field direction. The tensile stress was analyzed further to understand the failure modes. The 1D line plot of the stress along the graphene beam is shown in the Figure 6f. The tensile stress profile of the deflected graphene beam at the pull-in voltage for center aligned and misaligned structure with 400 nm is shown in Figure 6f. It is well known from the figure that center aligned structure has very symmetric stress profile and the misaligned structure has very asymmetric stress profile. The stress reaches the maximum value at pull-in condition. The tensile stress profile was very symmetry with respect to the center of the graphene beam for C-C aligned structure and the maximum value is ≈30 GPa. The asymmetric stress behavior was obtained for misaligned structure. Furthermore, tensile stress reached the maximum value ≈60 GPa, which is comparable to breaking strength of the CVD graphene reported in the literature. [56] During the continuous mechanical switching of the device, the asymmetric and high magnitude stress in the suspended graphene beam can facilitates the rolling-up of the graphene beam and subsequently leads to fracture of the graphene beam. Figure 6g shows the electric field strength for both C-C structure as well as C-E misalignment structure. The electric field reaches the maximum value at the edges of the graphene beam for both of the cases but evidently the magnitude of the electric field is reduced for the misaligned case. To understand further, we consider the single point integration of the electric field strength at the midpoint of the graphene beam. Figure 5f shows the electric field strength, and the pull-in voltage obtained experimentally as function of the misalignment. It is obvious that the electric field strength at Figure 6. FEM simulation of the G-G NEM contact switches. a) Initial structure of the G-G NEM contact switch with tetrahedral meshing. Each material layer is color coded with unique colors, red, magenta, blue, and yellow, respectively, for Si, SiO 2 , graphene, and gold. b) Schematic illustration of C-C and C-E structures. c) 1D line plot of the beam displacement. d) The 3D profile of the double-clamped graphene beam displacement at 1.3 V; Color legend shows the displacement in meter. e) The potential distribution (3D) of the switch across the center of the graphene beam; Arrows illustrate the electric field direction; Color legend shows the potential in voltage. f) The 1D line plot of the principal stress (tensile) for C-C aligned and C-E misaligned structures. g) The 1D line plot of the electric filed strength for C-C and C-E structures. midpoint of the graphene beam reduces as the misalignment increases; this is consistent with the increased pull-in voltage of the device. The FEM analysis of the G-G NEM switches with misalignment of the top actuation electrode reveals that the asymmetric electric field around the graphene beam can be one of the plausible reasons for the observed pull-in voltage deviation in the G-G NEM contact switches with identical device dimensions. The two terminal graphene NEM contact switches designed with graphene as contact material can possibly utilized as an alternative for the conventional CMOS transistor. Further-more, the low pull-in voltage with very high contact lifetime endurance has always been expected of NEM contact switches and the G-G NEM contact switches are reported with better switching performance in terms of low pull-in voltage and high reliability compared to the NEM switches reported in the literature (Refer Supporting Information Section S13 for more information on literature review of the high reliable NEM contact switches). Our device can be potentially used for energy efficient era of ultralow-power CMOS and CMOS-NEM hybrid integrated circuits.

Conclusion
We developed an electro-statically enabled graphene based NEM contact switch with graphene as the contact material. The fabricated G-G NEM contact switches demonstrated with an ultra low pull-in voltage of < 0.5 V as well as high contact lifetime of > 1.5 million switching cycles. The switches also showed an excellent switching performance, includes low off-state leakage current, stable on-state current and high on/off ratio. The G-G NEM contact switches demonstrated with stable electrical contact and maintained with an on-state current of 5 μA for more than five million switching cycles. Various types of failures that occurred in the G-G NEM contact switches were quantitatively analyzed. The role of an asymmetric electric field in the switching characteristics was also investigated in detail. The work presented here demonstrates that G-G NEM contact switch can be a potential candidate for achieving a reliable ultra-low power energy efficient switching applications.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.