Static capacitive pressure sensing using a single graphene drum
Dejan Davidovikj, Paul H. Scheepers, Herre S. J. van der Zant, Peter G. Steeneken
SStatic capacitive pressure sensing using a single graphene drum
D. Davidovikj , ∗ P. H. Scheepers , H. S. J. van der Zant , and P. G. Steeneken , Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands Department of Precision and Microsystems Engineering,Delft University of Technology, Mekelweg 2, 2628 CD, Delft, The Netherlands
To realize nanomechanical graphene-based pressure and gas sensors, it is beneficial to have amethod to electrically readout the static displacement of a suspended graphene membrane. Capac-itive readout, typical in micro-electro-mechanical systems (MEMS), gets increasingly challenging asone starts shrinking the dimensions of these devices, since the expected responsivity of such devicesis below 0.1 aF/Pa. To overcome the challenges of detecting small capacitance changes, we designan electrical readout device fabricated on top of an insulating quartz substrate, maximizing thecontribution of the suspended membrane to the total capacitance of the device. The capacitance ofthe drum is further increased by reducing the gap size to 110 nm. Using external pressure load, wedemonstrate successful detection of capacitance changes of a single graphene drum down to 50 aF,and pressure differences down to 25 mbar.
Nanomechanical devices from suspended graphene andother two-dimensional materials have been receivinggrowing interest in the past few years for their potentialas sensitive pressure and gas sensors. To realize in-tegrated, small and low-power devices, it is necessary tohave all-electrical on-chip transduction schemes, in con-trast to the currently often employed laser interferometrytechniques for the readout of their dynamic motion andstatic deflection.Reports on electrical readout of graphene membranenanomechanical devices have employed readout schemesbased on electrical transconductance and piezoresis-tivity . Both of these rely on the change in the con-ductance of the membrane as a function of deflection,which is then used to sense the motion of the mem-brane. Although these methods can be very sensitive, thegraphene conductance can also be affected by variationsin gas composition, humidity, light intensity and temper-ature. Moreover, the conductance is not only related tothe deflection of the graphene membrane, but dependsalso on material parameters like the electron mobilityand piezoresistive coefficients. These approaches there-fore require calibration and a high degree of stability ofthe graphene and insensitivity to variations in its sur-roundings.In contrast, the capacitance between a graphene mem-brane and a bottom electrode is, to first order, a func-tion only of the geometry of the system and, therefore,the deflection of the membrane. A measurement of thecapacitance of the membrane can therefore be used tocalculate its deflection, which makes capacitance detec-tion an interesting alternative method for electrical read-out of nanomechanical graphene sensors. Dynamic (on-resonance) capacitive readout has been demonstrated onsuspended graphene bridges . Measurements usingstatic capacitive readout of the deflection of graphenemembranes have been conducted on a voltage tunable ca-pacitor array comprised of thousands of graphene mem-branes in parallel .Here we extend on this work by capacitive detectionof the deflection of a single graphene drum and demon- (a) P out > P in C = C C > C C < C (b) quartz s p i n - on g l a ss AuPdgraphene A u P d P out < P in P out = P in FIG. 1. (a) A 3D schematic of the device: a capacitor isformed between a graphene drum suspended over a metalliccavity and a bottom metallic electrode that runs underneathan insulating (spin-on glass) oxide layer. The entire device isfabricated on top of an insulating quartz wafer. (b) Actuationprinciple: external pressure load is applied. Depending on thepressure difference between the cavity and the outside envi-ronment, the nanodrum will bulge upwards or downwards re-sulting in a decrease or increase of the measured capacitance. strating its performance as a pressure sensor.The total capacitance of a single graphene drum and theunderlying electrode in typical sample geometries (cir-cular drum 5-10 µ m in diameter, suspended over a 300nm-deep cavity), ranges from 0.5 to 2 femtoFarads. Adisplacement of such a drum of 1 nm would result in acapacitance change of only 2-6 attoFarads. Fabricatingreadout circuitry sensitive enough to detect such changesis faced with a few challenges. (i) Very shallow gapsare needed in order to maximize the capacitance of the a r X i v : . [ phy s i c s . a pp - ph ] A ug AuPd Q ua r t z Q ua r t z Q ua r t z S OG AuPd AuPd Q ua r t z S OG AuPd Q ua r t z S OG AuPd (a) (b) (c)(d) (e)e-beam lithography + AuPd evaporation lift-off SOG spin-coating + curing1h500°CN PMMA spin-coating + e-beam lithography +AuPd/Cr evaporation + lift-off reactive ion etching
CHF + ArAuPdCr (g) Cr etch Q ua r t z S OG AuPdAuPdPMMA 950KPMMA 495KAuPdCr (h)
500 nm5 μ m FIG. 2. (a) - (g) Fabrication steps. (h) A false-colored SEM image of the device showing the top and bottom electrodes (yellow)and the separating SOG layer (blue). A zoomed-in image of the AuPd/SOG/AuPd interface is shown in the bottom panel. device; (ii) Parasitic capacitances between the readoutelectrodes need to be as small as possible to improve thesignal to noise ratio; (iii) The surface should be flat andadhesive to facilitate the transfer of graphene; (iv) Ad-ditionally, to keep the pressure in the reference cavityconstant, the cavity needs to be hermetically sealed bythe graphene membrane. To address these challenges,we develop a device with electrical readout fabricated ontop of a quartz substrate, which substantially reducesthe parasitic capacitance of the electrical circuitry. Todemonstrate the sensing concept, we transfer a few-layergraphene flake on top of the device and we use externalgas pressure load to deflect the drum, reading out thecorresponding change in the capacitance.A 3D schematic of the proposed device is shown inFig. 1(a). The capacitor consists of a circular electrodeon the bottom and a suspended graphene drum on top,forming a sealed cavity. The bottom electrode runs un-derneath a dielectric layer of spin-on-glass (SOG), whichseparates it from the top metal electrode. The graphenedrum is mechanically supported by the top electrode,which also serves as an electrical contact to the graphene.Figure 1(b) shows the sensing principle: when the pres-sure inside the cavity ( P in ) is equal to the outside pressure( P out ), the capacitance of the device is given by the par-allel plate capacitor formed by the graphene and the bot-tom electrode: C . When the outside pressure is higherthan the pressure inside the cavity, this results in a pos-itive pressure difference across the membrane, causing itto bulge downwards, which manifests itself as an increaseof the measured capacitance. Conversely, if the pressureinside the cavity is higher than the outside pressure, thedrum bulges upwards, resulting in a decrease of the mea-sured capacitance.Fabrication requires two e-beam lithography steps forthe bottom and top electrodes. Both lithographic stepsuse two layers of PMMA resist (A6 495K [300 nm] andA3 950K [100 nm]) in order to create sloped resist walls which facilitate the lift-off. To minimize charging effectsduring the e-beam patterning, a 10 nm layer of Au issputtered on top of the resist prior to the e-beam expo-sure. The Au layer is removed before developing the re-sist using KI/I gold etchant. Figure 2(a) shows a sketchof the sample after developing the resist in MIBK:IPA(1:3) and evaporating 5 nm of titanium (5 nm) and 60nm of gold-palladium (Au . Pd . ) to form the bottomelectrode. The titanium is used as a thin adhesion layerand is not shown in the figure.After lift-off (Fig. 2(b)), a layer of FOX XR-1451spin-on-glass (SOG) is spin-coated on the chip. Inorder to improve the conformity of the SOG layer tothe underlying surface, the SOG layer is baked in twostages: 3 minutes at 150 ◦ C and 3 minutes at 250 ◦ C. Subsequently, the chip is placed in a N furnaceat 500 ◦ C at 1 atm, which cures the SOG, makingit mechanically harder and also improving its surfacesmoothness and step coverage (Fig. 2(c)). The bakingand curing processes are essential for obtaining a flatand smooth surface, which is important, as it largelyinfluences the roughness of the electrode evaporatedon top of it. Smooth surfaces enhance adhesion andthereby facilitate the transfer of graphene. The currentprocess flow results in a cavity depth of 110 nm. Thetop electrodes are fabricated on top of the SOG layer,following the same steps of Fig. 2(a-b), with a differentcombination of metals: Ti/Au . Pd . /Cr (5 nm/90nm/30). This is shown in Fig. 2(d). The top layerof chromium is used as a hard mask for the followingetching step, to avoid contamination of the underlyingAuPd.Fig. 2(e) shows the formation of the cavities by usingreactive ion etching (RIE) of the SOG everywherearound the top electrodes. This is done at 7 µ bar inCHF :Ar (50:2.5 sccm). The remaining Cr is thenetched away using Cr etchant, which results in the finaldevice (Fig. 2 (g)). The cavity depth can be easilytuned by changing the thickness of the top layer ofAuPd. In Fig. 2(h) we show a false-colored SEM imageof the device after the removal of the Cr. The bottompanel shows a zoom-in of the interface between the twoelectrodes (yellow) and the SOG layer in between (blue).After the device has been fabricated, graphene flakes aretransferred on top of the cavities using a dry transfertechnique. The resulting graphene drums are 5 µ m indiameter.The measurement setup is shown in Fig. 3. The deviceis mounted in a vacuum chamber connected to a mem-brane pump and a pressure controller. The pressure con-troller is connected to a N gas bottle (purity 99.999 %)and the pressure of the gas inside the chamber can becontrolled linearly by using a 0-10 V input voltage. Thepressure controller has a voltage output, which enablesa direct readout of the pressure inside the chamber. Inthis configuration, the pressure can be regulated between1-1000 mbar (0-10 Volts on the input) with a resolutionof ≈ V p = 100 mV. The integration time for the capacitancereadout is 1500 ms. The inset of Fig. 3 shows an opticalimage of the measured device: a 6 nm - thick graphenedrum.The measurement scheme is sketched in Fig. 4(a). Al-though graphene hermetically seals off the cavity, slowgas permeation usually takes place through its edges orthrough the underlying oxide . We make use of this ob-servation and keep the sample in vacuum for 48 h prior toeach measurement to ensure that the gas from the cavityis completely evacuated ( P in ≈ P out ) is changed pressure chamberLCR meterpressurecontrollerN DC sourceV 1 2C fFpump
FIG. 3. Schematic of the measurement setup: the device ismounted in a pressure chamber connected to a pressure con-troller with a DC voltage control input. The voltage outputof the pressure controller is proportional to the actual pres-sure inside the chamber ( P out ). The capacitance of the drumis read out using an LCR meter. The figure shows an opticalimage of the device. C apa c i t an c e ( f F ) P ou t P max P max = 250 mbarP max = 600 mbar Δ C = 400 aF Δ C = 270 aF(a)(b) Time (s)200 400 600 800 1000 12000
FIG. 4. (a) Experimental procedure: the pressure is changedfrom vacuum to P max with a period of 120 s. (b) Capacitanceof the device as a function of time for two different runs using P max = 600 mbar (top) and P max = 250 mbar (bottom). Bothpanels show the extracted capacitance step height (∆ C ) at themoment of changing the pressure. in a step-like fashion. Fig. 4(b) shows the measuredcapacitance of the device as a function of time for twovalues of the pressure step height: P max = 600 mbar(blue) and 250 mbar (red). Both graphs show that thecapacitance rises when the pressure inside the chamber P out increases, and jumps back to the initial value uponpumping down. Despite the care taken to eliminate par-asitic capacitances, by using a quartz substrate and localgate, the total capacitance of the device is ≈
590 fF,mostly stemming from the parasitic capacitance of thewiring and the on-chip inter-electrode capacitance, sincethe contribution of the graphene drum is calculated tobe only 1.58 fF.Starting from P in = P out and assuming an abruptchange in P out , such that permeation effects can be ne-glected, the expected capacitance change can be calcu-lated using an implicit relation between the pressure dif-ference across the membrane (∆ P ) and the deflection ofthe membrane’s center ( z ) :∆ P = 4 n R z + 8 Eh R (1 − ν ) z , (1)where n is the pre-tension of the membrane, R and h areits radius and thickness respectively and E and ν are theYoung’s modulus and the Poisson’s ratio of the material.Knowing z and the spherical deformation shape of themembrane ( U ( r ) = z (1 − r R )), the capacitance can becalculated using the parallel plate approximation as: C = 2 πε (cid:90) R rg − U ( r ) dr, (2)where ε is the vacuum permittivity and g is the gapsize. Using n = 0.1 N/m and E = 1 TPa, the valueof the extracted capacitance steps matches well with the (a)(b) datamodel (E = 1 TPa )
90 s25 mbar10 s output P ou t ( m ba r)
48 hinput P out P out Time (s) Δ C ( a F ) Δ C ( a F ) -500 0 500 1000-1000 Δ P (mbar)
FIG. 5. (a) Capacitance change as a function of time (top).Bottom: Starting from 1 bar, pressure is changed in a stepwisefashion with increasing steps of 25 mbar (10 % pulse durationwith a 100 s period). The chamber is then pumped down to1 mbar and a similar procedure is repeated with the chamberbeing pumped to vacuum after each step. (b) The extractedvalues for ∆ C as a function of the applied pressure differenceacross the membrane (gray dots). A theoretical fit to the datausing the device dimensions, with pre-tension n = 0.1 N/mand Young’s modulus E = 1 TPa. numbers expected from theory. On top of the measuredsignal, we also measure a slow drift of the capacitanceover time (see Fig. 4). The cause of the drift is not wellunderstood and it might be due to a combination of slowgas leakage and condensation of humidity on the elec-trodes .Using pressure pulses of increasing height we can traceout a dependence of the capacitance change on the pres-sure difference across the membrane. To do so, we employa measurement protocol sketched in the inset of Fig. 5(a).The sample is kept at vacuum and short pressure steps(10 s) are applied to the sample chamber, followed by 90s of pumping, to ensure that the cavity underneath thegraphene is pumped down to vacuum before applying thenext pressure step. This way, it can be safely assumedthat the height of the pressure step corresponds to the ac-tual pressure difference felt by the graphene membrane.The opposite applies for the left side of the graph (bluecurve): the sample is kept at 1 bar and pressure steps ofthe opposite sign are applied, followed by 90 s of ambientpressure. The measured capacitance change ∆ C is plot-ted in the bottom panel of Fig. 5(a). The aforementioned drift was subtracted for this dataset after fitting it witha polynomial (see Supporting Information Section I).The capacitance change is recorded as the height ofthe step in the measured capacitance immediately afterapplying the pressure pulse. Doing this for the entirespan of ∆ P (from -1 to +1 bar) we get a ∆ C vs. ∆ P curve, plotted in Fig. 5(b). The error bars at each pointcorrespond to the RMS noise of the signal in the vicinityof the pressure step, as a measure of the uncertainty ofthe step determination. The black curve is the modeledresponse of the system for a 6 nm thick graphene mem-brane with a Young’s modulus of 1 TPa. The model isin a good agreement with the measured response, pro-viding further evidence that the signal is indeed comingfrom the displacement of the membrane. Thanks to therelatively low parasitic capacitances, despite the presentdrift, capacitance changes down to 50 aF could be dis-tinguished.The resolution of the measurement setup is limited bythe resolution of the LCR meter, which is 10 aF. Thiscorresponds to a pressure resolution of ≈
360 Pa (or 0.36mbar) for ∆ P ≈ P = 1bar. For potential application of such device as a pressuresensor, it is interesting to look at the sensitivity of the de-vice around ∆ P = 0. By design, the sensitivity of the pre-sented device peaks at around 0.1 aF/Pa at 0 mbar pres-sure difference (see Supporting Information Section II).The root-mean-square (RMS) noise of the measurementsetup is 25 aF/ √ Hz. However, due to the drift presentin the measurements, the minimal step height that couldbe resolved was 50 aF. The relative error of the pressuremeasurement ranges from 0.6 % (for ∆ P ≈ −
100 mbar < ∆ P <
100 mbar. Theaccuracy of the sensor can also be influenced by morpho-logical imperfections of the membrane itself.There are multiple ways to increase the sensitivity ofthe device: decreasing the thickness of the graphene ( h ),decreasing the pre-tension of the membrane ( n ), increas-ing its radius ( R ), or connecting N such devices in paral-lel. A detailed analysis of the influence of each parameteron the sensitivity of the device are shown in the Support-ing Information Section II. According to the calculations,changing the thickness h does not drastically influencethe sensitivity. Increasing R or decreasing n improvethe sensitivity by one or two orders of magnitude. Wenote that controlling the pre-tension is challenging, sinceit largely depends on the transfer process and usually re-sults in large spreads . Moreover, making devices withlarger radii and low pre-tension would impair the yield ofthe devices and reduce their dynamic range (due to col-lapse of the membrane at high ∆ P ). However, increasingthe number of drums in parallel N linearly increases theresponsivity of the device. With more than 1000 drumsin parallel (resulting roughly in a chip size of 100 x 100 µm ) one could push the responsivity to values higherthan 100 aF/Pa, resulting in a 0.1 Pa resolution usingour current measurement setup.To demonstrate the feasibility of capacitive readoutof the graphene sensor with an integrated circuit, we re-placed the LCR meter with an Analog Devices (AD7746)capacitance-to-digital converter chip (with dimensions5x5 mm ) which we interfaced through using the built-inI2C protocol library of an Arduino. We show an exam-ple of such measurement in the Supporting InformationSection III. Even though the signal-to-noise of this mea-surement is worse than the one using the LCR meter, itstill serves as a proof-of-principle that on-chip detectionof small capacitance changes can be realized using com-mercial electronic devices.The drift in the measurement together with the poorhermeticity of the membrane hamper the long-term sta-bility of the device. For its commercial application asa pressure sensor, the hermeticity of the device needsto be improved (e.g. by properly sealing the membraneedges) and the cross-sensitivity to the environment (hu-midity/gas composition) needs to be investigated morethoroughly.In conclusion, we demonstrate on-chip capacitive read-out of a single suspended graphene drum. To obtain theresponsivity required for sensing such small capacitancechanges, the entire fabrication is done on an insulatingquartz substrate, minimizing the parasitic capacitance ofthe readout electrodes. We use uniform pressure loadto statically deform the membrane, which results in acapacitance change of the device. Using this method, we are able to read out capacitance changes down to50 aF and detect pressure steps down to 25 mbar. Theheight of the steps is consistent with predictions from thetheoretical model. We also traced out a force-deflectioncurve by pulsing the pressure in the chamber with pulsesof increasing height. The measured ∆ C vs. ∆ P curvematched well with theory, based on a graphene mem-brane with a Young’s modulus of 1 TPa. We also mea-sured a temporal drift in the capacitance, possibly origi-nating from residual humidity in the chamber. This workis aimed at probing the limit of static capacitive detec-tion of graphene nanodrums. We optimized the deviceto enable detection of very small capacitance changes ofdown to 50 aF. By combining this device design with anon-chip capacitance-to-digital converter we show a proof-of-concept demonstration of the feasibility of integratingsuspended 2D membranes into next-generation pressuresensors. Acknowledgements
This work was supported by the Netherlands Organisa-tion for Scientific Research (NWO/OCW), as part of theFrontiers of Nanoscience (NanoFront) program and theEuropean Union Seventh Framework Programme undergrant agreement n ◦ ∗ [email protected] A. Smith, S. Vaziri, F. Niklaus, A. Fischer, M. Sterner,A. Delin, M. ¨Ostling, and M. Lemme, Solid-State Elec-tronics , 89 (2013). R. J. Dolleman, D. Davidovikj, S. J. Cartamil-Bueno,H. S. J. van der Zant, and P. G. Steeneken, Nano Let-ters , 568 (2016). A. D. Smith, F. Niklaus, A. Paussa, S. Schroder, A. C.Fischer, M. Sterner, S. Wagner, S. Vaziri, F. Forsberg,D. Esseni, et al. , ACS nano , 9879 (2016). Y.-M. Chen, S.-M. He, C.-H. Huang, C.-C. Huang, W.-P.Shih, C.-L. Chu, J. Kong, J. Li, and C.-Y. Su, Nanoscale , 3555 (2016). R. N. Patel, J. P. Mathew, A. Borah, and M. M. Desh-mukh, 2D Materials , 011003 (2016). S. P. Koenig, L. Wang, J. Pellegrino, and J. S. Bunch,Nature Nanotechnology , 728 (2012). L. Wang, L. W. Drahushuk, L. Cantley, S. P. Koenig,X. Liu, J. Pellegrino, M. S. Strano, and J. S. Bunch, Na-ture nanotechnology , 785 (2015). R. J. Dolleman, S. J. Cartamil-Bueno, H. S. J. van derZant, and P. G. Steeneken, 2D Materials , 011002 (2016). C. Chen, S. Rosenblatt, K. I. Bolotin, W. Kalb, P. Kim,I. Kymissis, H. L. Stormer, T. F. Heinz, and J. Hone,Nature Nanotechnology , 861 (2009). Y. Xu, C. Chen, V. V. Deshpande, F. A. DiRenno,A. Gondarenko, D. B. Heinz, S. Liu, P. Kim, and J. Hone,Applied Physics Letters , 243111 (2010). C. Chen,
Graphene NanoElectroMechanical Resonatorsand Oscillators (Columbia University, 2013). M. AbdelGhany, F. Mahvash, M. Mukhopadhyay,A. Favron, R. Martel, M. Siaj, and T. Szkopek, 2D Ma-terials , 041005 (2016). J. S. Bunch, S. S. Verbridge, J. S. Alden, A. M. van derZande, J. M. Parpia, H. G. Craighead, and P. L. McEuen,Nano Letters , 2458 (2008). J. S. Bunch,
Mechanical and electrical properties ofgraphene sheets (Cornell University Ithaca, NY, 2008). L. Ford, Journal of the Institution of Electrical Engineers-Part II: Power Engineering , 709 (1948). C. Lee, X. Wei, J. W. Kysar, and J. Hone, Science ,385 (2008). S. J. Cartamil-Bueno, A. Centeno, A. Zurutuza, P. G.Steeneken, H. S. J. van der Zant, and S. Houri, Nanoscale , 7559 (2017). Supporting InformationI. Background drift subtraction
In this section we present an example of drift subtraction using the raw data of the ∆ C vs. ∆ P measurementpresented in Fig. 5 (a) of the main text (for P in ≈ th order) polynomial through these minima (red line) and subtracting it from the data (blueline). The resulting signal represents the ∆ C vs. ∆ P curve plotted on the right-hand side of Fig. 5 (a) of the main text. Time (s) C apa c i t an c e ( f F ) raw datafitted background FIG. S1. Drift subtraction: the blue curve represents the raw capacitance data, while the red curve is a polynomial fit of thelocal minima of the capacitance after each pumping step.
II. Pressure sensitivity
In this section we lay out the theoretical predictions of the expected capacitance change as a function of the pressuredifferences across the membrane, together with the expected responsivity of the device.Using eqs. (1) and (2) from the main text we model the predicted capacitance response as a function of the pressuredifference across the membrane. The pressure difference ∆ P is defined as ∆ P = P out − P in , where P out is the pressurein the chamber and P in is the pressure inside the cavity.We start with the dimensions of the device described in the main text, namely, a single graphene membrane witha thickness of h = 6 nm, pre-tension n = 0 . N/m and a radius R = 2 . µ m. For each of the panels (a-d) we keepall parameters fixed, while varying only (a) the thickness, (b) the pre-tension, (c) the radius and (d) the number ofidentical drums connected in parallel. The top panels represent the absolute value of the expected capacitance change( | ∆ C | ) for a given pressure difference (∆ P ). The bottom panels represent the calculated responsivity ( ∂C∂ ∆ P ) of thedevice, expressed in aF/Pa. -400 -200 0 200 40001234-400 -200 0 200 40010 -3 -2 -1 -1000 0 100000.10.20.30.40.50.60.7-1000 0 100010 -4 -3 -2 -1 -100 0 100051015202530-100 0 10010 -3 -2 -1 -400 -200 0 200 400010002000300040005000-400 -200 0 200 40010 -4 -2 Δ P (mbar) Δ P (mbar) Δ P (mbar) Δ P (mbar) Δ P (mbar) Δ P (mbar) Δ P (mbar) Δ P (mbar) r e s pon s i v i t y ( a F / P a ) r e s pon s i v i t y ( a F / P a )r e s pon s i v i t y ( a F / P a ) r e s pon s i v i t y ( a F / P a ) | Δ C | ( f F ) | Δ C | ( f F ) | Δ C | ( f F ) | Δ C | ( f F ) varying thickness (h) varying number ofdrums in parallel (N)varying pre-tension (n ) varying radius (R)SLGBLG1 nm3 nm6 nm 0.01 N/m0.1 N/m1 N/m 2.5 μ m5 μ m7 μ m 110100100010000(a) (b) (c) (d) FIG. S2. Calculated capacitance change (top panels) and responsivity (bottom panels) as a function of pressure difference forvarying (a) thickness, (b) pre-tension, (c) radius of the membrane and (d) number of graphene drums in parallel.
To be able to use such a device as a pressure sensor in mobile devices, considering the day-to-day fluctuations inthe atmospheric pressure, one needs to look at the -100 mbar ≤ ∆ P ≤
100 mbar region of the responsivity plots,where the responsivity is usually the highest.It can be seen from the plots that by varying the membrane’s thickness or pre-tension, one does not benefit a lotwhen it comes to the responsivity of the device. A larger radius, which results in a higher initial capacitance ( C ),naturally, increases the overall responsivity of the device. It has to be noted, however, that making the membraneslarger, thinner or with a lower pre-tension also significantly reduces the dynamic range of the device, because it makesit easier for the membrane to collapse at a high pressure difference ∆ P . Mathematically this follows directly from eq.1 of the main text.One solution for increasing the responsivity of the device is increasing the number of drums connected in parallel(Fig. S2 (d)). This way, the dynamic range of the sensor is unchanged, whereas the overall responsivity of the systemscales proportionally to the number of drums N . Using 1000 drums in parallel (resulting roughly in a 100 x 100 µm device), the responsivity peaks at 100 aF/Pa, which would, with a capacitance resolution of 10 aF, enable a pressureresolution of 0.1 Pa. Such a device could be used in next-generation pressure sensors and could have a comparable orbetter performance than the current state-of-the-art pressure sensors. III. Readout using an AD7746 chip interfaced with an Arduino
To demonstrate the integrability of the proposed device, we replaced the LCR meter with a compact, low-costAnalog Devices (AD7746) 24-bit capacitance-to-digital converter chip. According to the specifications, this chip isable to handle up to 4 pF parasitic capacitance and has a resolution of down to 4 aF. The dimensions of the readoutchip are ≈ and it has a built-in I C interface.
Time (s) C apa c i t an c e ( f F )
650 aF pressure chamberPCI C bus Cin+EXC A 500 10000589588587586585590 t (s) P ou t
01 bar(b)(a)
FIG. S 3. (a) Alternative measurement setup using an AD7746 capacitance-to-digital converter chip (1). The AD7746 isconnected to the sample (2) using two wires (EXC A and Cin+). The sample together with the chip carrier is mounted insidethe pressure chamber. The AD chip is interfaced with I C communication using an Arduino Uno, which is connected directlyto the measurement computer to read out the capacitance of the device. The 10 cent euro coin is shown for scale (20 mm indiameter). (b) Capacitance signal measured using the AD7746 chip following the same measurement scheme from Fig. 4 inthe main text. The pressure step is 1 bar and the corresponding capacitance step is 650 aF, consistent with theory.
The entire measurement setup (without the pressure chamber and the pressure controller) is shown in Fig. S3 (a).The AD chip is mounted on a printed circuit board (PCB) together with a few resistors and capacitors, necessaryfor its basic operation. Two wires from the chip (EXC A (red) and Cin+ (blue)) are connected directly to the topand bottom electrode of our device respectively. Our entire chip (marked with (2) in the image) is mounted on achip carrier inside the pressure chamber. The AD chip is interfaced using an Arduino Uno, which has a built-inI C protocol library. The Arduino is connected to a measurement computer, which records the capacitance valuemeasured by the AD7746 chip.To test the setup, we employ a measurement scheme similar to the one shown in Fig. 4 of the main text. Afterkeeping the device for 48 hours in vacuum, we apply 1 bar pressure steps with a duration of 60 s, followed by 60 s ofpumping. The measured capacitance signal is shown in Fig. S3 (b). We observe a capacitance change of ∆ C ≈ P = 1 bar (see Fig. 5 of the main text).The noise level of these measurements is much higher (79 aF/ √√