Letter pubs.acs.org/NanoLett
A New Switching Device for Printed Electronics: Inkjet-Printed Microelectromechanical Relay Eung Seok Park, Yenhao Chen, Tsu-Jae King Liu,* and Vivek Subramanian* Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, California 94720-1770 United States
ABSTRACT: Printed electronics employing solution-processed materials is considered to be the key to realizing low-cost largearea electronic systems, but the performance of printed transistors is generally inadequate for most of the intended applications due to limited performance of printable semiconductor materials. We propose an alternative approach for a printed switch, where the use of semiconductors can be avoided by building mechanical switches with printed metal nanoparticle-based inks. In this work, we detail the first demonstration of inkjet-printed microelectromechanical (MEM) switches with abrupt switching characteristics, very low on-state resistance (∼10 Ω), and very low off-state leakage. The devices are fabricated using a novel process scheme to build three-dimensional cantilever structures from solution-processed metallic nanoparticles and sacrificial polymer layers. These printed MEM switches thus represent a uniquely attractive path for realizing printed electronics. KEYWORDS: Printing, nanoparticles, metals, switches, MEMS
P
rinted electronics has received a great deal of attention in recent years as a means of realizing low-cost electronic systems such as displays, sensors, and RFID tags. Most efforts to date have focused on printed thin film transistors (TFTs) based on organic semiconductors. The performance of printed TFTs has been improving steadily due to improvements made to printable semiconductor materials, which have shown some promising results recently with respect to switching speed and hysteresis.1−6 On the other hand, further improvement is needed before printed TFTs can be adequate for a wide range of applications, both in terms of the effective carrier mobility and off-state leakage current. We propose that an alternative switch design may be extremely interesting in this regard, particularly if the use of semiconductor materials can be avoided. Interestingly, while printed semiconductors trail their conventional counterparts, the printed metals already come very close to achieving the properties of conventionally processed metallic thin films. Metallic nanoparticle inks have been used to form metallic features with conductivity >70% of bulk metal conductivity. Therefore, an alternative path for realizing printed electronics is clear; implement devices via solution-processed metals exclusively. Specifically, mechanical switches may be used rather than electrical switches, thus delivering the desired functionality with substantially enhanced performance. Mechanical switches can be made without semiconducting materials because they operate by electrostatic actuation of a movable (conducting) electrode structure, rather than by modulating the conductivity of fixed (semiconducting) channels. Indeed, there have been some initial demonstrations of mechanical constructs realized using printed materials,7−11 © 2013 American Chemical Society
suggesting that electromechanical switches may be a viable path for realization of printed electronics. Herein, we describe printed microelectromechanical (MEM) switches fabricated using nanoparticle inks and show on-state resistance (RON) and off-state leakage (IOFF) far better than that of printed transistors, attesting to the promise of this candidate switch for printed electronics.12 Indeed, we show that printed MEM switch technology has the potential to address virtually all the shortcomings of printed transistors. All this is achieved leveraging the best class of printed electronic materials available today, metals. In other words, printed mechanical switches could represent a uniquely attractive path for realizing printed electronics. The structure and operation of the printed MEM switch is illustrated in Figure 1. It comprises three terminals: the movable cantilever beam (the source electrode) is electrostatically actuated downward into contact with a drain electrode when a suitably large voltage difference exists between it and an underlying gate electrode. When the switch is in the OFF state, there is an air gap separating the beam and the drain electrode so that ideally no current can flow (IOFF = 0). Indeed, we find leakage to be at the noise floor of our measurement setup, attesting to the advantages of MEM switches in this regard. In these switches, the position of the source beam depends on the balance between the electrostatic force (Felec) and the spring Received: July 31, 2013 Revised: September 16, 2013 Published: October 3, 2013 5355
dx.doi.org/10.1021/nl4028632 | Nano Lett. 2013, 13, 5355−5360
Nano Letters
Letter
Figure 1. Printed MEM switch fabricated using nanoparticle ink. Scanning electron micrographs of (a) multiple printed MEM switches and (b) close-up view of one switch. The source beam is anchored to the source pad; the two other electrodes (actuating gate and contacting drain) are located underneath the source beam. Schematic cross-sectional illustration of the three-terminal switch structure (c) in the OFF-state and (d) in the ON-state.
Figure 2. Illustration of the process used to fabricate the first MEM switches. (a) Gate electrode is printed onto an electrically insulating substrate using silver nanoparticle ink. (b) Thin film of cross-linked PVP is spin-coated to form the gate dielectric. Drain electrode and source pad are printed in alignment with the gate electrode. Additional drops of ink can be printed in the drain-contacting region to define a contact dimple. (c) Sacrificial layer of PMMA is spin-coated and source beam is inkjet-printed. (d) The source anchor region is defined by printing acetone to dissolve PMMA underneath one end of the source beam. (e) The source anchor is formed by filling the hole in the PMMA with inkjet-printed silver nanoparticle ink, and the structure is released by removing PMMA film with acetone. (f) Optical plan-view micrograph of cantilever beam after the anchor-region hole is defined by inkjet-printing PMMA; (inset) surface profile of the hole along XX′.
restoring force of the beam (Fspring). While Felec increases superlinearly with increasing downward beam displacement, Fspring increases only linearly with displacement. Thus, there is a critical displacement beyond which Felec is always larger than Fspring, so that the gap will eventually close abruptly as the applied voltage across the gap is increased. This phenomenon is referred to as “pull-in,” and the voltage (VPI) at which it occurs is dependent on the switch design parameters13 VPI =
3 8 keff g0 27 ε0A a
We now discuss fabrication process innovations that were required for solution-based MEM relay fabrication. First, MEM switches require very good control of gap thickness to minimize variation in VPI; we achieved this by developing a sacrificial layer process wherein a highly uniform spin-coated organic sacrificial layer is formed over the source pad, gate, and drain electrodes to define the gap thickness prior to printing the source beam. Second, MEM switches require strong mechanical anchor formation on the source pad. We form the anchors by printing solvent to dissolve away the sacrificial material in the beam-pad overlap regions and then print metallic nanoparticle ink to fill the holes formed in the sacrificial layer. Finally, MEM switches require robust structures and a highly selective etch process to remove the sacrificial material and thereby release these structures; we achieve this by using sintered printed metallic nanoparticles to form the cantilever source beams and by etching away the organic sacrificial layer in solvent to release
(1)
where keff is the effective spring constant of the beam, g0 is asfabricated actuation gap thickness, ε 0 is the vacuum permittivity, and Aa is the actuation area (i.e., beam-to-gate overlap). To prevent the possibility of source-to-gate leakage, an insulating gate-dielectric layer covers the gate electrode. 5356
dx.doi.org/10.1021/nl4028632 | Nano Lett. 2013, 13, 5355−5360
Nano Letters
Letter
dimensions; thinner beams will collapse upon release due to stiction14 while thicker beams will have high VPI and hence have large dynamic power consumption. The thickness of the beam in this work was controlled by adjusting the number of printed layers (∼450 nm per layer) with interspersed drying steps between the layers. This multistep process ensures that thick films can be built up without any bulging and cracking from excessive solvent volumes.15 The beam was dried at 150 °C and annealed afterward at 180 °C. It was found that the beam thickness (H) must be >1.6 μm to avoid collapse or stiction upon release (Figure 2c). The anchor for the beam was then formed as follows: first, the anchor region was defined by inkjet-printing acetone to etch a hole in the PMMA16 underneath one end of the beam (Figure 2d); then, this hole was filled by printing silver nanoparticle ink, connecting the beam to the underlying source pad (Figure 2e). The use of inkjet printed solvent to form the anchor hole has some unique consequences. This is not a true “etching” process in that the resulting waste material is not removed but is rather deposited at the edge of the hole. Thus, the hole formation process results in a craterlike structure (Figure 2f,) where the boundary is thicker than the center of the pattern, due to the so-called “coffee-ring” effect.16 Since the “crater” is formed after the cantilever beam is printed and annealed, the crater does not affect the actuation gap; this is a key benefit of the two-step beam-anchor formation process used herein. Finally, the anchor hole is filled with nanoparticle metal ink and sintered to form the anchor and connect the cantilever beam to the source pad. The peak temperature for the entire process is 220 °C, which makes this device attractive for use on a wide range of substrates. Since the PVP is a dielectric material, other materials with lower cross-linking temperatures potentially could be used to reduce the peak process temperature requirements even further. The measured switch drive current (ID) versus gate voltage (VGS) characteristic is shown in Figure 3a, using the measurement setup illustrated in Figure 3c (VDD = 1, 10, 100 mV, RL = 6.74 kΩ). It shows immeasurably low IOFF (below the noise floor of the semiconductor parameter analyzer) and abrupt switching behavior; this attests to the excellent switching characteristics of MEMS switches, making them attractive for the aforementioned candidate applications of printed electronics. Note that the switch turns on when VGS increases above VPI, but turns off when VGS is lowered below the release voltage VRL. Hysteresis (VRL105 cycles.
than Fspring for values of VGS < VPI when the source is pulled in, and it is exacerbated by surface adhesive force (Fadhesion) in the contacting region. Note that hysteresis is not a problem for multiplexing/pass-gate applications such as active-matrix displays, provided that the release voltage is greater than 0 V so that the switch turns off properly, as is achieved herein. Figure 3b shows the measured switch current (ID) while the voltage on the drain electrode (VDS) is varied, for a fixed VGS. For VGS > VPI, ID linearly increases with VDS; thus, in the ON state the switch can be modeled as a simple resistor, RON (ONstate resistance). The inverse slope of the ID−VDS plot gives a value of RON ∼ 10 Ω. Considering that there is a metal−metal junction at the contact between the source beam and the drain electrode, RON can be further modeled as two resistors connected in series; RON = Rcontact + Rbulk, where Rcontact is the contact resistance at the drain contact and Rbulk is that from the rest of the switch. The sheet resistance of the printed nanoparticle film used in this work is about 0.2 Ω/□, and the switch comprises 14 squares, which results in Rbulk ∼ 3 Ω. Thus RON is dominated by Rcontact ∼ 7 Ω. To investigate the dynamic behavior of the cantilever-type switch, transient characteristics were measured. Figure 4a shows ac characteristics, where an input voltage pulse signal (frequency of 1 kHz and amplitude of 21 V) is applied to the gate and the output voltage is measured at the drain electrode, using a similar setup to that shown in Figure 3c. In the OFF state, the switch remains open and the potential at the drain electrode follows VDD; on the other hand, when the switch closes in the ON-state, the drain potential approaches that of the grounded source electrode since RON ≪ RL. There is a delay between the input and output voltage signals, which consists of mechanical and electrical delays. Mechanical delay arises from the fact that it takes time for the beam to traverse the as-fabricated contact gap in response to the electrostatic force induced by the applied gate voltage. Furthermore, electrical signals have intrinsic RC delays when they propagate
through a conducting medium. Therefore, the total delay is the sum of the mechanical and electrical delay. Figure 4b shows this turn-on delay, which is mainly mechanical delay since resistance and capacitance from the switch itself are small. As expected, this delay can be reduced by increasing the applied electrostatic actuation force on the beam, since larger Felec will result in larger acceleration causing the beam to travel faster. Specifically, turn-on delay is inversely proportional to the applied gate voltage17 in the range of 10−16 μs, depending on the gateoverdrive voltage, as shown in Figure 4c. The delay can be further reduced by scaling device dimensions, especially by reducing the contact gap (gd), and hence shortening the travel distance of the beam. Importantly, since the mechanical delay is dominant with negligible electrical RC delay, a substantial design opportunity exists to work around the mechanical delay and realize ultrafast electrical operation (we estimate the contribution of RC delay to be ∼RON(ε0Aa/g0) ∼ (10Ω)(87fF) ∼ 0.87 ps. Previous design analyses have shown that MEM switch-based circuits can be optimized by utilizing complex logic gates such that all the switches move simultaneously and only one mechanical delay is incurred per operation.17 Endurance of the MEM switches was studied using a similar test setup, by monitoring RON throughout the test. The switch can endure ∼105 on/off hot switching cycles, wherein there is a significant voltage difference between the drain and source whenever the switch is turned on, and current continually flows through the switch in the ON state without stiction- or welding-induced failure (Figure 4d). (This operating condition is harsher than in a digital logic circuit, where current flows momentarily through the switch to charge/discharge the load capacitance.) To study the mechanical properties of the printed cantilever beam, nanoindentation18 was employed. In a nanoindentation measurement, a sharp mechanically strong tip (often diamond) of a specific geometry is used to apply a controlled load onto the surface of a film such that a nanoscale indentation is formed 5358
dx.doi.org/10.1021/nl4028632 | Nano Lett. 2013, 13, 5355−5360
Nano Letters
Letter
Figure 5. Structural analysis of the printed cantilever by nanoindentation. (a) Force-displacement curve obtained by applying a nanoindenter load to the end of the cantilever beam; (inset) unloading portion of the curve is shown with effective force and distance values to extract the intrinsic properties of the printed nanoparticle film using the Oliver-Pharr model.18 (b) Schematic diagram illustrating the two different regimes where (I) the beam is free-standing and bends downward toward the drain electrode, and (II) the beam is in contact with the drain electrode and the nanoindenter tip starts to make an indentation into the surface of the beam. (c) Measured elastic modulus value of inkjet-printed silver nanoparticle film sintered at various temperatures. (d) Atomic force microscope (AFM) images of the printed nanoparticle films annealed at T = 120 and 180 °C show that grainsize increases with sintering temperature (scale bar = 300 nm).
can be estimated from the ratio of beam length (Lb) over the distance from the anchor to the center of the gate electrode (Lg), (Lb/Lg)3 ∼ (650 μm/450 μm)3 ∼ 3.0. This is within ∼10% of the ratio of measured spring constants, attesting to the accuracy of the modeling and material property estimates. In regime (II), intrinsic properties of the printed nanoparticle film can be extracted using the conventional Oliver−Pharr model18 by fitting the unloading portion of the curve, shown in the inset of Figure 5a, to the load equation: P = A(h − hf)m, where hf is final depth, h is the tip displacement, and A and m are fitting parameters. The contact stiffness S and reduced elastic modulus Er of the material can be obtained using S = dP/dh|h=hmax and Er = (π1/2)/{2[Ac(h)]1/2}, respectively, where Ac(h) is a contact area function that depends on the geometry of the indenter tip. From the curve, S is calculated to be 260 N/ m2 and Er is calculated to be 47.5 GPa, which is somewhat lower than the value for bulk silver. This can be explained by the fact that the printed nanoparticle film is still somewhat porous after annealing, resulting in reduced elastic modulus.15 Separate measurements were performed to verify this, wherein silver ink was printed on a blank substrate and sintered at various temperatures. The elastic moduli was found to increase with the sintering temperature (Figure 5c) since grain size and densification increases at higher sintering temperatures (Figure 5d). This sintering-dependent mechanical property offers added design versatility for printed MEMS; the sintering temperature of solution-processed nanoparticles can be used to tune the elastic modulus (Er), which determines the effective spring constant of the beam. Since the boundary between region (I) and region (II) corresponds to the tip displacement value at which the beam
while the applied load on the tip and the corresponding displacement of the tip are recorded. By analyzing the measured load−displacement curve, mechanical properties such as contact stiffness, hardness, and elastic modulus can be extracted. For movable structures, nanoindentation offers the additional advantage of enabling the extraction of the effective beam stiffness. We applied a nanoindenter load to the end of the cantilever beam and obtained a load−displacement curve as shown in Figure 5a. Two regimes are apparent: (I) the beam is free-standing and bends downward toward the drain electrode, and (II) the beam is in contact with the drain electrode and the nanoindenter tip starts to make an indentation into the surface of the beam (Figure 5b). In regime (I), the displacement is purely the beam displacement, which increases with the applied load. The load−displacement curve shows a linear response and the effective spring constant of the beam, keff, can be extracted from the slope to be 9.7 N/m (Figure 5a, segment a). It is worth comparing this measured value of keff to that calculated using the electrical measurement of VPI, as expressed in eq 1. The pull-in voltage was measured to be VPI ∼ 16.0 V (Figure 3a). From the fabrication and design of the switch, g0 ∼ 1.65 μm, Aa ∼ 180 × 90 μm2 (gate width ∼180 μm, beam width ∼90 μm), which results in keff ∼ 27.6 N/m. The difference in effective spring constant values, one from mechanical (keff,m) and the other from electrical (keff,e) measurements, is due to the difference in location of the applied load. The electrically applied load was distributed across a region of the beam whereas the mechanically applied load was applied only at the end of the beam. Since the effective spring constant of the beam depends on the distance from the anchor to the loading point (L),13 k ∝ 1/L3, the ratio of spring constants keff,e/keff,m ∼ 2.8 5359
dx.doi.org/10.1021/nl4028632 | Nano Lett. 2013, 13, 5355−5360
Nano Letters
Letter
(14) Bustillo, J. M.; Howe, R. T.; Muller, R. S.; Fellow, L. Proc. IEEE. 1998, 86, 1552. (15) Greer, J. R.; Street, R. A. J. Appl. Phys. 2007, 101, 103529. (16) Kawase, T.; Sirringhaus, H.; Friend, R. H.; Shimoda, T. Adv. Mater. 2001, 13, 1601. (17) Kam, H.; Liu, T.-J. K.; Stojanović, V.; Marković, D.; Alon, E. IEEE Trans. Electron Devices 2011, 58, 236. (18) Oliver, W. C.; Pharr, G. M. J. Mater. Res. 2011, 19, 3.
contacts the drain electrode, the contact gap size also can be obtained from the load−displacement curve, to be ∼1.2 μm. This is somewhat smaller than the thickness of the sacrificial PMMA, even after taking account of the difference between gd and go, because the silver nanoparticle ink is not perfectly orthogonal to PMMA and dissolves a small amount of PMMA during the cantilever printing process. In summary, we have developed a new switch technology based on inkjet-printed MEM cantilevers. The printed MEM switches offer excellent on-state and off-state characteristics and show promise for printed electronics applications. Care was taken to ensure precise control of the actuation gap, and the resulting devices showed low on-state resistance, immeasurably low off-state leakage, good switching speed, and lower voltage operation than many other demonstrated printed electronic devices.
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail: (T.-J.K.L)
[email protected]. *E-mail: (V.S.)
[email protected]. Author Contributions
E.S.P., T.-J.K.L., and V.S. designed the experiments. E.S.P. performed the experiments and wrote the manuscript. Y.C. contributed the transient characteristics measurement. All authors discussed results and edited the manuscript. Notes
The authors declare no competing financial interest.
■
REFERENCES
(1) Smaal, W.; Kjellander, C.; Jeong, Y.; Tripathi, A.; Putten, B.; van der Facchetti, A.; Yan, H.; Quinn, J.; Anthony, J.; Myny, K.; Dehaene, W.; Gelinck, G. Org. Electronics 2012, 13, 1686−1692. (2) Chung, D. S.; Lee, J.-S.; Huang, J.; Nag, A.; Ithurria, S.; Talapin, D. V Nano Lett. 2012, 12, 1813−1820. (3) Cho, S. Y.; Kang, Y. H.; Jung, J.-Y.; Nam, S. Y.; Lim, J.; Yoon, S. C.; Choi, D. H.; Lee, C. Chem. Mater. 2012, 24, 3517−3524. (4) Zschieschang, U.; Hofmockel, R.; Rödel, R.; Kraft, U.; Kang, M. J.; Takimiya, K.; Zaki, T.; Letzkus, F.; Butschke, J.; Richter, H.; Burghartz, J. N.; Klauk, H. Org. Electronics 2013, 14, 1516−1520. (5) Baeg, K.-J.; Jung, S.-W.; Khim, D.; Kim, J.; Kim, D.-Y.; Koo, J. B.; Quinn, J. R.; Facchetti, A.; You, I.-K.; Noh, Y.-Y. Org. Electronics 2013, 14, 1407−1418. (6) Jacob, S.; Abdinia, S.; Benwadih, M.; Bablet, J.; Chartier, I.; Gwoziecki, R.; Cantatore, E.; van Roermund, A. H. M.; Maddiona, L.; Tramontana, F.; Maiellaro, G.; Mariucci, L.; Rapisarda, M.; Palmisano, G.; Coppard, R. Solid-State Electron. 2013, 84, 167−178. (7) Fuller, S. B.; Wilhelm, E. J.; Jacobson, J. M. J. Microelectromech. Syst. 2002, 11, 54. (8) Packard, C. E.; Murarka, A.; Lam, E. W.; Schmidt, M. A.; Bulović, V. Adv. Mater. 2010, 22, 1840. (9) Kato, Y.; Sekitani, T.; Noguchi, Y.; Yokota, T.; Takamiya, M.; Sakurai, T.; Someya, T. Large-Area Flexible Ultrasonic Imaging System With an Organic Transistor Active Matrix. IEEE Trans. Electron Devices 2010, 57, 995−1002. (10) Nakano, S.; Sekitani, T.; Yokota, T.; Someya, T. Appl. Phys. Lett. 2008, 92, 053302. (11) Ahn, B. Y.; Duoss, E. B.; Motala, M. J.; Guo, X.; Park, S.-I.; Xiong, Y.; Yoon, J.; Nuzzo, R. G.; Rogers, J. A.; Lewis, J. A. Science 2009, 323, 1590−1593. (12) Park, E. S.; Chen, Y.; Liu, T.-J. K.; Subramanian, V. 2011 IEEE International Electron Devices Meeting (IEDM) 2011, 29−2. (13) Rebeiz, M. RF MEMS Theory, Design, and Technology; John Wiley and Sons, Inc.: Hoboken, New Jersey, 2003. 5360
dx.doi.org/10.1021/nl4028632 | Nano Lett. 2013, 13, 5355−5360
