Design and Fabrication of Addressable Microfluidic Energy Storage MEMS Device

1392 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 21, NO. 6, DECEMBER 2012

Design and Fabrication of Addressable MicrofluidicEnergy Storage MEMS Device

Victor A. Lifton, Member, IEEE, Steve Simon, Member, IEEE, Johan Holmqvist,Thorbjörn Ebefors, Member, IEEE, David Jansson, and Niklas Svedin

Abstract—Design and fabrication of microfluidic energy storagedevices that are based on the control of the liquid electrolyteinside a power cell are presented. A 12-cell array of individu-ally addressable reserve microbatteries has been built and tested,yielding ∼10-mAh capacity per each cell in the array. Lithiumand manganese dioxide or carbon monofluoride (Li/MnO2 andLi/CFx) have been used as anode and cathode in the batterywith LiClO4-based electrolyte. Inherent power management ca-pabilities allow for sequential single cell activation based on theexternal electronic trigger. The design is based on the super-lyophobic porous membrane that keeps liquid electrolyte awayfrom the solid electrode materials. When power is needed, batteryactivation (a single cell or several cells at once) is accomplishedvia electrowetting trigger that promotes electrolyte permeationthrough the porous membrane and wetting of the electrode stack,which combines the chemistry together to release stored electro-chemical energy. The membrane and associated package elementsare prepared using microelectromechanical system fabricationmethods that are described in details along with the assemblymethods. [2011-0312]

Index Terms—Array, electrowetting, lithium, membrane, mi-croelectromechanical systems (MEMS), microfluidic, reserve bat-tery, superhydrophobic.

I. INTRODUCTION

ENERGY storage devices are a vibrant and widely re-searched topic that spans photovoltaic and grid storage

batteries, batteries for hybrid electric vehicles, portable elec-tronics devices, and, recently, for various sensors and unat-tended sensor networks [1]–[4]. Obviously, the requirementsfor each application differ vastly, and that stimulates the de-velopment of many types of energy storage devices based ona multitude of chemistries, as well as fuel cells and super-capacitors. Each type of energy storage technology has manyvariations, developed for a specific function and performance.

Microelectromechanical system (MEMS) devices and sen-sors are receiving a world-wide attention for their promise as

Manuscript received October 22, 2011; revised March 22, 2012; acceptedJune 24, 2012. Date of publication August 1, 2012; date of current versionNovember 27, 2012. Portions of this work were accomplished under a PhaseII STTR DoD award; the views expressed in this paper are not intended as anendorsement by the funding agency. Subject Editor A. M. Shkel.

V. A. Lifton and S. Simon were with mPhase Technologies, Inc., Little Falls,NJ 07424 USA (e-mail: [email protected]).

J. Holmqvist, T. Ebefors, D. Jansson, and N. Svedin are with Silex Mi-crosystems AB, 175 26 Järfälla, Sweden (e-mail: [email protected];[email protected]; [email protected]; [email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JMEMS.2012.2208218

the enablers of the next generation devices and services. Thediversity of MEMS devices is at least as great as the diversityof the energy storage devices [5], [6]. A subclass of MEMSdevices that uses various forms of transduction to move andmanipulate fluids on micro and nanoscale is called microfluidicdevices [7], [8]. These devices can move droplets over certaindistances, dispense, mix and analyze liquids, and perform a hostof other functions.

Electrochemical batteries often contain liquids in the formof electrolytes. The electrolyte is normally added to the activeelectrode materials at the manufacturing stage, and that leadsto a commonly observed reduction in the capacity of a batteryover time due to unwanted, uncontrollable chemical reactionsbetween the electrodes and the electrolyte, coupled with theelectrolyte degradation and evaporation. Therefore, even whenbatteries are not in use or in storage, they are losing capacityand may become too depleted when they are deployed. Weset out to add the functionality of the microfluidic devices tothe electrochemical batteries to impart the ability to meter,valve, and manipulate its electrolyte on demand using a specialtype of batteries, called reserve batteries, as our inspiration.Reserve batteries address the unwanted degradation when notin use or in storage. They are designed so that the electrolyte isstored separately from the electrode materials in a hermeticallysealed vial or bellow [1]. When electricity generation is needed,the vial is broken which releases the electrolyte and forces itinto the electrode space to complete the battery and to beginelectricity generation. In many applications, a long life batterythat guarantees that power will be available when the enduser or the application needs it, a reserve battery, becomes anattractive or even a must-have option.

In the sections to follow, we will describe a particulartype of the microfluidic battery that uses a superlyophobic1

membrane with electrically controlled permeability to separateliquid electrolyte from the electrodes. When power is needed,the battery can be manually or remotely activated, for example,by an electronic signal from an external circuit to combine thechemistry together and initiate the electrochemical reaction forenergy generation.

Conceptually, we envision a multi-cell battery where the usermay trigger (turn on) only a subset of cells contained within a

1The term “superhydrophobic” is used when we discuss a membrane thatrepels aqueous solutions, the term “superlyophobic” is used when referring tothe membranes repelling not only high-surface-tension (aqueous) liquids, butalso low-surface-tension organic liquids such as the electrolytes employed inLi-based batteries.

1057-7157/$31.00 © 2012 IEEE

LIFTON et al.: DESIGN AND FABRICATION OF ADDRESSABLE MICROFLUIDIC ENERGY STORAGE MEMS DEVICE 1393

single battery pack. By addressing the appropriate cell within abattery, the user will move the appropriate electrolyte into theappropriate cell and trigger it. Moreover, individual cells can bewired into an addressable matrix to connect cells in parallel orin series depending on the load and power requirements.

We have previously reported on two different approaches tobuilding energy storage devices using MEMS technology. Oneapproach is based on superhydrophobic surfaces (“nanograss”)where a Zn/MnO2 reserve battery has been demonstrated [9].Several shortcoming of this design (such as limited amountof active electrode materials and complex processing) havebeen resolved in the second approach based on the superhy-drophobic porous membrane with controlled permeability andZn/MnO2 chemistry [10], [11]. In this design, the electrodeswere prepared on a separate substrate, while the electrolyte wassupported by a porous membrane. The membrane has tunablepermeability that can be changed from complete repellencyof the electrolyte, to complete permeation of the electrolytethroughout its thickness. Once the electrolyte starts permeatingthrough the membrane, it will be wicked from the top surfaceto the bottom surface and exit the pores. Typical capacitiesachieved by using electroplated electrodes were in 200 to400 μAh/cm2 and were limited by the thickness of Zn andMnO2 films. Both designs are based on aqueous Zn/MnO2

chemistry that suffers from the low energy and power densitiesbut is easy to handle in open air.

To address the issue of increasing the energy density,we attempted to build a MEMS-based battery using Libattery technology (using Li-based electrodes and organicelectrolytes). In our prior publication [12], we described amethod of preparation of the superlyophobic membrane only,and in this manuscript we summarize our work on full batteryfabrication and assembly, including preliminary electricalcharacterization data.

During inactive state of the cells, the electrolyte does notcome into contact with the electrodes; therefore, no electro-chemical reactions that lead to self-discharge can occur nor cancorrosion occur to reduce the amount of active materials. Thisis the key feature of all reserve battery designs including theone presented in this manuscript.

II. EXPERIMENTAL

A. Overall Battery Construction

As an extension of our earlier concept of using a porousmembrane as a microfluidic element [12], we designed amulti-cell membrane structure that results in a battery with12 cells, each cell acting as an individual microfluidic battery.It uses Li/MnO2 or Li/CFx battery chemistry, given that bothchemistries can be implemented using solid Li foil as an anodeand cast tape of MnO2 or CFx and liquid electrolyte. Forthe electrolyte, we chose to use 1M LiClO4 in 1 : 1 ethylenecarbonate and dimethyl carbonate.

The design of this microfluidic battery is based on therequirements of our target application (such as memory backup)and is geared toward very low discharge currents (less than1 μA) over very long periods of time. For example, if each

reserve cell in the 3 × 4 cell array (12 cell in total) lasts for3 years, based on the anticipated electrical load of this specificimplementation, it would provide a backup for up to 30 years ofuninterrupted service, whereas a conventional primary coin cellbattery with the same amount of active electrode materials willnot survive that long due to self-discharge and corrosion. A lifetime of 3 years is not unexpected given that many conventionalbatteries (e.g., coin or button batteries) may have a projectedlifetime of 5–10 years. The main feature of our design is theability to sequentially trigger or “turn-on” a single cell ondemand while keeping the rest of the cells in the inactive “inert”state, without self-discharge, thus, assuring the lifetime of 20years or longer, of the time deferred, activation of remainingcells of the multi-cell-based implementation. Such lifetime isunattainable in conventional batteries.

The battery consists of five main layers [Fig. 1(a)]. Theporous membrane is located in the middle of the stack. Itserves to separate liquid electrolyte from the electrode ma-terials. Backside metallization on the membrane serves as acurrent collector for the cathode [positive terminal of a battery,Fig. 1(a) and (b)]. The top side of the membrane is bonded to aglass reservoir layer with the electrolyte wells that contain theelectrolyte in the reserve (inactive) state. The glass reservoirlayer is bonded to the top cap layer. The top cap serves thepurpose of capping (sealing) the electrolyte wells to preventtheir contents contacting with the environment. This structurealso contains fill holes used to fill the assembled battery withthe electrolyte. Top side metallization on the top cap containscontacts to the trigger circuitry for battery activation [Fig. 1(b)].The electrodes are contained in the bottom structure that isformed by bonding a glass grid (fourth layer in the battery) andthe bottom cap structure. The glass grid defines the size of theelectrode cavity, while the bottom cap seals the electrode mate-rials and provides for the negative terminal of the battery (anodecontact). The cathode is tape cast MnO2 or CFx and the anodeis Li foil. To prevent shorting, the electrodes are separatedfrom each other by a layer of a battery separator (either glassfiber filter material APFC by Millipore or Celgard materialby Celgard). Top and bottom cap silicon wafers use “throughsilicon insulator (TSITM)” technology to dielectrically insulateindividual cells from each other.

A schematic process flow for the main elements of the batteryconstruction (membrane, top cap, and bottom cap) is presentedon Fig. 2. The accompanying legend identifies the processingsteps to help guide the reader in the description that follows.

B. Superlyophobic Membrane Design

In order to extend our microfluidic technology into Li-basedchemistries, we need to replace cathode/anode/electrolyte com-bination (e.g., replace Zn/MnO2 electrode pair and aque-ous ZnCl2 electrolyte with Li/MnO2 electrode pair andLiClO4 organic electrolyte). A more common electrolytebased on LiPF6 was found to be detrimental to the sta-bility of the silicon membrane and its elements. We spec-ulate that hydrofluoric acid formed as a result of LiPF6

hydrolysis and dissociation is responsible for etching SiO2

films deposited on the surface of the membrane. A fur-


Fig. 1. (a) Pictorial cross-section through the battery showing its components (layers), their thickness, and bonding methods used during assembly. (b) Electrolytemovement and vapor displacement during activation of the battery. Battery terminals as well as trigger contacts are shown.

ther complication arose when it was determined that theexisting porous membrane structure would not be able tosupport the electrolytes used in Li battery technology withoutletting them spontaneously go through the pores. The reason for

this behavior is the electrolytes’ low surface tension, roughly30–40 mN/m versus 72 mN/m for aqueous electrolytes previ-ously used in Zn/MnO2 battery. The solution was found in anovel approach in making universally repelling surfaces that not


Fig. 2. Simplified process flow of the three main elements of the battery: superlyophobic membrane, top cap, and bottom cap waters. All elements are madeon Si wafers. The insert shows that a fine-tuned deep reactive-ion etching etch produce the overhang nanofeature responsible for the superlyophobic behavior ofthe membrane. Process flow for the glass-based elements (electrolyte reservoir and bottom grid) is not shown as their processing involves only through waferultrasonic machining of glass wafers.

only repel water-based liquids but also organic liquids such asoils and alcohols. By using an approach developed in our workon “nanonails” [13], we designed porous structures geomet-rically similar to the porous membranes described earlier butwith an addition of a nanoscale feature along the inner edge ofthe pore (overhanging into the pore) [12]. Such nano-sized fea-tures enable universally repelling behavior (sometimes calledsuperlyophobic, superoleophobobic, omniphobic). Discussionof the effects of the re-entrant surface structures on the wettingbehavior is beyond the scope of this paper, and the interestedreader should consult [12]–[14].

Overall, we designed and fabricated Si-based membranesthat successfully supported low-surface-tension liquids that canbe used as solvents for the electrolytes in Li-based batteries.The details of the membrane fabrication have been described inour previous publication and on Fig. 2.

We created a process based on the regular (not silicon-on-insulator) Si wafers that produces 300-um-thick membranes,with the die size 31 × 33 mm and multiple porous regions(5 × 5 mm each), on the same die (Fig. 3). The size of theporous region is not a limiting factor and in our work; itwas dictated by the footprint required for our target applica-tion. As was established in our previous work [12], overhangs100–300 nm long give the optimum combination of the stabilityagainst self-triggering (spontaneous membrane wetting by theliquid sitting on top) as well as voltage required for the elec-trowetting transitions.

Fig. 3. Optical photo of a fully processed 6-in membrane wafer and a closeup of a single 3 × 4 multi-cell membrane chip.

When treated with an appropriate hydrophobic coating, aproperly etched Si membrane demonstrates superhydrophobicbehavior that distinguishes itself from a regular surface bysubstantially higher contact angle of a liquid on such a surface(e.g., 150◦ versus < 90◦ for water on a flat surface treated withthe same hydrophobic coating). Various hydrophobic coatingscan be used to render the membrane superhydrophobic andare effective at repelling liquids placed on it, without liquidpenetrating inside the pores. We have successfully used vapordeposited coating such as CFx [15], silane-based coatings, flu-oropolymer dip coatings such as Teflon (Du Pont) and CYTOP(Asahi Glass Co., Ltd), as well as liquid coatings obtainedfrom Cytonix LLC and vapor-deposited coating, Repellex, fromIntegrated Surface Technologies, Inc. In the majority of the testsdescribed here, we used 1% Teflon AF solution to prepare ahydrophobic coating of choice. Reliability and stability data


of such coatings on Si-based membranes have been reportedpreviously [12].

Electrowetting gives the ability to change the contact angle ofthe solid–liquid–vapor interface by applying voltage betweenthe liquid and the substrate. With the application of a voltagepulse, the liquid begins to wet the surface, collapses, and perme-ates into the porous space. Previously, electrowetting has beensuccessfully applied to create a variety of optical devices suchas lenses, diffraction gratings, and drug delivery devices (so-called lab-on-a-chip devices [16]–[20]. An important feature ofelectrowetting is that it is based on the capacitive charging ofthe liquid–solid interface, and hence no direct dc current flowsthrough the device upon trigger (limited to a small leakagecurrent in the nA range for the duration of the trigger pulse of afew hundred milliseconds); therefore, only a very small amountof energy is expended during trigger process.2

To trigger an electrowetting transition, one contact is made tothe electrolyte itself using a contact point made in the top cover[Fig. 1(a) and discussed later in Section II-F], and the secondcontact is achieved via a backside metallization to the backsideof the membrane [Fig. 1(b)]. In this particular implementation,the trigger voltage was supplied via a power source, mimickinga primary source of power that senses when it reaches its end-of-line and switches over to the reserve power source. We canalso visualize our reserve battery being triggered by an externalevent such as a vibration event disturbing a piezoelectric sensorgenerating voltage for trigger.

C. Membrane Fabrication

The final membrane design consists of an array of hexagonalpores 30 μm in diameter, 300 μm deep, separated by the solidwalls 15 μm thick (Fig. 3). The process is based on the throughwafer deep reactive-ion etching (DRIE) etching and it starts offon a pre-thinned 4 or 6-in wafer with 300 μm thickness [12].The choice of this particular thickness is rather arbitrary and isbased on the mechanical robustness of the wafer to make sureit can be easily handled during processing and that the finaldevice can be handled during post-clean room processing andintegration into the reserve battery.

First, a layer of silicon nitride is deposited across the wafer(280 nm thick); it will later serve as the structural layer toform the overhang. A lithography step defines the pores, thesize of the porous area (12 squares 5 × 5 mm in this partic-ular design), and the overall dimension of the membrane chip(31 × 33 mm). SiN layer also serves as a hardmask for theDRIE through-wafer etching. An RIE step removes the exposedSiN layer and is followed by the DRIE processing. This stepforms two important features, the pores and the overhang. Thepores are formed as a result of the through-wafer etch, and thesilicon nitride overhang is formed as a result of controlled un-dercutting of the silicon nitride layer on the Si wafer (that is, Sisubstrate immediately underneath SiN is etched away to a cer-tain depth and width, creating an overhanging SiN structure).Depending on the length of time for the etch and careful cycle

2In our previous work [10], we estimated that the trigger process uses∼60 nJ of energy, whereas a cell may contain at least 1 J or more.

control of passivation/etching, overhangs of various lengths canbe formed. DRIE etch development yielded a stable recipe en-abling creation of SiN overhangs uniformly across 6 in waferswithin 100 nm of a targeted dimension, typically 250–350 nmfor a targeted 300-nm undercut. The DRIE etch developmentand processing was monitored, evaluated, and verified byscanning electron microscopy (SEM) cross-sectional and top-view inspection of samples from wafer edge and wafer centerlocations, respectively.

A thermal oxidation step is subsequently performed to growa conformal SiO2 layer required for the electrowetting transi-tions. A blanket 2500-A Au/200-A TiW sputter deposition stepon the backside of the membrane wafers, intended as a commoncurrent collector for the battery cathode component, completesthe membrane wafer processing. Similar metallization processand materials have been employed to create seal rings onthe front side of the membrane used for thermocompressionbonding (described in Section II-H).

D. Electrolyte Reservoir and Bottom Glass Grids

Both elements are made out of borosilicate glass [layers“glass reservoir” and “bottom glass” on Fig. 1(a)] machinedout of 6-in glass wafers 3 and 1.5 mm thick, respectively. Thecavity machining is done using so-called ultrasonic machiningprocess. To enable wafer level thermocompression bonding(described in the later section), each glass wafer features anelectroplated and patterned Au seal ring layer. Each cell is sep-arately enclosed by a seal ring thereby enabling separate func-tionality of each one of the 12 battery cells in the 3 × 4 array.

The electroplating is performed onto a patterned seed layerprepared by sputter deposition of Au layer, followed by thephotoresist deposition and patterning. The patterned photoresistfilm acts as a mold during the Au electroplating step. Theoutlines of the parts are given in Fig. 4. Their thickness andthe internal cavity dimensions are determined by the amount ofthe electrolyte required. In this particular design, 60 μL is theminimum required, and we designed it for 100 μL (60 + 40 μLintentionally overfilled to account for the electrolyte trappedinside the pores of the membrane and the separator). Theamount of active electrode materials and consequently theirdimensions are based on balancing the amount of materialsparticipating in the electrochemical reaction.

E. Bottom Cap

This element is made out of a low resistivity Si wafer(< 0.0015 Ohm cm). It consists of 3 × 4 array of rectangularstructures that are dielectrically separated from each otherusing TSITM technology developed by Silex Microsystems[Fig. 1(a)]. The purpose of this array is to create 12 individualpower cells in a single battery package, so that each cell can beindependently addressed (triggered) and used.

Each cell has a metal contact on the inside (Ti) that will be incontact with anode material (Li foil) and a metal contact on theoutside (TiW) that is used as a negative terminal of the battery.Conductive Si substrate proves an electrical path between thetwo contacts [Fig. 1(b)]. The inside Ti contact is also used as


Fig. 4. Images of the glass reservoir and bottom grid. Reservoir size: 31×33× 3 mm, 5.5 × 5.5 mm—electrolyte chamber, 5.5 × 1 mm vent chamber.Bottom glass size: 31× 31× 1.5 mm, 9 × 6 mm—electrode cavity.

a diffusion barrier to prevent Li diffusion into the Si substratethat is known to induce very large volumetric expansion in thehundred percent range that would destroy the part. We usedcyclic voltammetry to establish that 1 μm thick coating of Ti issufficient to prevent Li diffusion into the underlying substrate.

Formation of the electrically insulated from each other con-ductive through-silicon contacts was accomplished using socalled “via trench” (Sil-Via®) process technology developedby Silex Microsystems.3 It utilizes DRIE etching for trenchcreation and subsequent dielectric low-pressure chemical vapordeposition to fill the trenches with dielectric. The concept con-sists of a closed loop of isolated vias, forming dielectric-filledisolating chains (Zero-CrosstalkTM) around the respective cellswithin the wafer.4 Once processed, bottom cap wafers wereanodically bonded to the bottom grid glass wafer.

F. Top Cap

Similar to the bottom cap, it is made out of a conductive Siwafer, separated into a 3 × 4 array of cells by TSI technology(Fig. 1). Each cell has its own fill hole for the electrolytefilling, created by through wafer DRIE etching that is sealedwith a layer of polyimide tape and epoxy in this particularimplementation [Fig. 1(b)]. In the future work, we plan toreplace the tape/epoxy seal with a truly hermetic soldered sealto prevent electrolyte evaporation and interaction with the air.Each cell also features a TiW contact on the inside and theoutside of the battery (front and backside side of the top capwafer), which are connected by the conductive through siliconvia [Fig. 1(a)]. They are used as the trigger points to the elec-trolyte to induce electrowetting transitions. In addition, the topcap design includes a microchannel system, specifically aimed

3http://silexmicrosystems.com/docs/developing_the_mid_end_foundry.pdf4http://www.silexmicrosystems.com/docs/Sil-ViaWithZeroCrosstalkFeature-

FirstHighVolumeViaProcessForPackagingAndIntegrationOfMEMS&CMOS.pdf

at facilitating vapor recirculation inside the battery package(described in more details in Section II-G below).

G. Vapor Recirculation Inside of the Battery Package

An important realization is that the vapors or inert gascontained within the battery after it is fully assembled mustbe displaced when the battery is triggered and the electrolytebegins to flow through the membrane [see Fig. 1(b)]. Our solu-tion was to incorporate special air-recirculation microchannelsthat were created for this purpose. This design is based on themicrofluidic principles which help identify the size of a porethat prevent fluid filling it while letting the air percolate throughit. We created an array of microchannels embedded into thecover (top cap) of the battery that caps the electrolyte reservoir.The reservoir consists of the ultrasonically machined glass withone wall separating the reservoir from the air recirculationchamber. The two chambers are connected with the microchan-nels that are covered on the top by a glass wall. The channelsare made longer than the width of wall to make sure theyare in contact with the internal atmosphere in the recirculationchamber or electrolyte in the reservoir. Since the channels arelater treated with the hydrophobic Teflon coating, they becomesuperlyophobic and will not allow the electrolyte to fill thechannels; therefore, only the inert gas and/or electrolyte vapors(internal atmosphere of the battery) are allowed to move fromone chamber to another.

To test this design on a Si wafer, we created a matrix ofmicrochannels with various width/length/depth for characteri-zation (width 2, 3, 5 μm; length 1.6, 1.7, 1.9, 2.5 mm, depth0.5 μm). The microchannel system was patterned usingpositive-tone photoresist in an ASML PAS 2500/40 step-per. The silicon was then RIE etched to a depth of 0.5 + /−0.1 μm. The wafer was bonded to a glass wafer machinedto form both the electrolyte reservoir and the recirculationchamber. For testing, we used two specific liquids, one wasthe electrolyte of choice (surface tension ∼30–35 mN/m), andthe second liquid was Fluorinert FC-84 by 3M, with surfacetension of ∼16 mN/m. Both liquids contained a UV fluoresc-ing dye to aid in flow visualization. Our estimates show thatFC-84 will penetrate and fill the microchannels regardless oftheir superlyophobic properties, since the surface tension in thiscase is too low and it is not supported by the surface structureof the microchannels. The electrolyte, on the other hand, hassufficient surface tension to be supported by the microchannelswithout penetration into the microchannels.

In a typical test, we filled the reservoir with FC-84 liquidfirst and observed how it moves across the microchannels byviewing it with the microscope through the transparent glasswall covering the microchannels. If the channels are pluggedby the Teflon coating or pinched off during anodic bonding,the liquid will not move across them, regardless of the surfacetension. On the other hand, if the channels are open, low-surface-tension liquids will move across them due to capillaryforce. As expected, FC liquid had no trouble wetting the mi-crochannels and passing across the channels from one chamberto another. The results of these initial tests showed that themicrochannels were not plugged by the dust particles, silicon


saw dust, or Teflon itself, and it was an important validationtest to ascertain that if we observed no electrolyte permeationthrough the microchannels, it was due to the intended functionof the microchannels and not due to the plugged pores.

Since we now knew that microchannels were not obstructed,we cleaned the structures and re-tested them using LiClO3

electrolyte as a test liquid. The cleaning was performed ina succession of ethanol and water baths. If the microchan-nels functioned as they should, no electrolyte flow across themicrochannels should be observed. We have tested multiplesamples from various wafers, corresponding to various depthsof the microchannels and have not seen any failure of themicrochannels. Only in one cell, on one sample, did we observeunwanted leakage across the microchannels after repeated test-ing and cleaning operations. We suspect this isolated incidentoccurred as a result of the failure of the Teflon coating due to therepeated exposure to various solvents and elevated temperatureduring drying cycles that occurred between the tests.

Overall, we believe these tests demonstrate that our approachat solving vapor displacement issue in the reserve microbatteryis a viable approach and will be implemented in final batteryassembly (w = 2 μm/l = 1.7 mm/d = 0.5 μm).

H. Wafer-Level Bonding Assembly

Anodic and thermocompression bonding have been used toattach various layers of the battery in order to achieve hermeticpackage. The need for such package is to prevent electrolyteevaporation, moisture absorption and to prevent Li chemicalreactions with water, oxygen, and nitrogen from the air. Metal-to-metal or glass-to-silicon bonds are known to be hermetic ornearly hermetic, and both are excellent choices for the batteryassembly. Fig. 1(a) shows battery layers and lists methods usedin assembly.

• The top side of the reservoir is bonded to the top cap usinganodic bonding.

• The top of side of the membrane is bonded to the backsideof the reservoir via thermocompression bonding.

• The top of the bottom grid is bonded to the metalized back-side of the membrane via adhesive bonding (described inSection II-J).

• The bottom side of the bottom glass grid is bonded to thetop side of the bottom cap via a modified anodic bondingprocess.

The choice of the bonding method is obviously related tothe type of materials being bonded. When clean surfaces of Siand glass were present, we used anodic bonding [21]; in othercases, we used thermocompression bonding to attach metallizedlayers to each other [22]. Such process was accomplished bydepositing Au seal rings on one of the parts and blanket met-allization of the counter surfaces. Wafer-level anodic and ther-mocompression bonding and bond alignment were performedusing commercial wafer bonding and bond alignment systems(EVG540 (bonding) and EVG610/620 (aligning), respectively.Thermocompression bonding on the wafer level was typicallyconducted at elevated temperatures (∼400 ◦C) and ∼4-kNapplied force.

Two wafer-level sub-assemblies have been prepared: “top cap/reservoir/membrane” and “bottom cap/glass grid” [Fig. 1(a)].In the next step, we would populate each cell with the electrodematerials and bond the two sub-assemblies together. However,given extreme reactivity and special handing procedures forLi, a MEMS foundry is normally not equipped to handle it.Therefore, the sub-assemblies have been diced into the indi-vidual chips and assembled on the chip level in a custom-builtbonder suitable for anodic, thermocompression, and adhesivebonding in a dry glove box, located in a dedicated Li processingfacility.

When the membrane was fabricated and integrated intothe sub-assembly “top cap/reservoir/membrane,” it is not su-perhydrophobic or superlyophobic. It must be treated with ahydrophobic coating in order to achieve this functionality.Therefore, after the sub-assembly has been prepared, we per-formed dip coating in 1 wt.% solution of Teflon AF 1600in Fluorinert FC-84 (perfluorinated liquid made by 3M). Dipcoating was performed using a bench-top dip coater, TL0-01(MTI Corp), at 0.1 mm/s pulling speed. The Teflon films werethen dried in an oven at 200 ◦C for 2 h in nitrogen atmosphere.A thin, uniform coating of approximately 100–500 nm wasproduced on the surface of the sub-assembly and the membrane.We have investigated dip-coated membranes in SEM and havenot seen any evidence of pore closing or narrowing as a resultof the polymer coating the inner walls of the structure.

I. Chip-Level Assembly

The final sealing step, after each cell in 3 × 4 array is pop-ulated with the electrode materials (cathode/separator/anode),was originally intended to be a thermocompression bondingbetween the bottom of the membrane and the top surface of thebottom glass grid. However, conventional thermocompressionbond performed at ∼400 ◦C is not possible as it would exceedthe melting temperature of Li (180.5 ◦C). A possible solutionto this limitation would be to use Li alloys such as Li–Sn,Li–Si that have melting temperatures above Li and therefore,more compatible with the bonding procedure. Such anodes arewell known and certainly offer a path forward. However, hightemperature may damage not only Li but also the cathode tapeas well as Celgard separator used in the battery. Therefore,we decided to explore the viability of reducing the assemblytemperature for thermocompression bonding to 150 ◦C.

We ran a series of tests preparing assemblies at varioustemperatures between 150 and 400 ◦C while checking thebond strength. However, we found that even 8 or 12 h longexperiments at 150 ◦C did not produce a strong enough bond.For example, small backpressure provided by the compressedelectrode stack inside of each cell that is intended to be kept un-der compression to assure electrical contact is enough to breakthe thermocompression bond. Therefore, in the short term, wehad to find a suitable replacement for the thermocompressionbond. We decided to use adhesive bond formed by Surlynpolymer by DuPont Co. Surlyn is well known in Li batterymanufacturing, as a resin that is often used to construct flexiblepouch-style batteries. It is chemically inert in the electrolyte andadheres strongly to metal and oxide surfaces.


We must point out that the use of the adhesive bond is only anintermediate step in our project while we are working to resolvethe thermocompression bond issues, by either implementinghigher-melting-temperature Li alloy anodes or by developinga low-temperature bonding process [23].

J. Surlyn Bonding

Overall, the assembly using Surlyn can be viewed as a typeof adhesive bonding [24] that in terms of equipment used isvery similar to the thermocomporession bonding, except that agasket made of Surlyn is placed in between two sub-assembliesprior to heat/pressure application. A sheet of Surlyn was cutinto the shape of the 3 × 4 grid to match the pattern of thesolid walls of the bottom glass grid and placed on top of thebottom assembly (pre-populated with the electrode stack). Itwas then covered and aligned with the reservoir/membrane sub-assembly and the bonding began by applying 25 psi pressureand 150 ◦C for 30 min. The resulting bond is strong enoughto survive manual handling during electrolyte filling andsealing.

Once the bond was formed, the battery was filled with theelectrolyte using a syringe (100 μL in each cell), coveredwith a polyimide tape and bonded to a printed circuit board(PCB) using silver epoxy. The PCB served two purposes: oneto provide electrical connections for triggering and testing andanother to provide the template for silver epoxy to preventshorting. Silver epoxy was applied as a puddle to completelycover the polyimide tape and the fill hole. Once the assemblywas air-cured, it was ready for testing. Note, that all of thebonding/filling/sealing operations have been conducted insidethe glove box purged with dry argon.

K. Electrode Stack

We created the electrode stack by manually inserting indi-vidual electrode elements into each cell of the 3 × 4 arraywhile it was positioned on the bonder tool. The ordering of thelayers was as follows, from the top down: CFx (680–705 μm)/Glass Separator (250 μm)/Lithium foil (300 μm)/Coppercurrent collector (25 μm)/Stainless steel compression spring(150 μm).

The functionality of the stainless steel compression spring(Snaptron part # F08150) was to apply upward force on theelectrode stack to keep the CFx electrodes in contact with thegold metallization on the bottom side of the porous honeycombmembrane. The gold contact on the membrane is designed tobe the common positive current collector for each of the cells.In addition, the compression spring is intended to compensatefor the reduction in size of the CFx material as it is consumedduring discharge, so that it remains in contact with the commongold contact on the membrane.

L. Battery Testing

After the battery was assembled and attached to PCBs, it wasready for triggering and discharging tests. The triggering hasbeen accomplished by applying a 90-V, ∼0.5-s long voltage

Fig. 5. Discharge of a single cell in 12-cell array in one of the fully assembledbatteries. Capacity of ∼10 mAh per cell is achieved. Overall, battery size is31 × 33 × 5.6 mm.

pulse to the desired cell. Activation has been observed by con-necting the cell to the digital voltmeter to monitor voltage riseto 3–3.4 V. Once the cell has been triggered, it was connectedto the Maccor tester to discharge at a known rate and to recordthe data as a function of time.

A typical discharge curve for one cell is shown in Fig. 5.The total capacity is ∼640 mAh/g of CFx (corresponding to∼10 mAh capacity per each cell) under ∼130 μA constantdischarge current. The total capacity of a cell is limited by theamount of the cathode material, CFx. Given that each cell is areplica of its neighbor, we only present data for one such celldischarge.

In a control test of a coin cell assembled with the samecathode and anode materials, we obtained the capacity of∼800 mAh/g of CFx, which is very close to the theoreticalvalue (∼850 mAh/g). The discrepancy may be explained bythe electrolyte starvation in the cell as not all electrolyte mayhave transferred through the membrane. We have been able totrigger and discharge multiple cells, with the yield of about50%. While such yield is certainly not acceptable from thetechnology implementation point of view, we remind that thiswork is a proof-of-concept demonstration. We are currently inthe next phase where we will implement several design changesto enable better performance of the battery.

An initial drop of the output voltage in Fig. 5 is related tothe electrolyte wetting and spreading inside the cell. Until theelectrodes are fully wetted, they may not be capable of carryingthe entire discharge load and the output voltage drops. Once theelectrodes are fully wetted, the voltage recovers to ∼2.5 V.

M. Wicking Problems Inside the Cell

In our tests we observed a much longer than expected ac-tivation of the cell (∼3–5 min versus expected 0.5–1 min).Such long trigger times are unacceptable for a real-time backupbattery, where a ramp up to full power should occur withinseveral seconds.

Several conditions have to be satisfied for the quick andsuccessful battery activation. First, the porous membrane hasto undergo an electrowetting triggering, second, the electrolytehas to be wicked quickly from the reservoir through the


membrane and through the electrode stack to fully wet theelectrodes, and third, a good mechanical contact has to existbetween the backside metallization of the membrane and thecathode and between the anode and the metallization inside thebottom cap to be able to discharge the cell.

We can eliminate the third condition as it is clearly satisfiedgiven the we did observe open-circuit voltage after a periodof time and have been able to discharge the cell to achieve∼80% of the theoretical capacity. In all, we believe that goodelectrical contact is maintained through the cells. The firstcondition can also be tested prior to the assembly by performingelectrowetting (EW) tests on the freestanding membranes. Insuch a test, “top cap/reservoir/membrane” chip is filled the elec-trolyte, placed on top of the filter paper (wick) and with theproper electrical connections made to the membrane and theelectrolyte, EW transition is induced. We have tested multiplesamples, and a consistent EW has been recorded at 90 V.Therefore, we conclude that EW transition is not responsiblefor the slow battery activation. This leaves us with the slowwicking inside the cell upon EW transition once the electrolytepermeates through the membrane and begins to fill the electrodecavity.

Once the electrolyte has penetrated through the membraneand reached the exit side of pores, it comes in contact with theCFx tape that acts as a “wick” that distributes the electrolyteinside the cell. We can speculate that the residual porosity in thetape material is not sufficiently high to promote fast wicking.Therefore, slow ramp up to full power may be attributed tothe slow wicking inside the cell. Potential solutions mightbe to make the cathode tape more porous or to integrate awicking material into the tape, for example, by laminating thetape to the glass fiber material or Celgard separator. However,both approaches would come at the expense of decreasing thevolumetric energy density of the cathode.

It can be noticed that when our reserve cell is assembled andsealed with the electrolyte inside, metallic Li anode will be ex-posed to the electrolyte vapors that can permeate freely throughthe porous membrane, and it may potentially lead to anodedegradation due to unwanted chemical reactions with vapors(so called gassing). However, we have not found evidence ofthe capacity reduction. We believe the key is that the reserve cellprevents self-discharge and electrochemical reactions betweenthe anode and cathode because only solvent may form thevapor and no salt evaporation occurs, therefore, no “electrolyte”formation is expected inside the cells as a result of evaporation.In addition the surface layer of Li anode will react with thevolatile solvent species to form a somewhat self-limiting solidelectrolyte interphase layer.

III. CONCLUSION

We introduced a concept of a microfluidic electrochemi-cal energy storage cell (battery), where control over liquidelectrolyte is achieved by combining superlyophobic MEMSstructures and electrowetting transitions. Several examples ofmicrofluidic reserve batteries that use superhydrophobic sur-faces and electrowetting to manipulate liquid electrolyte andto switch a dormant battery into the active state have been

demonstrated. Zn/MnO2 and Li/MnO2 batteries have been builtand tested using two different micro and nanostructures. In onedesign, the liquid is kept separate from the electrodes in theinactive state by a dense forest of nanoposts. In the secondapproach, a porous membrane with controlled permeability isused to suspend the liquid above the electrode materials. Simplevoltage trigger using electrowetting phenomenon can switchthese reserve batteries into active power generation. A multi-cell (3 × 4 array) battery is described, and its fabrication pro-cess is outlined. In such configuration, a single cell or severalcells can be triggered to provide required levels of power.Using external wiring, parallel or serial connectivity is alsopossible for flexible power management. We envision that in anadvanced configuration, a multi-cell battery can contain cellswith various electrode/electrolyte combinations to give rise toa unique battery performance that is tuned to environmentaland load conditions. All of the aforementioned functionalityis only possible because of the precise microfluidic controlover the electrolyte in each cell of the battery. Given that thetriggering is accomplished by a voltage pulse, it is easy toimplement by a simple logic monitoring circuitry that can bepreprogrammed for triggering based on a voltage threshold,elapsed time, environmental conditions, and other parameters.We created a process flow for fabrication of the individualcomponents of the battery and successfully assembled themto obtain a fully functional reserve microbattery. Normallyincompatible with Si MEMS wafer level processing, metalliclithium has been successfully integrated into MEMS deviceson a chip level. It was determined that 1 μm thick layer of Timetallization functions as an effective lithium diffusion barrierpreventing silicon lithiation and associated volume expansionand cracking. Surlyn resin has been used as an adhesive layerbetween two sub-assemblies enabling battery assembly at lowtemperature of 150 ◦C thus, avoiding lithium melting and cath-ode tape damage. Further design refinements and improvementshave been identified for the implementation in the next round ofinvestigations.

ACKNOWLEDGMENT

Continued support by mPhase Technologies is greatlyappreciated. Sil-Via® is a registered U.S. trademark of SilexMicrosystems. TSITM and Zero-CrosstalkTM are trademarksof Silex Microsystems. Lithium battery chemistry support viaa collaborative research agreement with Rutgers University(Energy Storage Research Group) is greatly appreciated.

REFERENCES

[1] D. Linden and T. B. Reddy, Handbook of Batteries, 3rd ed. New York:McGraw-Hill, 2001.

[2] Z. Yang, J. Zhang, M. C. W. Kintner-Meyer, X. Lu, D. Choi, J. P. Lemmon,and J. Liu, “Electrochemical energy storage for green grid,” Chem. Rev.,vol. 111, no. 5, pp. 3577–3613, Mar. 2011.

[3] M. Winter and R. J. Brodd, “What are batteries, fuel cells, and superca-pacitors?” Chem. Rev., vol. 104, no. 10, pp. 4245–4270, Oct. 2004.

[4] B. Scrosati and J. Garche, “Lithium batteries: Status, prospects andfuture,” J. Power Sources, vol. 195, no. 9, pp. 2419–2430, May 2010.

[5] M. Gad-el-Hak, The MEMS Handbook, Second Edition—3 Volume Set,2nd ed. Boca Raton, FL: CRC Press, 2005.

[6] R. Bogue, “MEMS sensors: Past, present and future,” Sens. Rev., vol. 27,no. 1, pp. 7–13, 2007.


[7] E. Verpoorte and N. F. De Rooij, “Microfluidics meets MEMS,” Proc.IEEE, vol. 91, no. 6, pp. 930–953, Jun. 2003.

[8] D. Erickson and D. Li, “Integrated microfluidic devices,” Anal. ChimicaActa, vol. 507, no. 1, pp. 11–26, Apr. 2004.

[9] V. A. Lifton, S. Simon, and R. E. Frahm, “Reserve battery architecturebased on superhydrophobic nanostructured surfaces,” Bell Labs Tech. J.,vol. 10, no. 3, pp. 81–85, Autumn(Fall) 2005.

[10] V. A. Lifton, J. A. Taylor, B. Vyas, P. Kolodner, R. Cirelli,N. Basavanhally, A. Papazian, R. Frahm, S. Simon, and T. Krupenkin,“Superhydrophobic membranes with electrically controllable permeabil-ity and their application to “smart” microbatteries,” Appl. Phys. Lett.,vol. 93, no. 4, pp. 043112-1–043112-3, Jul. 2008.

[11] V. A. Lifton, S. Simon, and F. M. Allen, “A chemistry-independent mi-crobattery with enhanced functionality,” in Proc. NSTI Nanotech. CleanTechnol.—Bio Energy, Renewables, Green Building, Smart Grid, Storage,Water, 2008, pp. 467–470.

[12] V. A. Lifton and S. Simon, “Robust Si-based membranes for fluid con-trol in microbatteries using superlyophobic nanostructures,” J. Microelec-tromech. Syst., vol. 20, no. 1, pp. 73–82, Feb. 2011.

[13] A. Ahuja, J. A. Taylor, V. Lifton, A. A. Sidorenko, T. R. Salamon,E. J. Lobaton, P. Kolodner, and T. N. Krupenkin, “Nanonails: A simplegeometrical approach to electrically tunable superlyophobic surfaces,”Langmuir, vol. 24, no. 1, pp. 9–14, Jan. 2008.

[14] A. Tuteja, W. Choi, J. M. Mabry, G. H. McKinley, and R. E. Cohen,“Robust omniphobic surfaces,” Proc. Nat. Acad. Sci., vol. 105, no. 47,pp. 18 200–18 205, Nov. 2008.

[15] A. K. Gnanappa, O. Slattery, F. Peters, C. O’Murchu, C. O’Mathuna,R. Fahey, J. A. Taylor, and T. N. Krupenkin, “Factors influencing adhesionof fluorocarbon (FC) thin film on silicon substrate,” Thin Solid Films,vol. 516, no. 16, pp. 5673–5680, Jun. 2008.

[16] F. Mugele and J.-C. Baret, “Electrowetting: From basics to applications,”J. Phys., Condens. Matter, vol. 17, no. 28, pp. R705–R774, Jul. 2005.

[17] J. Heikenfeld and M. Dhindsa, “Electrowetting on superhydrophobic sur-faces: Present status and prospects,” J. Adhesion Sci. Technol., vol. 22,no. 3, pp. 319–334, Jun. 2008.

[18] T. N. Krupenkin, J. A. Taylor, T. M. Schneider, and S. Yang, “From rollingball to complete wetting: The dynamic tuning of liquids on nanostructuredsurfaces,” Langmuir, vol. 20, no. 10, pp. 3824–3827, May 2004.

[19] T. Krupenkin, S. Yang, and P. Mach, “Tunable liquid microlens,” Appl.Phys. Lett., vol. 82, no. 3, pp. 316–318, Jan. 2003.

[20] V. A. Lifton and S. Simon, “Preparation and electrowetting transitionson superhydrophobic/hydrophilic bi-layer structures,” J. Porous Mater.,vol. 18, no. 5, pp. 535–544, Sep. 2011.

[21] K. M. Knowles and A. T. J. van Helvoort, “Anodic bonding,” Int. Mater.Rev., vol. 51, no. 5, pp. 273–311, Oct. 2006.

[22] C. H. Tsau, S. M. Spearing, and M. A. Schmidt, “Fabrication of wafer-level thermocompression bonds,” J. Microelectromech. Syst., vol. 11,no. 6, pp. 641–647, Dec. 2002.

[23] A. Decharat, J. Yu, M. Boers, G. Stemme, and F. Niklaus, “Room-temperature sealing of microcavities by cold metal welding,” J. Micro-electromech. Syst., vol. 18, no. 6, pp. 1318–1325, Dec. 2009.

[24] F. Niklaus, G. Stemme, J.-Q. Lu, and R. J. Gutmann, “Adhesive waferbonding,” J. Appl. Phys., vol. 99, no. 3, pp. 031101-1–031101-28,Feb. 2006.

Victor A. Lifton (M’96) received the M.S. degree from Moscow Institute ofSteel and Alloys, Moscow, Russia, in 1993, and the Ph.D. degree in materialsscience and engineering from Stevens Institute of Technology, Hoboken, NJ,in 1999.

From 1999 to 2004, he held various R&D positions in semiconductor pro-cessing and microelectromechanical system fabrication at Measurement Spe-cialties, Inc., Kulite Semiconductor Products, and Bell Laboratories. In 2004, hejoined mPhase Technologies, Little Falls, NJ, as a Senior Member of TechnicalStaff and in 2006 became a Chief Scientist. His job responsibilities includedbasic and applied R&D activities in the company’s efforts in nanotechnologyand commercialization of nanotechnology-based products such as reservemicrobatteries.

Steve Simon (M’97) received the B.A. degree in multimedia productions fromNew York University, New York, NY, and the M.S. degree in computer sciencefrom the City College of New York, New York, NY.

He was the Executive Vice President for research and development at mPhaseTechnologies, Inc., Little Falls, NJ. He was responsible for managing theresearch and development of mPhase’s portfolio of nano- and microelectrome-chanical system-based products. Prior to joining mPhase Technologies, he heldpositions as a Distinguished Member of Technical Staff and TechnologiesConsultant at AT&T Labs and Lucent Bell Labs. He is the holder of eightpatents in the telecommunications and nanotechnology areas.

Mr. Simon was awarded the Frost and Sullivan 2005 award for companyinnovation for bringing a novel nanobattery to the marketplace.

Johan Holmqvist received the M.S. degree in engineering biotechnology fromLinköping Institute of Technology, Linköping, Sweden, in 2008, after havingfocused on applied surface biotechnology and nano-cellulose thin film sensorapplications.

Later in 2008, he joined Silex Microsystems AB, a pure-play MEMS foundrylocated in Stockholm, Sweden, where as a Product Development ProductManager, he manages customer MEMS-projects through Silex Microsystems’product development and manufacturing technology platform.

Thorbjörn “Toby” Ebefors (S’95–M’00) received the M.Sc. degree in appliedphysics and electrical engineering from Linköping Institute of Technology,Linköping, Sweden, in 1994, and the Ph.D. degree in MEMS and microroboticsfrom the Royal Institute of Technology (KTH), Stockholm, Sweden, in 2000.

He is the Chief Technologist and Corporate VP R&D at Silex MicrosystemsAB, Stockholm, Sweden. He is one of the five founders of Silex and has beenwith the company since the start in 2000. He has managed several key customerprojects and is involved as Purchasing Director and Sourcing Manager. In 2005,he became a Director of Design with responsibility of computer-aided designand wafer layout of customers designs with tape-outs as well as building the runcard process templates for this. The key focus in his current position is R&Dactivities with focus on new MEMS materials, new 3-D MEMS and packagingtechnologies, and strategic research partnerships. He is also responsible formanaging the Silex IP portfolio and patent-related topics. He is an inventorto more than 25 granted patents and pending patent applications in the MEMSareas.

David Jansson, photograph and biography not available at the time ofpublication.

Niklas Svedin received the M.Sc. degree in electrical engineering from theRoyal Institute of Technology, Stockholm, Sweden, in 1995, and the Ph.D.degree in microelectromechanical systems from the Royal Institute of Tech-nology, Stockholm, in 2003.

He is Chief Engineer and Corporate VP at Silex Microsystems AB, Järfälla,Sweden. He is one of the founders of Silex and has been with the company sincethe start in 2000. He has managed several key customer projects, and in 2006,he became the Director of Engineering with responsibility for technical reviewand support for new foundry products. The key focus in his current position isto merge customer needs with the foundry capabilities and to promote the useof Silex standard process modules and process blocks.

Documents

Design and Fabrication of Addressable Microfluidic Energy Storage MEMS Device