Journal of Power Sources 171 (2007) 575584

Fabrication of polymer electrolytpries Hd, WAly 20007

Abstract

Utilizing air pstrated some that tmembrane e usedprint cartridg wereabrasion tes on ofor similar p ale fapparatus, o cm2A commercially available membrane under identical, albeit not optimized test conditions, showed about 7% greater power density. The objectiveof this work was to demonstrate some of the processing advantages of drop-on-demand technology for fabrication of MEAs. It remains to bedetermined if inkjet fabrication offers performance advantages or leads to more efficient utilization of expensive catalyst materials. 2007 Published by Elsevier B.V.

Keywords: Inkjet; MEA; PEM; Fuel cell; Printing

1. Introdu

Polymeemerged adue to theconstructiocapabilitiestive use reqfew thousation applicaindustry caa cost less taside, redumaterials crials and thmembranereducing th

CorresponE-mail ad

0378-7753/$doi:10.1016/jction

r electrolyte membrane fuel cells (PEMFCs) haves promising future power sources for automobilesir high efficiency, high energy densities, modularn, low operating temperatures and quick start-up[1,2]. However, large-scale application for automo-

uires transitioning an industry capable of producing and PEMFC stacks for both stationary and transporta-tions, at a real cost of greater than $1000 kW1, to anpable of producing hundreds of thousand of stacks athan $50 kW1. Issues of durability and performancecing the manufacturing costs must involve reducedosts, especially as regards the PFSA membrane mate-e platinum catalyst content, and simplification of theelectrode assembly (MEA) fabrication process bye number of processing steps.

ding author. Tel.: +1 509 372 6563; fax: +1 509 375 2167.dress: silas.towne@pnl.gov (S. Towne).

Inkjet printers are well-known devices that utilize drop-on-demand technology to deposit various materials or inkswithout contact between the printer head and the substrate. Thistechnology has become a popular manufacturing technique usedaround the world in desktop printers, and has been applied todeposition of various materials and the construction of uniquedevices such as ceramics, organic light emitting diodes, assem-bly of periodic structures, structural polymers and more [3,4].

Drop-on-demand printers use piezoelectric transducers typ-ified by the Epson (Nagano, Japan) line of home/office marketprinters or thermal resistors typified by the Canon (Tokyo,Japan) and Hewlett-Packard (Palo Alto, CA, USA) printers toexpel ink droplets only when needed. Both technologies printdrops at a rate of one drop per 80200s. This cycle time perdrop includes initiating a current pulse to the respective plate toeject a droplet and refilling the ink cavity; the latter step con-sumes approximately 90% of the cycle time [5]. Nozzle sizesfor these printers are usually 2030m with ink droplets of1020 pL [3]. Technology is constantly improving to achievebetter resolution with smaller nozzle sizes and ink droplets,with some printers delivering droplets as small as 1.5 pL. These

see front matter 2007 Published by Elsevier B.V..jpowsour.2007.07.017cell MEAs utilizing inkjetSilas Towne , Vish Viswanathan, Jam

Pacic Northwest National Laboratories, RichlanReceived 20 April 2007; received in revised form 9 Ju

Available online 17 July 2

drop-on-demand technology, we have successfully fabricated hydrogenof the processing advantages of this technology and have demonstratedlectrode assemblies (MEAs). Commercial desktop inkjet printers werees onto Nafion polymer membranes in the hydrogen form. The layers

ts and did so without any post-deposition hot press step. The eliminatirocessing technologies may provide a route to less expensive large-scpen circuit voltages up to 0.87 V and power densities of up to 155 mWe membrane fuelnt technologyolbery, Peter Rieke99352, United States

07; accepted 10 July 2007

olymer electrolyte membrane fuel cells (PEMFC), demon-he performance is comparable to conventionally fabricatedto deposit the active catalyst electrode layer directly fromwell-adhered and withstood simple tape peel, bending andthis processing step suggests that inkjet-based fabricationabrication of PEMFCs. When tested in our experimentalwere obtained with a catalyst loading of 0.20 mg Pt cm2.

576 S. Towne et al. / Journal of Power Sources 171 (2007) 575584

printers offer high resolution (dots per inch), as well as the abil-ity to print intricate features, gradients, and 3D structures ina repeatablhave been

In addittinuous inkthe flight ohis reviewhave replacin the homgiven prodquantities,advantages

Some othe reductioricate the Mthan batchadaptabilityalyst layerpure hydroability to raduced, (5)manufacturmatching (rate-limitin

The relation must bfabricationity to transnext job. Iset-up andscale, highthe depositexample, itwhile miniarea as man

loadings frloss in perfing manufareduce Pt loity [10]. Ththe thickneto vary theports.

Anotherufacturingunique abiing designsachieve theable levelsfabricationrequired foary power aLarge-scaleof holdinging applicaas across a

up inkjet printers will still have the ability to instantaneouslychange printing patterns, sized resolutions and thicknesses with-

jor reir adprodev

e-scaametc to, theon, as. Oloidasitionk cm prid ininkergycouse lowition, it wowerthe

timause

humne glicall

articuk prg can

e/ond 4ave

en pd anmpartzles

her nlity tnt, a

primhat etionA itrica

the trrks.regitech

ranelogyy one manner. Recently memjet-based printer systemstouted for their very high print speeds [6].ion to drop-on-demand systems, there are also con-jet systems that use an electrostatic field to directf ink droplets. It has been proposed by Le [5], inof inkjet technology, that drop-on-demand systemsed many of the continuous jet printers, especiallye/office market. In industrial applications where auct, such as an MEA, may be produced in multipleit is not clear that drop-on-demand necessarily holdsover more continuous printing or coating techniques.f the essential criteria for MEA fabrication are (1)n in the number of processing steps required to fab-EA, (2) wherever possible have continuous rather

processing from material input to stack assembly, (3)to fabricate different MEAs (e.g. compare the cat-

composition needed for a direct methanol fuel cell,gen-based and reformate-based fuel cells), (4) thepidly change size, shape and quantity of MEAs pro-

very high product quality control with the associateding metrics and (6) be capable of production speedsbut not necessarily greatly exceeding) that of otherg fabrication and assembly processes.tively high complexity of drop-on-demand fabrica-e contrasted to the simplicity of roll coating. Inkjetis much more versatile than roll coating in the abil-ition from one small- or medium-scale job to the

n comparison, roll-coating facilities are difficult tooptimize for a given job and are better suited to large--speed production. Further, inkjet fabrication allowsion of complex structures or if you will, images. Foris critical to optimize the performance of the MEAs

mizing the Pt loading. Progress has been made in thisy researchers have demonstrated the reduction of Pt

om greater than 4 mg cm2 to 0.4 mg cm2 withoutormance [79]. However, with increased control dur-cturing, precision-deposited Pt catalyst could furtherading while maintaining power density and durabil-

is can be accomplished by grading the loading withinss of the catalyst layer but it may also prove usefulloading density from fuel or oxidant inlet to outlet

important attribute of drop-on-demand-based man-is the exceptional reproducibility combined with thelity to adapt the manufacturing process to evolv-or applications. This characteristic may be used to

very low cell failure rates needed to maintain accept-of failure in assembled stacks. Further, an individualline can be readily adapted from production of MEAsr automotive applications to that required for station-pplications and even to direct methanol applications.inkjet printers can utilize multiple cartridges capable

a variety of solutions (various colors) for chang-tions or for variation of MEA composition as wellspecific printed electrode area. In addition, scaled

out malose thnologydesignto larg

Parspecifibilitiesprecisiprocesthe colcompospan. Iuniforby drieare theThe enthe vismust bIn addnozzleping, laroundand ulwill caThus,ethyle

Typof a pmeet inprintinthe hom1 cP aand an

Whavoidethan, 1eratedthe noz[3]. Otthe abiresista[5].

Thetions tfabricaful MEof elecallownetwoof thiscationmembtechnodirectletooling or setup. Although many technologies mayaptability and versatility when scaled up, inkjet tech-

vides versatility useful in small-scale single run MEAelopment to medium-scale job-shop production runsle dedicated fabrication lines.ers that impact the utility of inkjet print technologya component include the inkjet array delivery capa-environment in which the unit operates, the systemnd the functional control of the entire manufacturing

ne critical step in this process is the formulation ofl suspension delivered to the substrate, including then and stability over the manufacturing process timeomposition must meet rigid requirements to maintainnt patterns and prevent clogging of the printer nozzlesk or large particles. The most important parametersviscosity and surface tension, as well as particle size.of each printed drop must be sufficient to overcomeflow and surface tension of the ink, yet the viscosity

enough to allow the ink reservoir to refill quickly., if the ink surface tension is not high enough in theill spontaneously drip [3]. In combination with drip-surface tension can lead to wetting of the faceplate

nozzle preventing formation of a stable droplet streamtely causing drying around the nozzle [3]. Dryingnozzles to become blocked and no longer functional.ectants, low volatility water-miscible liquids such asycol, are added to prevent drying [3].y, inks are designed to match the desired performancelar printer (or vice versa), so novel solutions must

operties specific to the given printer before successfulbe accomplished. For the inks tested from printers in

ffice market, these properties are a viscosity betweencP, surface tension in the range of 3035 mN m1,rage particle size of approximately 0.2m.rinting colloidal solutions, sedimentation must be

d particle sizes should ideally be at, or preferably lessfor typical desktop printers. Large particles, agglom-icles, settling of unstable colloids, or even dust clog, increase the viscosity and lead to short ink shelf-lifeecessary characteristics of inkjet compatible inks areo correctly wet the substrate, be quick drying, smearnd permit easy clean-up with limited maintenance

ary objective of this study was to create print solu-xhibit the above characteristics, while optimized forof highly functional MEAs. The key to a success-

s creating an ideal three-phase boundary, a networkl and ionic conductivity with adequate porosity toansport of gas, protons and electrons throughout thePrevious research has focused on the optimizationon through ideal ionomer to Pt/carbon ratios, fabri-niques, and selective conditioning processes of theand finished MEA [8,1013]. In our study, inkjetwas utilized to print layers of catalyst solutionsthe membrane. Despite the use of the membrane

S. Towne et al. / Journal of Power Sources 171 (2007) 575584 577

in proton form, a highly adherent, mechanically stable cata-lyst layer could be formed without any hot pressing process[14,15]. Thcomparedcommerciastrate supebut to showformance tdistinct pro

2. Experim

2.1. Cataly

Catalystcarbon-supNJ) with Ntle, DE) anglycol andsurface tensonicated (netic stir bato select inkprinter usedsize (Horiblyzer LA-9tension (Bby an actuaby the duraprinted laying of the pquality, stamined thatto commercmaintained

2.2. Prepa

Nafionfor 1 h in 3water, boilionized waMilli-Q graprinting, thlow heat (ureturned to

2.3. Printi

Off-the-were usedwere opencatalyst inneighborintion. The mand fed thrIllustrator

Electrode dimensions were changed by altering the shape andsize of the bitmap image on the computer. Electrode thickness

solud lup dle frto 2to

tedlessn prlignctrodan 30acetd.prin

urester ecarr

ing

reparaph

mea

on c

ted.easu

logint woomm dmicSEMtheVSMparaple

ingaturOL

ratioic (E

uel c

As pavaiC-1stem

ancroximminH2 (erate

ance electrochemical properties of the electrodes werefor various processing conditions and compared tolly available MEAs. Our purpose was not to demon-rior or state-of-the-art electrochemical performance

that inkjet-printed MEAs showed comparable per-o commercially available MEAs while offering somecessing advantages.

ental

st formulation

solutions were prepared by thoroughly mixing aported catalyst (Vulcan XC-72R, E-TEK, Somerset,afion solution (EW1100, Ion Power Inc., New Cas-d Milli-Q grade de-ionized water. Water, ethyleneisopropanol were added to adjust the viscosity and

sion. To form a homogeneous ink, the solutions wereBranson Sonifier) for 5 min and stirred with a mag-r for 24 h. A design of experiments protocol was usedcompositions that were compatible with the desktopin this study. For each ink composition, the particle

a Laser Scattering Particle Size Distribution Ana-20), viscosity (Kruss K12 tensiometer) and surfaceohlin VOR Rheometer) were determined, followedl printing test. While the print quality, as determinedbility, continuity (lack of visual banding within the

er) and smearing of the image or by clogging or cak-rint head, was subjective, this method resulted in highble inks that printed well. As expected, it was deter-the ink with the best printability had similar attributesial inks designed for a specific printer. Good inks alsocolloidal stability for at least a few days.

ration of Naon membranes

membranes (N117, Ion Power, Inc.) were washedwt% H2O2, rinsed and boiled for 1 h in de-ionized

ed in 0.5 M H2SO4 for 1 h, rinsed, and boiled in de-ter for 60 min [17]. The membrane was then stored inde de-ionized water until ready for use. Just prior toe membrane was gently dried on a vacuum table undernder 70 C) until wrinkle free and the membrane hadits original dimensions.

ng of catalyst inks on Naon membranes

self piezo (Epson) and thermal (HP) inkjet printerswith only minor modification. The ink cartridges

ed, cleaned of the original ink and filled with thek using a syringe. All ink was removed from theg cartridges to remove the possibility of contamina-

embrane was secured to a cellulose acetate sheetough the printer using the original paper feed platen.software was used to control the print parameters.

and retion anthe droavailab

Upprintedconductions,betweeto re-athe eleless thto therepeate

Theprocedand wacol inkimprov

2.4. Pmicrog

Toprinteddeposiwas m

TechnogradiestereozPAXcaoptical

Fordle ofJEOLple prethe samsectiontemperin a JEacceletroscop

2.5. F

MEciallyLightFtest syperformof app0.05 L100%was opperformtion were controlled by modifying the hue, satura-minescence. These parameters are directly related toelivery rate and spacing but this relationship is notom the manufacturer.4 successive layers of varying compositions werebuild the desired electrode structures. Printing wasat ambient room temperature and under these condi-than 1 min was required to achieve adequate dryingints and was determined primarily by the time takenand reload the sample. After all layers were printede was dried by gently warming with a heat gun fors. The substrate was then turned over and re-secured

ate medium, and the printing and drying processes

ted electrodes were then processed by a variety ofthat included hot pressing at various temperatures

xtraction. Water extraction removes the ethylene gly-ier while hot pressing was used with the intention ofthe integrity and adhesion of the films.

ration of MEA sample for conductivity, opticals, and SEM and TEM measurements

sure the electrical conductivity, a single layer wasellulose acetate using a grey scale to vary the amount

The resistivity as a function of thickness of the layerred using a Seebeck Measurement System (MMRes, Inc., Mountain View, CA). This same densityas also viewed top down through an Olympus SZH

microscope (Center Valley, PA) with an attachedigital microscope camera (Villa Park, IL) to obtainrographs.

, 20 mm 5 mm samples were cut from the mid-MEA and the cross-section was examined using a-5900LV scanning electron microscope. TEM sam-

tion consisted of cutting a small sample, embeddingin epoxy resin and curing overnight at 60 C, then

with a diamond knife via an ultra-microtome at roome. The 100 nm sections were subsequently examined2010 high-resolution TEM equipped with 200 kV

n voltage equipped with an energy dispersive spec-DS) detector.

ell performance measurements

roduced for this study were compared to commer-lable MEAs. Initial tests were conducted using a(Fuel Cell Store, Boulder, CO) single cell fuel cell, designed for demonstration and low power outpute and with 2.75% H2 (97.25% Argon) at flow ratesately 0.2 L min1 and air at rates of approximately1. Later studies used an in-house fabricated cell and

35 mL min1 and 50 mL min1) and air. The systemd between ambient temperature and 75 C, with peake occurring at 75 C.

578 S. Towne et al. / Journal of Power Sources 171 (2007) 575584

In all cases, both gases were 100% humidified using anImmersion Circulator (VWR model 1112A) to control the tem-perature of the water bath and Perma Pure MD Gas Humidifiersto humidify the gases. Three way valves were utilized forboth the hydrogen and air gas lines allowing the selectionof either humidified or non-humidified gas streams. Carboncloth backings (GDL ELAT 1200-W, Fuel Cell Store) wereused as gas diffusion layers. Commercial MEAs tested werethree-layer MEAs (SL-117, Fuel Cell Store) with loadings of0.3 mg cm2 Pt for both the anode and cathode on Nafion 117membranes.

To test the performance a load bank was used consisting ofresistors in the 0.120 range. By connecting various resis-tances across the fuel cell, the current was varied, and a VIcurve generated. The current was measured by a digital mul-timeter connected in series with the circuit, while the voltagewas measured directly across the fuel cell terminals by anotherdigital mul

3. Results

3.1. Ink co

While nthan 98%shows theprinting. Tand a largetained collcartridge fiwas 35.5 mfound for ccosity wasOEM inks

Suitabledesign andbest electrotems is thisadvanced pMEA fabriacteristics.

Fig. 1. Particl

EMhe arr. To tey).

emb

. 2 reyer,sub400

eposess ot andWhen insufficient material was deposited, the print dropst fully coalesce and showed a pattern consistent with theof single drops. These results demonstrate the potential

p-on-demand technology to print evenly distributed cat-ayers as desired and shows the importance of substrateility on the distribution of material.conductivity was characterized at various locations to

ine if the carbon network was deposited in an effectiver. In addition to the above printed layer, conductivity mea-ents were also compared for various thinner printed layers.tal ink density was controlled by changing the lumines-n the illustrator program in approximately 4% increments

ch layer. Increasing the luminescence creates a lighterthat reduces the amount of ink deposited, resulting iner layer. In all, 14 different thicknesses were printed andosits were clearly visible and the optical density as deter-qualitatively by eye correlated with the luminescence

.seen in Fig. 3, the resistivity was about 1 cm for

esses above 580 nm, indicating reasonable connectivityhout the layer. Below this thickness the resistivity rapidlysed into the mega ohm range and was too high to be accu-measured. These results suggest that segregation of thetimeter.

and discussion

mpatibility

o best ink was identified, suitable inks have betterof the particles with a size less than 2m. Fig. 1particle size distribution of a typical ink suitable forhough this ink showed greater particle size disparityr average size than commercial inks, the ink main-oidal stability and consistently passed through thelter and print head. The surface tension of this inkN m1. This value is at the high end of the rangeommercial inks (3035 mN m1). Finally, the vis-found to be 3.35 cP, again within the range of the

(1.54 cP).ink characteristics are governed by the print headoperation and are not necessarily related to the

de composition. A drawback to off-the-shelf sys-limited range of suitable ink characteristics. More

rinting systems that might be used in large-scalecation have a much wider range of suitable ink char-

e size distribution of catalyst solution after sonication and stirring.

Fig. 2. Sstrate. Tlocation(dark gr

3.2. M

Figalyst laacetatethin asrials dthicknpercenfilms.did nodryingof droalyst lwettab

Thedetermmanne

surem

The tocence o

for eaimagea thinnall depminedsetting

Asthicknthrougincrearatelycross-section of a single catalyst layer on a cellulose acetate sub-ows measure the thickness of the catalyst layer at their prospectivehe right of the catalyst layer is the substrate. To the left is epoxy

rane/electrode assembly

plicates a cross-section SEM image of a single cat-approximately 1020 nm thick, printed on a cellulosestrate. Depending upon the printer system, layers asnm could be printed. Providing the mass of mate-

ited was sufficient, the layers were of very uniformver large areas with a variation in thickness of a fewa roughness much smaller than the thickness of the

S. Towne et al. / Journal of Power Sources 171 (2007) 575584 579

Fig. 3. Variation of resistivity of carbon supported catalyst layer with filmthickness on cellulose acetate support.

carbon occurred, or because of the printer characteristics, thematerial was not uniformly distributed. With this particular print-ing system, a deposit of at least 1m was required to maintaingood electr

By usineters suchnumber ofcannot be ming to thestudy, evideBanding ochead do no(dots per inlessened. Adisappeare3 of the 14representsthickness athere is stilconnectivitthe seventhconnectivitvertical andwith the higsamples, sh

EM cer isft.

plettivit

expeellulthe per isster ambr[16

t laywas

thinner electrode thickness than the 1020m thicknessof commercially available electrodes. In additional work

hicker electrodes were deposited. It is apparent from thesethat inkjet printing allows excellent control over the indi-layer thickness but still allows many layers and ultimatelythicker electrodes to be deposited.

Fig. 4. Opticaand (c) low. (aical conductivity.g a commercially available printer, printing param-as droplet size, jetting velocity, print head speed,nozzles, distance between nozzles, and platen speedanually controlled. Therefore, without exact match-

OEM inks, some discontinuities will occur. In thisnce of banding appeared when printing a single layer.curred when drops printed in one pass of the printt coalesce with the next print pass. With higher dpich) and low luminescence the appearance of bandinglso, with the printing of additional layers these lines

d as ink fills in the areas with less ink. Fig. 4 showssingle layers printed to test conductivity. Fig. 4a

the lowest luminescence and therefore, the highestnd best connectivity of the 14 samples. In the figurel some evidence of banding, but there was adequatey to give a negligent resistivity. Fig. 4b representsthickest print, a print that had low conductivity. Lessy is seen in this figure as the drop overlap in both the

horizontal direction lessens. In Fig. 4c, the samplehest luminescence, and therefore the thinnest of theowed the greatest degree of discontinuity. Individ-

Fig. 5. Salyst layon the le

ual droconduc

Asthan cmatchAnoththe wathe mesurfacecatalys3.2ma muchtypicalmuch tresultsvidualmuchl micrographs taken at 15 magnification showing evidence of banding in three sam) has high conductivity whereas both (b) and (c) have no conductivity due to low coross-section of eight printed catalyst layers on Nafion. The cat-labeled in the middle with the membrane on the right and epoxy

s start to become visible and there is no recordabley.cted, printing on Nafion proves to be more difficultose acetate as commercial printers are optimized torint speed to specific ink and substrate characteristics.ue with Nafion is swelling of the membrane fromnd alcohol in the catalyst ink. Water alone can swellane by as much as 32%, causing an uneven printing]. Fig. 5 shows a cross-section SEM image of eighters printed directly on Nafion. A thickness of up tomeasured. This first generation of printed MEAs hadples of different thickness (drop amounts): (a) high, (b) medium,nnectivity.

580 S. Towne et al. / Journal of Power Sources 171 (2007) 575584

Fig. 6. SEM 0 C.membrane on m.

Also visline locatedink. Withthe catalyscally embeadhesion orepresent aan oven foual alcoholthe good cthis imagethe catalysareas reprein Fig. 6b,on the leftMEA crosstinuous andelaminatio

The natuinterface bebe identifiepresumed tregions ancarbon partthis occursadhesion w

To testchemical tthrough boof materialpeel test. Blight rubbidamaged bsharp objecthan sufficiage to thechemical mversion to

ssinhe eat thls aftalysnifo

s. 5 ain-hoto bog onontrle imn imrollto mg thmim

n catthernetFigcross-section of eight printed layers on Nafion cured in oven for 10 min at 12the left and epoxy on the right. The scale bars represent (a) 10m and (b) 1

ible in the cross-section of the MEA is a thin darkat the interface of the membrane and the catalyst

no processing on the MEA, such as hot pressing,t layer is adhered to the membrane but not mechani-dded. This gave rise to concerns about the mechanicalf the catalyst layer to the membrane. Fig. 6a and bcross-section view of Pt/C on Nafion and cured inr 10 min at 120 C to promote evaporate the resid-s. Fig. 6a shows a magnification at 1000, showingonsistency of the catalyst layer over large areas. Inthe Nafion membrane is on the left separated fromt layer by a thin dark line. The darker grey and blacksent epoxy. A magnified view of this MEA is depictedwith a thin dark line again separating the Nafionfrom the catalyst layer. These SEM images of the-section indicate the catalyst layer is uniformly con-

d is adhered to the membrane with no evidence ofn.re of the materials comprising the thin dark line at thetween the catalyst layer and the membrane could notd. As this line appeared darker in the SEM, it could be

hot pretion. There ththe celthe cagas ma

Figslightis dueprintinity to cPossibwith aprinterbranesswellinshouldan eve

Furcarbonview ino be more electrically conductive than the adjoiningd this could be explained by phase segregation oficles to the interface. The answers to questions of howand whether this effect is responsible for enhancedill have to await further study.adhesion, MEAs were subjected to mechanical andreatments. Catalyst layers remained adhered eveniling in water or sulfuric acid. No significant amountwas removed by the simple yet standard scotch tapeending of the sample did not result in flaking nor didng with a tissue. The catalyst layer could be visiblyy a determined scratch such as with a finger nail ort. From this we determine that the adhesion is moreent for assembly of a cell without mechanical dam-catalyst layer. We note that this is achieved with noodification of the Nafion membrane such as con-

the sodium form prior to printing and without any

imately 10(representetures has ncondition.continuityclear evideof approximtwo differiof residualof the inkjadditional

Withoutelectronic adetrimentatechniqueswere tried.to compacThe catalyst layer is represented with an arrow with the Nafion

g after deposition as is typically used in MEA fabrica-ffect of hot pressing is discussed below. We also notee MEA remained mostly intact upon disassembly ofter electrochemical testing. However, some parts oft layer adhered to the gas diffusion layers or to theld.nd 6a show that printing on Nafion results in somemogeneities within the catalyst layer thickness. Thisth the uneven surface and the effect of swelling whena Nafion membrane, as well as the lack of abil-

ol the operating conditions of a commercial printer.provements with a commercial printer can be made

proved technique for securing the membrane to theer base during printing or by using reinforced mem-

inimize the effects of swelling. With reduction ofe catalyst layer printed on the Nafion membraneic that of the cellulose acetate substrate, resulting in

alyst layer surface.analysis of the MEA with TEM shows an incompletework as seen in Fig. 7a and b. The cross-sectional. 7b shows carbon particles with diameters of approx-

0 nm and catalyst particles between 3 nm and 10 nmd by black dots). The MEA represented in these pic-ot been hot pressed and represents the as-printedThe carbon particles in Fig. 7a do not show theexpected from the SEM images. Instead, there isnce of a layered morphology of alternating layers

ately 0.5m thickness. It is hypothesized that theng layers are a carbonionomer layer and a layerethylene glycol, both deposited with a single print

et printer. Each successive print therefore adds twodiffering layers with a similar morphology.

a continuous carbonionomer network, both thend ionic conductivities will decrease and presumably

lly affect the electrochemical performance. Differentfor removing this impeding layer of ethylene glycolThese included hot pressing at various temperaturest and strengthen the catalyst layer and using water

S. Towne et al. / Journal of Power Sources 171 (2007) 575584 581

Fig. 7. TEM image of printed catalyst layers on Nafion before processing. The arrow represents carbon particles. The small black dots on the carbon are platinumparticles. The scale bars represent (a) 0.5m and (b) 100m.

extraction to remove the ethylene glycol as seen with glycerolremoval in previous studies [17]. As earlier discussed, this studycompares hcombinatio

Fig. 8afor 5 min abon has beThese sampity within tThe blackgrey carbonof the layeto be at leawith the imthat perform

In summuniform anphase bounan automatchased prinmanufactur

cost effective process as compared to other MEA manufacturingtechniques.

erfor

l cerintef eacintert are

forloadmica

g Ps on

hilerisonof p

tingo pral re

Fig. 8. TEMot pressing with water extraction as well as with then of water extraction and hot pressing.and b shows images of an MEA pressed at 2045 psit 125 C. Clearly the interconnectivity of the car-

en increased over that shown in the previous figure.les showed increased electrical and ionic conductiv-

he section, as well as evidence of the porous network.dots within the image represent the Pt on the darker

particles. After pressing there is no longer evidencers seen in Fig. 7a. The ethylene glycol was assumedst mostly removed by heating. It was expected thatproved electrode structure seen after hot pressing,ance of the MEA would also increase.

ary, the MEAs depicted in these images are relativelyd repeatable, showing porous and distributed three-daries manufactured by printing catalyst layers viaed process. Using only a modified commercially pur-ter, these results show the merits of drop-on-demanding as a precise, repeatable, and potentially more

3.3. P

Fuewith parea o

cial pra prinprintedthe Pttroche0.094 mreason

ies. Wcompaeffects

Teswith nidenticimage of printed catalyst layers on Nafion after pressing MEA at 2045 psi at 125 Cmance of single cells

ll performance was measured on fabricated MEAsd layers for both the anode and cathode. The activeh printed MEA was 2.25 cm2. With the commer-an optimized dpi of 2880 1440 was printed. Overa of 1.5 cm 1.5 cm about 1.4 million drops wereeach layer with a drop size of 3 pL [18]. From this

ing can be determined and the MEAs used for elec-l test described below have a platinum loading oft cm2 unless otherwise noted. For laboratory safetyly 2.75% H2 could be used for these initial stud-this does not give optimal results that allow detailed

of different electrodes it did allow evaluation of therocessing on electrode performance.began by measuring the IV response of the MEAsocessing. Two separate printed MEAs gave nearlysults when tested with no post-processing as exem-for 5 min. The scale bars represent (a) 0.2m and (b) 100 nm.

582 S. Towne et al. / Journal of Power Sources 171 (2007) 575584

Fig. 9. Polarization curves comparing printed MEAs hot pressed at variouspressures for 5 min at 125 C to an unprocessed MEA. Cells were tested with2.75% H2/air at 75 C, ambient pressure and 100% RH.

plified by the baseline curves in Figs. 9 and 10. In Fig. 9, thecurrent density as a function of cell voltage of a MEA withno post-proent pressur2.75% H2/made a larprinted MEseen with epressure ofsures and tthe catalyspower denat 100 C,gain was atures are neelectrode pthe membrat 0.401 V,due to increand comprcontinuous

Fig. 10. Polarpressed at varwere tested w

Fig. 11. Polasoaked and hoMEA. Cells wRH.

inates ththylig. 1omp

. Alsand trnedis suand pME

of 4ver

watoutp

essedimumthatttle iter extraction. This is important as hot pressing not only

n additional manufacturing step, but also can permanentlycessing is compared against hot pressing at differ-es at 125 C for 5 min; all tested at 75 C and withAr and air. Hot pressing at a temperature of 125 Cge difference in improving the performance of theAs. Higher current densities at all voltages were

ach increase in pressure reaching a maximum with a2045 psi. This indicates that specific hot press pres-

emperatures have a significant and positive effect ont layer structure and improve MEA performance andsity. This experiment was also done with pressingbut showed very different results. No performancechieved with pressing, indicating elevated tempera-eded to compact the catalyst layers and improve theerformance. After a pressure of 2045 psi and 125 C,ane delivered a maximum output of 78.5 mA cm2representing a power of 31.5 mW cm2. This wasased contact between the catalyst and the membrane

ession of the multiple printed layers to form a morecarbon network as seen in Fig. 8a and b. Apparently

a combsqueezsome e

In Fing is cglycolwaterbe leaglycolaturessoakedpowerment oand nopowerand pra max

showsvery lijust waadds aization curves comparing printed MEAs water soaked and/or hotious pressures for 5 min at 125 C to an unprocessed MEA. Cellsith 2.75% H2/air at 75 C, ambient pressure and 100% RH.

damage thebility and l

Electrodadhesion eous layeredresult. Rathin extensivwould be atime we caresult.

After dehot pressinH2 and airAs can be sthe performperformancrization curves comparing printed MEAs water soaked and watert pressed at various pressures for 5 min at 125 C to an unprocessedere tested with 100% H2/air at 75 C, ambient pressure and 100%

ion of sufficient temperature and increasing pressuree ethylene glycol from the catalyst layer. Presumablyene glycol remained in the catalyst layer.0, the IV response of another MEA with no process-ared to a MEA soaked in water to leach out ethylene

o compared in Fig. 10 is a MEA that is first soaked inhen pressed at 950 psi for 3 min at 125 C. What canfrom Fig. 10 is that water leaching of the ethyleneperior to just simple hot pressing at elevated temper-ressures. For a direct comparison to hot pressing, theA delivered 106 mA cm2 at 0.401 V, representing a2.4 mW cm2. This represents nearly 35% improve-the power output of the MEA with only hot pressinger extraction. The soaked MEA recorded a maximumut of 44.5 mW cm2 at 0.340 V. The MEA soakedrecorded a power of 44.6 mW cm2 at 0.340 V andpower output of 45.0 mW cm2 at 0.324 V. This

hot pressing in addition to water extraction leads tof any additional electrochemical improvement overproton exchange membrane leading to reduced dura-ifetime of the MEA.es soaked in hot water continued to show excellent

ven without a hot press step. Given the discontinu-structure shown in Fig. 7, this was an unexpected

er it was expected that water extraction would resulte delamination of the interface and that hot pressingnecessary first step. This was not the case and at thisnnot offer a good explanation for this advantageous

termining that water extraction was superior to justg, the samples from Fig. 10 where tested using 100%to make a better comparison to commercial MEAs.een in Fig. 11, use of 100% H2 dramatically improvesance of all MEAs. These curves translate into peake of 77.4 mW cm2 at 0.350 V for the soaked MEA

S. Towne et al. / Journal of Power Sources 171 (2007) 575584 583

Fig. 12. Polarization curves comparing a commercial MEA with printed MEAswith different processing techniques and different loadings. Cells were testedwith 100% H2/air at 75 C, ambient pressure and 100% RH.

and 78.9 mW cm2 at 0.332 V for the soaked and pressed MEA.This represents nearly a 75% increase in peak power for bothcurves usin

After cothe fabricacially soldcurrent denfour selectprocessingprinted mewith a loadwith just oving of 0.20glycol, andloading ofties versusit can be soutperformprinted MEpower densrecorded a

Fig. 13. Powedifferent proc100% H2/air

senting a power density of 167 mW cm2. At that same voltage,the printed MEA recorded a performance of 333 mA cm2 anda power de 2a maximumfore, with acommerciapeak powe

These rinkjet-printMEAs. Thcurves, repohmic andThis is impGDLs, gasthickness.mal performWe made oadequatelynot necessamance requ

wor

fers srable

nclu

ew m

ped us toink-lougheata

ve e

urre

g Ptranenly

chniqve an

uesg 100% H2 over 2.75% H2.mparing the effects of processing for the MEAs,

ted membranes were compared against a commer-MEA tested using 100% H2 and air. Fig. 12 reportssity as a function of input voltage of the following

ed MEAs: (1) the printed membrane with no post-and 0.094 mg Pt cm2 for anode and cathode, (2) thembrane soaked in water to extract ethylene glycoling of 0.094 mg Pt cm2, (3) a printed membraneer twice the amount of printed layers to give a load-mg Pt cm2, also soaked in water to extract ethylene(4) the commercially fabricated membrane with a

0.3 mg Pt cm2. Fig. 13 compares the power densi-cell voltage for the same four MEAs. In these figureseen that though the commercial MEA significantlyed the MEA with a much lower catalyst loading, theA with 0.2 mg Pt cm2 has a comparable maximumity to the commercial MEA. The commercial MEApeak performance of 365 mA cm2 at 0.462 V, repre-

of thistion ofcompa

4. Co

A ndeveloprinterin anfed thrand repan actiup to c0.20 mmembwhile otion teeffectitechniqr curves comparing a commercial MEA with printed MEAs withessing techniques and different loadings. Cells were tested withat 75 C, ambient pressure and 100% RH.

electrodes.Our obj

the basic pare remainreliability oexample, rscale producompositioa more agadapted toinkjet fabriinum catalyfrom gas inreduction insity of the croll-coatingnsity of 154 mW cm . The printed MEA reachedpower density of 155 mW cm2 at 0.459 V. There-33% reduction in catalyst loading compared to the

l MEA, the printed MEA recorded only a 7% lowerr density.esults show that despite a lower catalyst loading,ed MEAs can compete with commercially availablee MEAs tested also had very similar polarizationresenting similar resistive losses, including kinetic,mass transport losses within the fuel cell system.ortant as both systems were tested with the sameconcentrations, humidity, and Nafion membrane

We do not represent these results as indicating opti-ance for either the printed or commercial electrodes.

nly the requisite effort to assure the membranes werehydrated. Operation at 75 C was convenient butrily ideal. Obtaining optimal electrochemical perfor-ires considerable effort and that was not the objectivek. Rather, we have demonstrated that inkjet fabrica-ignificant processing advantages and appears to offerelectrochemical performance.

sion

ethod for fabricating MEAs for PEMFCs has beensing relatively inexpensive home or office type inkjet

deposit successive layers of Pt/C catalyst dispersedike solution. Nafion is attached to a support and

the printer where the catalyst is printed in an evenble manner to produce uniform catalyst layers. With

lectrode area of 2.25 cm2, the cell has been operatednt densities of 155 mA cm2 with a loading of justcm2. These results are within 10% of a commercialwhile operating with 33% less catalyst loading andin the initial stages of production. This novel fabrica-ue, as an automated process, is both simple and timed has the potential for scaling to larger productionwhile still sustaining the ability to structure MEA

ective here was to demonstrate in a simple mannerrocess of inkjet fabrication of MEAs. Clearly thereing questions concerning the economics, speed andf this method compared to established methods. For

oll-coating techniques would be simpler for large-ction of MEA with very simple morphologies andns. However, inkjet fabrication can be the basis ofile and flexible manufacturing method that can bevarious sizes of MEA and/or compositions. Further,cation can allow the precise location of costly plat-st both through the thickness of the catalyst layer andlet to outlet. This represents a potential for significant

overall catalyst loading by optimizing the local den-atalyst. That is a process not easily achieved throughtechniques. Finally, we re-emphasize that the meth-

584 S. Towne et al. / Journal of Power Sources 171 (2007) 575584

ods presented here offer substantial simplification of the overallMEA fabrication process. We need make no use of chemicalconversion to the membrane and do not require any hot pressingoperations. However, it remains to be demonstrated that inkjet-based fabrication can equal or better the performance of MEAfabricated by now current well established, well understood andoptimized processes.

In initial assessments, fabrication of PEM fuel cells withinkjet technology has proven to be very promising. With the firstgeneration of printed and processed MEAs, power densities offabricated cells are close to that of commercial MEAs in ourfuel cell test apparatus. Future experimental work will focus onbetter understanding the three dimensional structure of a printedelectrode and building new structures with inkjet technology toimprove electrode efficiencies while reducing costs compared toelectrodes made with previously known techniques [8,10,12].

Acknowledgments

A portion of the research described in this paper was per-formed in the Environmental Molecular Sciences Laboratory,a national scientific user facility sponsored by the Departmentof Energys Office of Biological and Environmental Researchand located at Pacific Northwest National Laboratory. Financialsupport was provided by the Laboratory Directed Research andDevelopment (LDRD) program.

References

[1] D. Bevers, N. Wagner, M.V. Bradke, Int. J. Hydrogen Energy 23 (1988) 57.[2] V. Mehta, J.S. Cooper, J. Power Sources 114 (2003) 32.[3] P. Calvert, Chem. Mater. 13 (2001) 3299.[4] James E. Smay, Joseph Cesarano III, Jennifer E. Lewis, Langmuir 18 (2002)

5429.[5] H.P. Le, J. Imaging Sci. Technol. 42 (1998) 49.[6] Anon, Economist.com, 2007.[7] E.A. Ticianelli, C.R. Derouin, S. Srinivasan, J. Electroanal. Chem. 251

(1988) 275.[8] M.S. Wilson, S. Gottesfeld, J. Appl. Electrochem. 22 (1992) 1.[9] M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 139 (1992) L28.

[10] M.S. Wilson, J.A. Valerio, S. Gottesfeld, Electrochim. Acta 40 (1995)355.

[11] G.S. Kumar, M. Taja, S. Parthasarathy, Electrochim. Acta 40 (1995) 285.[12] M. Uchida, Y. Aoyama, N. Eda, A. Ohta, J. Electrochem. Soc. 142 (1995)

463.[13] Wilson, M.S., Membrane catalyst layer for fuel cells, US Patent 5,234,777

(1993).[14] Y.-G. Chun, C.-S. Kim, D.-H. Peck, D.-R. Shin, J. Power Sources 71 (1998)

174.[15] S.S. Kocha, in: W. Vielstich, H.A. Gasteiger, A. Lann (Eds.), Handbook

of Fuel CellsFundamentals, Technology and Applications, vol. 3, JohnWiley & Sons, 2003, p. 538.

[16] J. Lin, J.K. Lee, M. Kellner, R. Wycisk, P.N. Pintauro, J. Electrochem. Soc.153 (2006) A1325.

[17] A.P. Saab, F.H. Garzon, T.A. Zawodzinski, J. Electrochem. Soc. 150 (2003)A214.

[18] E. Store, http://www.epson.com/cgi-bin/Store/consumer/consDetail.jsp.

Fabrication of polymer electrolyte membrane fuel cell MEAs utilizing inkjet print technologyIntroductionExperimentalCatalyst formulationPreparation of Nafion membranesPrinting of catalyst inks on Nafion membranesPreparation of MEA sample for conductivity, optical micrographs, and SEM and TEM measurementsFuel cell performance measurements

Results and discussionInk compatibilityMembrane/electrode assemblyPerformance of single cells

ConclusionAcknowledgmentsReferences