+ All Categories
Home > Documents > Integrated plasma synthesis of efficient catalytic...

Integrated plasma synthesis of efficient catalytic...

Date post: 10-Feb-2018
Category:
Upload: lydang
View: 218 times
Download: 0 times
Share this document with a friend
9
IOP PUBLISHING NANOTECHNOLOGY Nanotechnology 18 (2007) 305603 (9pp) doi:10.1088/0957-4484/18/30/305603 Integrated plasma synthesis of efficient catalytic nanostructures for fuel cell electrodes A Caillard 1,2 , C Charles 1,3 , R Boswell 1 and P Brault 2 1 Space Plasma, Power, and Propulsion Group, Research School of Physical Sciences and Engineering, The Australian National University, Canberra, ACT 0200, Australia 2 Groupe de Recherche sur l’Energ´ etique des Milieux Ionis´ es UMR6606, Universit´ e d’Orl´ eans—CNRS, BP 6744, F-45067 Orl´ eans Cedex 2, France E-mail: [email protected], [email protected], [email protected] and [email protected] Received 29 March 2007, in final form 29 May 2007 Published 29 June 2007 Online at stacks.iop.org/Nano/18/305603 Abstract A single plasma process involving three consecutive steps has been developed for producing high gas flow catalytic nanostructures on the electrodes of proton exchange membrane (PEM) fuel cells (FC). Using a high density helicon radio frequency (13.56 MHz) plasma, nickel is sputtered onto a porous carbon support. Changing the background gas from argon to methane/hydrogen allowed 2 μm long, 37 nm diameter carbon nanofibres (CNFs) to be grown by diffusion through the nickel clusters in a ‘tip growth’ mechanism at the relatively low temperature of 400 C. The third step involves plasma sputtering of platinum onto the CNFs, resulting in nanoclusters (3–8 nm) being formed on the periphery of the CNFs. Four FC cathodes were synthesized on carbon paper and PTFE/carbon loaded cloth (known as gas diffusion layer, GDL), both with and without CNFs, with the Pt/CNFs nanostructures grown on PTFE/carbon loaded cloth having the best FC performances. Compared with conventional FCs, the efficiency of sputtered platinum in the Pt/CNF based cathode is much higher than in a chemically deposited system over the entire range of operating current. This indicates that combination of different, simple, plasma techniques is an effective method for preparing highly efficient catalyst layers. (Some figures in this article are in colour only in the electronic version) 1. Introduction The development of fuel cells is considered to be an integral part of a sustainable hydrogen economy. Today, the proton exchange membrane fuel cell is promising in the mass market for automotive, stationary and portable applications [1, 2]. A PEMFC is an electrochemical cell fed with hydrogen fuel, which is oxidized at the anode, and oxygen that is reduced at the cathode. The protons released during the oxidation are conducted through the membrane to the cathode, whereas the electrons travel along the external electrical circuit. Each electrode is made up a porous mixture of electron and proton 3 Author to whom any correspondence should be addressed. conductor (typically carbon and proton polymer). Part of the optimization of an electrode design is to determine the optimum partition between the transport media (carbon, proton polymer and pores) for each of the three phases (electrons, protons and gases). This leads to a reduction of both critical transport losses, and the amount of high cost materials used for the catalyst (e.g. platinum). Decreasing the amount of Pt while increasing the Pt utilization efficiency has been one of the major concerns during the past decade [3–5]. A conventional, chemically produced, electrode is usually prepared by ink processes and consists of a gas diffusion layer, typically carbon Vulcan and PTFE particles and an active layer. The latter has a mixture of carbon Vulcan particles coated 0957-4484/07/305603+09$30.00 1 © 2007 IOP Publishing Ltd Printed in the UK
Transcript

IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 18 (2007) 305603 (9pp) doi:10.1088/0957-4484/18/30/305603

Integrated plasma synthesis of efficientcatalytic nanostructures for fuel cellelectrodesA Caillard1,2, C Charles1,3, R Boswell1 and P Brault2

1 Space Plasma, Power, and Propulsion Group, Research School of Physical Sciences andEngineering, The Australian National University, Canberra, ACT 0200, Australia2 Groupe de Recherche sur l’Energetique des Milieux Ionises UMR6606,Universite d’Orleans—CNRS, BP 6744, F-45067 Orleans Cedex 2, France

E-mail: [email protected], [email protected],[email protected] and [email protected]

Received 29 March 2007, in final form 29 May 2007Published 29 June 2007Online at stacks.iop.org/Nano/18/305603

AbstractA single plasma process involving three consecutive steps has beendeveloped for producing high gas flow catalytic nanostructures on theelectrodes of proton exchange membrane (PEM) fuel cells (FC).

Using a high density helicon radio frequency (13.56 MHz) plasma, nickelis sputtered onto a porous carbon support. Changing the background gasfrom argon to methane/hydrogen allowed 2 µm long, 37 nm diameter carbonnanofibres (CNFs) to be grown by diffusion through the nickel clusters in a‘tip growth’ mechanism at the relatively low temperature of 400 ◦C. Thethird step involves plasma sputtering of platinum onto the CNFs, resulting innanoclusters (3–8 nm) being formed on the periphery of the CNFs. Four FCcathodes were synthesized on carbon paper and PTFE/carbon loaded cloth(known as gas diffusion layer, GDL), both with and without CNFs, with thePt/CNFs nanostructures grown on PTFE/carbon loaded cloth having the bestFC performances. Compared with conventional FCs, the efficiency ofsputtered platinum in the Pt/CNF based cathode is much higher than in achemically deposited system over the entire range of operating current. Thisindicates that combination of different, simple, plasma techniques is aneffective method for preparing highly efficient catalyst layers.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

The development of fuel cells is considered to be an integralpart of a sustainable hydrogen economy. Today, the protonexchange membrane fuel cell is promising in the mass marketfor automotive, stationary and portable applications [1, 2]. APEMFC is an electrochemical cell fed with hydrogen fuel,which is oxidized at the anode, and oxygen that is reducedat the cathode. The protons released during the oxidationare conducted through the membrane to the cathode, whereasthe electrons travel along the external electrical circuit. Eachelectrode is made up a porous mixture of electron and proton

3 Author to whom any correspondence should be addressed.

conductor (typically carbon and proton polymer). Part ofthe optimization of an electrode design is to determine theoptimum partition between the transport media (carbon, protonpolymer and pores) for each of the three phases (electrons,protons and gases). This leads to a reduction of both criticaltransport losses, and the amount of high cost materials usedfor the catalyst (e.g. platinum). Decreasing the amount of Ptwhile increasing the Pt utilization efficiency has been one ofthe major concerns during the past decade [3–5].

A conventional, chemically produced, electrode is usuallyprepared by ink processes and consists of a gas diffusion layer,typically carbon Vulcan and PTFE particles and an active layer.The latter has a mixture of carbon Vulcan particles coated

0957-4484/07/305603+09$30.00 1 © 2007 IOP Publishing Ltd Printed in the UK

Nanotechnology 18 (2007) 305603 A Caillard et al

by Pt catalyst and a proton conductive polymer (typicallyNafion) spread on either carbon cloth or carbon paper madeof weaved or stacked carbon fibres (mean diameters of 10and 5 µm, respectively) [6–8]. In such a structure, the Ptcatalyst will be active if it is simultaneously in contact with thegas, the continuous electron-conducting media (porous carbonsupports) and the continuous proton-conducting polymer.Even with the most advanced conventional electrodes, thiscatalyst layer nanostructure does not ensure an optimumPt utilization [9]. Firstly, the Nafion polymer additionencapsulates Pt/C particles to form agglomerates (more or lessspherical with a diameter in the range 200–1000 nm) whichconsequently isolate some Pt/C particles from the electrodesupport by cutting off the electron pathways. Secondly, theseagglomerates themselves are flooded in a low gas diffusivityproton polymer. As a result, hydrogen and oxygen cannotdiffuse through the whole volume of these agglomerates andPt catalyst far from the agglomerate surface becomes inactive.Finally, gas diffusion, electron and proton conduction andwater removal are quite difficult in such a porous electrode dueto its complex shape.

Carbon nanotubes (CNT) or nanofibres (CNF) have beenpreviously proposed for electrochemical devices [10] as theycan replace traditional black carbon particles in PEMFCelectrodes because of their high electrical conductivity andchemical inertia. These nanostructures were used in powdersspread on the carbon backing [11–13] but the results didnot show many advantages over conventional carbon particlepowders due to the formation of the agglomerates mentionedabove. The feasibility of growing CNT directly on carbonpaper has been demonstrated using a silica gel to catalyzeCNT growth [14–16] or using a process of electro-depositionof Pt on CNTs grown by chemical vapour deposition oncarbon paper after an initial electro-deposition of cobaltcatalyst [17–20]. In these experiments, the subsequentmembrane electrode assembly (MEA) performance testsremained low, possibly due to a degradation of the electriccontact between carbon paper, CNT and Pt particles, and/orfrom the large diameter of the resulted Pt nanoclusters.

In previous publications, we demonstrated that plasmasputtering of Pt on carbon GDLs is a suitable techniquefor improving the catalyst utilization efficiency in aPEMFC [21–23]. These studies have shown typical diffusionlengths of Pt atoms into GDL between 400 nm and 2 µmand the formation of Pt nanoclusters with diameters in therange of 2–10 nm, depending on plasma deposition parameterssuch as argon pressure. At low current density, the electricalperformance of MEAs based on such Pt sputtered anodes issimilar to a conventional electrode based MEA but the plasma-prepared electrode has a Pt loading five times lower. Recently,a novel open structure active layer [9, 24–26] was presentedas an alternative to the conventional active layer especiallyfor fuel cell cathodes. This structure is composed of 30 nmdiameter, rod-like, electron conductors grown perpendicularlyto the membrane that are coated with well dispersed platinumnanoclusters and a 10 nm thin layer of proton conductor.In this active layer, where electron/proton conductors andgas channels are highly oriented, the catalyst utilizationmay approach 100% and performance can be dramaticallyimproved [9].

In the present paper a detailed description of the threeplasma processes is presented along with the preparation andthe characterization of the novel efficient active layer. A lowpressure (5 mTorr) radio frequency (RF) helicon plasma is usedas it can produce high density plasmas with great efficiency inrather large volumes. The design of the system is such thatmany sputter targets can be installed, allowing a number ofsequential processing steps to be developed without bringingthe substrate up to atmospheric pressure, thus eliminatingthe risk of possibly poisoning the layers. Additionally, byusing sequential steps, the processing time can be considerablyreduced.

As a first step, the helicon plasma is ignited in argongas and a nickel target is used to sputter a thin nickel filmonto the carbon electrode. This film comprises clusters thatserve to act as nuclei for the growth of CNFs. The secondstep substitutes the argon gas with a mixture of methane andhydrogen. The substrate is heated to around 400 ◦C and CNFsgrow perpendicularly to the substrate surface to a distance limitof about 2 µm. The third and last step uses argon again butanother target containing platinum serves to sputter-deposit thenanoclusters on the CNFs. Four FC cathodes based on differentcatalyst nanostructures were plasma-synthesized and testedin four MEAs. The electrical performances of the plasma-synthesized MEAs compare very favourably to conventionalchemically prepared MEAs.

2. Experimental details

2.1. Plasma-controlled synthesis of oriented active layerbased on well aligned carbon nanofibres

Carbon nanofibres are grown in a plasma processing chambercalled Southern Cross (shown in figure 1) in which a typicalbase pressure below 5 × 10−6 mbar is achieved using aturbo-molecular pump. The helicon plasma is generated bya 13.56 MHz radio frequency powered double saddle antennaplaced around the 15 cm diameter glass source tube attachedto a diffusion chamber shaped as a cross (55.5 cm × 55.5 cm).Three solenoids surround the source and the diffusion chamberto produce a magnetic field of about 80 G along the verticalaxis between the helicon source and the 72 mm diametersubstrate holder. The electrode holder can be DC biased andmanually rotated and moved along the vertical axis. It isusually placed 10 cm below the plasma source during helicon-PECVD and 18 cm below the plasma for the helicon plasmasputtering processes, respectively. A spiral heater can bringthe substrate temperature to a maximum of 400 ◦C. A stainlesssteel mask with two squares having 5 cm2 open area is fixedto the electrode holder. Each square is marked with a grid of9 smaller squares of 0.44 cm2 open area allowing uniform Pt/Cnanostructure growth over 4 cm2.

This is the first report of the growth of CNFs usinghigh plasma density, low gas pressure helicon-PECVD. Earlierplasma methods involved arc discharge [27], vaporizationusing laser [28], pyrolysis [29], chemical vapour deposition ofhydrocarbons using metal catalysts [30–32] and, more recently,chemical vapour deposition methods enhanced by DC plasma,RF plasma and microwave plasma [33–38].

Three consecutive plasma deposition steps have beendeveloped in order to deposit:

2

Nanotechnology 18 (2007) 305603 A Caillard et al

Figure 1. Schematic of the helicon PECVD reactor ‘SouthernCross’.

• the nickel catalyst used to grow the CNFs,• the CNFs themselves,• the nanoclusters of platinum.

First, an argon plasma is used to provide ions needed tosputter a negatively −300 V biased Ni target and to deposita nickel catalyst layer on the carbon paper or GDL. Theargon pressure and the plasma power are fixed at 5 mTorr and500 W. In these conditions, the plasma density measured by auncompensated Langmuir probe is close to 1011 cm−3 on thesubstrate holder. The amount of deposited nickel measured byRutherford backscattering spectroscopy (RBS) is 240×1015 Ptatoms cm−2, which corresponds to a dense layer of 25 nm. Thisis somewhat thicker compared to other reports and is requiredto prevent nickel sputtering of the biased Ni/electrode duringthe subsequent CNFs growth steps (which can occur at therelative low pressures used here).

Second, the substrate temperature is gradually increasedto 400 ◦C over a period of 15 min following which the sourceis fed with methane and hydrogen at 5 mTorr with a ratio of1:4 and a 500 W plasma was ignited. The real Ni catalysttemperature induced by the additional heating of the plasma

was not measured during CNF growth. It was thought to beslightly higher than 400 ◦C due to plasma heating reported byother authors [39]. With the substrate biased at −100 V, CNFswere grown for a period of 90 min.

Third and last, Pt nanoclusters are grown on the substratesusing a −300 V biased Pt target sputtered by ions generatedby an argon plasma of pressure and power 5 mTorr and500 W, respectively. The deposition rate of platinum has beenpreviously measured by RBS [40]. It is assumed that the Ptdeposition rate on silicon is equivalent to the Pt deposition rateon GDL, carbon paper and a carpet of CNFs. In the presentstudy, the deposition time has been adjusted to deposit therequired amounts of platinum.

Following the plasma processes, a thin film of Nafionpolymer is spread on this carpet of Pt/CNF cylindricalnanostructures to ensure proton access from the membrane tothe Pt catalyst nanoclusters at the cathode. Four different typesof catalytic layer were deposited: two on GDL and two oncarbon paper, one of each having carbon nanofibres deposited.All four have the same catalyst loading of 0.02 mgPt cm−2.A fifth electrode was made of a GDL covered by a carpet ofPt/CNF nanostructures as a catalytic layer with a Pt loading of0.1 mgPt cm−2. Displayed in figure 2 is a schematic of the fourelectrode structures and a conventional, chemically prepared,electrode configuration.

2.2. Microscopy measurements and assembly of PEM fuel celland testing

A cold field scanning electronic microscopy (SEM, HitachiS4500) was used to characterize the sputtered Ni film on thetwo porous carbon supports while atomic force microscopywas used to characterize the sputtered nickel film on theplane silicon support. Deposition times were identical forall microscopy analysis. The CNFs and Pt covered CNFson carbon paper and GDL were analysed by the HitachiS4500 SEM and also by transmission electronic microscopy(TEM, Hitachi H7100FA). For TEM investigation, the CNFs(or Pt covered CNFs) were scraped from the carbon paperand deposited on a copper microscopy grid covered by a thincarbon film. One drop of distilled water is then spread on thespecimen to disperse the CNF nanostructures.

Considering that most of the efficiency losses come fromthe cathode side [41], the five fabricated (4 plasma and1 conventional) electrodes were successively tested on the

Figure 2. Schematics of plasma prepared electrodes ((A)–(D)) and chemically prepared electrode (E): (A) Pt/CNF/GDL, (B) Pt/CNF/carbonpaper, (C) Pt/GDL, (D) Pt/carbon paper, (E) conventional chemically prepared electrode.

3

Nanotechnology 18 (2007) 305603 A Caillard et al

Figure 3. SEM images of (A) GDL, (B) a carbon fibre and a nickelplasma-coated carbon fibre: (C) before and (D) after annealing:25 nm nickel sputtering: power generator: 500 W; target bias:−300 V; pressure: 5 mTorr; deposition time: 16 min.

cathode side of a MEA in a fuel cell station. Standard ETEKelectrodes (LT140EW) with a Pt loading of 0.5 mgPt cm−2

were used at the anode side. A Nafion 5 wt% solution wasspread onto each electrode using a pipette to ensure protonaccess from the membrane to the Pt nanoclusters. The Nafionloading covering is 1 mg cm−2 for the five prepared electrodesand 1.5 mg cm−2 for the five conventional ETEK electrodes.MEAs were prepared by hot pressing a humidified Nafion115 membrane with the electrodes at 130 ◦C for 120 s under apressure of 40 kg cm−2 to ensure good contact between the cellcomponents. Teflon gaskets, carbon bipolar plates, and coppercurrent collectors were added to make the final cell tests. Thefuel cell tests are carried out by recording current–voltagecurves (also called polarization curves) in a single 5 cm2 fuelcell test station at a cell temperature of 80 ◦C and back pressureof 3 bar. The flow rates of pure hydrogen and pure oxygenare fixed at 100 standard cubic centimetres per minute (sccm).Before entering the fuel cell, the hydrogen and oxygen werehumidified at 80 ◦C. The same procedure is followed for eachMEA test, even if the conditions are not necessarily optimumfor each assembly. We preferred not to change the test andassembly conditions for each complete fuel cell so as to allowcomparisons to be made under comparable conditions.

3. Results and discussion

3.1. Ni catalyst layer and CNF morphology

Figure 3 shows a SEM image of the uncoated GDL surface(A), of a single carbon fibre in a sheet of carbon paper before(B) and after (C) Ni sputter deposition. Figure 3(D) showsa SEM micrograph of nickel on carbon paper after thermalannealing. The dimensions of the outermost (most visible) of

Figure 4. AFM images of nickel clustered film: (A) before annealingand (B) after annealing: 25 nm nickel sputtering: power generator:500 W; target bias: −300 V; pressure: 5 mTorr; deposition time:16 min.

the uncoated carbon particles of the GDL are in the range of20–45 nm with a average size of 30 nm based on examining100 particles randomly chosen from the SEM micrograph offigure 3(A). Some larger non-spherical particles of PTFE arealso visible. The Ni morphology is difficult to analyse using theSEM because the Ni deposition is continuously affected by theelectron beam of the microscope during the time the surface isbeing imaged. Agglomerations of Ni nanoclusters are observedon the carbon fibre after plasma sputtering (figure 3(C)). Aftera 15 min long thermal anneal, figure 3(D) shows that the Nideposition has become denser and the Ni nanoclusters havedisappeared.

In order to further investigate this process, the Nisputtered on silicon was analysed by AFM, which was notexpected to change the Ni morphology during the investigation.Figure 4(A) displays the nickel layer sputtered on siliconbefore thermal annealing, showing nickel nanoclusters in therange of 10–20 nm (as displayed on the corresponding sizedistribution) with an average diameter of 15 nm. Afterthe 15 min thermal annealing, 30 nm diameter clustersform (figure 4(B)). Nevertheless, some 15 nm diameter Ninanoclusters are still evident on this micrograph as shown bythe corresponding statistical size distribution.

From the SEM images shown in figure 5, it can be seenthat CNFs grow normal to the surface of the Ni catalyzedcarbon paper or GDL. Any amorphous carbon produced bythe plasma process is removed by hydrogen that is dissociated

4

Nanotechnology 18 (2007) 305603 A Caillard et al

Figure 5. SEM images of CNF on a gas diffusion layer (GDL): (A) side view, (B) top view; and on a carbon fibre: (C) top view, (D) sideview: (i) 25 nm nickel sputtering: power generator: 500 W; target bias: −300 V; pressure: 5 mTorr; deposition time: 16 min. (ii) CNFs grown:power generator: 500 W, CH4:H2 = 1:4; pressure: 5 mTorr; substrate temperature: 400 ◦C; substrate bias: −100 V; deposition time: 90 min.

and thus activated in the plasma phase. No CNF appear onthe carbon fibre side which is neither exposed to the CH4:H2

plasma nor coated with a Ni cluster based film. From astatistical analysis of figure 5(B), the CNFs have a meandiameter of 37 nm with a statistical distribution showing twopeaks centred around 25 and 38 nm, as can be seen in the insetof figure 5(B). These peaks correspond to the short and thelong CNFs that can be seen in the background and foregroundof figure 5(B). Their size distribution may correspond to thedouble peak distribution of the Ni nanoclusters observed infigure 4(B) after the annealing step. The majority of theCNFs are about 2 µm in length as shown in figure 5(D) andthe corresponding mean growth rate is close to 20 nm min−1

which is quite low compared to other reports using plasmaor thermal CVD processes, which are typically ten timeshigher. This is possibly a result of the low CH4:H2 pressure(5 mTorr) which results in a low gas phase density of carboncompared to the higher pressures generally used [42]. Thisinitial growth rate gradually decreases during deposition, aspreviously reported [43], thereby limiting the CNF length to2 µm long for the present deposition parameters.

The effective surface area of the CNF/carbon paper andCNF/GDL nanostructure shown in figure 5 was measured to bein the range 15–20 m2 g−1 by nitrogen adsorption (Brunauer–Emmett–Teller method), which is higher than that of carbonpaper (2 m2 g−1) and similar to that of GDL (15 m2 g−1).The CNF density on the GDL surface is estimated to 2.2 ×1010 cm−2 from SEM micrograph analysis.

A TEM image of the CNF before platinum deposition isshown in figure 6 and simultaneous energy-dispersive x-rayspectroscopy (EDS) analysis showed that the black particleat the CNF tip is nickel. This result suggests a ‘tip’ growthmodel [44, 45], where the Ni is transported away fromthe carbon fibre (or GDL). Most of the particles adopt acharacteristic faceted or biconal shape at the upper end of theCNF, as is clear in figure 6(A). Most Ni particles can be seen

Figure 6. TEM images of nickel particle in triangular shape (A) andproposed ‘tip’ growth model for CNF (B). (i) 25 nm nickelsputtering: power generator: 500 W; target bias: −300 V; argonpressure 5 mTorr; deposition time 16 min. (ii) CNFs grown: powergenerator: 500 W, CH4:H2 = 1:4; pressure 5 mTorr; substratetemperature: 400 ◦C; substrate bias: −100 V; growth time: 90 min.

to be surrounded by a thin carbon layer on the face exposedto the CH4:H2 plasma, an important advantage for the fuelcell application. This carbon film prevents both the nickeldissolving in the fuel cell’s corrosive environment, and thenickel’s diffusion throughout the hydrated membrane, which

5

Nanotechnology 18 (2007) 305603 A Caillard et al

Figure 7. Images of a CNF surrounded by Pt nanoparticles: (A) SEM image, (B) TEM images of the upper end of a CNF, (C–F) close up ofPt nanoclusters along the CNF. Process parameters: (i) 25 nm nickel sputtering: power generator: 500 W; target bias: −300 V; argon pressure:5 mTorr; deposition time: 16 min. (ii) CNFs grown: power generator: 500 W, CH4:H2 = 1:4, pressure: 5 mTorr, substrate temperature:400 ◦C; substrate bias: −100 V; growth time: 90 min. (iii) 0.02 mgPt cm−2 platinum sputtering: power generator: 500 W; target bias: −300 V;argon pressure: 5 mTorr; deposition time: 4 min.

could result in blocking proton conduction channels. Theseobservations of the carbon nanofibre growth and previousstudies [46] add further weight to the hypothesis that a tipgrowth mechanism is operating.

The heat released from the ion bombardment of the biasednickel coated substrate and the exothermic reaction of methanedecomposing in the plasma phase or on the Ni cluster layerresults in a temperature gradient being established across theNi layer. Carbon atoms arriving at the nickel surface can thendiffuse into the Ni layer through the grain/particle boundary orthe specific substrate plane. Segregation and precipitation ofcarbon atoms on the other side of the Ni layer occurs becausethe solubility of carbon atoms decreases inside the Ni layerwhen the temperature decreases, resulting in the formation ofwell defined micro-cracks between Ni clusters where graphenelayers form. During CNF growth, the Ni nanocluster istransported away from the substrate; graphene layers near theunderside of the Ni tips are parallel to the lateral plane of thefaceted Ni tip, thus forming a typical stacked structure [47].On the outer side of the CNF wall, the graphene layers becomeperpendicular to the growth direction. At the end of thedeposition process when the plasma is turned off, a carbonfilm forms on the exposed Ni cluster face due to the nickeltemperature decreasing in the presence of CH4:H2 gas. Theschematic on figure 6(B) shows the main steps correspondingto this ‘tip’ growth mechanism from a Ni cluster base layer.

Around the bottom of such 2 µm stacked CNFs someshorter nanofibres can be seen on TEM micrographs. Theseshort nanofibres have spherical Ni particles surrounded by a

thicker carbon film on the face exposed to the plasma withno clear evidence of graphene layers on their side wall. Thenumber of such short nanofibres increases when the plasmadeposition parameters have not been optimized, typically forlower substrate temperature (below 350 ◦C). For a thickerinitial Ni layer some small Ni metallic inclusions can be seeneither inside or on the surface of CNF. In this case, no growthof secondary small CNF is observed on the side wall of theCNF.

3.2. Oriented active layer based on aligned Pt/CNFnanostructure morphology

A controlled amount of Pt was sputtered on these verticallyaligned CNFs by using two deposition times of 4 and 20 min,which resulted in Pt loadings of 0.02 and 0.1 mgPt cm−2,respectively. Each CNF is coated by well dispersednanoclusters of platinum determined from EDS spectra takenduring the TEM investigation. SEM pictures in figure 7show the upper end of the CNF after 4 min Pt deposition(0.02 mgPt cm−2). From figures 7(C) to (F), the mean diameterof these Pt nanoclusters decreases from 8 to 3 nm between 100and 600 nm from the top of the CNF. Assuming that each Ptnanocluster is perfectly spherical, the size dispersion can beestimated by dividing the number of Pt atoms on the sphericalsurface by the number of total Pt atoms in the spherical volume.From the Pt nanocluster diameters measured in figure 7(B), theestimated dispersion increases from 18% to 50% from the topto bottom of the CNF.

6

Nanotechnology 18 (2007) 305603 A Caillard et al

The average diameter of these nanoclusters all along these2 µm CNFs is close to 5 nm which is slightly higher than that ofcommercial electrodes. The platinum loading on such a CNFbased layer is around 6 wt%, (assuming that carbon density ina CNF is 2.25 g cm−3). The Pt nanoclusters density is higher atthe top of the CNF which will be in contact with the electrode–membrane interface where the Pt catalyst is the most activeduring fuel cell operation [48].

Unidirectional aligned nanofibre structure is expected toallow a free flow of fuel gas when exposed to an effectiveamount of catalyst as compared to the convoluted conventionalactive layer composed of randomly stacked Nafion/Pt/Cagglomerates. The use of carbon nanofibres directly grownon carbon backing and the resulting longitudinal electronicpathway prevents the carbon particles that form the electrodebacking from being isolated by the Nafion. This non-inkprocess ensures a good electrical contact between sputteredPt catalyst nanoclusters and the electrical support. Therefore,all nanoclusters are theoretically able to be in contact withthe external circuit of a PEMFC. Moreover, this catalyticconfiguration avoids a large part of the platinum catalystbeing located deep inside the Nafion flooded agglomerates andconsequently poorly accessible for the fuel gas.

3.3. Influence of the plasma based cathode nanostructure onfuel cell efficiency

The current–voltage and power density curves of the fourMEAs corresponding to the four cathode configurations offigure 2 are shown in figure 8: Pt/CNF/GDL (MEA1,figure 2(A)), Pt/CNF/carbon paper (MEA2, figure 2(B)),Pt/GDL (MEA3, figure 2(C)) and Pt/carbon paper (MEA4,figure 2(D)). Each cathode has an ultra-low Pt loading of0.02 mgPt cm−2, which has been sputtered either on carbonpaper, GDL or CNFs. This comparison clearly shows that thepresence of CNFs as catalyst support significantly improves thefuel cell performance.

Although the relative increase of performance withCNF on carbon paper is dramatic over the whole range ofinvestigated current, the voltage is consistently lower than thatobtained for the electrodes on carbon cloth with a GDL. Eachconventional ETEK GDL is composed of PTFE particles tomake it more hydrophobic and contribute to water evacuationduring operation, whereas the carbon paper based cathode iswetted easily due to its poor hydrophobicity [24]. In thelatter case, the catalytic layer pores are filled with liquid waterwhich prevents oxygen from getting to the catalytic sites anddecreases the resulting performance. Two other mechanismsmay play a role in the improved performance resulting fromthe presence of GDLs. First, the GDL surface is relativelyflat compared to that of porous carbon paper; this could resultin a better mechanical contact and a higher catalytic surfacearea at the membrane/electrode interface. Additionally, thesetwo carbon supports have very different porosities (about 25%for the GDL versus 80% for the carbon paper). Thus, Nafionmay cross the carbon paper and cover its back face, therebyinterfering with the electrical contact between the bipolar plateand the synthesized electrode.

Improvement of the cell performance with CNFs grownon GDLs is more evident at higher current densities: at

Figure 8. Cell voltage E versus current density j curves and powerversus current density j in a single 5 cm2 surface area PEMFC withdifferent Pt/C cathode structures, ultra-low Pt loading of0.02 mgPt cm−2: cathode based on Pt/CNF/GDL (MEA1),� cathode based on Pt/CNF/carbon paper (MEA2), � cathode basedon Pt/GDL (MEA3) and • cathode based on Pt/carbon paper(MEA4); open symbols stand for power density. (PH2 = 3 bar; H2

flow = 100 sccm; PO2 = 3 bar; O2 flow = 100 sccm, Tcell = 80 ◦C,membrane Nafion 115.)

600 mA cm−2 for example, the power densities displayed infigure 8 are close to 170 and 210 mW cm−2 for the Pt/GDLand Pt/CNF/GDL cathode, respectively. The maximum powerdensities achieved are 170 for MEA3 (without CNF) and230 mW cm−2 for MEA1 (with CNF). At high flow rates,the porous, well aligned, structure in the CNF based electrodewould allow a much improved flow of fuel gas and water andthus contribute to the performance improvement.

At low current density, CNFs grown on GDLs result in aslightly improved cell performance. The slope of polarizationcurve for medium current density is smaller with CNFsindicating a lower cell resistance. These results are likelyrelated to the increase of platinum active sites. Previous workon the structure of Pt/GDL plasma sputtered electrodes withoutCNF support have reported a decrease of platinum active sites.Indeed, these studies showed the existence of a practical limitfor dispersing platinum in a GDL. Pt deposition by plasmasputtering on GDL leads to Pt nanoclusters (mean diameterclose to 3.0 nm) which agglomerate and form a relatively densePt layer surrounding each carbon particle for increasing Ptcatalyst loading (depending on plasma deposition parameters).

Moreover, these studies have shown most of the effectivePt catalyst loading is contained in a GDL thickness of 400 nm,even if the maximum diffusion lengths of Pt atoms into GDLcan reach 2 µm (depending on plasma deposition parameters).On the other hand, a CNF support improves the Pt nanoclustersdispersion and avoids the formation of a Pt nanocluster bindinglayer. The unidirectional porosity in the aligned CNF layer isthought to contribute to a higher diffusion length of Pt atomsduring plasma sputtering compared to direct Pt sputtering onGDL [13–15].

3.4. Comparison between a MEA based on a Pt/CNF/GDLcathode and MEA based on two conventional electrodes

Two MEAs based on two Pt/CNF/GDL cathodes (figure 2(A))with different Pt loading and two ETEK anode have been com-

7

Nanotechnology 18 (2007) 305603 A Caillard et al

Figure 9. Cell voltage E versus current density j curves and powerversus current density j in a single 5 cm2 surface area PEMFC withPt/C cathodes with different sputter-deposited loadings of platinumon CNF/GDL composite and comparison with a chemically preparedcathode. cathode based on Pt/CNF/GDL with 0.02 mgPt cm−2

(MEA1); � cathode based on Pt/CNF/GDL with 0.1 mgPt cm−2

(MEA5), � chemical cathode Pt/Vulcan 0.5 mgPt cm−2 (MEA6).(PH2 = 3 bar; H2 flow = 100 sccm; PO2 = 3 bar; O2

flow = 100 sccm, Tcell = 80 ◦C, membrane Nafion 115.)

pared to a MEA with two ETEK electrodes (figure 2(E)). Thetest and assembly procedure is the same for each fuel cell.Figure 9 shows the polarization and power density curves ob-tained with these three MEAs: MEA1 (Pt/CNF/GDL cathodewith 0.02 mgPt cm−2) and MEA5 (Pt/CNF/GDL cathode with0.1 mgPt cm−2). The two CNF based cathodes lead to lowerelectrical performances than that of a ETEK electrode basedassembly MEA6 in terms of measured cell voltage or powerdensity at a given current density. However, the cathodic Ptloadings of MEA1 and MEA5 are 25 and 5 times lower thanthat of MEA6 and the maximum achieved power densities areonly 40% and 20% lower: 230, 300 and 370 mW cm−2 for thethree assemblies MEA1, MEA5 and MEA6.

This clearly indicates that the ratio of platinum used inthese two oriented active layers is at least 15 and 4 times higherthan that of the conventional one. Considering that plasmafabrication is a well established, clean and well-controlledprocess used by microelectronic manufacturers, it could be agood way for efficient and cost effective fuel cell electrode andMEA fabrication.

4. Conclusion

Vertically aligned carbon nanofibres surrounded by welldispersed platinum nanoclusters has been successfully grownon carbon paper and gas diffusion layers by combininghelicon-PECVD and helicon plasma sputtering processes.Such unidirectional Pt/CNF nanostructures on an open-poreconductive substrate were used as an oriented active layer formass and charge coupled transport in a fuel cell electrode.

The electrical performances of four MEAs based ondifferent cathode configurations having an ultra-low Pt loading(0.02 mgPt cm−2) were compared in a single 5 cm2 fuel cell teststation. The MEA based on Pt/CNF nanostructures grown onhydrophobic GDL leads to the best maximum power density(230 mW cm−2). The maximum achieved power density is40% lower without CNF presence on GDL and 65% lower

without hydrophobic GDL as catalyst support. Subsequently,a Pt/CNF/GDL electrode with a Pt loading of 0.1 mgPt cm−2

was compared with a higher Pt loaded standard chemicallyprepared electrode (0.5 mgPt cm−2) on the cathode side ofa MEA. The activity of the sputtered catalyst on CNF islower than that obtained with a standard cathode in termsof maximum achieved power density (300 mW cm−2 versus370 mW cm−2). However, the platinum utilization efficiencyin the oriented Pt/CNF nanostructures based cathode is atleast four times higher than that of the conventional one.This study has shown that plasma techniques are an effectivemethod for preparing thin oriented active layers for fuel cellsbased on vertically aligned carbon nanofibres surrounded by Ptnanoclusters. It allows a good control of fabrication parametersin order to correctly distribute the volume of the catalystlayer between the catalyst supported on carbon nanostructureand the unidirectional pores which allow adequate flow offuel gas while exposing the fuel to an effective amount ofcatalyst. In addition, these techniques allow the fabrication ofPEMFC electrode with lower platinum loading (compared toa chemically based electrode) without a decrease in efficiency.Hence, they can be used for the optimization of other fuel cellssuch as direct methanol fuel cells by combining platinum andanother catalyst (to protect platinum poisoning from CO).

Acknowledgments

AC gratefully acknowledges the ARC Australian ResearchNetwork for Advanced Materials, CNRS and ANU forfinancial support. We would like to thank P Alexanderfor precious technical help on Southern Cross. F Llewelynfrom the Electronic Materials Engineering Department (ANU),F Brink and S Stowe from the Electron Microscopy Unit(ANU) are acknowledged for assistance with the EDS analysis.

References

[1] Hoogers G 2002 Fuel Cell Technology Handbook (Boca Raton,FL: CRC Press)

[2] Cleghorn S J C, Ren X, Springer T E, Wilson M S,Zawodzinski C, Zawodzinski T A and Gottesfeld S 1997 Int.J. Hydrog. Energy 12 1137

[3] Borgward R 2001 Transp. Res. D 6 199[4] Costamagna P and Srinivasan S 2001 J. Power Sources

102 242[5] Litster S and McLean G 2004 J. Power Sources 130 61[6] Shin S J, Lee J K, Ha H Y, Hong S A, Chun H S and Oh I H

2002 J. Power Sources 106 146[7] Mehta V and Cooper J S 2003 J. Power Sources 114 32[8] Lobato J, Rodrigo M, Linares J and Scott K 2006 J. Power

Sources 157 284[9] Middelman E 2002 Fuel Cells Bull. 11 9

[10] Baughman R, Zakhidov A A and De Heer W A 2002 Science297 787

[11] Li W, Liang C, Zhou W, Qiu J, Zhou Z, Sun G and Xin Q 2003J. Phys. Chem. B 107 6292

[12] Steigerwalt E, Deluga G A, Cliffel D E and Lukehart C M 2001J. Phys. Chem. B 105 8097

[13] Steigerwalt E, Deluga G A and Lukehart C M 2002 J. Phys.Chem. B 106 760

[14] Smiljanic O, Dellero T, Serventi A, Lebrun G, Stanfield B L,Dodelet J P, Trudeau M and Desilets S 2001 Chem. Phys.Lett. 342 503

[15] Sun X, Li R, Villiers D, Dodelet J P and Desilets S 2003 Chem.Phys. Lett. 379 99

8

Nanotechnology 18 (2007) 305603 A Caillard et al

[16] Sun X, Li R, Stanfield B, Dodelet J P and Desilets S 2004Chem. Phys. Lett. 394 266

[17] Thostenson E T, Li W Z, Wang D Z, Ren Z F and Chou T W2002 J. Appl. Phys. 91 6034

[18] Wang C, Waje M, Tang J M, Haddon R C and Yan Y 2004Nano Lett. 4 345

[19] Waje M M, Wang X, Li W and Yan Y 2005 Nanotechnology16 395

[20] Wang X, Waje M and Yan Y 2005 Electrochem. Solid-StateLett. 8 42

[21] Brault P, Caillard A, Thomann A L, Mathias J, Charles C,Boswell R W, Escribano S, Durand J, Roualdes S andSauvage T 2004 J. Phys. D: Appl. Phys. 37 3419

[22] Caillard A, Brault P, Mathias J, Charles C and Boswell R W2005 Surf. Coat. Technol. 200 391

[23] Brault P, Roualdes S, Caillard A, Thomann A L, Mathias J,Durand J, Coutanceau C, Leger J M, Charles C andBoswell R W 2006 Eur. Phys. Appl. Phys. 34 151

[24] Du C Y, Cheng X Q, Yang T, Yin G P and Shi P F 2005Electrochem. Commun. 7 1411

[25] Du C Y, Yin G P, Cheng X Q and Shi P F 2006 J. PowerSources 160 224

[26] Du C Y, Yang T, Shi P F, Yin G P and Cheng X Q 2006Electrochim. Acta 51 4934

[27] Ebessen T W and Ayayan P M 1992 Nature 358 220[28] Thess A et al 1996 Science 273 483[29] Amelinck S, Zhang X B, Bernaerts D, Zhang X F,

Ivanov V and Nagy J B 1994 Science 265 635[30] Li W Z, Xie S S, Qian L X, Chang B H, Zou B S, Zhou W Y,

Zhao R A and Wang G 1996 Science 274 1701[31] Fan S, Chapline M G, Franklin N R, Tombler T W, Cassel A M

and Dai H 1999 Science 283 512

[32] Thomann A L, Salvetat J P, Breton Y, Andreazza-Vignolle Cand Brault P 2003 Thin Solid Films 428 242

[33] Delzeit L, McAninch I, Cruden B A, Hash D, Chen B,Han J and Meyyappan M 2002 J. Appl. Phys. 91 6027

[34] Ren Z F, Huang Z P, Xu J W, Huang J H, Bush P,Siegal M P and Proventio P 1998 Science 282 1105

[35] Boskovic B O, Stolojan V, Khan R U A, Hag S, Ravi S andSilva P 2002 Nat. Mater. 1 165

[36] Boskovic B O, Golovko V B, Cantoro M, Kleinsorge B,Chuang A T H, Ducati C, Hofmann S, Robertson J andJohnson B F G 2005 Carbon 43 2643

[37] Minea T M, Point S and Granier A 2004 Appl. Phys. Lett.38 1244

[38] Hofmann S, Ducati C, Robertson J and Kleinsorge B 2003Appl. Phys. Lett. 83 135

[39] Teo K B K et al 2004 Nano Lett. 4 921[40] Caillard A 2006 PhD Universite d’Orleans and the Australian

National University[41] Chen L H, AuBuchon J F, Chen I C, Dariao C, Ye X R,

Jin S and Wang C M 2006 Appl. Phys. Lett. 88 033103[42] Bower C, Zhu W, Jin S and Zhou O 2000 Appl. Phys. Lett.

77 830[43] Bower C, Zhou O, Zhu W, Werder D J and Jin S 2000 Appl.

Phys. Lett. 77 2767[44] Endo M 2002 Appl. Phys. Lett. 80 1267[45] Baker R T K and Harris P S 1978 Chemistry and Physics of

Carbon vol 14 (New York: Dekker) p 83[46] Han J et al 2002 Appl. Phys. Lett. 91 483[47] Baker R T K, Barber M A, Harris P S, Feates F S and

Waite R J 1972 J. Catal. 26 51[48] Antoine O, Bultel Y, Ozil P and Durand R 2000 Electrochim.

Acta 45 4493

9


Recommended