Mixed potential NOx sensors using thin film electrodes and electrolytes for stationary reciprocating engine type applications

A

phsmdwemw©

K

1

pbdtai0NUSiu

n

0d

Sensors and Actuators B 119 (2006) 398–408

Mixed potential NOx sensors using thin film electrodes and electrolytesfor stationary reciprocating engine type applications

Eric L. Brosha ∗, Rangachary Mukundan, Roger Lujan, Fernando H. GarzonLos Alamos National Laboratory, MS D429, Electrochemical Materials and Devices Group, Los Alamos, NM 87545, USA

Received 8 August 2005; received in revised form 15 December 2005; accepted 19 December 2005Available online 24 January 2006

bstract

Mixed potential sensors using dense, thin film metal oxide working electrodes, Pt counter electrodes, and thin film YSZ electrolytes on Al2O3

olycrystalline and sapphire substrates were prepared and studied between 450 and 650 ◦C. Their response to NO, NO2, CO, and C1 and C3ydrocarbons in 10.4% O2/N2 balance and in air atmospheres was characterized. The lanthanum chromite-based sensors showed preferentialensitivity to NO2 with cross sensitivity to CO and non-methane hydrocarbons such as C3H6 and C3H8 with the highest NOx sensitivity andinimal CO/HC cross sensitivity exhibited by a sensor prepared with a La0.8Sr0.2CrO3 working electrode. Studies of the Mg-doped LaCrO3

evices conducted for up to 800 h at 600 ◦C showed minimal aging in these devices. In contrast, sensors fabricated with spinel NiCr2O4 or ZnFe2O4

orking electrodes showed preferential sensitivity to HCs and minimal sensitivity to NO2. Sensors fabricated with Y0.16Tb0.3Zr0.54O2−∂ working
lectrodes required operating temperatures in excess of 600 ◦C and, like the spinels, showed minimal NOx sensitivity. A prototype three-electrodeixed potential device for the detection of multiple gases was also demonstrated. The use of a heated Pt black pre-catalyst upstream to the sensoras demonstrated to effectively remove the CO and hydrocarbon response.2005 Elsevier B.V. All rights reserved.
ofaTectataa[fI

eywords: Electrochemical sensor; NOx sensor; Mixed potential

. Introduction

Reciprocating gas engines are well suited for small-scaleower generation. They provide high reliability, inexpensiveack-up power, power for remote locations, and power pro-uction during peak times to reduce business costs. They arehe fastest-selling and least expensive distributed energy gener-ting systems in the world today [1]. Commercial reciprocat-ng engine systems using natural gas to produce power from.5 kW to 10 MW have efficiencies between 37 and 40%, andOx emissions as low as 1 g/bhp-h. Several of the goals of theS Department of Energy’s Advanced Reciprocating Engineystem (ARES) program are to improve fuel efficiency and flex-

bility, and reduce emissions of reciprocating engine systems
sed for distributed energy generation applications.
Advanced reciprocating engines require new sensor tech-ologies to reduce and control emissions and to improve

∗ Corresponding author. Tel.: +1 505 665 4008; fax: +1 505 665 4292.E-mail address: [email protected] (E.L. Brosha).

etoepN

925-4005/$ – see front matter © 2005 Elsevier B.V. All rights reserved.oi:10.1016/j.snb.2005.12.044

verall combustion efficiency. Sensors are required for engineeedback control, NOx emissions monitoring and control,nd sensors to control the regeneration of reduction catalysts.he NOx emissions target for natural gas-fired reciprocatingngines is 0.1 g/bhp-h (roughly a factor of ten reduction fromurrent levels). A new NOx sensing technology is necessaryo meet the sensor requirements for low cost, sulfur tolerance,nd a long, stable lifetime. Since sensors developed overhe last decade for lean-burn gasoline engines will not meetll of these requirements, we have worked to develop COnd HC mixed potential sensors based on doped zirconia2–4] and ceria [5] electrolytes and metal oxide electrodesor automotive on-board diagnostic applications (OBD-I).

Mixed potential sensors based on oxygen ion conductinglectrolytes have been investigated extensively for the detec-ion of CO, H2, HCs, and NOx. The competing reactions of
xygen reduction (oxidation) and analyte oxidation (reduction)stablish the mixed potential at an electrode. For NOx mixedotential sensors, the competing reactions of NO oxidation andO2 reduction establishes the mixed potential. Eqs. (1) and (2)

d Act

i

N

Ilt(epa

sasgHmtssaue

mcfaptLvYemdhdHs

ifipo

2

2

mgwfl

abh

ucttmsmtsA1

tbsthsttoaac

2b

usMwmpsTtpartmsspC

2

E.L. Brosha et al. / Sensors an

llustrate this for NO/NO2.

12 O2(g) + VO

•• + 2e− = OO (1)

O(g) + OO = NO2(g) + VO•• + 2e− (2)

n these equations, VO•• is an oxygen vacancy and OO is a

attice oxygen in the electrolyte. Depending on the rates of thewo reactions, a non-equilibrium potential that is more negativepositive for NO2 reduction) than that predicted by the Nernstquation develops under oxygen rich conditions [6]. This mixedotential, being dependent on kinetic factors, is expected to bestrong function of the electrode material [7].

The early reported mixed potential sensors utilized yttria-tabilized zirconia or Na beta alumina electrolytes with goldnd platinum metal electrodes [8]. Many other groups haveubsequently investigated mixed potential sensors based on oxy-en ion-conducting solid electrolytes for the detection of CO,2, HCs, and NOx [2–19]. We have focused on the develop-ent of mixed potential sensors based on metal oxide elec-

rodes for automotive applications [2–4]. Our work on NOx

ensor development expands on this research with emphasis onelecting refractory metal oxide electrode/electrolyte materialsnd sensor designs that are simple and inexpensive to man-facture and will withstand the rigors of the engine exhaustnvironment.

In our work presented in this paper, we begin to studyixed potential sensors based on Mg and Sr-doped lanthanum

hromite and Pt electrodes with thin film YSZ solid electrolytesor NO/NO2 sensitivity, cross sensitivity to gas species suchs CO, methane, and non-methane hydrocarbons (propane andropylene), and sensor voltage level stability. Thin films ofhe lanthanum chromite perovskites La0.8(Sr,Mg)0.2CrO3 andaCr0.8Mg0.2O3 were prepared using RF magnetron sputteringia a mixed phase, fluoride precursor route [20]. Thin films ofSZ electrolyte (<15 �m) were prepared using electron beam

vaporation. Planar sensor configurations using polished alu-ina or single crystal sapphire substrates were studied because

evices of this nature will facilitate incorporation of a thin filmeater in the future. Also, as will be shown in this work, planarevice structures may potentially allow simultaneous NOx andC/CO sensing from a single device for applications beyond

tationary reciprocating applications.In addition, we present in this paper some preliminary work

n the suitability of doped fluorite and spinel metal oxide thinlms as potential NOx-specific working electrodes for mixedotential sensors and we will report methods for preferentialxidation (reduction) of interference gas species [21].

. Experimental

.1. Preparation of perovskite metal oxide electrodes

Sr and Mg doped LaCrO3 thin films were prepared using RF
agnetron sputtering via fluoride/metal composite (off-axis) tar-
ets. Direct sputtering of the oxide from a stoichiometric targetas not used; rather, electrodes were prepared using an indirectuoride and metal multiphase powder target route. A complete

rp

uators B 119 (2006) 398–408 399

nd thorough description of this method and its advantages haseen covered in detail elsewhere [20]; only a summary is givenere.

Powder sputter targets (50.8 mm diameter) were preparedsing stoichiometric amounts of LaF3, MgF2 (or SrF2) andhromium metal. A 1 cm2 Al2O3 substrate was masked suchhat a 4.3 mm × 10 mm stripe of the substrate was exposed tohe sputter target. To measure thickness and deposition rates, a

asked piece of polished sapphire was mounted next to the sub-trate. The step created on this witness sample from the shadowask was then used to measure the film thickness produced in

he PVD run using a DEKTAK profilometer. The sputter depo-ition was carried out at 400 ◦C. The films were grown in a UHPr atmosphere at 40 mTorr pressure and at a power setting of25 W.

The fluoride/chromium composite films were removed fromhe deposition system after sputtering, placed into an aluminaoat, and transferred to a tube furnace. A switching system waset-up to switch from dry Ar to humidified Ar (gas bubbledhrough water at 25 ◦C) using a three-way valve. The films wereeated at 10 ◦C/min to 800 ◦C in dry Ar. At 800 ◦C, the gas waswitched to humidified Ar. This action caused the decomposi-ion of the lanthanum and alkali earth fluoride phases and, withhe presence of the Cr metal, lead to the formation of the per-vskite phase. The sample was annealed at 800 ◦C for 15 min,fter which the sample was then heated to 1000 ◦C in dry Arnd held for an hour to allow for grain growth in the lanthanumhromite film.

.2. Direct oxide RF magnetron sputtering and electroneam evaporation

Direct R.F. sputtering from stoichiometric oxide targets wassed to produce films with the desired spinel and fluorite crystaltructures. NiCr2O4 and ZnFe2O4 powders were obtained from

arketech International, Inc. The commercial spinel powdersere packed and uni-axially pressed at 1760 N into a speciallyade 50.8 mm dia. Cu target sputtering cup. The targets were

laced into the sputtering system and off-axis RF magnetronputtered at 125 W and 40 mTorr pure UHP Ar atmosphere.he substrate heater temperature was maintained at 700 ◦C for

he spinel thin film growth. Y0.16Tb0.3Zr0.54O2−∂ powder wasrepared by grinding stoichiometric mixtures of oxides in anlumina mortar and pestle and subsequent firing at 1400 ◦C andepeating to achieve homogeneous, single-phase material. Theargets were placed into the sputtering system and off-axis RF

agnetron sputtered at 125 W and 40 mTorr pure UHP Ar atmo-phere. The substrate temperature was set at 650 ◦C. Powderputtering targets for Y0.16Tb0.3Zr0.54O2−∂ were prepared byacking and uni-axially pressing at 1760 N into 50.8 mm dia.u sputter cups.

.3. Electrode characterization and sensor fabrication

A Siemens D5000 X-ray diffractometer (XRD) using Cu K�adiation was used to determine the film crystal structure, phaseurity and lattice parameters after direct sputter deposition (for

400 E.L. Brosha et al. / Sensors and Act

Fu

tp

AsTt3omosaabse

e5sbsssppp

Ppcm

3

Lotbr

otfofiocpsttatlobs

wtfcpitappcinoNtsmPlvsaaw1srCsaoNSb

ig. 1. Two electrode (left) and three electrode (right) sensor configurationssed in this work [22].

he oxide targets) and post anneal (for the perovskite films pre-ared from the fluoride targets).

After XRD characterization of the metal oxide electrode,5000 A thick Pt counter electrode was sputtered onto the

ubstrate, parallel and co-planar to the metal oxide electrode.he same mask that was used to pattern the metal oxide elec-

rodes was used for the Pt electrodes. At this time, a small,mm × 3 mm Pt pad approximately 5000 A thick was sputteredn a section of the metal oxide electrode to facilitate the attach-ent of a Pt lead wire to make electrical contact with the metal

xide, sensing electrode. Prior to the application of this Pt pad, aputter etch was performed to clean the substrate and film surfacefter which a flash coating (<100 A) of Ti metal was sputtereds an adhesion layer. Three-electrode devices may be fabricatedy the addition of another working electrode – by repeating theteps outlined above – co-planar and parallel to the first twolectrodes.

Finally, the substrate was transferred to an electron beamvaporation chamber to partially cover the electrodes with a–10 �m thick film of 8 mole% Y2O3–ZrO2 electrolyte. Theource material for the electrolyte film was crushed CeraFlexTM

rand YSZ obtained from MarketTech International. The sub-trate was heated to 800 ◦C during the deposition. A smallapphire substrate, partially masked, was included next to theensor to act as a witness. After the deposition, the witness sam-le was used to determine grown rate and film thickness usingrofilometry techniques while XRD was used to ensure phaseurity. A picture of the completed sensor is shown in Fig. 1.

Pt lead wires, 0.04′′ thick, were attached to the Pt pad andt counter electrode using a parallel gap wire bonder. The com-leted sensor was transferred to a tube furnace for testing. Aomputer-controlled Keithley 2400 source meter was used toeasure voltage levels at open circuit.

. Results and discussion

The first device made was a two-electrode sensor using aaCr0.8Mg0.2O3 working electrode and a Pt counter electrode
n polycrystalline alumina substrates from Marketech Interna-ional (see Fig. 1). XRD characterizations after post-anneal andefore counter electrode and electrolyte deposition showed onlyeflections from the substrate and the film. Since the intensity
gt1i

uators B 119 (2006) 398–408

f the principal LaCr0.8Mg0.2O3 reflection was only 6.6% ofhe principal substrate reflection, it was not possible to useull profile fitting methods to refine the film reflections to anrthorhombic unit cell as in previous work [20]. Instead, thelm reflections were fitted to a primitive pseudo-cubic unit cellf approximately 3.88 A. This compares very well to commer-ially prepared (Praxair) bulk La0.8Sr0.2CrO3 refined with arimitive pseudo-cubic unit 3.877 A (0.002) (error in parenthe-es). Although unit cell dimension for the film does not reflecthe incorporation of a smaller dopant cation (Mg) into the lat-ice, X-ray fluorescence spectroscopic analysis verified that Mgnd not Sr was present in the target used to make this elec-rode. Refinement of Mg-doped LaCrO3 samples prepared usingonger sputter depositions (producing a film thickness on therder of 2 �m) showed a smaller pseudo-cubic unit cell thanulk Sr-doped LaCrO3 of 3.8724 A (0.0001) consistent with aubstitution of a smaller dopant cation.

The sensor was mounted on a specially made sample holder,hich was then placed into a tube furnace and heated to 650 ◦C

o begin testing. Only two terminal measurements were per-ormed in this work; a reference electrode with a fixed oxygenhemical potential such as an air reference was not used unlike inrevious work [2]. The sensor configuration used in these exper-ments has an inert substrate and not YSZ and this precludeshe use of a two-compartment set-up. Furthermore, the use of

dense Pt electrode under a YSZ electrolyte film rather thanorous Pt electrode applied to a dense YSZ electrolyte leads to aseudo-reference counter electrode [26]. The source-meter wasonnected to the sensor such that the Pt counter electrode wasnstrument positive and the metal oxide electrode was instrumentegative. In this configuration, a mixed potential generated at thexide electrode that is more negative than the thermodynamic,ernst potential is recorded as a positive voltage. After initial

esting was performed, the sensor was left to soak at 650 ◦C intatic air for a period of time. The initial NOx/HC/CO response,easured a few hours after initial heating to 650 ◦C and 1%O2, is shown in Fig. 2a. The small “NO” response was most

ikely due to residual NO2 trapped in the mixing system as theoltage level returned to baseline after 5 min. Also evident is alow response time to 90% of level (approximately 34 s for NO2)nd longer period of time to return to baseline (4.5 min for NO2nd in excess of 5 min for C3H6). The voltage levels did changeith time; e.g. the voltage level for 100 ppm NO2 decreased by0 mV to −69.5 mV after 3 months of soaking at 650 ◦C. Clearlyome initial device aging was occurring. Fig. 2b is the sensoresponse to NO2 (100 ppm), C3H6 (500 ppm), C3H8 (500 ppm),O (500 ppm) and CH4 (500 ppm) after an approximate 2100 h

oak at 650 ◦C with ambient room air (flowing at 500 sccm)s the base gas. Several desirable sensor characteristics can bebserved in this plot. First is the preferential sensitivity towardsO2; e.g. 0.16 mV/ppm NO2 compared to 0.05 mV/ppm C3H6.econd, there is little sensitivity towards CH4 at 650 ◦C and thisehavior is desirable for a sensor that may be used for natural
as-fueled, stationary power generating applications. And last,here is the quick response time; e.g. for NO2 is approximately0 s to 90% of level. Since the actual oxygen partial pressuren the exhaust gas of a reciprocating engine falls between these

E.L. Brosha et al. / Sensors and Actuators B 119 (2006) 398–408 401

Fig. 2. (a) Initial sensor response of a LaCr0.8Mg0.2O3/YSZ/Pt device to NO2

NO, C3H6, and CO versus time at 650 ◦C. Base flow is 500 SCCM of 1% O2/N2

balance. (b) Sensor response of a LaCr Mg O /YSZ/Pt device to NO C H ,CB

tm

otiCwnsAcgatNo

FCa

str2osNioppsThe device still responded preferentially to NO2; e.g. approx-imately 0.401 mV/ppm NO2 versus 0.071 mV/ppm C3H6. Thesensitivity of the response to NO compares well with propy-lene at 0.076 mV/ppm NO. The response to CH4 is still low

0.8 0.2 3 2 3 6

3H8, CO, and CH4 versus time at 650 ◦C after an extended anneal (2100 h).ase flow is 500 SCCM of ambient air.

wo PO2 extremes, the sensor response with respect to PO2 waseasured next.Fig. 3 shows the sensor response at three different levels of

xygen partial pressure for the same test gases and concentra-ion levels. The magnitude of the voltage response for all gasesncreases with decreasing PO2. However, the voltage level for

3H6 does not follow this trend. NO2 shows the largest changeith PO2. Since the exhaust of a typical, modern, lean burnatural gas engine contains roughly 10% O2 by volume, sub-equent data were acquired in either this PO2 range or in air.

cylinder of 10.4% O2, UHP N2 balance was mixed and theoncentration measured using a Rosemount OxymitterTM oxy-en gas analyzer. Fig. 4 is the sensor response to NO2 at 650 ◦Cnd 10.4% O2 flowing at a base flow of 500 sccm. Fig. 5 illus-
rates that the preferential sensitivity for this device is towardsO2 with negligible sensitivity to CH4. Similar behavior wasbserved for both air and 1% O2/N2 base gases.
Fr

ig. 3. Sensor response of a LaCr0.8Mg0.2O3/YSZ/Pt device to NO2, C3H6,

3H8, CO and CH4 in ambient room air, 10.4% O2, and 1% O2 with N2 balancet 500 SCCM base flow rate at 650 ◦C.

The temperature of the device was lowered to 600 ◦C forubsequent testing. Ideally, any change in the electrodes or elec-rode/electrolyte interface that could lead to a change in deviceesponse over time, should have occurred during the extended100 h anneal at 650 ◦C. Fig. 6 is the sensor response to oxidesf nitrogen at 600 ◦C in air. At this time, NO was added to theeries of test gases. The direction of the voltage response forO is similar to that of CO or a hydrocarbon and is positive

ndicating that it is being electro-oxidized. The voltage levelf the device increased for all test gases at 600 ◦C when com-ared to the 650 ◦C data. This is normal behavior for mixedotential sensors [4,6–8,10]. The response time increased onlylightly for the NO2 to approximately 12 s to 90% of level.

ig. 4. NO2 response at 650 ◦C, 10.4% O2/N2 balance flowing at a base flowate of 500 SCCM [25].

402 E.L. Brosha et al. / Sensors and Actuators B 119 (2006) 398–408

Fg

arc

wfeoohcdaid

F6a

Fig. 7. Plot of sensor voltage measured at 100 ppm of NO and NO2 and 500 ppmca

tb

isfp“to

ig. 5. NO2 and CH4 sensor voltage levels versus concentration at 650 ◦C. Baseas is 10.4% O2 delivered at a flow of 500 SCCM.

t 0.0053 mV/ppm. Generally the sensor response exhibitedeproducibility when test gases were cycled between differentoncentrations levels.

Following this characterization, the temperature of the deviceas maintained at 600 ◦C for 800 h in order to measure any

urther device aging, and response behavior. The device wasxposed to either 10.4% O2 or ambient room air over the coursef the experiment. Fig. 7 is a plot of the voltage level for 100 ppmf NO2 and NO at 10.4% O2 versus time in hours – CO andydrocarbon responses are also plotted in Fig. 7. The rate ofhange of NO2 taken from the slope of a linear fit to these
ata points corresponds to a decrease (increase in negative volt-ge) in sensor voltage of 0.17 mV/day. The change in NO is anncrease of 0.016 mV/day, while the HC’s and CO range from aecrease in sensor voltage between −0.0017 mV/day for C3H8
ig. 6. Sensor response to NO2, NO, and selected HC’s and CO versus time at00 ◦C after device testing at 650 ◦C. Base gas is ambient room air delivered atflow rate of 500 SCCM [25].

BapsH

3

pfi(tTssBattfioATs

oncentration levels of C3H6, C3H8, CO, and CH4 versus a time period of 800 ht 600 ◦C in 10.4% O2.

o −0.043 mV/day for C3H6. The nature of these changes wille investigated in future work.

A total NOx sensor with minimal sensitivity to hydrocarbonss desired for engine control applications. While this first devicehowed excellent selectivity with respect to methane, cross inter-erence from C3 and presumably C2 and C4 hydrocarbons mayresent a problem. In our previous work on non-thin film-based,bulk” mixed potential sensors, we found that the magnitude ofhe mixed potential generated from C3 hydrocarbons on metalxide electrodes made from A-site doped La0.8Mg0.2CrO3 and-site doped LaCr0.8Mg0.2O3 exhibited similar CO and HC volt-ge levels per ppm of hydrocarbon but different voltage levels perpm of NO and NO2 [23]. This suggested that a three-electrodeensor might be fabricated to simultaneously measure NOx andC/CO gases with a single device.

.1. Three electrode device

A three-electrode device was fabricated to measure the mixedotential voltage levels for NOx and HCs/CO between a thinlm metal oxide electrode sputtered from an A-site doped targetLa1−xMgxCrO3) and a thin film metal oxide electrode sput-ered from a B-site doped (LaCr1−xMgxO3) target (see Fig. 1).o fabricate this device, a 2 cm × 1 cm polycrystalline Al2O3ubstrate was masked such that a 3 mm × 10 mm stripe of theubstrate was exposed to the sputter target. A composite film of-site doped lanthanum chromite was grown at 400 ◦C using off-xis magnetron sputtering from the LaF3/MgF2/Cr compositearget used in the 2-electrode sensor fabrication. After comple-ion, the mask was moved 5 mm laterally and parallel to the firstlm. The heater assembly and substrate was moved to a sec-
nd sputter gun that contained the A-site doped sputter target.
second 3 mm × 10 mm film was grown adjacent to the first.he substrate was removed and transferred to a metallizationputter system so that a 7000 A thick 3 mm × 10 mm Pt counter

d Actuators B 119 (2006) 398–408 403

ee

toLnatgta(ttwdbtocb8oaw

s2Iebrlcetamh

5tncessbaerratir

Fefl

(detatwoNOx with respect to hydrocarbons. Furthermore, at a higher tem-perature (550 ◦C) and after annealing and device break-in (90 h),the difference in mixed potential response between the two dif-ferent perovskites becomes almost negligible. Fig. 10a shows the

E.L. Brosha et al. / Sensors an

lectrode could be deposited 2 mm from the edge of the A-sitelectrode.

The substrate was placed into an annealing furnace as beforeo synthesize the perovskite phases. The thickness of the metalxide electrodes was 4 and 2 �m for the A-site and B-site dopedaMgCrO films, respectively. The difference in thickness wasot intentional and was due to slight variations in factors suchs target density and position of the substrate with respect tohe sputter gun, etc. Sputter times/conditions could be altered torow electrodes of the same thickness. X-ray diffraction showedhe presence of the desired metal oxide and substrate phaseslthough it was not possible by measuring lattice parameter shiftusing full profile XRD refinement techniques) to directly verifyhe substitution of the Mg for La (in the A-site case) and Mg forhe Cr (in the B-site case) given the low intensity of the thin filmsith respect to the substrate reflections. However, to identify theifferent sensor working electrodes, “A-site” and “B-site” wille used in this text and will refer to the target material used inhe sputtering process and not necessarily to the crystal structuref the resulting metal oxide film. The electrodes were partiallyovered with an Al2O3 shadow mask and placed into an electroneam evaporation system to deposit the YSZ solid electrolyte. A.75 �m thick film of YSZ was grown such that the Pt and metalxide electrodes were partially covered with the solid electrolytend Pt pads were subsequently applied so that Pt wires could beire-bonded to make electrical contact with the electrodes.The device was placed into the same automated sensor test

tation used for initial characterization and lifetime testing of the-electrode device. Testing for this device was carried out in air.nitial testing temperature started at 500 ◦C to minimize agingffects and to maximize NOx voltage. Therefore, the lengthyreak-in anneal performed on the 2-electrode device was notepeated with this device. The sensor was connected to a Keith-ey 2400 sourcemeter such that the Pt counter electrode wasonnected to the positive terminal and one of the metal oxidelectrodes was connected to the negative terminal. Using thishree-electrode configuration, the mixed potential magnitudend lifetime stability of two different electrodes can be directlyeasured and studied with identical atmosphere and thermal

istories.Fig. 8 is the initial response of the three-electrode device at

00 ◦C in ambient air. The response data for the B-site elec-rode is only slightly smaller than the A-site and it exhibits moreoise; the increase in noise may be due to the larger impedanceaused by the relative geometry with respect to the Pt counterlectrode although this was not determined by AC impedancepectroscopy. Perhaps a more symmetrical arrangement is pos-ible in that the Pt counter electrode could be placed equidistantetween the metal oxide working electrodes and this would bebetter configuration to work with in the future. The differ-

nce in NO and NO2 sensitivity between the two electrodes withespect to the Pt counter electrode is not that large. Fig. 9 is theesponse of the A and B-site electrodes versus the Pt electrode
nd the response of the A-site electrode versus the B-site elec-rode to NOx, CO and HCs. As before, a mixed potential responses obtained for both oxides of nitrogen and hydrocarbons witheducing gases yielding a positive response and oxidizing gases
FPBb

ig. 8. Initial NO and NO2 response for the A-site electrode versus Pt counterlectrode and B-site electrode versus Pt counter electrode at 500 ◦C. Base gasow is air delivered at 500 SCCM.

NO2) producing a negative response. Also plotted is the voltageifference between the two metal oxide electrodes. The differ-nce in the magnitude of the mixed potential response betweenhe two metal oxide electrodes did not produce the anticipatednd desired results; while the hydrocarbon and CO mixed poten-ials were reduced, commensurate with this the NOx sensitivityas also reduced. Inspection of Fig. 9 shows that using two metalxide electrodes actually decreases the preferential sensitivity to

ig. 9. Initial NOx, hydrocarbon and CO response for the A-site electrode versust counter electrode and B-site electrode versus Pt counter electrode at 500 ◦C.ase gas flow is ambient air delivered at 500 SCCM. The voltage differenceetween metal oxide electrodes is also plotted.


Fig. 10. (a) Response characterization to 100 ppm NO, NO2 and 500 ppm con-centrations of C3H8, CO, CH4, C3H6 at 550 ◦C and in ambient room air for A-siteawb

stCeTtto

3pic

b

Fsw

asia

neistopmtLsf

stossoepatil

nd B-site doped metal oxide electrodes versus the Pt counter electrode. Dataere taken after 90 h of operation at 550 ◦C. (b) Difference in sensor responseetween A-site and B-site electrodes at 550 ◦C after 90 h of annealing.

ensor voltage of A-site and B-site electrodes both with respecto Pt counter electrode after the break-in period. The NO andH4 response are almost completely nulled out (Fig. 10b), how-ver there is still cross sensitivity to CO, propane and propylene.herefore, in order to capitalize on the potential advantages of

his multi-electrode concept, either variation in composition ofhe lanthanum chromites must be studied or other suitable metalxide thin film materials must be found.

.2. Comparison of mixed potential voltages fromerovskite, fluorite, and spinel metal oxide electrodes andnvestigations into the use of pre-catalyst to mitigate
ross-interference
The remainder of the work in this study focused on possi-le methods to alter or control selectivity using two different

Ldrd

ig. 11. Plot of sensor voltage (versus Pt counter electrode, baseline voltageubtracted) for two similarly prepared devices comprised of LaCr0.8Mg0.2O3

orking electrode on Al2O3 substrate at 550, 600, and 650 ◦C in flowing air.

pproaches. First we compared the mixed potential response ofeveral possible metal oxide electrode materials suitable for usen mixed potential sensors. Second, we explored the use of a cat-lyst in front of the sensor to selectively oxidize CO and HCs.

The prerequisite criteria for the metal oxide are thermody-amic, chemical, and mechanical stability in the exhaust gasnvironment, sufficient electronic conductivity to control devicempedance, and the ability to generate non-Nernstian/steady-tate electric potentials at elevated temperatures. With regard tohis latter point, the sensitivity and selectivity to NOx over COr hydrocarbons is desirable from a sensor-engineering view-oint. Ideally, complete sensitivity to only NO2 and/or NO withinimal PO2 dependence is preferred. We have studied the rela-

ive NOx/CO/HC response of La0.8Sr0.2CrO3, La0.8Mg0.2CrO3,aMg0.2Cr0.8O3, a fluorite Y0.16Tb0.3Zr0.54O2−∂, and twopinel (ZnFe2O4 and NiCr2O4) electrodes to investigate the dif-erence in gas selectivity.

All sensors were tested on the same automated sensor testtation with identical flow conditions using ambient air. Afteresting was complete, the voltage levels were recorded and thepen circuit voltage recorded with no test gas flowing to theensor was subtracted. Fig. 11 shows the response of two sen-ors with “B-site” doped LaMg0.2Cr0.8O3 working electrodesn polished polycrystalline Al2O3 between 550 and 650 ◦C. Thelectrode and electrolyte films for both these sensors were pre-ared simultaneously. The metal oxide thickness was 0.5 �mnd the YSZ thickness was 5.6 �m. The temperature overlap inesting conditions between these two data sets is 600 ◦C. Theres excellent similarity in response behavior as well as voltageevels produced on both of these devices.

Measurements made on a device using an “A-site”
a0.2Mg0.2CrO3, working electrode show that there was littleifference in the recorded voltage levels between the two mate-ials despite doping the lanthanum chromite perovskite on twoifferent crystallographic sites. For example, Fig. 10a show the

E.L. Brosha et al. / Sensors and Actuators B 119 (2006) 398–408 405

Fig. 12. Plot of sensor voltage (versus Pt counter electrode, baseline voltages5

riatbXo“eTpcMsad

tpoeaTnslrtc6s

NsC

Fs6

1weCiNdtst

A sensor utilizing a ZnFe2O4 working electrode was also pre-pared. Fig. 14 is the response data for the device between 450and 600 ◦C. The ZnFe2O4 electrode was 2.5 �m, the YSZ thick-ness was 8 �m, and the Pt counter electrode was 5000 A thick.

ubtracted) of the La0.8Sr0.2CrO3 working electrode on Al2O3 substrate at 500,50, and 600 ◦C in flowing air.

esponse data at 550 ◦C after annealing both working electrodesn excess of 90 h. There is little difference in the mixed potentialt 90 h and the voltage difference decreased with annealing tohe point that after 900 h, the NO2 mixed potential level differedy only 2 mV. As mentioned earlier, it was not possible through-ray diffraction to measure a unit cell change that would verifyn which site the Cr cation was substituting. The “A-site” andB-site” labels used here merely reflect the stochiometric differ-nces between the two sputter targets used to produce the films.herefore, it is unclear whether the two working electrodes thatroduced the mixed potential response in Fig. 10a are reallyrystallographically distinct. Figs. 10a and 11 also show that theg-doped lanthanum chromite working electrodes produce a

ignificant HC and CO response in air and between the temper-tures of 550 and 650 ◦C. This is the opposite of what one wouldesire for engine control.

A sensor was prepared with a La0.8Sr0.2CrO3 working elec-rode grown using the fluoride precursor sputter method onolished polycrystalline Al2O3. The XRD trace of the metalxide electrode before deposition of the counter electrode andlectrolyte thin films showed reflections from only the substratend from the perovskite in agreement with previous work [20].he metal oxide electrode thickness was 1.2 �m, the YSZ thick-ess was 5.1 �m, and the Pt was 5000 A for this device. Fig. 12hows a large preferential sensitivity to NO2 and NO particu-arly at lower temperatures (500 ◦C). There is a much smalleresponse to HCs and almost no CO and CH4 response comparedo the response from the sensors prepared with an Mg dopedhromite-working electrodes. There was no response to NO at00 ◦C yet a clear response to 100 ppm NO2 was observed. Aensor of this configuration is primarily a NO2-sensing device.

Fig. 13 is the NiCr O spinel electrode sensor response to
2 4Ox and select HCs at 550 and 600 ◦C. There was preferential
ensitivity to C3H6 and C3H8 with negligible sensitivity towardsO, CH4, and to NOx. The NiCr2O4 electrode thickness was

Fs5

ig. 13. Plot of sensor voltage (versus Pt counter electrode, baseline voltageubtracted) of the NiCr2O4 working electrode on sapphire substrate at 550 and00 ◦C in flowing air.

.5 �m, the YSZ thickness was thick at 13–14 �m, and the Ptas once again 5000 A. Contrary to previous reports in the lit-

rature [5] there is very little sensitivity to NOx compared to3H6. A second NiCr2O4-based sensor was prepared in a sim-

lar fashion produced similar results. The XRD traces of bothiCr2O4 electrodes were the same in agreement with literatureata; the XRD traces were acquired off-axis so that only reflec-ions from the spinel film were seen in the trace. Both sensorspinal electrodes had identical thicknesses (1.5 �m), however,he YSZ thickness of the second sensor was only 8.5 �m.

ig. 14. Plot of sensor voltage (versus Pt counter electrode, baseline voltageubtracted) of the ZnFe2O4 working electrode on sapphire substrate at 450,00, 550 and 600 ◦C in flowing air.


Fig. 15. Plot of sensor voltage (versus Pt counter electrode, baseline voltages6

OsbiNctmtbbba

ittCsfib5

Tihrfitmw5mc

Fig. 16. Mixed potential response of A-site metal oxide electrode versus Ptcounter electrode at 550 ◦C for 100 ppm NO, NO2, and 500 ppm of C3H8,CO, CH , and C H . Base gas flow is air at 500 SCCM flowing through apo

tls

rtswhwtwwsHosb

dtrathaHcbs

ubtracted) of the Tb-doped YSZ working electrode on sapphire substrate at00 ◦C in flowing air.

ff-axis XRD characterization of the spinel electrode showed aingle phase film in excellent agreement with literature data forulk Franklinite. The ZnFe2O4 electrode shows strong sensitiv-ty towards non-methane HCs with lower sensitivity to NO2 andO. At 450 ◦C, there is almost no NO2 sensitivity but a signifi-

ant response to NO. As the sensor temperature was increased,he mixed potential due to NO2 exposure increased with a com-

ensurate decrease in NO response. At all temperatures tested,he HC response was quite substantial. The negative responseehavior for CO with respect to hydrocarbons was reproducedetween sensors. The electrode kinetics for CO oxidation maye more favorable on the Pt electrode than the spinel electrodet lower temperatures.

The large NO response at 450 ◦C is of interest because thiss one of the highest NO responses achieved in our investiga-ion. This is an important finding because 450 ◦C is a sufficientemperature to use a precatalyst to completely oxidize HCs andO. These data suggest that it is possible to engineer a NO-only

ensor using a configuration of this type combined with a thinlm precatalyst. Likewise, a sensor of this configuration woulde a NO2-only sensor if it were heated to a higher T such as50 ◦C if the CO/HCs were preferentially oxidized.

The last electrode material studied was terbium-doped YSZ.b-YSZ is a fluorite structure with p-type electronic conductiv-

ty at elevated temperature. It is resistant to sulfur attack and weave demonstrated its use as an oxygen sensor electrode mate-ial for high sulfur applications in previous work [24]. A thinlm was grown under similar sputter conditions used to grow

he spinel films. Fig. 15 is the sensor response at 600 ◦C. Theetal oxide electrode was 1.6–1.8 �m thick, the YSZ thickness
as 8–9 �m thick, and once again the Pt counter electrode was000 A thick. Fig. 15 shows negligible sensitivity to NOx withoderate sensitivity to HCs and CO. These data were once again
onfirmed with a second sensor. There was very little response

ta6p

4 3 6

re-treatment catalyst of a commercial catalyst consisting of 1% Pt supportedn gamma Al2O3 heated to 200, 500, and 600 ◦C.

o any gases below 600 ◦C. Characterization at 700 ◦C showedittle sensitivity gain to NOx however this device would be welluited as a high temperature HC/CO sensor.

The use of pre-catalysts in conjunction with electrode mate-ials selection and optimization may also improve sensor selec-ivity [21,23]. Pt black (ETEK 6 nm particle size) and 1% Ptupported on high surface area gamma-phase aluminum oxideere investigated as hydrocarbon oxidation catalysts. Severalundred milligrams of each catalyst was packed between quartzool plugs in a separate quartz tube furnace. This configura-

ion insured that the carrier gas containing the analyte speciesould pass through the packed catalyst particles. Two furnacesere employed to permit independent temperature control of

ensor and catalyst. Fig. 16 is the sensor response to NOx andCs/CO after the test gas passed through a 1% Pt supportedn gamma-Al2O3 at 200, 500, and 600 ◦C. For comparison, theensor response to these gases is also plotted without the catalysteing present.

At 200 ◦C, the catalyst removes most of the C3H6 and CO butoes not reduce the concentration of C3H8 or CH4. At 500 ◦Che C3H8 is reduced almost half (assuming here a linear sensoresponse) and there is only a small response to CH4. At 600 ◦C,ll of the HCs and CO are oxidized by the catalyst. The effect onhe NOx concentrations can also be seen. When the catalyst iseated to 200 ◦C, the NO is converted, at least partially, to NO2s indicated by the negative voltage generated by the device.owever, 100 ppm of NO is not completely oxidized to NO2 by

omparison to the 100 ppm NO2 signal level when the catalyst isypassed. Likewise, when the catalyst is in place and at 200 ◦C,ome of the NO2 is reduced to NO. As the catalyst is heatedo 500 ◦C, the voltage levels for 100 ppm of NO and NO are
2bout the same and are negative. At a catalyst temperature of00 ◦C, more NO is produced as the sensor voltage is small andositive.

E.L. Brosha et al. / Sensors and Act

Fig. 17. Mixed potential response of A-site metal oxide electrode versus Ptcounter electrode at 550 ◦C for 100 ppm NO, NO2, and 500 ppm of C3H8,Ct[

afc1pCtc

4

fiwbtTgistm(mdToaphstt

ianpt4bmr

A

cPtD

R

[

[[[

[[[

[

[[

[

[

[

[

O, CH4, and C3H6. Base gas flow is air at 500 SCCM flowing through a pre-reatment catalyst of a commercial catalyst consisting of unsupported Pt black25].

Fig. 17 shows effect of gas pre-treatment using Pt black at 200nd 400 ◦C. At 200 ◦C, all the CO and HCs are removed exceptor C3H8, which undergoes only partial combustion. The NO isonverted mostly to NO2 and the sensor response to both of the00 ppm levels is close to the sensor response without catalystre-treatment. At a 400 ◦C catalyst temperature, the remaining3H6 is removed and the NO is a mixture of NO and NO2 given

he increase in sensor voltage (less negative compared to noatalyst pre-treatment).

. Conclusions

Mixed potential sensors based on alkali-earth doped thinlm lanthanum chromite perovskites, doped fluorite, and spinelorking electrodes, Pt, and YSZ electrolyte were fabricated onoth alumina and sapphire substrates using sputtering and elec-ron beam evaporation physical vapor deposition techniques.he sensors were studied primarily in air and at reduced oxy-en level around 10% to simulate lean burn conditions foundn stationary reciprocating engine applications. Mixed potentialensors based on La0.8Sr0.2CrO3 working electrodes exhibitedhe largest mixed potential voltage to NOx preferential to non-ethane hydrocarbons and CO. Sensors with Mg-doped LaCrO3

prepared from both La0.8Mg0.2CrO3 and LaCr0.8Mg0.2O3 RFagnetron sputter targets) exhibited NOx voltage levels interme-

iate to Sr-doped chromite and to mixed potential sensors usingb-doped YSZ and spinel working electrodes. This latter groupf sensors tested showed preferential sensitivity towards COnd hydrocarbons with minimal sensitivity to NOx. The mixedotential voltage level generated for a particular gas species is
eavily dependent on sensor temperature essentially permittingome control over adjusting selectivity. A planar multi-electrode,hin film based mixed potential sensor was fabricated and testedhat suggests the possibility that, with proper selection of work-
[

uators B 119 (2006) 398–408 407

ng electrode combinations, gas selectivity may be controlled orarray of mixed potential sensors can be fabricated to simulta-eously measure NOx, HCs, and CO. The use of a catalyst toreferentially oxidize CO and HCs upstream of the mixed poten-ial sensor was also demonstrated. A Pt black catalyst heated to00 ◦C completely removed methane, CO, and C3 hydrocar-ons. Thus, catalytic pre-treatment combined with a total NOx

ixed potential sensor may be one possible solution for recip-ocating engine control applications.

cknowledgements

The authors acknowledge the DOE office of Energy Effi-iency and Renewable Energy, Distributed Energy Resourcesrogram for funding this work. Los Alamos National Labora-

ory is operated by the University of California for the U.S.epartment of Energy under contract W-7405-ENG-36.

eferences

[1] DOE Advanced Reciprocating Engines Program, http://www.eren.doe.gov/der.

[2] E.L. Brosha, R. Mukundan, D.R. Brown, F.H. Garzon, Sens. Actuators B87 (2002) 47–57.

[3] E.L. Brosha, R. Mukundan, D.R. Brown, F.H. Garzon, Solid State Ionics148 (2002) 61–69.

[4] F.H. Garzon, E.L. Brosha, R. Lujan, R. Mukundan, Solid State Ionics 175(2004) 487–490.

[5] R. Mukundan, E.L. Brosha, D.R. Brown, F.G. Garzon, J. Electrochem. Soc.147 (4) (2000) 1583–1588.

[6] W.J. Flemming, J. Electrochem. Soc. 124 (1977) 21.[7] A. Vogel, G. Baier, V. Schule, Sens. Actuators B 15/16 (1993) 147.[8] D.E. Williams, P. McGeehin, B.C. Tolfield, in: Proceedings of the Second

European Conference on Solid State Chemistry, Veldhoven, The Nether-lands, June 7–9, 1982, p. 275.

[9] H. Okamoto, H. Obayashi, T. Kudo, Solid State Ionics 1 (1980) 317.10] P.T. Moseley, in: P.T. Moseley, B.C. Tofield (Eds.), Solid State Gas Sensors,

Adam Hilger, 1987, p. 144.11] N. Li, T.C. Tan, H.C. Zeng, J. Electrochem. Soc. 140 (1993) 1068.12] Y. Tan, T.C. Tan, J. Electrochem. Soc. 141 (1994) 461.13] Z.Y. Can, H. Narita, J. Mizusaki, H. Tagawa, Solid State Ionics 79 (1995)

344.14] N. Miura, T. Raisen, G. Lu, N. Yamazoe, Sens. Actuators B 47 (1998) 84.15] G. Lu, N. Miura, N. Yamazoe, Sens. Actuators B 35–36 (1996) 130.16] N. Miura, G. Lu, N. Yamazoe, H. Kurosawa, M. Hasei, J. Electrochem.

Soc. 143 (1996) L33.17] T. Hibino, Y. Kuwahara, S. Wang, S. Kakimoto, M. Sano, Electrochem.

Soc. Lett. 1 (4) (1998) 197.18] T. Hibino, S. Kakimoto, M. Sano, J. Electrochem. Soc. 146 (9) (1999) 3361.19] N. Miura, T. Shiraishi, K. Shimanoe, N. Yamazoe, Electrochem. Commun.

2 (2000) 77.20] E.L. Brosha, R. Mukundan, D.R. Brown, Q.X. Jia, R. Lujan, F.H. Garzon,

Solid State Ionics 166 (2004) 425–440.21] E.M. Logothetis, R.E. Soltis, US Patent no. 4,840,913 and P.K. Dulta, N.F.

Szabo, US Patent no. 6,764,591.22] F.H. Garzon, E.L. Brosha, R. Mukundan, Thin Film Mixed Potential Sen-

sors, US patent applied for.23] R. Mukundan, E.L. Brosha, F.H. Garzon, Method for Forming a Potential

Hydrocarbon Sensor with Low Sensitivity to Methanol and CO, U.S. Patent
no. 6,656,336.
24] E.L. Brosha, R. Mukundan, D.R. Brown, F. H. Garzon, Sulfur resistantoxygen sensors for industrial boiler control, in: Proceedings of the 195thMeeting of the Electrochemical Society, Seattle, Washington, May 2–7,1999.

http://www.eren.doe.gov/der

http://www.eren.doe.gov/der

4 d Act

[

[

B

Et(HPAcoxr

Ralhrtcim

Ftootg

08 E.L. Brosha et al. / Sensors an

25] E. Brosha, R. Mukundan, R. Lujan, F. Garzon, Thin Film Mixed PotentialNOx Sensor Development for Stationary Reciprocating Engine Applica-tions, from the proceedings of the 206th Meeting of the ElectrochemicalSociety October 3–8, 2004, Honolulu, Hawaii, reproduced by permissionof The Electrochemical Society Inc.

26] R. Mukundan, E.L. Brosha, F.H. Garzon, J. Electrochem. Soc. 150 (12)(2003) H279.

iographies

ric Brosha is a staff member at Los Alamos National Laboratory at the Elec-ronic and Electrochemical Devices Group (MST-11). He received his B.A.Summa Cum Laude) in physics from Rider College, Lawrenceville, NJ in 1989.e was awarded an Ashton Fellowship from the University of Pennsylvania,hiladelphia, PA, in 1989 and received his Ph.D. in Materials Engineering in
ugust 1993. Currently, his research interests include synthesis of PEM fuel cell
atalysts, electrochemical gas sensors, materials chemistry and electrochemistryf high-temperature solid-oxide fuel cell electrolyte and electrode materials,-ray diffraction, x-ray fluorescence spectroscopy and thermal analysis of mate-ials.

scBTS

uators B 119 (2006) 398–408

angachary Mukundan graduated from the University of Roorkee (India) withbachelor’s degree in Metallurgical Engineering in 1991. He was a research fel-

ow at the University of Pennsylvania, Philadelphia, Pennsylvania, and receivedis Ph.D. in Materials Science and Engineering in February 1997. Currently, hisesearch interests include fuel cell materials, electrochemical gas sensors, highemperature proton-conductors and permeation membranes. Dr. Mukundan isurrently the secretary/treasurer of the “Sensor Division” of “The Electrochem-cal Society”. He has authored over 20 papers in peer-reviewed journals and has

ade several presentations at international conferences.

ernando Garzon is the Technical Project Leader for high temperature elec-ronic and ionic materials in the Electronic and Electrochemical Materials Groupf Los Alamos National Laboratory. His research interests include: the devel-pment of micro-electrochemical sensors, solid oxide fuel cell technology, thehermochemistry of electronically conducting transition metal oxides, thin filmrowth of oxide materials and ceramic membrane technology for sensor, gas
eparation and fuel cell applications. Dr. Garzon holds five patents in electro-hemical technology and has three more pending. Dr. Garzon also serves on theoard of Directors of the Electrochemical Society, is the Chairman of the Highemperature Materials Division of the ECS and is a member of the Internationalociety for Solid State Ionics.

Mixed potential NOx sensors using thin film electrodes and electrolytes for stationary reciprocating engine type applications

Documents