Surface Engineering

Proceedings of the 4th International Surface Engineering Congress

August 1–3, 2005 Radisson Riverfront Hotel St. Paul, Minnesota, USA

Edited by

Mark J. Jackson Waqar Ahmed

Sponsored by

Published by ASM International

Materials Park, Ohio 44073-0002 www.asminternational.org

© 2006 ASM International. All Rights Reserved.Proceedings of the 4th International Surface Engineering Congress (#05146Z)

www.asminternational.org

Thermodynamically-driven Reconstitution of Ceramics to Create Nanoscale Features

A.-M. Azad and M. Hammoud Department of Chemical and Environmental Engineering

The University of Toledo, 3052 Nitschke Hall, Toledo, OH 43606-3390, USA. aazad@eng.utoledo.edu, (419) 530-8103.

INTRODUCTION One aspect of current interest and great relevance to the fundamental understanding of the behavior of materials is the role of dimensionality and size on their optical, chemical and mechanical properties for application in a wide range of devices. Owing to the nanoscale features, one-dimensional systems exhibit novel physical and chemical properties that can be exploited in optics, catalysis and data storage devices. Hence, such systems are being synthesized and studied in great details. For instance, polymer nanofibers are used as selective gas separation membranes, filters, biomedical materials (drug carriers and wound dressings), protective clothing, space mirrors, and precursor platforms or scaffolds for the nanotube/nanowire synthesis. Thus, they become model systems to study and correlate the theoretical explanations that are still in progress. Such behavior is almost nonexistent in the bulk material where the particle size is in the micron level. There is growing interest in introducing such attributes in nanoscale inorganic materials as well. The most obvious advantage of doing so is the possibility of their application as quantum dots in a host of devices, such as MEMs, lab-on-a-chip sensors/detectors, structural elements in artificial organs and arteries, reinforced composites, micro solar cell electrodes, micro fuel cells, photocatalysts (splitting of water and in the deactivation of chemical and biological weapons), and electrocatalysts, to name a few. One area where nanofeatures in the materials are of immense relevance is the field of solid-state ceramic-based chemical sensors. High selectivity, enhanced sensitivity and short response time are some of the key features sought in these devices. Since the sensing mechanism and catalytic activity of ceramics are largely microstructure-dominated,

benign surface features such as small grain size, large surface area, high aspect ratio and, open/connected porosity are required to realize a successful sensor material [1]. A novel technique employed to impart such attributes by modifying the microstructural artifacts of ceramic-based sensor materials is described in this paper. The effect of the variation in the ambient oxygen partial pressure across the metal/metal oxide boundary on the microstructure and gas sensing characteristics (viz., enhancement of sensitivity and shortening of response time) of some oxides such as WO3, MoO3 and TiO2 were studied. In this paper, however, the results in the case of one such oxide (viz., WO3) only will be highlighted.

THEORETICAL RATIONALE The methodology adopted in this work stems from the simple fact that at a given temperature and standard pressure (ambient; 1atm.), the oxidation of a metal to its oxide or reduction of an oxide to metal or its suboxide occurs at a well-defined finite partial pressure of oxygen. If the two phases (metal/metal oxide or metal suboxide/metal oxide) are in equilibrium, the incumbent oxygen partial pressure is recognized as the equilibrium partial pressure. According to the Gibbs phase rule, on either side of this unique oxygen pressure, at a given temperature, one of the two coexisting phases must disappear. This is illustrated in the following by considering a hypothetical metal oxidation reaction:

( ) ( ) ( )sMOgOxsM x→+ 22 (1)

For which 2

1

2.

x

OM

MO

pa

aK x= (2a)

Proceedings of the 4 International Surface Engineering Congress, August 1–3, 2005, Radisson Riverfront Hotel, St. Paul, Minnesota, USAth

Copyright © 2006 ASM International 231®

And 1)1( ln. KRTGR −=∆ (2b)

where 1K and )1(RG∆ is the reaction equilibrium constant and the standard Gibbs’ energy change, respectively, for reaction (1). Equations (2a) and (2b) can be combined and simplified to give:

( )

∆=

xRTG

pO RMOM x

)1(2

.2exp (2c)

Equation (2c) gives the thermodynamic equilibrium oxygen partial pressure for the coexistence of a metal with its adjacent oxide at a given temperature in terms of the standard Gibbs energy change for reaction (1); in the present case it is also equal to the standard Gibbs energy of formation of the oxide, MOx. A plot of

2Op or 2

ln Op vs. temperature, gives the contour of the path of the M/MOx coexistence. Using reliable Gibbs energy data [2-3], the loci of

210log Op as a function of temperature in the range 400-800°C, for various metal/metal oxide couples are plotted in Fig. 1.

Fig. 1. Temperature dependence of the equilibrium oxygen partial pressure in: 1. Mo/MoO3; 2. W/WO3; 3. Mo/MoO2; 4. W/WO2; 5. Zn/ZnO; 6. Ti/TiO2 and 7. Ti/TiO. The variation in

2Op by changing the CO2/CO ratio in the range 10-5-105 at 450, 600 and 800°C is also shown (open triangles).

Obviously, in every case, at a 2Op lower than this

‘curve of existence’, metal oxide would be reduced either to corresponding metal or a suboxide. On the

other hand, at 2Op values above the curve, a metal

(or its suboxide) would be completely oxidized to the corresponding stable oxide. Such a thermodynamically feasible redox process can well be carried out in a reducing atmosphere of hydrogen or carbon monoxide followed by oxidation in static or dynamic air. For example, a metal oxide of interest can be reduced by H2 or CO to a suboxide or the corresponding metal. If such a reduced surface is simply heated in air (static or dynamic; .21.0

2atmpO = ), it leads to bulk oxidation,

forming an oxide whose morphological features may or may not be very different from the starting material. Contrary to that, if the reduced surface is exposed to a well-defined

2Op that is only a few orders of magnitudes higher than the theoretical value for the M/MO coexistence, interesting processes ensue. Since the prevailing oxygen potential is only slightly above that established by virtue of thermodynamic equilibrium between M and MOx or between MOx and MOy, this allows the formation and growth of new oxide surface on an atomic/ molecular level, under conditions of ‘oxygen starvation’. Similarly, by exposing the oxide to a precisely controlled

2Op regime that is below theoretical line of the metal oxide stability, one can modulate the extent of reduction of the said oxide either to a suboxide or ultimately to the metal. In any event, it can be envisaged that such a

2Op manipulation will deplete oxygen in a manner so as to cause atomic or submolecular level chemical variations. Hence, upon exposure to an environment that is only slightly rich in oxygen, new material build-up takes place layer-by-layer, thereby creating whole new morphological features that are alien to the starting bulk oxide. Adopting this scheme, we have successfully created novel microstructures in a host of ceramic oxide systems with a view of imparting benign surface features that are paramount in accentuating their functional behavior in gas sensing applications which will be discussed subsequently.

232

METHOD OF ESTABLISHING THE 2Op

DESIRED FOR THE REDOX REACTIONS As can be seen from Fig. 1, the equilibrium oxygen partial pressure in a typical metal oxide is rather low. An oxygen potential in the vicinity of these equilibria could potentially be generated by manipulating the ratio of two gaseous species in a buffer mixture, such as CO2/CO or H2O/H2. At high temperatures, CO and CO2 can exist in equilibrium with traces of oxygen.

2221 COOCO =+ (3)

For which

⋅−=∆

21

2

2ln

OCO

COo

pp

pRTG (4a)

This gives

22

12

2

⋅

= °∆−

RTG

CO

COO

epp

p (4b)

Where ( ) TJG 81.862824000 +−=∆ (4c) Therefore, by controlling the ratio of the concentration of CO2 and CO, it is possible to control the partial pressure of oxygen. Mixing CO2 and CO in the ratio that ranges from 10-5 to 105, provides good buffered systems. In this range, the theoretical

2Op varies between 10-35 and 10-15 atm. at 600°C and between 10-29 and 10-9 at 800°C. These are represented as open triangles in Fig. 1. Similar

2Op can be generated by an equilibrium established in a mixture of H2/H2O. Such oxygen potentials could also be obtained with relative ease via establishment of the following equilibrium:

OHOH 222 21

→+ (5)

For which

−=∆21

22

2ln

OH

OHo

pp

pRTG (6a)

This gives

22

0

2

2

2

1.

=

∆−RT

GH

OHO

epp

p (6b)

Where ( ) TTTJG 25.9ln14.82395000 −+−=∆ (6c) Thus by varying the water vapor to hydrogen ratio in the range of 10-5 to 105, the corresponding equilibrium oxygen partial pressure at 600°C can be conveniently varied between 10-34 and 10-14 atm. However, owing to the ease of mixing gaseous components and to eliminate the possibility of water condensation in the cooler section of the sensor set-up, a buffer mixture of CO/CO2 was employed in this work.

EXPERIMENTAL PROCEDURE The materials investigated in this work included thin foils of Mo and W as well as the powders of MoO3, WO3 and TiO2, all from Alfa-Aesar (99.8% or better). The vendor-specified average particle size of the oxides was between 20-45 µm. The metal foils were used to demonstrate the authenticity of the proposed concept while the oxides were used to fabricate the sensor films whose sensing behavior towards different levels of carbon monoxide in a10% O2-bal. N2 background was monitored before and after these films were subjected to the reduction-oxidation processes described above. A set of these films were also reduced in hydrogen and oxidized in air to differentiate the morphological attributes obtained in the proposed approach versus those resulting from classical redox method. Structural and microstructural examination of the as-received foils, raw powders and the products after each of the redox reactions was also conducted by X-ray diffraction (XRD), scanning and transmission electron microscopy (SEM/TEM) to corroborate the observed enhancement in sensing characteristics of the thick film sensors.

233

The thick film sensors in chemiresistor mode were fabricated as follows. The powders were first ball-milled using 10mm spherical zirconia milling media (Tosoh, NJ) in 2-propanol for 8h, dried and sieved through a 325-mesh stainless steel screen. Each powder was mixed with V-006 (an organic-based resinous vehicle with dispersant, from Heraeus, PA) and α-terpineol (Alfa-Aesar) in an appropriate weight ratio (~70% solid loading) and stirred well so as to form a uniform slurry of adequate rheology. In order to improve the adhesion of the film to the substrate, 2 wt% of tetraethoxysilane (TEOS, Alfa-Aesar) was also added to the slurry and homogenized prior to printing. The film was screen-printed on high density α-alumina substrates (14mm x 14mm) pre-fabricated with interdigitated gold electrodes (12mm x 12mm) and contact pads. The films were first dried in an air oven at 150°C followed by firing in the range of 500-900°C, depending upon the physico-chemical properties of the oxides (stability, volatility, phase transformation, etc.) for 1-2h in air. Gold lead wires (0.25 mm diameter, Alfa-Aesar) were attached to the contact pads via silver paste which was cured in three different stages between room temperature and 350°C so as to form good ohmic contacts. The sensor was placed on a flat platform in an all-quartz experimental set-up which was located in the uniform temperature zone of a compact horizontal Lindberg furnace (MiniMite) and the lead wires were taken out of the furnace through a twin-bore alumina tube. A type-K thermocouple was also placed just above the sensor to monitor the temperature and its variation (if any) during the test. The ends of the gold wires were connected to a high impedance Agilent 34220A digital multimeter, which in turn was connected to a desktop PC via HPIB interface card. Sensor resistance data was acquired and displayed in real-time with the help of IntuiLink software. A gas stream consisting of 10% O2-90% N2 (v/v) mixture was obtained by blending dry compressed air with high purity nitrogen to obtain the background (reference) gas. The sensor was first heated to a selected temperature in the background ambient, allowed to equilibrate at that temperature till a steady baseline resistance (Rb) was established.

Given amount of CO from a CO/N2 tank was then bled in and allowed to blend. Sensitivity of a given film was measured by recording change in film resistance with respect to Rb upon introduction of a given amount of CO in the stream. The sensor behavior was monitored both with increasing and decreasing level of CO in the ambient to confirm the reversibility attribute of the sensor. The response time (t90) was calculated by discerning from the recorded data, the time it took for the signal to attain 90% of the difference between the two steady states, viz., in the background (Rb) and that after CO was introduced (Rg).

RESULTS AND DISCUSSION The above-described hypothesis pertaining to the variation in the morphological artifacts in materials when they are subjected to redox reactions in precisely controlled oxygen potential regimes was first verified in the case of a pure metal. Small (25 mm x 12.5 mm) coupons cut from a tungsten foil were subjected to several redox schemes and were characterized at the end of each treatment by XRD and SEM. Figure 2 compares the XRD signatures of the as-received W foil and the foil subjected to oxidation (

2Op ~10-15) and reduction (2Op ~10-29) by CO/CO2

buffers at 800°C for 24h.

Fig. 2. XRD patterns of pure W foil and that subjected to redox reaction by the CO/CO2 buffer.

234

Two obvious inferences can be immediately drawn: (i) elemental W is regenerated after oxidation and reduction in CO/CO2 atmospheres, and, (ii) the regenerated metal has preferred orientation (higher intensity of the strongest <110> peak in the treated sample). The comparative morphological features of these samples are shown in Fig. 3.

Fig. 3. Electron micrographs of: (a) pristine W foil and (b) the foil oxidized and by a CO/CO2 mixture at 800°C/24h. The fractured morphology of the surface in Fig. 3b clearly indicates that the material has undergone major phase and structural (metal (cubic) → oxide (triclinic)) changes prior to re-conversion to the metal again. This is corroborated by the EDS spectra collected after each event (shown in Fig. 4). As shown in Fig. 5, the XRD signature of a W foil directly oxidized in air at 800°C/2h is identical with that of the foil subjected to an oxidation-reduction

cycle in CO/CO2 stream at 800°C/36h, followed by oxidation in air. Phase analysis of the two patterns shows that in both cases, WO3 is the dominant phase. This is understandable, since despite the intermediate heat-treatments in low

2Op regimes, the phase evolution is ultimately dictated by the final processing parameters and the ambient conditions which is the same in the two cases.

Fig. 4. EDS spectra of W foil subjected to various oxidation and reduction treatments (see the text).

Fig. 5. XRD pattern of (a): W foil after bulk oxidation in air at 800°C/2h, and (b): that oxidized and reduced in CO/CO2 mixtures at 800°C/36h followed by air oxidation at 800°C/2h. Accordingly, as shown in Fig. 6 the microstructural features of the two samples are also similar albeit with some grain growth and compaction clearly

a

b

235

visible in the second case, probably due to several heat-treatments leading to some sintering as well.

Fig. 6. Microstructural features of W foil after (a) heating in air for 2h at 800°C, and (b) oxidized and reduced for 36h in CO/CO2 mixtures followed by air oxidation for 2h at 800°C. In contrast to this, the microstructural features seem to undergo drastic changes, when the W foil is oxidized in CO/CO2 mixture in one case and oxidized, reduced and re-oxidized in manipulative

2Op regimes generated by different CO/CO2 ratios. This is shown in Fig. 7. The microstructure shown in Fig. 7a results when pure W foil (Fig. 3a) is heated at 800°C/24h in CO/CO2 mixture in a

2Op range that is above the W/WO3 line in Fig. 1, while the morphology shown in Fig. 7b evolves when the foil in Fig. 3b is subjected to an identical treatment.

A comparison of these features with those shown in Fig. 6, lends credibility to the proposed notion that depending upon the location of the equilibrium oxygen potential across the M/MO line of coexistence, changes on microscopic levels are caused in the bulk oxides.

Fig. 7. Scanning electron micrographs of W foil after: (a) single-stage oxidation and, (b) cyclic oxidation-reduction-oxidation by CO/CO2 buffer mixtures at 800°C/24h. This perhaps leads to atomic/submolecular level non-stoichiometry – that is nanoscale in nature and hence undetected by bulk techniques such as XRD. It is these defect sites that act as the nucleation and growth centers for the new oxide phase whose growth, due to the limited access to oxygen (defined by a

2Op that is designed to be only slightly above the theoretical line of M/MO coexistence), is slow. Evidently, the reduced phase is subject to a concentration strain in terms of

2Op . The need for

a

b

a

b

236

24h long dwell at 800°C for oxidation in the CO/CO2 buffer compared to 2h for air oxidation supports the diffusion-like controlled buildup of the new oxide phase almost on atomic scale - one monolayer after another. These observations are corroborated by the XRD patterns of the samples obtained under conditions leading to the microstructures in Fig. 7a and 7b. These are shown in Fig. 8a and 8b, respectively. Detailed analysis of the X-ray peaks reveals that such treatment yields a mixture of WO2/WO3, rather than either of the two oxides being the predominant phase. Thus, the role of controlled oxygen potential in the ambient and its effect on microstructural variation of a given phase is clearly brought out.

Fig. 8. XRD of W foil after: (a) single-stage oxidation and, (b) cyclic oxidation-reduction-oxidation by CO/CO2 buffer mixtures at 800°C/24h. Indeed, in many cases where bulk oxides were subjected to redox treatment by CO/CO2 mixtures for shorter duration, it was rather difficult to discern quantitative phase changes based simply on the XRD signatures. For example, when a WO3 thick film is subjected to the reducing and oxidizing sequence in appropriate oxygen potential regimes created by CO/CO2 buffers in one case, and another film after the above-mentioned treatment is finally heated in air, no difference in their XRD signatures could be discerned, as shown in Fig. 9. All the three patterns could be indexed as those belonging to triclinic WO3 [4].

On the contrary, the morphological features are significantly affected – both in terms of shape and size of the particles as well as in terms of the uniform porosity in the sample, as seen from the SEM pictures in Fig. 10.

Fig. 9. Comparative XRD patterns of: (a) WO3 film calcined in air at 800°C/2h, (b) a, subjected to redox in CO/CO2 mixtures at 800°C/12h and (c) b, heated in air at 800°C/2h.

Fig. 10. Scanning electron micrographs of WO3 thick films: (a) as-prepared, (b) a reduced in CO/CO2, (c) b oxidized in CO/CO2 and (d) b oxidized in air at 800°C for various periods of time. This vividly illustrates that the ambient oxygen potential in the vicinity of an oxide phase has profound effect on its morphological features,

a b

c d

237

which could be tailored to accentuate the sensing behavior of a potential semiconducting oxide. The response characteristics (such as resistance, sensitivity and response time) of WO3 thick film sensors to CO gas (14-100 ppm) at 450°C are shown in Fig. 11.

Fig. 11. Response of a WO3 thick film sensor to CO at 450°C. a: as-prepared, b: a reduced and oxidized in CO/CO2 mixtures, and c: a reduced in CO/CO2 mixture and oxidized in air. As can be seen from Fig. 11, surface modification by the proposed scheme has certainly brought about

marked changes in the behavior of a WO3-based CO sensor. The sensor could be operated in a cyclic fashion without compromising the signal, with a concomitant enhancement in sensitivity and significant shortening of the response time. In order to examine if such exotic microstructural features could be developed in bulk oxides, WO3 thick films formed by heating the slurry at 600°C/1h in air were subjected to reduction in a H2/N2 mixture at 600°C for ½ h, cooled to room temperature and again heated in air up to 500°C for ½ h. The evolved microstructures are shown in Fig. 12.

Fig. 12. Evidence of microstructural modification in bulk oxides (a) via H2 reduction-air oxidation (b); b at higher magnification is shown in c.

a

c

b

238

In such cases also, the regeneration of the parent oxide with different morphological features can be explained in a way similar to that summarized above. The only notable difference is that in the later case, the oxygen potential in the ambient is rather high (0.21 atm.) – a parameter which could kinetically favor the process, leading to the faster attainment of thermodynamically most stable (M/O) stoichiometry upon regeneration. Similar morphological changes have been observed in the case of sensors made with composite mixtures, viz., ZnMoO4-MoO3 (MZM). It was found that the rod-like MoO3 grains in the original mixture are regenerated as highly oriented thin platelets upon

exposing the MZM film to a gas mixture containing 1%CO at 450°C for 1h followed by natural cooling in air. It was also observed that as a result of this bulk redox reaction, the morphological features of the major phase (ZnMoO4) have undergone noticeable variation (from regular near spherical grains to triclinic habits with well-defined sharp edges) without any chemical degradation [5, 6]. These results are shown in the SEM pictures (Fig. 13) and, the EDS spectra (Fig. 14) collected in different pockets of the composite after subjecting it to the above-mentioned redox treatment.

Fig. 13. Microstructural features of the MZM composite sensor film: a. as-prepared phase pure ZnMoO4 (ZM), b. as-prepared MZM composite; morphological changes in ZnMoO4 and MoO3 phases after the composite was exposed to 1%CO at 450°C/1h following by natural cooling in air, are shown in c and d, respectively.

a b

c d

239

CONCLUSIONS High selectivity, enhanced sensitivity, short response time and long shelf-life are some of the key features sought in the solid-state ceramic-based chemical sensors. Since the sensing mechanism and catalytic activity of ceramics are predominantly surface-dominated, benign surface features in terms of small grain size, large surface area and, open and connected porosity, are sought to realize a successful material. We have shown that these features could be incorporated in a given semiconducting oxide, such as WO3, via a novel gas phase redox scheme, thereby resulting in better sensor performance. A plausible mechanism of such microstructural development under precisely controlled oxygen potential across the M/MO proximity line is suggested. In the light of structural and microstructural data shown here, the formation and growth of new oxide surface on an atomic/ submolecular level, under conditions of oxygen starvation appears to be the most likely pathway.

ACKNOLWEDGMENTS The authors wish to thank the College of Engineering and the Chemical Engineering Department at the University of Toledo for the new faculty start-up funding and a graduate research assistantship to AMA and MH, respectively.

REFERENCES

[1] Yoo, S., Akbar, S.A. and Sandhage, K.H., “Nanocarving of titania (TiO2): a novel approach for fabricating chemical sensing platform,” Ceramic International, Vol. 30 (2004), pp. 1121-1126. [2] Kubaschewski, O. and Alcock, C.B., Metallurgical Thermochemistry, 5th Edition, Pergamon (New York, 1979), pp. 378-384. [3] Pankratz, L.B., Stuve, J.M. and Gockcen, N.A., Thermodynamic Data for Mineral Technology, US Department of Interior Bureau of Mines, Bulletin # 677 (Washington, D.C., 1984) pp. 261-295. [4] ICDD reference card # 32-1395. [5] Azad, A.-M., unpublished research (2004). [6] Abrahams, S.C., “Crystal structure of zinc molybdate ZnMoO4,” J. Chem. Phys., Vol. 46 (1966) 2052-2063.

Fig. 14. EDS spectrum of ZnMoO4 grains before and after the redox reaction in 1%CO at 450°C/1h followed by natural cooling in air (top), and that of the regenerated platelets of MoO3 shown in Fig. 13d (bottom).

240