Nanoporous Titania Gas Sensing Films Prepared in a Premixed

30 Sep 2011 - Nanoporous Titania Gas Sensing Films Prepared in a Premixed. Stagnation Flame. Erik Tolmachoff, Saro Memarzadeh, and Hai Wang*...
0 downloads 0 Views 3MB Size
ARTICLE pubs.acs.org/JPCC

Nanoporous Titania Gas Sensing Films Prepared in a Premixed Stagnation Flame Erik Tolmachoff, Saro Memarzadeh, and Hai Wang* Department of Aerospace and Mechanical Engineering, University of Southern California, Los Angeles, California 90089-1453, United States ABSTRACT: We examine the conductometric CO sensing of TiO2 nanoparticle films prepared with a recently developed flame technique. Porous films of crystalline TiO2 nanoparticles were grown directly on interdigitated electrodes by repeatedly translating electrodes over a premixed stagnation flame doped with titanium tetraisopropoxide as the titanium precursor. Flame-deposited electrodes with particle diameter around 9 nm show enhanced sensitivity to CO by up to an order of magnitude compared to sensing films prepared using a commercial TiO2 powder with the particle diameter around 25 nm. A gas-surface model is used to examine chemical kinetic and equilibrium behaviors and explain the sensor responses. The analysis shows that the nature of the gas-surface reactions is similar between these films. The desirable feature of flame-deposited sensing film is attributed to the smaller particle size which provides a greater surface area and a more electrically sensitive conduit.

1. INTRODUCTION Semiconducting metal oxides have become one of the most widely used gas sensing materials since the discovery of the chemiresistive effect about fifty years ago.1 Typical conductometric semiconductor metal oxide (SMO) gas sensors or chemiresistors are made of porous networks of lightly sintered nanoparticles that complete a circuit between metal electrodes.2,3 A bias is applied to the electrode and current percolates through the network in a manner similar to a network of resistors. The changes in the electric resistance of the film arise from changes in the surface composition resulting from gas-surface reaction kinetics and equilibrium.3 In the case of an n-type SMO, the free electron concentration near the semiconductor surface may be enriched by a reducing gas upon reaction with and removal of surface oxygen, thus changing the resistance of the nanoparticle network.24 Conversely, a change in film resistance signals a shift of the gas/surface chemical kinetics and equilibrium and therefore indicates a change in the gas-phase composition. There are three important factors in semiconductor metal oxide (SMO) gas sensing. These are sensitivity, selectivity, and stability.5 Sensitivity is defined as the magnitude of the transducer response to the target gas it is measuring. Selectivity is based on how the sensor responds (or fails to respond) to a particular species when other species are present. Stability is the ability of a sensor to maintain its sensitivity and selectivity over a long period of time. Sensing stability is largely related to the resistance of the particles against sintering when operated at high temperatures. It is generally thought that the sensitivity of SMO gas sensor made of a given material can be enhanced by decreasing the size of its constituent particles.2,610 There are two reasons for this. First, as particles become smaller the surface area to volume r 2011 American Chemical Society

ratio increases providing a greater number of available sites for gas/surface reactions to occur. Second, because the change in available charge carriers is localized, the paths through which current may pass should be small so that the change in resistance is amplified as the particle size is decreased, causing the current pathway to become choked off.68,11 Moreover, as particle size decreases, interparticle connection sizes or “particle necks” become smaller and more numerous. These grain boundaries contribute to a critical portion of the overall sensor resistance,2,10,12 leading to an enhanced sensitivity.1015 Finally, for porous networks with a sub-3-dimensional nature, the electric current is forced to traverse barriers or more tortuous paths caused by the depleted regions. This leads to an increase in transduction.16 Selectivity is the ability of a sensor to transduce a signal from a specific gas species when others are present. Enhancement of selectivity can be achieved through several general routes: specific data acquisition, sensor structure, morphology, doping, or variations in crystalline structure.8,17 Methods of specific data acquisition include combination of AC and DC sources as well as three or four point contact methods, capable of determining where the signal transduction is taking place—the bulk, the surface, or the metal-SMO interface—as well as specific temperature operation and pattern recognition.8,1820 More common, though, is the specific alteration of the SMO with either homogeneous bulk or surface dopant or composite metal oxides which specifically catalyze a certain reaction.2,8,2023 Of particular interest to titania-based sensors is the selectivity of the anatase and rutile polymorphs to target species. Previous work Received: June 27, 2011 Revised: September 29, 2011 Published: September 30, 2011 21620

dx.doi.org/10.1021/jp206061h | J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C has shown that under certain conditions, anatase and rutile titania show opposite responses to reducing gases and composite mixtures of the two polymorphs are able to show selective sensitivity to carbon monoxide (CO) while showing little response to methane (CH4).24 Stability requires that the sensor material remains both mechanically and chemically unchanged over a long period of time—that grain size, defect, and dopant densities remain steady and that the SMO surfaces are not poisoned. The latter issues are not yet completely understood at this time.8,20 Chemical routes to achieve stability include doping the SMO to stabilize its crystalline structure and surface energy.20,25,26 To achieve stability it is critical that the SMO is annealed before its use. The hightemperature annealing typically causes some grain growth and the concentration of defects to become somewhat stable so that the material maintains these properties during its lifetime of operation.20,2729 Other means to achieve long-term stability include low operation temperature and signal drift compensation through pattern recognition.8 Flame synthesis of metal oxide particles is one of the preferred routes for producing nanoparticles in industrial settings because of its high throughput continuous nature with little or no need for post processing or liquid-waste removal compared to wet synthesis methods.30 It has been demonstrated that direct deposition of SnO2 sensing films is possible through a diffusion flame setup in which the gaseous fuel and oxidizer undergoes diffusive mixing during combustion.31,32 Close control over particle size and size distributions of flame generated nanoparticles, however, remains challenging. Particle size distributions depend on individual particle time and temperature histories as well as metal precursor concentrations in the unburned gas—properties which are coupled with fluid dynamics and chemical reactions. As a result, flame processes often produce particles with wide and even bimodal size distributions as aggregates of primary particles with different degrees of sintering begin to form. Control over the flame synthesis of particles can be achieved through uniformity of velocity and temperature profiles such that particles experience similar conditions during nucleation and growth. Porous plug flames, aerodynamically quenched flames, and electrophoresis have all been used as a means to provide uniform particle growth time in a flame.33,34 In recent years, we proposed the method of flame stabilizing on a rotating surface (FSRS) to synthesize metal oxide nanoparticles and deposit them into a thin film in a single step.35,36 As a novel synthesis method, FSRS uses an aerodynamic nozzle to generate a laminar premixed flat flame stagnating on a rotating surface due to aerodynamic flow stretch.37 The setup extends from what we reported earlier,3840 in which the flow stagnation surface was held still and cooled by a flow of water. In both cases, the method produces a stable, quasi one-dimensional, round flame sheet resting several centimeters from the nozzle exit, but a few millimeters from the flow stagnation surface. The unburned gas may be doped with an organometallic precursor, and in the case of titania, titanium tetraisoproproxide (TTIP) was used. The precursor stays intact in the gas flow until it reaches the flame sheet, in which it undergoes rapid decomposition and oxidation.41 The resulting vapor-phase TiO2 undergoes nucleation and coagulates to form particles. Since the flame is typically over 2000 K and the surface is cooled to around 450 K either by convective heat transfer37 due to its rotation or by cooling water, the large temperature gradient produces a strong thermophoretic force, pushing the freshly

ARTICLE

synthesized particles toward the chilled surface, limiting the growth time to no longer than a few milliseconds. Particles in the flame experience similar growth histories due to the quasi one-dimensional nature of the flame.37 By rotating the stagnation surface, the particles are captured into a thin-film and growth is quenched. Titania particles produced in this manner are crystalline phase-pure anatase with their sizes usually narrower than self-preserving size distribution.42 The median particle size can be controlled as desired (in the range of 3 to 25 nm) by adjusting the precursor concentration while the size distributions are only negligibly affected by perturbations to the flame,35,42 resulting in a highly reproducible, scalable process resilient to flame and other perturbations. The phase purity, tunable median size along with narrow size distribution is especially desirable for SMO sensors in terms of their sensitivity and selectivity.24,4345 These qualities also ensure the phase stability of titania under sensor operating conditions, as this stability is size dependent. Particles smaller than ∼15 nm in diameter are known to be thermodynamically stable as anatase while larger particles are stable in the rutile phase.46 In an X-ray and electron diffractometric study reported earlier,40 we demonstrated that the anatase particles prepared with the aforementioned method are resistant to polymorphic changes and sintering at or below ∼900 K in air and helium, suggesting that they are able to maintain long-term stability. More recently, we showed that the crystal phase of the titania particles may be tuned by altering flame stoichiometry.36 Of particular interest is the fact that the rutile phase, although only metastable below 10 nm, can be readily synthesized in fuelrich flames. Another advantage of the FSRS method is that as the stagnation surface rotates and a substrate affixed into the rotating surface repeatedly passes over the flame, particles may be deposited into porous films on the substrate, resulting in a single-step gas-to-film deposition. Film thickness is easily controlled by varying the time of deposition.35 The versatility, reliability and scalability of the FSRS method would be desirable to produce inexpensive sensor films, especially if the film fabrication can be completed in a single step. The purpose of the present work is to establish the baseline of sensing performance for the anatase titania nanoparticle films prepared by the FSRS method. We characterize this performance by focusing on the sensitivity and stability for quantitative detection of CO in air. For comparison, commercial anatase titania powders (Degussa P25) were also used for testing.

2. EXPERIMENTAL SECTION Thin films of titania particles were prepared using the FSRS technique schematically shown in Figure 1 and detailed earlier.35 Briefly, the unburned premixed gas (4.0% C2H4-26.5% O2Ar, equivalence ratio ϕ = 0.45) was doped with 1070 PPM of titanium tetraisoproproxide (TTIP, Aldrich, 97%), delivered by a syringe pump (PHD 22/2000, Harvard Apparatus), and injected and vaporized in the preheated premixed gas flow at 423K. The flat flame was established by issuing a laminar gas jet through an aerodynamic nozzle 1 cm in the exit diameter. The flow speed was 429 cm/s. The stagnation surface is a stainless steel disk 30.5 cm diameter, placed 3.4 cm from the nozzle exit. The disk rotated at 300 rpm, corresponding to a linear velocity of 380 cm/s over the round flame sheet. The flame sheet was established at ∼3 mm above the stagnation surface. The gas jet and the flame were isolated from the surrounding air by a shroud 21621

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

of argon issuing through a tube concentric to the nozzle at a rate of 11 LPM (STP). Particles produced under these conditions have been shown to be nearly phase-pure anatase with a lognormal distribution of size with a mean diameter of 8.8 nm and a geometric standard deviation of 1.31, narrower than the selfpreserving value of 1.45.35 Sensor electrodes were acquired from the Case Western Reserve University Electronics Design Center. The active area is 1  1 cm2, made of a series of interdigitated gold films printed on CoorsTek ADS-996 alumina with an area of 1.5  1.5 cm2. The digits are 250 μm wide, spaced by 250 μm. A total of four sensor films were prepared and tested in the current work (Table 1). Two of the films were prepared by deposition onto the electrode directly using the FSRS method. The electrode was fastened to the stagnation plate with a high temperature tape. The as-grown film has a porosity exceeding 90%,35 and was densified by wetting it with a solution of ethyl cellulose (15 wt %) in ethanol. The second method used the traditional technique of doctor blading. FSRS and P25 powder was suspended individually in ethylene glycol (VWR, 99% pure) (25 mg TiO2 in 1 mL ethylene glycol). The solution was milled with a glass rod to disperse agglomerates. Scotch tape (50 μm thick) was placed around the electrode and the solution was delivered by pipet onto the electrode. The solution was dried and the tape surrounding the film was removed. All films were heated in ambient air to a temperature of 873 K for a period of 4 h to oxidize the organic components of the film and to sinter it to create interparticle necking. Sensor tests were conducted in a homemade stagnation flow reactor enclosed in a glass joint, as shown in Figure 2. This experimental approach, known as the particle-on-substrate stagnation flow reactor, was developed earlier for studies of heterogeneous reaction of deliquesced inorganic particles with trace gasphase component.4749 The reactor design allows for uniform exposure of the sensor surface to an incoming, fresh stream

of test gas of a prescribed composition. In the current study, the reactor was improved to include electric feed through for both sensing measurement and temperature programming of the sensor. The sensor chip rests on a flat fixture, which is heated from below by cartridge heaters (Watlow FIREROD C1J6) aligned parallel to the electrode and cast in ceramic. A 200 μm K-type thermocouple is attached to the heater surface, flush with the electrode to monitor temperature and provide the feedback necessary for surface temperature control. The temperature of the sensor stage is programmable and can reach 873 K. Test gas stream is fed through a circular glass tube 1.0 cm in diameter (ID) and forms a laminar stagnation flow above the sensor chip. The distance between the glass tube exit and the sensor surface was 0.8 cm. Flow rates of the target gas (CO 10 000 PPM in Ar, Scott Specialty Gases) were controlled using a Porter 201 series mass flow controller while the balance gas at 1 LPM (STP) dry air (Gilmore Liquid Air Company, < 7 PPB H2O) was monitored by a mass flow meter (Omega FMA 17001800). The flow setup allows reliable CO metering down to 5 PPM. A bias DC voltage was supplied to the electrodes and the current passing through the sensor was measured using a Keithley 2001 multimeter. Data acquisition, temperature, and mass flow control of the CO mixture were achieved through the LabVIEW interface. For TEM characterization FSRS particles collected from the rotating surface were diluted in ethanol, and sonicated and pipetted onto a holley carbon TEM grid. Images were analyzed to collect size statistics as well as crystalline quality. Images were taken using a 300 kV Tecnai F30 field emission TEM. XRD analysis was performed using a Rigaku Ultima IV X-ray diffractometer with a Cu Kα radiation source (λ = 1.54 Å) on pristine FSRS particles as well as particles annealed at 873 K for 4 h. Sensing film thickness was measured using a stylus profilometer (Ambios XP-2).

Figure 1. Schematic and image of Flame-Stabilized on a Rotating Surface (FSRS).

Figure 2. Schematic of the temperature-programmable particle-onsubstrate stagnation flow reactor for sensor testing. Inset: photo of a sensor on the heater.

Table 1. Summary of Titania Sensor Films Prepared and Tested sensor

a

particle source

median particle diameter,a (nm)

film preparation

film thickness (μm)

FSRS-1

FSRS

8.8 (1.31)

direct flame deposition (10 min)

1

FSRS-2

FSRS

8.8 (1.31)

direct flame deposition (20 min)

2

FSRS-DB P25

FSRS Degussa P25

doctor blading doctor blading

6 6

8.8 (1.31) ∼25

The value in the parentheses is the geometric standard deviation of the particle size distribution. 21622

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

Figure 4. XRD patterns of FSRS produced TiO2 prior to annealing (“as grown”) and after annealing at 873 K for 4 h. “A” denotes anatase feature and “R” denotes rutile feature.

Figure 3. (a) Particle size distributions (symbols: experimental data; lines: log-normal fit to data); (b) TEM image and diffraction pattern of titania particles; and (c) SEM image of an undensified film, all grown by the FSRS technique.

3. RESULTS The size of the titania particles prepared by the FSRS method is distributed lognormally with a mean diameter of 8.8 nm and a geometric standard deviation of σ = 1.31, as shown in the top panel of Figure 3. Both TEM and XRD results confirm the asgrown FSRS particles to be phase-pure anatase (see, the middle panel of Figure 3 and Figure 4). Comparison of the X-ray diffraction patterns before and after annealing at 873 K for 4 h shows that annealing produced a small fraction of rutile as evidenced by the stronger R(110) peak shown in Figure 4. The rutile phase is no larger than 5% by weight after annealing. In comparison, Degussa P25 consists of primary particles with a mean diameter of 25 nm and is a mixture of ∼75% anatase and 25% rutile.50 The thickness of films prepared by doctor blading was measured to be 6 μm after sintering. These films are labeled FSRS-DB and P25 (see, Table 1). The thicknesses of the two sensor films from direct FSRS deposition (samples FSRS-1 and FSRS-2) was around 1 and 2 μm after densification and sintering. Sensor electric resistance was measured to determine the sensitivity to CO exposure. The heater was turned on and air was introduced until the temperature reached a desirable steadystate value. After the signal baseline was determined to be steady, a trace amount of CO was introduced for a period of 1000 s, then CO was turned off and the sensor recovered to its baseline resistance. Here the sensitivity S is defined here as follows: S¼

R0 1 R

ð1Þ

where R0 is the baseline resistance in absence of CO and R is the sensor resistance with CO introduced. Sensor responses were measured at 723, 773, and 823 K. Typical sensor responses to different CO exposure are shown in Figure 5. The test used the FSRS-2 sensor film repeatedly exposed to CO with concentrations ranging from 5 to 280 PPM. As seen, response to CO presence is almost immediate. Resistance of the film drops quickly due to an increase in available charge carriers coming from the reaction of CO with the surface oxygen resulting in an increased free electron concentration.3,12,14,51 Recovery of the sensor baseline is initially rapid with the resistance approaching the baseline values in just a few seconds, although full recovery to baseline took on the order of 10 to 100 s. There exist notable differences in the baseline stability and sensitivity toward CO sensing between films prepared by FSRS direct deposition and by doctor blading. As we can see from Figure 6, the baseline of the DB-FSRS film is about a factor of 1.7 lower than that of the FSRS-1 film. This difference is caused at least in part by the difference in the thickness of the film. The thickness of the DB-FSRS film is 6 μm and that of FSRS-1 is roughly 1 μm. The fact that the baseline resistance differs only by a factor of 1.7, whereas the thickness differ by a factor of 6 suggests that the FSRS film has a smaller porosity, since the particle material of the FSRS film is identical to that of DBFSRS. One of the drawbacks of the doctor blading technique is that the baseline resistance of the resulting film usually undergoes fairly significant drift over time, in comparison to the FSRS direct deposition. Interestingly the sensing sensitivity of the two films is quite similar. For example, the sensitivity of the FSRS-1 film is 1.5 at 280 PPM CO exposure and that of the DB-FSRS film is 1.6 at 250 PPM. We observed that the FSRS film thickness has a notable effect on the sensitivity to CO sensing, as shown in Figure 7. Comparison of the sensitivity measured for FSRS-1 and FSRS-2 shows that the 1 μm FSRS-1 film exhibits a smaller sensitivity to CO than the 2 μm FSRS-2 film, by as much as an order of magnitude. Additionally, while the sensitivity of FSRS-2 film shows very little influence from temperature, the FSRS-1 film exhibits a notably lower sensitivity at 723 K than FSRS-2. The rather unexpected 21623

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

Figure 5. Responses of directly deposited FSRS sensor (sample FSRS-2) to CO exposure at 773 K. Labels 1 through 6 correspond to 280, 140, 93, 46, 18, and 5 PPM CO in air, respectively. The signal overshoot at 18 and 5 PPM CO is due to overshoot of the mass flow controller.

Figure 6. Comparison of responses to CO exposure (in air) at 773 K for sensor films directly deposited by FSRS (top panel) and by doctor blading using the FSRS TiO2 powder (bottom panel).

Figure 7. Variations of sensitivities to CO detection for sensors prepared directly by the FSRS method as a function of temperature. Symbols are experimental data; lines are drawn to guide the eye. The error bars represent one-standard deviation.

sensitivity drop-off is probably indicative of film defects in FSRS-1. The film is too thin to ensure uniform coating. For this reason, the discussion will be focused on results obtained with the FSRS-2 film hereafter. Sensors prepared with the FSRS particle films show significantly greater sensitivity than films prepared with Degussa P25 at all temperatures tested. Figure 8 shows that the sensitivity of FSRS-2 toward CO is larger than that of P25, by as much as a factor of ten under comparable conditions. In addition, the FSRS films show significantly improved sensitivity toward low CO concentration. The sensitivity of the FSRS film is of the order of unity at 5 PPM CO suggesting that it is possible to achieve sub-PPM level CO detection. This is at least an order of magnitude improvement over

P25 at that concentration. Lifetime stability tests were not directly performed here, however, tests on the FSRS sensor show that it remains stable over an extended period of time at 773 K (Figure 9). A somewhat interesting observation is that the recovery time of the FSRS film toward the baseline resistance is substantially longer than that of the P-25 film. Figure 10 shows the resistance of these films relative to that of the baseline as a function of time. The zero time represents the point at which the test gas switches from CO containing to CO free. As seen, the recovery time of the FSRS-2 film is at least an order of magnitude longer than that of the P-25 film. This difference is consistent with the early suggestion that the FSRS film is denser than the P-25 film, leading to smaller pore sizes for gas diffusion. 21624

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

Figure 8. Variations of sensitivities to CO detection. Symbols are experimental data; lines are drawn to guide the eye. The error bars represent one-standard deviation.

Figure 9. Signal stability for the FSRS-2 sensor film for CO detection at 773 K. Symbols are experimental data; lines are drawn to guide the eye. The error bars represent one-standard deviation.

Figure 10. Resistance recovery to the CO-free baseline.

4. DISCUSSION Signal from a conductometric gas sensor is a result of free charge carrier concentration and only indirectly related to surface

ARTICLE

Figure 11. Pseudo Arrhenius plot of conductance of the FSRS-2 sensor film. The apparent activation energy is Ea = 25.2 ( 0.2, 25.9 ( 0.2, 25.9 ( 0.2, and 27.4 ( 0.3 kcal/mol for 0, 5, 46, and 280 PPM CO, respectively.

chemistry. Limiting the understanding of these gas sensors is the incomplete knowledge of the surface chemistry, charge transfer with adsorbed species and the coupling between them. Although rutile (110) is the most studied of all metal oxide surfaces, much less is known about anatase and still less is known about how these surfaces act at high temperatures. The current understanding of the mechanism of low-temperature oxygen adsorption on the rutile (110) surface involves molecular oxygen sticking to an oxygen vacancy, then dissociating with one atom filling the vacancy and the other adsorbed as an atom on a nearby site.52,53 Particle size and phase may also affect surface chemistry. Experimental evidence has shown that size affects desorption rates and that oxygen desorption rates are greater for rutile than anatase.54,55 Since nanoparticle populations are likely to be an amalgam of different crystalline phases and surfaces, a detailed understanding of nanophase titania chemistry may rest in developing mechanisms for different surfaces. Moreover, the chemical kinetic parameters are likely to be non constant and in this study there is certainly some evidence of this —with a period of rapid oxygen adsorption the further the system can be perturbed from its equilibrium to some extent (cf, Figure 5). The coupling of free charge carriers and surface chemistry is not completely understood, however, there is evidence that an increase in charge carriers can catalyze reactions and, in fact, free charge carrier concentration is central to the current understanding of both photocatalyzed and strong metalsupport interaction (SMSI) catalyzed reactions. Additionally, energetically heterogeneous surfaces, and even varying sticking coefficients are thought to influence reaction rates.56 Theory tying the coupling of charge carriers and chemistry together for gas sensors—the electroadsorptive effect—has been discussed for some time15,5759 and modeling the transduction for large grain materials is understood through Schottky Barrier theory.60,61 Unfortunately, these theories breakdown for small particles with size close to the Debye length, such as the ones investigated here in which the distinction between bulk and surface properties becomes blurred.62,63 Despite the various uncertainties just discussed, the experimental results obtained here are rather expected and may be explained by a simple model to be discussed later. Here, let us first examine the changes in the film resistance with respect to 21625

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

where As is the specific surface area, F is the mass density of the particle material, ns is the surface density of oxygen deficiency, ϕ is the porosity, and the factor of 2 accounts for the fact that surface bound oxygen is doubly coordinated with surface Ti sites. The reaction of gas-phase CO and O2 on the titania surface may be modeled by considering the surface to consist of two species only: oxygen atoms and oxygen vacancies. The surface reactions are expressed as the reversible oxygen adsorption/ desorption, forming/depleting the surface oxygen vacancy (s), and the reaction of gas-phase CO with the surface oxygen O, ka

O2ðgÞ þ 2s f 2O kd

2O f 2s þ O2ðgÞ kCO

Figure 12. Pseudo Arrhenius plot of conductance of the P25 sensor film. The apparent activation energy is Ea = 27 ( 2 kcal/mol at 0 PPM CO and 27 ( 5 kcal/mol for both 64 and 640 PPM CO, respectively.

temperature. Figures 11 and 12 present the pseudo-Arrhenius plots of reciprocal film resistance measured for the FSRS and P25 films, respectively. The pseudo activation energy of the FSRS film under a steady-state impinging air flow is Ea = 25 kcal/mol in the absence of CO presence, 26 kcal/mol at 5 PPM CO, and 27 kcal/ mol at 280 PPM CO. These values are all quite close to 27 kcal/ mol measured for the P25 film, even though the uncertainty for the measured Ea value is substantially larger than that of the FSRS film. The similarity in the apparent activation energy values suggests that the surface chemistry is similar with or without CO presence and between particles of drastically different origins and sizes. The apparent activation energy values just discussed may be explained by a simple model based on the Drude theory, in which the conductivity σ is expressed as follows: σ ¼ Nc eμ

ð2Þ

where Nc is the volumetric number density of free electrons, e is the elementary charge, and μ is the electron mobility. Electron mobility is typically quite small for porous films (∼1010 m2V1s1) and is heavily influenced by film morphology.6466 For the present study, we assumed that μ is constant. We look, instead, at the changes in donor-like oxygen vacancies and free carrier concentrations relative to the baseline signal. It is not currently possible to know the density of energy states, which would vary with particles size, disorder and defect concentration, well enough to precisely determine charge carrier concentrations and although some unintentional impurities must exist, their type and concentrations cannot be determined. Therefore, the free charge carrier concentration relies on two assumptions: (1) all vacancies are assumed to be ionized and (2) all free carriers are the result of vacancies since the samples are not intentionally doped. It follows that the concentration of free carriers (Nc) is proportional to the concentration of donor-like defects due to oxygen deficiencies (Nd). The model begins by examining the film surface area in order to determine the ratios of surface area to volume. The volumetric number density of donor like defects due to oxygen vacancies is Nd ¼ 2As Fns ð1  ϕÞ

ð3Þ

COðgÞ þ O f s þ CO2ðgÞ

ð4Þ ð5Þ ð6Þ

where ka, kd, and kCO are the rate coefficients for oxygen adsorption, desorption, and CO reaction with the surface oxygen, respectively. The corresponding rate equation is as follows: dns ¼ 2ka n2s ½O2ðgÞ   2kd ðn  ns Þ2  kCO ðn  ns Þ½COðgÞ  dt ð7Þ where n is the surface number density of oxygen atoms on a fully saturated surface (n = 1.56  1015 cm2).67 For an equilibrium process in absence of CO, we find, θ20 =

θ20 kd 2 ¼ k ½O ð1  θ0 Þ a 2ðgÞ 

ð8Þ

or kd = θ20 ½O2ðgÞ  ka

ð9Þ

where θ0 = ns,0/n is the baseline, steady-state, fractional surface oxygen deficiency without CO presence, which is expected to be ,1 for the condition of interest. Equation 9 suggests that at a given gas-phase oxygen concentration, the surface vacancy is directly related the equilibrium constant of the oxygen adsorption/desorption reaction. Combining eqs 2, 3, and 9, and using the Nc ≈ Nd proportionality discussed earlier, we find, !1=2 1 kd 1 ≈ σ ≈ θ0 ≈ ð10Þ R0 ka ½O2ðgÞ  The above equation points to the fact that the apparent activation energy measured for the reciprocal film resistance (Figures 11 and 12) is related to the reaction enthalpy of the oxygen adsorption/desorption process. To a first approximation, this reaction enthalpy is twice that of the apparent activation energy, ΔHr = 2 Ea. In absence of CO, ΔHr = 50.4 ( 0.4 kcal/mol for the FSRS film and 54 ( 2 kcal/mol for the P25 film at around 773 K. The standard-state value may be calculated by considering the sensible enthalpy of molecular oxygen (H°(773)H°(298) = 4 kcal/mol, which gives ΔHdes°(773) = 54 and 58 kcal/mol for the FSRS and P25 films, respectively. Both values are quite close to the literature value of 59 kcal/mol, which was determined for a mixture of rutile and anatase in a powder form.68 The fact that the desorption energy of the FSRS film is somewhat smaller than that 21626

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

of the P25 film may be explained, in light of the recent observation that nanophase transition metal oxides show large thermodynamically driven shifts in oxidationreduction equilibria.69 Specifically, metaloxygen bond energies are expected to weaken toward small particle sizes, leading to a reduced reaction enthalpy for the adsorption/desorption process. Here we note that the FSRS particle size is about a factor of 3 smaller than the P25 particle size. Under steady-state CO presence, the apparent activation energy remains relatively unchanged, as shown in Figures 11 and 12. In the case of FSRS-2 sensor film, Ea increases from 25.2 ( 0.2 kcal/mol without CO exposure to 27.4 ( 0.3 kcal/mol with CO exposure. This result may be understood by manipulating eq 7 to show the following: θ2  θ20 =

1 kCO 1 1½COðgÞ  2 ka n ½O2ðgÞ 

ð11Þ

where θ is the fractional oxygen deficiency with CO presence. Hence, the right-hand side of the above equation constitutes a correction to the temperature dependence of the film conductivity when oxygen adsorption/desorption is perturbed by CO. The weak sensitivity measured for the activation energy with respect CO concentration suggests that kCO and ka have similar activation energies. Both are substantially smaller than the reaction enthalpy of oxygen desorption and as such, the presence of CO should cause only a small change in the apparent activation energy. The fact we see an increased activation energy with CO presence indicates that reaction (6) has a larger energy barrier than reaction (4), as one would expect. The above discussion suggests that the nature of the gassurface reactions is similar between the FSRS and P25 film. Hence, the drastically improved sensitivity of the FSRS film to CO sensing may be readily explained by the size of the sensor particles; FSRS particles have a mean diameter of 8.8 nm, while P25 particles have a mean diameter of 25 nm. As discussed earlier, since the smaller FSRS particles comprising the film network have a greater surface area to volume ratio, they offer a 2-fold advantage over the P25 sensor: (1) there is more surface area on which reactions may occur and (2) the change in charge carrier concentration per unit volume is more pronounced in the smaller particles.

5. CONCLUSIONS We demonstrated the benefit of FSRS synthesized TiO2 particles for gas sensing applications, compared to Degussa P25 powder. The FSRS-derived sensor showed greater sensitivity to CO sensing at all CO concentrations and sensor temperatures tested. The enhanced sensitivity in principle allows for sensors to operate at low temperatures, a characteristic important to creating long lasting sensors with low power requirements. Unoptimized FSRS sensors with films synthesized directly by the FSRS approach are capable of sensing CO at the PPM level. Tests also show that the sensor is both accurate and precise over a long period of time at elevated temperatures. The desirable feature of FSRS sensing film is attributed to the smaller particle size which provides a greater surface area for CO to react with as well as a more electrically sensitive conduit. ’ AUTHOR INFORMATION Corresponding Author

*Phone: (213) 740-0499; E-mail: [email protected].

’ ACKNOWLEDGMENT The authors wish to thank Professor Andrea Hodge for her help with profilometry. The work was supported by the Combustion Energy Frontier Research Center (CEFRC), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award No. DE-SC0001198. ’ REFERENCES (1) Seiyama, T.; Kato, A.; Fujiishi, K.; Nagatani, M. Anal. Chem. 1962, 34, 1502–1503. (2) Matthias, B.; Ulrike, D. Phys. Chem. Chem. Phys. 2007, 9, 2307–2318. (3) Williams, D. E. Sensor. Actuator. B-Chem 1999, 57, 1–16. (4) Barsan, N.; Koziej, D.; Weimar, U. Sens. Actuators, B-Chem. 2007, 121, 18–35. (5) Barsan, N.; Weimar, U. J. Phys.-Condens. Matter 2003, 15, R813–839. (6) Sberveglieri, G. Sensor. Actuator. B-Chem 1995, 23, 103–109. (7) Azad, A. M.; Akbar, S. A.; Mhaisalkar, S. G.; Birkefeld, L. D.; Goto, K. S. J. Electrochem. Soc. 1992, 139, 3690–3704. (8) Gopel, W.; Schierbaum, A. K. Sens. Actuators B-Chem. 1995, 2627, 1–12. (9) B^arsan, N. Sens. Actuators B-Chem. 1994, 17, 241–246. (10) Ansari, S. G.; Boroojerdian, P.; Sainkar, S. R.; Karekar, R. N.; Aiyer, R. C.; Kulkarni, S. K. Thin Solid Films 1997, 295, 271–276. (11) Yamazoe, N. Sens. Actuators B-Chem. 1991, 5, 7–19. (12) Barsan, N.; Weimar, U. J. Electroceram. 2001, 7, 143–167. (13) Simon, I.; B^arsan, N.; Bauer, M.; Weimar, U. Sens. Actuators B-Chem. 2001, 73, 1–26. (14) Dutta, P. K.; Ginwalla, A.; Hogg, B.; Patton, B. R.; Chwieroth, B.; Liang, Z.; Gouma, P.; Mills, M.; Akbar, S. J. Phys. Chem. B 1999, 103, 4412–4422. (15) Geistlinger, H. Sens. Actuators. B-Chem. 1993, 17, 47–60. (16) Kim, I.; Rothschild, A.; Lee, B. H.; Kim, D. Y.; Jo, S. M.; Tuller, H. L. Nano Lett. 2006, 6, 2009–2013. (17) Toohey, M. J. Sens. Actuators B-Chem. 2005, 105, 232–250. (18) Heilig, A.; Barsan, N.; Weimar, U.; Schweizer-Berberich, M.; Gardner, J. W.; Gopel, W. Sens. Actuators B-Chem. 1997, 43, 45–51. (19) Hoefer, U.; Steiner, K.; Wagner, E. Sens. Actuators B-Chem. 1995, 26, 59–63. (20) Eranna, G.; Joshi, B. C.; Runthala, D. P.; Gupta, R. P. CRC Cr. Rev. Sol. State 2004, 29, 111–188. (21) Kappler, J.; Tomescu, A.; Barsan, N.; Weimar, U. Thin Solid Films 2001, 391, 186–191. (22) Frank, M. L.; Fulkerson, M. D.; Patton, B. R.; Dutta, P. K. Sensor. Actuator. B-Chem 2002, 87, 471–479. (23) Yamada, Y.; Seno, Y.; Masuoka, Y.; Nakamura, T.; Yamashita, K. Sens. Actuators B-Chem. 2000, 66, 164–166. (24) Savage, N.; Chwieroth, B.; Ginwalla, A.; Patton, B. R.; Akbar, S. A.; Dutta, P. K. Sens. Actuators B-Chem. 2001, 17–27. (25) Sheikh, A.; Prabir, D.; Chonghoon, L. Int. J. Appl. Ceram. Technol. 2006, 3, 302–311. (26) Ozaki, Y.; Suzuki, S.; Morimitsu, M.; Matsunaga, M. Sens. Actuators B-Chem. 2000, 62, 220–225. (27) Barsan, N.; Schweizer-Berberich, M.; Gopel, W. Fresenius. J. Anal. Chem. 1999, 365, 287–304. (28) Fleischer, M.; Meixner, H. Sens. Actuators B-Chem. 1997, 43, 1–10. (29) Sharma, R. K.; Chan, P. C. H.; Tang, Z.; Yan, G.; Hsing, I. M.; Sin, J. K. O. Sens. Actuators B-Chem. 2001, 72, 160–166. (30) Pratsinis, S. E.; Vemury, S. Powder Technol. 1996, 88, 267–273. (31) Madler, L.; Roessler, A.; Pratsinis, S. E.; Sahm, T.; Gurlo, A.; Barsan, N.; Weimar, U. Sens. Actuators B-Chem. 2006, 114, 283–295. (32) Liu, Y.; Koep, E.; Liu, M. Chem. Mater. 2005, 17, 3997–4000. (33) Tsantilis, S.; Pratsinis, S. E. Aerosol Sci. Technol. 2004, 35, 405–420. (34) Wegner, K.; Stark, W. J.; Pratsinis, S. E. Mater. Lett. 2002, 55, 318–321. 21627

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628

The Journal of Physical Chemistry C

ARTICLE

(35) Tolmachoff, E. D.; Abid, A. D.; Phares, D. J.; Campbell, C. S.; Wang, H. Proc. Combust. Inst. 2009, 32, 1839–1845. (36) Memarzadeh, S.; Tolmachoff, E.; Phares, D. J.; Wang, H. Proc. Combust. Inst. 2011, 33, 1917–1924. (37) Law, C. K. Combustion Physics; Cambridge University Press: New York, 2006. (38) Zhao, B.; Uchikawa, K.; McCormick, J. C.; Ni, C. Y.; Chen, J. G.; Wang, H. Proc. Combust. Inst. 2005, 30, 2569–2576. (39) Zhao, B.; Uchikawa, K.; Wang, H. Proc. Combust. Inst. 2007, 31, 851–860. (40) McCormick, J. R.; Zhao, B.; Rykov, S.; Wang, H.; Chen, J. G. J. Phys. Chem. B 2004, 108, 17398–17402. (41) Tolmachoff, E. D.; Abid, A. D.; Phares, D. J.; Campbell, C. S.; Wang, H. Proc. Combust. Inst. 2009, 32, 1839–1845. (42) Tolmachoff, E.; Wang, H. Modeling and Sensitivity Analysis of TiO2 Nanoparticle Formation in a Premixed Stagnation Flame. In 6th U.S. National Combustion Meeting; Ann Arbor, MI, 2009. (43) Lazzeri, M.; Vittadini, A.; Selloni, A. Phys. Rev. B 2001, 63, 155409. (44) Zhang, H.; Banfield, J. F. Chem. Mater. 2005, 17, 3421–3425. (45) Ranade, M. R.; Navrotsky, A.; Zhang, H. Z.; Banfield, J. F.; Elder, S. H.; Zaban, A.; Borse, P. H.; Kulkarni, S. K.; Doran, G. S.; Whitfield, H. J. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 6476–6481. (46) Zhang, H.; Banfield, J. F. J. Mater. Chem. 1998, 8, 2073–2076. (47) Liu, Y.; Cain, J. P.; Wang, H.; Laskin, A. J. Phys. Chem. A 2007, 111, 10026–10043. (48) Liu, Y.; Gibson, E. R.; Cain, J. P.; Wang, H.; Grassian, V. H.; Laskin, A. J. Phys. Chem. A 2008, 112, 1561–1571. (49) Liu, Y.; Minofar, B.; Desyaterik, Y.; Dames, E.; Zhu, Z.; Cain, J. P.; Hopkins, R. J.; Gilles, M. K.; Wang, H.; Jungwirth, P.; Laskin, A. Phys. Chem. Chem. Phys. 2011, 13, 11846–11857. (50) Ohno, T.; Sarukawa, K.; Tokieda, K.; Matsumura, M. J. Catal. 2001, 203, 82–86. (51) Akbar, S. A.; Younkman, L. B. J. Electrochem. Soc. 1997, 144, 1750–1753. (52) Bikondoa, O.; Pang, C. L.; Ithnin, R.; Muryn, C. A.; Onishi, H.; Thornton, G. Nat. Mater. 2006, 5, 189–192. (53) Du, Y. G.; Dohnalek, Z.; Lyubinetsky, I. J. Phys. Chem. C 2008, 112, 2649–2653. (54) Thomas, A. G.; Flavell, W. R.; Mallick, A. K.; Kumarasinghe, A. R.; Tsoutsou, D.; Khan, N.; Chatwin, C.; Rayner, S.; Smith, G. C.; Stockbauer, R. L.; Warren, S.; Johal, T. K.; Patel, S.; Holland, D.; Taleb, A.; Wiame, F. Phys. Rev. B 2007, 75, 035105. (55) Huber, B.; Brodyanski, A.; Scheib, M.; Orendorz, A.; Ziegler, C.; Gnaser, H. Thin Solid Films 2005, 472, 114–124. (56) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.; Diebold, U. J. Phys. Chem. B 1999, 103, 5328–5337. (57) Wolkenst., T; Peshev, O. J. Catal. 1965, 4, 301–309. (58) Morrison, S. R. Sens. Actuators 1982, 2, 329–341. (59) Geistlinger, H. Surf. Sci. 1992, 277, 429–441. (60) Fort, A.; Mugnaini, M.; Rocchi, S.; Serrano-Santos, M. B.; Spinicci, R.; Vignoli, V. IEEE Trans. Instrum. Meas. 2006, 55, 2107–2117. (61) Fort, A.; Rocchi, S.; Serrano-Santos, M. B.; Spinicci, R.; Vignoli, V. IEEE Trans. Instrum. Meas. 2006, 55, 2102–2106. (62) Schierbaum, K. D.; Weimar, U.; Gopel, W.; Kowalkowski, R. Sensors Actuators B-Chem. 1991, 3, 205–214. (63) Hagfeldt, A.; Gratzel, M. Chem. Rev. 1995, 95, 49–68. (64) Dittrich, T.; Weidmann, J.; Timoshenko, V. Y.; Petrov, A. A.; Koch, F.; Lisachenko, M. G.; Lebedev, E. Mater. Sci. Eng. B-Solid 2000, 69, 489–493. (65) Duzhko, V.; Timoshenko, V. Y.; Koch, F.; Dittrich, T. Phys. Rev. B 2001, 64, 075204. (66) Konenkamp, R. Phys. Rev. B 2000, 61, 11057–11064. (67) Pan, J. M.; Maschhoff, B. L.; Diebold, U.; Madey, T. E. J. Vac. Sci. Technol. B 1992, 10, 2470–2476. (68) Halpern, B.; Germain, J. E. J. Catal. 1975, 37, 44–56. (69) Navrotsky, A.; Ma, C.; Lilova, K.; Birkner, N. Science 2010, 330, 199–201. 21628

dx.doi.org/10.1021/jp206061h |J. Phys. Chem. C 2011, 115, 21620–21628