Вы находитесь на странице: 1из 10

Thin Solid Films 446 (2004) 238247

Surface phase transitions upon reduction of epitaxial WO3(1 0 0) thin films


M. Lia,*, E.I. Altmana, A. Posadasb, C.H. Ahnb
a

Department of Chemical Engineering, Yale University, New Haven, CT 06520, USA b Department of Applied Physics, Yale University, New Haven, CT 06520, USA

Received 3 June 2003; received in revised form 10 September 2003; accepted 1 October 2003

Abstract The surface structure and morphology of WO3 (1 0 0) thin films were studied using scanning tunneling microscopy (STM) and low-energy electron diffraction. The films experienced a net-reducing environment when annealed in oxygen at 800 K leading to surface phase transitions from p(2=2) to p(4=2), and from p(4=2) to a mix p(4=2) and p(3=2). Increasing the annealing temperature to 830 K in ultra-high-vacuum (UHV) led to a fully p(3=2) reconstructed surface. Continued UHV annealing above 800 K caused (1=1) islands to appear on the p(3=2) surface and the film color to darken. Eventually, prolonged UHV annealing led to a (1=1)-terminated surface with straight steps oriented in w0 0 1x or w0 1 0x directions due to crystallographic shear planes. The randomly spaced steps on the (1=1) surface indicated variations in the local stoichiometry in the film. An added row model proposed for the p(4=2) structure is also shown to be consistent with the p(3=2) structure. The formation of the p(4=2) structure from the p(2=2) structure was attributed to W5q migration into the bulk to form the troughs between the added rows. Reduction of the p(4=2) structure caused the troughs to narrow rather than deepen, suggesting that W5q or added row surface diffusion competes with migration of reduced W ions into the bulk when the p(3=2) structure forms. The STM images also show evidence that the (1=1) structure forms through coalescence of the added rows. 2003 Elsevier B.V. All rights reserved.
Keywords: Tungsten oxide; Epitaxial oxide films; Surface reconstructions; Surface phase transitions; Scanning tunneling microscopy

1. Introduction Applications of WO3 thin films in gas sensing w1,2x, catalysis w3x and electrochromic devices such as displays w4,5x and smart windows w6,7x have been studied extensively. The surface morphology and structure of WO3 films, especially epitaxial films, however, have been paid less attention, even though fundamental understanding of gas sensing and catalytic mechanisms is closely related to surface structural changes on oxidation and reduction. This is partially due to the restricted availability of WO3 single crystals w8x and the reliability of sample preparation for surface characterization by techniques such as scanning tunneling microscopy (STM). On the other hand, the bulk structure of WO3 has been well characterized w914x. The crystal structure of stoi*Corresponding author. Tel.: q1-203-4324332; fax: q1-2034324387. E-mail address: min.li@yale.edu (M. Li).

chiometric WO3 is composed of corner-sharing WO6 octahedra arranged in an ReO3 framework; distortions and tilting of the octahedra lead to deviations from the ideal cubic structure. At least five phase transitions are observed between 100 and 1000 K. A monoclinic g5 phase belonging to the space group P21 y n (C2h ) with lattice constants as0.7297 nm, bs0.7539 nm, cs 0.7688 nm and bs90.918 is favored between 290 and 603 K w12x. At higher temperatures an orthorhombic phase belonging to space group Pmnb with as0.7341 nm, bs0.7570 nm and cs0.7754 nm is seen w10x, which then converts to a tetragonal phase belonging to space group P4 y nmm at 1010 K w11x. Each monoclinic and orthorhombic unit cell contains eight WO6 octahedra. As a wide bandgap metal oxide (2.6 eV) w15x, WO3 becomes an n-type semiconductor after bulk reduction by sputtering or vacuum annealing and thus is amenable to surface characterization by electron and ion spectro-

0040-6090/04/$ - see front matter 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2003.10.002

M. Li et al. / Thin Solid Films 446 (2004) 238247

239

scopies as well as by STM. For w0 0 1x-oriented WO3 single crystals several surface phases were reported depending on the surface treatment. A c(2=2) structure w17,18x with half of the on-top oxygen atoms removed was considered to be fully oxidized as this surface is autocompensated with all W ions in the 6q oxidation state. Reduction led to p(2=2) w16,17x, c(6=2) and c(4=2) w18x structures with only 1 y 4 of the on-top oxygen atoms left and half of the exposed W ions reduced to 5q. Further reduction resulted in a fully reduced (1=1) surface with all on-top oxygen atoms removed and all surface W ions reduced to 5q w18x. In addition, a longer range superstructure attributed to arrays of W6O18 clusters was observed as a precursor to the c(2=2) structure w18x. All of these structures were defined with respect to an idealized ReO3 cubic unit cell with a lattice constant of ls0.375 nm. In keeping with this precedent, we will use the same terminology. Meanwhile, bulk phase transitions and related surface morphology changes have been reported for WO3 films. When annealed in air at elevated temperatures, asdeposited films on sapphire that contained small monoclinic crystallites exhibited sintering to form larger grains w19,20x. As the grain size increased hydrated WO31 y 3H2O, hexagonal WO3, reduced WO3 with wellordered crystallographic shear (CS) planes and eventually monoclinic WO3 were seen. The microdomain orientation of the monoclinic grains was resolved on the film surface with atomic force microscopy w19x. The granular structures of polycrystalline tetragonal and metastable hexagonal phases, as well as the step-terrace morphology of the monoclinic phase, have also been studied on WO3 films on sapphire substrates as a function of growth temperature and substrate preparation conditions w21x. Our current interest is focused on the atomic scale surface structure of epitaxial monoclinic WO3 films. Recently, we showed that w1 0 0xoriented g-WO3 epitaxial thin films can be grown on LaAlO3(1 0 0) w22x. A p(4=2) surface reconstruction was induced on the film surface by reducing a p(2=2) surface. Despite the strong similarity between the WO3(1 0 0) and (0 0 1) surfaces, in cubic ReO3 the surfaces are identical, no similar reconstruction was seen on (0 0 1)-oriented single crystal surfaces. The p(4=2) reconstruction was composed of strands running along the w0 1 0x and w0 0 1x directions; defects in the structure included strands spaced 3= and 5= apart. The p(4=2) phase was metastable and eventually converted to a fully reduced (1=1) surface after prolonged reduction. The close relationship between surface structure and bulk oxygen deficiency motivated our further investigation of the surface of the WO3 films. In this paper, we present a detailed study of the surface structural transitions that g-WO3(1 0 0) thin films grown on LaAlO3(1 0 0) exhibit as the films are progressively reduced. The film experiences bulk as well

as surface reduction as indicated by the darkening of the film. Meanwhile, the surface structure changes from a partially oxygen-terminated p(2=2) surface to stranded p(4=2) and p(3=2) reconstructed surfaces without oxygen on top of the strands and eventually to a fully reduced (1=1) surface with steps due to CS planes intersecting the surface. These structural changes are expected to play a strong role in catalysis and gas sensing where the type of surface oxygen species and the coordination of the exposed metal cations are generally considered important parameters in determining heats of adsorption and reaction pathways. 2. Experimental details The WO3 films were grown by sputtering a 99.95% pure tungsten target in a process gas of 50% O2 y 50% Ar onto (1 0 0)-oriented LaAlO3 substrates. The total gas pressure and substrate temperature were maintained at 3=10y3 Torr and 873893 K (as measured by an optical pyrometer), respectively, during the growth. The deposition rate was ;3 nm y min, and 40100 nm thick films were used for the study. After deposition, the films were cooled in an atmosphere of 3=10y2 Torr oxygen. X-ray diffraction indicated that the WO3w1 0 0x direction was perpendicular to the film surface w22x. The STM, low-energy electron diffraction (LEED) and Auger electron spectroscopy (AES) measurements were performed using an ultra-high-vacuum (UHV) system with a base pressure of 1=10y10 Torr w23x. The temperature was measured by clamping a K-type thermocouple housed in a thin Ta tube against the sample surface w23x. Samples were cleaned either in a side chamber by heating to 820 K in oxygen pressures up to 10y4 Torr or annealing in the main chamber with NO2 (;1=10y6 Torr) between 770 and 820 K. In the main chamber gases were dosed using a directed doser that provides an enhancement factor of roughly 45 over the background system pressure; the exposures have been corrected using this factor. Samples were considered clean when atmospheric contaminants such as C and S were below the AES detection limit. Traces of potassium impurities were always detected with AES after the films were cleaned. Tungsten tips for STM were cleaned by electron beam bombardment prior to use. The tunneling current was 0.5 nA in all of the STM measurements. Only unoccupied state images corresponding to positive sample biases are presented in this paper. Stable filled state imaging was not possible because oxygen-deficient WO3 is an n-type semiconductor with donor states in the gap that pin the Fermi level to the bottom of the conduction band w9x. The large WO3 bandgap then pushes the valence band too far below the Fermi level to be accessed by STM.

240

M. Li et al. / Thin Solid Films 446 (2004) 238247

Fig. 1. LEED patterns of a WO3(1 0 0) film obtained after annealing above 800 K in either O2, NO2 or in UHV. (a) The p(4=2) pattern after O2 (3=10y5 Torr) annealing at 800 K for 7 h; (b) the p(3=2) pattern after UHV flashing at 970 K; (c) the (1=1) pattern after UHV annealing for 5 h at 850 K and additional 17 h at 830 K. Weak streaks can be resolved between the primary 1= spots and (d) the (1=1) after further UHV annealing for 15 h at 810 K and 45 h between 820 and 850 K. The streaks between 1= spots become strong. The weak spots in the dashed circles in (a) and (b) represent weak spots at {3y4, 1y2} and {2y3, 1y2}, respectively.

3. Results The WO3 films were oxidized in a furnace at 840 K in flowing oxygen at 1 atm before introduction into the vacuum chamber. This treatment produced a pale yellow film. The sample was then annealed in oxygen (3=10y5 Torr) at 800 K for 7 h, which induced a p(4=2) surface reconstruction as shown in the LEED pattern in Fig. 1a. No change in color was observed at this point. The spots at {3 y 4, 1 y 2} were weak as shown in the dashed circle. Spot splitting was previously observed on (0 0 1)-oriented WO3 single crystals due to monoclinic domain twinning along the N1 1 0M direction w17,18x. For (1 0 0)-oriented films, twinning by a 908 rotation in the surface plane is still possible, the monoclinic angle with respect to the surface normal, however, is fixed. The spot splitting caused by the difference between the b and c lattice constants is too small to be resolved by LEED. Both p(4=2) and p(2=4) domains rotated 908 from each other were resolved in Fig. 1a. Corresponding STM images of this surface are provided in Fig. 2a and b. The image in Fig. 2a shows p(2=2) areas on the surface which can explain the relatively

strong intensity of the 1 y 2 order spots in Fig. 1a; the p(2=2) unit cell is highlighted close to the strands in Fig. 2a. The distance between the strands was measured to be 1.48 nm, approximately twice that of the p(2=2) structure or 4=, consistent with a p(4=2) registry. Meanwhile, a 2= spacing was observed on top of the strands as highlighted by arrows. Most areas of this surface were completely covered by the strands as shown in Fig. 2b. Further annealing in 2=10y5 Torr of O2 at 820 K for 2 h produced domains with two different strand spacings as shown in Fig. 2c. While a 4= spacing was observed on the right side of the image, pointed to by the arrows, a spacing of 1.09 nm or approximately three times the distance between neighboring WO6 octahedra was measured on the left side of the arrowlabeled strand. This surface still yielded a p(4=2) pattern similar to Fig. 1a but with weaker intensity at {3 y 4, 1 y 2}, presumably because of the areas covered by 3= domains. After UHV flashing at 970 K, the surface exhibited a p(3=2) LEED pattern as shown in Fig. 1b. Two p(3=2) and p(2=3) domains rotated 908 from each other can be seen in the figure. The relatively weak spots at

M. Li et al. / Thin Solid Films 446 (2004) 238247

241

Fig. 2. STM images of WO3 (1 0 0) thin films annealed above 800 K (ad). (a and b) 7-h O2 (3=10y5 Torr) annealing at 800 K; (c) additional 2 h of annealing in O2 (2=10y5 Torr) at 820 K; (d and e) UHV flashing at 970 K followed by 75-min annealing in O2 (3=10y6 Torr) at 820 K. The imaging biases were 2.6 V for (a); 1.85 V for (b) and 2 V for (ce).

{2 y 3, 1 y 2} can be resolved as highlighted by the circles in Fig. 1b. The film changed from light yellow to light blue after this treatment. An additional 75 min of annealing in O2 (3=10y6 Torr) at 820 K yielded a strong p(3=2) pattern similar to Fig. 1b. At this point, the surface was completely covered by strands separated by 3= as shown in the STM image in Fig. 2d. The observed LEED pattern suggests a 2= periodicity along the strands. The image also shows large flat terraces with straight step edges oriented along the w0 1 0x and w0 0 1x directions. The very few scattered white clusters were attributed to surface impurities. In addition, Fig. 2e shows p(3=2) and p(2=3) domains on the same

terrace. This indicates that the minor structural differences between the w0 1 0x and w0 0 1x directions do not play a role in the reconstruction. The structural similarity between 3= and 4= strands can be seen in Fig. 3a. The small inset shows a p(4=2) domain lined up with an anti-phase domain boundary in a p(3=2) terrace where the strands are 4= apart as highlighted by the white line. The strands appear identical in the two images. These images were recorded at lower biases (1.25 V for the inset and 1.5 V for the larger image). Under these conditions, we previously showed that a 1= periodicity can be resolved along the strands of the p(4=2) structure w22x; the figure shows

242

M. Li et al. / Thin Solid Films 446 (2004) 238247

Fig. 3. (a) A small-scale STM image of p(3=2) strands showing a 1= periodicity along the strands at the bias of 1.5 V; the inset is a p(4=2) surface where the 1= periodicity was also resolved on the strands at 1.25 V; the white line shows a 4= spacing on the p(3=2) surface. (bd) The added row model of the strands viewed along the w1 0 0x, w0 0 1x and w0 1 0x directions which allows both 3= and 4= spacings between strands.

that the same is true for the p(3=2) structure. In contrast at higher biases, such as in Fig. 2a, a 2= periodicity can be seen along the strands. The previous study of the p(4=2) structure also revealed that the height difference between p(4=2) strands and p(2=2) terraces is close to that between (1=1) and p(2=2) terraces, and that the stoichiometry of the reconstruction is intermediate between the p(2=2) and (1=1) structures w22x. On the basis of this information, an added row model for the p(4=2) structure was proposed w22x. The similarities between the p(3=2) and p(4=2) strands suggest that the same basic added row structure can also explain the p(3=2) reconstruction as shown in Fig. 3bd. In the schematics, W and O atoms in gray octahedra are located at their bulk positions with all on-top O atoms removed, as viewed along the w0 0 1x direction in Fig. 3c. Oxygen atoms are removed from the edges of every other octahedron to create a local (2=2) unit cell along the strand as viewed along the w1 0 0x direction in Fig. 3b. This also gives the strands the same stoichiometry as (1=1) terraces and follows the general requirement that half the oxygen be removed from {1 0 0} step edges on ReO3{1 0 0} surfaces to maintain the stoichiometry of the adjacent terrace. The strand height in the model equals that of (1=1) islands above p(2=2) terraces, and the distance between the two outermost O atoms equals the p(2=2) unit cell dimension. This model

allows both 3= and 4= strand spacings as illustrated in Fig. 3d along w0 1 0x. The existence of trough oxygen atoms in Fig. 3d gives the structures stoichiometries between the p(2=2) and (1=1) surfaces, which is consistent with the observation that continued reduction eventually converts the stranded surfaces into (1=1) surfaces as described below. After an additional 5 h of UHV annealing at 850 K, the surface in Fig. 2d changed from light blue to blue. The STM images of such a film showed that large (1=1) islands (;9 nm wide) had grown across the p(3=2) surface as pictured in Fig. 4a. The (1=1) island step edges appeared rough with many kinks; the predominant mean step orientation in Fig. 4a was 2x. The appearance of (1=1) islands led to a more w0 1 diffuse p(3=2) LEED pattern, similar to Fig. 1b but with very weak {2 y 3, 1 y 2} spots. The film was then further reduced by UHV annealing at 830 K for 17 h, after which the film turned a deep blue. As shown in Fig. 4b the area covered by the p(3=2) reconstruction decreased while more (1=1) islands interconnected by strands appeared. The 3= spots in the LEED pattern were replaced by weak streaks between the 1= spots as shown in Fig. 1c. The 1= spots were streaked along both the w0 1 0x and w0 0 1x directions. Annealing for another 15 h in UHV at 810 K caused no noticeable change in the color of the film. The

M. Li et al. / Thin Solid Films 446 (2004) 238247

243

sample was then annealed in 3=10y5 Torr O2 for 50 min at 820 K before annealing in UHV at 820 K for 40 min, after which a streaky (1=1) LEED pattern similar to Fig. 1c was observed. As shown in Fig. 4c, this increased the density of (1=1) domains. The (1=1) unit cell is highlighted in the small-scale image in Fig. 5a, which shows a different area of the surface in Fig. 4c. The contrast was enhanced in Fig. 5a by cycling through the gray scale on each terrace; as a result the dark grooves in Fig. 4c appear white in Fig. 5a. The images also show that the (1=1) islands often terminate in p(3=2) strands. Meanwhile, the (1=1) islands appear as wider strands. Interestingly, the terrace steps and the domain boundaries between strands and (1=1) islands on the same terrace oriented favorably along the w0 2 1x direction as highlighted by the arrows in Fig. 4c. As highlighted by arrows in Fig. 5b, multiple domain

Fig. 4. STM images of WO3(1 0 0) thin films annealed in UHV. (a) After annealing the surface in Fig. 2d for 5 h at 850 K; (b) an additional 17 h of annealing at 830 K; (c) further 15 h of annealing at 810 K. The imaging biases were 2 V for (a), (b) and 1.8 V for (c).

Fig. 5. Small-scale STM images of the surface in Fig. 4c. (a) The (1=1) unit cell highlighted on a (1=1) domain; (b) (1=1) islands interconnected with strands. Arrows highlight anti-phase domain boundaries between strands. The imaging biases were 1.5 V for both images.

244

M. Li et al. / Thin Solid Films 446 (2004) 238247

6a were due to the interception of CS planes with the (1 0 0) surface w24x. It was noticed that traces of potassium were always present on the surface. In a previous study on rutile, it was suggested that the CS plane formation was related to Ca segregation to the surface w25x. Here, however, K was always present on the surface but the half-height steps were only seen following reduction by prolonged annealing in UHV, indicating that the shear planes in this case are due to reduction. Occasionally, w0 1 0x- or w0 0 1x-oriented troughs (Fig. 6b) similar to those separating the strands were seen on the (1=1) terraces while no w2 1 0xoriented steps were observed. 4. Discussion At annealing temperatures above 800 K the WO3(1 0 0) film experienced a net-reducing environment even in the presence of O2 or NO2. The surface phase transitions upon reduction can be summarized as follows: (1) a transition from a p(2=2) surface with 1 y 4 monolayer of on-top oxygen to a p(4=2) stranded reconstruction with no oxygen on top of the strands; (2) a change in the strand spacing from 4= to 3= to create a p(3=2) structure; (3) a conversion of the p(3=2) surface to a (1=1) surface with no on-top oxygen and (4) introduction of bulk CS planes that create half-height steps where they intersect the (1=1) terraces. Since the color of the film can be associated with the bulk density of W5q w26x, the darkening that accompanied the surface phase transitions indicates a link between the bulk oxygen deficiency and the surface structure. Similar effects have previously been reported for rutile TiO2(1 1 0) single crystals, where numerous metastable surface structures were induced by oxidizing crystals with different initial bulk oxygen deficiencies indicated by the color of the crystals w27x. The transport of interstitial-reduced cations (i.e. Ti3q in TiO2) has been shown to play an important role in the oxidation

Fig. 6. STM images of WO3(1 0 0) thin films after further reduction in UHV. (a) After annealing the surface in Fig. 4c between 820 and 850 K for 45 h; (b) a small-scale image of the surface in (a). The (1=1) unit cell is highlighted on the terrace. The imaging biases were 2 V for (a) and (b).

boundaries across p(3=2) strands can also be resolved in a small-scale image of the same surface in Fig. 4c. The appearance of (1=1) islands and the small size of the p(3=2) domains can explain the disappearance of the 1 y 3 order spots in the LEED pattern in Fig. 1c. The film was then further reduced by UHV annealing between 820 and 850 K for a total of 45 h, after which the film was noticeably darker and a strongly streaked (1=1) LEED pattern was observed as shown in Fig. 1d. The STM image in Fig. 6a shows that the surface was covered by narrow (1=1) terraces (typical width: ;5 nm) with irregularly spaced steps that can account for the streaking in the LEED pattern. The atomically resolved (1=1) structure can be seen in Fig. 6b. Straight steps oriented in w0 1 0x or w0 0 1x directions terminated the (1=1) terraces. As illustrated in the profile in Fig. 7 taken along the line labeled CD in Fig. 6a, the step heights were 0.18 and 0.55 nm or 1 y 2 and 3 y 2 the distance between neighboring WO6 octahedra along the w1 0 0x direction. Meanwhile, the step height measured along AB in Fig. 4c across a step separating p(3=2) and (1=1) terraces was 0.39 nm, consistent with the expected monatomic step height. The half-height steps and the dark color suggest that the step edges in Fig.

Fig. 7. Line profiles giving the mono-step height for the p(3=2)y(1=1) surface (AB from Fig. 4c) and the CS terminated step height (CD from Fig. 6a) on the (1=1) surface.

M. Li et al. / Thin Solid Films 446 (2004) 238247

245

and reduction of transition metal oxides w18,28,29x. For (0 0 1)-oriented WO3 single crystals, a metastable surface phase composed of W6O18 clusters was observed when atomically rough surfaces created by cleaving in air were slowly reduced w18x. These clusters gave way to large, smooth c(2=2) terraces upon continued reduction. These results were explained by reduced W5q migrating into the bulk to leave behind a smooth surface. The formation of p(4=2) strands was suggested to proceed in a similar way w22x: as the p(2=2) surface is reduced, the W5q concentration builds up on the surface. When the chemical potential of W5q on the surface exceeds that in the bulk, the reduced cations migrate into the bulk causing the terraces to etch away. The resulting vacancies organize into ordered troughs to produce the p(4=2) and p(3=2) structures. The favored orientations of the domain boundaries between (1=1) islands, and strands in the w0 2 1x, w0 1 0x and w0 0 1x directions suggest comparable formation energies. As intermediate phases between p(2=2) and (1=1), both p(3=2) and p(4=2) have on-top oxygen atoms in the troughs. This is consistent with the appearance of an exclusively p(3=2)-terminated surface after reducing a p(4=2) surface; i.e. narrowing the troughs reduces the surface. This also suggests that W5q surface diffusion or diffusion of the strands, competes with bulk migration and dominates as the troughs become narrow, leading to narrower rather than deeper troughs as more oxygen is removed. The formation of (1=1) islands from p(3=2) terraces then proceeds through strand, or trough, coalescence as suggested by the STM images in Fig. 4b and c and Fig. 5b that show wider strands, occasional troughs in the midst of (1=1) terraces, strands connecting (1=1) islands, and strands extending out of the edges of (1=1) islands. Strands of different lengths coalescing can explain the latter. It is not clear if W5q migration into the bulk continues once the (1=1) structure forms, however, the appearance of bulk shear planes when the surface is completely (1=1)terminated suggests that the bulk becomes saturated with interstitial W5q causing a change in the way bulk oxygen deficiency is accommodated. It is well known that the continued loss of O atoms from heavily reduced metal oxides leads to planar defects such as bulk CS planes w24,30x. For rutile TiO2(1 1 0), STM studies revealed steps as terminations for the {1 3 2} w31x and {1 1 2} w25x series of CS planes. For WO3, it has been shown that the elimination of bulk point defects leads to well-defined CS planes oriented in {10m} directions w32x. In this case, cornershared WO6 octahedra become edge-sharing along CS planes leading to a reduction in the O to W ratio. Thus, the CS planes can be used as a measure of the local stoichiometry. When the ordering of the CS planes is complete, the crystals are members of the homologous series of WnO3ny(my1). Slightly reduced crystals favor

the series with ms2, giving ordered {1 0 2} CS planes with the stoichiometry of WnO3ny1, while ms3 and 4 are expected for more heavily reduced crystals with the CS plane orientations of {1 0 3} and {1 0 4} and the compositions WnO3ny2 and WnO3ny3, respectively. The series W11O32, W15O43 and W19O54 was proposed accordingly with ordered {1 0 2}, {1 0 3} and {1 0 4} CS planes spaced 9l y 2 apart (1.69 nm) across the w1 0 0x direction w24x. The measured CS plane spacing on surfaces similar to Fig. 6a varied between 3 and 15 nm indicating a variation in the local stoichiometry. Apparently, the annealing time or temperature was insufficient for the shear planes to order or the film to separate into WO3 and WnO3ny1 phases. A structural 0 1) CS planes is illusmodel for steps induced by (2 trated in Fig. 8, where the edge-sharing WO6 octahedra form the CS planes. The model illustrates how a CS plane creates a half-height step when it intersects a surface and how the combination of a monatomic step with a CS plane intersecting the surface leads to a 3 y 2height step as seen in Fig. 6a and Fig. 7. Identical surfaces can be created with any {10m} CS plane, and so it is impossible to determine m from the STM images or to tell whether CS planes propagate through the entire film or are restricted to the surface region. Therefore, the observed half-height steps are attributed to interceptions of scattered {10m} CS planes with the surface. It has been reported that the incorporation of CS plane steps on rutile stabilizes the step edges by reducing the step edge free energy w31x. Fig. 8 shows that on WO3(1 0 0), unlike monatomic steps, the half-height CS plane steps do not reduce the coordination of the step edge atoms, suggesting that the CS plane steps are also lower in energy on this surface. This can explain the higher density of half-height vs. 3 y 2-height steps. Interestingly, well-ordered bulk CS planes corresponding to W40O118 were observed when monoclinic WO3 films on sapphire were annealed in air at 520670 K w20x. The sapphire substrate was claimed to favor the growth of large grains during annealing which is essential for such long-range structures to form. Despite lateral grain sizes of at least 3.3 mm (the scan range of the STM) in the films studied here, no well-ordered CS planes were observed even though the conditions were more strongly reducing, higher temperatures and no oxygen. Gas sensing relies on conductivity increases (decreases) when tungsten oxide films are exposed to a reductant (oxidant). The results presented here show that oxidizing and reducing WO3 films not only changes the conductivity, which is related to the film color through the W5q concentration, but also the surface structure in a dramatic way. In particular, the models in Fig. 3 suggest that reduction to the stranded p(4=2) and p(3=2) surfaces introduce fourfold coordinated W ions while

246

M. Li et al. / Thin Solid Films 446 (2004) 238247

0 1) CS planes giving half- and 3y2-height steps on the (1=1) surface. The bulk (2 0 1) CS Fig. 8. A structural model of steps induced by (2 planes are formed by edge-sharing WO6 octahedra in gray.

Fig. 7 shows that these sites are eliminated when the (1=1) surface forms. Since the coordination of the exposed transition metal cations is generally considered to be an important factor in adsorption and catalysis on oxide surfaces, the surface chemistry would be expected to change when the p(4=2) and p(3=2) structures appear on the surface. This suggests a complex dependence of the sensor sensitivity to different species on the bulk film stoichiometry. To address this issue, we are currently studying how the p(4=2) and p(3=2) structures affect adsorption and oxidation of simple organic molecules on WO3. 5. Summary In net-reducing environments at temperatures above 800 K, epitaxial WO3(1 0 0) thin films display a series of surface phase transitions. Reduction of the p(2=2) surface causes W5q migration into the bulk leaving vacancies on the surface that organize into ordered troughs to form the p(4=2) strand-terminated surface. Continued reduction causes the troughs to narrow to produce a p(3=2) surface composed of the same sorts of strands. On more heavily reduced films, the strands coalesce to form (1=1) islands that eventually cover the entire surface. When the surface is (1=1)-terminated, a second bulk reduction mechanism appears: crystal shear. Crystal shear affects the surface through the introduction of half-height steps due to the CS planes intersecting the surface. After prolonged annealing at 800 K, the CS planes are randomly spaced indicating that no well-defined CS phases form under these conditions. Acknowledgments The authors acknowledge the help of Weiwei Gao and Jun Wang in carrying out this work. A. Posadas

and C.H. Ahn acknowledge the support from NSF (DMR-0134721). This project is supported by the Department of Energy through Basic Energy Sciences grant number DE-FG02-98ER14882. References
w1x M. Akiyama, Z. Zhang, J. Tamaki, N. Miura, N. Yamazoe, Sensors Actuators B 1314 (1993) 619. w2x M.J. Madou, S.R. Morrison, Chemical Sensing with Solid State Devices, Oxford University Press, Oxford, 1989. w3x F.A. Cotton, G. Wilkinson, Advances in Organic Chemistry, Wiley, New York, 1988. w4x J.S.E.M. Svensson, C.G. Granqvist, Solar Energy Mater. 11 (1984) 29. w5x M. Green, Electrochromism and Electrochromic Displays, University of Rome, Italy, 1982. w6x T. Lambert, Solar Energy Mater. 11 (1984) 1. w7x J. Svenson, C. Granqvist, Appl. Phys. Lett. 45 (1984) 828. w8x M. Sundberg, R.J.D. Tilley, J. Solid State Chem. 11 (1974) 150. w9x A. Hamnett, J.B. Goodenough, in: O. Madelung (Ed.), Semiconductors, vol. 17g, Springer, Berlin, 1984. w10x E. Salje, Acta Crystallogr. B 33 (1977) 574. w11x E. Salje, K. Viswanathan, Acta Crystallogr. A 31 (1975) 356. w12x B.O. Loopstra, P. Boldrini, Acta Crystallogr. 21 (1966) 158. w13x E.K.H. Salje, S. Rehmann, F. Pobell, D. Morris, K.S. Knight, T. Herrmannsdorfer, M.T. Dove, J. Phys.: Condens. Matter 9 (1997) 6563. w14x R. Diehl, G. Brandt, Acta Crystallogr. B 34 (1978) 1105. w15x R.A. Dixon, J.J. Williams, D. Morris, J. Rebane, F.H. Jones, R.G. Egdell, S.W. Downes, Surf. Sci. 399 (1998) 199. w16x F.H. Jones, K. Rawlings, J.S. Foord, P.A. Cox, R.G. Egdell, J.B. Pethica, B.M.R. Wanklyn, Phys. Rev. B 52 (1995) 14392. w17x F.H. Jones, K. Rawlings, J.S. Foord, R.G. Egdell, J.B. Pethica, B.M.R. Wanklyn, S.C. Parker, P.M. Oliver, Surf. Sci. 359 (1996) 107. w18x R.E. Tanner, E.I. Altman, J. Vac. Sci. Technol. A 19 (2001) 1502. w19x M. Gillet, A. Al-Mohammad, C. Lemire, Thin Solid Films 410 (2002) 194.

M. Li et al. / Thin Solid Films 446 (2004) 238247 w20x A. Al-Mohammad, M. Gillet, Thin Solid Films 408 (2002) 302. w21x S.C. Moulzolf, L.J. Legore, R.J. Lad, Thin Solid Films 400 (2001) 56. w22x M. Li, E.I. Altman, A. Posadas, C.H. Ahn, Surf. Sci. 542 (2003) 22. w23x C.Y. Nakakura, V.M. Phanse, G. Zheng, G. Bannon, E.I. Altman, K.P. Lee, Rev. Sci. Instrum. 69 (1998) 3251. w24x E. Iguchi, R.J.D. Tilley, Philos. Trans. R. Soc. Lond. Ser. A, Math. Phys. Sci. 286 (1977) 55. w25x H. Norenberg, R.E. Tanner, K.D. Schierbaum, S. Fischer, G.A.D. Briggs, Surf. Sci. 396 (1998) 52. w26x M.J. Sienko, J.M. Berak, The Chemistry of Extended Defects in Non-Metallic Solids, North-Holland, Amsterdam, 1970.

247

w27x M. Li, W. Hebenstreit, U. Diebold, A.M. Tyryshkin, M.K. Bowman, G.G. Dunham, M.A. Henderson, J. Phys. Chem. B 104 (2000) 4944. w28x M. Li, W. Hebenstreit, L. Gross, U. Diebold, M.A. Henderson, D.R. Jennison, P.A. Schultz, M.P. Sears, Surf. Sci. 437 (1999) 173. w29x H. Onishi, Y. Iwasawa, Phys. Rev. Lett. 76 (1996) 791. w30x L.A. Bursill, B.G. Hyde, Prog. Solid State Chem. 7 (1972) 177. w31x R.A. Bennett, S. Poulston, P. Stone, M. Bowker, Phys. Rev. B 59 (1999) 10341. w32x K. Kosuge, Chemistry of Non-Stoichiometric Compounds, Oxford University Press, Oxford, 1994.

Вам также может понравиться