Unconventional grain growth suppression in oxygen-rich metal oxide nanoribbons.

Nanograined metal oxides are requisite for diverse applications that use large surface area, such as gas sensors and catalysts. However, nanoscale grains are thermodynamically unstable and tend to coarsen at elevated temperatures. Here, we report effective grain growth suppression in metal oxide nanoribbons annealed at high temperature (900°C) by tuning the metal-to-oxygen ratio and confining the nanoribbons. Despite the high annealing temperatures, the average grain size was maintained at ~6 nm, which also retained their structural integrity. We observe that excess oxygen in amorphous tin oxide nanoribbons prevents merging of small grains during crystallization, leading to suppressed grain growth. As an exemplary application, we demonstrate a gas sensor using grain growth–suppressed tin oxide nanoribbons, which exhibited both high sensitivity and unusual long-term operation stability. Our findings provide a previously unknown pathway to simultaneously achieve high performance and excellent thermal stability in nanograined metal oxide nanostructures.

INTRODUCTION

Stabilizing nanoscale grains in polycrystalline materials is important for numerous applications that use surface reactions, such as catalysis (–) and gas sensing (–). For example, the activity and endurance of metal oxide–based devices are determined by the size and stability of small grains and surface and grain boundaries that interact predominantly with target molecules (–). For high-performance gas sensors, the formation of small grains and a porous structure is desirable for maximum exposure and easy transport of gas molecules to surface reaction sites. However, metal oxides usually require annealing steps to obtain high crystallinity, which inevitably causes undesirable grain growth and coarsening, as well as deterioration of porous nanostructures. Moreover, metal oxide devices operated at elevated temperatures and under chemical exposure for prolonged time often show instability, drift in response, and eventual failure due to the grain growth and coarsening. Therefore, maintaining their small grain structures while ensuring high crystallinity during the fabrication of metal oxide nanostructures remains a substantial challenge.

Although several mechanisms, such as kinetic pinning or thermodynamic stabilization, have been shown to stabilize nanoscale grains (–), they involve incorporation of immobile secondary phase or segregation of solutes at grain boundaries, and the resulting grains are relatively large (usually in the range of several tens of nanometers). Using in situ transmission electron microscopy (TEM), we previously observed that grain growth of Pt-based metallic glass nanorods during crystallization can be significantly modulated by nanoscale confinement (–). In addition, crystallization kinetics, such as the nucleation rate and grain growth, can be extensively modulated by local chemical compositions, which can promote or suppress the crystallization kinetics (–). However, the effects of the nanoscale confinement and local chemical compositions on the grain growth of metal oxides are unclear, and the underlying mechanism has not been explored because of the difficulty in the simultaneous modulation of size and chemical composition.

Here, we report that systematic engineering of the nanoribbons and the collaborative effects of nanoscale confinement and chemical composition can successfully suppress the grain growth of metal oxides. Specifically, we show an average grain size of 6 nm in tin oxide (SnOx) nanoribbons even with 900°C annealing for 36 hours. To elucidate the mechanism of this unusual phenomenon, we precisely controlled the dimension and chemical composition of as-prepared amorphous nanoribbon structures and thereupon demonstrate that the final grain size decreases with a decreasing ratio of Sn/O and the size of nanoribbons. In situ TEM irradiation experiments on the amorphous SnOx nanoribbons show that crystallization can be induced by knock-on sputtering of oxygen at room temperature, which directly demonstrates the effects of the Sn/O ratio on the kinetics of grain growth. In situ TEM experiments also reveal that the excess oxygen atoms in amorphous SnOx nanoribbons act as a barrier to merging of grains and create nanoscale pores as they leave the nanoribbons during annealing. Last, we demonstrate excellent gas sensing performance of small-grain nanoribbons with unusually stable operation of more than 5 days at 350°C.

RESULTS

Fabrication and characterization of amorphous metal oxide nanoribbons

For the controllable fabrication of metal oxide nanoribbons, we used solvent-assisted nanotransfer printing (S-nTP), a strategy that we previously developed and reported (). Figure 1A schematically illustrates the fabrication steps for the metal oxide nanoribbon array, using the S-nTP method. S-nTP uses a polymeric transfer medium with solvent-controlled, switchable adhesion to transfer nanostructures with a high resolution and a high transfer yield (, ). First, a poly(methyl methacrylate) (PMMA) thin film was spin-coated onto the master template of topographic line patterns to form a PMMA replica film. The PMMA replica film was peeled off using a polyimide (PI) adhesive film (Fig. 1A, ii), and amorphous metal oxide was deposited on the replica PMMA/PI film by electron beam (e-beam) evaporation (Fig. 1A, iii), forming an array of metal oxide nanoribbons in the grooves of the PMMA/PI film. The sample was then exposed to a mixed solvent vapor of acetone and heptane to reduce the adhesion between the PI film and PMMA replica, followed by transfer onto a target substrate (Fig. 1A, iv). After washing the PMMA layer with acetone, only the array of the metal oxide nanoribbons remained on the target substrate (Fig. 1A, v).

Fig. 1.

Fabrication of metal oxide nanoribbon arrays and their stability upon annealing.

(A) Schematic illustrations of S-nTP process to fabricate a tin oxide nanoribbon array: (i) a master template, (ii) peeling of a PMMA (green)/PI (yellow) film from the master template, (iii) evaporation of metal oxide (gray) onto the PMMA/PI replica film, (iv) transfer of the film onto a target substrate (purple), and (v) metal oxide nanoribbon array (gray) after removing the PMMA. (B) Scanning electron microscopy (SEM) image of a (left) transfer-printed nanoribbon array on a target substrate and (right) annealed at 900°C in air for 6 hours. Scale bars, 200 nm. Region enclosed by rectangles at higher magnification. Scale bars, 100 nm. (C) Bright-field TEM image of an as-transferred, amorphous SnOx nanoribbon and high-angle annular dark-field scanning TEM (HAADF-STEM) images of a polycrystalline SnO2 nanoribbon after ex situ annealing at various temperatures (as indicated) in air for 6 hours. Scale bars, 50 nm. (D) XRD patterns of SnOx nanoribbons (width of 50 nm and thickness of 20 nm) before and after annealing. a.u., arbitrary units. (E) High-resolution TEM and HAADF-STEM images of a SnOx nanoribbon annealed at 900°C in air for 6 hours. Scale bars, 5 nm.

Using S-nTP, we successfully fabricated an array of perfectly aligned amorphous nanoribbons that were 50 nm in width and 20 nm in thickness on the SiO2/Si substrate, as shown in the scanning electron microscopy (SEM) image in Fig. 1B. The amorphous state of the as-transferred nanoribbons was confirmed by TEM and x-ray diffraction (XRD) analyses (Fig. 1, C and D). We examined the crystallization of amorphous SnOx nanoribbons annealed at various temperatures in air for 6 hours using high-angle annular dark-field scanning TEM (HAADF-STEM). It showed that grain growth was effectively suppressed, even at annealing temperatures as high as 900°C (Fig. 1C). The overall alignment and the external shape of the nanoribbon array were retained after the annealing process (Fig. 1B). However, the continuous, amorphous microstructures of the nanoribbons were transformed to porous structures composed of nanoscale grains, as shown in the TEM and HAADF-STEM image (Fig. 1E). XRD analysis indicates that polycrystalline SnO2 formed after annealing (Fig. 1D), and a high-resolution TEM image shows a lattice spacing of 3.3 Å, in agreement with the SnO2 (110) lattice spacing (Fig. 1E).

Unexpectedly, the grains remained on a nanoscale without any notable grain coarsening after annealing for 6 hours. We annealed the amorphous SnOx nanoribbons in air for 6 hours at different temperatures and obtained the average grain size of crystallized SnO2 to be 5.2 ± 1.1, 5.5 ± 0.9, and 6.2 ± 0.8 nm at the annealing temperatures of 500°, 700°, and 900°C, respectively (fig. S1). Thus, we observed that, despite high-temperature annealing for a sufficiently long time, the SnO2 grains in nanoribbons did not grow larger than 7 nm in diameter. In contrast, the grain size of an amorphous SnOx thin film with the same thickness (20 nm) increased rapidly with an increase in annealing temperature, reaching ~50 nm in lateral dimensions after annealing at 900°C in air for 6 hours (fig. S2).

Effects of nanoribbon size on grain growth

To investigate confinement effects on the observed suppression of grain growth, we fabricated 50-nm-wide SnOx nanoribbons with varying thicknesses of 10, 20, and 30 nm and annealed them at 900°C in air for 6 hours (Fig. 2, A to C). After annealing, the average grain sizes were 6.0 ± 0.8, 6.2 ± 0.9, and 12.7 ± 1.9 nm for increasing thickness of 10, 20, and 30 nm of the nanoribbons, respectively. In addition, the grains appeared to be individually isolated in the 10- and 20-nm-thick nanoribbons while they were fully connected in the 30-nm-thick nanoribbon. Figure 2D provides a summary of the grain growth: Increasing the initial thickness of the nanoribbon results in an increase in the grain size. We also annealed the nanoribbons with varying thicknesses at different temperatures (500°, 700°, and 900°C) and observed the same phenomenon (fig. S3). Thus, we demonstrate that the suppression of grain growth is most pronounced in the thinnest SnO2 nanoribbons.

Fig. 2.

Characterization of grain growth with varying thicknesses of nanoribbon.

Fifty-nanometer-wide amorphous SnOx nanoribbons with thicknesses of (A) 10 nm, (B) 20 nm, and (C) 30 nm were annealed at 900°C in air. Scale bars, 20 nm. Cross sections of the nanoribbons are shown in the top schematics in (A) to (C). (D) Average grain size as a function of annealing temperature for different thicknesses of nanoribbon. (E) X-ray photoelectron spectroscopy (XPS) spectra of the amorphous SnOx nanoribbons with different thicknesses, showing the O 1s and Sn 3d peaks for the composition analysis.

To test whether the observed suppression of grain growth is purely due to nanoscale confinement, we measured the chemical compositions of the initial amorphous SnOx nanoribbons, using x-ray photoelectron spectroscopy (XPS). The Sn-to-O ratios of the nanoribbons were analyzed as a function of the nanoribbon thickness (Fig. 2E), and different chemical compositions were observed for different thicknesses. Thinner nanoribbons were observed to be more oxygen rich; with decreasing thickness, the O/Sn atomic ratio increased from 3.5 to 8.1. The rather high O/Sn atomic ratio can be attributed to the formation of adsorbed oxygen on the surface, which is more prominent for thinner nanoribbons. For the 10-nm-thick nanoribbon, the O 1s peak at 532 eV suggests the presence of nonlattice oxygen (e.g., O22−) on the nanoribbon. By contrast, the oxygen peak of bulk SnO2 is located at 530 eV, which reflects the lattice oxygen ions (O2−) in the metal oxide (see detailed analysis of XPS Sn 3d and O 1s peaks in fig. S4) (, ). The XPS depth profile experiments revealed a depth-dependent compositional change in a ~80-nm-thick film, with ~5-nm-thick oxygen-rich top and bottom surfaces (fig. S5), which can explain the high O/Sn ratio for the 10-nm-thick nanoribbon.

We note that the observed suppression of grain growth is not specific to tin oxide but is broadly applicable to other metal oxides (NiO and CeO2) at the nanoscale. We observed that amorphous NiOx nanoribbons that were 50 nm in width and 10 nm in thickness resulted in an average grain size of ~6 nm after annealing at 900°C for 6 hours in air (fig. S6).

Effects of chemical compositions on grain growth

The results presented in Fig. 2 show the coupled effects of physical confinement and chemical composition on the suppression of grain growth. To investigate only the effect of the chemical composition on suppressing grain growth, we kept the dimension of the SnOx nanoribbons to be 50 nm wide and 20 nm thick and systematically varied either the annealing environment or the O/Sn ratio of the initial SnOx nanoribbon. Figure 3 (A to C) shows SnOx nanoribbons that were annealed at 900°C for 6 hours in three different reducing environments: oxidation (air), inert (Ar), and reduction (H2/Ar). The average grain sizes after annealing were 6.1 ± 0.75 nm annealed in air, 6.9 ± 0.61 nm annealed in Ar, and 11.3 ± 2.69 nm annealed in H2/Ar (Fig. 3D), indicating that the concentration of oxygen in annealing environment influences the resulting grain size. The grains were observed to be smallest when excess oxygen was present during annealing. Thus, we hypothesize that chemical composition is important for controlling the grain growth kinetics and that preserving high oxygen content promotes suppression of grain growth.

Fig. 3.

Characterization of grain growth with different annealing environments and nanoribbon compositions.

Fifty-nanometer-wide and 20-nm-thick amorphous SnOx nanoribbons were annealed at 900°C in different annealing environments of (A) air, (B) Ar, and (C) H2/Ar. Scale bars, 20 nm. (D) Average grain size as a function of the annealing environment. Fifty-nanometer-wide and 20-nm-thick amorphous SnOx nanoribbons with different concentrations of Sn (E) 11%, (F) 16%, and (G) 21% were annealed at 900°C in air. Scale bars, 20 nm. (H) Average grain size as a function of the initial Sn concentration. (I) The average grain size after annealing at 900°C for 6 hours in air as a function of the cross-sectional area and initial Sn concentration of nanoribbon, where the color saturation indicates the grain size.

To further verify our hypothesis, we fabricated nanoribbons of a fixed dimension (50 nm in width and 20 nm in thickness), but with different initial O/Sn ratios, and annealed them at 900°C for 6 hours in air (see details in Materials and Methods). The compositions of the as-transferred, amorphous nanoribbons were analyzed by XPS to show initial atomic percent Sn of 11, 16, and 21% for three batches of nanoribbons. Figure 3 (E to G) exhibits TEM images of the annealed nanoribbons, which give average grain sizes of 6.1 ± 0.8, 7.4 ± 1.1, and 12.3 ± 3.5 nm for the nanoribbon with 11, 16, and 21% Sn, respectively (Fig. 3H). Increasing the Sn fraction leads to a larger grain size, and, concurrently, the nanoribbons also demonstrate a transition from disconnected grains to connected grains. Thus, the observed suppression of grain growth was clearly influenced by the composition of the initial SnOx nanoribbons. The results shown in Fig. 3 agree with those in Fig. 2, which presents that oxygen-rich nanoribbons contain smaller grains. In addition, we analyzed the C/Sn ratio of the nanoribbons, which remains unchanged as a function of nanoribbon size and Sn content using the XPS data (fig. S7), ruling out the possibility that carbon may affect the grain growth.

While the compositional effect clearly suppresses the grain growth, nanoscale confinement also plays a role. We observe that a 30-nm-thick SnOx nanoribbon with initial 13.5% Sn crystallizes with an average grain size of 13.3 ± 3.9 nm (fig. S8), while a 20-nm-thick SnOx nanoribbon with initial 16% Sn crystallizes with a grain size of 7.4 ± 1.1 nm (Fig. 3F). Thus, grains were smaller for the thinner nanoribbon despite its higher Sn content, which would lead to larger grains if the compositional effect were the only factor. In addition, for the same initial Sn % content, a SnOx film shows significantly larger grains than SnOx nanoribbons at the same thickness (fig. S2). In summary, suppression of grain growth is observed when the initial Sn % is low and the nanoribbons are small, which is summarized in Fig. 3I.

Grain growth kinetics and mechanism

The grain size versus annealing temperature curve shows that grain growth was effectively suppressed in the 50-nm-wide and 10-nm-thick oxygen-rich (Sn 11%) nanoribbons even after being annealed in air for 36 hours at temperatures as high as 900°C (Fig. 4A). The grain growth–suppressed (GGS) SnO2 nanoribbons (10/50 nm, Sn 11%) exhibited a stable grain size of ~6 nm that was essentially unchanged. In contrast, the grain size of the non-GGS SnO2 nanoribbons (30/50 nm, Sn 21%) increased with the annealing temperature.

Fig. 4.

Grain growth kinetics.

(A) Average grain sizes of the GGS and non-GGS SnO2 nanoribbons annealed at different temperatures for 36 hours. (B) The average grain size of the GGS SnO2 nanoribbons with varying annealing temperature and duration. (C) The Arrhenius plots for calculating the grain growth activation energy of the GGS SnO2 nanoribbons, non-GGS SnO2 nanoribbons, and 500-nm-thick SnO2 film.

The kinetics and underlying mechanism of grain growth suppression in the SnO2 nanoribbons were determined quantitatively (Fig. 4B). As controls, the grain growth in a 500-nm-thick SnO2 film and non-GGS SnO2 nanoribbons were also investigated (fig. S9, A and B). A classic equation describing the relationship between the grain size G and the annealing time t, G − G0 = k (t − t0) (), was used, where t0 was 2 hours, G0 is the grain size at t0, and n is an exponent related to the growth mechanism in the range of 1 to 4. Setting n at 3, a good linear relationship was found between (G3 − G03) and (t − t0), and the grain growth rate constant (k) was obtained from a linear fitting (fig. S9, C and D).

Specifically, we estimated the k values for the GGS and non-GGS SnO2 nanoribbons as well as the 500-nm-thick SnO2 film at different temperatures. By plotting the logarithm of k against the reciprocal of absolute temperature as shown in Fig. 4C, the apparent activation energy of grain growth could be derived from the slope of the linear fittings (see calculation details in Materials and Methods). The activation energy of the non-GGS SnO2 nanoribbons (34.6 kJ/mol) was 21% lower than that of the 500-nm-thick SnO2 film (43.17 kJ/mol). The reduced activation energy of the non-GGS SnO2 nanoribbons indicates that grain growth in the non-GGS SnO2 nanoribbons is easier than in the thick film counterpart because of their high surface-to-volume ratios but had a similar range because of similar amorphous state for starting conditions. In contrast, the GGS SnO2 nanoribbons showed significantly low k’s up to 900°C and did not follow the usual classical growth, indicating a different grain growth mechanism that is related to suppression of the driving force for grain growth.

In situ TEM for direct visualization of grain growth

To understand the suppressed driving force for the grain growth in GGS nanoribbons, we carried out in situ TEM experiments on the amorphous SnOx nanoribbons and directly visualized the grain growth in the nanoribbon. We used e-beam irradiation to induce crystallization in the amorphous nanoribbons at room temperature. Figure 5 (A and D) shows low-magnification TEM images of a 10- and 30-nm-thick SnOx nanoribbon before and after e-beam irradiation, respectively. We observe that, under similar e-beam irradiation conditions, more nuclei formed in the 30-nm-thick nanoribbon, which agrees with ex situ heating results shown in Fig. 2. Figure 5 (C and F) shows TEM snapshot images from in situ movies of the 10- and 30-nm-thick SnOx nanoribbons (movies S1 and S2). For the 10-nm-thick nanoribbon, we observed that grains did not grow much beyond 480-s e-beam irradiation. We tracked five grains from the in situ TEM movie and measured their cluster size as a function of irradiation time (Fig. 5B); their size remains at ~5 nm throughout the movie. By contrast, the 30-nm-thick nanoribbon showed more grain growth, which is evident in the time series of TEM images (Fig. 5F). Individual grains slowly grew larger over time, and two grains actually merged to make a bigger grain (Fig. 5E). Because the grain growth jumps significantly when two grains merge, the main mechanism of grain growth is attributed to merging of grains.

Fig. 5.

Grain growth monitoring of SnOx nanoribbons with varying thicknesses using in situ TEM.

(A) TEM images of a 10-nm-thick SnOx nanoribbon before (left) and after (right) an in situ TEM movie (movie S1). Scale bars, 5 nm. (B) Evolution of the grain size for five different grains, tracked as a function of time from the in situ movie. (C) TEM images from the in situ TEM movie, showing the grain growth of particle 1 (P1) and P2. Scale bars, 5 nm. The images are from the region marked by the red dotted box in (A). (D) TEM images of a 30-nm-thick SnOx nanoribbon before (left) and after (right) an in situ TEM movie (movie S2). Scale bars, 5 nm. (E) Evolution of the grain size for five different grains, tracked as a function of time from the in situ movie. (F) TEM images from the in situ TEM movie, showing the grain growth of P1. Scale bars, 5 nm. The images are from the region marked by the blue dotted box in (D).

We hypothesize that the e-beam irradiation induced crystallization because of knock-on damage on oxygen (–), which would increase the relative Sn concentration in the nanoribbon to create favorable conditions for crystallization. To confirm the knock-on sputtering of oxygen, we acquired energy-dispersive x-ray spectra of the nanoribbons before and after the in situ experiments, which showed a significant decrease of the O peak after the e-beam irradiation (fig. S10). Thus, we directly showed that the local Sn/O composition is important to control nucleation and growth of SnO2 nanoscale grains. During thin SnOx nanoribbon crystallization, excess oxygen surrounding the SnO2 grains would act as a barrier to merging of the grains, leading to suppression of grain growth (fig. S11). In addition, the high concentration of oxygen in thin SnOx nanoribbons would lead to a porous structure after annealing as the excess oxygen would evaporate as O2 during annealing (fig. S12). This explains why the porous structures were observed for thin nanoribbons but not for thicker nanoribbons (Fig. 2, A to C).

High-temperature stability as gas sensor due to small grains

A chemoresistive gas sensor detects target gas molecules by charge transfer from the gas molecules adsorbed on the metal oxide surface of the sensor (). Thus, most gas sensors use porous metal oxides with nanoscale grains to increase the surface interactions with the gas molecules (–). In the case of the small SnOx nanoribbons with high oxygen content, the SnO2 grains did not grow larger than ~7 nm in size despite harsh annealing at 900°C for 36 hours. In addition, the excess oxygen in the SnOx nanoribbons resulted in a porous structure after annealing. Thus, an aligned array of small SnOx nanoribbons could provide a high-sensitivity gas sensor with long-term stability.

To investigate the gas sensing properties of the GGS nanoribbon, we fabricated a resistance-type sensor using an electrode with 20-μm channel width on the nanoribbons (Fig. 6A). The sensing mechanism of the GGS nanoribbon is attributed to the microporous structure with large reactive surface area and superior grain growth suppression properties, which retains small grain size for long-term operation (Fig. 6B, i). Figure 6 (C to E) shows the resistance changes and gas response of the GGS nanoribbon and the Sn-rich nanoribbon (denoted as dense nanoribbon), and the thin-film sensor tested as a control (Fig. 6B, ii and iii). Figure 6C presents the dynamic ethanol sensing characteristics of the various SnO2 nanostructures (GGS nanoribbon, dense nanoribbon, and thin film), as the concentration of ethanol decreased from 100 to 5 parts per million (ppm) balance air at 350°C with 1 and 5 days’ operation (see details of gas sensing measurements in Materials and Methods). Figure 6 (D and E) shows the resistance and response changes of the sensors, which were continuously examined for 120 hours at 350°C. The baseline resistance was measured as a function of time. Initially, the sensor resistance increased rapidly within the first 10 hours. Including the initial burn-in time, the resistance of the thin film and dense nanoribbon increased from the initial value by 13 and 3.35 times, respectively, while the GGS nanoribbon increased by only 1.8 times (Fig. 6D). Therefore, the drift in the resistance response over time is substantially smaller for the GGS nanoribbon sensor compared to the dense nanoribbon and thin-film sensor.

Fig. 6.

Gas sensing performance of a tin oxide nanoribbon array with nanoscale grains.

Illustrations of (A) the gas sensor with (B) three different nanostructures: (i) GGS nanoribbons with schematic illustration of micropore formation and enhanced sensing due to the interactions with the gas molecules at micropores, (ii) dense nanoribbons, and (iii) thin film. (C) Dynamic gas sensing properties of (top) SnO2 GGS nanoribbon array, (middle) SnO2 dense nanoribbon array, and (bottom) SnO2 thin film. The nanoribbons were 20 nm thick and 50 nm wide, and the film was 20 nm thick. The sensors were operated at 350°C in ethanol gas concentrations ranging from 100 to 5 ppm for 1 and 5 days. The nanoribbon arrays were annealed at 500°C in air before gas sensing. Comparisons of (D) baseline resistance drift in air and (E) response change at 5 ppm ethanol between the GGS nanoribbons, dense nanoribbons, and thin film operated at 350°C.

For the gas response comparison between the GGS nanoribbon, dense nanoribbon, and thin film (Fig. 6E), the GGS nanoribbon array exhibited a 2.5 and 9.3 times higher response than the dense nanoribbon and thin film, owing to the higher surface-to-volume ratio of the porous structure. After 120 hours of operation at 350°C, the change of the gas response from the initial value (ΔS/S0, S = ΔR/Rair, and ΔS = Smax − Smin) is within 13.3% for the GGS nanoribbon sensor, while it is 70.6 and 31.1% for the dense nanoribbon and thin film, respectively. Last, the response of the nanoribbon array remained stable over 120 hours, while the responses of the dense nanoribbon and thin film were found to decrease over time, indicating significant degradation of the sensor performance. Thus, by grain growth suppression, we demonstrate markedly improved thermal stability while maintaining high sensitivity using metal oxide nanoribbons.

DISCUSSION

In summary, we reported the unusual suppression of grain growth in amorphous tin oxide nanoribbons, induced via control of the excess oxygen in the initial amorphous nanoribbons and their nanoscale confinement. The excess oxygen prevents merging of small grains, which is the primary growth mechanism, as revealed by in situ TEM. Thus, by tuning the Sn/O ratio away from the SnO2, the nucleation and growth of the SnO2 grains are greatly suppressed in the nanoribbons, leading to stable preservation of nanoscale grains despite high-temperature annealing. As a result, we demonstrated more than 120 hours of continuous operation of gas sensing using an array of tin oxide nanoribbons at elevated temperature with high sensitivity and stability. We therefore present oxygen-rich compositional tuning as an effective method to suppress grain growth and stabilize nanoscale grains in metal oxide nanostructures. We also note that our approach is broadly applicable to other metal oxides such as NiO and CeO2. Our results are expected to be extended to other applications that are based on surface reactions of metal oxides, including gas sensors, emission catalysts, and fuel cells.

MATERIALS AND METHODS

Fabrication of amorphous metal oxide nanoribbons

Figure 1A schematically shows the fabrication process of an aligned array of metal oxide nanoribbons onto a receiver substrate using a previously reported method of S-nTP (see references for details) (). PMMA (molecular weight = 100 kg mol−1; Sigma-Aldrich Inc.) was dissolved in a mixed solvent of toluene, acetone, and heptane (4.5:4.5:1 by volume) to yield a 4 weight % solution. The solution was spin-casted onto a patterned Si master mold, which was treated with a polydimethylsiloxane brush (Polymer Source Inc.). A PI adhesive film (3M Inc.) was then attached to the surface of the PMMA film and pulled to peel off from the Si mold the PMMA film, which was an inverted image of the surface topography of the master mold. Metal oxide nanoribbons were formed through metal oxide deposition onto the PMMA replica, which was tilted 80°, using an e-beam evaporator. To deposit the metal oxide, SnO2 (99.99%; iTASCO, Korea) and NiO (99.99%; iTASCO, Korea) were used as the source. To tune the chemical composition of amorphous tin oxide nanoribbons deposited into the grooves of the PMMA replica, we controlled the opening time of the shutter in the e-beam evaporator. For oxygen-rich amorphous tin oxide nanoribbons [~11 Sn atomic % (at %)], the shutter was opened when the deposition rate reached 1 Å/s. For metal-rich nanoribbons with ~20 Sn at %, the shutter was opened after 20 nm of tin oxide was deposited, as indicated by the crystal monitor. After deposition, the metal oxide–deposited PMMA replica film was placed in a preheated chamber (55°C) that was saturated with acetone/heptane vapor. After 20 s, the film was removed from the chamber and brought into contact with the target substrate. Mild pressure was applied for uniform contact between the nanoribbon/replica/adhesive film and the target substrate. The adhesive film was selectively detached from the substrate, while the nanostructure/replica film remained on the target substrate.

Characterization of nanoribbons

Structural characterizations of the nanoribbons were carried out using a SEM and TEM. A field-emission SEM (Hitachi S-4800) was used at an acceleration voltage of 10 kV and a working distance of 5 mm. TEM experiments were performed with a FEI Tecnai Osiris 200-kV TEM, and high-resolution HAADF-STEM images were obtained using a 200-kV accelerating voltage TEM (FEI, Talos F200X). For the chemical compositions of the nanoribbons, XPS data were acquired using a multipurpose x-ray photoelectron spectrometer (Sigma Probe, Thermo Fisher Scientific). For XPS analysis, the metal oxide nanoribbons were transferred on Si substrate to avoid potential oxygen signal from the substrate. Bulk SnO2 [tin(IV) oxide, 325-mesh powder, 99.9%; Sigma-Aldrich] was used for reference. The XRD measurements were carried out using a multipurpose thin-film x-ray diffractometer (D/Max 2500, Rigaku).

Grain size calculation

We calculated the average crystallite size in the SnO2 thick film by using the Scherrer equation from XRD peaks, and the average grain size of the nanoribbons was determined through direct TEM observation. For the TEM measurements, we used ImageJ to quantify the grain size.

Activation energy calculation

The grain growth rate constant k is calculated by a linear fit between the grain size and the annealing time at a certain temperature. The data are fitted to the grain growth kinetic law of the following type: G − G0 = k (t − t0), where the value of k follows the typical Arrhenius behavior, k = k0·exp(−Ea/RT), where G and G0 are the grain diameters at time t and at t0, T is temperature, n is grain growth exponent, k is grain growth rate constant, k0 is constant, R is the gas constant, and Ea is the activation energy for grain growth. Thus, the logarithm of k linearly scales with the reciprocal of absolute temperature and the Ea can be obtained from the slope of the linear fit.

In situ TEM experiments of crystallization of nanoribbons

All in situ TEM experiments were performed using a FEI Tecnai Osiris 200-kV TEM. Nanoribbons were transferred to in situ TEM grids via a S-nTP method. We used commercially available in situ TEM heating grids (E-FHBC) and a heating holder (Aduro 300DT System) manufactured by Protochips Inc. The nanoribbons were video-recorded in real time under a bright-field TEM imaging mode (Snagit software).

Resistance measurement of a gas sensor and gas delivery system

To measure the resistance of the metal oxide nanoribbons on the SiO2/Si substrates, two electrodes were fabricated using conventional photolithography. Photoresist (AZ-5214E) was spin-casted on an array of nanoribbon/SiO2/Si substrate, and electrode regions were defined with a ∼20-μm spacing (MDA-8000B). After development of the photoresist, Ni (10 nm)/Au (100 nm) electrodes were deposited by e-beam evaporation, followed by lift-off with acetone. The fabricated metal oxide sensors were mounted on a sensing chamber, which was designed to measure resistance signals using a data acquisition module (34970A, Agilent Technologies) in a tube furnace. The sealed gas sensing chamber is approximately 5 cm in diameter and 20 cm in length. A gas delivery system built in-house was used to control the gas flow into the sensing chamber to measure the sensor response under ambient pressure. The test gas used in this study contained 100 ppm of ethanol (balance air). The serial dilution system used to control the concentration consisted of a mass flow controller, Teflon-PFA tubing (1/4″), lock-type fitting, and a valve system. Air was used as the reference gas, and the total flow rate for the reference gas and tested ethanol was fixed at 500 cm3/min. The sensing tests were performed at 350°C.