Letter pubs.acs.org/NanoLett
Self Assembly of Nanoislands on YSZ-(001) Surface: A Mechanistic Approach Toward a Robust Process Haris M. Ansari,† Vikas Dixit,† Lawrence B. Zimmerman,‡ Michael D. Rauscher,† Suliman A. Dregia,† and Sheikh A. Akbar*,† †
Department of Materials Science and Engineering and ‡Department of Chemical and Biomolecular Engineering, The Ohio State University, Columbus, Ohio 43210, United States ABSTRACT: We experimentally investigate the mechanism of formation of self-assembled arrays of nanoislands surrounding dopant sources on the (001) surface of yttriastabilized zirconia. Initially, we used lithographically defined thin-film patches of gadolinia-doped ceria (GDC) as dopant sources. During annealing at approximately one-half the melting temperature of zirconia, surface diffusion of dopants leads to the breakup of the surface around the source, creating arrays of epitaxial nanoislands with a characteristic size (∼100 nm) and alignment along elastically compliant directions, ⟨110⟩. The breakup relieves elastic strain energy at the expense of increasing surface energy. On the basis of understanding the mechanism of island formation, we introduce a simple and versatile powder-based doping process for spontaneous surface patterning. The new process bypasses lithography and conventional vaporsource doping, opening the door to spontaneous surface patterning of functional ceramics and other refractory materials. In addition to using GDC solid-solution powders, we demonstrate the effectiveness of the process in another system based on Eu2O3. KEYWORDS: Self-assembly, nanoislands, self-patterning, nanoimprinting, YSZ, GDC
S
island arrays were also found far from the remnant patches, and it was proposed that film delamination left the exposed surface with a modified composition due to implantation mixing during the initial growth of the film. The remnant patches of GDC were thought to be under stress, which provided the driving force for initiation of island formation on their margins. Previous work on nanoisland arrays in the GDC/YSZ system leads to several questions about the formation mechanism. First, do the GDC patches act as sources of stress, or sources of solute diffusion, or both? Second, is a premodified YSZ surface required for the formation of self-assembled nanoislands? Finally, do the islands form during the isothermal hightemperature hold or upon subsequent cooling? Answering the latter question would allow us to differentiate between two possible mechanisms, namely, interdiffusion with a concurrent morphological instability versus a two-step mechanism of dopant dissolution at high temperature followed by surface precipitation of islands on cooling. Here, we present new results that shed light on the mechanism of nanoisland formation in the GDC/YSZ system and its extension toward a highly simplified powder-based process. Single crystal 8 mol % YSZ-(001) substrates were purchased from MTI Corporation (Richmond, CA). The substrates were 5.0 mm × 5.0 mm × 0.5 mm in dimensions,
elf-assembly of nanostructures on surfaces is a well-known phenomenon in the vapor deposition of lattice-mismatched semiconductor thin films, for example, Ge on Si, where nanoislands spontaneously nucleate on the substrate surface and exhibit order in their size, shape, separation, and spatial alignment.1,2 The islands form due to a strain-energy driven instability similar to the Asaro−Tiller−Grinfeld (ATG) instability.3−5 A high degree of control over array characteristics (island size uniformity and ordered spatial distribution) can be achieved by depositing multilayers,6,7 and by depositing a film on a regular template, such as surfaces modified by periodic strain fields from buried dislocation arrays8 or lithographically predefined mesas.9,10 Surfaces with self-assembled nanostructures can be used for a wide range of applications such as stamps to transfer nanopatterns to polymers by soft imprinting techniques,11 as templates for depositing nanostructured magnetic thin films,12 and for protein adsorption studies,13 to name a few. Recently, Rauscher et al.14 have discovered that self-assembled nanoislands can also be produced by solid-state processing in the gadolinia-doped ceria (GDC)/yttria-stabilized zirconia (YSZ) system without lithography. They observed nanoisland arrays on the surfaces that had been exposed by partial delamination of a GDC film (sputtered at room temperature) on YSZ-(001) substrates, upon annealing at 1150 °C. The nanoisland arrays formed preferentially on the margins of unspalled patches of the GDC on the YSZ surface, and there was remarkable alignment of island rows along the elastically compliant ⟨110⟩ directions of the YSZ surface. Some © XXXX American Chemical Society
Received: February 9, 2013 Revised: March 25, 2013
A
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 1. (a) Nanoisland arrays correlated with lithographically defined GDC patches, (inset, a trench is visible around two trails of nanoislands), (b) bright-field TEM image of a typical nanoisland (inset, a SAD pattern with a ⟨110⟩ zone axis), (c) EDS spectrum obtained at the point marked by the white dot in (b), (d) elemental profiles obtained along the line in (b), and (e) elemental profiles obtained across the GDC patch/YSZ substrate interface.
in acetone, the samples were air-dried then annealed at 1125 °C for 4−36 h in a box furnace in air atmosphere. For the powder suspension method, we used 0.1−0.3 g/L suspensions of 5 and 10 mol % GDC (Nextech Materials, Columbus, OH) and similarly concentrated suspensions of pure Eu2O3 (Alfa Aesar, Ward Hill, MA) powders in distilled, deionized water. The suspensions were ultrasonicated for 5 min to yield a milky suspension and then applied to the YSZ substrate surface via an eyedropper. The droplet size was controlled such that the substrate was immersed completely in it and was allowed to dry in air in order to leave powder particles dispersed on the YSZ surface. The samples were heat treated at 1000−1150 °C for 0−25 h in air. Scanning electron microscopy was performed on FEI Sirion FEG−SEM (Hillsboro OR, U.S.A.). Transmission
and as supplied they were chemically polished to reduce the surface roughness below 5 Å. In order to mimic the unspalled GDC remnants in a more controlled manner, we used liftoff photolithography to pattern RF-sputtered GDC films on YSZ(001) substrates (60 W, 4 h, 5 mTorr Argon atmosphere, using a Discovery-18 DC/RF magnetron sputter deposition system (Denton Vacuum, Moorestown, NJ)). The resulting samples consisted of arrays of large (200 × 200 μm) GDC patches separated by YSZ surfaces that had been shielded by photoresist during deposition. The sputtering target was made from 5 mol % GDC powder (purchased from Nextech Materials, Columbus, OH) by Sputtering Target Manufacturing Co., LLC (Westerville OH, which has since been purchased by Kurt J. Lesker Company). After photoresist removal by dipping B
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 2. (a) SEM micrograph of a powder particle acting as a source of GDC to produce nanoislands in its proximity (annealed at 1100 °C for 5 h with 10 °C/min heating and 1 °C/min cooling rates), and (b) small powder particles with good contact with the substrate surface almost totally consumed (held at 1150 °C for 5 h with 5 °C/min heating rate and water quenched).
rapidly than they did in the depth direction, by about 3 orders of magnitude. This disparity in diffusion distances suggests that surface diffusion is much faster than bulk diffusion at this temperature, and this explains lateral uniformity of island composition versus the nonuniformity in the depth direction. As an island mounds up by local surface diffusion, it is also receiving dopant from the distant source, and it is useful to think of it as a stack of layers added sequentially at different points in the transient diffusion process. Rapid lateral surface diffusion homogenizes the instantaneous top layer of a growing island, but once a layer is buried its composition is essentially preserved because it can only be altered by the much slower bulk diffusion mechanism. Thus, the disparity between lateral versus vertical composition uniformity is caused by the large disparity between surface versus bulk diffusion. Finally, at a given distance from the source the supply of diffusing dopants is expected to increase with time, and this might also explain why the island top, which is formed later in the process, is richer in dopants than the bottom that is buried early in the mounding process and shielded from later enrichment. The correlation of the nanoislands with the GDC patches suggests that there is something special about the surface near the patches to initiate nanoisland formation. Boyne et al.18 have shown through a Phase Field simulation model that surface structures, such as the GDC patches in our work, can cause extensive stress-induced sequential formation of islands on the surface next to them, but in order to sustain island formation far away from the patch, the surface must have a local source of stress, such as a chemical inhomogeneity caused by doping; the stresses in the patch itself could not be active over a long-range. This modeling result narrowed the possibilities for the role of the GDC patches (and hence the GDC film remnants in Rauscher’s work), namely, that they were predominantly acting as a source of dopant on the YSZ surface, and their own stress state may be important in initiating the breakup but not in sustaining it at long-range. Therefore, the islands were expected to form regardless of how the GDC source is applied on the substrate surface. In order to investigate this hypothesis, we developed a very simple process whereby the source material can be dispersed as a solid powder on the surface. An aqueous suspension of 5 mol % GDC powder was prepared and applied to the sample surface via an eyedropper. This powder was from the same batch that had been used to prepare the sputtering target for lithographic samples. The samples were dried in air, leaving powder particles on the YSZ surface. The dry samples were heat treated in air between 1000−1150 °C in a high-
electron microscopy (TEM) samples were prepared by focused ion beam (FIB, FEI Helios, FEI Company, Hillsboro OR) and imaging was performed on Phillips CM200 TEM (Phillips Eindhoven, Holland), wheras energy dispersive spectroscopy (EDS) analysis was performed on FEI Tecnai F-20 STEM (FEI Company, Hillsboro OR). Figure 1a is an SEM micrograph of a lithographically patterned sample after annealing for 25 h at 1125 °C in air (with 10 °C/min heating and cooling rates). The nanoislands are found around the edges of the GDC patch, and they align along the ⟨110⟩ directions of the substrate. It is also noteworthy that the YSZ surface between the GDC patches had been masked by photoresist during film deposition, and therefore had no contact with GDC prior to annealing. Figure 1b shows a bright-field TEM image of a typical island cross-section, including a diffraction pattern with a ⟨110⟩ zone axis, which confirms the parallel epitaxy of the islands reported by Rauscher et al. Figure 1c shows the EDS spectrum obtained from the middle of the island shown in Figure 1b (location marked with a dot) confirming that the islands are indeed solid solutions of GDC and YSZ, rich in YSZ content. The transport of YSZ into the islands is also revealed in the SEM image inset in Figure 1a, where a trench around two trails of nanoislands is visible, suggesting that material was drawn up from the substrate when the islands formed. However, an EDS line scan obtained along the line marked in Figure 1b shows that the island composition is nonuniform (Figure 1d). The Gd and Ce content starts at a maximum near the top of the island and decreases with depth, tapering off to zero in the substrate. On the other hand, lateral EDS line scans, obtained across the island width, show a flat composition profile. Since Gd3+ and Ce4+ have larger ionic radii15,16 as compared to Zr4+,17 assuming island formation via a strain-based mechanism, it is plausible that the species causing strain (Gd3+ and Ce4+) lie as far away from the island/substrate interface as possible and closer to the free surface where some strain relaxation is feasible in order to lower the strain energy. Although dopant enrichment near the surface could explain the decaying depth profiles in Figure 1d, it fails to explain the lateral uniformity of composition profiles through the island, which leads us to consider the relative rates of diffusion on the surface versus through the depth. An EDS depth scan obtained across the GDC patch/YSZ interface shows that the Ce4+ and Gd3+ diffusion penetration across this interface is ∼10 nm (Figure 1e). On the other hand, the margin of nanoislands formed in the same time around the GDC patch is ∼2 μm wide, which means the dopants from the GDC have diffused laterally more C
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 3. (a) SEM image of cracks in the sample due to thermal shock experienced during water quench (different area on the sample shown in Figure 2b, and (b) a quench crack splits a nanoisland colony in half showing that the colony formed during the soak before the crack propagated through it.
temperature furnace, using 10 °C/min heating and 1 °C/min cooling rates. Figure 2a shows an SEM micrograph of a typical sample after a 5 h anneal at 1100 °C. Another powder suspension was prepared in distilled water using 10 mol % GDC powder. Figure 2b shows an SEM micrograph of a typical sample after a 5 h soak at 1150 °C, followed by quenching in water. It is evident from Figure 2 that both 5 and 10 mol % GDC powder particles (brightest contrast) serve a function similar to the patches in the lithography samples as dopant sources. Also, the powder-based process is versatile for testing various powder compositions as sources for nanoisland formation in an inexpensive manner. It must also be noted that the powder particles only have physical contact with the substrate surface prior to annealing, and this contact is insufficient to load the surface with stress as does a sputtered thin film. Although there might be a different stress state near the GDC film remnants and the lithographically defined patches, and that might be catalyzing nanoisland formation, it is unlikely that the same stress state would be achieved around irregularly shaped powder particles. Thus, stresses in the dopant source maybe playing a role locally, but they are not required. Furthermore, experiments with powder confirm that a modified YSZ surface is not required prior to annealing and that this modification can happen during annealing. The area coverage of the nanoisland array depends on the area coverage of the GDC powder particles, their distribution and contact with the substrate surface. Smaller particles with good contact might disappear completely during annealing and leave a footprint of large islands. The island arrays start off from around the powder particles, and the pattern spreads outward as diffusion continues. Since there are numerous dopant sources (corresponding to each powder particle), the periodicity is affected when the island growth fronts from two different sources impinge on one another. It is pertinent to mention here that we are only presenting proof of principle in this communication and not ways to optimize and control island periodicity and area coverage. Because the process relies on surface diffusion of the dopants on the substrate, there exist multiple combinations of time and temperature for nanoisland formation. Annealing at lower temperatures for longer times may result in island morphology (and area coverage) similar to one obtained at higher temperatures and shorter times, although bulk diffusion might become significant at higher temperatures. We have observed island formation for temperatures as low as a 1000 °C for soak times of 5−25 h. Generally, we have found that islands formed at lower soak temperatures are, on average, smaller in size as compared to those formed at higher temperatures for
the same soak time. Also, the width of the margin (around a particular source of GDC) converted into nanoisland arrays increases as the soak time increases at a particular temperature. Fast ramps to the desired temperature and furnace cooling without any soak showed very limited extent of nanoisland formation suggesting that the islands form either during or after the soak but not during the ramp up. Thus far, we have established that the GDC patches or powder particles primarily act as dopant sources. Another important aspect of the mechanism is determining whether the islands form during the high-temperature hold or on subsequent cooling, as these possibilities would imply two different mechanisms. Formation at the soak temperature would imply morphological transformations are induced by the dissolution of dopants, whereas formation on cooling would imply precipitation of dopant-rich islands from a solution formed at the soak temperature. It was necessary to study the effect of cooling rate on island formation and to test whether the transformation could be suppressed by rapid quenching from the soak temperature. For quenching experiments, the samples were placed on an alumina platform that was hung by a platinum wire in a vertical tube furnace. The samples used for this purpose were prepared via the powder suspension method. A control sample was given a standard anneal in the vertical furnace at 1150 °C for 5 h in air, and it revealed nanoisland arrays similar to those observed in box-furnace treated sample, for example, as shown in Figure 2. For the quench, a sample was held at 1150 °C (5 °C/min heating rate) for 5 h and then the platinum wire was cut to allow the sample (along with the alumina platform) to drop into a beaker of chilled water (10 °C). The sample was taken out within a minute and dried in air. A typical sample had dimensions of approximately 2.5 mm × 2.5 mm × 0.5 mm. We estimate the cooling rate to be ∼18 °C/ s which is comparable with cooling rates achieved in quenching of steel.19 SEM revealed that the surface of the quenched sample was covered with nanoislands similar to those observed previously on the slowly cooled samples (Figure 2b). A striking feature of the quenched samples is the quench cracking due to thermal shock (Figure 3). These cracked samples remained intact for handling. The cracks were mostly aligned along the ⟨110⟩ directions of the substrate surface and provided evidence that the island arrays formed during the soak and not on subsequent cooling. In Figure 3b, we show an example of a crack that runs through an existing island colony and bisects some of the islands, which could only have happened had the islands already been present on the surface during the soak, before the quench cracking. D
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
although changing the substrate is another possibility, we used the same substrate to exploit its known elastic anisotropy. Nanoislands in the YSZ-(001) form via a strain-based mechanism similar to the ATG instability, whereby the stress accumulated in the doped surface of the YSZ substrate is relieved by creation of self-assembled nanoislands. A premodified surface layer is not required prior to annealing, that is, this modification can occur during annealing by surface diffusion of dopants from the solid dopant sources with simultaneous breakup, which occurs at the hold temperature independent of the subsequent cooling. On the basis of this understanding, we introduce a very simple powder based process of producing nanoislands which bypasses lithography and thin film deposition processes. Experiments with the powder based process confirm that stress that might be present near the GDC remnant patches in Rauscher’s work14 and lithographically defined patches in the current work is not required to catalyze the island formation process. The powderbased process offers an easy and versatile way of testing various powders and powder compositions for potential to nanoisland formation on YSZ. Overall, the type of self-assembly we demonstrate for functional ceramics in this paper is similar in its underlying physical phenomena to ordering of semiconductor islands on 2D surfaces,1 arrays of nickel aluminide precipitates in 3D Ni-base superalloys22 and 3D semiconductor island arrays in multilayer systems.23
The evidence presented supports the conclusion that the islands form via a strain based mechanism such as the one applicable in semiconductor heteroepitaxial systems. In those systems, the surface layer/film is created from a vapor source while in the case of GDC/YSZ, this surface layer is created via surface diffusion from solid sources dispersed on the YSZ surface. Clearly, the solid-source process has an advantage in patterning functional ceramics with multicomponent dopants that cannot readily be prepared in the form of vapor. Stress buildup as the doping level of the surface layer increases causes the surface to break up into nanoislands (via surface diffusion) to relieve stress by a mechanism similar to the ATG instability. Here, the instability is more localized at an advancing front because of concentration gradients; that is, doping levels are higher in regions closer to the dopant source than far away from them. It is a dynamic process and happens first near the GDC sources and spreads outward as time progresses, or in other words, as more and more area is doped by diffusion. Doping and breakup occur simultaneously during the soak, independent of the subsequent cooling. A typical island is 50− 75 nm tall. Once an island is nucleated and is morphologically evolving by mounding up of material from the doped surface layer, a trench is dug around it. The remarkable alignment of the islands along the ⟨110⟩ directions of the substrate comes from the elastic modulus anisotropy of the YSZ substrate and island−island elastic interactions, as previously suggested by Rauscher et al.14 The ease of making these patterned substrates and the fact that the process can be scaled up to cover large surfaces make them amenable to be used as master patterns for nanoimprinting to soft polymers such as polydimethylsiloxane (PDMS), ethyleneglycol dimethacrylate (EGDMA), and polystyrene (PS).11 The patterned substrates can also be used directly for cell attachment and proliferation studies.20 To demonstrate the versatility of the powder-based process and to show that nanoisland formation on YSZ-(001) surfaces is a general phenomenon not limited to GDC as dopant, we repeated the process successfully with many other oxide21 dopants, and a manuscript is in preparation for a future publication. Here, we show an example of Eu2O3 as dopant that upon annealing at 1125 °C for 25 h revealed self-assembled nanoisland arrays (on YSZ-(001) surface) similar to the ones observed in the GDC/YSZ system (Figure 4). It should be noted that changing the dopant requires some process optimization in terms of the soak time and annealing temperature, which provides flexibility for controlling average size, shape, and areal density of the nanoislands. Moreover,
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Tel: +1-614-292-6725 Fax: +1-614292-1537. Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors are indebted to Joe Atria for help with quenching experiments and Orton Ceramic Foundation (Westerville, Ohio, U.S.A.) for their generosity in allowing use of their facilities. This work was supported in part by the OSU Institute for Materials Research (IMR) and the Fulbright − HEC Pakistan Scholarship Program. S.A.D. dedicates this paper to his uncle Mohamed S. Dregia, who had the creative vision of selfreplicating ceramic tiles over 80 years ago.
■
REFERENCES
(1) Eaglesham, D. J.; Cerullo, M. Phys. Rev. Lett. 1990, 64, 1943− 1946. (2) Mo, Y.-W.; Savage, D. E.; Swartzentruber, B. S.; Lagally, M. G. Phys. Rev. Lett. 1990, 65, 1020−1023. (3) Asaro, R. J.; Tiller, W. A. Metall. Trans. 1972, 3, 1789−1796. (4) Grinfeld, M. A. Phys.-Dokl. 1986, 31, 831−834. (5) Srolovitz, D. J. Acta Metall. 1989, 37, 621−625. (6) Woll, A. R.; Rugheimer, P.; Lagally, M. G. Mater. Sci. Eng., B 2002, 96, 94−101. (7) Pinczolits, M.; Springholz, G.; Bauer, G. Phys. Rev. B 1999, 60, 11524−11529. (8) Voigtlander, B.; Kawamura, M.; Paul, N.; Cherepanov, V. J. Phys.: Condens. Matter 2004, 16, S1535−S1551.
Figure 4. SEM micrograph of self-assembled nanoislands with Eu2O3 as dopant source on YSZ-(001) substrate surface (annealed at 1125 °C for 25 h with 10 °C/min heating and cooling rates). E
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
(9) Kamins, T. I.; Williams, R. S. Appl. Phys. Lett. 1997, 71, 1201− 1203. (10) Jin, G.; Wan, J.; Luo, Y. H.; Liu, J. L.; Wang, K. L. J. Cryst. Growth 2001, 227−228, 1100−1105. (11) Zimmerman, L. B.; Rauscher, M. D.; Ellis, J.; Boukany, P.; Lee, L. J. Nanotechnology 2010, 21, 045304. (12) Teichert, C. Appl. Phys. A 2003, 76, 653−664. (13) Riedel, M.; Muller, B.; Wintermantel, E. Biomaterials 2001, 22, 2307−2316. (14) Rauscher, M. D.; Boyne, A.; Dregia, S. A.; Akbar, S. A. Adv. Mater. 2008, 20, 1699−1705. (15) Taylor, M. A.; Argirusis, C.; Kilo, M.; Borchardt, G.; Luther, K.D.; Assmus, W. Solid State Ionics 2004, 173, 51−56. (16) Bekale, V. M.; Legros, C.; Sattonnay, G.; Huntz, A. M.; Lesage, B.; Argirusis, C.; Jomard, F. Defect Diffus. Forum 2006, 258−260, 46− 51. (17) Kilo, M.; Taylor, M. A.; Argirusis, C.; Borchardt, G.; Weber, S.; Scherrer, H.; Jackson, R. A. J. Chem. Phys. 2004, 121, 5482−5487. (18) Boyne, A.; Rauscher, M. D.; Dregia, S. A.; Wang, Y. Scr. Mater. 2011, 64, 705−708. (19) Fernandes, F. M. B.; Denis, S.; Simon, A. Mater. Sci. Technol. 1985, 1, 838−844. (20) Parikh, K. S.; Rao, S. S.; Ansari, H. M.; Zimmerman, L. B.; Lee, L. J.; Akbar, S. A.; Winter, J. O. Mater. Sci. Eng., C 2012, 32, 2469− 2475. (21) Ansari, H. M. Ph.D. Dissertation, The Ohio State University, Columbus, OH, 2012 (22) Khachaturyan, A. G. Theory of Structural Transformations in Solids; Wiley: New York, 1983. (23) Springholz, G.; Holy, V.; Pinczolits, M.; Bauer, G. Science 1998, 282, 734−737.
F
dx.doi.org/10.1021/nl4005282 | Nano Lett. XXXX, XXX, XXX−XXX
