NANO LETTERS
Ordered Mesoporous Silicon through Magnesium Reduction of Polymer Templated Silica Thin Films
2008 Vol. 8, No. 9 3075-3079
Erik K. Richman, Chris B. Kang, Torsten Brezesinski, and Sarah H. Tolbert* Department of Chemistry and Biochemistry, and the California NanoSystems Institute, UCLA Los Angeles, California 90095-1569 Received June 18, 2008; Revised Manuscript Received July 18, 2008
ABSTRACT This paper describes the process of making ordered mesoporous silicon (Si) thin films. The process begins with mesoporous silica (SiO2) thin films that are produced via evaporation induced self-assembly (EISA) using sol-gel silica precursors with a diblock copolymer template. This results in a film with a cubic lattice of 15 nm diameter pores and 10 nm thick walls. The silicon is produced through reduction of the silica thin films in a magnesium (Mg) vapor at 675 °C. Magnesium reduction preserves the ordered pore-solid architecture but replaces the dense silica walls with 10-17 nm silicon crystallites. The resulting porous silicon films are characterized by a combination of low and high angle X-ray diffraction, combined with direct SEM imaging. The result is a straightforward route to the production of ordered nanoporous silicon.
Porous silicon is a technologically significant material with a variety of applications in sensors, optics, and biocatalysts that result from its unique combination of physical and electrical properties.1-7 The first production of porous silicon dates back to 1956 as an accidental product in a substrate cleaning process.8 Using anodic etching, the usual route to mesoporous silicon is to etch under a significant bias (50-100 V).9 However, lower bias can also be used with the addition of oxidants.10-12 Tuning conditions for the etch process can produce an ordered structure on the macroporous scale (>50 nm) but fails as smaller structures are targeted.13,14 Pore formation is limited to current densities below a critical value depending on the solution, while porosity generally increases with current density. In the mesopore scale (2-50 nm), location is random and close packing is lacking due to the tendency for branching during the etching process.9,15 In the mesopore regime, porous inorganic materials can readily be produced by surfactant templating methods, resulting in materials with close packed porosity and narrow pore size distributions. The first templated silica materials were formed in powder form and introduced by Yanagisawa et al. of Japan in 1990 and by Kresge et al. of Mobil Corporation in 1992.16,17 Mesoporous silica can now be made with a variety of structural symmetries including cubic packing of spheres, 2D hexagonally packed cylinders, and lamellar phases.18-22 Templated versions of many other mesoporous metal oxides have been made,23-27 yet despite potential interesting physical and electrical characteristics, * To whom correspondence should be addressed. E-mail: tolbert@ chem.ucla.edu. 10.1021/nl801759x CCC: $40.75 Published on Web 08/15/2008
2008 American Chemical Society
there exists few non-oxide polymer or surfactant templated mesoporous materials.28-30 The initial work on powders was followed by the formation of thin films using evaporation induced self-assembly to generate mesoporous thin films with pores sizes in the 2-30 nm range and film thicknesses ranging from 15 to 500 nm by simple solution phase routes.25,26,31-33 These films demonstrate tunable materials properties, symmetry, and pore size.34 Such films can be arbitrarily stacked with independently determined porosity for more complex structures or for control over pore orientation.35,36 For a variety of applications, ranging from solar cells to electronics, it is desirable to translate this flexible solution processing approach from silica to silicon. Brute force reduction of silica has generally required high temperature processing with structurally destructive results.37-39 A promising method for reprocessing silica without losing its structure via magnesiothermic reduction, however, has recently been published by Sandhage et al.40 Magnesium vapor in the presence of silica produces the reaction in eq 1 2Mg(g) + SiO2(s) f 2MgO(s) + Si(s)
(1)
This reaction is spontaneous at temperatures above the melting point of magnesium.41 The magnesia can be removed quickly with an acid solution to leave only trace silica and structured silicon. Further washing with HF eliminates the remaining silica. Recent application of this method to a diatom system effectively reproduced the three-dimensional ∼10 µm sized bulk structures in silicon; approximately, 100 nm pores were retained, and the final structure was composed of ∼13 nm silicon crystallites.40
The 13 nm crystallites imply a possible lower size limit for the structures that can be retained upon magnesium reduction. Indeed, we have found that, when mesostructured thin films with 13-15 nm periodic repeat distances and 3-5 nm thick walls were reduced,33 the nanoscale structure was not well preserved and in most cases dense films were formed. In this work, we show that, by moving to a templated silica structure with a larger repeat distance and thicker wall, the nanoscale structure can be retained upon reduction. We produce porous cubic silica precursor films with a 25 nm pitch and 10 nm thick walls using the large diblock copolymer KLE (H(CH2CH2CH2(CH)CH2CH3)89(OCH2CH2)79OH).42Upon reduction, conversion of the silica structure to silicon is possible with preservation of the mesoscale porosity and much of the nanoscale periodicity. Experimental Section. Synthesis of the Mesoporous Silica Framework. Synthesis of the silica framework is accomplished using a 0.23:2:10:1 by mass mixture of KLE, TEOS, ethanol, and 0.01 M HCl. First, 0.07 g of KLE was dissolved in 2 mL of ethanol while 600 mg of TEOS and 300 mg of 0.01 M HCl were mixed with 2 mL of ethanol in a separate container. The solutions were combined and stirred for 1 h; the solution was then left to age for 1 day. Films were dip coated from the aged solution onto silicon wafers in a humidity controlled chamber set to 30% relative humidity. The withdrawal rate was 2 cm/min. The films were dried overnight and then heated to 60 °C for 24 h. The films were calcined at 450 °C in air for 1 h using a 1 °C/min heating ramp. This produced a 15 nm diameter pore with a pore-to-pore spacing of 25 nm. Reduction of Silica to Silicon. A 1 cm2 film was placed into a brass chamber, which has an inner volume of 5 cm3, and 5 mg of Mg was placed ∼1 cm away from a film. The chamber was then sealed in a glovebox with an argon atmosphere. To generate Mg vapor, the chamber was heated above the melting point of Mg in an oven in a glovebox. Temperature was ramped to 675 °C in 2 h and soaked for 5 h. To remove magnesia, the film was immersed in 1 M HCl for 10 min. Finally, to remove any unreacted silica, the film was immersed in 0.5% HF for 10 min. Sample Characterization. Low angle 1-D XRD, high angle XRD, 2-D SAXS, SEM, and EDAX measurements were taken to verify sample structure and elemental composition. Low angle XRD was measured on a Panalytical X’Pert PRO instrument; the sample was sealed under argon using a polyester film. High angle XRD was measured using a Bruker AXS D8 Discover instrument; here the sample was sealed under argon using a kapton film. Two-dimensional SAXS data were collected on beamline 1-4 at the Stanford Synchrotron Radiation Laboratory. SEM was obtained using a JEOL 6700F with an EDAX attachment; the sample was loaded within a minute to minimize oxygen exposure. Results and Discussion. To verify the success of the reaction, we use high angle X-ray diffraction (XRD) to examine the magnesium treated films. The substrate is oriented silicon, which does not produce peaks in our θ-θ reflection geometry (Figure 1 is background subtracted; substrate not shown). Any silicon peaks are from 3076
Figure 1. High angle XRD. Part (a) shows a film after magnesiothermic treatment. Peaks from both silicon and magnesia can be seen. Part (b) shows a film after magnesiothermic treatment and an HCl wash. Finally, part (c) shows a film after magnesiothermic treatment and washes with both HCl and HF. As in the (b) pattern, all diffraction peaks can be indexed to diamond structure silicon (according to JCPDS reference card #27-1402).
polyoriented crystallites only. Figure 1a shows the high angle XRD pattern after treatment with magnesium vapor. We see peaks for both magnesia and silicon. After washing with HCl, Figure 1b shows only peaks from silicon. Finally, after the wash with HF, we continue to see silicon peaks in Figure 1c, and the film remains visibly intact. Scherrer analysis43 of the domain size from the diffraction peak width indicates that the crystallite diameter is ∼13 nm, in good agreement with the Sandhage observations.40 This size is slightly larger than the original silica wall and implies either elongated grains within the walls or some restructuring. The Scherrer width is strongly dominated by the largest contributors, however, so the repositioning of the wall lattice is not necessarily drastic. Interestingly, the repeated appearance of 13 nm silicon domains in materials generated from silica structures with very different size structures suggests that this may be the critical nucleation size for silicon under these thermal and chemical conditions. Low angle one-dimensional X-ray diffraction (1D-XRD) shows that the mesostructure is preserved upon magnesium reduction, but some contraction is observed due to the thermal processing. Figure 2a shows the original diffraction of the KLE-based silica film with a cubic lattice. These data were collected on the mesoporous film after calcination to 450 °C. Processing the film to 675 °C, as shown in Figure 2b, results in contraction but otherwise preserves the order of the film. Figure 2c shows 1D-XRD of the film after exposure to Mg vapor at 675 °C in an otherwise inert atmosphere. The nanoscale periodicity is clearly retained after the reduction process. Though some degradation of diffraction is observed, the structure certainly remains after removal of the MgO via washing with HCl, as shown in Figure 2d. Finally, the film is also observed to retain diffraction after a wash with HF to remove remnant silica (Figure 2e). The final material is pure polysilicon with mesostructured pores. Nano Lett., Vol. 8, No. 9, 2008
Figure 2. One-dimensional low angle XRD. Part (a) shows a film before magnesiothermic treatment (calcined at 450 °C in air). Part (b) shows a film that was heated to the Mg reaction temperature (675 °C) in air. The result of this thermal treatment is some peak broadening and a marked shift to higher angle. Part (c) shows a film after magnesiothermic treatment, while (d) shows the reduced films after an HCl wash. Finally, part (d) shows low angle XRD after the final HF wash to remove unreacted silica. For all samples, clear low angle diffraction peaks and overtones can be observed, indicating the preservation of nanometer scale periodicity.
Figure 3. Two-dimensional small-angle X-ray scattering. The patterns were taken at an angle of incidence of 8°. Here the scattering vector s is given in nm-1 and defined as s ) 2/λ sin θ. Part (a) shows a KLE-templated silica thin film after thermal treatment at 450 °C. Part (b) shows the same film after exposure to Mg vapor at 675 °C. Part (c) shows mesoporous polysilicon obtained after subsequent HCI and HF washes of the film used in (b). Treatment by Mg vapor, acid, and HF clearly leaves behind an ordered mesoporous silicon film.
One-dimensional XRD results are further supported by two-dimensional small-angle X-ray scattering (2D-SAXS) data. Figure 3 shows 2D-SAXS patterns of a KLE-templated SiO2 film before (Figure 3a) and after (Figure 3b) Mg treatment as well as of a purified polysilicon film after acid and HF washing (Figure 3c). For small angles of incidence β (defined as the angle between the X-ray beam and the plane of the substrate), the films produce patterns with distinct maxima, which can be indexed in terms of a face-centered cubic system with (111) orientation relative to the plane of the substrate. One can clearly see that the mesoscale order is retained after both Mg treatment and removal of unreacted silica. However, the process leads to some restructuring as shown in the broadened peaks in Figure 3b and c. The scattering patterns further indicate unidirectional lattice contraction perpendicular to the plane of the substrate. This Nano Lett., Vol. 8, No. 9, 2008
Figure 4. SEM characterization of the various mesoporous films used in this work. Part (a) shows a top view of a film before magnesiothermic treatment. The inset shows a 2-D Fourier transform of the image, indicating the highly periodic hexagonal nature of the top surface. Part (b) shows a top view image of a pure silica film after magnesiothermic treatment and washing with both HCl and HF. While the structure is not as clean as the one shown in part (a), the 2-D FFT (inset) again shows clear hexagonal ordering of the pores. Part (c) shows an SEM image of a cut in a film similar to that shown in (b). Porosity within the film is clearly visible. Finally, EDAX characterization of silicon films before and after HF treatment is presented in part (d). A small amount of oxygen is present in the films before HF treatment, which is almost completely removed after the HF wash, indicating that any residual silica can be efficiently removed.
result underscores the power of substrate stabilization in maintaining nanometer scale periodicity through a transformation process like thissthe out-of-plane scattering is almost gone in the final silicon material, but the in-plane maxima remain almost unchanged. On the basis of the relative position of the out-of-plane reflections, a contraction of about 65% is determined for the pure polysilicon films. Patterns taken in transmission mode (i.e., at β ) 90°, not shown) produce isotropic diffraction rings, which confirm the presence of a random distribution of mesostructured domain orientations within the plane of the film. Figure 4a shows an SEM of the original film before magnesium treatment. The high quality spherical lattice is characteristic of oxide phases templated with the diblock copolymer KLE.42 The inset 2-D fast Fourier transform emphasizes the symmetry of the ordered surface. After reduction by magnesium and the subsequent HCl and HF washes, the surface retains the continuous lattice as seen by the inset in Figure 4b. The pores are lower quality but definitely retain the basic shape enforced by the template during the original solution phase assembly process. Figure 4c shows a razor blade cut in the reduced film. The top shows retention of ordered porosity, while the side shows pores of similar size to the original template. The debris field has components of 10-20 nm crystallites, which corresponds well with the crystallite size determined from analysis of the high angle XRD. We note that, while the porosity in the silicon film looks less ordered than the original silica film, the sample still diffracts quite well, suggesting that the 3077
disorder may be due to added porosity caused by the loss of oxygen. The matter that remains, however, apparently sits on a subset of lattice sites very similar to those found in the original silica-based material. It is interesting to note that the film has changed mechanically upon reduction. Normally cubic porous silica or titania films of the same thickness tend to tear or smash when subjected to a razor blade. Most of the time, the tearing allows for clean views of the film in cross section. These films showed almost exclusively debris fields as though a more brittle film had crumbled under the razor blade. Clean cuts are nearly nonexistent, making cross-sectional SEM a dubious venture. Added to this, the framework has a crystalline granularity not present in amorphous films. Despite this fact, the basic structure remainssthe polymer template has clearly left its imprint on the final structure. Figure 4d shows EDAX analysis of the film. After the magnesium treatment and HCl wash, there is some oxygen signature, but no magnesium. After the HF treatment, there is only a miniscule amount of oxygen; this small residual oxygen level is not surprising and is probably due to a small amount of surface oxidation caused by exposure to air during introduction into the sample chamber. Nevertheless, the key points are complete removal of magnesium using the acid wash and nearly complete removal of oxygen from using HF. The EDAX data present an interesting quandary. Compared to the work of the Sandhage group,41 we have used excess magnesium, longer reaction times, and higher temperatures, but unlike their work, we do not detect the formation of Mg2Si. One of the problems of magnesium reductions is that oxygen is not the only substance that can be removed during the reaction. In the presence of excess magnesium, silicon is consumed as well. 2Mg(s) + Si(s) f Mg2Si(s)
(2)
In our work, we are working with an equilibrium Mg vapor pressure of ∼5 Torr at 675 °C. This should be sufficient to drive the reaction to Mg2Si. However, the lattice of these templated films is under an additional strain because of adhesion to the substrate that is not present in powder samples. We believe this film strain reduces the rate of Mg2Si production. We note that Mg2Si is a less-than-useful side product, but the real structural loss likely comes from a Mg-Si eutectic liquid that forms and flows above 638 °C.44 This liquid can effectively solidify the film if it does not escape. Our results, and lack of magnesium in the final film, however, show the feasibility of using this reaction on ordered, porous thin films. Conclusions. We have demonstrated that a combination of the evaporation induced self-assembly and magnesiothermic reduction can be used to produce mesostructured silicon from an ordered silica framework with ∼15 nm pores and 10 nm thick walls. XRD shows both the formation of silicon crystallites and the retention of mesostructural order, even after complete removal of all MgO and unreacted SiO2. SEM illustrates the fine scale damage resulting from the reduction, but it also clearly illustrates that the periodicity produced by the KLE template is preserved in the final material. The 3078
silicon walls are composed of crystallites in the 10-17 nm range. The distortion to the lattice appears to be primarily a result of thermal contraction and not system collapse or phase change. The method appears to have a lower size limit for retention of structure set by the size of the silicon crystallites. This conclusion is based mostly on the fact that smaller templates that generate silica films with thinner walls did not produce films that could be consistently reduced under these reaction conditions. Potentially, this relatively modest temperature process can take advantage of the diverse chemical control developed for templated mesostructured thin film processing.36,35 On the most basic level, we open the way for new geometries for mesoporous silicon. The use of anodic etching to produce mesopores generates pits with branches but not close packed spherical pores.15 The geometry presented here is isotropic, three-dimensional, bicontinuous, and can be placed anywhere. Solution processing is effectively independent of the substrate properties, needing only a relatively smooth surface. Extending this system to a 2D hexagonal honeycomb geometry should make it possible to produce ordered anisotropic polysilicon frameworks with pores in the plane of the substrate.45 Extension of this reaction to other metal oxides presents another opportunity. On the basis of the Gibbs free energy, the formation of metals from metal oxides, in the presence of hot magnesium gas, is favored in many cases.46 Titanium oxide reduction has a free energy of -303 kJ/mol at 675 °C, while tin oxide reduction has a free energy of -627 kJ/ mol. Also, this method begs the investigation of cesium (Cs) reduction for less stable oxides as plastics like kapton might survive at the lower temperatures required to provide Cs vapor, leading to flexible templated porous metal films. Acknowledgment. This work was supported by the Petroleum Research Fund of the American Chemical Society under Grant ACS PRF# 46107-AC5, and by National Science Foundation under Grant CHE-0527015. Portions of this research were carried out at the Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. T.B. acknowledges the support of a DFG postdoctoral fellowship. References (1) Mizsei, J. Thin Solid Films 2007, 515, 8311. (2) Barratoa, C.; Faglia, G.; Sberveglieri, G.; Boarino, L.; Rossi, A. M.; Amato, G. Thin Solid Films 2001, 391, 261. (3) Weiss, S. M.; Fauchet, P. M. Phys. Status Solidi A 2003, 197, 556. (4) Watanabe, K.; Okada, T.; Choe, I.; Sato, Y. Sens. Actuators B 1996, 33, 194. (5) Bengtsson, M.; Ekstrom, S.; Drott, J.; Collins, A.; Csoregi, E.; MarkoVarga, G.; Laurel, T. Phys. Status Solidi 2000, 182, 495. (6) Pacholski, C.; Sartor, M.; Sailor, M. J.; Cunin, F.; Miskelly, G. M. J. Am. Chem. Soc. 2005, 127, 11636. (7) Lin, V. S.-Y.; Motesharei, K.; Dancil, K. P. S.; Sailor, M. J.; Ghadiri, M. R. Nature 1997, 278, 840. (8) Uhlir, A. Bell System Tech. J. 1956, 35, 333. (9) Lehmann, V.; Stengl, R.; Luigart, A. Mater. Sci. Eng. 2000, B69B70, 11. (10) Ogata, Y. H.; Koyama, A.; Harraz, F. A.; Salem, M. S.; Sakka, T. Electrochemistry 2007, 75, 270. (11) Bao, X. Q.; Jiao, J. W.; Wang, Y. L.; Na, K. W.; Choi, H. J. Electrochem. Soc. 2007, 154, D175. Nano Lett., Vol. 8, No. 9, 2008
(12) Ouyang, H.; Christophersen, M.; Fauchet, P. M. Phys. Status Solidi A 2005, 202, 1396. (13) Lehmann, V. J. Electrochem. Soc. 1993, 140, 2836. (14) Granitzer, P.; Rumpf, K.; Polt, P.; Reichmann, A.; Krenn, H. Physica E 2007, 38, 205. (15) Fo¨ll, H.; Christophersen, M.; Carstensen, J.; Hasse, G. Mater. Sci. Eng. R 2002, 39, 93. (16) Yanagisawa, T.; Shimizu, T.; Kuroda, K.; Kato, C. Bull. Chem. Soc. Jpn. 1990, 63, 988. (17) Kresge, C. T.; Leonowitz, M. E.; Roth, W. J.; Vartuli, J. C.; Beck, J. S. Nature 1992, 359, 710. (18) Mou, C. Y.; Lin, H. P. Pure Appl. Chem. 2000, 72, 137. (19) Fan, H.; Lu, Y.; Stump, A.; Reed, S. T.; Baer, T.; Schunk, R.; PerezLuna, V.; Lopez, G. P.; Brinker, C. J. Nature 2000, 405, 56. (20) Goltner, C. G.; Henke, S.; Weissenberger, M. C.; Antonietti, M. Angew. Chem., Int. Ed. 1998, 37, 613. (21) Ciesla, U.; Schuth, F. Microporous Mesoporous Mater. 1999, 27, 131. (22) Tate, M. P.; Eggiman, B. W.; Kowalski, J. D.; Hillhouse, H. W. Langmuir 2005, 21, 10112. (23) Brezesinski, T.; Groenewolt, M.; Antonietti, M.; Smarsly, B. Angew. Chem., Int. Ed. 2006, 45, 781. (24) Yang, P.; Zhao, D.; Margolese, D. I.; Chmelka, B. F.; Stucky, G. D. Nature 1998, 396, 152. (25) Alberius, P. C. A.; Frindell, K. L.; Hayward, R. C.; Kramer, J. E.; Stucky, G. D.; Schmelka, B. F. Chem. Mater. 2002, 14, 3284. (26) Mamak, M.; Coombs, N.; Ozin, G. AdV. Mater. 2000, 12, 198. (27) Brezesinski, T.; Groenewolt, M.; Gibaud, A.; Pinna, N.; Antonietti, M.; Smarsly, B. M. AdV. Mater. 2006, 18, 2260. (28) Sun, D.; Riley, A. E.; Cadby, A. J.; Richman, E. K.; Korlann, S.; Tolbert, S. H. Nature 2006, 441, 1126. (29) Armatas, G. S.; Kanatzidis, M. G. Science 2006, 313, 817. (30) Riley, A. E.; Korlann, S. D.; Richman, E. K.; Tolbert, S. H. Angew. Chem., Int. Ed. 2006, 45, 235.
Nano Lett., Vol. 8, No. 9, 2008
(31) Alonso, B.; Balkenende, A. R.; Albouy, P. A.; Amenitsch, H.; Rager, M.-N.; Babonneau, F. J. Sol-Gel Sci. Technol. 2003, 26, 587. (32) Urade, V. N.; Bollmann, L.; Kowalski, J. D.; Tate, M. P.; Hillhouse, H. W. Langmuir 2007, 23, 4268. (33) Zhao, D.; Yang, P.; Melosh, N.; Feng, J.; Chmelka, B. F.; Stucky, G. D. AdV. Mater. 1998, 10, 1380. (34) Brezesinski, T.; Antonietti, M.; Smarsly, B. AdV. Mater. 2007, 19, 1074. (35) Richman, E. K.; Brezesinski, T. B.; Tolbert, S. H. Submitted to Nat. Mater. (36) Brinker, C. J.; Lu, Y.; Sellinger, A.; Fan, H. AdV. Mater. 1999, 11, 579. (37) Yasuda, K.; Nohira, T.; Takahashi, K.; Hagiwara, R.; Ogata, Y. H. J. Electrochem. Soc. 2005, 152, D232. (38) Nohira, T.; Yasuda, K.; Ito, Y. Nat. Mater. 2003, 2, 397. (39) Nagamori, M.; Malinsky, I.; Claveau, A. Metall. Mater. Trans. B 1986, 17, 503. (40) Bao, Z.; Weatherspoon, M. R.; Shian, S.; Cai, Y.; Graham, P. D.; Allan, S. M.; Ahmad, G.; Dickerson, M. B.; Church, B. C.; Kang, Z.; Abernathy, H. W., III; Summers, C. J.; Liu, M.; Sandhage, K. H. Nature 2007, 446, 172. (41) Cai, Y.; Allan, S. M.; Sandhage, K. H. J. Am. Ceram. Soc. 2005, 88, 2005. (42) Smarsly, B.; Antonietti, M. Eur. J. Inorg. Chem. 2006, 6, 1111. (43) Cullity, B. D. Elements of X-ray Diffraction 101-102; AddisonWesley: Reading, MA, 1978. (44) Okamoto, H. Phase Diagrams for Binary Alloys; ASM International: Materials Park, OH, 2000. (45) Richman, E. K.; Senter, R.; Tolbert S. H. Manuscript in preparation. (46) CRC: Handbook of Chemistry and Physics http://www.hbcpnetbase. com/articles/05_02_87.pdf.
NL801759X
3079
