Catherine
Marichy
a,
Jean-Francois
Dechézelles
b,
Marc-Georg
Willinger
a,
Nicola
Pinna
*ac,
Serge
Ravaine
b and
Renaud
Vallée
*b
aDepartment of Chemistry, CICECO, University of Aveiro, 3810-193 Aveiro, Portugal. E-mail: pinna@ua.pt
bCentre de Recherche Paul Pascal (UPR 8641, CNRS), 115 avenue du docteur Albert Schweitzer, F-33600 Pessac, France. E-mail: vallee@crpp-bordeaux.cnrs.fr
cWorld Class University (WCU) program of Chemical Convergence for Energy & Environment (C2E2), School of Chemical and Biological Engineering, College of Engineering, Seoul National University (SNU), Seoul 151-744, Korea. E-mail: pinna@snu.ac.kr
First published on 15th March 2010
Combining both electromagnetic simulations and experiments, it is shown that the photonic pseudo band gap (PPBG) exhibited by a silica opal can be fully controlled by Atomic Layer Deposition (ALD) of titania into the pores of the silica spheres constituting the opal. Different types of opals were assembled by the Langmuir–Blodgett technique: homogeneous closed packed structures set up of, respectively, 260 and 285 nm silica spheres, as well as opal heterostructures consisting of a monolayer of 430 nm silica spheres embedded within 10 layers of 280 nm silica spheres. For the stepwise infiltration of the opals with titania, titanium isopropoxide and acetic acid were used as metal and oxygen sources, in accordance with a recently published non-aqueous approach to ALD. A shift of the direct PPBG, its disappearance, and the subsequent appearance and shifting of the inverse PPBG are observed as the opal is progressively filled. The close agreement between simulated and experimental results is striking, and promising in terms of predicting the properties of advanced photonic materials. Moreover, this work demonstrates that the ALD process is rather robust and can be applied to the coating of complex nanostructures.
This article reports on the application of atomic layer deposition (ALD) for the highly controlled infiltration of titania in a silica direct opal through the conformal shell-like coating of the interstitial sites in the face-centered cubic arranged silica spheres. ALD has already proven to be the most suitable technique for the deposition of conformal films with submonolayer control, allowing one to reach filling fractions as good as 88% of the pore volume.12–17 The deposition process applied in this work was recently developed and is based on a non-aqueous sol–gel reaction18 between a metal alkoxide and acetic acid as metal and oxygen sources, respectively.19,27 It proved to be rather robust and well suited for the coating of various nanostructured materials, such as carbon nanotubes.20,21 This report further points out the robustness and the wide applicability of this new ALD process. While the complete GaAs ALD infiltration of a Langmuir–Blodgett engineered silica opal was already reported by Povey et al.,22 the combination of the layer by layer Langmuir–Blodgett deposition method for the production of complex photonic crystals (e.g., opals containing defect layers)23–25 with the subsequent and progressive coating of the nanostructure by ALD is reported here for the first time. This allows precise control of the photonic properties of the resulting heterostructure. The shifting of the direct PPBG, its disappearance, the subsequent appearance and shifting of the inverse PPBG that is observed as the direct opal is progressively filled, are perfectly reproduced by electromagnetic simulations. Similar results are obtained for the pass band within the PPBG of a progressively filled heterostructure (i.e., an opal containing an intentional defect layer). This controlled modulation of the PPBG can be used, for example, to tune the range of an optical filter.
Fig. 1 Simulated band structures of the direct silica opal (a), of a titania partially (b) or completely (c) conformally filled silica opal and of a pure titania inverse opal, obtained by conformal growth of titania around the air pores (d). The filling fractions of titania in the anatase phase of these last three structures are r = 4.1%, r = 13.1% and r = 13.1%, respectively. |
Fig. 2 Evolution of the frequency shift (solid lines) and width (vertical interdistance between the dash or dot lines) of the ΓL PPBG as a function of the filling fraction r, expressed as the coating thickness divided by the sphere diameter in the cases of amorphous (solid line between dot lines) and anatase (solid line between dash lines) titania. |
Fig. 3 SEM images of the (a,b) 285 nm sphere diameter direct silica opal filled with titania, (c) 260 nm sphere diameter inverse titania opal after removal of the template. (d,e) TEM images of the 285 nm sphere diameter direct silica opal filled with titania (the arrows denote the contact point of the silica spheres during deposition) and (f) energy filtered TEM images at the Ti 2p edge showing the conformality of the titania coating around the silica spheres. TEM images of cross section cuts of the annealed opal (260 nm sphere diameter) before (g,h) and after removal of the silica template (i). Inset: power spectrum of the HRTEM image in (h). |
Fig. 4 Experimental (a, solid lines) and simulated (b, dash lines) extinction spectra of 285 nm sphere diameter direct silica opal progressively filled with titania by ALD. The blue lines depict the progressive red shift and reduction of the direct opal photonic pseudo-band gap. The black lines depict its disappearance. The red lines depict the progressive red shift and increase of the titania-filled silica inverse opal PPBG. |
In agreement with the theoretical calculations, the first step (disappearance of the peak) corresponds to the modification of the band structure of the direct opal in such a way that the refractive index contrast between the constitutive spheres and the surroundings vanishes. In the present system, this happens at a filling fraction r = 6%. In the second step, the index contrast gets reversed as the surrounding pores are further filled with a higher index dielectric material. The band gap of the resulting inverse opal appears and converges once the opal is completely filled. It is important to notice that the increased extinction at wavelengths below the position of the PPBG originates from the increased incoherent diffusion as titania is progressively introduced in the structure.26
Comparison between the theoretical and experimental plots in Fig. 4 reveals two things. Firstly, the good quality of the grown photonic crystals becomes evident from observing the low wavelength range of the extinction spectra. The secondary periodic oscillations, so-called “Fabry–Pérot” fringes, are due to interferences of light propagating among various optical paths back and forth in the structure. They appear in a spectral region where the dispersion relation is linear.28 In this region, the crystal can be assimilated to a homogeneous transparent medium with an effective refractive index. Only good quality crystals, with a homogeneous optical thickness, can exhibit “Fabry–Pérot” fringes. It is important to point out the excellent matching between the spectral positions of the fringes observed on both theoretical and experimental extinction spectra. Secondly, Fig. 4 also reveals that the theoretical spectrum shows a slightly reduced shift in the spectral position of the peaks at maximum filling. This discrepancy is due to the fact that the opal obtained by the Langmuir–Blodgett deposition is not fcc close packed, as simulated here. It is better described as a (2 + 1)D photonic crystal resulting from the stacking in a one-dimensional (1D) lattice of prepacked two-dimensional (2D) colloidal crystals.29 The lack of a strict correlation between successive monolayers induces a greater d111 of the LB films as compared to the equivalent photonic crystal grown by controlled evaporation (fcc).22 Consequently, a higher void volume fraction of the overall structure leads to a significantly increased composite dielectric constant, and so to a larger shift in the Bragg peak after ALD infiltration. Evidence for this is provided by the spectra recorded from the fully filled samples after annealing (Fig. 4a, black curve). It shows a blue shift of the extinction spectrum, which is opposed to what would be expected solely from the transformation of the amorphous titania phase into anatase, where the increased refractive index would lead to a red shift of the PPBG (cf.Fig. 2 for the evolution of the adimensional frequency in the amorphous and anatase cases). This is an indication of the fact that the relaxation/densification of the structure overcompensates for the increase of the refractive index upon annealing, underlining the explanation given above.
In Fig. 5a, the comparison between the extinction spectrum of the annealed sample and the simulated spectra of silica opals conformally filled with anatase is shown. The best matching of the extinction spectra, with a maximum at λ = 740 nm, occurs for a theoretical filling ratio of 9.7%. Such a filling fraction is larger than the theoretical maximum expected from the geometrical arrangement of silica spheres in perfectly close packed (111) planes, for which the filling would end, because of the closing of the structure, at a ratio r = 7.75%.12 The larger filling ratio obtained through compacting of the structure by the annealing process is another indication of the non-fcc close packing of the direct opal.
Fig. 5 Experimental and simulated extinction spectra of 285 nm (a) or 260 nm (b) sphere diameter direct silica opal fully filled with titania by ALD. (a) The black crosses curve pertains to the extinction spectrum of the opal after annealing. The red crosses curves pertain to the simulated extinction spectra for anatase titania with filling fractions r ranging from 9.7 to 10.2 and 13.1% (from left to right). The best agreement between experiment and theory is found for r = 9.7%. (b) The red solid line (black crosses curve) pertains to the extinction spectrum of the opal prior (after) annealing. The black circles curve pertains to the extinction spectrum of the annealed opal after HF treatment, inducing removal of the silica particles. The red curves pertain to the simulated extinction spectra of anatase titania-filled silica direct opal (crosses) and pure anatase titania inverse opal (circles) with a common filling fraction r = 10.2%, matching the best the corresponding experimental spectra in both cases. |
The whole study was repeated on a silica template constituted of spheres with a diameter D = 260 nm. Fig. 5b shows the experimental extinction spectra obtained for the fully filled silica template before and after annealing, as well as the extinction spectra of a pure titania inverse opal. The latter was obtained after removal of the silica spheres by HF treatment (cf. HRTEM image, Fig. 3i). In agreement with above, the extinction spectrum after annealing is blue-shifted. After removal of the silica template, the extinction spectrum is further blue-shifted due to the reduction of the effective refractive index. For comparison, Fig. 5b also presents the theoretical extinction spectra calculated for a conformal coating of the silica opal with anatase and the corresponding inverse opal. In excellent agreement with the results obtained in the case of the larger silica spheres, the best matching is obtained for a theoretical filling ratio of r = 10.2%.
The controlled insertion of a defect causes a rupture in the periodicity of the photonic crystal and induces the appearance of localized states for photons within the gap. Recently, different techniques were developed to introduce a planar defect into 3D PCs.24,30–32 The obtained heterostructures present a dip, also called a pass band, inside the stop band in the transmission spectrum. To the best of our knowledge, neither the progressive titania filling by ALD of the direct template nor the inversion of such a heterostructure have been reported so far. Moreover, the inversion process presents the advantage of increasing the quality factor of the pass band. Accordingly, due to the versatility of the Langmuir–Blodgett technique, allowing control of the deposition layer by layer, a photonic heterostructure made of two stacks of five layers of 280 nm diameter silica host particles surrounding a defect layer constituted by a monolayer of 430 nm diameter silica guest particles was produced. In the next step, the silica template was filled with titania by ALD. Fig. 6 shows the extinction spectra obtained both experimentally (a) and by FDTD simulations (b). The SEM transversal view of the heterostructure is also shown as insets in Fig. 6a and a cut in the yz-plane of the FDTD simulated heterostructure in Fig. 6b. The good ordering of the structure is clearly visible. A slight positional disorder generated in the layer situated on top of the guest particles layer is also observed, as already reported previously.25 The insets of Fig. 6a and b show the main difference between the experimentally engineered and simulated heterostructures: while the defect layer is constituted of host silica particles arranged in a hexagonally packed monolayer in the built opal, it is approximated to a block of the same thickness but with an effective refractive index material (n = 1.35) in the simulations. Although such an approximation was very successful in modeling the direct heterostructures,25 this is not the case for the filled opal. Indeed, on the experimental side, the conformal growth will take place everywhere, including around the host particles, while in the simulated case this conformal growth is effective only around the guest particles, as exemplified in the inset of Fig. 6b. Such a difference will have an impact on the optical properties of the system. As predicted by theory and simulations, the introduction of a defect layer introduces a pass band in the PPBG of the direct opal. This pass band manifests as a dip at 612 nm, as seen in the extinction spectra (Fig. 6b). The progressive titania ALD filling of the direct template induces a red shift of the PPBG together with its pass band (blue lines), followed by its disappearance (black line), and the reappearance and red shift of the PPBG together with its pass band (red lines) in the inverse opal geometry. Note here the monotonous red shift of the pass band as a function of filling. Fig. 6a shows a similar tendency but with slight variations: while the pass band and high wavelength edge of the PPBG shifts to the red (blue lines) as a function of filling before the disappearance of the pass band (black line), the low wavelength edge of the PPBG did not significantly move. In contrast to what happens in the simulations, further filling induces a blue-shifted reappearance of the pass band within the PPBG in the inverse geometry (red lines) prior to progressively shifting more to the red as the filling fraction further increases. In the inverse geometry, both low and high wavelength edges of the PPBG move together with the pass band. As a whole, the main differences between simulations and experiments are a reduced and non-monotonous red shift of the pass band as a function of the filling fraction, together with a non-complete disappearance of the PPBG (e.g. the low wavelength edge persists, cf. black line) in the experimental extinction spectra. We tentatively attribute these discrepancies to the approximation made in order to simulate the defect layer. The increased intensity observed in the blue region of the experimental spectra is due to diffusion by unintentional defects introduced in the structure (inset of Fig. 6a), which are not taken into account in the simulations. In the next step, the experimental results presented here will be compared to new types of simulations combining molecular dynamics and FDTD taking more precisely into account the actual composition of the defect layer and the (2 + 1)D instead of the fcc structure of LB-obtained colloidal crystals. Furthermore, an experimental architecture consisting in a block defect layer instead of the hexagonally packed monolayer of host particles will also be produced and compared to the herein presented calculations.
Fig. 6 Experimental (a) and simulated (b) extinction spectra of a heterostructure constituted by a defect layer of 430 nm diameter guest silica particles comprised between two stacks of five layers of host silica particles of 280 nm diameter progressively filled with titania by ALD. The blue lines depict the progressive red shift and reduction of the direct opal photonic pass band. The black lines depict its disappearance. The red lines depict the progressive red shift and increasing of the titania filled silica inverse opal pass band. The insets show a SEM transversal view (a) and a cut in the yz-plane (b) of the engineered and simulated heterostructure, respectively. |
Simulations were performed in two steps. Firstly, in order to determine the photonic band structure of the various samples, fully-vectorial eigenmodes of Maxwell's equations with periodic boundary conditions were computed by preconditioned conjugate-gradient minimization of the block Rayleigh quotient in a planewave basis, using a freely available software package.34 The eigenfrequencies up to the twelfth band were computed for 100 equally spaced k points along each high frequency direction (XULΓXWK) of the first Brillouin Zone. The irreducible Wigner–Seitz cell (WSC) was divided into 32768 segments, defining the spatial resolution. Similar simulations were performed to calculate the photonic local density of states in direct silica opals.35 Secondly, in order to calculate the optical extinction spectra, finite-difference time-domain (FDTD) simulations were performed,36 using a freely available software package with subpixel smoothing for increased accuracy.37 The computational cell, in which the incoming wave propagates along the z direction, has been implemented with periodic boundary conditions in the x and y directions, and perfectly matched layers (PMLs) in the z direction. The resolution of the grid has been refined such that the convergence of the results was ensured. In all simulations, the direct opals have been represented by monodisperse spheres of a given size (285 or 260 nm) and dielectric constant (ε = 2.1) in an arrangement corresponding a face-centered cubic (fcc) lattice.25 In the FDTD simulations, the lattice has been limited to 10 planes normal to the direction [111], in agreement with the planned experimental conditions. The resulting crystal was placed on a substrate with a dielectric constant of 2.3. In order to simulate the growth of titania (amorphous with ε = 5.52, anatase with ε = 7.29) in the pores left by the direct opal during the ALD process, the original silica spheres were surrounded by interpenetrating titania shells of higher radius.
This journal is © The Royal Society of Chemistry 2010 |