Optical properties and refractive index sensitivity of reactive sputtered oxide coatings with embedded Au clusters

Optical properties and refractive index sensitivity of reactive sputtered oxide coatings with embedded Au clusters N. M. Figueiredo, 1,2,a) T. Kubart, 2 J. A. Sanchez-Garc ıa, 3 R. Escobar Galindo, 4 A. Climent-Font, 5 and A. Cavaleiro 1 1 SEG-CEMUC – Department of Mechanical Engineering, University of Coimbra, Rua Lu ıs Reis Santos, 3030-788 Coimbra, Portugal 2 The ˚ Angstr€ om Laboratory, Uppsala University, P.O. Box 534, SE-751 21 Uppsala, Sweden 3 Energy and Environment Division, TECNALIA, San Sebastian 20009, Spain 4 Instituto de Ciencia de Materiales de Madrid, ICMM-CSIC, Campus Cantoblanco, 28049 Madrid, Spain 5 Centro de Micro-An alisis de Materiales, Universidad Aut onoma de Madrid, 28049 Madrid, Spain (Received 31 August 2013; accepted 16 December 2013; published online 14 February 2014) In the present study, nanocomposite coatings of Au clusters embedded in two different oxides, TiO 2 and Al 2 O 3 , were synthesized using pulsed DC magnetron sputtering. The depositions were carried out in three steps, by depositing the oxide, the Au clusters, and again the oxide. The deposition time of the Au clusters was varied in order to achieve different cluster sizes, morphologies, and nanocomposite topographies. The structure, microstructure, morphology, and the optical properties of the coatings were studied. With the increase in Au content, red-shifted surface plasmon resonance (SPR) peaks with higher intensity and increased widths were observed due to changes in the metal clusters sizes and morphology and due to interparticle effects. In order to relate the peculiar SPR extinction bands with the different clusters shapes and distributions, a simulation of the optical properties of the nanocomposites was performed making use of the Renormalized Maxwell-Garnett approach. A theoretical study concerning the refractive index sensitivity was made in order to predict the optimal coatings parameters for sensing experiments. The increased surface area and the strong SPR extinction bands make these coatings suitable for gas sensing and also catalysis, albeit many other application fields can be envisaged. V C2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4861136] I. INTRODUCTION Thin composite films consisting of metal nanoparticles in an oxide matrix are of interest for applications in electronics, sensors, catalysis, and biology. 1 This increasingly interest is mainly due to the exhibition of Localized Surface Plasmon Resonances (LSPRs) by the nanoparticles, optically induced oscillations of free electrons at the surface of the noble metal nanoparticles. The excitation of LSPRs results in strong light extinction effects, which are heavily dependent on the nanoparticle’s dielectric constant, size, and shape but also on the dielectric constant of the surrounding medium. 2,3 Materials exhibiting LSPR are thus attractive for a number of applications, including gas sensors. When metallic clusters exhibiting LSPR at a suitable wavelength are embedded in chemically reactive dielectric matrixes, significant change in the optical properties of the nanocomposites (e.g., reflectivity) can occur as a result of the interaction between the metal oxide and the gas molecules (either by chemical reaction or by physical adsorption). Since the LSPR is highly dependent on the dielectric properties of both the nanoparticles and the surrounding medium, molecule-induced changes in the vicinity of the nanoparticles can be detected by simply monitoring the position of the LSPR extinction peak. The detection of small concentrations of gas molecules can thus be carried out in a novel way. 4,5 Optical sensors have several inherent advantages as compared to other sensor technologies including the ability to monitor different optical properties of the sensing material (transmission, reflection, luminescence, etc.), compatibility with optical fibre based remote sensing approaches and the multi-gas detection capability (by using differences in the intensity, wavelength, phase, and polarization of the output light signals). They also offer more robust operation thanks to higher resistance to electromagnetic noise than conventional resistive sensors. The LSPR gas sensors further allows simultaneous temperature and gas sensing by monitoring of the SPR peak width and position. 6–8 The short (and tunable) characteristic electromagnetic field decay length is responsible for the enhanced sensitivity of the LSPR nanosensor. 9 Moreover, the LSPR nanosensors possess smaller footprint and cost at least 30–60 times less than commercialized thin film SPR sensors. 3 The addition of noble metals in the form of nanoparticles can enhance the response of many metal oxides to different gases by increasing the active surface area, improving the gas diffusion inside the film, and allowing, in certain cases, the selective adsorption of the target gas. LSPR sensors can thus be a suitable option for sensing in control and measuring systems with excellent performance, reliability, and low price. 10 Nanostructured materials exhibiting LSPR have been prepared mostly by colloid chemical methods, vacuum a) Author to whom correspondence should be addressed. Electronic mail: [email protected]t. 0021-8979/2014/115(6)/063512/13/$30.00 V C2014 AIP Publishing LLC115, 063512-1 JOURNAL OF APPLIED PHYSICS 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf evaporation, lithographic techniques, laser ablation, and electrodeposition. 11 Among all, the technique of magnetron sputtering has been widely used due to its low cost, simplicity, and versatility. 12 In this study, nanostructured coatings suitable for refractive index LSPR gas sensors were deposited by alternating sputtering. The coatings consisted of nanoclusters of Au embedded in a TiO 2 or Al 2 O 3 matrix and followed a simple but effective design, carefully chosen to allow optimization of the light extinction and the overall gas adsorption of the system. Au is one of the most inert materials known today, and it exhibits LSPR in the visible light spectrum. TiO 2 has many possible applications such as in photo-catalysis, heterogeneous catalysis, as gas sensors, solar cells, and bio-compatible materials; 13 as a gas sensor, TiO 2 is sensitive to O 2 ,H 2 , CO, CH 4 , and to several alcohols. 14 Al 2 O 3 is widely used in mechanical, optical, and micro-electronic applications because of its excellent chemical resistance, good mechanical strength, high hardness, transparency, high abrasive, and corrosion resistance as well as high insulating strength. 15 Al 2 O 3 is an excellent sensing material for humidity, and sensors based on Al 2 O 3 operate mainly at near room temperature making use of porous microstructures. At higher temperatures, they can also sense H 2 ,CO 2 , and O 2 . 14 The deposited nanocomposite thin films of Au@TiO 2 and Au@Al 2 O 3 were characterized with respect to the structure, microstructure, surface morphology, and optical properties. Results of this characterization are summarized in Sec. IV. Then, a theoretical analysis of the nanocomposite’s optical response was performed in order to identify coating parameters that allow maximizing their sensing performance. These results are presented in Sec. V. II. EXPERIMENTAL DETAILS Experiments were performed in a Lesker CMS-18 deposition system equipped with four magnetron sputtering sources. Ti and Al targets (99.995% purity) with a thickness and diameter of 6 and 100 mm, respectively, were used for depositing TiO 2 and Al 2 O 3 under reactive atmosphere. Au target (99.995% purity) with a thickness and diameter of 3 and 50 mm, respectively, was utilized for depositing the nanoparticles. Au was deposited using a magnetron Torus 2HV by Kurt J. Lesker equipped with high strength magnetic assembly. Base pressure of the system was below 3 10 5 Pa. The depositions were carried out at constant pressure by using a pressure-controlled gate valve in the pumping system. Au layers were deposited at a constant pressure of 1 Pa using 40 sccm of Ar. For the oxide layers, 15 sccm of O 2 was introduced in the chamber along with the 40 sccm of Ar, and the deposition pressure was set to 0.5 Pa. All depositions were carried out in pulsed DC mode with a Pinnacle Plus Advanced Energy generator. The Au was deposited at a constant power of 65 W, frequency of 250 kHz, and the off time of 0.5 ls. For TiO 2 and Al 2 O 3 , the same pulsing configuration was used at a constant power of 800 W. All the depositions were carried out with substrate rotation at a rotation speed of 20 r.p.m. The substrates were kept at floating potential, and no intentional heating was used. The deposition rates of Au and TiO 2 or Al 2 O 3 were calculated after depositing thicker layers and determining their thicknesses using a Dektak 150 profilometer. The depositions were carried out in a three-step process, depositing titania (or alumina), Au, and again titania (or alumina). The thickness of each oxide layer was set to about 20 nm. Au layers with thickness corresponding to 1, 2, and 3 nm were selected. Although Au does not form a continuous film at such low thickness, the nominal thickness is used to identify the studied samples. The structure was analysed by X-ray diffraction (XRD) using a Philips (PANalytical) diffractometer with Co-K a radiation in grazing incident configuration (GIXRD), with an incident angle of 2. The XRD parameters, such as the peak position (2h 0 ), peak intensity, and full width at half maximum (FWHM) were evaluated after fitting the XRD patterns using a Voigt function. The surface morphology of the coatings was analysed by high resolution (HR) field emission gun scanning electron microscopy (FE-SEM) using a Zeiss Leo 1550 microscope. The ImageJ 1.45 s program was then employed for image analysis. Additional characterization of the Au islands concerning their surface morphology was performed by atomic force microscopy (AFM) using a Veeco-Innova microscope operating in contact mode and using ultra sharp silicon nitride tips (with 2 nm tip radius). Rutherford backscattering spectroscopy (RBS) experiments were performed with a 5 MV HVEE Tandetron accelerator. 16 RBS spectra were collected using a 3.035 MeV He þ beam in order to improve the sensitivity to oxygen at the non-Rutherford cross section resonance 16 O(a,a) 16 O. The data were acquired simultaneously with two silicon surface barrier detectors located at scattering angles of 170with an energy resolution of 16 keV and an ion dose of 10 lC per detector. The experimental spectra were fitted using the program RBX. 17 The optical transmittance of the films was recorded in a UV-Vis-NIR spectrophotometer, using a Shimadzu SolidSpec-3700. III. THEORETICAL BACKGROUND A. Optical properties of nanocomposite thin films The most common way to model the optical properties of a nanocomposite system is to use the concept of effective dielectric function (EDF). One of the most used mean-field theories for the EDF calculation is the Maxwell-Garnett (MG) approximation, which is valid in the low concentration limit (volume fraction of the particles f1, always lower than 0.1 18 ). The MG approach can be improved and extended to higher fby taking into account the dipole-dipole interactions between particles, 19 a formalism that can be referred to as renormalized MG (RMG) approximation. 20 The RMG approach was first derived in Ref. 21 and has been generalized to nonspherical particles in Ref. 19. Considering the classical MG approach, under the mean field approximation, the effective dielectric function e eff of 063512-2 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf the composite system containing a small volume fraction fof separate inclusions on a host medium is given by 22 eef f eh eef f þ2eh¼4p 3N a;(1) where e h is the dielectric function of the host medium, N ¼f=Vis the particles concentration (Vrefers to the particle volume), and  a¼Piai=3(a i refers to the polarizability of the individual particle following a principal axis). An individual spherical particle of radius akpolarized by an external electromagnetic field, without interacting with other particles, is described by the polarizability 23,24 ai¼a3emeh emþ2eh ;(2) where emis the dielectric function of the metal inclusion. For ellipsoidal particles with semi axes a,band c, an analogous expression can be found in the quasistatic approximation via introducing geometrical depolarization factors Lialong these axes 23,24 ai¼abc 3 emeh ehþLiemeh ðÞ ;i¼1;2;3;(3) where RLi¼1. For spherical particles, L1¼L2¼L3¼1=3 and the expression (3) resumes to Eq. (2). If the particle possesses axial symmetry, two of the depolarization factors coincide. Prolate (“cigar”-shaped) spheroids, for which b¼cand L2¼L3, can be generated by rotating an ellipse about its major axis and have the following analytical expressions for L 1 and L 2 as a function of the eccentricity e: L1¼1e2 e21þ1 2eln 1þe 1e  ;e2¼1b2 a2;(4) L2¼1L1 2:(5) Oblate (“pancake”-shaped) spheroids, for which a¼b and L1¼L2, are generated by rotating an ellipse about its minor axis and have the following expressions for L 1 and L 3 as a function of e: L1¼1 2e2ffiffiffiffiffiffiffiffiffiffiffiffi 1e2 e2 rp 2tan1ffiffiffiffiffiffiffiffiffiffiffiffi 1e2 e2 r ! 1 2 1e2 e2  ; e2¼1c2 a2;(6) L3¼12L1:(7) The plasmon resonance thus splits into a strongly redshifted long-axis mode (polarization parallel to the long axis) and a slightly blue-shifted short axis mode (polarization perpendicular to the long axis). 24 Considering incoming light with normal incidence and ellipsoids with one of their axes perpendicular to a plane but with otherwise random orientations, the following simplification can be used for the average polarizability:  a¼a1þa2 2¼abc 6 emeh ehþL1emeh ðÞ þemeh ehþL2emeh ðÞ  : (8) The dipole-dipole interaction between the particles renormalizes their average polarizability, which then becomes 19 a¼2 a j1ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1j1d ðÞ p2ffiffiffiffiffiffiffiffiffiffiffi 1 pþarcsin1=2 1=2  "# ; (9) where ¼3jd=1j1d ðÞ½ ,j¼f4p a=3V ðÞ 2, and d ¼a?ak ðÞ =a?þak ðÞ is an anisotropy parameter that accounts for the differences in polarizability (between the axis perpendicular a?and the axis parallel akto the axis of symmetry of the spheroid). When f!0, then a! a. The effective dielectric function within the RMG approach is calculated by replacing  awith ain Eq. (1). Once the effective dielectric function e eff is known, the optical properties of the nanocomposite thin films can be determined. For simulating light propagation in planar multilayer thin films, including the effects of multiple internal reflections and interference, the “Transfer Matrix Method” 25 was implemented in Mathematica. This allowed calculating the total reflectance (R) and transmittance (T) of the nanocomposites. The absorbance (A) of the system was afterwards calculated through the relation A¼logT. B. Refractive index sensitivity of LSPR sensors Three material parameters can be responsible for changing the plasmon resonance peak as a function of changes in operating temperature and ambient gas atmosphere: 8 refractive index of the host matrix (nh), free carrier density of Au (N), and damping frequency of free carriers in Au (C). Gas phase interactions that cause changes in nhand Nresult predominantly in a shift of the LSPR absorption peak while interactions that modify Care expected to predominantly create a broadening or narrowing of the peak. The changes in the refractive index of the metal oxides have been proposed as the primary mechanism causing the shifts in the plasmon resonance peaks position of Au nanoparticles incorporated into a number of different metal oxide matrixes, including Au-TiO 2 ,Au-WO 3 , Au-NiO, Au-CuO, and Au-Co 3 O 4 . 7,26–29 Such changes are due to the physisorption or chemisorption of the analyte molecules onto the surface of the metal-oxide and into the porous matrix of the films. 7 Metal oxides are usually slightly off-stoichiometric resulting in either an excess [M mþx O] or a deficiency [M m-x O] in metal ions (M) compared to oxygen ions (O), originating either nor p-type semiconductor, respectively. 30 Atmospheric oxygen, for instance, is reported to adsorb preferentially on the surface defects of these oxides, resulting in changes of the conductivity and optical properties according to the nor p-nature of the oxide vacancies. 31 Some recent studies suggest that at high temperatures, the variations in the effective electron density of the Au 063512-3 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf nanoparticles can be the main cause for explaining the SPR peak shifts of some systems. 8,32 Concerning such cases, it was recently pointed out that the reducing gases that oxidize over metal oxide nanocomposites containing Au nanoparticles should donate electrons to the oxide matrix (that afterwards can interact with Au) and not directly to the Au nanoparticles. 31 Such changes in the matrix will also likely induce a change in the polarizability, changing its dielectric constant as well. 33 Thus, for the cases where Nseems to play a definite role in the change of the optical property of interest (which seems likely at least for the higher operating temperatures of certain metal oxides), a more complex panorama of interactions is present that requires further studies in order to clearly identify and quantify each individual contribution responsible for the overall observed change in the optical signal. In this work, the sensitivity of an SPR sensor will be evaluated in terms of refractive index sensitivity, which is valid for many Au-oxide systems operating between room temperature (R.T.) and moderately higher temperatures (300–400 C). Two main strategies are currently being used to quantify the variations in the SPR peak position: either by calculating the difference in the peak maxima position or by determining the variation in the peak height at a convenient point in the curve. The former strategy will be used here. The refractive index sensitivity of an SPR sensor with spectral interrogation is defined as 34–37 S¼dkres dnh¼ de0mðkÞ dnh  k¼kres de0mðkÞ dk  k¼kres ;(10) where e0mis the real part of the complex metal dielectric constant (em¼e0mþie00m). If the refractive index of the sensing medium is altered by dnh, then the resonance wavelength shifts by dkres. The refractive index sensitivity of the resonance is determined by the same two relations that determine the resonance wavelength kres, the resonance condition and the wavelength dependence of the dielectric function. 34 Resonance occurs at poles of the polarizability. At wavelengths where the imaginary part of the metal dielectric function is small or slowly varying, the pole of the polarizability is found by setting the real part of the denominator in Eq. (3) to zero. This results in the resonance condition 34 e0mðkresÞ¼1L Ln2 h;(11) which determines the value of the real part of the dielectric function at resonance. For the case of a sphere, L¼1=3, and Eq. (11) reduces to the familiar expression e0mðkresÞ¼2eh. The resonance wavelength kres is complex because the LSP is lossy due to material and radiative losses. Since the complex permittivity of a metal is a holomorphic function, for metals fulfilling ImfkresgRefkresg, the following relationship between the complex resonant wavelength kres and the real wavelength of maximum extinction at which extinction peak is observed kpeak holds: dkpeak ¼Re dkres fg :(12) Thus, the changes in the experimental data can be directly related to the analytical variations. 38 The LSPR phenomenon originates in the collective excitation of free-electrons of the metal particles, which can be described as a three-dimensional free-electron gas by the Drude theory (the first term of Eq. (21)). The real and imaginary parts of the metal dielectric constant near the resonance are thus given by e0mðx;RÞ¼1f0x2 p x2þC2 0ðRÞ and e00mðx;RÞ¼ f0x2 pC0ðRÞ x3þC2 0ðRÞx: (13) The previous expressions can be written in terms of wavelength by using the equality x¼2pc=k(where cis the speed of light in free space). For the real part of the metal dielectric constant, it reads e0mðk;RÞ¼1f0x2 p 2pc=k ðÞ 2þC2 0ðRÞ:(14) Substituting Eqs. (11) and (14) into Eq. (10) results in the general-shape expression for LSPR sensitivity S¼nh1L ðÞ =L  2pc=kres ðÞ 2þC2 0ðA;RÞ hi 2 x2 p2pc ðÞ 2=k3 res :(15) The ability of a spectroscopic SPR sensor to resolve small refractive index changes is directly proportional to the refractive index sensitivity and indirectly proportional to the width of the resonant feature. 35,36,38,39 In order to further infer the sensing potential of the coatings, the following figure of merit can thus be introduced: 35,36,38 v¼ S w ;(16) where Sis the sensitivity and wis the FWHM of the extinction peak. This figure of merit does not explicitly account for another important factor that is the SPR peak intensity. However, the variations in SPR peak intensity are indirectly accounted by the wterm; the narrowing of an SPR peak is usually accompanied by an increase in the peak intensity, and the broadening of an SPR peak is accompanied by a decrease in the peak intensity. IV. RESULTS This section starts with the chemical analysis of the coatings, then describes the structural and morphological analysis, and finally the optical properties and theoretical refractive index sensitivity are evaluated. 063512-4 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf A. Au concentration in the coatings The evolution of the elemental composition of the films with Au thickness was studied by RBS. From this analysis, it was possible to determine the atomic density of Au in the coatings. For the TiO 2 case, a total of 8:81015, 1:35 106, and 1:82 1016cm 2 were obtained for the 1, 2, and 3 nm thick Au layer, respectively (Table I). Assuming continuous Au layers, the corresponding thicknesses would be 0.74, 1.74, and 2.47 nm, taking into account the Au bulk density and molar mass. These values are roughly 20% lower than the theoretical values, which might be due to errors in the determination of the deposition rate from the thicker Au film. B. Structural properties The titanium oxide can crystallize in different structures. At ambient conditions, four structures are known: rutile, anatase, brookite, and srilankite. 15 The most common and stable structure of TiO 2 is rutile. 40 TiO 2 thin films with all of the four structures can be synthesized although for deposition temperatures below 150 C, amorphous TiO 2 thin films are usually obtained. Regarding the Al 2 O 3 , it can appear as the thermodynamically stable a-alumina as well as crystallized under several other metastable polymorphs. For depositions with substrate temperatures below 300 C, sputtered alumina coatings are typically amorphous. 41 For all the Au@TiO 2 and Au@Al 2 O 3 coatings, no XRD peak concerning any of the TiO 2 or Al 2 O 3 phases was found. This was expected considering the low substrate temperature during depositions. On the other hand, for both systems, Au was found to be increasingly crystalline with increasing Au content. In Figure 1, a close up of the XRD spectra concerning the Au (111) peak for the Au@TiO 2 and Au@Al 2 O 3 cases is shown. The particle sizes estimated from Scherrer formula 42 are presented as well. It can be seen in Figure 1that for the different oxide systems and Au contents, the XRD peaks become always wider after the encapsulation step. This trend suggests slightly smaller clusters after the top oxide deposition. This apparent diminution in the clusters sizes might be simply the result of the nucleation of smaller Au clusters (<1–2 nm) that were initially trapped in the oxide defects (being undetectable either by HRSEM or XRD) and that are now able to contribute to the XRD spectra. The possible changes in the morphology of the clusters could also add to the changes in the XRD peak widths. Another more remote possibility is that the bigger clusters consisting of aggregates of two or more smaller clusters get cleaved during the bombardment of the top oxide layer, thus moving the size distribution curve towards the smaller values. It was found that, for all the Au contents, the clusters are smaller in the Al 2 O 3 system. This can be explained by the higher surface mobility of Au on the TiO 2 surface. C. Surface morphology SEM analysis was performed before and after the deposition of the top oxide layer of TiO 2 or Al 2 O 3 , Figure 2and Figure 3, respectively. From Figure 2, it is apparent that the size and shape of the Au nanoparticles strongly depends on the thickness of the sputtered Au layer. When increasing the Au thickness from 1 to 3 nm, the Au clusters shape goes from rather circular to fiber-like and their average particle size increases considerably from 6 to 10 and to 17 nm (TiO 2 ) and from 4 to 7 and to 12 nm (Al 2 O 3 ). These three different cases illustrate well the growth modes expected for Au on oxide surfaces: 43–45 clusters at 1 nm thick layer picture corresponds to the end of the first stage of nucleation of nanometric 3D hemispherical Au clusters (Figure 2(a)); at 2 nm, a second stage of lateral growth and some coalescence of the 3D clusters occurs (Figure 2(b)); and at 3 nm, third stage of pronounced coalescence of the 3D clusters takes place (Figure 2(c)). At even higher thicknesses, a vertical growth of the coalesced clusters would finally be observed. By comparing the micrographs of the two oxides, it becomes apparent that the Au clusters are much bigger and the cluster density much lower in the TiO 2 case, for all the Au TABLE I. Surface analysis results of Au nanoparticles over Al 2 O 3 and TiO 2 . Oxide Al 2 O 3 TiO 2 Nominal Au thickness (nm) 1 2 3 1 2 3 Ellipse major axis—a(nm) 5.9 9.9 16.9 6.7 12.8 22.9 Ellipse minor axis— b(nm) 3.9 6.3 9.3 4.7 8.1 12.4 Au area fraction—A(%) 34.2 41.7 47.1 36.0 44.1 49.7 Particle density (/lm 2 ) 17 044 8096 3622 13 702 5254 2110 Au atomic density (RBS) (/cm 2 ) … … … 8.8 10 15 1.4 10 16 1.8 10 16 Cluster’s height (RBS/SEM)— c(nm) 3.7 6.7 8.4 3.3 6.2 8.1 Average size of the bigger axes (oblate)—xab (nm) 4.8 7.9 12.5 5.6 10.2 16.7 Average size of the smaller axes (prolate)—xbc (nm) 3.8 6.5 8.9 4.0 7.1 10.0 Aspect ratio (oblate shape)—AR obl 0.76 0.84 0.67 0.59 0.60 0.48 Aspect ratio (prolate shape)—AR pro 0.64 0.65 0.52 0.59 0.55 0.44 Equivalent diameter of a sphere (nm) 4.4 7.5 11.0 4.7 8.6 13.2 Interparticle distance—D(nm) 7.8 11.6 17.4 8.9 15.0 23.8 Au average height (XRD)—h0(nm) 3.6 5.1 5.3 4.3 5.3 6.4 Au average height (AFM)—h(nm) 2.8 3.5 3.9 3.1 3.4 3.6 Volume fraction (optical)—f00.10 0.13 0.13 0.08 0.10 0.10 063512-5 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf concentrations. The higher surface mobility of Au in the TiO 2 explains this difference. From the FE-SEM micrographs of each nanocomposite film (Figure 2), the Au area fraction, the average major and minor axis of fitted ellipses, the Au mean particle size, and the number density of Au nanoparticles were obtained by image analysis. The results are given in Table I. From the Au atomic density determined by RBS and the particle density determined by SEM, it was possible to calculate the average clusters height assuming the presence of spheroidal particles. The particle volume of a whole spheroid is defined by Vparticle ¼1 6pabc;(17) where a,b, and care the equivalent main axes. Knowing the average particle volume through the relation  Vparticle ¼MAu qAu atomic density particle density ;(18) where M Au is the molar mass in [g=atom], qAu is the density in [g=nm3], the atomic density is in [atom=lm2] and the particle density is in [number of particles=lm2], makes then easy to calculate the average height cof the cluster. Since the deposition conditions used for depositing Au were identical for the two oxides, the Au atomic density measured on TiO 2 was used for the Al 2 O 3 case as well. The interparticle distance was evaluated with the Au nanoparticle neighbourhood defined by the Voronoi boundary in the Voronoi diagram of each film. 46 For both systems, each Au nanoparticle had six neighbours on average. For the hexagonal lattice, the average interparticle separation can be written as 45,46 D¼p1=2 x 121=4A1=2:(19) Finally, and for comparison purposes, the diameter of a sphere with an equivalent volume to the spheroid can be calculated through the following equation: Dsphere ¼ffiffiffiffiffiffiffi abc 3 p:(20) All the results of these calculations are summarized in Table I. In most cases, the calculated clusters heights are close but smaller than the minor lateral dimensions of the FIG. 1. GIXRD patterns of the films with increasing Au interlayer thickness concerning the two systems: (a) Au@TiO 2 , (b) Au@Al 2 O 3 . (c) Nanoparticle sizes estimated by the Scherrer formula, before (open symbols) and after the final encapsulation (full symbols) step. 063512-6 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf FIG. 2. HRSEM images of the nanocomposites with increasing Au thickness, for the TiO 2 þAu: (a1) 1 nm, (b1) 2 nm, (c1) 3 nm and for the Al 2 O 3 þAu: (a2) 1 nm, (b2) 2 nm, (c2) 3 nm. Histograms representing the Au clusters size distributions are shown in inset. FIG. 3. HRSEM images of the TiO 2 þ Au þTiO 2 samples with increasing Au interlayer thickness: (a) 1 nm, (b) 2 nm, (c) 3 nm. 063512-7 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf clusters. In order to determine the aspect ratio of the spheroidal particles, an average of either the two bigger axes (oblate) or the two smaller axes (prolate) was considered. Higher aspect ratios (ARs) were achieved for the oblate shapes, suggesting better agreement with experiments. Additional information about the clusters morphology and its variation with the deposition time was obtained by AFM. From the AFM images of both the Au@TiO 2 and Au@Al 2 O 3 systems, the average height of the deposited clusters was determined (Table I). For both systems, it was found that the clusters heights calculated by AFM were lower than calculated from the RBS/SEM data and even than calculated from XRD. Since the measured AFM image is a convolution of the tip geometry and the geometry of the clusters, lower values for the heights can be measured due to the close proximity of the clusters. On the other hand, the presence of smaller clusters in between the SEM detected clusters could contribute as well to the lower values found for the clusters height. A few line profiles were performed in zones of the samples containing minor scratches (case where the clusters were absent from the oxide surface). The average values for the clusters heights were greater than the averaged ones from the intact zones (the values were closer to the ones obtained from the XRD technique). In general terms, the data from the different characterization techniques are consistent, especially when considering the small sizes of the nanoparticles being studied. The current coatings design allows optimization of coatings topography for gas sensing. Due to the presence of Au nanoparticles with different sizes and shapes in the interlayer and to the chosen thickness of the metal-oxide overcoat, it is possible to attain a top oxide layer with increased roughness and surface area (high density of surface defects)—a topography following the one of the Au clusters underneath. In Figure 3, the existence of TiO 2 caps separated with clear interfaces (corresponding to the zones without Au) is observed. The size and shape of these caps closely follows the ones of the Au nanoparticles. Both the reactivity and gas permeability of the thin TiO 2 and Al 2 O 3 films are expected to be enhanced by the present coatings design. D. Optical properties Optical transmission spectroscopy measurements were carried out in order to study the SPR effect with the Au@TiO 2 and Au@Al 2 O 3 coatings (Figure 4). In Figure 4, it is observed that the overall transmission decreases with the increase in the Au thickness. This corresponds to an enhanced absorption due to the higher volume fractions of Au (Table I). There is also a red-shift and a broadening of the SPR absorption peaks with increasing Au thickness, due to a combined effect of changes in the size and shape of Au nanoparticles as well as in the interparticle distances (Figure 2). In order to elucidate the connection between the structural and microstructural properties and the extinction signal of the nanocomposite films, a theoretical modelling of the optical properties was performed using the concept of effective dielectric function (EDF). As an approximation for e h , the optical constants (refractive index nand extinction coefficient k)ofapureAl 2 O 3 and TiO 2 thin film (deposited under the same conditions as in the nanocomposites) were estimated numerically from the transmittance results by means of an unconstrained optimization technique describedindetailinRef.47. These constants were then substituted in the complex expression eh¼ðnþikÞ2. The thickness of each oxide layer was also estimated with the application of this method. For the pure Al 2 O 3 and TiO 2 films, the refractive index at a wavelength of 550 nm was 1.68 and 2.43 and the thickness of the thin films was 14 and 16 nm, respectively. To represent e Au , an analytical expression determined by Rakic et al. 48 based on the Brendel-Bormann (BB) model was used. It was shown by the authors that, for Au, the BB model allows better description of experimental data than the more frequently used Drude-Lorentz (DL) model, especially where the lowest-laying interband transitions occurs. The BB model accounts for the transitions within the conduction band, by using a Drude term, as well as for the interband transitions from lower-lying bands into the conduction band or from the conduction band into higher unoccupied levels, by using a superposition of an infinite number of oscillators. The resulting expression for the dielectric function of the metal is the following: 48 eðxÞ¼1 X2 p x2þiC0x þX k j¼1 1 ffiffiffiffiffiffi 2p prjðþ1 1 exp xxj ðÞ 2 2r2 j "# fjx2 p x2x2 ðÞ þixCj dx;(21) FIG. 4. Transmission spectra of the (a) Au@TiO 2 and (b) Au@Al 2 O 3 samples on glass with increasing Au interlayer thickness. 063512-8 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf where xis the frequency of light, xpis the plasma frequency of the metal, Xp¼ffiffiffiffi f0 pxpis the plasma frequency associated with intraband transitions with oscillator strength f0and damping constant C0, while kis the number of BB oscillators used to interpret the interband part of the spectrum with frequency xj, strength fj, lifetime 1=Cj, and line shape parameter rj. The number of harmonic oscillators per frequency interval is determined by a Gaussian function. In the present work, the analytical solution of the second term of Eq. (21) given in Ref. 48 was used for the calculations. In order to account for the intrinsic size effects of the nanoparticles, the BB model was modified by correcting the plasmon damping parameter (C) according to the relation 49 C¼C0þAF R;(22) where C0is the relaxation constant of the bulk metal (given in Ref. 48), Ais a phenomenological parameter that is a function of the geometry, of the order of unity, and which in practice is adjusted to provide the best fit of the data, 50  F is the Au Fermy velocity (1.39 10 15 nm/s), and Ris the radius of the nanoparticle (in nm). One way of estimating the volume fraction fof metal islands from two-dimensional pictures is by using the concept of “optical thickness” t opt as in terms of effective depolarization factors. An appropriate definition would be to use 45 topt ¼4 3D;(23) where Dis the average particle’s center-to-center distance. The volume fraction follows then from the relation 45 f¼t topt ;(24) where tis the nominal thickness of the film. The results of these calculations were compiled into Table I. The simulations of the optical absorbance spectra were performed considering a stack of three layers on a glass substrate. The stack consisted of a top oxide layer, followed by an intermediate nanocomposite layer and another bottom oxide layer with characteristics identical to the first layer. Different parameters were varied in order to achieve a good fit to the experimental data, namely, the Au volume fraction f, the phenomenological parameter A, the average diameter 2R, the shape (spherical, oblate, and prolate) and aspect ratio AR of the clusters, and the thickness of each layer d i . The simulated absorbance spectra are presented in Figure 5, and their corresponding fitting parameters are listed in Table II. The simulated spectra were divided into two major groups A and B, following the two types of spheroidal particles that were tested, oblate and prolate, respectively. From all the samples, only in the 1 nm Au@TiO 2 case, it was possible to consistently describe the optical absorbance spectrum by means of using perfectly spherical shapes in the simulations (AR ¼1). For all the other cases, both the oblate and the prolate shapes were successfully implemented in the simulations. The main differences in parameters between the two spheroidal shapes were found in the volume fraction and in the aspect ratio of the clusters. They were higher for the prolate shapes, and this difference in values became more pronounced with the increase in the nominal Au thickness. The size effect parameters and particles sizes used to correct the Au dielectric function were similar in both cases. When comparing simulation results between oxide systems, it was clear that higher ARs and lower fs were needed for the TiO 2 case, irrespectively of the shape being considered. When simulating the optical response of the metallic spheroidal nanoparticles, two of the axes were considered to be parallel to the thin film surface. For the particular case of the prolate shape, this resulted in the appearance of two distinct resonances, the less intense blue-shifted one being somewhat masked for the case of higher ARs, as shown in Figure 5(a), but becoming evident for lower ARs, as it is seen in Figures 5(b) and 5(c). In the experimental data, the higher energy resonance is clearly absent, which again suggests the better agreement of the oblate shape with the experiment. Still, for the two 2 nm Au samples and the 3 nm Au@TiO 2 sample, it was possible to satisfactorily simulate the absorbance spectra by assuming the presence of prolate shapes. Overall, the simulations performed considering particles with oblate shapes (group A) showed better (very good) agreement with the experimental data of both the oxides, even for the 3 nm Au samples, case where the particles were more irregular shaped. Aditionally, the simulations using oblate shapes gave as well Au volume fraction values closer to the ones determined from SEM images using Eq. (24). The results from the simulations concerning the clusters heights (given in this case by the parameter 2R) seem to agree better with the XRD results than with the combined RBS/SEM results (which give higher values). When estimating previously the cluster heights using RBS and SEM data, all the Au atoms were considered to be in the SEM-detected clusters. Both the optical and the structural results suggest that even after the encapsulation step, there are still smaller Au atoms (<1–2 nm) present in the coatings. E. Sensitivity of the LSPR sensors In Figure 6, the sensitivities and figure of merit calculated for the different coatings based on the previous simulation results are presented. The shape that showed better agreement with experiment, i.e., the oblate shape (simulation group A) was used in the calculations. Figure 6(a) shows that, for both oxides, there is an increase in the sensitivity with the nominal thickness of Au. This is mainly due to the decrease in the depolarization factor L(Table II) with the Au increase. When comparing results between oxides, overall the Al 2 O 3 matrix shows lower sensitivity values, despite the lower depolarization values achieved in the simulation. This is mainly due the lower refractive index of the Al 2 O 3 . The only exception is in the 2 nm Au layer, case where the difference in the 063512-9 Figueiredo et al. J. Appl. Phys. 115, 063512 (2014) Downloaded from http://pubs.aip.org/aip/jap/article-pdf/doi/10.1063/1.4861136/13379663/063512_1_online.pdf