Tetraethoxysilane (TEOS), Pluronic F127 ([EO]₁₀₆ [PO]₇₀ [EO]₁₀₆; EO = ethylene oxide, PO: propylene oxide), cetyltrimethylammoniumbromide (CTAB), acetylacetone (ACAC), dihexadecyl phosphate (DHDP), and chlorotrimethylsilane (TMCS) were supplied by Sigma-Aldrich. Zirconium isopropoxide (Zr(OCH₂CH₂CH₃)₄), N,N-diisopropylethylamine (DIPEA), benzyl alcohol, and hydrochloric acid were purchased from Merck. Ethanol was purchased from Cicarelli. Tetrahydrofurane (THF) was supplied by Biopack, and toluene by Anedra. E-pure water (18 MΩ.cm) was used as a reagent.

Mesoporous oxide thin films were deposited by dip coating on glass substrates, combining the sol-gel process and evaporation induced self-assembly of surfactants (Brinker et al., 1999).

For the production of mesoporous SiO₂ films (named SC), a prehydrolysis solution was formulated by refluxing TEOS for 1 h in a water-ethanol solution, with [H₂O]:[Si] = 1 and [EtOH]:[Si] = 5. SC films were obtained with the sol comprised of TEOS:CTAB:H₂O:EtOH:HCl with a molar ratio 1:0.1:5:40:0.004. Mesoporous silica films synthesized using this surfactant exhibit an ordered and interconnected pore array with three-dimensional hexagonal (P6₃/mmc) structure, with a characteristic pore diameter of around 3 nm (Fuertes et al., 2008).

On the other hand, a solution containing Zr(OCH₂CH₂CH₃)₄:F127:H₂O:EtOH:HCl:ACAC with molar ratios 1:0.005:20:40:1:1 was used for deposition of mesoporous ZrO₂ films (named ZF). The mesoporous zirconia films synthesized using this surfactant usually present ordered arrangements of pores in an Im3m cubic structure, with pore diameters between 8 and 10 nm (Zelcer and Soler-Illia, 2013).

Three different types of samples were prepared: single-layer films, bilayer films, and PCs. The single-layer SC and ZF films were deposited onto bare glass by dip-coating. The deposition was carried out using an extraction speed of 2 mm/s, and was performed in an environment with a relative humidity (RH) of 20%. This extraction speed was chosen in order to obtain thick films that are more appropriate for studying functionalization effects, as they provide a larger amount of material and surface area for interaction with functionalization agents and analytes. SC samples were thermally treated for 30 min periods at room temperature, 60 °C, 130 °C and 200 °C; ZF films were treated directly at 200 °C also for 30 min (Zelcer and Soler-Illia, 2013). These thermal treatments consolidate the films, but are mild enough to preserve the surfactants inside the pores. Therefore, a new layer can be deposited on top of these films without filling the mesopores with the precursor solution (Fuertes et al., 2007).

For bilayers, successive deposition of SC and ZF layers was carried out following alternating steps of deposition and thermal treatment for each material. Both possible material order configurations were built, with either the SC film or the ZF film adjacent to the substrate. To remove the surfactant, the same treatment was applied to all systems: the samples were heated to 350 °C with a temperature ramp of 1 °C/min, and held at that temperature for 2 h.

In the case of the PCs, alternating deposition of the SC and ZF mesoporous oxide layers by dip-coating was performed with extraction speeds of 1 mm/s. This extraction speed was selected to obtain layer thicknesses that yield a device with a photonic band gap in the UV region of the spectrum. The same thermal treatments used for the single and bilayers were applied to the PCs. However, in this case, fused silica was used as a substrate since this substrate is transparent in the UV, the range of the electromagnetic in wich the photonic band gap is expected to be located. The synthesized photonic crystals were named 4X (SC/ZF), where 4 represents the number of bilayers deposited and the films that comprise each one of them are indicated in parenthesis, being the silica film the one adjacent to the substrate.

A flowchart showing the steps involved in the preparation of single layers, bilayers and PCs is presented in Supplementary Figure S1.

Mesoporous systems were functionalized using TMCS and DHDP. TMCS acts as a silanizing agent for SiO₂, reacting with surface silanol groups to form OSi(CH₃)₃ species (Zhao and Lu, 1998; Venkateswara Rao et al., 2009; Sun et al., 2021). In contrast, DHDP anchors to transition metals by surface complexation (Angelomé and Soler-Illia, 2005). A scheme of the functionalization reactions is presented in Supplementary Figure S2.

To perform the silanization, the samples were first cleaned with ethanol, dried, and then immersed for 10 min in a 9% w/w solution of TMCS in toluene. To accelerate the functionalization reaction, a few drops of DIPEA were added to the solution. Functionalization with DHDP was carried out by immersing the samples in a 0.01 M DHDP solution in THF. For single and bilayer samples, the immersion time was 2 h, following previously established protocols (Angelomé and Soler-Illia, 2005; Fuertes, 2009). In the case of multilayer systems, the immersion time was extended to 24 h to ensure uniform functionalization throughout the porous network. When dual functionalization was required, TMCS modification was performed prior to DHDP functionalization, in order to avoid the potential removal of DHDP molecules by toluene. A flowchart showing the steps involved in the functionalization of the PCs is presented in Supplementary Figure S3. The nomenclature adopted for the multilayers analyzed in this study is summarized in Table 1.

Porosity and thickness were determined using X-ray reflectometry (XRR) with a Panalytical Empyrean X-ray diffractometer and CuK_α incident radiation (λ = 1.54 Å) (van der Lee, 2000). A specially designed chamber was used to control the RH in the sample environment in order to obtain the porosity. A full description of the method can be found elsewhere (Fuertes et al., 2009; Steinberg et al., 2019).

Infrared spectroscopy was used to characterize the chemical modifications introduced by surface functionalization. Measurements were performed using NICOLET Magna 560 or NICOLET 8700 instruments equipped with liquid N₂ cooled MCT-A detectors. Transmission mode was employed to acquire spectra of samples deposited on glass or fused silica substrates in the 3,500–2,500 cm^-1 range. In this case, the samples were directly mounted onto the sample holder. Diffuse reflectance mode (Diffuse reflectance infrared Fourier transform spectroscopy, DRIFTS), on the other hand, was used to determine the samples full spectra (4,000–600 cm^-1 range). In this case, the samples were scratched from the substrate and mixed with KBr to perform the measurements.

Modifications in the hydrophobicity of the synthesized materials were determined by liquid-solid contact angle (CA) measurements. A Ramé-Hart 290 equipment was used, and the acquired images were processed with DROPimage Advanced V25 software. All samples were dried at 130 °C before each measurement.

The multilayer structure and chemical composition before and after post-synthesis modifications was characterized by transmission electron microscopy (TEM) coupled with Energy Dispersive Spectroscopy (EDS), using a Philips CM 200 microscope operated at an accelerating voltage of 180 kV coupled with an EDAX^® detector. Samples were prepared by scratching the photonic crystals and dispersing the resulting fragments onto carbon-coated copper grids.

The performance of the different photonic crystals as optical devices for the detection of non-polar organic vapors was evaluated. The transmission spectra of the multilayer systems were measured by UV-visible spectroscopy using an Agilent 8453 diode array spectrophotometer. Measurements were taken on samples subjected to a low RH atmosphere (dry conditions, between 5% and 10% RH) and in a H₂O (n₆₃₀ = 1.333; P_v = 15.7 kPa at 55 °C, being n₆₃₀ the refractive index at λ = 630 nm and P_v the vapor pressure) (Sienko and Plane, 1967; Haynes, 2003) saturated atmosphere, and exposed to organic solvents saturated atmospheres. Measurements were carried out both under static equilibrium conditions and dynamically at regular time intervals, to evaluate the response kinetics in environments containing either pure vapor solvents or mixtures of polar and non-polar solvents. The organic solvents used were toluene (n₆₃₀ = 1.496; P_v = 15.3 kPa at 55 °C) and benzyl alcohol (n₆₃₀ = 1.540; P_v = 0.13 kPa at 55 °C). These organic compounds were selected because they exhibit similar refractive indices but markedly different vapor pressures and physicochemical properties. Water was included to assess the impact of ubiquitous ambient humidity on the device response. Prior to use, the organic solvents were dried using 3A and 4A molecular sieves.

The performance was evaluated by analyzing the response times for each pure solvent and for the solvents’ mixture, defined as the minimum time in which the greatest shift of the photonic band gap (λ_B) is reached (Sansierra, 2019). To carry out the experiments, the samples were placed inside a quartz cuvette mounted in a sample holder at 55 °C. For the pure solvent tests, a filter paper soaked with the solvent of interest was placed inside the cuvette. In the case of solvent mixtures, two filter papers were introduced simultaneously, one imbibed with water and the other with the non-polar solvent. Prior to each experiment, the samples were rinsed with ethanol and dried in an oven at 130 °C for 10 min to ensure the complete removal of the solvents and water from the mesopores.

TEM images of the multilayers presented in Figure 4; Supplementary Figure S7. Since Zr has a higher electron scattering cross-section than Si, the ZF layers appear as dark stripes in the images, while SC layers appear lighter. Layer thicknesses are very uniform, as shown in our previous studies of multilayer structures fabricated using similar methods (Fuertes et al., 2007; Sansierra et al., 2019; Morrone et al., 2021). Typical roughness is below 2 nm. The mesopores are visible within the stripes as brighter regions. The images reveal well-defined stacked SC and ZF layers, with no evidence of interpenetration between the materials. Additionally, no film cracking is observed.

For all layers, the thicknesses estimated from the TEM images are approximately (40 ± 8) nm, which corresponds to the target values for constructing PCs with a band gap in the UV region using this combination of porous oxides. After sequential functionalization with TMCS and DHDP, the layered structure remains intact. In fact, the micrograph of the PC-TMCS + DHDP sample shows no significant changes in layers’ thicknesses or porosities compared to the unmodified PC (see Figure 4).

The incorporation of surface chemical modifications was studied by EDS and infrared spectroscopy. Figure 5 shows the DRIFTS spectra of the multilayer system before functionalization, after treatment with TMCS or DHDP, and after sequential functionalization with both reagents. Upon reaction with TMCS, a new absorption band appears at 850 cm^-1, corresponding to the ν_Si–C vibrations characteristic of the trimethylsilyl group. Additionally, a band at 2,960 cm^-1 is observed, associated with the ν_C–H stretching vibrations of CH₃ groups. These features confirm the successful surface modification of the mesoporous PC. Notably, this latter band is not detected in single- or bilayer systems, likely because the multilayer system contains four stacked layers, increasing the overall signal intensity. For DHDP functionalization, the appearance of ν_C–H stretching bands characteristic of alkyl chains is observed in the 2,800–3000 cm^-1 region. After further modification of the PC–TMCS sample with DHDP, the 2,960 cm^-1 band overlaps with the DHDP signals in this region. The simultaneous presence of all these bands confirms the successful incorporation of both functionalizing agents.

To confirm the penetration of DHDP within the multilayer structure, additional FTIR measurements in transmission mode were performed on single layers, bilayers, and multilayers deposited on glass or fused silica substrates (Supplementary Figure S6). The results show that, when normalizing by the layer’s thicknesses, the amount of DHDP incorporated into the 8-layered device (containing 4 ZF layers) is nearly three times higher than that observed for a single ZF layer. This indicates that DHDP molecules infiltrate nearly all the entire accessible pore volume. However, the total DHDP content in the multilayer is slightly lower than the theoretical maximum (assuming all ZF layers were filled to the same extent as a single layer) likely due to diffusion limitations within the more complex multilayer architecture. As the upper layers begin adsorbing DHDP, the accessibility of deeper ZF layers decreases, making it progressively more difficult for DHDP to diffuse and reach the bottom layers.

Once the presence of the organic functions throughout the entire PC was confirmed, the specificity of the functionalization was addressed. To this end, EDS spectra were acquired coupled with TEM imaging. Spectra from specific regions of the PCs before and after functionalization were obtained, as shown in Supplementary Figure S7. The incorporation of DHDP can, in principle, be tracked by detecting phosphorus. However, in these samples, the presence of P in ZF layers could not be unambiguously identified due to the overlap between the K-α line of P and the L-α line of Zr, combined with the relatively low P content compared to Zr. Nevertheless, no P signal was detected in the SC layers after functionalization. Since the presence of DHDP was independently confirmed by infrared spectroscopy, it can be concluded that this molecule is specifically attached to the ZF layers. Unfortunately, due to the resolution limits of EDS and the small thickness of the ZF layers, isolated spectra of individual layers could not be obtained. As a result, silicon appears in both functionalized and non-functionalized samples, preventing the confirmation of TCMS functionalization specificity by this technique.

Subsequently, the surface physicochemical properties were evaluated through CA measurements (Table 3). The non-functionalized multilayer exhibits a higher CA (52°) than the single-layer films. This is expected, as the thicker porous PC is less polarizable than the glass substrate, which lies closer to the fluid–air–solid interface in single-layer systems. Functionalization of the PC with TMCS does not result in a significant change in the CA. In contrast, derivatization with DHDP leads to a 44% increase in CA. These results are consistent with the fact that the topmost layer is ZF, which reacts selectively with DHDP. Finally, when the PC is sequentially functionalized with both TMCS and DHDP (PC–TMCS + DHDP system), the resulting CA is comparable to that of the PC–DHDP system. Thus, the hydrophobicity of the multilayer system can be enhanced through appropriate surface functionalization, thereby improving the selectivity of the device.

Finally, the devices’ optical behavior was evaluated. The as-prepared multilayer exhibits a photonic band gap centered at λ_B = 270 nm, with an average full width at half maximum of approximately 100 nm and a maximum reflectance close to 90% at λ_B, as can be seen in Figure 6. This remarkably high reflectivity is achieved with only 8 layers, owing to the high refractive index contrast between the oxides. A precise determination of the band gap width is limited by the onset of UV absorption below 225 nm from the constituent materials of the PC. All systems retain their photonic band gap after functionalization, with no changes in either the central wavelength or the bandwidth. The minor differences in λ_B values observed among all the dry PCs are similar to those found when sampling different areas of the same PC and are attributed to minor thickness variations inherent to the synthesis method. Therefore, to compare the response of these different PCs to vapors, the total displacement of the band gap position (Δλ_B) is analyzed rather than the absolute initial and final value of λ_B.

The optical response of the PCs to different analytes was evaluated using transmission UV-Vis spectroscopy in controlled atmosphere. Spectra were acquired under dry conditions and in atmosphere saturated with the vapors of different solvents. Measurements were performed both at equilibrium conditions and dynamically, to elucidate the response kinetics. In equilibrium experiments, the partial pressure of each solvent corresponds to its equilibrium P_v at the studied temperature (55 °C for all experiments). In the dynamic experiments, the partial pressure of the solvent gradually increases until it reaches the equilibrium value.

First, the water uptake in the different devices was determined. The results are presented in Figure 6 and Supplementary Table S2, SI. Upon exposure to water vapor, the non-functionalized 4X (SC/ZF) system is responsive, and the photonic band gap shift is approximately 16 nm. After derivatization, the systems remain responsive; however, the shift of λ_B upon water vapor exposure is significantly reduced. In systems functionalized only with DHDP, the λ_B shift decreases by about 50% compared to the non-functionalized PC. In contrast, systems modified with TMCS, either alone or in combination with DHDP, exhibit an even more pronounced reduction, with the λ_B shift diminishing by approximately 90%.

Therefore, functionalization has a marked effect on water condensation within the films. DHDP limits water condensation, presumably within the ZF layers, restricting it mainly to the SC layers. This partial condensation leads to a reduced refractive index contrast and an increase in the average refractive index, resulting in a redshift of the band gap and a decrease in reflectivity. In contrast, functionalization with TMCS produces a much stronger effect, significantly limiting water condensation and resulting in minimal band gap shifts.

Capillary condensation in mesopores can be understood using the Kelvin equation (Boissiere et al., 2005; Butt and Kappl, 2010). The equilibrium vapor pressure of a liquid condensed in a pore is governed by the curvature of the meniscus formed. This curvature is determined by the characteristic pore size and by the contact angle of the liquid-solid interface. Therefore, liquids will condense at lower partial pressures on smaller pores and more wettable surfaces than in larger pores and surfaces that show less favorable interactions with the condensing phase (see Supplementary Figure S8).

ZF layers have a larger pore size than SC layers, as previously described; therefore, SC pores are expected to fill with liquid water at a lower water partial vapor pressure (Fuertes et al., 2008). After functionalization with DHDP, the surface wettability of ZF layers decreases, producing a synergistic effect with pore size that enhances the differences between both layers. As a result, water condenses preferentially in SC layers, while capillary condensation does not occur in the non-wetting, functionalized ZF layers. In contrast, functionalization with TMCS has an antagonistic effect relative to pore size, as the layers with smaller pores now exhibit non-wetting behavior. Furthermore, the contact angle on ZF surfaces is also affected by dehydration, as previously discussed, effectively preventing pore filling.

Overall, the obtained results indicate that the incorporation of the tested functional groups prevents water condensation and, consequently, may render the PC response more resilient to ambient humidity. Thus, once the water uptake behavior under equilibrium conditions was determined, the dynamic optical response of the different multilayer systems to water and two non-polar organic vapors was evaluated. The PCs were exposed to atmospheres containing water, toluene, or benzyl alcohol vapors, and UV-Vis transmission spectra were recorded at regular intervals. The results are presented in Figure 7 and Table 4.

First, the response times were compared, revealing clear differences depending on the solvent used. Notably, the response to all vapors was very fast before the chemical modification: the maximum photonic band shift occurred in under 2 min. After the functionalization, responses times as short as 9 s were measured for toluene. In the case of exposure to water vapor, the PC-DHDP system showed a 65% increase in response time compared to the non-functionalized PC while for TMCS-functionalized system the response was undetectable due to the significantly reduced accessibility of water molecules to the multilayered structure. Nevertheless, the functionalized multilayers remained responsive to toluene and benzyl alcohol vapors, confirming that the chemical modifications did not block the pores and that the porosity remained accessible.

When comparing the two organic solvents, the response to toluene was consistently faster than to benzyl alcohol (see Table 4). Upon exposure to toluene vapors, all systems exhibit very short response times (around 10–20 s). In contrast, exposure to benzyl alcohol results in response time between 1 and 2 min, depending on the functionalization. Since the optical response in these mesoporous systems originates from capillary condensation within the pores, the characteristic response time is associated with the time required for the solvent’s partial pressure inside the pores to reach the threshold for capillary condensation. From the solvent perspective, lower P_v and higher molecular weight reduce diffusion rates, also extending the time required for condensation: this is the case of benzyl alcohol compared to toluene. This behavior is consistent with previous observations in mesoporous multilayer systems based on titania and silica films with photonic response in the visible region of the spectrum (Sansierra et al., 2019).

Besides, from the perspective of the porous material, less wettable surfaces (as can be obtained using the functionalization agents) and larger pore diameters (in these multilayers, in the ZF layers) demand a higher partial pressure for condensation, thus requiring longer times to complete the filling with the solvent.

After functionalization with DHDP, the strong interactions between the alcohol group of the benzyl alcohol and the ZrO₂ surface are replaced by weaker interactions with the aliphatic chains of the phosphate. As a result, higher solvent partial pressures are required to trigger capillary condensation, leading to longer response times: from 1 to 2 min.

By the other hand, TMCS reacts selectively with the SC films, which have smaller pore sizes and therefore require lower partial pressures for condensation; in this case, the changes in surface wettability are not sufficient to outweigh the pore size effect. Therefore, the observed difference in the time response between benzyl alcohol and toluene detected using both modified PCs can be attributed to the presence of the interaction between the DHDP and the hydroxyl group in benzyl alcohol.

It is noteworthy that the maximum shifts of the photonic band gap for both organic solvents are equivalent for each device (Figure 7E). These solvents exhibit similar refractive indices and thus produce comparable photonic band shifts in this type of optical device (Sansierra et al., 2019).

After analyzing the behavior of the devices detecting the pure solvents, the dynamic displacement of the photonic band gap in a mixture of solvents was studied. These tests aimed to evaluate the PCs responsiveness in a mixed vapor environment, simulating the detection of a non-polar organic compound under high humidity conditions. The results obtained for the mixture of water and benzyl alcohol are presented in Figure 8.

First, the change in the optical properties of the dry non-functionalized PC exposed to an atmosphere containing both water and benzyl alcohol vapors is shown in Figure 8A. During the initial seconds, the system responds predominantly to water, given its higher P_v, resulting in a behavior similar to that observed under pure water vapor. As the environment becomes saturated with benzyl alcohol (approximately after 20 s), the solvents adsorption dynamics change slightly. Nevertheless, the final shift in the photonic band gap is close to that obtained under pure water vapor. This indicates that, in the absence of surface functionalization, the presence of water in the environment diminishes the photonic crystal’s response to benzyl alcohol vapors compared to the response observed under pure benzyl alcohol.

The same experiment was then performed using the double-functionalized PC (Figure 8B). In this case, the optical response in the mixed solvent environment closely resembles that observed under exposure to the pure non-polar solvent. The amount of water condensed within the functionalized PC is minimal, allowing benzyl alcohol to be detected within just a few seconds from the start of the experiment. Notably, the presence of water vapor in the atmosphere does not hinder the detection of the non-polar solvent, even though benzyl alcohol has a lower P_v than water.

Figure 8C shows the shift of the photonic band gap measured once equilibrium is reached for both pure vapors and their mixture, comparing non-functionalized and functionalized PCs. For the solvents mixture, the maximum shift is slightly lower than that observed with pure benzyl alcohol. This reduction is attributed to the presence of a small amount of condensed water, which has a lower refractive index than the non-polar solvent.

A similar behavior was observed for the water-toluene mixture, as shown in Supplementary Figure S9. In this case, both solvents exhibit comparable vapor pressures, further supporting the conclusion that water does not interfere with the detection of non-polar solvents in this type of functionalized detectors.

Finally, these UV optical devices showed excellent chemical and mechanical robustness. Each one was tested several times (at least 40) under different conditions, and consistently delivered reliable results without any observable damage or loss of function. This highlights a clear advantage of using mesoporous oxides in their construction: their strong resistance to chemical and mechanical stress (Sansierra et al., 2019; Ramallo et al., 2022). Furthermore, even after repeated use, the devices maintained a fully reversible optical response simply by drying them in an oven at 130 °C for 15 min between exposures to different solvents. This makes them especially promising for long-term use in UV sensing applications, where stability, durability and UV resistance are critical.