Introduction

Photonic-integrated devices are very attractive for different applications such as biosensing [1,2,3] or telecommunication [4]. Therefore, their fabrication technologies have been extensively studied and developed in recent years. Considerable progress has been made in this field, though there are still many challenges that researchers have to face, mainly connected to low-cost fabrication of optical components, including active ones, with low optical losses [5, 6]. This also applies to microstructures such as microdiscs or microgoblets (under-etched microdiscs), which are particularly attractive in light amplification [7,8,9,10,11,12,13] and sensing [14] applications. Developing new optical materials and their structurization techniques can address the above issues.

The structurization protocols are well-established for polymer materials [7, 9, 12,13,14] or silica-based materials [8, 10, 11] enabling for the fabrication of microstructures doped with organic dyes or rare earth ions, respectively. However, organic- and silica-based materials are usually characterized by relatively low RI, which can decrease the light confinement in photonic structures on commonly used substrates (e.g., soda-lime glass, silica). In the case of organic materials, an increase of the RI may require complex synthesis of sophisticated materials [15]. On the other hand, silica-based systems enable relatively simple tuning of the RI in the range of 1.2–2.2, by the addition of high refractive index metal oxides such as TiO2 [6].

Sol–gel materials based on TiO2 and SiO2 have attracted much attention in terms of their potential application in planar photonics. Their advantages, when compared to such platforms as InP, silicon on insulator (SOI), and SiN, include low cost, simple preparation, transparency in the visible spectral range, and relatively low optical losses [5, 6]. However, fabrication of an active luminescent component with a high RI is still not an easy task due to the high-temperature treatment required to obtain the SiO2-TiO2 materials (usually above 500 °C), which in most cases leads to the degradation of an organic dye.

A number of papers describing the fabrication and optical properties of organic dye-activated sol–gel layers based on TiO2 and SiO2 precursors and prepared at lower temperatures is very limited [16,17,18,19]. Materials based on TiO2 alkoxides and ORMOSIL precursors were often used in the preparation of luminescent layers [20,21,22,23,24,25,26]. In our recent paper, it was shown that the use of a sol–gel matrix based on TiO2 alkoxide and ORMOSIL precursor can improve the stability of the RhB dye under heat treatment conditions compared to the layer with non-modified SiO2 alkoxide precursor [25]. Moreover, our fabricated layers were suitable for transferring periodic structures (serving as grating couplers) using nanoimprint lithography (NIL).

Generally, in the case of sol–gel materials, different types of patterning techniques, such as inductively coupled plasma reactive ion etching (ICP-RIE) and electron cyclotron resonance plasma etching (ECR) [27, 30], wet-chemical etching [30,31,32], direct photolithography (PL) [33,34,35], electron-beam lithography (EBL) [36, 37] or X-ray lithography (XRL) [36, 38,39,40,41,42,43], and other such as XeF2 chemical etching [7,8,9, 11, 44], micromolding in capillaries (MIMIC) [45], and NIL [25, 46,47,48,49], can be implemented.

In the past, layers based on (3-glycidyloxypropyl)trimethoxysilane (GLYMO) and 20 mol% of TiO2 alkoxide were subjected to etching using RIE. However, some irregularities in the form of chain-like structures [27] or significant roughness [30] appeared on the waveguides formed after the etching processes. Using the same technique, the fabrication of channel waveguides with much better overall quality was possible on films based on GLYMO and GeO2 precursors [28]. In the other report, wet-chemical etching using diluted hydrofluoric acid (HF) was implemented to obtain ridge waveguides in layers based on TiO2 alkoxide and GLYMO [30, 31].

Alternatively, other materials were employed directly as negative photoresists [33, 34]. The waveguides with rounded sidewalls were obtained after photoprocessing the sol–gel films based on tetraethyl orthosilicate (TEOS), titanium(IV) butoxide, and vinyltriethoxysilane [34]. In the second article, sol composed of GLYMO, GeO2 alkoxide precursor, and photoacid generator exhibited additionally a UV-induced change of the RI and was used for the transfer of surface relief gratings using the thermal nanoimprint technique (TNIL) [34]. Finally, ring resonators were fabricated using UV-curable sol–gel by so-called dual-step soft NIL (DSS-NIL) [50]. However, the lack of the metal alkoxides or their low content in the above-mentioned materials resulted in their relatively low RI not exceeding 1.55.

Higher RI microstructures were obtained from amorphous TiO2 layers prepared using low-temperature heat treatment not exceeding 200 °C and wet-chemical etching combined with chemical etching using XeF2 gas [44]. The resulting under-etched microdisc and microgoblet structures successfully served as optical resonators. Recently, we also studied the direct patterning of sol–gel-derived layers based on TiO2 precursors using an electron beam, obtaining dense periodic structures that may be potentially used for grating couplers [37]. These materials were not activated with organic dyes.

Patterning using lithographic and etching techniques was also presented for luminescent sol–gel layers but only for materials not exceeding an RI of 1.53. The rhodamine 6G-doped layers deposited from sols composed of GLYMO and GeO2 precursors were used as negative tone resists for EBL and XRL, among others, for the fabrication of fluorescent microstructures, e.g., pillars [36]. XRL was also implemented for the structurization of silica layers doped with rhodamine 6G. In this case, the exposition of the dye-doped layer to the X-rays caused its simultaneous densification and degradation of the dye. Alternatively, patterning of the surface with functional groups (–NH2), which allowed for selective binding of fluorescein isothiocyanate in the non-exposed areas, was performed [40]. Round-wall channels with a width of 30 µm and depth of 12 µm were prepared by deposition of ZrO2-ORMOSIL films containing near-infrared (NIR) emitting styryl dyes (LDS) in wet chemically etched glass substrates [32]. Finally, simple strip waveguide structures were fabricated using ICP-RIE of rhodamine 6G-doped sol–gel silica layers [29].

All in all, research on structurization techniques of sol–gel layers is still required to obtain high-quality under-etched photonic microstructures with high RI and detected luminescent properties, which is of great importance for developing low-cost technologies for photonic integrated circuits with active optical components.

Therefore, we continue the topic of the fabrication of luminescent photonic microstructures derived from sol–gel materials of relatively high RI. Particularly, this work is focused on the investigation of the structurization potential of the luminescent hybrid sol–gel layers previously explored by us [25]. For that, wet-chemical etching of the luminescent RhB-doped layers based on titanium(IV) alkoxide and ORMOSIL is studied. Firstly, the stability of the layers against common solvents and reagents used in the photolithography process is established. The impact of their annealing at 200 °C on the thickness, RI, luminescent properties, etching rate, and selectivity of the etching is determined. Moreover, the stability of the layers against aging under ambient conditions is investigated. The structurization protocol of these layers using photolithography and wet-chemical etching with BHF has been successfully implemented, leading to the fabrication of microdisc and microgoblet structures.

Materials and methods

Sol synthesis and film preparation

Synthesis of the sol

All chemicals were used without further purification. Sols were synthesized using TET (Sigma-Aldrich, for synthesis), acetylacetone (AcAc, Fluka Analytical, ≥ 99.5%), GLYMO (Fluorochem), EtOH (Chemsolute®, 99.9%), deionized H2O with conductivity not higher than 0.1 μS/cm from water deionizer HLP 5UV (Hydrolab), HCl (Chempur, 35–37%), and RhB (J&K Scientific Ltd., 95%). The sol–gel preparation protocol was based on our previous report [25]. Briefly describing, the sol was prepared from ORMOSIL sol, where a molar ratio of GLYMO: EtOH: H2O: HCl equaled 1: 4: 1.5: 0.036 and titanium(IV) alkoxide sol. Here, a molar ratio of TET: EtOH: AcAc was equal to 1: 10: 0.5. The final sol was prepared by mixing those sols at the final molar ratio of TET: GLYMO was equal to 1: 1. RhB was added at the stage of preparation of ORMOSIL sol as EtOH solution (10−2 M in EtOH) in the amount providing molar ratio of RhB to the sum of the TET and GLYMO precursors in the final sol equal to 0.05%.

The sol–gel film preparation

Sol–gel layers were deposited on soda-lime glass or silicon substrates using a DC single M dip-coater (Bioling Scientific). Before deposition, the substrates were subjected to a cleaning procedure, which (depending on the type of the substrate) was as follows:

  • For soda-lime glass: the substrates were rinsed for 1 min in deionized water, then gently cleaned with a sponge with Trilux detergent (Analab), rinsed again for 1 min in deionized water, and dried with the stream of nitrogen gas;

  • For silicon and silica on silicon: the substrates were rinsed for 1 min in deionized water, then rinsed in acetone and isopropyl alcohol, and again for 1 min in deionized water, and dried with the stream of nitrogen gas.

Both types of substrates were then immersed for 10 min in Piranha solution (H2SO4: H2O2 = 4: 1, V: V), rinsed with water, dried with nitrogen gas, and subjected twice to the O2 plasma cleaning for 5 min (PVA TePla 300 W). Please note that Piranha solution should be prepared and handled with particular caution (under the fume hood only in glass or Pyrex containers). During preparation, H2O2 should be always slowly added to H2SO4 (not in reverse). Additionally, such solution must be used for cleaning only surfaces, which are free of organic solvents. Fast mixing of Piranha solution components or its use for samples contaminated with large amounts of organic solvents can lead to exothermic reactions causing uncontrolled boiling of the solution. After its use, the solution should be allowed to cool overnight under the fume hood in an open container. Then, it should be transferred to a clean, empty glass bottle and loosely capped to avoid possible build up of the pressure inside.

For the initial thickness test, the films were deposited using a withdrawal speed in the range of 50–100 mm/min from the freshly prepared sol and heated on the hot plate set to 200 °C for 10 min. Samples designated for investigation of the heating time effect on their properties (for the spectroscopic measurements, photolithography, and wet-chemical etching processes) were withdrawn at 80 mm/min from the sol–gel and heated in a laboratory dryer at 200 °C for various times from 1 to 24 h.

Structurization

Photomask design

The photomask for photolithography was designed based on numerical calculations. 2D finite difference time-domain (FDTD) method implemented in Lumerical [51] was used to study electric field resonances in the spectra of sol–gel (RI = 1.67) microdisc resonators (surrounded by air) of varying radii (Fig. S1). The resulting design of the photomask is presented in Fig. S2.

The samples cleaning

After the deposition of sol–gel layers and their heat treatment, they were subjected to a cleaning procedure involving rinsing in deionized water for 1 min, rinsing in acetone and isopropyl alcohol for around 5 s, and again in water for 1 min. Finally, the samples were dried in the stream of nitrogen gas and then dried in the laboratory dryer at 150 °C for 30 min. After drying, the samples were cooled to room temperature (RT) and used in the photolithography process.

Photolithography process

The sol–gel layers on the soda-lime glass substrates were homogenously coated with an adhesion promoter AR 300–80 (Allresist GmbH), followed by a heat treatment on the hot plate (2 min at 180 °C). A negative tone photoresist AZ 15nXT diluted in AZ EBR solvent in a volume ratio of 1: 1.2 (both provided by MicroChemicals) was then coated on the samples (4000 rpm for 40 s) and soft-baked at 110 °C for 2 min. The photolithography process was carried out in a hard contact mode using a mask aligner (Suss MicroTec MA8/BA8 with 1000 W UV lamp) and a chromium photomask. After the exposure (D = 60 mJ/cm2), the samples were baked at 120 °C for 1 min and developed for 12–14 s in AZ 2026 MIF developer (MicroChemicals), followed by immersion in deionized water for 1 min. The typical height of the photoresist structures after the process was around 420 nm.

Wet-chemical etching

The wet-chemical etching was performed in the Teflon Petri dish using around 40 ml of a buffered hydrofluoric acid (Buffer HF improved, Sigma-Aldrich). After the sample immersion, it was held in the etching solution for a desired time without stirring. Next, the sample was taken out and immediately immersed in the deionized water for 3–5 s. Then, it was rinsed with a stream of deionized water for 1 min and dried with nitrogen. Particular care should be taken when working with BHF solution, including the usage of proper personal protective clothing and HF-resistant labware, as well as the availability of calcium gluconate gel and HF spill kit close nearby.

To remove the photoresist after wet-chemical etching of the sol–gel layer, the samples were subjected to an ultrasonic bath (US) in dimethyl sulfoxide—DMSO (anhydrous, VWR Chemicals BDH®).

Films and structures characterization

Microscopy images

Optical microscopy images of the samples were registered using a DM8000 optical microscope (Leica) in reflection mode using white or UV light illumination. SEM images were acquired using the Helios NanoLab microscope (FEI, currently Thermo Fisher) equipped with a focused ion beam (FIB) column and gas injection system (GIS). The SEM imaging was performed with low accelerating voltage (typically 2 kV). The through-the-lens detector (TLD) was applied to obtain high-resolution images. The FIB module was used to prepare cross sections of selected samples. Platinum layers were deposited locally using the GIS and electron-beam patterning to protect microstructures during FIB cross-sectioning.

Thickness and RI determination

The thicknesses of the deposited photoresist and the sol–gel layers, as well as the height of the structures after wet-chemical etching, were determined using the contact profilometer DektakXT (Bruker). Additionally, the dispersion of the RI, along with the thickness of the sol–gel layers, was evaluated using SE 850 DUV spectroscopic ellipsometer (Sentech) and SpectraRay software by employing Cauchy’s formula. Measurements were performed in the spectral range of 400–930 nm. The RI value was given for λ = 632.8 nm in the results and discussion section.

Photoluminescence spectroscopy

Photoluminescence spectra of the sol–gel films were collected using an FLS980 spectrofluorimeter (Edinburgh Instruments) with a xenon arc lamp light source.

Fourier-transform infrared spectroscopy (FT-IR)

Infrared (IR) absorption spectra were acquired for the sol–gel layers deposited on Si substrate in the range of 4000–400 cm−1 using Nicolet 6700 FT-IR Spectrometer (Thermo Fisher Scientific) equipped with Smart Orbit ATR Accessory. For clarity, the spectra are presented in the two spectral regions of interest 3600–2400 cm−1 and 1400–400 cm−1. Spectra of the selected layers after BHF etching were recorded after etching for 10 s and subsequent rinsing with a stream of deionized water for 1 min and drying with nitrogen.

Results and discussion

Technological process of the sol–gel layer structurization

In the dip-coating deposition process of the sol–gel, the thickness of a layer was controlled by changing the withdrawal speed of a substrate from the sol. The resulting characteristic for the films deposited on soda-lime glass substrates together with a photograph of one of the obtained films, are presented in Fig. 1.

Figure 1
figure 1

Dependence of the thickness of the sol–gel film on withdrawal speed. The inset shows a photograph of the layer obtained on a soda-lime glass substrate at 80 mm/min (approx. dimensions of the sample were 2.5 × 4 × 0.1 cm3).

Full size image

The above dependence was determined by measuring the thicknesses of the layers obtained from a freshly prepared sol on soda-lime glass and heated at 200 °C on the hot plate for 10 min. The sol–gel layer prepared under similar conditions (withdrawal speed of 80 mm/min) but on a silicon substrate with a 3-µm-thick SiO2 layer (Si/SiO2) was of comparable thickness, i.e., around 450 nm. However, after around a month of the sol aging under ambient conditions, the thickness of layers (deposited with the speed mentioned above on both types of substrates) reached around 540 nm. It was connected with the increased sol viscosity resulting from the ongoing hydrolysis and condensation reactions. This effect is known for the sols [52], yet very important from the technological point of view. The shelf life of the sol is long enough to allow its repeated use for layer deposition. Long shelf life can be connected with the presence of AcAc in the sol. This complexing agent is known for its ability to react with titanium(IV) alkoxides, thus reducing their reactivity [53]. What is crucial is that the overall homogeneity of the layer’s thickness deposited from the aged sol was very good, which was estimated from the uniform interference color of the film. For further structurization experiments, only the samples obtained with the withdrawal speed of 80 mm/min were used. The technological process of microstructure fabrication consisting of several steps is schematically presented in Fig. 2. In the shown procedure, first the luminescent layers were coated with an adhesion promoter and then with a photoresist. After UV exposure of the photoresist layer, it was developed to dissolve its non-exposed areas.

Figure 2
figure 2

Scheme of the technological process: I photoresist layers deposition; II selective photoresist exposure through photomask and development; III sol–gel layer wet-chemical etching and the photoresist removal; A–C microstructures of different depth obtained by varying the etching time.

Full size image

Exposure dose and developing time optimization enabled the fabrication of photoresist structures with a separated waveguide and disc structures (process II in Fig. 2). Etching of the sol–gel layer was possible using BHF (process III in Fig. 2). When the non-masked areas of the sol–gel film were exposed to the BHF etching solution, they were easily removed due to the chemical reaction—similarly as for silica glass, despite the presence of ORMOSIL precursor in the sol–gel matrix. However, as a consequence of using the above precursor, the removal of the remaining photoresist could not be done in the same manner as in the case of processing inorganic materials (using Piranha solution) because of the degradation of the sol–gel film. Thus, the photoresist removal process had to be optimized to ensure its complete stripping from the sol–gel layer while preserving the latter intact. This was done by soaking the wet chemically etched samples with the sol–gel and photoresist test structures („gap1”) in a DMSO bath with ultrasonication (US). The results are presented in Fig. 3.

Figure 3
figure 3

Microphotographs of the sol–gel layer after the photolithography and wet-chemical etching processes: a before photoresist removal, b after DMSO treatment (without US), c after DMSO treatment in US bath, and d thickness and RI of the sol–gel layers heat-treated at 200 °C for different duration time (measured before and after the treatment with DMSO for 15 min).

Full size image

As presented above, the soaking of the etched samples with photoresist in DMSO did not result in a complete removal of the photoresist (Fig. 3b). This observation was valid even for longer soaking periods of up to 30 min. However, the photoresist was entirely stripped after the application of the DMSO bath for at least 15 min together with US (Fig. 3c). Structures of overall good quality were obtained after wet-chemical etching of the sol–gel layer and removal of the photoresist using DMSO with the application of US. Any undesirable layer dissolution was not observed, even during prolonged DMSO treatment (up to 30 min). Moreover, the thickness and RI of sol–gel layers deposited on Si and heated at 200 °C (in a laboratory dryer) for different times were investigated before and after the DMSO treatment (for 15 min) (Fig. 3d). It can be seen that even the layer which was not subjected to any heat treatment (prepared at RT) had good stability against DMSO and no dissolution of this layer was observed. Its thickness and RI did not change significantly after being taken out of the DMSO bath. Similarly, the determined thickness of the other samples heat-treated at 200 °C for one, two, and four hours did not change after keeping them in DMSO.

Wet-chemical etching

The mask implemented for the structurization of the sol–gel layer contained disc-shaped structures of various diameters and strip waveguides together with disc structures (with or without a gap between them). Initial structurization experiments were performed on the sol–gel layers of thickness around 450 nm (heated at 200 °C for 10 min). Depending on the etching time in BHF, disc-shaped structures with various etching depths could be obtained, as presented in Fig. 4.

Figure 4
figure 4

SEM images (tilted view) of the disc-shaped sol–gel structures obtained after etching in BHF for different times: a 28 s, b 50 s, c 120 s; d dependence of the sample etching depth on the BHF etching time, and e SEM–FIB cross-sectional view of the sol–gel disc etched for 120 s.

Full size image

Depending on the etching time (28 s, 50 s, 120 s), three different etching scenarios were found (Figs 4a–c). For the shortest etching time (28 s, Fig. 4a), the etching depth was lower than the thickness of the sol–gel layer. This allowed us to obtain structures partially etched in the layer. Elongation of the etching time led to further active layer removal, and at the etching time of around 50 s (Fig. 4b), it was observed that the etching depth was similar to the thickness of the layer. In the third case, the prolonged etching time (120 s, Fig. 4c) led to the complete removal of the active layer in non-masked areas and further etching of the soda-lime glass substrate. It is worth underlining that the surface and sidewalls of the etched sol–gel disc are rather smooth without any visible irregularities. This is in contrast to the sol–gel layers of similar composition (based on GLYMO and TiO2 precursors) etched using RIE, where chain-like structures [27] or roughness [30] were visible after the etching process, reducing their homogeneity.

However, from Fig. 4d, it can be deduced that the etching rates of the sol–gel layer and the substrate are not significantly different. As it can be seen from the FIB cross-sectional view of the sample etched for the longest time (Fig. 4e), the under-etching (i.e., lateral etching under the layer) is not visible, which confirms that the difference between the etching rates of the substrate and the sol–gel layer was small and thus not enough to obtain goblet-shaped microstructures. When looking at the fabricated waveguide structures, we observed that they were preserved only in the case of the etching for 28 s and 50 s (Figs 5a and S3a–b). Due to the isotropic nature of the wet-chemical etching of glass and the sol–gel material, the waveguide structures detached from the sample etched for the longest time (Fig. S3c).

Figure 5
figure 5

a SEM image (tilted view) of the sol–gel waveguide and disc on the sample etched in BHF for 28 s. b Emission spectra of the sol–gel layers after BHF etching for 28 s, 50 s, and 120 s. Insets show the same spectra after normalization, the emission spectrum of the layer before etching (a gray area), and photographs of the samples under UV lamp excitation (λmax = 366 nm).

Full size image

The fluorescent dye present in the sol–gel matrix was protected from the aggressive etchant, and no leaching or degradation of the dye was noted in the etched samples. To confirm that, the samples were investigated using photoluminescence spectroscopy. The emission spectra and photographs of the samples under UV light are presented in Fig. 5b.

The samples’ fluorescence was still visible (under UV light) after the etching. For the sample etched for the shortest time (28 s), the fluorescence was visible on the whole area of the sample, which is consistent with the determined depth of the etching, which was lower than the layer thickness. On the contrary, the fluorescence of the samples etched for longer times (50 s and 120 s) was visible only on the patterned areas, which confirmed that the sol–gel active layer was retained only in areas masked with the photoresist during the wet-chemical etching. When comparing the emission spectra measured using the spectrofluorimeter, it can be seen that the signal registered for the sample, etched for the shortest time, is the highest. These results should be treated more as an illustrative example rather than a quantitative analysis. It showed well that the emission signal was higher because of the higher area of the luminescent sol–gel layer remaining on the sample. A comparison of normalized emission spectra is presented in the inset in Fig. 5b. It can be seen that despite different etching times, the shape of the spectrum did not undergo significant changes, confirming that the dye inside the sol–gel layer did not degrade or react even upon the prolonged BHF treatment.

Tuning of the etching rate of the sol–gel layer

To fabricate more complex photonic components, e.g., under-etched disc resonators, we attempted to increase the difference between the etching rates of the sol–gel layer and the soda-lime glass substrate. Additional etching experiments were performed on the layers that were heat-treated in the laboratory dryer for different time at 200 °C. Here, layers with slightly higher initial thickness were used because of the sol–gel aging (as indicated in Fig. 1).

To check the effect of the annealing time on the thickness and RI of the sol–gel layers, they were prepared on the Si substrates under different conditions—at RT and heat-treated at 200 °C for varied times in a range of 1–24 h. Results of the spectroscopic ellipsometry, as well as emission spectroscopy measurements, are presented in Fig. 6.

Figure 6
figure 6

a RI and b thickness of the sol–gel layer as a function of its heat treatment time, c emission spectra of the sol–gel films heat-treated at different times (normalized spectra are presented in the inset).

Full size image

It can be observed that the RI of the layer increased, and the thickness decreased with the heat treatment time (Fig. 6a, b respectively). Similar behavior was observed for the layers based on GLYMO and titanium(IV) butoxide precursor in a molar ratio 7: 3 [35] and prepared with a heating time of up to 17 h at 150 °C. In our case, the most significant changes could be observed in the initial drying process—the thickness of the layer changed from 730 nm to around 520 nm in the first hour of the heat treatment, whereas a further decrease of the thickness to around 330 nm took place in the next 15 h. Similar changes can be observed for the RI, where the increase from around 1.56–1.61 occurred in the first 1–2 h of the heat treatment, while an increase to 1.67 required heating for another 22–23 h. This is connected with the gradual evaporation of remaining solvent and water, which leads to the collapse of the sol–gel framework, thus leading to reduction of the porosity [54]. One may expect that due to the relatively low temperature of the heat treatment, annealing longer than 24 h would further slightly reduce the thickness of the sol–gel layer and increase its RI.

Despite this continuous change of the thickness and RI of the layers upon prolonged heat treatment at 200 °C, almost all of the prepared sol–gel layers were stable when stored under ambient conditions (Fig. S4). Only in the case of the layer prepared at RT (without any additional heat treatment), the thickness decreased over time reaching around 91% of its initial value after around 60 days. The RI of this layer is increased from 1.56 to 1.57. However, it did not reach the higher values characteristic for the layers heat-treated for longer times. The changes in the properties of the not heat-treated layer are most probably due to the stabilization of the layer connected with the evaporation of the remaining solvent, which occurs even at RT.

When the heat treatment was performed for a longer time, no obvious changes in the properties of the layer were observed upon storage at RT. The layers heat-treated for 1, 2, and 4 h did not change significantly after almost two months, maintaining their initial thickness (Fig. S4). Only minor deviations in the thicknesses and RI of these layers could be detected on different measurement days in contrast to significant changes detected for the not heat-treated layer. Thus, we can conclude that the layers prepared at 200 °C are stable over the presented time, independent of their heat treatment time.

When selecting a proper heat treatment duration, one can think about the longest one because of the highest obtained RI. However, due to the organic dye content, it is essential to determine the stability of the dye against the heat treatment. When using a particular dye (RhB), we previously observed that its stability in the hybrid sol–gel layer was superior compared to the sol–gel layer without the use of the ORMOSIL component [25]. However, in this research, we additionally checked how the prolonged heat treatment time can affect the luminescent properties of the dye in the sol–gel matrix. Figure 6c shows the emission spectra of the samples after different heat treatment times.

The emission intensity coming from the samples decreased with the heat treatment duration. This effect can be connected in the first place with the gradual degradation of the RhB dye. From the thermogravimetry analysis at temperature below 100 °C, the mass loss of the dye can be connected with loss of adsorptive water. Further steps of mass loss could be observed in ranges of 196–220 °C and 330–457 °C [55]. However, the effect of emission intensity decrease cannot be treated quantitively as many factors can contribute to it, such as, e.g., a level of scattering of the light, which can be slightly different between the samples due to small differences in sample placement or increased roughness. On the normalized emission spectra presented in the inset in Fig. 6c, it can be seen that the emission maximum shifted toward a shorter wavelength from around 581 nm (at RT and 1 h at 200 °C) to 552 nm after 16 h of heating. A similar effect was observed by us previously for the same dye in a matrix composed of TET and TEOS but at lower temperature and even for a relatively short heat treatment time [25]. Other researchers reported a blue shift of the emission for rhodamine 19 and rhodamine 6G dyes in silica matrix after heat treatment [29], which was correlated with the partial thermal decomposition of the dyes in the temperature range of 200–300 °C.

Additionally, the FT-IR technique provided insight into structural changes, which occurred in the sol–gel layer upon heat treatment and exposure to BHF. The detailed results are presented in Figs. S5 and S6. The spectrum of the non-heat-treated layer (Fig. S5) has visible bands at around 2975 cm−1, 2930 cm−1, and 2865 cm−1 related to the aliphatic CH2 group symmetric stretching vibrations and a band at around 1195 cm−1 connected with the CH3 group rocking vibrations [26]. These bands are connected with GLYMO and are visible also for the heat-treated layers. However, with increasing heat treatment time, these bands gradually disappeared, which can be a result of condensation of the precursor but also gradual degradation of the organic chain from the ORMOSIL.

Distinguishable bands are also visible at around 1090 cm−1 and 1050 m−1 for the non-heat-treated layer (Fig. S5). Their position shifted upon prolonged heat treatment to around 1110 cm−1 and 1035 cm−1, respectively. Generally, the bands in the range 1000–1200 cm−1 are assigned to the Si–O–R stretching vibration of ethoxy groups and overlap with the Si–O–Si range [26]. Thus, the observable shifts in this range upon heat treatment may be connected with the proceeding formation of the Si–O–Si network. Additionally, the appearance of a broad band at 930 cm−1 for the heat-treated samples in favor of that at 895 cm−1 observable for non-heat-treated sample might be ascribed to the gradual formation of Si–O–Ti network, which is often connected with the band at 950 cm−1 [26, 56].

The IR absorption spectra were measured also for the selected sol–gel layers, after wet-chemical etching. For this experiment, non-heat-treated layer and layers annealed at 200 °C for 4 h and 24 h were selected. They were exposed to the BHF for the same time equal to 10 s. Unfortunately, the etching was too long for the non-heat-treated layer, which was fully consumed in the BHF after a few seconds, and the IR spectrum of this layer after etching was not registered. The samples annealed for 4 h and 24 h were etched in BHF, which led to a decrease in their thicknesses. What is important, the remaining sol–gel material on both samples did not experience any structural changes, e.g., by absorption of BHF or water, which was confirmed by identical IR spectra registered before and after wet-chemical etching (Fig. S6).

The next important question stated in these studies was how the heat treatment at 200 °C influences the etching rate of the sol–gel layer in BHF. To address this problem, the sol–gel layers were prepared on soda-lime glass instead of the Si substrates. The layers fabricated on glass and heat-treated for 1, 2, or 4 h were smooth and without significant defects (Fig. S7). Similarly, the photoresist structures obtained on them after the photolithography process were characterized by overall good quality, including the presence of distinct separation between the waveguides and the discs (Fig. S8). Properties of the layers on the glass substrates, such as thickness, RI, and position of the maximum of the emission spectra, were similar to those prepared on the Si substrates (Figs. S9, S10).

The sol–gel layers on soda-lime glass substrates were subjected to a photolithography process using the photoresist and subsequent wet-chemical etching using BHF. The etching depths, as well as etching rates of samples heat-treated and etched for different times, together with the etching selectivity (calculated as a ratio between etching rates of the glass and of the sol–gel layer), are presented in Fig. 7.

Figure 7
figure 7

a Etching depth of the samples etched for different times versus heat treatment time (at 200 °C), b resulting etching rates, and c etching selectivity of the corresponding samples.

Full size image

The first batch of samples was etched for a short time (20 s) to remove only the sol–gel layer and determine its etching rate (Fig. 7a, orange points). It can be seen that the etching depth decreased with the increase in heat treatment time. For the sample heat-treated for 1 h, it was slightly above 500 nm, which was equal to the sol–gel layer thickness. The etching rate of the film decreased from around 27 nm/s to 5 nm/s for the samples heat-treated for 1 and 4 h, respectively (Fig. 7b).

To ensure complete removal of the layer in the areas not coated by the photoresist and further etching of soda-lime glass substrate, time in BHF for the second batch of the samples was selected based on the obtained sol–gel layers’ etching rates. As expected, samples treated in BHF for longer times (50 s, 83 s, and 116 s) were characterized with a much larger total etching depth—above 1000 nm. Based on these results and obtained etching rates for the sol–gel layer, the etching rate of the soda-lime glass substrate in this experiment was calculated (Fig. 7b, gray points). The calculated rate of around 15 nm/s was rather independent of the exposition time of the samples to BHF and their heat treatment time. The etching rate of the sol–gel layer was higher than that of the soda-lime glass substrate, only in the case of the sample heat-treated for 1 h. The calculated etching selectivity increased with increasing the heat treatment time, exceeding 1 for the sample heat-treated for 2 h and reaching around 3.3 for that heat-treated for 4 h (Fig. 7c). It suggests that only in the case of the samples with longer heat treatment the under-etching of the disc structures would be possible due the high selectivity of the etching.

For this reason, a third batch of the samples was prepared (with 2 and 4 h of heat treatment time). These samples were soaked in BHF for 80 s more than in the previous batch. Results showed that the total depth of the obtained structures was above 2000 nm, which was over two to times higher compared to the previous batch (Fig. 7a).

To assess the quality of the structures after wet-chemical etching and whether the sol–gel layer has been under-etched, the SEM images of the samples etched for the longest time (heat-treated for 2 and 4 h) were acquired. The results are presented in Fig. 8.

Figure 8
figure 8

SEM images (tilted view) of the microstructures etched in sol–gel layers heat-treated at 200 °C and their FIB cross-section view in the insets: a the layer heat-treated for 2 h and etched for 163 s and b the layer heat-treated for 4 h and etched for 196 s.

Full size image

Microstructures in both of the above-presented samples were characterized by good overall quality. However, despite the etching depth in both cases being the same (around 2200 nm), the lateral etching depth of the glass beneath the sol–gel structure was different in both cases (due to the different etching selectivity). Only small lateral etching under the sol–gel layer was noted for the sample heat-treated for 2 h (inset in Fig. 8a), while for the second sample under-etching of around 800 nm was observed (inset in Fig. 8b), enabling fabrication of microgoblets. It clearly shows that the longer heat treatment duration at 200 °C is beneficial when the under-etching of the sol–gel disc is desirable. Although BHF solution was in the past implemented for the structurization of sol–gel layers based on GLYMO and ZrO2, the protocol involved the etching of the glass substrate and deposition of the sol–gel layer [32]. More complex microstructures such as under-etched discs would not be possible to fabricate using this strategy. On opposite, our methodology of direct etching of the sol–gel layer using BHF enables to fabricate under-etched structures, taking advantage of different etching speeds of the sol–gel layer and soda-lime glass substrate. It can be a low-cost and less technically demanding alternative to the XeF2 isotropic etching widely implemented for achieving under-etched sol–gel and polymer structures on Si substrates [7,8,9, 11, 44].

Moreover, to verify etching rate repeatability, we compared the etching depths of several samples heat-treated under selected conditions (4 h at 200 °C) and subsequently etched for the same period. The etching depth of the sol–gel layers deposited on two separate soda-lime glass substrates and etched on different days for around 200 s was comparable, and only a small difference of around 100 nm could be observed between them, which is around 4.6% of the total etched thickness (Fig. S11a). Elongation of the etching time up to 300 s led to a further increase in the depth—above 3000 nm. Disc structure was still retained after such a long etching process, which can be seen in the microphotographs of the sample presented in Fig. S11b. Moreover, we can conclude that even larger under-etching was obtained for this sample, as evidenced by observing the dark region on the optical microscopy image around the disc structure. It was more pronounced compared to the samples etched for the shorter time (196 s), where under-etching was already observed in the SEM image.

Conclusions

Luminescent films based on TET and GLYMO, which are resistant to chemicals used in the photolithography process, were fabricated via a simple sol–gel synthesis method and dip-coating technique. Their potential for micropatterning by photolithography and wet-chemical etching processes was confirmed. Depending on the etching time, the microdisc structures with various depths could be obtained with only partial or complete etching of the sol–gel layer. Due to the protective character of the sol–gel matrix, the RhB dye, after chemical etching, still exhibited luminescent properties. Thus, the fabrication of fluorescent microstructures was possible. It was found that a longer annealing time applied for the layers increased the etching selectivity of the soda-lime glass substrate and led to over-etching of the sol–gel layer. In turn, this made it possible to obtain luminescent under-etched microgoblets, which are potentially attractive for active photonics applications. To substantiate the potential application of these luminescent microstructures in photonic devices, additional characterization is essential and will be a part of further studies. Simultaneously, it was proven that the prolonged time of annealing influenced the luminescent properties of these films by shifting the emission maximum toward a shorter wavelength, indicating some changes in the structure of the dye. Luminescent disc microstructures with relatively high RI were fabricated for the first time from sol–gel materials. However, an improvement in terms of dye thermal stability is still needed. Further progress in this area may enable the integration of this component in more complex planar photonic systems such as microlasers or sensors.