We present a method of post-deposition tuning of the optical properties of thin film dielectric filters and mirrors containing chalcogenide glass (ChG) layers by thermally adjusting their refractive index. A common challenge associated with the use of ChG films in practical applications is that they suffer from slight run-to-run variations in optical properties resulting from hard-to-control changes in source material and deposition conditions. These variations lead to inconsistencies in optical constants, making the fabrication of devices with prescribed optical properties challenging. In this paper, we present new work that takes advantage of the large variation of a ChG films’ refractive index as a function of annealing. We have carried out extensive characterization of the thermal index tuning and thickness change of arsenic selenide (As2Se3) ChG thin films and observed refractive index changes larger than 0.1 in some cases. We show results for refractive index as a function of annealing time and temperature and propose a model to describe this behavior based on bond rearrangement. We apply thermal refractive index tuning to permanently shift the resonance of a Fabry-Perot filter and the cutoff wavelength of a Bragg reflector. The Bragg reflector, consisting of alternating As2Se3 and CaF2 layers, exhibits high reflectance across a ∼550 nm band with only five layers. Modeling results are compared with spectroscopic measurements, demonstrating good agreement.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Thin films of chalcogenide glasses (ChGs) are of interest for a variety of photonics applications . They exhibit high linear refractive indices, enabling high index contrast structures; they have large optical nonlinearities, making them useful for tunable devices and nonlinear frequency conversion; and they have wide transmission windows extending from the visible through the midwave and longwave infrared.
Thin film dielectric stacks that include chalcogenide layers have long been of interest for both mirrors [2–5] and filters [6–9]. The high index contrast that they enable results in high efficiency with relatively few layers. Additionally, the infrared transmission bands of ChGs permit devices with high transmission in bands where traditional thin film materials are highly absorbing.
While there are clear potential benefits to using ChGs in thin film devices, a concern remains that the precise properties of such structures may be difficult to control. Because these materials are multicomponent glasses, deposited films are subject to slight run-to-run deviations in composition and bonding that may lead to variations in their optical properties. The composition of chalcogenide-containing films is especially challenging to control because some fraction of the chalcogen element can be lost to the vapor phase , a phenomenon that can be sensitive to both source material conditions and deposition parameters. In addition to composition, the structural properties of ChG films may also vary. Most vacuum deposition methods are non-equilibrium processes and result in ChG films with bonding that differs in character from that of an annealed bulk sample. For example, it has been shown that evaporated films of both As2S3 and As2Se3 are deposited in a non-equilibrium state and can undergo a structural transformation akin to polymerization upon thermal annealing or light exposure [11,12]. This process results in a significant, nonreversible change in refractive index.
Often, to mitigate this effect, ChG films are thermally annealed or exposed to light until their refractive index approaches that of an annealed bulk sample. Alternately, with carefully controlled annealing or light exposure, it is possible to take advantage of this one-way adjustment in a film’s refractive index and tune its index to a desired value. It has been shown that a photo-induced index change of a ChG film can shift the center wavelength of a dielectric thin film filter [6,7,9]. Shifting the resonance of a ChG metasurface via thermal tuning has also recently been demonstrated .
Here, we present results for refractive index change, Δn, and thickness change, Δd of thermally evaporated thin films of As2Se3, measured in situ during thermal annealing. Stable values of Δn in excess of 0.1 were observed at a wavelength of λ=1548 nm, with annealing times varying by multiple orders of magnitude based on the annealing temperature. A model based on the Hill equation is proposed to describe changes in the films’ refractive indices. Its large value of Δn, together with a wide infrared transmittance window, makes As2Se3 a suitable choice from among ChG materials for the tunable devices described here.
Fabry-Perot filters and Bragg reflectors were then fabricated and annealed in a furnace. Shifts in the center of the transmittance peak of filters and the long wavelength edge of Bragg reflectors were measured and found to be up to ∼40 nm in magnitude in both cases. Further annealing was found to result in a decrease in thickness, in some cases reversing the direction of the shift in optical response. The magnitude of the observed shifts is substantial, making the technique a potentially useful way to achieve tunability in thin film devices for real applications. While the devices described here are designed for the SWIR, the materials’ transparency extends into the LWIR, so with modifications to the designs, performance could be extended into the MWIR or LWIR.
There are several advantages to the controlled thermal annealing approach described in this paper. Unlike tuning achieved by light exposure, thermal tuning can be applied to multi-layer structures without regard to changes in irradiance caused by layers above or below a given layer. In addition, with thermal tuning it is straightforward to anneal a large-area device with good uniformity. Initially, there were concerns about the stability of As2Se3 films under atmospheric conditions, but it was shown that while unprotected films degrade in air, aided by the presence of light and moisture, it is possible to passivate these films with a passivation layer as thin as 10 nm . This passivation, together with avoiding exposure of the films to above band gap light, keeps them pristine for long periods without any measurable degradation in their optical properties. The fact that unannealed films can be made stable under atmospheric conditions is key because it implies that the partially annealed films described here can be made stable as well. While the films described in this paper have not been passivated, robust, stable devices for practical use could be achieved by using this stabilization approach described in .
In addition to dielectric thin film filters and mirrors, the tuning approach described here has the potential to impact a variety of other photonic applications that use ChG films. Examples in the linear regime include passive waveguides , sensors , and nonmechanical beam steerer devices . A number of nonlinear applications that incorporate ChG films and could use this approach include waveguides for supercontinuum generation [18–20] and integrated optic devices based on Brillouin scattering . It is hoped that the study of thermal annealing behavior of As2Se3 will lend itself to future developments in a variety of photonic devices that include ChG films.
Several different types of samples were designed and fabricated, as shown in Fig. 1. Figure 1(a) shows a neat film sample, used to measure refractive index (n), thickness (d), and film transmittance (T). The sample consists of a 1.5 µm thick As2Se3 film on a sapphire substrate. A Fabry-Perot resonator, shown in Fig. 1(b), consists of a 277.8 nm thick As2Se3 film sandwiched between 30 nm thick layers of Al on a silica substrate. A Bragg reflector sample, shown in Fig. 1(c) consists of alternating high (H) and low (L) index layers with a design of HLHLH on a silica substrate. The H layers are As2Se3, with a thickness of 138.9 nm, approximately a quarter wave at λ=1550 nm, and the L layers are CaF2—chosen because its low refractive index (n=1.426 at λ=1550 nm) provides high index contrast with As2Se3—with a thickness of 271.7 nm, also approximately a quarter wave at λ=1550 nm.
Devices were modeled using the matrix method as implemented by FilmStar software (FTG Software Associates). Dispersion was accounted for using Sellmeier coefficients obtained from bulk samples. For partially annealed samples, the As2Se3 films exhibited refractive indices less than those of bulk glass, thus Δn was negative. It was treated as a constant offset; i.e. the shape of the dispersion curve was taken to be the same independent of annealing time, but shifted by a constant offset that was determined experimentally. Given that the dispersion of As2Se3 has been well-characterized previously, this approach permits a straightforward method of accounting for dispersion in annealed films.
Bulk As2Se3 samples were batched from purified precursors in a nitrogen-purged glovebox. Stoichiometric quantities of purified arsenic and selenium were added to a quartz ampoule. The ampoule was then sealed under high vacuum and transferred to a rocking furnace. The melt was then formed and homogenized by heating the precursors at 750 °C for 10 hours. The glass melt was removed from the furnace, quenched in water, and annealed, and the boule was removed from the quartz ampoule. The boule was broken, and small pieces were used as deposition sources.
Films of As2Se3 were deposited via thermal evaporation while maintaining the substrate temperature at approximately 25 °C. The resulting films were spatially uniform with low surface roughness (root mean squared roughness <10 Å) and with a composition similar to those of the deposition material. Films of CaF2 and Al were both deposited by electron beam evaporation.
Refractive indices of neat film samples were measured using a prism coupler (Metricon 2010/M) with a rutile prism at a wavelength of λ=1548 nm. Both the piston and prism are heated so that heat is applied from both sides of the sample, as shown in Fig. 2. This method has previously been applied to measure the refractive index for ChG films at elevated temperature . Samples were annealed in situ at four different temperatures, 80 °C, 100 °C, 120 °C, and 150 °C, temperatures chosen to span the range where n is tuned but changes slowly enough to be measured periodically, and n was measured periodically using TE polarized light by fitting a minimum of two strongly coupled modes.
Fabry-Perot filter and Bragg reflector samples were fabricated according to the designs described above. Samples of each type were annealed together in a furnace, rather than in the prism coupler, and removed periodically for measurement. A reference neat film sample was annealed together with these samples, and n for this sample was measured in the prism coupler at room temperature. Measurements of transmittance, T, and reflectance, R, were performed in the visible through the SWIR in a spectrophotometer (Cary 7000).
3.1 Thin films
A series of neat As2Se3 films on sapphire substrates was used to measure film properties as a function of annealing time and temperature, and the results are shown in Fig. 3. The film thickness of 1.5 µm for these samples was chosen to result in at least two coupled modes at λ=1548 nm, enabling the prism coupler measurements of n. Figure 3(a) shows T in the visible through SWIR for each anneal time. We observe a shift in fringe position and an increase in fringe contrast with respect to annealing time, consistent with a change in refractive index. We do not, however, observe a decrease in the maxima of the intensity modulation, indicating that absorption has not changed. In addition, we can observe that the band edge has not shifted significantly.
For Figs. 3(b)-(d) data obtained for the furnace-annealed reference sample is shown together with data obtained for in situ-annealed samples. For the furnace-annealed sample, anneal time, t, refers to cumulative anneal time (with time out of the furnace for measurement not included). Figure 3(b) shows measured n at λ=1548 nm; Fig. 3(c) shows d/d0 where d0 is the as-deposited film thickness; and Fig. 3(d) shows optical thickness. In each case, data for the furnace-annealed reference sample used here fits the general trend of the in situ-annealed samples. Slight deviations are likely caused by the fact that the reference sample is repeatedly removed from and placed back in the furnace, resulting in uncertainty in anneal time, and also the fact that the sample is repositioned between prism coupler measurements, resulting in additional error.
For n, as plotted in Fig. 3(b), for each temperature we observe a clear sigmoidal increase from a starting refractive index of ∼2.73 that asymptotically approaches a final value of n >2.8. However, the times required to reach equivalent refractive indices differs by orders of magnitude based on anneal temperature. For example, n=2.8 is obtained after 2.5 minutes at 150 °C, while achieving the same index takes over 2000 minutes at 100 °C. We note that the annealing rate at 150 °C is so rapid that the first data point we are able to capture after temperature stabilization (∼2 min ramp from room temperature to 150 °C) already shows a significant index change over an unannealed film. The values obtained here also correspond well to values measured via ellipsometry, with the as-deposited As2Se3 having an index value of n=2.730, and n=2.766 after a one hour furnace anneal at 100 °C.1. The decrease in k is consistent with the increase in reaction speed at higher temperatures. With regards to the Hill coefficient, when h<1, reactive binding slows further binding, a process known as negatively cooperative binding. The Hill equation offers an intuitive understanding of annealing behavior by considering the reactivity of unsatisfied bonding sites within the material; the sites are satisfied through annealing, and as the sites are satisfied, the annealing rate slows due to increased distance between unsatisfied sites. As unsatisfied bonds reorganize under thermal annealing to form a more satisfied bonding structure, the reaction speed of bonding rearrangement slows. This is analogous to the slowing of reactive binding over time in the context of the Hill equation.
The normalized film thickness d/d0, plotted in Fig. 3(c), indicates that higher temperatures lead to much faster changes in thickness, whereas the change is much more gradual at lower temperatures. For thin film devices, we are also concerned with the change in optical path length as a function of annealing. The increase in n with anneal time is counteracted somewhat by the decrease in d in terms of fractional optical thickness change, n×d/d0. The net effect, however, as shown in Fig. 3(d) is positive, leveling off for the furnace-annealed samples between 10 and 100 min. and decreasing slightly for anneals beyond 100 min. The total increase in n×d/d0 for the furnace-annealed samples appear to be less than that of the in situ-annealed samples.
The change in optical thickness is due to three competing mechanisms: refractive index increase due to bond rearrangement, physical thinning due to the applied pressure in the prism coupler, and film densification due to annealing which has been reported in other chalcogenide films . The roll off in optical thickness that is observed at longer time scales for the 100 °C and 120 °C samples is most likely due to a combination of applied pressure at elevated temperatures and reduced change in refractive index as the annealing process slows.
3.2 Fabry-Perot filters
Fabry-Perot filters, as shown in Fig. 1(b) were fabricated and characterized. Results are shown in Fig. 4. Measured data for T, along with simulated results generated as described in Section 2 for several anneal times are shown in Fig. 4(a). The thickness for the As2Se3 layer used in the model, 277.8 nm, was determined by obtaining a best-fit to data for T for an unannealed sample. This thickness matched the target thickness, controlled by a crystal monitor with appropriate tooling during deposition, to within the ±5% error of the crystal monitor. Measured data is shown as points, and simulated results are shown as solid lines. Simulated results account for Δd. We observe a maximum in T of approximately 0.2% at a wavelength of 1600 nm prior to annealing, with a maximum red shift of 40 nm after 1380 min. of annealing. The model predicts the resonance position and peak value of T accurately. It underestimates the width of the transmittance peak by approximately a factor of two, possibly due to inhomogeneities in the optical thickness of the film across the measured area.
Figure 4(b) shows measured data for the peak shift. The plot also shows two sets of simulated results, one accounting for Δd and the other neglecting it. Before accounting for Δd, the model overestimates peak shift, predicting a change of 53.5 nm. After accounting for thickness change, the model underestimates peak shift, predicting a maximum shift of 28.5 nm. It is possible that the Al reflector layers explain this discrepancy; some of the thickness change in the reference sample is likely due to Se loss during annealing, but in the case of the Fabry-Perot sample, however, the Al layer may act as a barrier, preventing Se loss. While the thickness still changes somewhat, the effect is not as pronounced as in the case of the uncoated reference sample, so the experimental results lie in between modeling results that do not account for thickness change and those that account for the full change observed in the reference sample.
Note that the Al cavity mirrors used here were chosen for the purpose of simplicity; in order to provide a clear demonstration of the tuning effect, we employed a design with the minimum required number of layers. In order to make a practical, robust device, one would likely add passivation layers to prevent oxidation of the Al or replace the Al mirrors with dielectric thin film Bragg mirrors to increase transmittance.
3.3 Bragg reflectors
Bragg reflector samples were fabricated based on the design shown in Fig. 1(c). Figure 5 shows measured and simulated values for R for the entire main reflectance peak of the device for an unannealed sample. The model, as described in Section 2, was used to generate simulated results. The thicknesses for the As2Se3 and CaF2 layers used in the model, 138.9 nm and 271.7 nm respectively, were determined by obtaining a best-fit to data for R for an unannealed sample. These thicknesses both matched the target thicknesses, controlled by a crystal monitor with appropriate tooling during deposition, to within the ±5% error of the crystal monitor. We observe a broad peak in R with a maximum of approximately 96% and a value >90% from approximately 1120 nm to 1670 nm, a span of 550 nm. We note that this is excellent performance for a Bragg reflector consisting of only five layers and results from the high index contrast between As2Se3 and CaF2. There is a red shift on the long-λ side of the reflectance curve that increases with annealing time. The model reproduces the value of peak reflectance and the general shape of the peak accurately. Again, the model underestimates the width of the distribution, predicting a short-λ edge position about 60 nm longer than its actual position. This discrepancy is possibly due to inhomogeneities in the optical thickness of the film across the measured area, but the explanation is not certain at this point and is still being investigated.
Measured and simulated data for the long-λ edge of the reflectance curve are shown in Fig. 6(a). Measured data is shown as points, and simulated results are shown as solid lines. From the plot, we can see that the simulation is in good agreement with the measured results. Figure 6(b) shows measured data for the shift in the edge of the reflectance band, measured at the point of 50% of peak R. Upon annealing, the long-λ edge is red shifted by up to 40 nm while the short wavelength edge position remains fixed. Before accounting for thickness change, the model again overestimates the shift, predicting a shift of 68 nm. When accounting for thickness change, the model’s prediction of 42.5 nm is close to the measured value. Furthermore, the model accurately predicts the maximum value of R and the fact that the short wavelength edge remains fixed after annealing. Note that the maximum reflectance and flatness of the peak could be improved by increasing the number of layers beyond 5.
In this work, we demonstrated the potential for the thermal tuning of As2Se3 to enable devices with permanent, one way tunability. In situ measurements of n as a function of t and temperature show that a large Δn of approximately 0.1 is possible in some cases. A model based on the Hill equation is able to explain the evolution in n based on the fact that annealing leads to the satisfaction of bonding sites within the material. The large change in Δn enables the thermal tuning of thin film devices. This effect was demonstrated first in Fabry-Perot filters, with a peak shift of ∼40 nm observed, then in Bragg reflectors, with a shift of ∼40 nm in the long wavelength edge of the reflectance curve. The Bragg reflector design used here exhibits high reflectance across a ∼1000 nm band with only five layers. Because both the high and low index materials, As2Se3 and CaF2 respectively, have wide infrared transmission bands, the design applied here could easily me modified for MWIR or LWIR applications. It is anticipated that the thermal tuning effect utilized for this work can be applied more broadly, for other ChG film materials and for other linear and nonlinear devices, to enable post-deposition tuning of the optical response for a variety of applications.
U.S. Naval Research Laboratory (6.1 base program funding); Office of Naval Research (N0001419WX01197).
The authors declare no conflicts of interest.
1. B. J. Eggleton, B. Luther-Davies, and K. Richardson, “Chalcogenide Photonics,” Nat. Photonics 5(3), 141–148 (2011). [CrossRef]
2. T. Kohoutek, J. Orava, T. Wagner, M. Hrdlicka, M. Vlcek, and M. Frumar, “Preparation of dielectric mirrors from high-refractive index contrast amorphous chalcogenide films,” J. Phys. Chem. Solids 69(8), 2070–2074 (2008). [CrossRef]
3. R. Todorov, J. Tasseva, T. Babeva, and K. Petkov, “Multilayer As2Se3/GeS2 quarter wave structures for photonic applications,” J. Phys. D: Appl. Phys. 43(50), 505103 (2010). [CrossRef]
4. P. Němec, J. Charrier, M. Cathelinaud, M. Allix, J.-L. Adam, S. Zhang, and V. Nazabal, “Pulsed laser deposited amorphous chalcogenide and alumino-silicate thin films and their multilayered structures for photonic applications,” Thin Solid Films 539, 226–232 (2013). [CrossRef]
5. S. Normani, J. Gutwirth, L. Bodiou, M. Baillieul, L. Joanny, J. Charrier, P. Nĕmec, and V. Nazabal, “Radio-frequency sputtering fabrication of chalcogenide-based Er 3+-doped vertical optical cavities for near-infrared operation,” Opt. Mater. Express 10(10), 2500 (2020). [CrossRef]
6. W. Shen, M. Cathelinaud, M. Lequime, V. Nazabal, and X. Liu, “Photosensitive post tuning of chalcogenide Te20As30Se50 narrow bandpass filters,” Opt. Commun. 281(14), 3726–3731 (2008). [CrossRef]
7. W. D. Shen, M. Cathelinaud, M. Lequime, F. Charpentier, and V. Nazabal, “Light trimming of a narrow bandpass filter based on a photosensitive chalcogenide spacer,” Opt. Express 16(1), 373–383 (2008). [CrossRef]
8. T. Kohoutek, J. Orava, J. Prikryl, T. Wagner, M. Vlcek, P. Knotek, and M. Frumar, “Planar chalcogenide quarter wave stack filter for near-infrared,” J. Non-Cryst. Solids 355(28-30), 1521–1525 (2009). [CrossRef]
9. A. Bourgade and J. Lumeau, “Large aperture, highly uniform narrow bandpass Fabry–Perot filter using photosensitive As2S3 thin films,” Opt. Lett. 44(2), 351 (2019). [CrossRef]
10. J. A. Frantz, J. D. Myers, R. Y. Bekele, V. Q. Nguyen, B. M. Sadowski, S. I. Maximenko, M. P. Lumb, R. J. Walters, and J. S. Sanghera, “Quaternary Sputtered Cu(In,Ga)Se2 Absorbers for Photovoltaics: A Review,” IEEE J. Photovolt. 6(4), 1036–1050 (2016). [CrossRef]
11. J. P. De Neufville, S. C. Moss, and S. R. Ovshinsky, “Photostructural transformations in amorphous As2Se3 and As2S3 films,” J. Non-Cryst. Solids 13(2), 191–223 (1974). [CrossRef]
12. D.-Y. Choi, A. Wade, S. Madden, R. Wang, D. Bulla, and B. Luther-Davies, “Photo-induced and Thermal Annealing of Chalcogenide Films for Waveguide Fabrication,” Phys. Procedia 48, 196–205 (2013). [CrossRef]
13. J. A. Frantz, R. Y. Bekele, A. R. Clabeau, J. D. Myers, V. Q. Nguyen, N. M. Litchinitser, and J. S. Sanghera, “Thermally tuned resonances in chalcogenide glass dielectric metasurfaces,” Proc. SPIE 11287, 1128701 (2020). [CrossRef]
14. J. A. Frantz, J. D. Myers, R. Y. Bekele, C. M. Spillmann, J. Kolacz, H. Gotjen, V. Q. Nguyen, C. C. McClain, and J. S. Sanghera, “Arsenic selenide thin film degradation and its mitigation,” Opt. Mater. Express 8(12), 3659 (2018). [CrossRef]
15. Z. Fan, K. Yan, L. Zhang, J. Qin, J. Chen, R. Wang, and X. Shen, “Design and fabrication of As2Se3 chalcogenide waveguides with low optical losses,” Appl. Opt. 59(6), 1564 (2020). [CrossRef]
16. E. Baudet, M. Sergent, P. Němec, C. Cardinaud, E. Rinnert, K. Michel, L. Jouany, B. Bureau, and V. Nazabal, “Experimental design approach for deposition optimization of RF sputtered chalcogenide thin films devoted to environmental optical sensors,” Sci. Rep. 7(1), 3500 (2017). [CrossRef]
17. J. A. Frantz, J. D. Myers, R. Y. Bekele, C. M. Spillmann, J. Naciri, J. Kolacz, H. G. Gotjen, V. Q. Nguyen, C. C. McClain, L. Brandon Shaw, and J. S. Sanghera, “Chip-based nonmechanical beam steerer in the midwave infrared,” J. Opt. Soc. Am. B 35(12), C29 (2018). [CrossRef]
18. X. Gai, D.-Y. Choi, S. Madden, Z. Yang, R. Wang, and B. Luther-Davies, “Supercontinuum generation in the mid-infrared from a dispersion-engineered As 2 S 3 glass rib waveguide,” Opt. Lett. 37(18), 3870–3872 (2012). [CrossRef]
19. Y. Yu, X. Gai, P. Ma, K. Vu, Z. Yang, R. Wang, D.-Y. Choi, S. Madden, and B. Luther-Davies, “Experimental demonstration of linearly polarized 2–10 µm supercontinuum generation in a chalcogenide rib waveguide,” Opt. Lett. 41(5), 958 (2016). [CrossRef]
20. J.-B. Dory, C. Castro-Chavarria, A. Verdy, J.-B. Jager, M. Bernard, C. Sabbione, M. Tessaire, J.-M. Fédéli, A. Coillet, B. Cluzel, and P. Noé, “Ge–Sb–S–Se–Te amorphous chalcogenide thin films towards on-chip nonlinear photonic devices,” Sci. Rep. 10, 11894 (2020).
21. B. J. Eggleton, C. G. Poulton, P. T. Rakich, M. J. Steel, and G. Bahl, “Brillouin integrated photonics,” Nat. Photonics 13(10), 664–677 (2019). [CrossRef]
22. B. Gleason, K. Richardson, L. Sisken, and C. Smith, “Refractive Index and Thermo-Optic Coefficients of Ge-As-Se Chalcogenide Glasses,” Int. J. Appl. Glass Sci. 7(3), 374–383 (2016). [CrossRef]
23. Q. Zhang, S. Bhattacharya, and M. E. Andersen, “Ultrasensitive response motifs: basic amplifiers in molecular signalling networks,” Open Biol. 3(4), 130031 (2013). [CrossRef]
24. C. Tsay, E. Mujagić, C. K. Madsen, C. F. Gmachl, and C. B. Arnold, “Mid-infrared characterization of solution processed As2S3 chalcogenide glass waveguides,” Opt. Express 18(15), 15523 (2010). [CrossRef]