Open Access
Add to BibSonomy
Abstract
The next frontier in photonics will rely on the synergistic combination of disparate material systems. One unique organic molecule is azobenzene. This molecule can reversibly change conformations when optically excited in the blue (trans-to-cis) or mid-IR (cis-to-trans). Here, we form an oriented monolayer of azobenzene-containing 4-(4-diethylaminophenylazo)pyridine (Aazo) on SiO2 optical resonators. Due to the uniformity of the Aazo layers, quality factors over 106 are achieved. To control the photo-response, the density of Aazo groups is tuned by integrating methyl spacer molecules. Using a pair of lasers, the molecule is reversibly flipped between molecular conformations, inducing a refractive index change which results in a resonant wavelength shift. The magnitude of the shift scales with the relative surface density of Aazo. To investigate reproducibility and stability of the organic monolayer, three switching cycles are demonstrated, and the performance is consistent even after a device is stored in air for 6 months.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Optically responsive organic small molecules have played an enabling role in numerous technologies ranging from organic light-emitting diodes to organic solar cells. One particularly interesting class of organic materials are photoswitchable molecules which switch their molecular conformations from trans-to-cis (or cis-to-trans) upon exposure to light [1–4]. Among these molecules, azobenzene is one of the most commonly used photoswitches due to its ease of synthesis and highly predictable reversible switching behavior [5]. While this molecule has been used in a wide range of materials-driven applications [6–9], only limited work has investigated its compatibility with integrated photonics devices [10].
One interesting optical device is the whispering gallery mode optical resonator. The cavity confines light in circular orbits at precisely defined resonant wavelengths (λ) that are determined by the optical and geometric properties of the cavity [11]. A small portion of the optical field forms an evanescent tail, leaking into the environment and interacting with any molecules or thin films bound to the device. Thus, any change in the index or geometry of the surface bound monolayer will change the resonant wavelength. The photon lifetime or storage time inside the cavity is defined as the quality factor (Q). The optical interaction strength between bound molecules and the circulating optical field is significantly enhanced due to the high Q as compared to a simple waveguide structure.
In previous work, two different types of optically responsive materials have been explored. To create tunable resonant cavities, disordered layers of biological [12,13] or azobenzene-based [10,14,15] photoswitchable molecules were formed on the surface of the cavity. While innovative, the previous efforts did not rigorously study the role of the surface chemistry in the device performance. For example, the role of the surface density of the photo-responsive groups in the device response was not investigated and the stability in different storing conditions was not verified.
It is important to note that the ability to control the concentration of the Aazo group was not investigated in the prior work, and controlling the surface density of chemical or biological functional groups on optical device surfaces is an under-studied area in general [16]. The Aazo group physically changes confirmation, controlling the spacing between Aazo groups is critical for two reasons: 1) enabling the photoswitching behavior and 2) controlling the magnitude of the device response. However, in other fields, like biodetection, the density of functional groups controls the detection signal [17,18]. Therefore, developing strategies to precisely control the surface density of functional groups is a critical technology in many fields [19–21].
In the present work, we demonstrate a hybrid optical device comprised of an integrated optical resonator with a grafted monolayer containing the photoswitchable azobenzene functional group in Fig. 1(a). The specific photoswitchable organic small molecule investigated was 4-(4-diethylaminophenylazo)pyridine (Aazo), and the density on the surface was varied by adding a non-photoisomerizable co-adsorbate methyl to the layer. The Aazo photoisomerizes from the thermodynamically stable trans-isomer to thermodynamically unstable cis-isomer at 410 nm and reverts to trans- by heating with a 10.6 µm CO2 laser in Figs. 1(b)–1(c). The photoswitching of the Aazo induces a refractive index change, the 1300 nm resonant wavelength is tuned, and the magnitude of the reversible frequency shift is dependent on the density of bound Aazo molecules. The device response is investigated over a range of incident powers and before and after 6 months of storage in air. Complementary density functional theory (DFT) and finite element method (FEM) calculations and ellipsometry data confirm that the observed wavelength shift is due to the photoswitching of the Aazo.
Fig. 1. (a) SEM image of Aazo-coated microtoroid resonant cavity. (b)-(c) Rendering of the region indicated by the box in (a) showing the reversible Aazo trans to (c) cis photoisomerization process.
2. Theory
The quality factor is defined as Q = λ/δλ where δλ is the resonance linewidth of the cavity. In the present cavities, the intrinsic cavity Q is limited by the material Q factor. As shown in the expression below, in addition to being dependent on the material loss, the material Q is wavelength and refractive index dependent according to the following expression [11]:
where neff is the effective refractive index and αeff is the effective material absorption. As a result, there are two cavity Q factors of interest: the cavity Q at 410 nm (Q410) and the cavity Q at 1300 nm (Q1300). Additionally, each Q plays a distinctly different role in this photoswitchable, pump-probe system.To trigger the trans-cis switching of the Aazo, light is coupled into a cavity resonance at 410 nm, and the optical power from this resonance initiates the photoswitching. To calculate the power circulating inside the cavity, the following general expression can be used [22]:
where Pcirc is the circulating power, R is the radius of the device, neff is the effective refractive index, λ is the cavity resonant wavelength, Pin is the input power, and K ≡ Qo/Qc where Qo is the intrinsic Q, and Qc is the coupling losses of the system.In the present system, there are two circulating powers: one at 1300 nm and the other at 410 nm. However, due to the selective absorption of the Aazo in the blue wavelength range, only the input power of the 410 nm laser will induce photo-switching. Therefore, in Eq. (2), the values of relevance are the Pin of the 410 nm laser and the Q and λ at 410 nm. From this analysis, it is clear that having a higher Q in the blue wavelength range significantly reduces the amount of input power needed to induce a resonant wavelength change. In the case of the near-IR, δλ determines the resolution of the resonant wavelength shift (Δλ). Therefore, a higher Q factor allows for smaller wavelength shifts to be detected.
The resonant wavelength position in an ultra-high Q optical cavity is defined by the cavity effective refractive index (neff) and R according to λ = 2πneffR/m, where m is the optical mode number. Therefore, changes in either parameter will induce a shift in the λ according to Δλ = λ(Δneff/neff + ΔR/R) where Δλ is the change of resonant cavity wavelength, ΔR is a change in the cavity radius, and Δneff is the change of the cavity effective refractive index [23]. Given this dependence on R and neff, the Aazo molecule studied in the present work has the potential to modify λ through both the ΔR and the Δneff terms.
To gain insight into the radius change and the refractive index of the Aazo layer, DFT calculations were performed. The DFT calculations used Q-Chem 5.1 Software (Q-Chem, Inc.). The gas phase ground state molecular geometry was optimized at the B3LYP/6-31G** level of theory [24,25] for two pairs of free standing Aazo molecules before calculating their dynamic polarizabilities at 1300 nm by time-dependent DFT. In this study, Aazo and [4-(chloromethyl)phenyl]-trichlorosilane (CMPS)-Aazo in both trans- and cis-isomers were considered in Fig. 2. A similar molecule, 4-[4-(N,N-dimethylamino phenyl)azo]pyridine (MAP), was used as a benchmark. MAP has the exact structure as Aazo except for two methyls on the amine group.
Fig. 2. The two Aazo structures in two conformations modeled using DFT: (a) trans-Aazo, (b) cis-Aazo, (c) trans-CMPS-Aazo, and (d) cis-CMPS-Aazo.
The length of each molecule was estimated by measuring the distance between the two ends of each Aazo in the optimized ground state geometry. From Table 1, the length of trans-Aazo and trans-CMPS-Aazo are 13.33 Å and 20.68 Å, respectively. The length of MAP and CMPS-MAP are previously reported as 10 Å and 18 Å, respectively [26]. This shows that the sizes of MAP and Aazo are very similar, further providing support for using MAP as the benchmarking molecule.
Table 1. Results from DFT calculations
The radius of the device is unaffected by the circulating optical field at low input powers therefore, the only contributor to the ΔR/R term is the change in the length of the molecule. Based on the calculations, as the molecule switches from trans to cis, the length change (ΔR) is ∼3–5 Å. As mentioned previously, the devices used in the present work have radii on the order of 20–30 μm. Therefore, the ΔR/R contribution to Δλ/λ will be approximately 1E-5.
The index of refraction for each Aazo was estimated by combining the DFT results with the Lorenz Model [27] that describes the relationship between polarizability and index of refraction. The expression is:
where nL is the calculated Lorenz refractive index of the specific Aazo, N is the average number of Aazo molecules per unit volume, and p is the mean dynamic polarizability at 1300 nm in all directions. N was calculated as the inverse of unit Aazo volume. Both p and N were estimated using the Polarizable Continuum Model [28] with air as the solvent. It is important to note that, because this calculation is based on closely packed molecules, it will over-estimate the refractive index values of the material. Thus, it should be viewed as providing an upper bound.Based on the results calculated using Eq. (3) and summarized in Table 1, several important conclusions can be drawn. As the Aazo photoswitches from trans to cis, the change of Lorenz index of Aazo (ΔnL) is approximately -(0.13-0.14), indicating the reduction of nL. This decrease will induce a blueshift in the resonant wavelength (λ). However, because the optical field does not solely reside in the Aazo monolayer, the optical device does not experience a change in index of this magnitude.
To accurately calculate the effect of the Aazo switching on the resonant wavelength, finite element method (FEM) modeling of the resonant wavelength is performed using COMSOL Multiphysics [29,30]. The model is comprised of a silica toroidal cavity with a monolayer of CMPS-Aazo. Two models are constructed to reflect the cis and trans states of the device, and the thickness and refractive index values are acquired from DFT calculations in Table 1. The rest of the model parameters were set to match the devices used in the experiments. We searched the fundamental modes with different azimuthal mode numbers (M). Mode with M = 181 has the wavelength closest to the mode we characterized in experiments, so this mode was studied in both simulation models to calculate the difference in resonant wavelengths. Based on the FEM calculations, the resonant wavelength change is (ΔλFEM) approximately -130 pm. Using this value, the Δλ/λ = 1E-4, which is an order of magnitude larger than the radius change alone. Therefore, the major contribution of Δλ/λ comes from Δneff/neff.
3. Results and discussion
3.1 Device fabrication
Si wafers (100) with 2 μm of thermal oxide (SiO2) were purchased from WRS Materials. The SiO2 microtoroids were fabricated according to a previously reported protocol [31]. The major radius (R) is around 20–30 μm while the minor radius (r) is around 2.5–4.5 μm (Fig. 1(a)).
To attach the Aazo molecules to the on-chip silica cavities, a three-step process is used as shown in Fig. 3(a). This silanization-based surface chemistry process results in a self-assembled Aazo monolayer and was based on methods developed for optical devices with several modifications [32]. Notably, in previous work, it was shown that the general techniques used would have negligible impact on the cavity Q. Briefly, the device surface was first activated with OH groups. In the second step, the OH groups covalently bound both silanization linker agents to the surface. However, only the CMPS is reactive with the Aazo moiety. In the third step, the Aazo moiety was attached. All reagents and solvents are used as received unless otherwise noted. Aazo (98%) was purchased from Tokyo Chemical Industry. Tetrahydrofuran (THF, 99%) was purchased from VWR International. CMPS (97%) and trichloromethylsilane (TCMS, 99%) were purchased from Sigma-Aldrich.
Fig. 3. The surface functionalization process of optical resonant cavities. (a) Scheme of surface functionalization of Aazo and CH3 on SiO2 microtoroid cavities (photoswitchable azobenzene group colored). The non-optically responsive spacer molecule CH3 is indicated. Different amounts of this molecule allowed for control of the surface density of the photoactive Aazo. (b) Representative optical microscope images of [CH3 only] control and [CH3:Aazo = 10:1] microtoroid cavities.
To start with, O2 plasma (SCE104, Anatech USA) was used to clean organic residues on the surface of SiO2 microtoroids, yielding a monolayer of OH groups grafted on the SiO2 microtoroids (Fig. 3(a)). Next, the CMPS which contains the reactive site (benzyl chloride) for the Aazo attachment was deposited. Chemical vapor deposition (CVD) for 10 min at room temperature was used to covalently bond the CMPS with the OH groups, resulting in a monolayer of CMPS on the SiO2 microtoroids. Lastly, to graft Aazo on the CMPS-grafted SiO2 microtoroids, Aazo in THF solution (∼8.0 mM) was spin-coated on the CMPS-grafted SiO2 microtoroids (7000 rpm, 30 s). To facilitate the covalent attachment between Aazo and the benzyl chloride on the CMPS-grafted SiO2 microtoroids, the samples were treated in vacuum for 20 min at 120 °C. After each functionalization step, the surface of the microtoroid samples were rinsed with acetone, methanol and isopropanol, and dried on a hotplate at 100 °C for 10 min. This rinsing step removed all physisorbed molecules, resulting in a covalently attached monolayer.
To control the surface density of the Aazo, TCMS was introduced along with the CMPS during the CVD process. The non-reactive CH3 groups on the TCMS act as spacer molecules (Fig. 3(a)). Moreover, because the CH3 group is relatively small, the photoisomerization of Aazo can occur unhindered, as shown. In this work, four [CH3:Aazo] ratios as well as a control are studied: 3:1, 5:1, 7:1, 10:1, and CH3 only. These ratios are defined according to the relative ratios of TCMS:CMPS in the solutions. Therefore, the actual relative ratio of [CH3:Aazo] on the devices might be slightly different. Images of coated devices were taken using an optical microscope and are shown in Figs. 3(b) and 3(c). Additional control samples consisting of surface functionalized SiO2/Si wafers are prepared in parallel with the optical cavities and used for surface characterization measurements.
3.2 Characterization of photoswitchable Aazo in solution
Before being anchored onto the device surface, the switching behavior of the Aazo was characterized in THF using a LAMBDA 950 UV/Vis Spectrophotometer (PerkinElmer). To verify the selectivity of the photoswitching to the 405 nm optical source as well as analyze the optical loss of the Aazo in the near-IR, a pair of measurements was made before and after exposing the solution to the 405 nm light (BlueWave LED DX-1000 VisiCure, Dymax) for 10 min. The scan range was from 300–1330 nm. Before exposure, the spectrum is dominated by the π-π* transition of the trans isomer of the Aazo moiety (Fig. 4(a)). After exposure, the 440 nm peak decreased, because the thermodynamically favored trans-isomer was switched to the relatively unstable cis-isomer [1]. It is important to point out that both isomers of Aazo have minimal absorption at 1300 nm. This is an important consideration for the device performance in the subsequent measurements.
Fig. 4. UV-Vis spectroscopy results. (a) Absorption spectra of trans-Aazo (black curve) and cis-Aazo (red curve). Inset: photoisomerization process of Aazo. (b) Kinetics study monitoring the photoisomerization progress. Excitation to induce switching occurred at 405 nm and absorbance was monitored at 410 nm.
To study the time dependence of the photoisomerization in THF, first the trans-to-cis photoisomerization of Aazo was initiated by illuminating the sample solution with the 405 nm blue lamp for 10 min. To allow the reverse cis-to-trans photoisomerization, the sample solution was placed in a dark environment for 10 min. The change in absorbance at 410 nm in both directions were recorded every minute until the absorbance at 410 nm became relatively stable. The results are shown in Fig. 4(b). The change is fit to a single exponential for both photoisomerization processes, and the rate constants for trans-to-cis and cis-to-trans are (0.5 ± 0.1) min-1 and (0.3 ± 0.1) min-1, respectively. These values are similar to previous work [33,34]. It is important to note that this timescale is expected to be faster than a molecule anchored to a substrate for several reasons. First, these molecules are able to freely diffuse. This unhindered motion allows both functional groups to contribute to the motion required for the trans-to-cis configuration change. Second, the concentration in solution is low, reducing molecule-molecule interactions. On a surface, there is a chance of steric interactions.
3.3 Surface chemistry analysis
A common approach to perform elemental analysis of surfaces and to confirm surface chemistry is X-ray photoelectron spectroscopy (XPS) [35,36]. This technique was used to verify the surface attachment of the Aazo. To confirm each step of the surface chemistry process shown in Fig. 3, three samples were characterized: SiO2/Si wafer, CMPS functionalized sample, and CMPS-Aazo functionalized sample. XPS data were taken from AXIS Ultra photoelectron spectrometer (Kratos Analytical Ltd) equipped with Al Kα X-ray source (1486.6 eV). XPS measurements penetrated 10 nm below the surface of the SiO2/Si wafer samples. Survey scans were performed 5 times for each sample across 1200–0 eV with the analyzer pass energy of 160 eV. For N 1s, high resolution scans were conducted 15 times for each sample ranging 410–390 eV with the analyzer pass energy of 40 eV.
The XPS results in Fig. 5 verified the success of the surface functionalization of Aazo on the SiO2/Si wafers. For the SiO2/Si wafer, there are O 1s peak at 530 eV, C 1s peak at 282 eV, and Si 2s peak at 152 eV. For the CMPS-SiO2 wafer, there is an additional Cl 2p peak at 197 eV compared to the SiO2/Si wafer, showing the success of the CVD process of the CMPS. Lastly, for the CMPS-Aazo coated wafer, an additional N 1s at 396.7 eV was confirmed by a high-resolution scan, demonstrating that Aazo is functionalized. The trend of peak intensities correlates to our previously reported work [32].
Fig. 5. XPS results verifying the two steps in the surface functionalization process. The Cl 2p which is present in CMPS-coated wafers indicates successful CMPS silanation, and the N 1s indicates successful attachment of the Aazo. The C 1s is contamination in the chamber.
The surface chemistry analyses were conducted using a spectroscopic ellipsometer (VASE, J.A. Woollam Co.) with variable wavelengths ranging from 1000 to 1700 nm and variable incident angles ranging from 65 to 75°. The wafers used in the ellipsometry study matched the composition as those in the resonant cavity studied, [CH3:Aazo = 3:1, 5:1, 7:1, 10:1] coated wafers and the control [CH3 only] coated wafer. Three samples of each wafer were analyzed. The amplitude component and phase difference data were fit using the Sellmeier equation to determine the refractive index at 1300 nm from ellipsometry (n1300 (ellipsometry)). A 450 nm laser (∼1.6 W/cm2) was applied for one minute to initiate trans-to-cis Aazo photoisomerization on each sample. For reversing the Aazo photoisomerization, the sample was either heat-treated for 30 minutes at 100°C or stored in the dark for over 24 hours. n1300 (ellipsometry) was measured initially, after 450 nm exposure, and after treatment (thermal or dark).
The Δn1300 (ellipsometry) induced by trans-to-cis photoisomerization ranges from 6.7E-5 for the [CH3 only] coated wafer to -5.7E-4 for the [CH3:Aazo = 3:1] coated wafer. While all Aazo functionalized samples showed a decrease in refractive index after blue laser exposure, the CH3 functionalized had a small positive index change after exposure. This increase is most likely due to the thermo-optic effect. The magnitude of index change was related to the relative concentration of Aazo, indicating that the surface density was able to be controlled. After storage in the dark for over 24 hours, the system fully recovered.
3.4 Device characterizations and optical Q
Given the importance of Q410 and Q1300 in the device performance, measuring the device Q at both wavelengths is the first step. Notably, the device is operated on-resonance at both 410 nm and 1300 nm simultaneously, which means that both wavelengths are coupled into the device at the same time. This dual wavelength excitation can be accomplished by either using two different waveguides or using one waveguide capable of coupling in both wavelengths. In the present work, the latter approach was taken.
The optical testing setup used is depicted in Fig. 6(a). The two fiber coupled lasers were combined into a single fiber using a 90% (410 nm):10% (1300 nm) coupler. The output from the coupler was sent into a tapered optical fiber waveguide capable of simultaneously coupling 1300 nm and 410 nm light into the cavity. However, the coupling mechanisms were different, as will be discussed. Additionally, the goal of using a single waveguide resulted in a decreased efficiency at both wavelengths: approximately 15% at 410 nm, and about 30-40% at 1300 nm. The optical taper waveguide was fabricated by pulling a coating-stripped optical fiber (single mode fiber F-SMF-28, Newport) on a two-axis stage controller (Sigma Koki) while being heated by a hydrogen torch. The coupling distance between the taper and the microtoroid cavity was precisely controlled by a 3-axis nanopositioning stage. Photodetectors (Thorlabs) were used to detect the optical signals, displaying each signal on an oscilloscope processed with a high-speed digitizer.
Fig. 6. Optical device characterization. (a) Optical device characterization testing setup and photoisomerization system with all components labeled. (b) Example of a Q spectrum at 1300 nm. Inset is a side view optical microscope image of an Aazo-coated device coupled by an optical taper. (c) Example of a Q spectrum at 410 nm.
At 1300 nm, light from a tunable, narrow linewidth laser (Velocity series, Newport) is coupled evanescently into the cavity [37]. The transmission vs λ spectra were fit to a Lorentzian, and the loaded Q1300 was calculated by Q =λ/δλ, where δλ is the spectral linewidth (Fig. 6(b)). By varying the coupling strength, the intrinsic Q1300 (Q0,1300) was calculated using a coupled cavity modeled [38].
At 410 nm, light from a tunable, narrow linewidth laser (Velocity series, Newport) is coupled into the optical cavity. However, because phase and index matching criteria are unable to be continuously met from 400-1300 nm, it is not possible to evanescently couple at both wavelengths. Therefore, a different coupling mechanism is used, specifically, Rayleigh scattering [39]. The loaded Q410 (Qloaded,410) was calculated using the same method described previously, and an example spectrum is shown in Fig. 6(c). Unfortunately, using this coupling mechanism, it is not possible to vary the coupling strength. Therefore, an intrinsic cavity Q at 410 nm cannot be determined.
All optical cavity measurements are performed at room temperature under ambient conditions (room temperature, pressure, air) with one exception. All measurements were performed in the dark, to reduce the effect of room light on the Aazo group. When not being tested, all devices are stored under ambient conditions. To reversibly switch the Aazo between trans and cis isomerization states, first, the 410 nm laser initiated the photoswitching behavior from trans-Aazo to cis-Aazo. A range of 410 nm input powers were studied to determine the response and sensitivity. The 1300 nm resonant wavelength was tracked in real-time as soon as the 410 nm laser started to couple light into a microtoroid sample.
Previous work demonstrated that by adding thermal energy to the system, it is possible to accelerate the reverse isomerization [10,40]. Therefore, to assist in the reverse isomerization, we use a CO2 laser to indirectly heat the molecule on the device. The 10.6 μm CO2 laser (48-1KAM, Synrad) was guided by a gold coated copper mirror (Electro Optical Components, EOC) and focused by a ZnSe lens (Thorlabs) onto the microtoroid samples. The CO2 laser is highly absorbed by silica, generating thermal energy. As a result, the resonant wavelength shifts due to a combination of the thermo-optic effect of silica and the molecular switching. Therefore, the cavity must thermally equilibrate in the dark to completely return to its initial position. The incident power used from the CO2 laser was approximately ∼0.22 W with a spot size of approximately 100 μm, which is over an order of magnitude lower than the power used to reflow the cavity. This reduced power was also confirmed in several of the measurements. The thermal red-shift observed was indicative of an approximate 4-5 °C temperature increase according to the resonant wavelength shift [41], even at the highest powers used.
In the present system, the thermal damage threshold of the organic monolayer is governed by the bond strength between the benzylchlorosilane linker molecule and the oxygen groups on the surface. The thermal degradation of a typical siloxy bond begins at 200 °C. This roughly corresponds to a thermal damage threshold of 200-300°C [42,43]. For comparison, the damage threshold of silica is about 1100 °C [44].
Both Q0,1300 and Qload,410 were determined on three different devices for each coating ratio. The Q0,1300 are plotted in Fig. 7(a), and the values range from ∼106 to ∼107. The Q0,1300 values of the control device (CH3 coated) are similar to previous works using silanization processes [32,45,46]. On the other hand, the Qloaded,410 of various devices were plotted in Fig. 7(b), ranging from ∼105 to ∼106. For all devices, the Qloaded,410 is less than an order of magnitude lower than the corresponding Q0,1300 on the same device. The relatively small variations between various Aazo-coated and CH3 control devices are evidence of the uniformity of the molecular surface functionalization at both Q0,1300 and Qloaded,410.
Fig. 7. Optical cavity quality factors. (a) Q0,1300 and (b) Qloaded,410 of Aazo-coated devices with different ratios. Each data point is a unique device, and the error is the error in the fitting of the linewidth. (c) Q0,1300 and (d) Qloaded,410 with respect to the input power of 410 nm laser. The measurements in part (c) and (d) were performed on the same Aazo-coated SiO2 microtoroid device.
Additionally, given the important role of Q at 410 nm and at 1300 nm both in inducing and in detecting the switching behavior, the dependence of Q as a function of 410 nm input power was also investigated. As can be observed in Figs. 7(c) and 7(d), the quality factor at both wavelengths is independent of the input power of the 410 nm laser over the range used in this work. The input power at 1300 nm was held constant throughout the measurements.
To study the effects of photoisomerization on Q, both Q0,1300 and Qloaded,410 of trans-Aazo and cis-Aazo samples were measured with the same [CH3:Aazo = 10:1] device (Table 2). While trans-Aazo has a slightly higher Q0,1300 and Qloaded,410 than those of cis-Aazo, the differences are negligible at both wavelengths; therefore, the photoisomerization process minimally affects Q.
Table 2. Q0,1300 and Qloaded,410 of [CH3:Aazo = 10:1] device on different Aazo isomers
The Qloaded,410 exhibits at least 105, indicating high Q modes are excited in our whispering gallery mode system, allowing for a resonantly enhanced absorption of 410 nm light by the Aazo monolayer on-device. However, given these Q values and the observed low coupling efficiency, it is clear that the coupling losses are the dominant losses in this system. Therefore, due to the uncertainty in coupling losses at 410 nm, the conventional approximation that K = 1 should not be applied by default. Nonetheless, the loaded Q values at 410 nm are less than an order of magnitude lower than the intrinsic Q values at 1300 nm. Given this data, the coupling losses cannot be significant. Therefore, in our following Pcirc calculations, we are using the conventional approximation.
3.5 Photoisomerization-induced wavelength shifts
The entire photoswitching cycle is shown in Fig. 8(a) for the [CH3:Aazo = 10:1] device with the three phases indicated. Initially, the 410 nm laser is coupled into the device. This initiates a trans-to-cis Aazo photoisomerization, inducing a blue-shift in the resonant cavity wavelength (λ1300) of approximately 7.8 pm (Pcirc = 420.86 mW). It took around 10 min to stabilize the blueshift. This timeframe is significantly longer than in solution; however, the two results should not be directly compared for a couple reasons. First, the laser excitation wavelength and the incident optical power are different in the two measurements. Second, the molecules are anchored to the device surface, which reduces their mobility.
Fig. 8. Tracing λ1300 shifts on two different [CH3:Aazo = 10:1] Aazo-functionalized optical cavities upon coupling 410 nm laser. (a) One cycle of reversible λ1300 shift. When the device was coupled by the 410 nm light (Pcirc = 420.86 mW), λ1300 blueshifted 7.8 pm. Upon exposure to the 10.6 μm CO2 laser, λ1300 immediately red-shifted; ultimately, it returned to the initial λ1300. (b) Repeated photoisomerization of Aazo on-device. The black, red and blue regions are the 1st, 2nd and 3rd cycle, respectively. The corresponding blueshifts for each cycle were 4.5 pm (Pcirc = 127.36 mW), 3.6 pm (Pcirc = 96.33 mW), and 3.0 pm (Pcirc = 92.35 mW), respectively. (c) Two reversible cycles on the same device as Figs. 9(a) and 9(b) with similar 410 nm input power and Δλ: black curve was the first study and red curve was tested after 6 months.
Then, the 410 nm laser was replaced by the CO2 laser. Due to the high optical absorption of silica at 10.6 μm, the device surface becomes hot within a couple of seconds. As a result, the thermal energy provided by the CO2 laser reversed the cis-Aazo back to trans-Aazo, allowing a transient redshifted λ1300. After the CO2 laser was turned off, λ1300 can equilibrate in the dark and return to its initial λ1300. As seen in Fig. 8(a), the net shift for this entire cycle was approximately zero, indicating the complete reversibility of the photoisomerization process.
A common concern when using organic molecules for optical applications is the molecular stability (or reproducibility of the response), particularly when they are operated and stored in ambient environments (environments containing oxygen). This concern is further compounded when they are exposed to high intensity optical fields, such as the ones used in the present work.
To study this potential degradation mechanism, a second device with the same ratio [CH3:Aazo = 10:1] is fabricated and is cycled multiple times in air. On this device, the results from three photoswitching events which were conducted sequentially are shown in Fig. 8(b). Based on Eq. (2), Δλ scales with Pcirc, and the net shift for each cycle is negligible, indicating complete reversibility of the photoisomerization process with minimal loss of response. This qualitative reproducibility of three cycles is notable as it confirms two important technical considerations. First, the surface density of Aazo on-device is sufficiently optimized to allow for free movement of the Aazo, and second, the optical and thermal powers being used to induce the switching do not damage the molecule. Due to the testing procedure used, which required measuring the power in the blue laser before and after each measurement, the noise and total Pcirc changed slightly in each cycle, as shown in Fig. 8(b).
In a complementary study, the same device was analyzed twice over a timespan of six months. During the interim period, the device was stored in an ambient environment (room pressure, temperature, and air). As can be seen in Fig. 8(c), even after 6 months of storing the device in air at room temperature, the device response did not noticeably change. This result provides further evidence that the surface chemistry protocol developed to attach the Aazo monolayer is environmentally stable, and the device generates a reproducible response.
The reason for the observed stability is rooted in the fundamental photoswitching mechanism. Unlike many organic molecules used in photophysics measurements, Aazo does not sensitize singlet oxygen which could then react with the molecule, leading to degradation. Instead, Aazo is changing its conformation on ultrafast timescales, many orders of magnitude faster than other degradation mechanisms may act. Previous work studying the fundamental photochemistry of various Aazo compounds observed high photostability in solid state. For example, for one type of azopyridine derivative, even after exposure to 50,000 cycles of UV pulses (355 nm, 10 mJ per pulse), the absorbance of the molecule was unchanged [47,48]. In our previous work involving a surface bound Aazo compound, individual measurements were up to 34 hours long, and a total of 8 measurements were performed on the same device with no degradation in response observed. Thus, the cumulative exposure was ∼2500 J/cm2 [10].
To investigate the effect of the surface density of Aazo on Δλ, similar measurements to those in Fig. 8 were performed with all [CH3:Aazo] ratios. Additionally, to more thoroughly understand the dependence of the generated response on the circulating optical power, the amount of Pcirc was changed. This data was taken continuously in a similar manner to Fig. 8(b) where each data point represents an entire switching cycle, except the input power was varied. Notably, all data for a single surface density is taken on the same device. The optical response is defined as the λ1300 blueshift for a given amount of circulating power (Δλ/Pcirc). In each [CH3:Aazo] ratio, at least five (Pcirc, Δλ) data points were taken before determining the optical response of the system from the slope of the data. In performing this analysis, it is important to account for variations in device performance and size among the different cavities. For this reason, the conventional strategy is to use Pcirc in Eq. (2) instead of Pin when comparing the optical response between devices.
A summary of the trans-to-cis induced λ1300 shift is shown in Fig. 9(a). Experimentally, Δλ is limited by two factors: Pcirc of the 410 nm laser and the scanning range over which the resonant wavelength can be monitored. Based on experimental results, for a given [CH3:Aazo] ratio, the device exhibited a linear response to the input power throughout the entire series of Pcirc investigated, indicating that higher Pcirc would continue to generate a larger Δλ. Based on classic sensor theory, the device response will saturate at some point [49]. However, this range is outside of our testing set-up capabilities. Additionally, this linearity in response indicates that the monolayer is not degrading even as the device is being cycled a higher number of times and with higher input powers. There is minimal response in the CH3 control therefore, the entire Δλ can be solely attributed to the Aazo photoisomerization. Therefore, it is anticipated that the surface density of photoswitchable Aazo groups would modify the Δλ/Pcirc of the cavity. In order to directly determine the effect of Pcirc and [CH3:Aazo] ratio on the change of the refractive index at 1300 nm (Δn1300), Δn1300 vs Pcirc curves at various [CH3:Aazo] ratio were plotted as Fig. 9(a) inset, where Δn1300 = Δλneff/1300 nm. Since neff and 1300 nm are constants, Δn1300 vs Pcirc curves exhibit exactly the same behaviors as Δλ vs Pcirc curves.
Fig. 9. Δλ as a function of Pcirc across various Aazo surface concentration. (a) For a given surface density of Aazo, as Pcirc increases, Δλ also increases. The slopes of the linear fits are the optical responses Δλ/Pcirc of devices which are characteristic to the surface density of Aazo. The error of each data point (each reversible cycle) was determined by the Gaussian fit on the histogram of the data points (N > 500) in the baseline. Inset: The magnitude of Δn1300 behaves in a similar manner to Δλ with respect to Pcirc among varied [CH3:Aazo] ratios. The error is directed related to the error in the primary data set. (b) Optical responses Δλ/Pcirc with respect to [CH3: Aazo] ratios. Negative values indicate the reduction of Δλ by 410 nm. The error was the error in the linear fitting in part (a). (c) Using the optical cavity refractive index changes from (a, inset) and setting Pcirc = 200 mW, the experimental results on change of optical cavity refractive index (black curve) can be compared to the change of refractive index from ellipsometry studies (red curve). The general trend of two curves is also similar. The error in the black curve was the error in the linear fit of Δn1300 when setting Pcirc = 200 mW. The error in the red curve was taken from the average error of Δn1300 (ellipsometry) among three wafers on each ratio.
To visualize the optical response on each [CH3:Aazo] ratio on the cavity, Fig. 9(b) was plotted. As seen, Δλ/Pcirc decreases as the surface density of Aazo in [CH3:Aazo] decreases, demonstrating the ability to control the surface density of the Aazo group through the integration of the CMPS. As expected, the control CH3-coated device does not show any response, and the optical response extends to nearly -0.3 pm/mW for Aazo-coated devices.
The Δλ in Fig. 9(a) blueshifted a few tens of pm, which is lower than FEM simulation prediction of approximately -130 pm. However, as mentioned, the FEM results should be viewed as an upper bound because the calculations most likely over-approximate both the density of Aazo molecules and the switching efficiency. A more accurate model system is the ellipsometry results. The ellipsometry data was acquired with a fixed power blue laser operating at 450 nm, so it is not possible to quantitatively compare the resonant cavity and ellipsometry results over the entire power range investigated. However, the general trends can be compared. To perform this analysis, the responsivity values from Fig. 9(b) are used to calculate a wavelength shift assuming a Pcirc of 200 mW. Using the previously defined expression, Δn1300 is computed. While this approach most likely results in an over-approximation for the higher Aazo surface densities, using a lower value would result in null or even negative responses for lower surface densities, which is not physically realizable.
Using these values, the resonant cavity index change values are plotted alongside the ellipsometry index change values (Fig. 9(c)). As can be seen, the response is similar. This provides further evidence that the blueshift can be solely attributed to the photoswitchable Aazo and, by varying the density of Aazo on the surface and the magnitude of the optical response can be controlled. Furthermore, Δn1300 (ellipsometry) has an excellent agreement on previous FEM simulations. While the CH3-coated sample indicates a negligible positive change on Δn1300 (ellipsometry) upon blue laser exposure, all Aazo-coated samples show the reduction of Δn1300 (ellipsometry) up to -6.0E-4. This is within the same order of magnitude of FEM Δneff (-2.1E-4), and it further strengthens that FEM is qualitatively reasonable despite an over-estimation.
4. Conclusions
By combining photoswitchable organic molecules with integrated optical devices, all-optical control of the resonant wavelength of integrated optical resonant cavities is demonstrated. Using a highly controlled process to attach the Aazo layer, the cavity Q at both 410 nm and 1300 nm is maintained, allowing low input power for resonant wavelength modulation. To further increase the dynamic range, a competitive surface chemistry approach was developed which enabled two molecules to be simultaneously anchored to the device surface, providing control over the density of the photoswitchable Aazo groups. A single waveguide was used to simultaneously excite resonant modes in the visible and the near-IR. The visible resonance was used to photoswitch the Aazo groups while the near-IR resonance probed the state of the molecules. The magnitude of the optical response of the laser-induced photoswitching is governed by the surface density of Aazo groups and the circulating power of the 410 nm laser. The change in refractive index measured by ellipsometry at 1300 nm (Δn1300 (ellipsometry)) with respect to [CH3:Aazo] ratio correlates to that of the cavity index change with respect to [CH3:Aazo] ratio at 1300 nm (Δn1300). Lastly, the observed switching behavior is stable over multiple switching cycles and is conserved, even after 6 months of storage in an ambient environment.
This work sets the stage for a plethora of future research directions in chemistry, physics, and integrated photonics. The development of fully integrated optically-triggerable photonic devices with photo-responsive organic monolayers is one clear path. The ability to anchor a monolayer of organic dye to an optical device surface at varying surface densities is the equivalent of being able to alter the doping levels in a device after fabrication. Before moving to such a platform, the fabrication procedure of the device must be modified so that only the resonator itself is functionalized with the organic molecule. One innovative fabrication approach recently demonstrated would enable this type of selective patterning by opening a window in the cladding layer directly over a waveguide-coupled ring [50]. On the other hand, using the device to investigate the dynamics of surface effects in photoswitchable organic materials could open new lines of investigations for organic chemists.
Funding
Army Research Office (W911NF1810033); Office of Naval Research (N00014-17-1-2270).
Acknowledgments
The authors would like to thank Mark Veksler (University of Southern California) for the renderings.
Disclosures
The authors declare no conflicts of interest.
References
1. H. M. D. Bandara and S. C. Burdette, “Photoisomerization in different classes of azobenzene,” Chem. Soc. Rev. 41(5), 1809–1825 (2012). [CrossRef]
2. T. Schultz, J. Quenneville, B. Levine, A. Toniolo, T. J. Martínez, S. Lochbrunner, M. Schmitt, J. P. Shaffer, M. Z. Zgierski, and A. Stolow, “Mechanism and Dynamics of Azobenzene Photoisomerization,” J. Am. Chem. Soc. 125(27), 8098–8099 (2003). [CrossRef]
3. C. R. Crecca and A. E. Roitberg, “Theoretical Study of the Isomerization Mechanism of Azobenzene and Disubstituted Azobenzene Derivatives,” J. Phys. Chem. A 110(26), 8188–8203 (2006). [CrossRef]
4. F. Castet, V. Rodriguez, J.-L. Pozzo, L. Ducasse, A. Plaquet, and B. Champagne, “Design and Characterization of Molecular Nonlinear Optical Switches,” Acc. Chem. Res. 46(11), 2656–2665 (2013). [CrossRef]
5. C. J. Barrett, J. Mamiya, K. G. Yager, and T. Ikeda, “Photo-mechanical effects in azobenzene-containing soft materials,” Soft Matter 3(10), 1249–1261 (2007). [CrossRef]
6. Z. F. Liu, K. Hashimoto, and A. Fujishima, “Photoelectrochemical information storage using an azobenzene derivative,” Nature 347(6294), 658–660 (1990). [CrossRef]
7. T. Ikeda and O. Tsutsumi, “Optical Switching and Image Storage by Means of Azobenzene Liquid-Crystal Films,” Science 268(5219), 1873–1875 (1995). [CrossRef]
8. K. G. Yager and C. J. Barrett, “Novel photo-switching using azobenzene functional materials,” J. Photochem. Photobiol., A 182(3), 250–261 (2006). [CrossRef]
9. A. A. Beharry and G. Andrew Woolley, “Azobenzene photoswitches for biomolecules,” Chem. Soc. Rev. 40(8), 4422–4437 (2011). [CrossRef]
10. A. Kovach, J. He, P. J. G. Saris, D. Chen, and A. M. Armani, “Optically tunable microresonator using an azobenzene monolayer,” AIP Adv. 10(4), 045117 (2020). [CrossRef]
11. A. Chiasera, Y. Dumeige, P. Féron, M. Ferrari, Y. Jestin, G. N. Conti, S. Pelli, S. Soria, and G. C. Righini, “Spherical whispering-gallery-mode microresonators,” Laser Photonics Rev. 4(3), 457–482 (2010). [CrossRef]
12. J. Topolancik and F. Vollmer, “All-optical switching in the near infrared with bacteriorhodopsin-coated microcavities,” Appl. Phys. Lett. 89(18), 184103 (2006). [CrossRef]
13. S. Roy, M. Prasad, J. Topolancik, and F. Vollmer, “All-optical switching with bacteriorhodopsin protein coated microcavities and its application to low power computing circuits,” J. Appl. Phys. 107(5), 053115 (2010). [CrossRef]
14. Y. Li, H. Zhang, B. Liu, W. Lin, D. Yan, C. Yang, and J. Wu, “Laser-tuned whispering gallery modes in a silica-based microsphere resonator integrated with ethyl-orange-doped polyvinyl alcohol coating,” Sens. Actuators, B 238, 98–104 (2017). [CrossRef]
15. K. Tabata, D. Braam, S. Kushida, L. Tong, J. Kuwabara, T. Kanbara, A. Beckel, A. Lorke, and Y. Yamamoto, “Self-assembled conjugated polymer spheres as fluorescent microresonators,” Sci. Rep. 4(1), 5902 (2015). [CrossRef]
16. H. K. Hunt and A. M. Armani, “Bioconjugation Strategies for Label-Free Optical Microcavity Sensors,” IEEE J. Sel. Top. Quantum Electron. 20(2), 121–133 (2014). [CrossRef]
17. J.-Y. Byeon, F. T. Limpoco, and R. C. Bailey, “Efficient Bioconjugation of Protein Capture Agents to Biosensor Surfaces Using Aniline-Catalyzed Hydrazone Ligation,” Langmuir 26(19), 15430–15435 (2010). [CrossRef]
18. S. Mehrabani, A. J. Maker, and A. M. Armani, “Hybrid Integrated Label-Free Chemical and Biological Sensors,” Sensors 14(4), 5890–5928 (2014). [CrossRef]
19. G. Lin, A. Coillet, and Y. K. Chembo, “Nonlinear photonics with high-Q whispering-gallery-mode resonators,” Adv. Opt. Photonics 9(4), 828–890 (2017). [CrossRef]
20. A. Kovach, D. Chen, J. He, H. Choi, A. H. Dogan, M. Ghasemkhani, H. Taheri, and A. M. Armani, “Emerging material systems for integrated optical Kerr frequency combs,” Adv. Opt. Photonics 12(1), 135–222 (2020). [CrossRef]
21. J. He, H. Chen, J. Hu, J. Zhou, Y. Zhang, A. Kovach, C. Sideris, M. C. Harrison, Y. Zhao, and A. M. Armani, “Nonlinear nanophotonic devices in the Ultraviolet to Visible wavelength range,” arXiv:2006.08041 [physics] (2020).
22. S. M. Spillane, T. J. Kippenberg, O. J. Painter, and K. J. Vahala, “Ideality in a Fiber-Taper-Coupled Microresonator System for Application to Cavity Quantum Electrodynamics,” Phys. Rev. Lett. 91(4), 043902 (2003). [CrossRef]
23. S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, “Hybrid microcavity humidity sensor,” Appl. Phys. Lett. 102(24), 241101 (2013). [CrossRef]
24. A. D. Becke, “Density-functional thermochemistry. III. The role of exact exchange,” J. Chem. Phys. 98(7), 5648–5652 (1993). [CrossRef]
25. D. Hadji and A. Rahmouni, “Theoretical study of nonlinear optical properties of some azoic dyes,” Mediterr. J. Chem. 4(4), 185–192 (2015). [CrossRef]
26. N. Peor, R. Sfez, and S. Yitzchaik, “Variable Density Effect of Self-Assembled Polarizable Monolayers on the Electronic Properties of Silicon,” J. Am. Chem. Soc. 130(12), 4158–4165 (2008). [CrossRef]
27. E. Ortyl and S. Kucharski, “Refractive index modulation in polymeric photochromic films,” Cent. Eur. J. Chem. 1(2), 137–159 (2003). [CrossRef]
28. J. Dokić, M. Gothe, J. Wirth, M. V. Peters, J. Schwarz, S. Hecht, and P. Saalfrank, “Quantum Chemical Investigation of Thermal Cis-to-Trans Isomerization of Azobenzene Derivatives: Substituent Effects, Solvent Effects, and Comparison to Experimental Data,” J. Phys. Chem. A 113(24), 6763–6773 (2009). [CrossRef]
29. H. S. Choi, X. Zhang, and A. M. Armani, “Hybrid silica-polymer ultra-high-Q microresonators,” Opt. Lett. 35(4), 459–461 (2010). [CrossRef]
30. A. Kaplan, M. Tomes, T. Carmon, M. Kozlov, O. Cohen, G. Bartal, and H. G. L. Schwefel, “Finite element simulation of a perturbed axial-symmetric whispering-gallery mode and its use for intensity enhancement with a nanoparticle coupled to a microtoroid,” Opt. Express 21(12), 14169–14180 (2013). [CrossRef]
31. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, “Ultra-high-Q toroid microcavity on a chip,” Nature 421(6926), 925–928 (2003). [CrossRef]
32. X. Shen, R. C. Beltran, V. M. Diep, S. Soltani, and A. M. Armani, “Low-threshold parametric oscillation in organically modified microcavities,” Sci. Adv. 4(1), eaao4507 (2018). [CrossRef]
33. A. Airinei, N. Fifere, M. Homocianu, C. Gaina, V. Gaina, and B. C. Simionescu, “Optical Properties of Some New Azo Photoisomerizable Bismaleimide Derivatives,” Int. J. Mol. Sci. 12(9), 6176–6193 (2011). [CrossRef]
34. N. R. Krekiehn, M. Müller, U. Jung, S. Ulrich, R. Herges, and O. M. Magnussen, “UV/Vis Spectroscopy Studies of the Photoisomerization Kinetics in Self-Assembled Azobenzene-Containing Adlayers,” Langmuir 31(30), 8362–8370 (2015). [CrossRef]
35. M. Onoue, M. R. Han, E. Ito, and M. Hara, “Step-wise decomposition process of azobenzene self-assembled monolayers,” Surf. Sci. 600(18), 3999–4003 (2006). [CrossRef]
36. W.-W. Zhang, X.-M. Ren, H.-F. Li, C.-S. Lu, C.-J. Hu, H.-Z. Zhu, and Q.-J. Meng, “Study on Self-Assembled Monolayers of Functionalized Azobenzene Thiols on Gold: XPS, Electrochemical Properties, and Surface-Enhanced Raman Spectroscopy,” J. Colloid Interface Sci. 255(1), 150–157 (2002). [CrossRef]
37. J. C. Knight, G. Cheung, F. Jacques, and T. A. Birks, “Phase-matched excitation of whispering-gallery-mode resonances by a fiber taper,” Opt. Lett. 22(15), 1129–1131 (1997). [CrossRef]
38. A. Yariv, “Universal relations for coupling of optical power between microresonators and dielectric waveguides,” Electron. Lett. 36(4), 321–322 (2000). [CrossRef]
39. J. M. Ward, F. Lei, S. Vincent, P. Gupta, S. K. Mondal, J. Fick, and S. N. Chormaic, ““Excitation of whispering gallery modes with a “point-and-play,” fiber-based, optical nano-antenna,” Opt. Lett. 44(13), 3386–3389 (2019). [CrossRef]
40. X. Yu, Z. Wang, M. Buchholz, N. Füllgrabe, S. Grosjean, F. Bebensee, S. Bräse, C. Wöll, and L. Heinke, “cis -to- trans isomerization of azobenzene investigated by using thin films of metal–organic frameworks,” Phys. Chem. Chem. Phys. 17(35), 22721–22725 (2015). [CrossRef]
41. B. A. Rose, A. J. Maker, and A. M. Armani, “Characterization of thermo-optic coefficient and material loss of high refractive index silica sol-gel films in the visible and near-IR,” Opt. Mater. Express 2(5), 671 (2012). [CrossRef]
42. Y. X. Zhuang, O. Hansen, T. Knieling, C. Wang, P. Rombach, W. Lang, W. Benecke, M. Kehlenbeck, and J. Koblitz, “Thermal stability of vapor phase deposited self-assembled monolayers for MEMS anti-stiction,” J. Micromech. Microeng. 16(11), 2259–2264 (2006). [CrossRef]
43. A. Chandekar, S. K. Sengupta, and J. E. Whitten, “Thermal stability of thiol and silane monolayers: A comparative study,” Appl. Surf. Sci. 256(9), 2742–2749 (2010). [CrossRef]
44. R. Brückner, “Properties and structure of vitreous silica. I,” J. Non-Cryst. Solids 5(2), 123–175 (1970). [CrossRef]
45. H. K. Hunt and A. M. Armani, “Recycling microcavity optical biosensors,” Opt. Lett. 36(7), 1092–1094 (2011). [CrossRef]
46. H. K. Hunt, C. Soteropulos, and A. M. Armani, “Bioconjugation Strategies for Microtoroidal Optical Resonators,” Sensors 10(10), 9317–9336 (2010). [CrossRef]
47. J. Garcia-Amorós, M. Díaz-Lobo, S. Nonell, and D. Velasco, “Fastest Thermal Isomerization of an Azobenzene for Nanosecond Photoswitching Applications under Physiological Conditions,” Angew. Chem., Int. Ed. 51(51), 12820–12823 (2012). [CrossRef]
48. T. Caronna, F. Fontana, B. Marcandalli, and E. Selli, “Photostability of substituted 4-diethylaminoazobenzenes,” Dyes Pigm. 49(2), 127–133 (2001). [CrossRef]
49. H. K. Hunt and A. M. Armani, “Label-free biological and chemical sensors,” Nanoscale 2(9), 1544–1559 (2010). [CrossRef]
50. C. Wang, M. Zhang, M. Yu, R. Zhu, H. Hu, and M. Loncar, “Monolithic lithium niobate photonic circuits for Kerr frequency comb generation and modulation,” Nat. Commun. 10(1), 978 (2019). [CrossRef]
