I. INTRODUCTION
Heat management and control of optical devices is of fundamental importance for their practical applications [
1,
2]. For integrated photonic devices with a compact footprint and tight mode confinement, and particularly for materials that do not exhibit second-order optical nonlinearities such as the Pockels effect [
3], the importance of precisely engineering their thermo-optic effects is even more pronounced [
4,
5]. Over the past decade, with the rapid advancement of integrated photonics, extensive research has been dedicated to investigating and harnessing thermo-optic effects to manipulate light in integrated photonic devices, particularly those based on centrosymmetric materials [
4,
6]. This has enabled the realization of a variety of functionalities such as mode-locking [
7,
8], optical switches [
9,
10], logic gates [
11], power limiters [
11,
12], and optical memories [
11,
13,
14].
As an important complementary metal–oxide–semiconductor (CMOS)-compatible integrated platform, high index doped silica (HIDS) has been extensively utilized for diverse linear and nonlinear integrated photonic devices for a range of applications [
15,
16,
17,
18,
19,
20,
21,
22]. HIDS possesses a host of attractive optical properties, such as low linear optical absorption over a broad band, a reasonably strong Kerr nonlinearity (about 5 times that of silica), and negligible nonlinear optical absorption [
23,
24,
25,
26]. The combination of these properties and its strong compatibility with the globally established CMOS infrastructure, contributes to the exceptional performance and versatility of HIDS devices in various applications within the field of integrated photonics.
Despite its proven success for many optical applications, investigation of the thermo-optic effects in HIDS devices has not been as extensive as in other integrated photonic devices made from other centrosymmetric materials such as silicon and silicon nitride [
14,
27,
28]. There remains a need for exploration and an understanding of the thermo-optic properties of HIDS devices to fully leverage their potential in integrated photonics. In this paper, we address this issue by providing a comprehensive experimental characterization and theoretical analysis of these effects in HIDS integrated devices. By fitting experimental results with theory, we obtain fundamental parameters that characterize the thermo-optic properties of HIDS devices, including the thermo-optic coefficient, the efficiency for the optically induced thermo-optic process, and the thermal conductivity. We also provide a comparison of these parameters with those of other materials used for CMOS-compatible integrated photonic platforms, such as silicon, silicon nitride, and silica. These findings provide a comprehensive understanding of the thermo-optic properties of HIDS devices, important for effectively controlling and engineering these devices in many applications.
II. DEVICE FABRICATION AND CHARACTERISATION
Figure 1(a) shows a schematic of an add-drop MRR made from HIDS. A microscope image of the fabricated device is shown in
Figure 1(b). The MRR was fabricated via CMOS-compatible processes [
15,
18]. First, the HIDS film with a refractive index of ~1.66 was deposited using plasma enhanced chemical vapor deposition (PECVD). Next, waveguides with exceptionally low sidewall roughness were formed by employing deep ultraviolet photolithography techniques and reactive ion etching. Finally, a silica layer with a refractive index of ~1.45 was deposited via PECVD as the upper cladding. The waveguide cross section of both the MRR and the two coupling bus waveguides was ~3 µm × ~2 µm. The MRR had a radius of ~592.1 µm, which corresponded to a free spectral range (FSR) of ~0.4 nm (
i.e., ~49 GHz). Note that although there were a number of concentric rings in
Figure 1(b), only the central ring was coupled with the through / drop bus waveguides to form an MRR with a radius of ~592.1 µm ‒ the rest were simply used to enable easy identification by eye. A similar MRR layout was used in our previous work on HIDS devices [
25,
26]. The input and output ports of the MRR were connected to specially designed mode converters that were packaged with fiber pigtails. The fiber-to-chip coupling loss was ~1.5 dB / facet, with this low value enabled through the use of on-chip mode converters to the pigtailed fibers.
Figure 1(c) shows the measured transmission spectra of a fabricated HIDS MRR for both transverse magnetic (TE) and transverse electric (TM) polarizations. The wavelength of a tunable continuous-wave (CW) laser was scanned at a constant input power of ~0 dBm to measure the transmission spectra, and a polarization controller (PC) was employed to adjust the input polarization. The input power here and in our following analysis refers to the power coupled into the device (i.e., the on-chip power), with the fiber-to-chip coupling loss being subtracted from the laser’s output power. The free spectral range (FSR) of the TE- and TM- polarized transmission spectra was ~0.4 nm, which corresponded to ~49 GHz. By tuning the PC, the maximum polarization extinction ratios for the TE- and TM- polarized resonances were > 30 dB.
Figure 1(d) shows zoom-in views of single TE- and TM- polarized resonances at ~1550.381 nm and ~1550.288 nm. There was no significant asymmetry in the measured resonance spectral lineshape, indicating that the thermal effect at the input power of ~0 dBm was negligible. The full widths at half maximum (FWHMs) of the TE- and TM-polarized resonances were ~0.0015 nm (~190 MHz) and ~0.0020 nm (~250 MHz), respectively, which corresponded to Q factors of ~1.0 × 10
6 and ~7.8 × 10
5, respectively. In addition, the -20-dB bandwidths of the TE- and TM- polarized resonances were ~1.75 GHz and ~1.87 GHz, respectively. By using the scattering matrix method [
30,
31] to fit the measured spectra in
Figure 1(d), we obtained the device parameters for the HIDS MRR that will be used for analysis in the subsequent sections. These parameters, together with the specific material and waveguide parameters, are summarized in Table I.
Table I.
DEVICE PARAMETERS OF THE HIDS MRR.
Table I.
DEVICE PARAMETERS OF THE HIDS MRR.
III. THERMO-OPTIC COEFFICIENT
The thermo-optic coefficient of a material is a fundamental parameter that indicates how its refractive index changes with environmental temperature, which plays an important role in the design and engineering of relevant devices [
32]. In this section, we characterize the thermo-optic coefficient of HIDS by measuring the transmission spectra of the HIDS MRR with varying chip temperature.
When there are changes in environmental temperature, the thermo-optic effect causes changes in the effective refractive index of the HIDS waveguides. Consequently, this leads to a shift in the resonance wavelengths of the HIDS MRR.
Figure 2(a) shows the TE- and TM- polarized transmission spectra of the HIDS MRR when the chip temperature changed from 23 °C to 30 °C, respectively. We measured the shifts of three resonances, including a TE-polarized resonance and two TM-polarized resonances (TM1 and TM2). Specially, the TE-polarized resonance was located between the two TM-polarized resonances. To adjust the temperature of the integrated chip mounted on a stage, a temperature controller was employed. The input power of the scanned CW laser was maintained as ~0 dBm (
i.e., the same as that in
Figure 1) in order to mitigate noticeable thermal effects. It is important to highlight that despite the changes in environmental temperature, no significant asymmetry was observed in the measured resonance spectral lineshape. This observation indicates that changes in environmental temperature induced by the temperature controller have a minimal impact on the asymmetry of the resonance spectral lineshape and does not induce significant optical bistability that will be discussed in the next section [
33].
Figure 2(b) shows the resonance wavelength shifts versus the chip temperature, which were extracted from the results in
Figure 2(a). The TE-polarized resonance redshifted at a rate of ∼14.2 pm / °C, whereas the two TM-polarized resonances exhibited a redshift rate of ∼13.4 pm / °C.
Figure 2(c) depicts the changes in the waveguide effective refractive indices versus the chip temperature. These results were calculated using the measured results in
Figure 2(b), along with the relationship between the resonance wavelengths and the waveguide effective refractive index given by [
34,
35]:
where
neff is the effective refractive index of the HIDS waveguide,
L is the circumference of the HIDS MRR, and
m represents the
mth resonance, with
λm denoting the corresponding resonance wavelength.
In
Figure 2(c), the TE mode displays a change in the effective refractive index at a rate of ~1.52 × 10
-5 /°C, while the TM mode changes at a rate of ~1.43 × 10
-5 /°C. The difference in these rates can be attributed to the asymmetric cross section of the HIDS waveguide. Based on these results, we further extract the thermo-optic coefficient of the HIDS material at various chip temperatures by using Lumerical FDTD commercial mode solving software. The results are presented in
Figure 2(d). In our simulation, the thermo-optic coefficient of silica is assumed to be ~1.09 × 10
-5 /°C [
36]. The thermo-optic coefficients of HIDS in
Figure 2(d) do not show significant temperature dependence. We also note that the average values of the thermo-optic coefficients of HIDS derived from the TE and TM modes exhibit remarkable similarity, at ~1.49 × 10
-5 /°C and ~1.44 × 10
-5 /°C, respectively. This close resemblance between the coefficients reflects that the HIDS does not exhibit significant anisotropy in terms of its thermo-optic coefficient.
IV. OPTICALLY INDUCED THERMO-OPTIC RESPONSE
When a material is illuminated with intense light, optical absorption leads to heat generation that raises the local temperature. This in turn modifies the material’s refractive index, thereby influencing the propagation of light through the material. In this optically induced thermo-optic process, the change in the material’s refractive index
n due to the temperature variation induced by the optical field can be modeled as [
27,
37]:
where
n0 is the material’s refractive index when not exposed to light, and?
n2 ·
I is the refractive index change due to the optically induced temperature change, with
I denoting the light intensity and?
n2 denoting the coefficient that characterizes the efficiency for this process. In this section, we characterize the?
n2 of HIDS by measuring the transmission spectra of the HIDS MRR at various input powers. It is worth noting that
Eq. (2) is the same as that used for modeling the nonlinear Kerr optical effect [
38,
39]. We also note that
is an effective response, in that it is device geometry dependent, including the MRR Q-factor, coupling coefficient etc..
For the optically induced refractive index change, in addition to the optically induced thermo-optic effect, there will also be a presence of the Kerr optical effect. Despite having the same mathematical modeling as shown in
Eq. (2), these two effects are associated with different physical processes that exhibit distinct characteristics. For example, compared to the Kerr optical effect that has an ultrafast time response on the order of 10
-15 s [
40,
41], the time response for the optically induced thermo-optic effect is much slower, typically on the order of 10
-6 ‒10
-3 s [
31,
42].
When the wavelength of incident light is on resonance with the MRR, the incident light power converts into heat more efficiently, being enhanced significantly by the ring resonance, leading to an efficient change in the effective refractive index of the HIDS waveguides caused by the thermo-optic effect. This refractive index change also results in a shift in the resonance wavelengths of the HIDS MRR.
Figs. 3(a-i) and
(a-ii) show measured transmission spectra of the HIDS MRR at different input powers for TE and TM polarizations, respectively. As the input power increased, a redshift in the resonance wavelengths was observed, accompanied by increasingly asymmetric resonance spectra. The spectra also exhibit a steepened transition edge, indicating the presence of the optical bistability [
43,
44].
Depending on the dominating nonlinear mechanism, the resonance wavelengths can experience either a blue or red shift. In previous work on bistability in silicon MRRs at room temperature, it was observed that the resonance wavelengths initially exhibited a blueshift and subsequently transitioned to a redshift as the input power increased [
45]. This is because the free-carrier dispersion (FCD) that results in a decreased refractive index of silicon dominates at low powers, whereas the thermo-optic effect that leads to an increased refractive index dominates at high powers [
45]. Here, we only observed a redshift in the resonance wavelengths, mainly due to the dominating thermo-optic effect and negligible FCD for the HIDS MRR [
26,
29] and the fact that the magnitude of the all-optical Kerr component of the index change tends to be much smaller for typical CW powers.
Figure 3(b) shows the shifts of the resonance wavelength versus the input power. For both the TE- and TM- polarizations, the positive Δ
λ (which indicates a redshift) exhibits a nearly linear relationship with the input power. By linearly fitting the measured results, we obtained the rates for the resonance wavelength shift, which were ~0.4655 pm / mW and ~0.3144 pm / mW for the TE and TM polarizations, respectively.
Figure 3(c) shows the changes in the waveguide effective refractive indices versus the input power for both TE and TM polarizations. These results were calculated based on Eq. (1), using the measured results in Figure 3(b). As the input power increased from ~2 mW to ~16 mW, the effective refractive indices of TE and TM modes displayed changes of ~6.533 × 10-6 and ~5.012 × 10-6, respectively. These changes correspond to average rates of ~4.985 × 10-7 /mW and ~3.335 × 10-7 /mW, respectively.
Figure 3(d) shows the waveguide effective?
n2, denoted as
n2, eff, versus the input power for both TE and TM polarizations, which were extracted from the results in
Figure 3(c). The?
n2, eff was calculated by [
27]
where Δ
n is the refractive index change, and
I is the light intensity in the MRR given by [
27]
In
Eq. (4),
Pin is the input power,
Aeff is the effective mode area [
26,
46], and
BUF is the intensity build-up factor of the MRR that can be expressed [
47,
48]
where
t1,2 and
a are the fit MRR parameters in
Table 1.
In
Figure 3(d), the average values of the extracted?
n2, eff’ for the TE and TM polarizations are ~4.514 × 10
-14 cm
2 / W and ~2.903 × 10
-14 cm
2 / W, respectively. The difference in these responses can be attributed to the asymmetric cross section of the HIDS waveguide that results in different optical field distributions for the two modes. Based on the results in
Figure 3(d), we further extract the
for the HIDS material, denoted as?
n2, HIDS, according to [
26,
49]:
where
D is the integral of the optical fields over the material regions,
Sz is the time-averaged Poynting vector calculated using Lumerical FDTD commercial mode solving software,
n0 (
x,
y) and?
n2 (
x,
y) are the linear refractive index and
profiles over the waveguide cross section, respectively. The value of?
n2 for silica used in our calculation was ~1.8 × 10
-14 cm
2 / W [
36,
50].
Figure 3(e) shows the extracted
n2, HIDS versus the input power. The average values of?
n2, HIDS derived from the TE and TM modes are ~3.1× 10
-14 cm
2 / W and ~2.7 × 10
-14 cm
2 / W, respectively. The close resemblance between them reflects that the HIDS does not exhibit significant anisotropy in terms of its?
n2. The results in
Figure 3(e) also confirm that the predominant cause of the observed nonlinearity is thermal in nature. This is also supported by the fact that the Kerr nonlinear coefficient of HIDS (~1.3 × 10
–15 cm
2 / W [
18,
51]) was over one order of magnitude lower. Although there are minor fluctuations in?
n2, HIDS across various input powers in
Figure 3(d), these variations are not significant. Considering the limited input power range (
i.e., ~2 mW to ~16 mW), it can be inferred that the?
n2, HIDS values will exhibit a relatively stable behavior [
15,
18]. Hence, the slight power-dependent variations in?
n2, HIDS are likely attributable to measurement errors.
V. OPTICAL BISTABILITY
Due to a steepened asymmetric transitional edge, optical bistability arising from nonlinear thermo-optic effects has been used for controlling light with light and achieving optical switches [
9,
10].
Figure 4 shows the measured output power as a function of the input power when it was progressively increased from ~1 mW to ~8 mW. For comparison, we also plot the downward output power as the input power was subsequently reduced back to ~1 mW. In
Figs. 4(a) ‒
(c), we show the results for three initial wavelength detunings of
δ = ~1.3, ~1.5, and ~1.7, respectively. The
δ is defined as:
where
λlaser is the wavelength of the input CW light,
λres is the resonance wavelength of the MRR measured at a low input CW power of ~0 dBm (
i.e., the same as that in
Figure 1 and does not induce significant asymmetry in the measured resonance spectral lineshape), and Δ
λ is the 3-dB bandwidth of the resonance. In our measurements, we chose a TE-polarized resonance centered at
λres = ~1550.3758 nm and a TM-polarized resonance centered at
λres = ~1550.2826 nm. During the measurements, the maximum polarization extinction ratios were kept > 30 dB.
In
Figs. 4(a) ‒
(c), redshifts of the resonance wavelengths can be observed for both TE and TM polarizations. During the upward sweeping, the output power first exhibited a steady and continuous increase, followed by a sudden jump towards higher output power. Conversely, during the downward sweeping with decreasing input power, there was a sudden jump toward lower output power after a gradual decrease in the output power. Clearly, the presence of a hysteresis loop resulting from the upward and downward wavelength sweeping provides evidence for the existence of optical bistability in the HIDS MRR [
52]. As
δ was increased from ~1.3 to ~1.7, the input power threshold for optical bistability increased, and the hysteresis loop became more open. These phenomena are similar to those observed in Refs. [
53,
54]. We also note that the TE-polarized resonance exhibits a more open hysteresis loop compared with the TM-polarized resonance at the same
δ. This observation shows agreement with the relatively large redshift of the resonance wavelength for the TE polarization in
Figure 3(b).
Figure 5 shows the measured and theoretical output powers versus the input power for both TE and TM polarizations. The theoretical curves were calculated based on the theory in Refs. [
33,
52,
55], using both the device parameters in
Table I and the fit?
n2, HIDS in
Figure 3(e). In principle, bistable behavior occurs in the resonator response because, under specific conditions, the output power yields multiple distinct solutions for a given input power. Consequently, the resonator can switch between these solutions due to the influence of noise [
33,
55]. In
Figure 5, the measured results show good agreement with the theoretical curves, providing further confirmation of the accuracy of the fit thermo-optic property parameters for the HIDS devices.
VI. THERMAL CONDUCTIVITY
Thermal conductivity, a parameter that defines a material’s ability to conduct heat, has been widely used for modeling thermal transport for applications related to thermal management and energy storage [
40,
56,
57,
58,
59,
60]. In this section, the thermal conductivity of HIDS is characterized by fitting the measured transmission spectra of the HIDS MRR at various input powers with theoretical simulations.
Figure 6(a-i) and
(a-ii) show the simulated TE and TM mode profiles for the HIDS waveguide. The corresponding effective refractive indices were
neff_TE = ~1.560 and
neff_TM = ~1.558 at 1550 nm. To further investigate the heat generated in the HIDS waveguide, we simulated the cross-sectional temperature distribution for both TE and TM polarizations.
Figures 6(b-i) and
(b-ii) show the steady-state temperature distributions at an incident power of 16 mW, which were obtained by solving the heat equation [
1]:
where
T is the steady-state temperature distribution,
K is the thermal conductivity, and
q is the heat source resulting from the absorption of light. In
Eq. (8), ∇
T denotes the gradient of
T, and ∇ acting on the vector function
K∇
T is the corresponding divergence operator. The heat source was calculated based on the TE and TM mode profiles in
Figs. 6(a-i) and
(a-ii) using [
61]:
where
σ is the electrical conductivity of the waveguide in
Table I and
E is the amplitude of the optical field simulated in
Figure 6(a). It is worth noting that the build-up factor
BUF in
Eq. (5) was taken into account when calculating the optical intensity in the MRR. In our simulation, the initial temperature
T0 was set to 23°C, which was the ambient temperature during the experiments.
When there are changes in the input power, the material conducts heat, leading to a rise in temperature and a redshift of the resonance wavelength. According to the results in
Figs. 2(c) and
3(c), we calculated the device temperature variation versus the input power. As shown in
Figure 6(c), at an input power of 16mW, the temperature variations for the TE and TM modes are Δ
T = ~0.4298 °C and ~0.3495 °C, respectively. By fitting these temperature variations with the temperature distributions in
Figure 6(b), we obtained the thermal conductivity for the HIDS, as shown in
Figure 6(d). For the TE and TM polarizations, the average values for the fitted thermal conductivity were ~0.66 W / (m · °C) and ~0.96 W / (m · °C). We note that the thermal conductivity of the HIDS is lower than that of silica (
i.e., ~1.4 W / (m · °C) [
61]). This can be attributed to the introduction of the doping material that slows down the lattice vibration coupling and the energy transfer. The specific values depend on the type and concentration of the doping element used, as well as the material’s fabrication method and structure. In our fabrication, the HIDS was developed by using the high-index-contrast materials such as Ta
2O
5 and polymeric systems [
62,
63]. We note that the thermal conductivity of Ta
2O
5 is much lower than that of silica (i.e., ~0.026 W / (m · °C) [
64,
65]), which could be a reason for the relatively low thermal conductivity of HIDS. Another possible reason is that the thermal conductivity of silica films grown by PECVD (
i.e., ~1.0 W / (m · °C) [
66]) is slightly lower than the previously mentioned value of silica. Based on
Eq. (8), the low thermal conductivity of the HIDS waveguide restricts heat propagation, leading to a higher concentration of thermal energy within the waveguide. Consequently, this amplifies the temperature increase, which, in turn, facilitates the attainment of more pronounced optical bistability.