Epsilon-near-zero medium for optical switches in a monolithic waveguide chip at 1.9 μm

1 Introduction

A complementary metal–oxide–semiconductor (CMOS)-compatible broadband saturable absorber (SA) is required for all-optical modulation in passive photonics circuits [1] and integrated on-chip laser sources [2], which are two critical components in realizing fully integrated hybrid photonics devices. Semiconductor saturable absorber mirrors (SESAMs) are currently the most commonly used absorbers in solid-state lasers and have dramatically pushed ultrafast lasers to their limits [3]. However, the modulation bandwidth of SESAMs is limited by the semiconductor bandgap, and the stringent fabrication processes of SESAMs can be challenging to be integrated with other on-chip optical elements [4]. In recent years, various low-dimensional materials have been exploited as the candidates, yet the material compatibility with the existing photonics platforms and fine-controlled material fabrications remain hurdles. An epsilon-near-zero (ε~0, ENZ) material, in which the real component of permittivity vanishes, exhibits distinct optical features such as large nonlinearity, near-zero refractive index, decoupling of electricity and magnetism, and infinite phase velocity [5], [6], [7], was first proposed in the case of metamaterials [7]. Large nonlinearity is of particular interest for the nonlinear optics and ultrafast photonics communities. Recently, degenerate semiconductor transparent conducting oxides (TCOs) with limited imaginary permittivity (low optical loss) have emerged as easily available, CMOS-compatible, and ENZ wavelength-tunable ENZ medium alternatives for integrated photonics [8], [9], [10].

Indium tin oxide (ITO), one of the TCOs, has an optical band gap of ~4 eV and thus features high transmittance for photons of lower energy (λ>~350 nm) [11]. However, ITO still absorbs photons with energy smaller than the bandgap via free-carrier absorption in its conduction band. Photon absorption induces hot-electron redistribution near the Fermi surface, which results in a change of the real and imaginary components of the permittivity, thus leading to optical nonlinearity of ITO [8], [12], [13]. Recently, large nonlinearities, including 170% of linear refractive index change and 30% of transmittance change, of ITO have been demonstrated at the ENZ wavelength (~1240 nm) [8] with an incident laser intensity as high as 66 GW/cm². In this study, we show that an ITO film can be an efficient all-optical modulator in its ENZ region with remarkably low light illumination. Optical amplitude modulation of 7.4% has been achieved under the incident light intensity of 306 MW/cm². This low bleaching threshold was further confirmed via implementing the ITO film into a low-gain waveguide chip for optical switches at 1.9 μm. Q-switched, Q-switched mode-locked, and mode-locked waveguide lasers have been demonstrated, and the operation mode could be shifted by altering the integrated three-surface interferometer (TSI). It is worth noting that analogous optical switching techniques via direct illumination or evanescent field with an ITO SA can also be applied to solid-state or fiber lasers.

2 Results

The ITO film was coated on a 0.5-mm-thick alkaline earth boro-aluminosilicate glass via physical vapor deposition (PVD). To demonstrate the optical modulation of ITO at 1.9 μm, different ITO samples with varying wavelengths of ENZ were systematically prepared and investigated in this study (see Figure S3). The ITO samples were first tested in an external waveguide resonator to check their laser performance [14] (see Figures S4, S5), and the optimized one was then incorporated into a compact integrated waveguide resonator to demonstrate the optical switches. In this preliminary study, the ITO sample features an ENZ wavelength of ~2000 nm (Figure 1A), which is slightly longer than the laser operating wavelength. However, efficient optical switches, including Q-switched, Q-switched mode-locked, and finally mode-locked laser operations, have been well demonstrated in the integrated waveguide resonator using the optimized ITO sample. Nevertheless, the ENZ wavelength of ITO can be further tuned as required by altering the doping concentration or the applied electric field [15], [16].

Figure 1:

Characterization of the ITO film.

(A) The electrical permittivity of the ITO film from 400 to 2100 nm, limited to the measurement apparatus. The ENZ wavelength of the ITO film is ~2000 nm. The blue shadow indicates the nonlinear transmission characterization and laser operation band. (B) Photo image of the 4-in. ITO film coated on a glass substrate and the element maps of the ITO film. KA and LA indicate the electron shells. Scale bar: 2 μm. (C) Linear optical transmittance spectrum of the ITO film. Fresnel reflection and absorption of the glass substrate are baselined with one bare glass. Inset: Optical bandgap of the ITO film extrapolated from the Tauc plot. (D) Optical transmittance of the ITO film as a function of the incident laser intensity at wavelength 1880 nm. Inset: schematic diagram showing the photon-induced electron redistribution near the Fermi surface.

The electronic resistivity and thickness of the optimized ITO film are 1.2×10⁻³ Ω·cm and 120 nm, respectively. Element mapping shows the film is homogeneously deposited and exhibits a uniform surface morphology without any voids or cracks (Figure 1B). High optical transmittance (>87%) of the ITO film is observed from 400 to 2500 nm (Figure 1C), which confirms its lossless wide spectral modulation window from the visible to the mid-infrared regime. The optical bandgap (E_g) is extracted by extrapolating the linear portion of the square of the absorption coefficient (α) versus the photon energy (hν) using the Tauc [17], [18] relation (αhν)²=C(hν–E_g), where C is a constant, with a value of 3.78 eV, which is in good agreement with the previously reported value of 3.5–4.0 eV [11].

The nonlinear optical transmission of the ITO film was characterized using an open aperture Z-scan approach [19], [20] at the wavelength 1880 nm (see SI #1). The detected optical transmission monotonously increases with the incident light intensity, and 7.4% optical amplitude modulation has been achieved under the light illumination intensity of 306 MW/cm² (see Figure 1D). The nonlinear optical transmission originates from the redistribution of hot electrons near the Fermi surface (see the inset of Figure 1D); more detailed mathematical evolutions can be found in Refs [8], [12], [21]. It is worth noting that other nonlinear absorption mechanisms, e.g. excited state absorption (ESA) and multiphoton absorption (MPA), may also happen in the meantime under light illumination, but not free-carrier absorption (FCA) [22]. Since both ESA and MPA result in increased absorption, the decreasing absorption with increasing laser intensity indicates that FCA is the dominating processes in this experiment. Unlike the interband transition and recombination processes in semiconductors, Alam et al. found that fast intraband transitions of the excited carriers in ITO have a short rise and recovery time of 200 and 360 fs, respectively, which suggests a high all-optical modulation speed up to the terahertz range [8].

The nonlinearity of ITO can be altered by an applied electric field or current, apart from incident light illumination, which has been well demonstrated in a compact passive silicon waveguide [23]. However, its use as an intracavity all-optical modulator (known as a saturable absorber) in an active waveguide has not been exploited yet. To illustrate the low bleaching threshold and high modulation speed of ITO, a Tm: ZBLAN waveguide is adopted here to realize a compact monolithic waveguide resonator by using ITO as an intracavity absorber (see Figure 2A).

Figure 2:

Q-switched WG laser performance using the ITO SA.

(A) Layout of the waveguide resonator. The 790-nm pump light (yellow color) is directed by a silver mirror and focused by a planoconvex lens (f=50 mm). The waveguide is tightly coupled to the input coupler (IC, HT@800 nm, HR@1.9 μm), while the ITO sample is glued on the output coupler (OC, R=80% @ 1.9 μm) and fitted on a piezo-actuator to finely control the air gap width (step: ~30 nm). Inset: the cross section of the waveguide. Pol.: polarization. The white dashed line shows the up section of the waveguide cladding where uneven stress is formed. Scale bar: 50 μm. (B) Average output power as a function of the pump power. Laser threshold P_th=123.2 mW, slope efficiency=16.2%, maximum output power P_max=97.2 mW. Inset: a temporal oscilloscope spectrum of the Q-switched pulse at the maximum pump power. (C) Repetition rate and pulse duration of Q-switching versus the pump power. (D) A typical optical spectrum of the Q-switching operation. The zoomed-in spectrum shows the longitudinal modes with a spacing of 78 pm.

The Tm: ZBLAN waveguide was fabricated via the femtosecond laser direct-write (FLDW) approach [24], [25]. Localized structure and index modification could be realized, which was triggered by multiphoton absorption or/and tunneling ionization of transparent dielectric under a tightly focused femtosecond laser, and thus could be utilized for footprint optical waveguide construction. The waveguide diameter and length are 60 μm and 14.2 mm, respectively. The propagation losses of the waveguide amount to ~0.2 dB/cm [26]. The interfaces between ZBLAN glass and air, air and the ITO film glass substrate, and the glass substrate and output mirror form a three-surface interferometer (TSI). The intracavity dispersion, the reflectivity of the TSI, and the incident laser intensity on the ITO film can be finely tailored by altering the width of the air gap. The reflectivity gives the signal feedback with extra dispersion to the gain medium. The laser illumination at the ITO position due to the standing wave effect alters the bleaching level of the absorber. To achieve mode-locked laser operation, two critical conditions must be satisfied: (1) the ITO SA is efficiently bleached to the initial pulsed operation, and (2) the generated intracavity dispersion from the TSI compensates the nonlinear self-phase modulation (SPM) of the mode-locked pulse. The geometric dispersion of the waveguide can be neglected because of the large core diameter and low contrast of the core–cladding index. The dispersion of the Tm: ZBLAN and the glass substrate was determined to be –487.8 fs² (see SI #3). Since the tuning range of the generated group delay dispersion (GDD) from the TSI can vary from –100 to 100 ps², it dominates the intracavity dispersion alternation.

The stable, pure Q-switching operation was successfully demonstrated from the monolithic waveguide resonator when the ITO SA was tightly coupled with the waveguide end facet (air gap width ~0), as shown in Figure 2. The clean single peak of the radio frequency (RF) spectrum with a signal-to-noise ratio (SNR) of 26 dB confirms the stable Q-switching operation. The Q-switching repetition rate monotonically increases as the pump power increases. Conversely, the pulse duration decreases in the meantime (Figure 2C). The maximum output power of the Q-switched laser pulse is as high as 97.2 mW with a slope efficiency of 16.2%. The maximum repetition rate and minimum pulse duration are 241 kHz and 526 ns, respectively, corresponding to a peak power of 0.77 W and a pulse energy of 403 nJ at the maximum pump. Figure 2D shows a typical Q-switched laser optical spectrum whose band center is at 1862.5 nm and has 2.49 nm spacing originating from the etalon effect of the 0.5-mm glass substrate. Two longitudinal modes can be observed at most in a single emission band. The spacing of the longitudinal modes is 78 pm at 1860 nm, which is consistent with the resonator optical path of 21.81 mm. The output laser is highly linearly polarized with a polarization extinction ratio of 97.3%, which is due to the uneven cladding stress originating from the bottom-up waveguide cladding construction processes, as shown in the inset of Figure 2A.

The output power at the Q-switching threshold is 18.8 mW, and the laser beam area on the ITO SA is 2826 μm². Q-switching typically arises from the intensity of noise spikes, but the Q-switching operation can be stable only when the peak intensity induces sufficient optical modulation. The peak intensity incident on the ITO SA at the Q-switching threshold is 33 kW/cm². Considering that the standing wave effect enhancement of the incident laser from the TSI is below four, the laser intensity is quite low for realizing stable intracavity optical modulation. The absorber is unnecessary to be totally bleached to obtain the pulsed laser output. A higher modulation is favorable for pure Q-switching, while it will induce instabilities for mode-locking based on Haus’s perturbation analysis [27]. Nevertheless, such low Q-switching threshold confirms the low bleaching threshold of the ITO SA and indicates its likely use for integrated nanophotonic devices.

When the air gap is assumed to be zero, the standing wave effect of the TSI gives an intensity enhancement of 2.59 normalized to the incident laser intensity at the ITO film position, which facilitates short pulse generation. The calculated GDD value of the laser wavelengths is in the normal regime with a value of 8.61±0.02 ps², which is not favorable for soliton-like mode-locking operation, so only pure Q-switching was obtained.

The ITO SA and the output coupling mirror were bonded together and mounted on a piezo-actuator (see SI #4). By using the piezo-actuator, the thickness of the air gap was finely controlled by changing the voltage applied to the piezo-ceramic. By moving the ITO SA to further positions, Q-switched or Q-switched mode-locked laser operation could be easily obtained. Laser operation could be confirmed with an air gap width of up to 3.2 mm at the maximum pump power. Pulsed laser operation mode is not continuous with the air gap width change, and a laser operation mode scan is shown in Figure 3. The air gap at the scan start position is 0.6±0.1 mm, as set by a coarse translation stage. The pump power is 569.1 mW, and the output laser power is 37.6 mW before the piezo-scan and 36.4 mW after the scan (air gap ~0.6 mm+31 μm). The voltage varies from 10 to 0 V with a step size of 0.01 V (displacement step ~30 nm). The observed laser mode of operation is shown in Figure 3B. The black bars indicate pulsed (Q-switched or Q-switched mode-locked) operation, whereas in the gaps in between, the laser is running in the continuous wave (CW) mode. It is apparent that the mode of operation strongly depends on the air gap width. This is because the illuminated laser intensity on the ITO film changes with the air gap width. Thus, low mechanical vibration is required for long-term operation, while altering the air gap distance can be an effective mechanism for shifting the laser operation mode. The pulsed laser signal bar at a smaller air gap is broader than that at a larger air gap because the laser beam divergence results in lower laser intensity at a more substantial air gap position. The pulsed laser signal peak positions (pi, i=1, 2, …, 29) in Figure 3B are summarized in Figure 3C and D, where the peak position linearly increases with the peak number. There is a large position jump near the DC voltage offset position. The adjacent peak position distance (Δi, i=1, 2, …, 28) is presented in Figure 3D. The average shift period of the laser operation mode is 913.8 nm, which suggests a half-wavelength optical path mode alteration period in the air medium and a good match with the numerical analysis of the TSI system (see SI #5).

Figure 3:

Laser operation mode scan.

(A) Open-loop displacement calibration of the piezo-actuator when changing the operation voltage from 10 to 0 V. The large displacement between 5 and 4.5 V is due to the applied voltage across the DC offset of the wave generator. (B) Laser operation mode change as the voltage is applied to the piezo-actuator. The pulsed mode (Q-switching or Q-switched mode-locking) is recorded as signal “1”, while CW laser is recorded as signal “0”. Thus vertical bars indicate regions where the laser is Q-switching or Q-switched mode-locking. In all regions in between, the laser is running CW. All experiments were performed at 569.1 mW of pump power. The significant gap at the middle position is due to the applied voltage across the DC offset. (C) The location of the pulsed signal peak center as shown in Figure 4B. (D) Air gap adjustment period for obtaining pulsed laser operation is as shown in Figure 4B.

Stable mode-locked laser operation was successfully achieved when the TSI air gap was 1.5±0.2 mm, as shown in Figure 4. The maximum output power of the mode-locked laser is 28.6 mW with a slope efficiency of 6.8%. The output laser beam profile is the fundamental Gaussian mode (Figure 4B), which is due to the higher propagation losses of the higher order modes [28]. Because of the short resonator optical path of 23.31 mm, the mode-locking repetition rate is as high as 6.42 GHz, which confirms the high modulation speed of ITO. The single clean peak without any parasitic peaks of the RF spectrum and an SNR of 27.6 dB confirm stable mode-locked operation. The optical frequency agrees with the reflection of the TSI, where eight longitudinal resonator modes can be observed with the spacing of 76 pm. Passive mode-locked waveguide lasers have been demonstrated by using carbon nanotubes (CNTs) [29], graphene [30], and SESAMs [31] as SAs at wavelengths of 1.06–1.55 μm. Very recently, Q-switched mode-locking at the wavelength of 1.9 μm was reported by using graphene as the SA from a Tm:YAG waveguide [32]. However, because of the high losses (~30%) and high saturation fluence (59 μJ/cm²) of the graphene SA, the maximum output power was 6.5 mW with a slope efficiency below 1%.

Figure 4:

Mode-locked WG laser performance using the ITO absorber.

(A) Temporal oscilloscope spectra of Q-switched mode-locked, unstable mode-locked, and stable mode-locked laser operations. The left-hand side shows the spectra on a 1000 ns time scale, and the right-hand side shows the zoomed-in spectra on 10-ns timescale. (B) Average output power as a function of the pump power. Inset: the output beam profile. Scale bar: 1 mm. (C) Radio frequency spectrum of the mode-locking. The inset shows the zoomed-in RF spectrum. (D) Calculated reflection of the TSI agrees well with the output laser wavelength of the mode-locking operation. The air gap width is assumed to be at the adjusting period of 1500471.55 nm. (E) Fast Fourier transform of the mode-locked optical spectrum. The resulting pulse duration is 18.9 ps. (F) Temporal oscilloscope spectra of the mode-locking operation over a time scale of 40 min.

Considering that the maximum output power of the mode-locking operation is 28.6 mW, and the high repetition of 6.42 GHz, the output pulse energy is around 4.5 pJ, with which it is infeasible to obtain a second-harmonic generation signal from a normal autocorrelator. Thus, the pulse duration was estimated via a fast Fourier transform of the laser spectrum to be 18.9 ps, under the assumption of CW operation (see Figure 4E). This value is close to the recently reported result for an Nd:YAG waveguide using graphene as SA [30]. The output power at the mode-locking threshold is 16.3 mW, and the beam area on the ITO film position is 5681 μm² in the diffraction-limit assumption. Thus, the peak laser intensity falling on the ITO film is calculated to be 25.3 kW/cm². This is similar to the Q-switching threshold discussed above, and strongly confirms the low bleaching threshold of ITO again. The mode-locking operation was stable during all the measurements over the course of more than 30 min. Another long-term stability check of the mode-locking operation was performed over a time scale of 40 min (see Figure 4F).

To study the influence of the TSI on the mode-locking operation, a vector analysis in the absence of losses was applied (see SI #5) [33]. Figure 5A shows the reflection of the TSI as a function of the air gap width and optical wavelength. The reflection of the TSI is mainly determined by the 0.5-mm glass substrate’s etalon effect, while the air gap modifies the maxima of the reflection peaks. For a given air gap, the reflection spectrum of the TSI agrees with the laser wavelength, where a series of air gap widths of ~1.5 mm (1500471.55 nm+k×948.5 nm, k=0, ±1, ±2, …) could be obtained. This indicates that the adjusting period of the air gap is 948.5 nm, which corresponds to a half-wavelength optical path in the air medium. The generated GDD from the TSI is calculated and shown in Figure 5B. The GDD value from the TSI for the laser wavelengths is determined to be –0.04±0.02 ps², which is in good agreement with previous reports from a Gires–Tournois interferometer (GTI) in realizing soliton-like mode-locked continuous wave (CWML) in monolithic waveguide resonators [30], [31]. Note that the calculation error of the GDD in a TSI system can be higher than in the regular GTI system because of one more layer interference, while the qualitative nature of the generated GDD is not affected. The standing wave effects of the TSI offers a normalized intensity enhancement of 2.59 at the ITO SA position, which indicates the high bleaching level of the ITO SA for short pulse initiation.

Figure 5:

Numerical analysis of the TSI.

(A) Reflectivity map of the TSI versus the air gap width and laser wavelength. The vertical blue and red line shows two emission band centers of the mode-locking operation. The horizontal black lines indicate the air gap adjusting period (1500471.55 nm+k×948.5 nm, k=0, ±1, ±2,…), where a maximum reflectivity is obtained at the lasing wavelength. (B) GDD map of the TSI as a function of the air gap width and the laser wavelength.

3 Conclusion

The present study shows the excellent nonlinear optical modulations of a nanometer ITO film, including low bleaching threshold, moderate modulation depth, high modulation speed, and low linear losses, all of which are appealing for integrated on-chip photonics and electro-optics applications. By adding the mature MOS fabrication processes that are compatible with existing silicon photonic platform, there is a tremendous potential application for all-optical light manipulation in both passive and active photonic circuits in the near future. Implementing the ITO film in an active waveguide chip illustrates a practical use for the optical switches of monolithic waveguide chip lasers. The gigahertz, mode-locked, mid-infrared, on-chip source can be beneficial in ecological monitoring [34], medicine [35], and chemical reactions [36].