1. Introduction

Two-dimensional (2D) materials, such as graphene and monolayers of transition metal dichalcogenides (TMDCs), have unique properties that enable active functionalities on otherwise passive photonic platforms. TMDCs are indirect bandgap semiconductors in their bulk form that transition to a direct bandgap material when they become monolayer [1,2]. On-chip high-speed modulation [3–6] and detection [7–11] based on integrating 2D materials with photonic platforms have been achieved. However, on-chip lasing remains challenging and has only been demonstrated by emitting into free space. Demonstrations of lasing with TMDC’s monolayers emitting into free space have been shown with different photonic cavities based on photonic crystal cavities, distributed Bragg cavities, standing disks, nanobeams, or spheres [11–20]. These cavities can have small modal volumes and high-quality factors that facilitate lasing, even at room temperature, but their emission is hard to integrate on a photonic chip. Previous efforts to integrate the light emission of 2D materials on-chip were limited by a trade-off between quality factor and cavity size, which determines the number of modes of the light source. These previous efforts did not show lasing features [21,22]. Here, we show the spectroscopic characterization of the emission of a 2D material gain medium coupled to an on-chip cavity and report the observation of laser-like emission properties.

2. Device structure

To achieve the coupled emission, we integrate a monolayer of tungsten diselenide, WSe₂, with a high-quality factor coupled cavity in a silicon nitride platform. The coupled cavity consists of a silicon nitride ring resonator coupled to a Mach-Zehnder interferometer (MZI) (Fig. 1(a)). The silicon nitride waveguide is transparent to the emission of WSe₂ since the transparency window of silicon nitride reaches from 400 nm to 2350 nm [23]. Furthermore, silicon nitride has been demonstrated to be an excellent platform for creating low-loss waveguides and high-quality on-chip cavities [24,25]. The monolayer is transferred on top and in contact with the ring resonator and interacts strongly with the waveguide’s optical mode (Fig. 1(b)). The cross-section of the waveguide is 500 nm by 300 nm. The ring’s radius is R = 13 µm. A gap of 350 nm separates the ring and bus waveguide, which results in critical coupling around the WSe₂ excitonic emission wavelength of 750 nm.

 

Fig. 1. a) Schematic of our device. WSe₂ monolayer is transferred on top of a ring resonator coupled to an MZI. b) Interaction of the ring’s mode with the monolayer c) Microscope image of a single ring device with a WSe₂ monolayer. Red, blue, and yellow lines enclose the WSe₂ monolayer, bilayer, and multilayer, respectively. d) PL mapping of the monolayer on top of the ring resonator shown in 1c. The regions of Si₃N₄ covered with the monolayer are dark in the mapping showing a strong interaction of the PL with the Si₃N₄. The ring shown in c and d is not coupled to an MZI and only exemplifies the architecture of our devices. e) Output spectrum of our device at room temperature showing the peaks corresponding to the resonances of the ring coupled to an MZI.

Download Full Size | PDF

The fraction of the waveguide’s mode energy contained within the monolayer, also called the confinement factor Γ, is 0.004. Integrating the monolayer on top of the device enables efficient light emission coupled to an on-chip waveguide. The WSe₂ monolayer covers approximately 95% of the ring, maximizing the monolayer’s interaction area with the cavity (Fig. 1(c)). We use photoluminescence (PL) mapping to further verify the successful transfer and interaction of the monolayer with the cavity (Fig. 1(d)). We choose WSe₂ as the emission material because of its high quantum yield compared with other TMDCs [11,12,26]. The monolayer is mechanically exfoliated [27–29] from a bulk crystal (2D Semiconductors), grown using the flux zone technique [30], and dry transferred [31] on top of the ring resonator.

We measure sharp peaks in the emission spectrum of our device when we optically pump the monolayer. These sharp peaks correspond to the resonant modes of the cavity (Fig. 1(e)). When we pump the monolayer, its PL couples to the ring. However, only the wavelengths resonant with the cavity are enhanced. A portion of the light circulating the ring couples to the bus waveguide and travels to the edge of the chip, where we collect it with an optical fiber. With our coupled emission scheme, the fiber collects light only from the PL coupled into the cavity modes, and we do not collect background light originating from PL coupling into free space modes. The fiber collection acts as a natural filter to the background PL arising from coupling into free space modes. When we use high enough pump power, the monolayer’s gain can overcome the cavity losses. The light enhanced by the cavity reaches threshold and becomes laser light.

3. Coupled-cavity scheme

We use a coupled cavity scheme to reduce the number of modes overlapping with the monolayer’s gain spectrum. The coupled cavity increases the spectral separation between the modes, Free Spectral Range (FSR), decreasing the number of resonant modes that overlap with the gain spectrum of the WSe₂ monolayer. If fewer resonant modes of the cavity fit within the emission spectrum of the monolayer, the pump power necessary to reach threshold decreases for each mode [21]. The PL emission of WSe₂ has a spectrum with a full-width half maximum (FWHM) of approximately 16 ± 0.13 nm (supplemental document), depending on the quality of the monolayer. The number of modes that can fit in this range depends on the FSR of the ring, which is inversely proportional to its radius. The FSR is given by $FSR = {\lambda ^2}/({{n_g}2\pi R} )$, where ${n_g}$ is the group index of the waveguide and R is the ring’s radius. The radius of the ring cannot be indefinitely decreased because a smaller radius increases the radiation losses of the ring thus raising the lasing threshold. To decrease the number of modes without decreasing the radius of the ring, we couple an MZI to the ring resonator (Fig. 2(a)). The MZI acts as a filter suppressing resonances of the ring [32–35]. Certain ring resonances can be suppressed depending on the length of the MZI arms (supplemental document). Our device is designed to suppress every other ring resonance, increasing its FSR by a factor of 2 (Fig. 2(b)).

 

Fig. 2. a) Ring resonator coupled to MZI. b) Simulated output spectrum of a single ring resonator (blue) and MZI coupled ring resonator (red). c) Ring resonance at 751.41 nm with Q = 350,000. Lorenztian fit indicated by blue line. d) Ring resonance after monolayer transfer with Q = 95,000. Lorentzian fit indicated by blue line. e) Optical pump focused on top of the ring resonator. e) Output spectrum of our device at room temperature for increasing pumping powers.

Download Full Size | PDF

Together with the large FSR created by the coupled cavity scheme, the device has a high intrinsic quality factor, Q, which further decreases the power necessary to reach threshold. We measure an intrinsic quality factor Q of 350,000 (Fig. 2(c)) and 95,000 (Fig. 2(d)) before and after the monolayer transfer, respectively. The quality factor is extracted by fitting a Lorentzian function to the data [36]. The change in quality factor after the transfer confirms the interaction of the ring with the monolayer, which decreases the quality factor because the monolayer absorbs the light in the ring and represents a new loss. The resonance in Fig. 2(c) shows that our device is close to critical coupling around 750 nm, which is the excitonic emission wavelength of WSe₂ at room temperature. The coupling regime of the ring changes when a monolayer is transferred. The monolayer introduces a new loss to the ring, and it becomes undercoupled, as the loss of the ring is greater than the power coupled to the ring (Fig. 2(d)). As the monolayer is pumped its absorption decreases and the loss decreases thereby increasing the quality factor. The coupling between the ring and the bus waveguide does not change with the pump but the intrinsic loss of the ring does. This changes the coupling regime of the ring resonator.

4. Spectroscopic characterization

We characterize the waveguide-coupled emission of our device at different pumping powers. As we increase the pump power, we observe an increase in the output intensity at the resonant wavelengths. (Figure 2(f)) We optically pump the device from the top to excite the monolayer. The pump is a CW Ti:Sapphire laser at 701 nm focused on the monolayer to a 15µm radius spot size. The focused spot covers the entire ring, exciting all the gain material interacting with the cavity (Fig. 2(e)). The 701 nm pump is close to the bandgap transition of WSe₂, but it is still above its bandgap. An excitation wavelength close to the bandgap transition is chosen to excite the excitons slightly above the valley of the conduction band. In this way, most of the excitons will decay to the valley corresponding to the direct transition and not to other undesired indirect transitions.

The collected spectra (Fig. 2(f)) show the peaks that correspond to the resonant modes of the ring that also overlap with the PL spectrum of WSe₂ centered at ∼750 nm. The resonances come in pairs. Each pair occurs approximately every 6.6 nm, and each pair’s resonances are separated by 2.1 nm. The first peak of each pair belongs to the fundamental TE mode. The TE modes are separated by 6.63 nm. This peak spacing corresponds to the doubled FSR of the ring resonator coupled to the designed MZI. The second peak of each pair corresponds to the TM modes. At a wavelength of 750 nm, the effective index of the fundamental TE and TM mode is n_eff= 1.7570 and n_eff= 1.6884, respectively. The group index of the fundamental TE and TM mode is n_g= 2.2031 and 2.2487, respectively. All the measurements are done at a room temperature of 22°C.

5. Laser-like properties

We report the observation of laser-like emission properties when we characterize the device’s power emission as a function of pumping power. We observe the two important properties of lasing: a ‘kink’ in the input vs. output power plot and linewidth narrowing. We fit a Lorentzian function to the emission peaks (Fig. 3(a)) to obtain their output intensity and linewidth. The linewidth is also called FWHM. First, we observe a ‘kink’ in the device’s input vs. output intensity curve for the 751.05 nm emission line. The plot of the input vs. output power is referred to as the “light input-light output” or “L-L” curve. At the threshold value of the excitation power, there is a rapid increase in the output intensity, indicating the beginning of stimulated emission or lasing. This rapid increase creates a ‘kink’ in the graph and is characteristic of the lasing behavior. Our device’s ‘L-L’ curve is shown in Fig. 3(b) and exhibits the rapid increase of output emission or kink characteristic of the laser behavior.

 

Fig. 3. a) Lorentzian function fitting for a peak at 751.05 nm at different input powers. Black arrows indicate the FWHM. b) L-L plot showing a ‘kink’ between 1 mW and 6 mW of input power. The blue line is a guide to the eye. c) Linewidth vs. Input power plot showing linewidth narrowing between 1 mW and 6 mW of input power. The blue line is a guide to the eye. d) Log-log plot of the output intensity of the device as a function of pump power. Solid lines are simulated results for different β factors. The red lines show the best fit to the measured results giving a β factor of 0.1.

Download Full Size | PDF

We observe a 30% linewidth narrowing as further evidence of lasing properties. This decrease in linewidth is another important laser property that appears when the device reaches threshold. The linewidth narrowing indicates the increase in the emission’s coherence. The FWHM as a function of the input power of the peak centered at ∼750 nm is shown in Fig. 3(c). For the same range of pump power as in Fig. 3(b), the input vs. linewidth plot in Fig. 3(c) shows progressive narrowing from 0.48 nm to 0.34 nm, followed by a slight increase and then a decrease to a plateau at 0.34 nm. This narrowing corresponds to a linewidth narrowing of 30%. Figure 3(a) also shows the linewidth decrease. The slight linewidth increase followed by a subsequent linewidth narrowing has been observed before in microlasers based in 2D materials. [11,12,16] This behavior remains an open question.

We measure a ratio of spontaneous emission coupled to the cavity mode over the total spontaneous emission of 0.5. This ratio is called β factor and is another figure of merit of a laser. The theoretical limit of β is one [37]. The β factor is obtained by fitting the log-log plot of the L-L curve to the theoretical laser rate equations (see supplemental document). Figure 3(d) shows the log-log plot of the L-L curve for different values of the β factor. The plot in red is the best fit for the experimental data giving a β factor of 0.5, in line with previous reports of 2D microcavity lasers. We obtained a threshold value of 2 mW or 283W/cm² with the rate equations fitting. This value agrees with previous threshold values reported for 2D nanolasers. [12] The threshold value can be improved by increasing the confinement factor of the device. An increase in confinement factor could be done by creating an architecture where the monolayer is located at the center of the waveguide instead of the top.

6. Conclusion

In conclusion, our device demonstrates the possibility of integrating the PL of monolayer TMDCs to on-chip cavities and producing integrated laser emission at room temperature, enabling a lasing platform that is scalable to arrays of on-chip lasers. We observe room temperature, on-chip laser light based on TMDC monolayers integrated into a photonic cavity. We demonstrate waveguide coupled, optically pumped lasing emitted at a central wavelength of 751.05 nm by a monolayer of WSe₂ integrated with an on-chip microring cavity. The laser emission was verified by the kink present in the L-L curve of the devices and the linewidth narrowing of the peaks. This demonstration, the variety of wavelength emission of TMDCs, and their tunability open the door for a broad range of on-chip lasers. Since integrated modulation and detection can also be achieved with 2D materials, this demonstration paves the way to a fully integrated and complete optical circuit using the properties of 2D materials. It also paves the way for the electrical pump of TMDCs lasers.