Room-temperature continuous-wave lasing from monolayer molybdenum ditelluride integrated with a silicon nanobeam cavity

In the format provided by the authors and unedited. DOI: 10.1038/NNANO.2017.128 Room-temperature continuous-wave lasing from monolayer molybdenum ditelluride integrated with a silicon nanobeam cavity Yongzhuo Li 1, Jianxing, Zhang 1, Dandan Huang 1, Hao Sun 1, Fan Fan 2, Jiabin Feng 1, Zhen Wang 1, and C. Z. Ning 1,2,* correspondence to: cning@asu.edu This PDF file includes: Supplementary Text Supplementary Figures S1 to S9 Supplementary Tables S1 to S3 I. Fabrication of Monolayer Semiconductor Nanolaser II. Measurement System and Pumping Configuration III. Optical Properties of Monolayer MoTe2 IV. Simulation of the Proposed Device V. The Laser Rate Equation Fitting VI. Effects of Si-absorption, Heating, and Linewidth Broadening VII. Further Results of Other Devices NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 1

I. Fabrication of Monolayer Semiconductor Nanolaser For a detailed explanation of the fabrication process, see Methods in the main text. Figure S1 shows the optical image of an exfoliated monolayer MoTe2 (a) and the main fabrication process of silicon nanobeam cavity (b). Figure S1. a, Large-area monolayer MoTe2 by using an improved mechanical exfoliation technique. b, Schematic of fabrication process for silicon nanobeam cavity. II. Measurement System and Pumping Configuration The optical properties of monolayer MoTe2 and our devices are measured in a micro-pl system, pumped by a continuous-wave He-Ne laser at 633 nm. The pump light is vertically incident onto the sample plane by a 100 objective, which also collects the emission light into a monochromator equipped with a LN cooled InGaAs detector. We note that the area of active monolayer MoTe2 and the size of the pumping beam are both much larger than that of the 2 nd cavity mode in our device, especially in the direction normal to the beam axis. Since the spontaneous emission from the region of monolayer MoTe2 far away from the nanobeam cavity is not coupled to the cavity modes, they mainly contribute to the background PL spectrum, leading to a continuous increase of PL with pumping. This is different from a typical semiconductor laser where one expects a depressed spontaneous emission background above the threshold due to carrier density clamping. In order to reduce area of MoTe2 monolayer being pumped and reduce the background emission generated by the monolayer MoTe2 far away from the nanobeam cavity, the monolayer is placed on the top of the nanobeam cavity to leave most of the excess area on one side of the cavity (lower side of the cavity in Fig. S2c). The pump beam is then incident normally on the other side (above the device in Fig. S2c) at a distance of 3 μm. Measurement of the Pump Laser Beam Size: To determine the pumping power density, we need to know more precisely the pump beam diameter. The diameter of the pump laser was measured using the scanning knife-edge method 1, as shown in Fig. S2a. We measured the Raman intensities of a sliding Si wafer at different slide positions, and a diameter of 2.44 μm was obtained by fitting the results of integrated Raman intensity, as demonstrated in Fig. S2b. NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 2

Figure S2. a, Schematic of the scanning knife-edge measurement. b, The normalized integral Raman intensity of Si at different positions and the fitting results. The inset is an example of the Si Raman spectrum. c, Modular electric field distribution of the pump laser (633 nm) in the X-Y plane. The pump beam is incident normally to the device plane at a distance of 3 μm which happens to be at the center of another nanobeam. Note that the nanobeam marked device is the nanolaser being studied with 2D gain material. Other two nanobeams above do not have 2D gain material and are not studied as devices. Their presence affects the pump beam profile at the location of nanolaser being studied. d, Modular electric field distribution of pump laser in the Y-Z plane, where the device is marked. Estimate of Pump Power Density: Due to the existence of other nanobeams and the detailed structure and sizes of these beams, there are more complicated modes and beam structures on the planes of devices. This is also complicated by the fact that we did not focus the center of the beam directly at the device we measured (rather than about 3 NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 3

microns away, which happens to be another nanobeam) to minimize the pumping of uncoupled 2D material, as shown in Fig. S2c. The power density of pump laser on the device was calculated through the FDTD method. As can be seen from Fig. S2c, the electric field distribution of pump laser in the X-Y plane on the device is no longer a Gaussian distribution any more, rather a complicated pattern structure. Figure S2d shows the electric field distribution in the Y-Z plane. Using the simulated mode structure, we can more accurately determine the pump beam power density, the ratio of pump power on the device (the size of nanobeam cavity, 7200 nm 365 nm) over the whole pump power is calculated at 0.2%. Thus, at the total threshold power of 97 μw, the pump power density on the device is 6.6 W/cm 2. III. Optical Properties of Monolayer MoTe2 MoTe2 undergoes a transformation from an indirect to direct bandgap semiconductor when the material is reduced from multilayers to a monolayer. Thus, the emission efficiency is much higher for a monolayer MoTe2. The preparation of monolayer MoTe2 has been described above in SI Section I. After the exfoliation, it is crucial to identify monolayers from multilayers of MoTe2 to ensure that the former is coupled to the optical cavity and used as an efficient gain medium for our laser demonstration. We use atomic force microscope (AFM) and Raman measurements for such identification. The height profiles of monolayer and bilayer MoTe2 were measured with the AFM and are overlaid on the corresponding microscope image in Fig. S3a, where the thickness of monolayer and bilayer MoTe2 are 0.7 nm and 1.35 nm, respectively. To further identify the monolayer MoTe2, we measured the Raman spectroscopy of MoTe2 samples by utilizing a commercial micro-raman setup (HORIBA Evolution) under a 532 nm pump laser. The typical phonon modes of MoTe2, A1g, E 1 2g, and B2g are observed in the Raman spectra, as shown in Fig. S3b. These modes correspond to out-of-plane, in-plane, and out-of-plane (Bulk-Raman inactive) modes, respectively 2. It s worth mentioning that the B2g mode becomes Raman active mode in few-layer MoTe2 due to the translation symmetry breaking. The absence of B2g mode is used as a unique identification of a monolayer 2 and thus Raman spectroscopy serves as an easy means to distinguish regions of monolayers from those of multi-layers. PL spectra were measured under the excitation of a He-Ne laser (633 nm) at different temperatures. The emitted light is collected by a 100 objective and coupled to a monochromator equipped with a liquid nitrogen cooled InGaAs detector. From the PL spectra, shown in Fig. S3c, we can see that the peak wavelength red-shifts from 1111 to 1151 nm as temperature increases from 77 to 300 K. The full-width at half-maximum (FWHM) of PL spectrum increases from 27 to 52 nm in the same range of temperature, as shown in Fig. S3d. In addition, we compared monolayer MoTe2 emission peak and Si bandgap 3 under different temperatures (see Fig. S3e). The peak energy of monolayer MoTe2 emission spectrum is lower than the Si bandgap by ~ 50 mev, assuring minimal absorption of MoTe2 emission by above-bandgap transitions in Si. This justifies our choice of Si-nanobeam cavity and renders MoTe2 as an important gain medium for Sibased photonic system. NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 4

Stability of MoTe2 has been a concern in the literature, especially with respect to PL quality. We have measured PL spectroscopy on different days after the monolayers were exfoliated from the bulk. Figure S4 shows the results where both PL intensities and the corresponding linewidths are shown. As is seen, within the time frame that we measured, not clear degradation tendency is observable, indicating decent stability of material qualities, even without any protection. Figure S3. a, Optical microscope image of monolayer, bilayer, and bulk MoTe2 and their corresponding AFM height profiles. The scale bar is 10 μm. b, Raman spectra of monolayer and bilayer MoTe2. c, PL spectra of monolayer MoTe2 under different NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 5

temperatures. d, Linewidths and peak wavelengths of PL spectra of monolayer MoTe2 under different temperatures. e, Silicon bandgaps and monolayer MoTe2 emission peaks under different temperatures. Figure S4. PL intensity (as measured by detector counts) (gray band, left axis) and the corresponding linewidth of the PL (red band, right axis) measured on different days after their exfoliations. The different data points on the same day refer to measurements on different samples exfoliated on the same day. IV. Simulation of the Proposed Device 1. Design of silicon nanobeam cavity The silicon photonic crystal nanobeam cavity provides one of the most effective means to confine optical field and to provide a high Q cavity for wavelengths that are transparent to Si such as the emission of MoTe2. For this purpose, it is necessary to carefully design the structural parameters of nanobeam cavity to assure the cavity modes to be within the gain spectrum of the monolayer MoTe2. Figure S5a shows the schematic diagram of nanobeam cavity, which consists of a waveguide and a series of air holes, where ri (i=1, 2, 3, 4, and n) is the radius of air hole i. aj (j=1, 2, 3, 4, and n) is the distance between adjacent air holes (center to center). W and L are the width and length of nanobeam cavity. The structural parameters of devices are shown in Table S3. The two-side segments with periodic air holes serve as mirror regions. Four air holes with radii gradually reduced from outside towards the center (tapered section) are NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 6

symmetrically positioned with respect to the center of nanobeam cavity. The nanobeam cavity is designed and optimized to have the maximum Q values with the modes within the gain spectrum. Mode profiles for the first, second and third mode are shown in Fig. 2. Figure S5b and S5c show the electric field (Ey) of top view and side view for the second mode, whose modal volume is calculated as 1.6 10 7 nm 3 or (252 nm) 3. The dash line in Fig. S5c indicates the monolayer MoTe2, which is closely placed on the top side of the silicon nanobeam cavity. Transmission spectrum of nanobeam cavity, calculated by FDTD method, is shown in Fig. S5d. The detailed structure parameters are shown in Table S3, marked by DM1. Calculated resonant wavelengths of the first, second, and third mode are 1054, 1132 and 1167 nm, while the corresponding calculated Q factors are 5.2 10 6, 6.5 10 5, and 1.4 10 3, respectively. Figure S5. a, Schematic diagram of photonic crystal nanobeam cavity, device DM1 (see Table S3 for a list of parameters). b, Electric field (Ey) distribution of the second mode (1132 nm) from the top view. c, Electric field (Ey) distribution of the second mode (1132 nm) from the side view. The dash line indicates monolayer MoTe2. d, Calculated transmission spectrum of nanobeam cavity. 2. Optical confinement factor of the device The optical gain confinement factor of our device is numerically studied, which is defined as: NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 7

2 MoTe E dv 2 MoTe2 = (S1) 2 E dv where ε is the dielectric constant, and E is the electric field. Here, we study two cases: with or without PMMA layer on the top of the MoTe2 layer. Table S1 presents the calculation results. We can see that the optical confinement factor of the second mode is the highest. For the devices covered with PMMA layers, the confinement factors can be improved about 10%. Table S1. Optical confinement factor of different modes. Mode Γ (Without PMMA cover) Γ (With PMMA cover) 1 st mode 0.0117% 0.0127% 2 nd mode 0.0132% 0.0145% 3 rd mode 0.0128% 0.0141% 3. Purcell factor The Purcell effect refers the modification of spontaneous emission in a cavity, and the Purcell factor is described as 3 3Q Fmax 2 (S2) 4 Veff n where Q is the quality factor of the resonant mode, Veff is the effective mode volume, λ is the resonant wavelength, and n=3.53 is the refractive index of Si at resonant wavelength. In our device, Q is determined from the measured linewidth of resonant mode below lasing threshold, λ is measured at 1132 nm, while Veff is calculated to be 2 2 3 Veff E dv max E 0.48 n for the second cavity mode, which is much smaller than other kinds of dielectric microcavities 4. Fmax from Eq. S2 is estimated to be 448. While, in our case, the active monolayer MoTe2 is not at the positon of maximum of electric field (Ey) intensity in the Z direction, as shown in Fig. S5c. Meanwhile, the direction of emitter dipole is random in the X-Y plane. Thus, the Purcell factor is modified as: 2 EMoTe2 2 F Fmax cos (S3) E where 2 MoTe2 E E max =0.34, max 2π 2 1 2 1 cos = cos 2π d, θ is the angle between polarization 0 2 of electric field (Ey) and direction of emitter dipole. Therefore, the Purcell factor is calculated as F=76. The large Purcell factor leads to a large spontaneous emission coupling into our lasing mode in our device. V. The Laser Rate Equation Fitting NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 8

For ultrasmall lasers such as ours, the spontaneous emission coupling factor β is very important. To better understand the behavior of our laser, especially its threshold, we performed analysis of the rate equations. The carrier density N and the photon density S satisfy 5 : dn P 1 0 N F0N vg gs (S4) dt V a sp sp ds vggs+ F N 0 S dt sp p (S5) The material gain g is a function of carrier density N, and can be approximated as: g N a N N (S6) where a=2.6 10-15 cm 2 is the absorption cross-section or linear gain coefficient of the gain material and Ntr is the transparency carrier density. Under continuous pump and dn ds steady condition, we can set 0 and 0. We obtain the stationary version of the dt dt rate equations for the CW solutions: P 1 0 N F0N vg gs 0 (S7) V a sp sp N S vggs+ F 0 0 (S8) sp p Eliminating S, we can get an equation of carrier density N: 2 Va10 vgpan Va10 vgpantrpspvgpa Va10F0 N (S9) P 1v an 0 sp g p tr From Eq. S9 and Eq. S7-S8, we can determine S and N in terms of other device and material parameters. In order to extract β0 from experiment data, we first carefully adjust a until the threshold pump intensity matches experimental measurement. Afterwards light-in vs light-out curves with different β are plotted in log scale (see Fig. 4b in the main text). We find that β0=0.0015 fits best to the experimental data (see Table S2 for other parameters in the laser rate equation). The spontaneous emission factor is defined as β=fβ0/[1+(f-1)β0], which represents the ratio of spontaneous emission coupled to the lasing mode to the total emission, as shown in Fig. S6a. Thus, the spontaneous emission factor β=0.1 is obtained. For a small laser with a large spontaneous emission factor β, a better way to define the threshold is to calculate the second-order derivatives of the lightin vs light-out curve 6. The maximum of the first-order derivative curve shows the threshold clearly, as shown in Fig. S6b and S6c. Finally, the threshold of the laser is estimated to be 0.097 mw (see the red dash line in Fig. S6), with a threshold modal gain of 51 cm -1. Table S2. Definitions and values of parameters used in the rate equations. Parameter Definition Value/Unit N Carrier density cm -3 tr NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 9

S Photon Density cm -3 P Pumping power μw η Absorption efficiency Va Active volume 2440 nm 0.7 nm 365 nm =6.2 10-4 μm 3 F Purcell effect factor 76 τsp Spontaneous emission lifetime 4 ps vg Group velocity 1 10 8 m/s g Material gain cm -1 Γ Confinement factor 0.0145% β0 Spontaneous emission factor without the Purcell enhancement 0.0015 τp Photon lifetime 1.7 ps Figure S6. a, Log-log plot of light-in vs light-out where solid squares are the experimental data. Solid line is the result of a rate-equation calculation with spontaneous emission factor β=0.1. b-c, The first and second order derivative curve (the red dash line indicates the threshold of our device). NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 10

VI. Effects of Si-absorption, Heating, and Linewidth Broadening The FWHM of the cavity mode is extracted through the Lorentzian fitting and shown in Fig. 4d in the main text. The linewidth shows a significant reduction from 0.4 to 0.202 nm when pump power increases from 0.087 to 0.173 mw. We notice a significant linewidth re-broadening with the further increase of pumping above the threshold, which can be explained as follows: in our device, at the higher pumping level, absorption of the pumping laser leads to a higher carrier density in Si-nanobeam. Such carrier-density increase leads to a change of the complex dielectric function of Si 7. The corresponding decrease in the refractive index of Si leads to the blue shift of resonant cavity modes, which can be calculated as a function of refractive index of Si by employing the simulation model in SI Section IV.1. The change of imaginary part of refractive index in Si leads to additional absorption of the cavity mode. In addition, there is pump induced heating of Si which leads to a variation of Si-refractive index, thus a red-shifting of the cavity mode. The heat equation was solved numerically with the heat source as the pumping laser by assuming 40% of the pump power becomes heat. To solve the stationary temperature distribution of the whole nanobeam structure with the two ends connected to the Si-wafer and heat dissipation from the SiO2 insulator layer and Si substrate, the initial value was set as the room temperature, and the outflow boundary was used. Thus, the temperature increase arisen from the heating of pump laser can be obtained, as shown in Fig. S7. The blue-shift of cavity mode caused by the carrier-density increase and red-shift caused by the heating effect both contribute to the change of resonant wavelength from 1132.32 to 1132.17 nm when pump power increases from 0.087 to 0.434 mw. Such a wavelength shift corresponds to a total refractive index change ( n) of -5 10-4 from our simulation of cavity modes of the nanobeam cavity. The temperature raises due to pump induced heating is shown in Fig. S7, where we can see a temperature increase of 7 o C from room temperature under the pump power of 0.434 mw. Based on the Si thermooptic coefficient of 1.86 10-4 K -1 8, the refractive index change ( nheat) of 1.3 10-3 is obtained. Thus, the refractive index change ( ncarrier) caused by carrier-density increase is ncarrier= n- nheat=-1.8 10-3. On the other hand, according to the electrooptical effects in Si 7, the change of refractive index and absorption at 1132 nm can be expressed as: -22-18 nn 0.8 e nh 4.710 Ne 4.510 N h (S10) -18-18 e h 4.510 Ne 3.210 Nh. (S11) Where ne and nh are refractive index changes due to electrons and holes, respectively, and α is the increased absorption of silicon caused by carrier-density increase with αe and αh corresponding to components due to electrons and holes respectively. Ne and Nh are the incremental densities of electrons and holes, respectively. Therefore, the estimated absorption increase for resonant wavelength at 1132 nm is around 8 cm -1 under the pump power of 0.434 mw. Such increase of absorption is equivalent of cavity Q decrease, leading directly to the linewidth broadening. The linewidth of resonant mode can be calculated as follow: 1 Q g (S12) NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 11

where λ, λ and ω are the FWHM, resonant wavelength and angular frequency of cavity mode, respectively. Thus, the linewidth is broadened to 0.251 nm through calculation, very close to the measured linewidth of 0.28 nm under the pump power of 0.434 mw. Figure S7. Temperature variation of the device versus pump power. (The inset is the schematic of positions of pump laser and device during the measurements). VII. Further Results of Other Devices Figure S8 and S9 show results of two more devices (DS1 and DS2) that show signature of lasing. The device designs are similar to that in Fig. 3 and Fig.4, but with different structural parameters as shown in Table S3. Figure S8 shows the spectral evolution measured on device DS1 at increasing levels of pumping power in a high-resolution measurement (grating: 600 g/mm, resolution: 0.09 nm) (a) and a selected spectrum at the pumping level of 54 W (b), showing the details of the spectral linewidth together with a Lorentzian fitting. The lasing wavelength of 1195.8 nm agrees well with the simulation result of DS1, which gives the wavelength of the second mode at 1195 nm. The lasing wavelength is longer than the lasing wavelength of 1132 nm in the main text (Device DM1). This is caused by reduced radius of air holes, as shown in Table S3. The narrowest linewidth of 0.28 nm is similar to that shown in Fig. 4. Somewhat broadening occurs at higher pumping levels. Figure S9a shows spectral evolution measured on device DS2 under the lower spectral resolution. The lasing mode appears at 1213 nm, in good agreement with 2 nd mode (1213 nm) of DS2 from the simulation based on the structural parameters given in Table S3. Besides reducing the radius of air holes and increasing the width of nanobeam, the distance between the air holes in the center of nanobeam cavity was extended from 206 to 850 nm, as shown in Table S3. Therefore, the lasing wavelength of device DS2 NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 12

further increases. The double-log plots of output-input relation (Fig. S9b) shows a clear S-like curve with the best fitting by =0.2, indicating a lasing threshold of (~ 4.2 W), corresponding to a power density of ~8.7 W/cm 2. Different from the measurement of device in the main text, the pump laser for the measurement of device DS2 was focused directly on the device, and the power density was estimated using the FDTD method, the same as in the case of SI Section II. Figure S8. a, PL spectra of device DS1 at increasing pumping power levels measured at high spectral resolution level (grating at 600 g/mm). b, A selected spectrum at pumping power of 54 W, showing the linewidth and fitting. Figure S9. a, PL spectra of device DS2 at increasing pumping power levels with 150 g/mm grating. b, The double-log plot of the output vs. input relation with the rateequations fitting. =0.2 gives the best fitting. Table S3. Structural parameters of devices DS1, DS2, and DM1 (the device shown in the main text). The parameters were measured from the scanning electron micrograms of the fabricated devices. Parameter (Unit: nm) Device DS1 Device DS2 Device DM1 NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 13

W 365 390 365 L 7200 7800 7200 a1 222 214 206 a2 233 231 221 a3 245 249 237 a4 256 267 252 an 268 285 268 r1 68 66 72 r2 71 72 77 r3 74 77 83 r4 78 83 88 rn 82 88 94 References 1. Suzaki, Y. & Tachibana, A. Measurement of the μm sized radius of Gaussian laser beam using the scanning knife-edge. Appl. Opt. 14, 2809 2810 (1975). 2. Yamamoto, M. et al. Strong enhancement of Raman scattering from a bulk-inactive vibrational mode in few-layer MoTe2. ACS Nano 8, 3895 3903 (2014). 3. Bludau, W., Onton, A. & Heinke, W. Temperature dependence of the band gap of silicon. J. Appl. Phys. 45, 1846 (1974). 4. Vahala, K. J. Optical microcavities. Nature 424, 839 846 (2003). 5. Ding, K. & Ning, C. Z. Metallic subwavelength-cavity semiconductor nanolasers. Light Sci. Appl. 1, e20 (2012). 6. Ning, C. Z. What is Laser Threshold? IEEE J. Sel. Top. Quantum Electron. 19, 1503604 1503604 (2013). 7. Soref, R. & Bennett, B. Electrooptical effects in silicon. IEEE J. Quantum Electron. 23, 123 129 (1987). 8. Cocorullo, G. & Rendina, I. Thermo-optical modulatin at 1.5 μm in silicon etalon. Electron. Lett. 28, 83 85 (1992). NATURE NANOTECHNOLOGY www.nature.com/naturenanotechnology 14