Photothermal Characterization of MoS 2 Emission Coupled to a Microdisk Cavity

Transcription

1 Photothermal Characterization of MoS 2 Emission Coupled to a Microdisk Cavity Jason C Reed 3, Stephanie C. Malek 3, Fei Yi 3, Carl H. Naylor 4, A. T. Charlie Johnson 4, and Ertugrul Cubukcu 1,2 * 1 Department of NanoEngineering, 2 Department of Electrical and Computer Engineering University of California, San Diego, La Jolla, CA 92093, USA 3 Department of Materials Science and Engineering, 4 Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104, USA All correspondence should be addressed to * ecubukcu@ucsd.edu Abstract: Integration of emerging two-dimensional direct bandgap semiconductors onto optical microcavities is important for nanophotonic light sources. In most cases, to achieve high quality factors, such microcavity designs require thermally isolated structures leading to pronounced photothermal effects. Here, we report experimental results on spectroscopic and time-domain characterization of photothermal response from MoS 2 monolayers coupled to microdisk resonators. We find that judicious utilization of pulsed laser excitation can circumvent irreversible photoabsorption induced material damage. Our results agree well with finite element method based thermal simulations. 1

2 Transition metal dichalcogenides (TMDCs) have garnered intense research interest in the past few years. The TMDCs are semiconductor materials with a bulk indirect bandgap electronic structure. Much like graphene, the interest for this class of material comes from the ability to isolate or synthesize single molecular layers. At the monolayer limit, the TMDCs transition to a direct bandgap material with broken inversion symmetry, leading to a range of interesting optical phenomena such as photoluminescence (PL) 1-5 and resonant optical nonlinearity 6-8 arising from in-plane polarized excitons with large binding energies. However, these optical phenomena, while substantial for a monolayer of material, still produce relatively weak overall response due to the total available material volume and need to be improved upon to compete with incumbent photonic technologies. The well-established scheme of utilizing optical resonators for enhanced light-matter interactions seems particularly well suited for optical materials of monolayer thickness. Recently, optically resonant structures such as plasmonic nanoparticles 9-11 have been explored to enhance or otherwise manipulate the optical effects in TMDCs. Similarly, incorporation of TMDCs and their excitonic PL emission into high-q optical cavities is interesting for enhanced light matter interactions for increasing absorption and emission probability of photons Various optical microcavities in Fabry-Perot, whispering gallery mode, and photonic crystal configurations have been explored in the context of 2D semiconductors including materials such as MoS 2, WSe 2, and WS 13,14, Such cavity designs usually require thermally isolated structures to achieve high refractive index contrast a key condition to achieve high optical Q-factors based on total internal reflection. For example, a typical microdisk can only dissipate heat through a stem like support structure leading to accumulation of heat locally on the absorbing active light emitting materials 17. Such photothermal heating effect can be detrimental for active optical devices. In this work, we study photothermal effects induced by pump photon absorption in TMDC materials coupled to optical microdisk cavities. Specifically, we explore, using timedomain pulsed diode laser spectroscopy, these effects in MoS 2 flakes integrated into a microdisk geometry. We show that the photothermal effect becomes significant for the design of photonic devices with 2D active layers potentially leading to damage of the emitting material. 2

3 For the fabrication of the MoS 2 -coupled optical microdisk cavity, we first mechanically exfoliated molybdenite onto a thin (285 nm) thermal oxide layer on silicon substrates 17. After optical identification of the few layer MoS 2 flakes, we protect the flakes by addition of a thin layer of SiO 2 through atomic layer deposition. Using electron beam lithography, the shape of the microdisk was patterned and then etched into the SiO 2 substrate. A subsequent undercutting Si etch step was performed to release the edges of the microdisk to form the microdisk cavity structures. Figures 1a and 1b show SEM images of the fabricated microdisks. The dimensions of the microdisk are 3.8 μm in radius and 285 nm in height. These dimensions were chosen to support high-quality (Q) factor whispering gallery modes in the microdisk while taking into account physical constraints in terms of the substrate oxide thickness and MoS 2 flake size. Additionally, a semicircular notch defect was incorporated into the microdisk design which helps with input coupling of excitation light and output coupling of the cavity emission. The notch was also positioned on the opposite side of the microdisk as the MoS 2 flake area (colored purple) to separate the cavity-coupled emission from the direct flake emission. First, we study the microdisk coupled PL emission from MoS 2 flakes using a continuouswave (CW) λ = 532 nm laser illumination. The laser illumination is directed to and focused at the microdisk notch with a spot size around 1 µm in an inverted microscope arrangement. Using the same objective, light is collected at the notch and passes through a set of edge long-pass filters to filter out the excitation wavelength. We observed spectrally narrow cavity mode emission collected at the microcavity notch on top of the background MoS 2 PL spectrum using a spectrometer equipped with an EMCCD. Figure 1c shows the observed emission spectrum at various levels of the excitation laser power. We note that the total power absorbed directly by the flake is significantly lower than that at the notch due to the low coupling efficiency. In order to observe the photothermal effects, we track the intensity of the 685 nm cavity mode peak as the MoS 2 flake is illuminated with increasing pump power as shown in Figure 1d (black curve). We chose this mode since it lies slightly on the red side of the exciton emission peak. We observe a monotonic increase in the peak intensity up to 12.3 mw of illumination power at the notch. We need to note that only a fraction of this power is actually absorbed by the flake due to the limited input coupling efficiency of the notch. Above this damage threshold, the intensity of the cavitycoupled emission peaks start to decrease. Subsequent cycling of the illumination power on the microdisk (Fig. 1d; red, blue and purple curves, respectively) yields a permanently lowered 3

4 response. Similar behavior is observed for other WGM peaks as well indicating that thisdecrease in the intensity is a clear indication of photothermal damage to the active light emitting material. Figure 2a shows the spectra at three different power levels: low power: 0.47 mw, damage threshold: 12.3 mw, and well beyond the damage threshold: 22.8 mw. By extracting and normalizing the background direct MoS 2 PL at these power levels, as seen in Figure 2b, we make two observations. First, there is a noticeable red-shift of the PL spectra from 670 nm to 690 nm between the low power and damage threshold illumination levels. Second, after illumination above the damage threshold, the red-shift of the PL spectra ceases. The first observation suggests that there is an absorption induced photothermal effect: as the intensity of the illumination increases, more light is coupled into the microcavity and absorbed by the MoS 2 flake. This leads to non-radiative recombination processes in the MoS 2 flake that generate local heating. 21,22 Figure 2c shows the PL peak wavelength shift for a monolayer MoS 2 flake on an un-patterned substrate as the entire sample is heated to elevated temperatures on a heat stage. From these data, as well as the spectra shown in Fig. 2d, we can see that there is a clear and reversible redshift in the PL spectra in MoS 2 with increasing temperature. Observations based on our data set matches well with other reports on temperature-dependent MoS 2 PL spectra. 23,24 Thus, we can confidently describe the shift in the background PL of the MoS 2 in this first observation as a result of the photothermal effect. The second observation that above the damage threshold there is no further redshift suggests that thermal oxidation of the MoS 2 takes effect as a permanent photo-induced damage mechanism. Oxidation of bulk MoS 2 has been reported to become significant above 500 K. 25 Extrapolating from the data gathered on the PL peak position with respect to temperature (Fig. 2c), we can determine the temperature of the MoS 2 flake that relates to the shift in the PL background. Temperature can be tracked from the low power to the damage threshold level and above. The damage threshold corresponds to a roughly 20 nm redshift (from 670 to 690 nm) in the background PL peak and a temperature of around 550 K as determined by linear extrapolation of the PL peak vs. temperature data in Fig. 2d. This is clearly in the range of material oxidation, and any further increase in the absorbed laser intensity will only serve to oxidize more of the emitting material. The reason for no further redshift (and thus material heating) occurs beyond the damage threshold level is that as the MoS 2 is oxidized, light 4

5 absorption also wanes. However, since the optical power distribution in this whispering gallery mode cavity is not uniform, 17 parts of the material that spatially overlaps with the electric field intensity maxima will absorb and subsequently oxidize first. No changes to the spectral position of the microdisk cavity modes have been observed over this temperature range; heating of the microdisk will likely only affect the fraction of the microdisk that contains the MoS 2. The geometry and material composition of the optical cavity are also important for the photothermal response of the cavity-coupled 2-D TMDC material. We modeled the thermal response of our MoS 2 -coupled microdisk cavity using the finite element method analysis software package, COMSOL. The model includes a 3.8 µm radius SiO 2 disk with a thickness of 285 nm, which is connected to a Si substrate acting as a heatsink via a Si stem of varying radii and can be seen in Fig. 3. This geometry closely resembles the fabricated structures after the Si undercut. For simplicity, we modeled the MoS 2 flake area as one third of the microdisk area and the absorption of the MoS 2 as a heat source confined to 500 nm wide radial segment within that third of the microdisk. Figure 3a shows the results for 0.55 mw of generated heat as a consequence of absorbed laser power and the resulting thermal distribution on the thermally isolated microdisk structure. The maximum local temperature on the top surface of the microdisk reaches 550 K at the edge. This heating level was chosen to match the approximate temperature at the previously discussed damage threshold level for MoS 2. The upper portion of Fig. 3a shows a temperature line scan along the diameter of the microdisk perpendicular to the flake orientation. In addition to the absorbed power dependence (Fig. 3b), the peak temperature depends also on the geometry as shown in Fig. 3c for different stem radii. From this, it is clear that the thermal conductance of the Si pillar that increases with the stem cross section (as seen in the inset of Fig. 3a) has a large effect on the thermal distribution and maximum observed temperature, since this is the only heat conduction path to the Si substrate acting as a heatsink. Therefore it is expected that a smaller stem diameter leads to more thermal isolation and a higher overall peak temperature. In the experiments, the stem radius is determined by the amount of Si undercut from the SiO 2 microdisk during the fabrication process. In order to further investigate and explore ways to minimize the photothermal effect, we analyzed the time-domain response of the cavity system. Figure 4a shows the COMSOL simulation results for the temperature response of the microdisk shortly after the onset of heating 5

6 when the local heat source is turned on. This system behaves very similarly to a thermal RC circuit in parallel where the heat capacitance and the inverse of the thermal conductance are reminiscent of capacitance and resistance in an electrical circuit, respectively. Here, the heat capacitance is determined by the volume and specific heat capacity of the SiO 2 microdisk, while the thermal conductance depends on the thermal conductivity of Si and the stem cross-section and height. For this thermal circuit, we determine the thermal time constant τ to be 395 ns based on our heat transfer simulations. The temperature reaches steady state commensurate with this time constant (Fig. 4a). We have also experimentally characterized the time-domain thermal response of microdisk coupled MoS 2 samples. Here, instead of the CW 532 nm laser used in our previous measurements, we used a pulsed 405 nm laser diode at a constant repetition rate of 30 khz. By varying the pulse width we can determine the time-domain thermal response of the microdisk system with the MoS 2 flake acting as a local heat source induced by optical absorption at the excitation laser wavelength. Since we do not have the means to probe the temperature of the microdisk at the sub-microsecond level, to experimentally study the heating of the cavity at this time scale, we must do so through indirect methods. For this purpose, we use the measured relationship shown in Fig. 2d between the PL peak position and the temperature of the MoS 2 flake. By measuring the PL peak position at different illumination pulse widths, we can infer the temperature of the MoS 2 for an illumination of that duration. In our experiments, we chose slow enough repetition rates, i.e. much longer period than thermal time constant, to allow efficient subsequent cooling of the microdisk upon absorption of the laser pulse before the next one arrives to avoid accumulation of heat. In this approach, by varying the pulse width we can determine the time domain response as shown in Fig. 4b. For pulses much shorter than the microdisk time constant, no significant accumulation of heat or a temperature increase is expected. However, when the pulse duration is comparable to thermal response time, heat accumulates leading to a situation similar to that of CW illumination. Therefore, varying the pulse width allows us to indirectly determine the timescale involved in the local temperature increase on the microdisk, which is of concern for thermal material damage in microscale or nanoscale photonic devices. This suggests that laser illumination with a pulse width shorter than the time constant mitigates the effects of heating and damage associated with a CW illumination scheme. 6

7 Figure 4b shows the data for the increase in temperature from the shortest pulse of 50 ns for increasingly longer pulses with a constant peak laser power. From this, we can effectively measure the thermal response of the microdisk in the sub-microsecond regime. For this system, steady state is reached within 1 μs and the thermal time constant is estimated to be 184 ns. This value is a factor of ~2 less than that predicted by the model. We attribute this difference to experimental deviations from ideality and especially to size dependent thermal conductivities on the nanoscale and chemical processing. These results give us insights for the best experimental conditions for testing and design of future active photonics devices such as lasers. For example, carrying out experiments with pulsed pump lasers seem to be an effective approach as demonstrated in Ref. 13. Similarly, designs based on materials with higher thermal conductivities such as GaP 15 and operating at cryogenic temperatures 19 have the potential to circumvent photothermal damage. In conclusion, we have characterized the photothermal effect that is inherent to optical cavity-coupled 2-D TMDC material systems. This effect has been explored for MoS 2 coupled to a microdisk cavity in this work. We have shown that cavity design and fabrication should consider thermal issues alongside optimizing optical performance. Pulsed excitation of the optical cavity system cannot only help mitigate this effect but also can be used as a tool for investigating the time domain thermal response. Acknowledgements This work was supported by NSF under the NSF 2-DARE program (EFMA ). 7

8 Figure 1. (a-b) SEM images of a silica microdisk at a tilted and top-down view, respectively. The black circular regions at the center correspond to the Si stem that supports the microdisk. The false-colored purple areas denote the location of the MoS 2. Scale bars are 2 μm. (c) Cavitycoupled emission spectra collected at the microcavity notch as a function of excitation laser power. (d) Peak intensity of the 685 nm cavity mode as the excitation power is cycled to and from 22.8 mw in two consecutive cycles (in order: black squares, red circles, blue upward triangles and purple downward triangles). At ~12.3 mw on the notch with only a fraction of this absorbed by the flake, irreversible photodamage of the monolayer initiates as indicated by the consecutive power cycling measurements. 8

9 Figure 2. (a) Emission spectra at low power (0.47 mw), at the damage threshold (12.3 mw) and high power beyond the damage threshold (22.8 mw) and the corresponding background PL fits overlaid. We note that only a fraction of the excitation power is absorbed by the flake. Inset: PL spectrum at 0.47 mw. (b) Normalized background PL extracted from the fits in panel (a) at the three different power levels. (c) The thermal evolution of the PL spectra of a bare MoS 2 flake on an unpatterned substrate during heating on a heat stage. The peak redshifts with increasing temperature. (d) The position of the PL peak wavelength for the sample in panel (c) during heating to and cooling down from elevated temperatures 9

10 Figure 3. (a) Simulated temperature distribution for a SiO 2 microdisk of 3.8 μm radius for 0.55 mw of generated heat (lower panel) The heat source is confined to the area outlined by the white dotted line representing the part of the monolayer flake that overlaps with the optical modes. (upper panel) The temperature profile along the center line (i.e. y=0) of the microdisk. Inset: Schematic cross-section of the microdisk geometry showing the stem radius. (b) Peak temperature observed on the microdisk of a 2.5 μm stem radius at different absorbed power levels. (c) Peak temperature observed for a microdisk of varying stem radius and fixed 0.55 mw of absorbed power. 10

11 Figure 4. (a) Simulated thermal time-domain response of the microdisk at short time scales. The temperature reaches a steady state with a rise time of 395 ns. The absorbed power was set to be 0.55 mw with a stem radius of 3.0 μm. (b) Experimental results for the estimated change in temperature in the microdisk after illumination with a pulsed laser diode for different pulse widths. The change in temperature was determined from the shift in the PL peak position of the MoS 2 on the microdisk. The change in temperature for the pulsed excitation becomes equivalent to continuous wave illumination for a pulse width corresponding to a decay time constant of 184 ns. 11

12 References A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C.-Y. Chim, G. Galli, and F. Wang, Nano Letters 10, 1271 (2010). K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, Physical Review Letters 105, (2010). P. Tonndorf, R. Schmidt, P. Böttger, X. Zhang, J. Börner, A. Liebig, M. Albrecht, C. Kloc, O. Gordan, D. R. T. Zahn, S. Michaelis de Vasconcellos, and R. Bratschitsch, Optics Express 21, 4908 (2013). S. Tongay, J. Suh, C. Ataca, W. Fan, A. Luce, J. S. Kang, J. Liu, C. Ko, R. Raghunathanan, J. Zhou, F. Ogletree, J. Li, J. C. Grossman, and J. Wu, Sci. Rep. 3 (2013). Y. Yu, Y. Yu, C. Xu, Y.-Q. Cai, L. Su, Y. Zhang, Y.-W. Zhang, K. Gundogdu, and L. Cao, Advanced Functional Materials 26, 4733 (2016). L. M. Malard, T. V. Alencar, A. P. M. Barboza, K. F. Mak, and A. M. de Paula, Physical Review B 87, (2013). H. Zeng, G.-B. Liu, J. Dai, Y. Yan, B. Zhu, R. He, L. Xie, S. Xu, X. Chen, W. Yao, and X. Cui, Sci. Rep. 3 (2013). F. Yi, M. Ren, J. C. Reed, H. Zhu, J. Hou, C. H. Naylor, A. T. C. Johnson, R. Agarwal, and E. Cubukcu, Nano Letters 16, 1631 (2016). A. Sobhani, A. Lauchner, S. Najmaei, C. Ayala-Orozco, F. Wen, J. Lou, and N. J. Halas, Applied Physics Letters 104, (2014). S. Butun, S. Tongay, and K. Aydin, Nano Letters 15, 2700 (2015). B. Lee, J. Park, G. H. Han, H.-S. Ee, C. H. Naylor, W. Liu, A. T. C. Johnson, and R. Agarwal, in ArXiv e-prints; Vol (2015), p K. J. Vahala, Nature 424, 839 (2003). Y. Ye, Z. J. Wong, X. Lu, H. Zhu, X. Chen, Y. Wang, and X. Zhang, in ArXiv e-prints; Vol (2015), p S. Wu, S. Buckley, A. M. Jones, J. S. Ross, N. J. Ghimire, J. Yan, D. G. Mandrus, W. Yao, F. Hatami, J. Vučković, A. Majumdar, and X. Xu, 2D Materials 1, (2014). X. Gan, Y. Gao, K. Fai Mak, X. Yao, R.-J. Shiue, A. van der Zande, M. E. Trusheim, F. Hatami, T. F. Heinz, J. Hone, and D. Englund, Applied Physics Letters 103 (2013). 12

13 S. Schwarz, S. Dufferwiel, P. M. Walker, F. Withers, A. A. P. Trichet, M. Sich, F. Li, E. A. Chekhovich, D. N. Borisenko, N. N. Kolesnikov, K. S. Novoselov, M. S. Skolnick, J. M. Smith, D. N. Krizhanovskii, and A. I. Tartakovskii, Nano Letters 14, 7003 (2014). J. C. Reed, A. Y. Zhu, H. Zhu, F. Yi, and E. Cubukcu, Nano Letters 15, 1967 (2015). X. Gan, Y. Gao, K. Fai Mak, X. Yao, R.-J. Shiue, A. van der Zande, M. E. Trusheim, F. Hatami, T. F. Heinz, J. Hone, and D. Englund, Applied Physics Letters 103, (2013). S. Wu, S. Buckley, J. R. Schaibley, L. Feng, J. Yan, D. G. Mandrus, F. Hatami, W. Yao, J. Vuckovic, A. Majumdar, and X. Xu, Nature 520, 69 (2015). X. Liu, T. Galfsky, Z. Sun, F. Xia, E.-c. Lin, Y.-H. Lee, S. Kéna-Cohen, and V. M. Menon, Nat Photon 9, 30 (2015). H. Wang, C. Zhang, and F. Rana, Nano Letters 15, 339 (2015). H. Shi, R. Yan, S. Bertolazzi, J. Brivio, B. Gao, A. Kis, D. Jena, H. G. Xing, and L. Huang, ACS Nano 7, 1072 (2013). S. Tongay, J. Zhou, C. Ataca, K. Lo, T. S. Matthews, J. Li, J. C. Grossman, and J. Wu, Nano Letters 12, 5576 (2012). R. Soklaski, Y. Liang, and L. Yang, Applied Physics Letters 104, (2014). M. T. Lavik, T. M. Medved, and G. D. Moore, A S L E Transactions 11, 44 (1968). 13

14

15

16

17