Integrated nanophotonics based on wire plasmons and atomically-thin material

Photonic integrated circuits are an enabling technology in modern communications systems. The continually increasing demands for higher-speed and lower operating power devices have resulted in the continued impetus to shrink photonic components. In this work, we demonstrate a primitive nanophotonic integrated circuit element composed of a single silver nanowire and single-layer molybdenum disulfide (MoS2) flake. We show that nanowire plasmons can excite MoS2 photoluminescence via direct plasmon-to-exciton conversion along the wire and plasmon-to-photon-to-exciton conversion at the MoS2-covered wire end. We also find that the reverse process is possible: MoS2 excitons can decay into nanowire plasmons that can then be routed via the nanowire on-chip. Finally, we demonstrate that the nanowire may serve the dual purpose of both exciting MoS2 photoluminescence via plasmons and recollecting the decaying excitons.

layer molybdenum disulfide (MoS 2 ) 34 , a semiconductor being explored for its photoluminescence 35 , valleyselective properties [36][37][38] , and potential as a transistor 39 and photodetector 40,41 , is an ideal choice to couple with nanoplasmonic circuitry. In this paper, we explore the nanophotonics of a MoS 2 /Ag nanowire hybrid structure. We demonstrate coupling between a single-layer MoS 2 flake and a single Ag nanowire. We show that a plasmon excited at the uncovered end of the nanowire can propagate and excite MoS 2 photoluminescence (PL), both by direct plasmon-to-exciton conversion along the wire and by absorbing photons rescattering from the end of the wire. We also demonstrate MoS 2 excitons can decay to generate Ag-nanowire plasmons. Finally, we show it is possible for the Ag nanowire to serve a dual role as both a channel for MoS 2 excitation and subsequent extraction of the decaying MoS 2 excitons.

Results
MoS 2 /nanowire hybrid device The charge-coupled device (CCD) image in Fig. 1c demonstrates plasmon propagation and photon reemission. Laser radiation (λ = 635 nm), polarized parallel to the wire axis, is coupled from the far-field into the nanowire at the end labeled "1" in Fig. 1b using a 100× oil-immersion objective with numerical aperture (NA) of 1.3. The power at the sample is 20 µW. To reduce scattering and eliminate leakage radiation, the sample was covered in index-matching (n = 1.515) oil. In order to convert a photon into an SPP, the laser must be focused onto one of the ends of the wire; this accounts for the momentum mismatch between the incoming photon and the plasmon 19 . Due to confinement of the SPP modes, smaller-diameter wires yield shorter 1/e propagation lengths 31,43 . In addition, the SPP 1/e propagation length increases as the optical excitation wavelength increases 6 . The wires used in our study support two lower-order modes. When the incident light is polarized parallel to the wire, the light couples to the lowest-order, m = 0 SPP mode  shows the normalized polarization dependence of the signal with a visibility of 21%. While the large-diameter Ag wire has modes that can be excited with an incident field polarized parallel or perpendicular to the wire, the observed modulation indicates that the coupling is stronger when the excitation is parallel. The largest visibility that we observed on a device was 80% (see Supplementary Fig. 6). Simulations using the finite- We anticipate that plasmon-excited MoS 2 PL is not limited to the end of the wire. To investigate this, the displacement between the laser excitation and collection was adjusted to be a fraction of the wire length.
The sample is then translated so that the laser excitation is at the uncovered end of the wire. For reference, Fig. 3c shows a fluorescence scan of the full sample with circles to mark the effective positions of the spectral collection along the wire. Figure 3d presents the spectra corresponding to each of these points, starting from the top circle, labeled "1", and walking downward in Fig. 3c. We observe that the PL is strongest near the end of the wire. However, we also obtain significant signal over the entire length that the wire is

Plasmon extraction from MoS 2 fluorescence
In addition to plasmons exciting the MoS 2 , the reverse process can also occur; that is, excitons in the MoS 2 can be converted to plasmons that propagate along the wire and are rescattered as photons. To demonstrate that MoS 2 fluorescence can couple to Ag nanowire plasmons, the excitation is aligned with the overlap region of the MoS 2 flake/nanowire end, and the collection focal volume is aligned to the uncovered nanowire end (the reverse configuration of Fig. 3). Figure 4a shows the resulting fluorescence image when the sample is scanned in this configuration with the excitation laser polarized parallel to the wire. Again, a CCD image of the MoS 2 /wire structure is overlayed on this image. Compared to the localized feature in Fig. 3a, the present image shows an attribute that extends beyond the end of the wire. This is suggestive of plasmonic excitation along the MoS 2 /wire interface, not just at the covered end. As the laser excitation scans over The petri dish with the wire sample was then filled with DI water, and a suitable wire was found using an inverted microscope with a long working distance objective. The PMMA film was transferred to the petri dish, and a post with a teflon-coated end attached to a micropositioner was brought into contact with the PMMA. Using a Harvard PhD 2000, the water was pumped from the petri dish; the objective focus could be adjusted to position the flake over the wire. The sample dried before immersion in an acetone bath to dissolve the PMMA.

Optical characterization
The samples were characterized with an inverted microscope equipped with an oil-immersion objective. A nanopositioning stage (Mad City Labs, Inc.) was used to scan and position the sample. The sample was characterized using a 532 nm wavelength laser for Raman spectroscopy or a 633 nm wavelength laser for photoluminescence and plasmon propagation measurements. Excitation polarization was controlled by a half-wave plate. The signal from the sample was sent to either an APD or a spectrometer. Longpass filters to block the laser line were used in front of both detectors. For some of the fluorescence images, a similar second set-up with a 635 nm wavelength laser was used. Laser power of 5 µW was used in Fig. 2, Fig. 3, and Fig. 4b and 4c. Laser power of 20 µW was used in Fig. 1c and Fig. 3a. Laser power of 70 µW was used in Fig. 1b Estimation of photon re-emission efficiency from wires Supplementary Figure 1a displays a bare wire of length 5 µm (wire in text was about 7 µm long).
Supplementary Fig. 1b shows a scanning electron microscope (SEM) image of a different wire to show the tapered end geometry. We show plasmon propagation resulting in photon reemission for the wire in Supplementary Fig. 1a in Supplementary Fig. 1c with the laser polarized parallel to the wire, and Supplementary Fig. 1d displays the image resulting from displacing the collection and excitation focal volumes by a transverse length equal to the wire length. These images are contrasted to Supplementary Fig. 1e and 1f, in which the excitation is perpendicular to the wire. The filters in front of the APD are an optical density filter to cut three orders of magnitude of light (OD3) and a 633/10 bandpass. From this, we can estimate the photon-tophoton conversion efficiency. For this wavelength and power, the input photon flux is approximately 6.4x10 13 photons/s. At the APD, the photon flux is approximately 2.5x10 6 photons/s for the parallel excitation. Accounting for the OD3 filter (1000x reduction) and a factor of 4-6x for losses in the system (including a beamsplitter and a pinhole apparatus), we arrive at an efficiency of about 0.015% to 0.023% for this wire. The ratio of photon flux out to photon flux in takes the form of , where represents a factor encompassing the photon-to-plasmon and plasmon-to-photon coupling efficiencies, and L o is the propagation length. Using simulation data, we estimate L o to be about 3.1 µm for our wires. Using this value and the calculated efficiency for the wire in Supplementary Fig. 1a, we calculate A to be between 7.37x10 -4 and 1.13x10 -3 . Applying these values to the wire in the main text, we can estimate the efficiency to be about 0.008% to 0.012%.