1. Introduction

Plasmonic crystal laser has received increasing attentions recently [1–10] due to their easy fabrication, small divergence angle, and wide tuning range of lasing wavelength. Different from the beyond-diffraction-limit plasmonic nano-lasers [11,12], the emission area of plasmonic crystal lasers sizes from tenths to hundreds micrometers so they can provide suitable power for practical use. Potential applications includes bio-sensing, 3-D imaging, high-power laser arrays, and high-speed fiber communications [6]. Although the first demonstration of plasmonic crystal lasers made of a semiconductor using bulk InGaAs on an InP substrate [1], most studies focus on organic gain materials [2–7,13,14] as they were room-temperature working devices. Organic material can be spin-coated directly on the nanoscale metal pillar/hole array so the near-field interaction between plasmonic mode and organic molecular is significantly enhanced. Taking this advantage, organic-based plasmonic crystal lasers made enormous progresses in this decade.

Semiconductor diode lasers have been everywhere in modern life thanks to their low cost, high efficiency, high reliability, and compactness. For the studies on semiconductor-based plasmonic lasers, all previously reported lasing were limited at cryogenic temperature [1,15,16] except the recent work by Kim and the co-workers [17]. They used 1D metallic gratings on bulk InGaAsP on an InP substrate and demonstrated a very wide tuning range of ∼ 400 nm at room temperature. The exact reason for realizing room temperature lasing in their work was not stated clearly. Comparing with the previous works using metal hole array, their devices using 1D grating benefited from a lower metal coverage (∼60%) and a thicker gain medium layer. The lower metal coverage may be a key factor as the metallic loss could be reduced. Their thick InGaAsP layer (∼700 nm) may not only provide more gain but also have guiding effect for the lasing mode. In the present work, the room temperature plasmonic crystal lasers on a GaAs substrate have been demonstrated. Our lasers deploy two-dimensional (2D) nano-pillar arrays and InGaAs/GaAs quantum wells (QWs) as the gain medium. The L-L curves, lasing spectra, and far-field patterns will be presented and discussed. We have also found an unintentionally-fabricated low-threshold laser arising from the plasmonic mirror cavity effect and the designed waveguide structure. This work reveals the great potential of these intriguing devices and paves the way for developing electrically-driven plasmonic lasers.

2. Sample growth, device fabrication, and measurement setup

The sample (RN1660) was grown on an n-typed GaAs (100) substrate by a solid-source molecular beam epitaxy (MBE) system (Veeco Gen II). Figure 1(a) shows the schematically epi-structure. The gain region is consisted of four 8-nm-thick In_0.2Ga_0.8As/GaAs QWs. The deepest QW locates at the depth of ∼100 nm to ensure the overlapping of gain medium and surface plasmon mode. The bottom cladding layer of 1-μm-thick Al_0.4Ga_0.6As prevents the mode from leaking to the substrate. The growth temperature was 585 °C except that was 510 °C for the QWs region. The sample being a p-i-n diode was originally designed for electrical injection but only the optical pumping devices are investigated in this work.

 

Fig. 1. (a) Schematic sample structure prepared by molecular beam epitaxy; (b) Top-viewed layout of the fabricated 56 devices with the various periods and e-beam doses.

Download Full Size | PDF

Two-dimensional square-lattice nano-pillar array is chosen for our plasmonic crystal lasers. The nano-pillar array on the sample surface was prepared by e-beam lithography with a double-layer resist (PMMA A3 and A5), the e-gun evaporation (2-nm Ti and 65-nm Au), and the following lift-off in acetone. The period of square lattice, denoted as P, ranges from 260–280 nm in a step of 10 nm and from 290 to 310 nm in a step of 5 nm. For each period, there are seven devices with different pillar diameters controlled by the exposed e-beam dose. The layout of all devices to be discussed is illustrated in Fig. 1(b). The size of single devices is 80 × 80 μm². The center-to-center distance between adjacent devices are 250 μm in horizontal and 500 μm in vertical. In total, there are fifty-six devices with the varied periods and pillar diameters. In this paper, the devices are named as Pxxx-Dx for clarity. As an example, the device with the period of 305 nm and the e-beam dose of 6 is named as P305-D6.

Figure 2 shows the scanning electron microscopy (SEM) images taken on the devices with the same fabrication condition on the same chip. The square-lattice pillar array was clearly pictured. The pillar sizes of the three devices, P270-D3, P270-D5, and P270-D7, are about 98 nm, 110 nm, and 127 nm, respectively. We found that the pillar size is independent to the lattice period P and ranges from about 80 nm to about 130 nm from dose 1 to dose 7. Therefore, the filling factors or metal coverage span approximately from 7% to 25%.

 

Fig. 2. Scanning electron microscopy images on the devices fabricated with the same condition of P270-D3, P270-D5, and P270-D7 in (a)–(c), respectively.

Download Full Size | PDF

The optical measurements were performed at room temperature. The devices were optically pumped by a normally-incident 532-nm pulsed laser (Matrix 532-7-30) with a repetition rate of 30 kHz and a pulse duration of 14 ns. The spot size of the excitation laser was ∼ 240 μm on the device. A dichroic mirror cut at 650 nm was transparent for the pumping laser and reflective for the light emitted by the device under test. The emission light from the device was either taken by a far-field camera (Hamamatsu FFP optics system of A3267-12 with the near infrared camera of C5840) or analyzed by a monochromator (iHR550) equipped with a thermal-electric (TE) cooled silicon photodiode. A linear polarizer in front of the camera or monochromator was used to check the polarization dependence.

3. Measurement results and discussion

3.1 Lasing characteristics

We first focus on one of the devices as an example. Figure 3(a) shows the L-L curves on the device of period P = 270 nm and dose D = 6, named as P270-D6. The lasing threshold P_th is about 15 kW/cm² which is 960 mW on the 80 × 80 μm² device area. In terms of pulse energy, the threshold is about 32 μJ, which is much higher than that in the previous report [17].

 

Fig. 3. Lasing characteristics of device P270-D6. (a) L-L curve; (b) Emission spectra taken respectively at the excitation power densities below and above its threshold; (c) Lasing spectra at three excitation power densities; (d) Polarization-dependent far-field patterns above the threshold.

Download Full Size | PDF

Figure 3(b) shows the emission spectra at the excitation power below threshold (12.1 kW/cm², ∼0.8 P_th) and above threshold (15.5 kW/cm², ∼1.03 P_th). It clearly confirms the lasing action of the device as the linewidth shrinkage is so dramatic. Besides, the emission spectra below the lasing threshold tells that the ground state is at ∼980 nm and the first excited state of electrons and heavy holes is at ∼900 nm, which have been observed repeatedly in our devices and are well consistent with the calculation [18]. The lasing of device P270-D6 peaks at 890.65 nm which is above the first excited state and could be one of the reasons for its high threshold power density. Figure 3(c) exhibits the normalized lasing spectra at three excitation power densities. The lasing peaks at 890.65 nm are independent of the excitation power density. However, the linewidth (the full width at half maximum, FWHM) increases from 0.54 nm to 0.96 nm with the increasing power densities. The linewidth braodening could be due to the state filling effect [19], the sequential lasing of two nearly degenerate and linearly-polarized states, carrier-induced refractive index change [20], and other reasons. The stable lasing peak and the very narrow linewidth advantage its future applications. The four pictures in Fig. 3(d) are the lasing far-field patterns (FFPs) without and with the polarizer. First, the upper left image exhibiting a cross-like FFP was taken without a polarizer in front of the camera. The vertically polarized (V-polarized) and horizontally polarized (H-polarized) images on the right half both show a dumbbell-like FFP, which together consist of the observed non-polarized image, which is confirmed by the lower left 45°-polarized FFP as it resembles the non-polarized one. It is worth noting that the divergence angle is small, about 4°, and the angle between the two maximums of the dumbbell is about 2.6°–2.8°.

3.2 Overall device lasing map

All devices have been measured to understand the overall lasing behavior. Figure 4(a) illustrates the lasing spectra for the devices P260-D5–P310-D5, together with the photoluminescence (PL) spectra taken in the unprocessed area on the same chip. Obviously, the lasing wavelength was determined by the period (or lattice constant) of the plasmonic crystal. For the devices of P = 260–310 nm, the lasing wavelength increase from 864.65 to 1001.0 nm. That is, the tuning range of the plasmonic crystal lasers is about 136 nm. Accordingly, the effective refractive index n_eff, calculated by the ratio between the lasing peak (λ_p) and the period P, decrease from 3.326 to 3.229 with the increasing period. The room-temperature PL spectra shows a wide distribution with the peak at about 986 nm. In the following discussions, the devices are divided into two groups based on the period P. Group A includes the devices with P = 260, 270, and 280 nm, and group B includes those with P = 295–310 nm.

 

Fig. 4. (a) Lasing spectra for the devices with the periods P = 260–310 nm and the dose D = 5, together with the PL spectra from the unprocessed region; (b) Overall device lasing mapping.

Download Full Size | PDF

Figure 4(b) illustrates the lasing mapping of all devices. Let us focus on group A devices first. We found that all seven devices with P = 260 nm lase but the number of lasing devices reduces to six for P = 270 nm and to four for P = 280 nm. This trend can be explained as follows. The higher dose in e-beam writing gives larger size of the nano-pillars, as we have seen in Fig. 2. The fundamental mode wavelength of the localized surface plasmon (LSP) supported by the individual nano-pillars decreases with the decreasing pillar size. It is known that LSP is crucial for the coupling efficiency between the emitter and the propagating surface plasmon polariton (SSP) whose characteristic wavelength is determined by the lattice period [6,21]. Therefore, for the large period devices such as P = 270 and 280 nm, the lasing wavelength could not be supported by the smaller pillar as its fundamental mode wavelength of LSP becomes too short for that period. For group B devices, their situations are more complicated than we first thought [22]. It appears in Fig. 4(b) that the devices with the dose D = 3–6 are lasing. The lasing from these devices, however, is actually not arising from the devices themselves as we shall discuss in detail in the next section.

3.3 Mirror cavity effect

It takes us quite a while to learn that the devices in group B did not lase but the unintentional device in the area between two devices in horizontal direction did. We named those devices as “mirror cavity” devices, similar to those observed at cryo-temperature very recently [23]. For convenience, the device located at the area between, for example, the devices P305-D4 and P305-D5 is denoted as device P305-M45 (‘M’ for mirror). As shown above in Fig. 1(b), the center-to-center distance between two devices in horizontal direction is 250 μm. Taking the device size of 80 μm into account, the width of no-pillar region between two devices is about 170 μm. When we did measurement on, for example, device P305-D4, the pumping laser was focused on its center but the area between P305-D4 and P305-D5 (that is device P305-M45) was also pumped because of the large pumping spot size (∼ 240 μm). It is difficult to exclude the emission from the area surrounding the device under test. To investigate this mirror cavity effect, we performed the measurement on two sets of the devices, one is from group A (P270-D4, P270-D5, and P270-M45) and the other one is from group B (P305-D4, P305-D5, and P305-M45).

Figure 5(a) shows the L-L curves of the three devices from group A. For device P270-D4 or P270-D5, the excitation laser spot was focused at the center of the device. For device P270-M45, the spot was aimed at the middle region between the two devices instead, as sketched in the inset on Fig. 5(a). The resultant L-L curves of devices P270-D4 and P270-D5 are nearly the same, with a threshold power densities around 30 kW/cm². For device P270-M45, the threshold is nearly the same but with a smaller slope efficiency. In Fig. 5(b), the power-dependent lasing spectra of the mirror-cavity device P270-M45 exhibit clear oscillations with a period about 0.5 nm, which is very different from those from P270-D4 and P270-D5 [not shown here, see Fig. 3(c) instead]. By using the formula of mode spacing in a Fabry-Perot cavity ($\Delta\lambda=\lambda^{2}/2nl$) and the cavity length of 170 μm (the distance between the two devices), the refractive index n is about 4.70. Although this value, as expected, is larger than the effective refractive index n_eff of 3.298 at this period. It is difficult to justify this value without the knowledge of the exact band structure of the plasmonic crystal. In addition, the effective cavity length may depend on the field penetration depth into the nano-pillar array so the refractive index n may be overestimated [23]. Although the exact physical model for the effective cavity length is not complete, the two adjacent devices do act like two mirrors for the device in the middle area to achieve lasing.

 

Fig. 5. (a) L-L curves of the devices P270-D4, P270-D5, and P270-M45. Inset: schematic for the pumping spot on the area between two devices; (b) Lasing spectra from device P270-M45 under various pumping power densities; (c) L-L curves of the devices P305-D4, P305-D5, and P305-M45. Inset: OM photo showing the scratched lines outside a device. (d) Lasing spectra from device P305-M45 under various pumping power densities. (e)–(h) respective far-field patterns from devices P270-D5, P305-D5, P270-M45, and P305-M45.

Download Full Size | PDF

However, the situation for the devices in group B is different. Figure 5(c) shows the L-L curves for devices P305-D4, P305-D5, and P305-M45. The threshold power density of the mirror cavity device P305-M45 is too small (<10 kW/cm²) to be measured with our current setup. Actually, we found that, in this case, the lasing from the devices P305-D4 and P305-D5 did not arise from the devices themselves. The emission in the L-L curves of P305-D4 and P305-D5 actually came from, mostly, the area between the two devices. Because the large spot size in our system, the excitation laser not just pumped the device but also the area surrounding it. The threshold of the mirror cavity laser (P305-M45) is much lower than the two nearby devices (P305-D4 and P305-D5) so the mirror cavity laser achieves lasing even the pumping laser was not aimed at the middle region and caused the false threshold behavior in the L-L curves of P305-D4 and P305-D5 in Fig. 5(c). We have also double checked this argument by using a scriber to make two scratched lines outside device P305-D4 after all measurements, as shown in the inset on Fig. 5(c) and found that it was not lasing anymore. The same method has been deployed on device P270-D4 in group A but no clear change of the lasing behavior was observed. Figure 5(d) shows the power-dependent lasing spectra of device P305-M45. The Fabry-Perot oscillations, probably overtaken by the widened spectra, were not observed. Furthermore, the FFPs in Figs. 5(e)–5(h) tell the same story about the actual devices and mirror cavity ones. Figure 5(e) shows the FFPs of device P270-D5 in group A. Similar to the other device P270-D6 in Fig. 3(d), the FFPs in Fig. 5(e) exhibits a cross-like shape consisted with two orthogonal and linearly polarized dumbbell-like components. In contrast, the FFPs of device P305-D5 in Fig. 5(f) is purely V-polarized because the FFPs was coming from the mirror cavity device P305-M45 (and/or P305-M56). The dumbbell-like V-polarized FFP arising from the horizontally propagating wave is also observed in the mirror cavity devices P207-M45 and P305-M45 shown in Figs. 5(g) and 5(h), respectively.

It is worth noting that the mirror cavity laser, although it was not fabricated intentionally and revealed by accident, could be of great potential. Its threshold is low and its linewidth is less than 1 nm as shown in Figs. 5(c) and 5(d). Similar trend has also been observed with the most devices in the respective group. In short, the device in group A achieves lasing by itself but those in group B does not. Its possible reason is that, for group B devices, the gain in the ground state is not high enough to compensate the metallic loss. In addition, the lasing wavelength of the mirror cavity devices in both groups is determined by the period of the two adjacent plasmonic crystals. We reckon that the plasmonic crystal serves as a mirror with a narrow reflection band [23]. Further investigations on its mechanism are certainly needed to clarify this explanation and to understand its potential for the applications on plasmonic lasers and other optoelectronic devices.

4. Conclusions

The room-temperature 2D plasmonic-crystal lasers on a GaAs substrate have been presented. The lasing peaks range from 865 to 1001 nm as the lattice constant of nano-pillar array increases from 260 to 310 nm. The surface-emitting plasmonic-crystal lasers have small divergence angle and extremely narrow linewidth. An unintentionally fabricated mirror-cavity lasers exhibit very low threshold power density and could be of great potential for making electrically-driven laser in the future. Our work suggests that plasmonic laser is a good candidate for various applications such as 3D imaging and bio-sensing.