1. Introduction

Semiconductor nanowires are attractive nanomaterial because they have a variety of optical and electrical functionalities. They have huge potential as innovative devices in high-density integration in high-performance transistors [1,2] for complementary metal oxide semiconductor (CMOS), photonics-electronics convergence platforms [3], ultimately high efficiency solar cells [4,5], quantum information devices [6], and sensors [7,8]. Nanowire photonics [9–11], one of the research fields in nanowires, has also been very active in the past decade. In fact, numerous applications using single nanowires, such as lasers [12–18], light emitting diodes (LEDs) [19–21], switches [22,23], detectors [24–26], single-photon sources [27–30], and wavelength converters [31,32], have been reported. In these nanowires, telecom-band nanowires [33,34] could be especially key components of future photonic on-chip processors because the wavelength in the telecom band is available for CMOS-compatible Si photonics. In addition, with recent advances in technology for III-V nanowire growth on Si [24,35,36], photonic processors based on monolithically integrated III-V nanowires on Si are becoming more realistic. Although telecom-band nanowire lasers are anticipated as on-chip light sources for future photonic on-chip processors, there are still several problems in applying them to devices. For example, to electrically drive the lasers, they should be implemented in a p-i-n injection in a III-V nanowire and designed using appropriate electrodes [37]. Good quality p-i-n junctions have been formed by using recent technology for selective growth on Si [24,35,36], and transparent electrodes could be a solution for integration in photonic circuits [26]. Another problem is the high lasing threshold of nanowire lasers due to poor cavity structures. A low threshold is required to reduce the energy cost per bit for future photonic on-chip processors. To help in reducing it, nanowires can be combined with other nanophotonic devices [34].

The most important requirement is room-temperature continuous-wave (CW) operation. However most of the nanowire lasers reported so far have been demonstrated under pulse-pumping conditions, and the few that have shown CW lasing did so only at cryogenic temperatures. The reason demonstrations of room-temperature CW lasing have been unsuccessful is thought to be the heating problem. That kind of effect is clearly observed at high pump condition [33]. The cross sections of nanowires are generally circular or hexagonal; therefore, the thermal resistance could be elevated because the contact region between the substrate and nanowire is minimal. In addition, low reflectance at the nanowire edge further increases the lasing threshold. Furthermore, the nanowire can be heated up by a pump laser or injection current before reaching the lasing threshold. However, this last heating effect has not been investigated experimentally thus far.

We previously reported InP/InAs nanowire lasers on a metal-coated substrate at room temperature; however, they showed pulsed lasing, not CW lasing [33]. In this paper, we systematically investigate the thermal effect on nanowire lasers on different optical platforms. First, we ascertain the characteristic temperature by changing the substrate temperature and estimate the nanowire temperature by changing the pumping laser repetition frequency. Then, we compare the lasing behavior of nanowires on Au, Au with Al₂O₃, a Si waveguide, and a SiN waveguide. From the results of the comparison, we conclude that the thermal resistance between the nanowires and substrate is largely responsible for nanowire heating.

2. Nanowire laser on Au

2.1 Temperature dependence

First, the temperature dependence of a nanowire laser on Au is investigated to estimate the characteristic temperature. InP nanowires were grown from indium droplets on an InP substrate by the vapor-liquid-solid method. There were 50 InAs/InP multiple quantum wells (MQWs) embedded (nanowire details are described in Supplement 1). The average diameter and length are 1 and 10 $\mathrm{\mu}$m. Figure 1(a) and (b) show a schematic and an optical microscope image of a nanowire on Au-coated substrate. The nanowires from the nanowire growth substrate were dispersed on a 50-nm-thick Au film deposited on SiO₂ substrate by direct physical contact. To reduce unnecessary sample heating by the pump laser in the photoluminescence (PL) measurement, we used a 1060-nm pump laser that can optically excite InAs layers (PL measurement details are described in Supplement 1). The pulse width was 13 ns and the repetition frequency was 250 kHz, which limits the heating effects on the sample from the pump laser. (Setting the pump laser to a frequency of 10 kHz with a 100-ns pulse width is a common way to neglect thermal effects caused by the laser. The laser properties measured under this condition are shown in Supplement 1.) The substrate with the samples was placed on a heater and measured after the temperature stabilized. Figure 1(c) shows the resultant lasing spectra. The lasing is observed at about 1280 nm at room temperature (the lasing range of our nanowires from the same growth substrate is from 1270 to 1330 nm). Figure 1(d) and (e) show the light-in vs light-out (L-L) curve and the cavity wavelength at different temperatures (296, 324, 344, and 364 K). The results show typical lasing behavior because the clear kink in the curve appears before and after the start of lasing (this is near the lasing threshold). The emission wavelength slightly shifts towards longer wavelengths (red shift). This indicates that pump laser is heating up the sample. Increasing the heater temperature shows changes in the lasing properties as the temperature changes. As shown Fig. 1(d), the lasing emission weakens as the temperature increases. We also estimated the emission wavelength [Fig. 1(e)]. The laser demonstrates a clear red shift with the increase in heater temperature. Form these results, it was found that the wavelength shift was 0.059 nm/K. The estimated characteristic temperature of the nanowire laser is approximately modeled as ${P_{\textrm{th}}} = {P_0}{e^{T/{T_0}}}$, where ${P_{\textrm{th}}}$, ${P_0}$, T, and ${T_0}$ are the lasing threshold, a constant, temperature, and characteristic temperature, respectively [38]. Both temperatures are given in degrees Kelvin. Figure 1(f) shows the theoretical and experimental threshold ratio normalized by the threshold at 296 K for different temperatures. From this result, it is found that the characteristic temperature is estimated as ∼80 K obtained from the linear regression. In general, characteristic temperatures of InGaAsP and AlGaAs are 70 K and 100–160 K, respectively. Therefore, an 80-K characteristic temperature for our InAs/InP nanowire is reasonable, and the temperature dependence of the laser is well represented in the predicted behavior.

 

Fig. 1. (a) Schematic of nanowire on Au-coated SiO₂ substrate. (b) Optical microscope image of nanowire on Au-coated SiO₂ substrate. (c) Emission spectrum of nanowire laser. (d) L-L curve at different temperatures (296, 324, 344, 364 K). (e) Cavity wavelength at different temperatures (296, 324, 344, 364 K). (f) Theoretical and experimental threshold ratio normalized by the threshold at 296 K for different temperatures.

Download Full Size | PDF

2.2 Frequency dependence

Next, we increased the pump laser frequency to determine the resultant heating effects in these conditions. Figure 2(a) and (b) show the L-L properties and cavity wavelength at 250 kHz, 500 kHz, and 1 MHz with 13 ns of pulse width. The pump pulse shapes are the same at these frequencies. To compare these results, we plot pump power vs emission intensity per pulse. As shown Fig. 2(a) and (b), emission intensity is reduced, and a larger red shift is observed in the high-frequency condition. Because the speed of heat conductivity is ∼1 $\mu $s, the observed effect is regarded as a heating effect. From the results in the previous section, the temperature dependence of the cavity shift was determined to be 0.06 nm/K; therefore, we estimated the temperature increase as 19 K when the pump power was 8 mJ cm⁻²/pulse in the 1-MHz condition. In the 1-MHz condition, thermal effects were significant, despite the Au-coated substrate’s high thermal conductivity. This implies there is a large thermal resistance due to imperfect contact between the nanowire and substrate. This would need to be overcome to achieve CW lasing.

 

Fig. 2. (a) L-L curve of nanowire laser on Au at different frequencies (250 kHz, 500 KHz, 1 MHz). (b) Cavity wavelength at different frequencies (250 kHz, 500 KHz, 1 MHz). (c) L-L curve and cavity wavelength of nanowire laser on Au after Al₂O₃ deposition.

Download Full Size | PDF

To reduce the thermal resistance between the nanowire and substrate, we coated the nanowires on Au with Al₂O₃ using atomic layer deposition (ALD) equipment. In general, ALD is excellent in filling narrow gaps between nanowires and a substrate because, unlike in physical vapor deposition processes, the chemical vapor can fill the gaps. Figure 2(c) shows the L-L curve and the cavity wavelength of a nanowire laser on Au with 50-nm-thick Al₂O₃ deposition. Although the external quantum efficiency is reduced and the lasing threshold is increased due to unexpected light scattering and accidental modifications of the mode profile by the Al₂O₃ deposition, the wavelength shift is clearly reduced. These results suggest the ALD deposition improves the thermal conductivity (Additional data and comparisons before and after ALD are put in Supplement 1).

3. Nanowire laser on SiN and Si waveguide

3.1 Substrates dependence

Lasing properties of other nanowire lasers on different platforms are also investigated to estimate the thermal effects on different substrates. In the previous section, the nanowires were dispersed on Au-coated SiO2. In this section, nanowires were transferred onto SiN and Si waveguide substrates, which have different thermal conductivities.

Figure 3(a) and (b) show a schematic image and an optical microscopic image of a nanowire on a SiN photonic crystal (PhC) waveguide. The nanowire is transferred from a nanowire-dispersed substrate to the SiN PhC waveguide with a microgripper (see Supplement 1). The PhC is implemented in a 350-nm-thick SiN layer deposited on the Si (the fabrication process for the SiN photonic platform is described in detail in Supplement 1). The PhC waveguides are W8 types and have an air-bridge structure. The lattice constant and diameter are 620 and 350 nm, respectively. PL measurements were carried out in the same measurement condition as in Fig. 2 (1-MHz pulse frequency with pulse width of 13 ns). Figure 3(c) shows the transmission spectrum of the waveguide, which has a non-transmission band, namely a photonic band gap, between 1330–1400 nm. The L-L curve and the cavity wavelength are plotted in Fig. 3(d). The lasing wavelength is ∼1334 nm, and it is in the photonic band gap as shown in Fig. 3(c). Therefore, the emission cannot couple to the waveguide mode and is easily detected from the top direction.

 

Fig. 3. (a) Schematic of nanowire on SiN PhC waveguide. (b) Optical microscope image of nanowire on SiN PhC waveguide. (c) Transmission spectrum of SiN PhC waveguide. (d) L-L curve and cavity wavelength of nanowire laser on SiN PhC waveguide. (e) Schematic of nanowire on Si waveguide. (f) Optical microscope image of nanowire on Si waveguide. (g) Transmission spectrum of Si waveguide. (h) L-L curve and cavity wavelength of nanowire laser on Si waveguide.

Download Full Size | PDF

The nanowire lasing on Si waveguide was also measured. A schematic image and an optical microscopic image of the nanowire are shown in Fig. 3(e) and (f). The Si waveguide is on SiO₂. The Si thickness is 220 nm and the width is 3 $\mathrm{\mu }$m. The Si waveguide has a wider transmission bandwidth as shown in Fig. 3(g). Figure 3(h) shows the L-L curve and the cavity wavelength. In this case, the emission intensity is weaker than the emission from the nanowire on the SiN waveguide because the emission is better coupled to the waveguide.

3.2 Comparison among different platforms

Finally, we compare the heating effects of nanowire lasers on different platforms. Figure 4(a) summarizes the cavity wavelength shift for different substrates. By comparing the wavelength shift for a nanowire laser on Au, Si, and SiN, it was found that the wavelength shift for SiN and Au was slightly larger than that of Si, despite Au and Si having a larger thermal conductivity than SiN. There is seemingly no strong relationship between substrate thermal conductivity and sample heating. One possibility is that the heat resistance between a nanowire and substrate could have large effect on the thermal conductivity. To estimate the roughness of the material surface, we carried out atomic force microscope (AFM) measurements. Figure 4(b) shows the AFM images of Au, Si, and SiN substrates and a nanowire to estimate Rq, which represents root mean square average of height deviations taken from the mean image data plane. The Rq values for Au, SiN, and Si are 1.5, 1.3, and 0.27 nm. In our samples, the surfaces of the Au and SiN substrate are rougher than the surface of the Si; therefore, their thermal resistances should be higher as well. This is consistent with the experimental results in Fig. 4(a) because the wavelength shift of a nanowire on Si is slightly smaller than that of one on Au and SiN due to the smaller thermal resistance.

 

Fig. 4. (a) Cavity wavelength shift for different substrates. (b) AFM images of Au, Si, and SiN substrates. (c) Temperature distribution of 10-µm-long nanowire on Au before (right) and after (left) ALD deposition.

Download Full Size | PDF

As mentioned in the previous section, conversely, a nanowire on Au coated with Al₂O₃ shows a much smaller cavity shift, which implies sample heating is smaller. This suggests that ALD deposition decreases the thermal resistance between the nanowires and substrates considerably. To estimate the temperature increase in nanowires, we used the finite element method (COMSOL5.5, heat transfer module) to simulate the temperature distribution of the nanowire on Au before and after ALD deposition [see Fig. 4(c)]. This module is based on the heat equation. The simulation area was 1 mm x 1 mm x 0.5 mm. A heat source of 0.4 mW was placed on the nanowire, with an initial temperature was 293K. This was done because the actual average input power of the 1 MHz pulse laser is 0.8 mW when the pump power is 8 mJ cm⁻²/pulse in the 1 MHz condition and we assume 50% of the pump power is kept in the nanowire, considering microsecond thermal diffusion rates. In this model, we also regard the heat source as the whole nanowire itself and the temperature is homogeneously distributed throughout the nanowire. We assumed the contact between the nanowire and substrate is approximately rectangular at 50 nm x 10 µm. The reason is that the nanowire cross section is not an ideal circle and <110> face is slightly flat (scanning electron microscope image is shown in Supplement 1) and the width of flat face is 50–100 nm although this can vary between samples. The temperature is increased to 303 and 314 K with and without Al₂O₃, respectively. This simulation result shows that our nanowire without Al₂O₃ can be heated by 17 K after the lasing threshold. Our nanowire with Al₂O₃ also shows lower heating than the nanowire without it. This helps to accurately describe our experimental results (19 K of estimated temperature increase) in section 2.2.

4. Conclusion and outlook

In this work, we clarified that the heating effects increase with the thermal resistance between nanowires and substrates. Nanowires placed on Au-coated substrate before and after Al₂O₃ deposition and on Si and SiN waveguides were further compared. Through these systematic measurements, it is found that Al₂O₃ deposition by the ALD machine can reduce the thermal resistance. This suggests that other materials capable being deposited, for instance AlN, which has a higher thermal conductivity than Al₂O₃, could further improve laser performance. Of course, improving the nanowire’s crystal quality to reduce the surface roughness is also required. In addition, nanowires with hexagonal cross-sections could contribute to reducing the thermal resistance because of the increased contact region between nanowires and substrates will be increased. Moreover, we should consider not only thermal management but also reduction of the lasing threshold. For example, implementing additional cavity structures could reduce the lasing threshold. Although there are still some problems that need to be solved, we believe that CW nanowire lasers can be realized for future photonic on-chip processors.