Introduction

The surface plasmon resonance (SPR) has great potential in improving the performance of subwavelength photonic devices because of its capability in enhancing the local electrical field surrounding subwavelength metallic structures [1]. III-V and II-VI compound semiconductor related optoelectronic applications such as wavelength-tunable filters [2, 3], optical modulators [4, 5] and light emitters and detectors [6–8] have benefitted from the SPR effect. Randomly placing a metallic particle on top of a single quantum emitter, such as a quantum dot made of InGaAs and GaAs, can lead to either a large PL enhancement when the dimensions of the metallic particles are properly designed [7, 9], or a PL quenching of the quantum dot [10, 11]. Using a systematically-patterned metallic structure on semiconductor quantum emitters offers another way to control the PL emission [12, 13]. The optical spectral response of the structure can be readily observed from the reflectance and transmittance, and controlled by changing the arrangement of the structure, for instance, pattern period and dimension of the metal units. PL emission enhancement can be achieved by using well-designed patterns. Electron beam lithography and focused-ion-beam lithography are normally used for the patterning and fabrication of the structures of different dimensions, but they are slow and difficult in large scale fabrication. Laser interference lithography, however, can do large area patterning in a fast way. It has been previously used to pattern two-dimensional periodic gold nano-patterns on InGaAs quantum well samples to enhance the PL emission. A dielectric layer with a lower refractive index than the substrate was used to couple the SPR and PL emission [14]. The wavelength of the SPR, however, is fixed in this method because of the fixed metallic structure fabricated by interference lithography and the fixed value of the medium’s refractive index. It is therefore highly desirable that there be a technique to render flexibility in SPR tuning.

In this work, we propose to fabricate a quantum well nanodisk (QWND) array capped with Au particles by using laser interference lithography and tune the SPR to enhance the PL emission of the QWNDs. The SPR wavelength can be easily tuned by adjusting the dimensions of the QWNDs during wet chemical etching. Our approach is free from the wavelength restriction of the SPR coupling from a fixed metallic pattern structure and a fixed medium’s refractive index. The wavelength relationships between the optical excitation, PL emission from QWNDs and SPR of the structure are studied by both simulation and experiment. This method provides the freedom to apply the same metallic structure onto quantum emitters with different emission wavelengths and quantum structures such as SPR coupled quantum wires and quantum dots.

Experiment and Results

Figure 1 shows a schematic diagram of a 2-D orderly assembled array of InGaAs QWNDs fabricated by a top-down wet chemical etching method. The gold cap on top of each QWND serves as an etching mask in forming the disk structure and is later used as a plasmonic structure to enhance the PL from the underlying InGaAs QWND. As each gold particle is located in the vicinity of a QWND, the optical energy density surrounding the gold and the disk is promoted to a higher level by SPR coupling [15]. This enhancement of optical density increases the absorption rate at certain excitation wavelengths depending on the dimensions of the Au/disk structures. Enhanced PL emission from a QWND occurs when the SPR mode matches the disk’s emission wavelength. In the structure, light interaction that contributes to the PL enhancement is only observed within individual Au/disk structures instead of through the surface lattice resonance created between the Au array particles [16]. This is discussed further in a later section.

 

Fig. 1 (a) A schematic diagram of the Au/InGaAs structure. The gold cap is 180 nm in diameter and 20 nm in height. The period of the array structure is 290 nm. The purple-colored material is the GaAs substrate and capping layer; the pink-colored material is the InGaAs QWND. (b) The cross section of the Au/InGaAs disk structure showing different diameters and heights after various etching times.

Download Full Size | PDF

The as-grown InGaAs/GaAs QW sample consists of a single 8 nm thick In_0.12Ga_0.88As QW layer sandwiched between a 20 nm thick GaAs cap layer and 500 nm buffer layer grown by molecular beam epitaxy (MBE). A two-dimensional (2D) square lattice array with a pitch of 290 nm was patterned on negative resist NR7-200P by interference lithography using a 325 nm wavelength He-Cd laser. Meanwhile, a comparison QWND sample was fabricated by positive resist S1805, with the same periodicity as in the negative resist sample. A 20 nm gold thin film was then deposited onto the top of the negative resist patterns by electron beam evaporation to form a 2D gold particle array with a particle diameter of 180 ± 10 nm after the lift-off process. Afterwards, the gold particles served as masks in the wet-etching process to form the QWND array by etching through the InGaAs QW layer using HCl/H₂O₂ solution [17]. Considering the high sensitivity of GaAs related materials to non-radiative surface recombination during the etching process, a wet etching method was used to reduce the potentially unwanted defects and surface states, which might be generated by dry etching methods [18, 19]. Based on the design of the Au/disk structure, the SPR wavelength is able to be directly tuned by simply changing the size of the Au particles or the dimensions of the supporting disks (Fig. 1(b)). As the etching time is gradually increased from 5 s to 15 s, the dimensions of the underlying disks change with an increasing height and a narrowing diameter. When the etching time reaches 20 s, a gold cap crash will occur as the supporting area of the underlying disk is too small to hold the gold particle. With the help of these processing steps, Au/disk structures were successfully fabricated.Figure 2 shows scanning electron microscope images of Au/disk samples corresponding to three different etching times, etching for 5 s (indicated as Structure A), 10 s (indicated as Structure B) and 15 s (indicated as Structure C).

 

Fig. 2 Scanning electron microscope images of Au/disk structures etched for (a) 5 seconds (Structure A), (b) 10 seconds (Structure B), (c) 15 seconds (Structure C).

Download Full Size | PDF

In the fabrication process, the periodicity of the pattern is designed to be 290 nm which is close to the smallest dimension with a good uniformity the laser interference set-up is able to reach. The reason behind using this periodicity is that the SPR between the Au and GaAs medium of which we want to take advantage lies in the near infrared range, but still has some distance to the InGaAs QW emission wavelength. Hence the smaller the Au particle that can be fabricated, the closer the SPR wavelength and the QW emission wavelength can be brought to each other. In addition to this method, changing the environment of the Au particles provides another way to help further tune the SPR of the structure to the PL emission wavelength. Adding a lower refractive index layer between Au and GaAs can help tune the SPR wavelength, but the increased distance between the SPR and GaAs may reduce the coupling intensity within the structure. Also, the refractive index of the additional layer has to be delicately designed to ensure effective coupling between the SPR and QW emission [14]. Another approach used here is removal of an area of underlying materials by an etching method in order to tune the SPR wavelength of the structure more easily.

Reflectance spectra of the Au/disk samples with different etching times are then measured by a CRAIC UV-visible-NIR microspectrophotometer, as shown in Fig. 3(a).The peak and the dip in the reflectance spectra in the near-infrared region are caused by the SPR. However, we should note that neither the peak nor the dip in the reflectance curve represents the accurate resonant frequency which lies somewhere between these two positions; this will be discussed later. It can be clearly seen that the reflectance dip blue-shifts from 1480 nm to 1060 nm as the structure is changing from A to C. Lumerical FDTD was used to numerically analyze the SPR properties and reflectance for Au/disk structures. In the simulation, a light source was placed above a single Au/disk structure with periodic boundary conditions on the lateral dimension and perfect matched layer (PML) conditions in the vertical direction. Since the In_0.12Ga_0.88As only contains 12% Indium, the default values of refractive index n and extinction coefficient k of GaAs in the software were used for the whole disk [20]. This approximation is acceptable as the differences of n and k between In_0.12Ga_0.88As and GaAs are small enough. For the n and k of Au, we chose values based on previous measured results for better appropriateness in the simulated wavelength range [21, 22]. The simulated reflectance spectra of the Au/disk structures of the same dimensions as the experimental ones are shown in Fig. 3(b), which indicates an SPR in the near-infrared wavelength range and a blue-shift of the SPR position as the structure is changing from A to C. A sharper shape of the peaks and the dips in the simulation spectra comparing to those in the experiment spectra is caused by the imperfection in uniformity and size of the 2-D disk arrays in the experiment. For Structure A, the SPR wavelength is in the 1400 nm range as the contiguous GaAs/InGaAs are high refractive index materials. The blue-shift of the SPR wavelength corresponding to the gradually decreasing disk width is caused by a reduction in overall effective refractive index of the surrounding dielectric medium, since the supporting GaAs disk radius is getting smaller (as seen in Fig. 3(c)). This SPR shift is able to tune the coupling of the SPR to the QWND when the SPR position moves close to the PL wavelength.

 

Fig. 3 (a) Measured reflectance spectra of the Au/disk structures of different etching times. (b) Simulated reflectance spectra of the same structure. Dimensions of the samples after etching for Structure A, Structure B and Structure C are diameter/height of 160 nm/20 nm, 130 nm/40 nm and 100 nm/55 nm respectively.

Download Full Size | PDF

Low temperature (5 K) photoluminescence (PL) spectra of Structure A-C with/without Au caps in the same scale are shown in Fig. 4. PL measurements were excited by a 514 nm Ar + laser with a 5 mW continuous incident power. A Helium cryostat was used to create the 5 K environment and the PL signal was detected by a setup of a photomultiplier tube (PMT) in conjunction with a monochromator. The PL spectra displayed are as-collected. We did not perform any normalization according to the active PL emission volume. When we compared the PL from different samples, we always measured the PL from individual samples with and without Au in the same settings. The aim is to compare the difference in PL intensity for sample with and without Au capping to see the plasmonic resonance effect. It is true that the active material volume is different in the three samples, and the plasmonic resonance effect is larger if we normalize the PL intensity according to the volume. However, the PL intensity may not be linearly proportional to the active volume. Consequently, to avoid any overestimation, we did not normalize the results with sample volume. The reason for the PL peaks’ blue-shifts up to 9 nm with respect to different etching times is not entirely known in this part of the work. A lateral confinement effect could be one reason for the peaks’ blue-shifts as the QWNDs are gradually etched away. However, the 9 nm shift (11 meV energy shift) only occurs when the InGaAs dot diameter reaches 50 nm or even smaller [23]. The formation of an optically inactive layer on the InGaAs sidewall induced by the wet etching can probably reduce the effective diameter of the QWND and lead to this blue-shift [23, 24]. The Au cap is another possible reason for this PL peak shift. Here, however, we are more interested in the PL intensity trend due to the coupling of different SPR energies.

 

Fig. 4 Photoluminescence (5 K) relative intensity for Structure A-C with/without Au caps of different etching time in the same scale. The inset shows the peak PL intensity for each etching time.

Download Full Size | PDF

It is obvious that Structure C with Au caps shows the highest PL intensity among the three cases, while the PL intensity of Structure B is the lowest. As the etching time is increased (as shown in the Fig. 4 inset), the reduced InGaAs disk area causes the PL to decrease between Structure A and B reasonably, whereas Structure C with the Au particle capped shows an abnormally high PL emission. This greatly enhanced PL emission of the Au capped Structure C is due to the result of good coupling between the SPR and PL emission. As mentioned above, when the SPR of the structure is far from the QWND’s PL emission like in Structures A and B, the Au particles contribute no positive effect to the spontaneous emission process. When a mismatch between SPR and PL occurs, a PL quenching of the two cases appears due to the power loss of the incident laser light and the PL emission caused by the Au particles. For the case of Structure C, a 4.5-fold enhancement is found when the SPR moves relatively closer to the PL emission wavelength. The coupling between the SPR and PL emission can cause an increase in the surface plasmon induced spontaneous recombination rate and lead to an enhancement in the QWND’s photoemission [25].

However, for Structure C, there is still a gap of about 125 nm between the QWND’s emission wavelength and the valley in the reflectance spectrum as seen from Fig. 3(a) and Fig. 4. In our samples, since InGaAs and GaAs are non-transparent materials in which transmittance is hard to measure, a reflectance measurement becomes the only guide to indicate the SPR position and shift of the structure. In order to find the relationship between the overall SPR absorption position and the PL wavelength of the disks, we did further study to investigate the effective SPR position of the Au/disk structure with the same conditions as the previous simulation. Figure 5(a) shows the spectra of transmittance, reflectance, absorption and localized electrical field intensity of the Au/disk Structures A-C. We can see clearly from Fig. 5(a) that the absorption position on the reflectance spectrum is apparently 84 nm away from the position of the overall absorption peak for Structure C. Similar discrepancies in the spectra between this absorption peak and reflectance dip are found for Structures A and B as well.

 

Fig. 5 (a) Absorbance, transmittance and reflectance spectra and the electric field intensity at the QWND position based on Structures A-C as simulated by Lumerical. (b) Cross sections of the electric intensity distributions for a single Au/disk structure measured at the InGaAs PL wavelength and the SPR wavelength of Structure A-C.

Download Full Size | PDF

Figure 5(b) presents cross-section images of the electrical field distributions for the different Au/disk structures at both their specific SPR wavelengths and at 935 nm, which is the InGaAs QWND’s PL emission wavelength. The electric field distributions for the different cases of Structures A-C show the blue-shift of the resonance wavelength and also reveal the state of the SPR coupling within the Au/disk structures. The hot spot region in each image represents coupling effects occurring near the interface of the Au particle and the semiconductor. In Structure C, where the SPR position is close to the PL wavelength, the coupling between the Au particle and the QWND causes a large electric field near the disk, which helps to enhance the PL emission of the QWND, while Structures A and B only show a hot spot coupling effect at their own SPR wavelengths. When these wavelengths are out of the PL emission range, the coupled electric field becomes weak and easily dissipates out of the Au/disk region. As shown in Fig. 5(b), this PL enhancement corresponds to the plasmon resonance within each individual nanostructure cell as the local field enhancement only occurs beneath each Au particle. The surface lattice plasmon resonance that occurs between the Au particles makes no contribution to the enhancement [16, 26]. Simulation results in Fig. 5(b) show that the light oscillation only occurs beneath the Au structure region at both the PL emission wavelength and the SPR wavelength of Structures A-C. If there were any interactions between the Au disks, light collection would be observed in the gap region. Because of this strong enhancement behavior in each individual cell at the PL wavelength, no photonic crystal effect is observed. Extraction efficiency, one of the factors which people usually consider in photoemission enhancement in surface patterned structures of this type, makes no positive contribution to this particular InGaAs PL enhancement [27, 28].

The designed Au cap and underlying QWND structure is similar to a classical optical cavity; the cavity quantum electrodynamic (QED) Purcell effect can help to understand this PL enhancement better [29, 30]. The Purcell factor shows the spontaneous emission rate and intensity simultaneously can be used only for a rough estimation for judging the PL enhancement, as the time-resolved decay rate of the emission is not shown here [31]. The structure forms an inherent sub-micron-cavity with a Q and an ultrasmall cavity volume V [30]. The Purcell factor enhancement of luminescence is defined as

Conclusion

In summary, we have successfully fabricated ordered arrays of plasmonic Au/disk structures and demonstrated PL enhancement by a SPR coupling effect between the Au and the QWND. The structure was fabricated by interference lithography and easily-controlled chemical etching. We found a blue-shift and an effective tunability of the SPR wavelength when the disk’s diameter was reduced with a fixed Au cap size. A 4.5-fold enhancement in PL was obtained when the plasmonic coupling was at its maximum, indicating good alignment between the Au/disk SPR oscillation and the intrinsic PL emission. The relatively strongest coupling spot was found to shift to shorter wavelengths corresponding to a blue-shift in the SPR wavelength. This work provides a promising way to fabricate plasmonic enhanced optoelectronic devices in InGaAs and other semiconductor materials.