Surface Plasmon Lasing & Dispersion

Surface Plasmon Lasing and

Dispersion

Erik de Vos

Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands

August 28, 2018

Abstract

This thesis comprises two research projects focused on investigating surface plasmon (SP) propagation on metal-dielectric interfaces. The characterization of a newly fabricated device, designed for further research, is also presented. This thesis first showcases a new form of SP lasing, which utilizes two metal hole arrays as cavity mirrors. Analysis of

the lasing mode provides a direct view of SP propagation and allows for determining the SP group velocity on any metal-dielectric interface. Second, high-quality measurements of the dispersion characteristics of SP-mediated emission from an actively pumped elliptical-hole array are

presented. These measurements resulted in the observation of an intriguing SP dispersion relation ω(k)and provide an insight into the

effect of symmetry-breaking of SP scattering in metal hole arrays. Looking into the future, this thesis also describes the production process as well as subsequent characterization of a newly fabricated device. The

Contents

1 Introduction 1

2 Theory 3

2.1 Surface Plasmons 3

2.2 Optical Gain Media 4

2.3 Distributed Bragg Reflectors 5

3 Samples & experimental setup 7

3.1 Samples 7

3.2 Experimental setup 8

4 A Surface plasmon laser with two arrays as cavity mirrors 11

4.1 Introduction 11

4.2 Lasing characteristics 12

4.3 Direct Imaging 14

4.4 Angle-Resolved Imaging 16

4.5 Conclusion 18

5 Surface plasmon scattering in elliptical-hole arrays 19

5.1 Introduction 19

5.2 Sample 20

5.3 Dispersion Measurements 20

5.4 Discussion 23

5.5 Conclusion 23

6 Characterization of newly fabricated device: RA3586 25

6.1 Production 25

6.2 Characterization 27

iv

6.4 Conclusion 34

7 Summary & Outlook 35

A Appendix A 37

A.1 Code for experimental setup 37

A.2 Code for creating hole patterns 52

B Appendix B 55

B.1 Map of device ’RA3586’ 55

iv

Chapter

1

Introduction

The goal of this thesis is to further understand, characterize and utilize the propagation of surface plasmons on metal-dielectric interfaces. Surface plasmons can be described as electromagnetic surface waves propagating along a conducting surface or thin film [1]. Their localization and field-enhancing properties make them ideally suited for applications in biosens-ing [2–5] , nanophotonics [6–8], integrated optical circuits [9, 10] and other areas benefiting from strong atom-field interaction and field confinement [11, 12].

Ohmic losses in the metal generally result in a short SP coherence length and limit their usage. In recent years, however, the introduction of semi-conductor and dye gain media has allowed for the compensation of these losses [13, 14]. Among other results, this has allowed for the usage of surface plasmon lasing as a means of studying the properties of SPs [15, 16]. By combining loss compensation with the structuring of metals on a nanometer size scale, SP lasing has been accomplished in geometries including metal-coated nanocavities [17, 18] and nanopillars [19], hybrid plasmonic waveguides [20] and - most important for the purpose of this thesis - in metal hole arrays [15].

Metal hole arrays are an important tool in utilizing and studying sur-face plasmon propagation. These arrays consist of patterned holes in the metal layer of the metal-dielectric interface. Surface plasmons, propagat-ing on this interface, scatter from these holes. These scatterpropagat-ing processes are not confined to the plane of the interface itself, but SPs can also scatter out-of-plane into photons. The characteristics - most importantly the dis-persion relation ω(k) - of this emission can be linked directly to those of the SPs.

2 Introduction

chapter 4 they will be used as cavity mirrors of an effective Fabry-Perot cavity for surface plasmons. When the area between the arrays is optically excited, SP lasing is observed from the edges of the arrays. The lasing characteristics as well as direct and angle-resolved imaging then results in the determination of both the group index of traveling SPs as well as an effective scattering rate for SPs impending on a metal hole array.

In chapter 5 metal hole arrays will be used in a more conventional man-ner, namely as regions - under optical excitation - for SPs to propagate. In-stead of circular holes, however, elliptical holes will be used. These tilted ellipses break the symmetry of SP-scattering. The emission, established by out-of-plane SP-scattering, is investigated and dispersion characteris-tics for SPs propagating in such arrays are presented.

Chapter 6 describes the production and characterization of a new de-vice, which was partly designed in order to improve on and generalize the measurements described in chapter 5. This device was also made as a means of studying metal hole arrays with a larger lattice spacing (a mul-tiple of the SP wavelength on the interface). Although the device did not function as desired, its transmission and emission properties lead to sev-eral interesting questions about the fabrication process.

Chapters 2, 3, 7 and the appendices A & B serve as sections for use-ful additional information such as a short theoretical background (chap-ter 2) and a description of the samples and experimental setup that were used (chapter 3). Chapter 7 summarizes the findings of the research pre-sented in this thesis and looks forward to suggest several improvements and possible further studies. The appendices contain information that is most likely only relevant to a researcher that will follow-up the research with the same experimental setup. They contain documentation of the extensive Python code that is used for obtaining and analyzing data (ap-pendix A) as well as an overview and map of the sample that is described in chapter 6 and contains 2763 hole arrays (Appendix B).

The presented investigations probe many interesting aspects of the prop-agation of SPs on structured metal-dielectric interfaces. They showcase an intriguing and - as of yet - unexplored form of SP lasing as well as an inves-tigation into the dispersion properties of SPs propagating in elliptical-hole arrays. Even though many questions still remain, the results are likely a valuable contribution to the field of SP physics.

2

Chapter

2

Theory

2.1

Surface Plasmons

Although Surface Plasmons (SPs) are intrinsically quantum mechanical, a classical description suffices for the purpose of this thesis. SPs will there-fore be defined as electromagnetic surface waves, on the interface of a material with negative and one with positive permittivity. They manifest themselves as longitudinal surface charge oscillations (in the material with ǫ1 < 0) strongly coupled to light (in the material with ǫ2 > 0). The term Surface Plasmon Polaritons (SPPs) is often used in order to emphasize this coupling.

We can describe the basic properties of SPs by considering a two-dimensional case with the x-axis associated with the propagation direction of the SPs on the interface. We choose the z-axis to lie along the surface normal. The charge oscillations, propagating along the interface, have a well defined frequency and dispersion relation ω(kx). They generate a mixed trans-verse and longitudinal electromagnetic field (E = E0exp[i(kxx±kzz−

ωt]), that exhibits a maximum at the interface and is strongly localized (orthogonal to the z-axis). kz is complex and different in the two media. Its imaginary part causes the E-field to decay exponentially along the z-axis.

Both the propagation length as well as the skin depth increase with increasing wavelength. Especially the increase in propagation length is useful for many applications based on SP propagation. Thus performing the experiments with infrared - rather than visible - light increases the functionality of these devices. This is partly the motivation for the choice of InGaAs (which emits fluorescence at telecom wavelengths λ =1400− 1700 nm) as an active layer in the work described in this thesis.

4 Theory

For SPs to couple out of the interface, surface roughness has to be in-troduced. For the experiments described in this thesis this was done with metal hole arrays, consisting of a patterned array of sub-wavelength sized holes in the metallic layer of the interface. The holes in the metallic layer act as scatterers of the SPs. The SPs do not only scatter in-plane, but can also scatter out-of-plane into photons. Photons, emitted at an angle θ with respect to the surface normal, will carry a momentum k_|| = (ω/c)sin θ parallel to the interface. This can be related to the wave-vector of the SPs traveling in an array with reciprocal lattice vectors G_i by ksp = k||+Gi, where kspis the (only relevant) in-plane momentum of the SPs. A first or-der approximation for a square lattice only takes into account scattering into the four fundamental lattice directions and thus|G_i| = 2π/a0, where a0 is the lattice spacing of the square array. In subsequent chapters, only these first-order scattering processes will be taken into account.

2.2

Optical Gain Media

For all experiments described in this thesis, surface plasmons are excited by fluorescence emanating from an actively pumped semiconductor. The physics behind this process is generally discussed in an introductory course in solid state physics. Nevertheless, it will be briefly summarized in this section.

The defining characteristic of semiconductors is the close proximity of their valence and conduction band. This means that electrons can easily be

Bandgap Valence band Conduction band k Energy Excitation pump E = ħω electron

Figure 2.1: Schematic indicating the process of spontaneous emission from an actively pumped semiconductor.

4

2.3 Distributed Bragg Reflectors 5

excited into the conduction band. For the research presented in this thesis, this happens via optical excitation (gold arrow in figure 2.1). After excita-tion, thermal relaxation causes the electron to fall down to a lower energy level in the conduction band (namely the bottom of the conduction band or the Fermi energy if this is raised to lie above the band gap). The electron will then spontaneously drop back to the valence band whilst emitting a photon whose frequency is defined by the energy difference of the drop (∆E= ¯hω).

Figure 2.2: Fluorescence spectrum of In53Ga47As through 100 nm thick gold

and 20 nm thick chrome. Maximum in-put power (100%) was 260 mW.

Pumping the semiconductor with a higher power raises the Fermi en-ergy. If the Fermi energy is in-creased to lie inside the conduc-tion band, the excited electrons fall down to this energy after thermal relaxation. This means that after spontaneous emission they expe-rience a larger energy difference which leads to emission with lower wavelengths. This bandfilling is an important means of characterizing the performance of a semiconduc-tor gain layer.

Previous measurements by Ralph

Lenssen and Frerik van Beijnum in the QO group in Leiden have deter-mined the wavelength spectrum of emission from a 130 nm thick actively pumped In_0.53Ga_0.47As-layer through a 100 nm thick gold layer and 20 nm of chrome. Their measurements also clearly indicate the bandfilling - and corresponding shift of the emission to lower wavelengths - that occurs as the input power is increased (see figure 2.2, which is a copy of figure 4.5 of the MSc thesis [21] of Ralph Lenssen ).

2.3

Distributed Bragg Reflectors

A Distributed Bragg Reflector (DBR) is a (Bragg-)mirror that is based on constructive interference of successive reflections of an incoming electro-magnetic wave from a periodic many-layer structure. As the name sug-gests, constructive interference (and thus high reflectivity) only occurs in the spectral region that (roughly) satisfies the Bragg condition for the pe-riodic structure. Typical examples of DBRs are stacks of dielectric slabs [22–24] or fiber Bragg gratings [25]. Via application of straightforward

6 Theory

Figure 2.3:Typical reflectivity of a DBR-structure as a function of wavelength.

transmission-matrix multiplication their spectral properties can be calcu-lated. This calculation, however, is beyond the scope of this section.

A typical qualitative result, however, is depicted in figure 2.3. The cen-tral - highly reflecting - region is called the stopband, because the trans-mission is effectively reduced to 0 for these wavelengths. Furthermore, by going through the calculations [26] one can also show that the angular (and frequency) dependent reflection of such a structure is approximately equal to that of a fixed mirror a distance dpen away from the front of the initial reflector. Here dpen is the penetration depth of the intensity of the incoming electromagnetic wave (i.e. the distance into the periodic struc-ture after which its intensity has reduced to a fraction 1/e of its original value).

In chapter 4 a new type of SP laser is discussed, whose spectral prop-erties are explained by considering metal hole arrays as DBRs of incom-ing SPs. In this case the reflection of the SPs is produced via successive backscattering of the SPs on layers of holes. The stopband will be asso-ciated with the spectral region where no emission from actively pumped individual arrays is found, due to the absence of optical modes.

6

Chapter

3

Samples & experimental setup

3.1

Samples

Figure 3.1: Sideview of the layer structure and SP field intensity (red curve) of the devices under investigation. Image adopted from [27]. The experiments described in this thesis were

conducted with two devices. Chapters 4 & 5 discuss measurements on the same device (named: RA2389-I), whereas the production of a second device (RA3586) is discussed in chapter 6.

Both devices have the same layer struc-ture. This layer structure consisted of a 100 nm thick gold layer on top of a 150 nm (RA2389-I) or 130 nm (RA3586) thick In0.53Ga0.47As gain layer. The InGaAs layer was lattice matched to an InP substrate. A 15-20 nm spacer comprising SiNx and InP was used to separate the gold from the gain layer. A 0.5 nm thick layer of Chromium was used in order to improve the adherence of the gold layer to the underlying spacer. On top of the gold layer, a 20 nm thick Chrome layer

damps the SPs on the gold-air interface for RA2389-I. This Chrome layer was absent for the newly produced sample RA3586. The layer structure is depicted in figure 3.1. The square of the magnetic field component of the SP field is depicted in red, illustrating the confinement of the SP field to the gold-semiconductor interface.

The purpose of the gain layer is to increase the amplitude of the SP field at the gold-semiconductor interface, which in turn compensates the

8 Samples & experimental setup 1064 nm Nd:YAG Sample Cryostat Objective 20x Polarizer Multimode Fiber (Ø 2. μm) Spectrometer CCD Filters LP 1000 nm LP 1100 nm (BP 1480 nm) Scanning stage X Y Z

Figure 3.2: Sketch of experimental setup, where the lens system depicts angle-resolved observation. By changing the final lens in front of the CCD or scanning stage, direct images of the emission can also be made.

Ohmic losses of propagating SPs. Care was taken to increase the thickness of the gain layer without the layer structure starting to exhibit waveguide characteristics for ordinary optical modes (with a wider EM-field profile), instead of SPs. The spacer layer was placed in order to prevent optical quenching of the fluorescence emanating from the excited gain layer.

For both devices square hole arrays have been patterned into the gold layer by using lift-off of sub-wavelength sized nanopillars that were pro-duced via nanolithographic method, as previously described in [28].

3.2

Experimental setup

The InGaAs gain layer is optically excited through the InP substrate by a Nd:YAG laser of wavelength 1064 nm. The pump spot was circular and gaussian with FWHM ≈ 30 µm for the measurements on elliptical hole arrays in chapter 5. An approximately elliptically shaped (FWHM 40x50 µm) excitation spot was used in the measurements described in chapter 4. For all measurements, the sample was cooled to 100 K in a helium flow cryostat in a turbo-pumped vacuum.

The emission is collected on the air-side of the device by a 20x micro-scope objective (N.A. = 0.4), combined with a f = 200 mm tube lens and produces a 20x direct image. Alternatively, by adding an extra lens, a far-field (angle-resolved) image can be produced.

Emission can be observed in two distinct ways; (i) either directly with an InGaAs CCD (Xenics Xeva 1.7-320) or (ii) by using scanning a multi-8

3.2 Experimental setup 9

mode fiber (62.5 µm core) through the image plane. A Newport Motion Controller ESP301 was used for scanning at sub-micrometer precision.

The light incident on the CCD is spectrally filtered by two long-pass filters (λ_{cuto f f} = 1000 & 1100 nm), in order to remove ambient light as well as the 1064 nm pump light shining through the sample. For the mea-surements described in chapter 4, an additional Gaussian band-pass filter (λc = 1480 nm, FWHM = 12 nm) selects part of the emission and lasing spectrum (see figure 4.1).

The emission collected by the scanning fiber is sent to a grating spec-trometer, allowing for simultaneous spectral and spatial imaging.

Chapter

4

A Surface plasmon laser with two

arrays as cavity mirrors

∗

4.1

Introduction

When one is introduced to the subject of lasers in an introductory physics course, this invariably happens by discussing a collection of atoms (the gain medium) being brought to an excited state by an external light source or electric field (excitation pump) in between two mirrors that provide the feedback needed for coherent lasing emission (optical resonator). A nat-ural step in investigating the properties of surface plasmons by surface plasmon lasing is to mimic this geometry. This is done in the work de-scribed in this chapter.

Both the excitation pump (1064nm Nd:YAG laser) and gain medium (In0.53Ga0.47As gain layer) of the experimental setup have been discussed in previous chapters. Metal hole arrays were used to create effective SP mirrors. By actively pumping the area between two arrays, the backscat-tering of the SPs on these arrays provides the necessary feedback.

Scattering processes on similar arrays as the ones used in this chapter have extensively been discussed in reference [27]. In this publication, all relevant scattering rates have been determined with a coupled-mode de-scription. In a first order approximation the only supported backscattering mode in these arrays is backscattering under θ = 180◦. This supports the usage of square metal hole arrays as mirrors, because backscattered SPs are unlikely to scatter into spurious directions that lack constructive inter-ference. SPs that aren’t scattered back into the mirror resonator will either

12 A Surface plasmon laser with two arrays as cavity mirrors

couple out of the metal-dielectric interface or they will scatter further into the arrays through successive sideward (θ = ±90◦) or forward (θ = 0◦) scattering until they decay.

In the work presented in this chapter SP lasing has been investigated in this geometry by focusing on SPs traveling back and forth between two arrays with lattice spacing a0 = 450 nm (≈SP wavelength). The lattices are separated by a patch of 50 µm of unpatterned gold. A different e-beam dose was used for both arrays, leading to slightly different hole sizes (diameter ≈ 180 & 205 nm). The pump spot was increased in size from optimal focus to an approximate ellipse of 40 x 50 µm at FWHM. This size was chosen, so that the pump spot covered the entire central area up to the edges of both arrays.

4.2

Lasing characteristics

Results

When the area between two arrays is excited by the pump laser, optical emission is observed from the edges of the arrays. Measurements obtained with a polarizer in front of the CCD showed that the lasing emission was largely transversely polarized along the direction of SP propagation on the sample (perpendicular to the edges of the arrays).

The near-field emission spectrum, obtained at the edge of an array, is depicted for multiple total incident pump powers in figure 4.1a. Several peaks, spaced with a period of 5.2± 0.1 nm, appear when the total inci-dent pump power is increased above a threshold of around 230 mW. In the following, the focus will be on the lasing modes associated with the spectral peaks at 1480 and 1485 nm. This will be combined with the data obtained by the CCD behind a bandpass filter (λc = 1480 nm, FWHM = 12 nm, see transmission curve in fig 4.1a).

In fig. 4.1a a shift of the spectral peaks with increasing pump power can be observed. As the power is increased the peaks shift to slightly higher (and thus less energetic) wavelengths

The intensity of the peaks also varied abnormally with temperature for measurements ranging from 80-100 K. For a single measurement: at 80 and 90 K the mode at 1490 nm was of significant higher intensity than the others, whereas for measurements at 85 and 100 K the 1485 nm peak was most dominant, as was the peak at 1480 nm at 95 K. Although the inten-sity seemingly varied periodically with temperature, the behavior was not found to be repeatable and will not be taken into account in this section. 12

4.2 Lasing characteristics 13 120 140 160 180 200 220 240 260 280 300 0 5 10 15 In te n s it y ( a .u .) Pump Power (mW) BP 1480 nm 1485 nm 1480 nm I_Threshold 1 150 200 250 3000 5 10 15 Pump Power (mW) In te n s it y ( a .u .) 1430 1440 1450 1460 1470 1480 1490 1500 1510 1520 0,0 0,2 0,4 0,6 0,8 1,0 Wavelength (nm) In te n s it y ( a .u .) (b) 10 A B C S 0 2 4 6 8 10 12 14 16 18 300mW 275mW 259mW 242mW 181mW 120mW In te n s it y ( a .u .) (a) 150 200 250 300 In te n s it y ( a .u .) Pump Power (mW) (c)

Figure 4.1: (a) Emission spectrum observed at various pump powers (vertically shifted by 0.2 for readability). The black dashed curve depicts the transmission properties of the bandpass (BP) filter. (b) Spontaneous emission spectrum of a single hole array (other array gave identical results). (c) Input-output curves of measurements spectrally resolved at 1485 nm, 1480 nm and with the BP filter centered around 1480 nm. Output intensity is scaled to the intensity at Pthreshold =

230 mW. Inset shows linear input-output relation above threshold for the lasing

peak at 1485 nm.

All measurements discussed in this thesis were conducted at 100 K. In fig. 4.1b the spontaneous emission spectrum of a single array un-der excitation is depicted for comparison. The spectrum is decomposed into four Lorentzian resonances, as discussed in reference [27]. The lasing threshold for this array was P_in ≈200 mW. Both individual arrays exhib-ited similar emission spectra and lasing thresholds.

Figure 4.1c depicts the power dependency of the integrated intensity observed with the CCD and the intensity of the lasing peaks at 1480 and 1485 nm. The inset indicates the approximate linear P_in/Poutcurve of the lasing mode at 1485 nm above a threshold of P_in ≈ 230 mW. This mode also shows the steepest increase in intensity at high input power.

Discussion

The kink in the input/output curve in figure 4.1c around Pin =Pthreshold ≈ 230 mW, the subsequent steep linear increase and the spectral narrowing

14 A Surface plasmon laser with two arrays as cavity mirrors

above this threshold are typical characteristics of laser emission. The emis-sion above threshold for this device will therefore be referred to as such throughout this thesis.

The spectral peaks in figure 4.1a can be associated with the Fabry-Perot modes of the mirror cavity. The spectral separation of these modes is 5.2(1) nm. This is as expected for a Fabry-Perot cavity with an effective length

of L_{e f f} = 50µm + dR_pen+dL_pen = 54−57 µm. Where dR_pen and dL_pen are the

penetration depths of the SPs into the right and left array respectively. The group refractive index of SPs in the Fabry-Perot cavity can be cal-culated with

ngroup = λ

2

2L_{e f f}∆_λ (4.1)

where ∆λ is the spectral separation of the lasing modes and λ ≈ 1480 nm. This yields a group refractive index of ngroup = 3.82(11) for the circulating surface plasmons, comparable to ngroup ≈ 3.73 calculated for In0.53Ga0.47As around λ =1480 nm.

A comparison between the emission spectrum of the mirror-cavity laser and the spectrum of the arrays themselves (see figure 4.1a,b) shows that the lasing peaks are positioned around the C mode of the individual ar-rays, with tails towards the B and S mode. This spectral selection can be explained by considering the spectral separation of the A-mode and the group of the B, C and S mode as a DBR stopband. As was discussed in section 2.3, the wavelength dependent reflection amplitude R(λ)(see fig-ure 2.3) exhibits a maximum in the stopband and decreases significantly at the edges (see figure 2.3). The structure will start lasing when the round-trip loss, associated with a low value of R(λ), is compensated by the gain g(λ) supplied by the InGaAs-layer. The reason why this geometry isn’t lasing in the middle of the stopband, where the reflectivity is maximal, is not yet fully understood.

4.3

Direct Imaging

Results

Figure 4.2a shows a direct image of the emission on the air side of the las-ing device at Pin ≈300 mW. The FWHM of the excitation spot is indicated by the red dashed line. The intensity of the SP field on the Au-InGaAs interface is expected to be maximal in the area between the arrays, but is not visible through the 100 nm gold layer on top.

14

4.3 Direct Imaging 15

In figure 4.2b, the emission intensity (black) and the intensity of the pump spot (red dashed) along the central green line are shown. The emis-sion intensity decays approximately exponentially into the arrays, as is further evidenced by the log-plots in figure 4.2c. Measurements of the pen-etration depth into the array with the CCD yielded values of dL_pen =3.9(6) µm and dR_pen = 3.2(5) µm for the left and right array respectively. Mea-surements with the scanning-fiber, spectrally resolved at 1480 and 1485 nm resulted in comparable values of dL_pen = 2.3(3) µm and dR_pen = 1.9(2) µm. The obtained penetration depth did not notably vary with input power and remained approximately unchanged below lasing threshold.

Discussion

In this experiment, the penetration depth is the parameter that contains the most information about the propagation and scattering of SPs. The values for dL_pen were consistently (across all measurements) ∼ 20% larger than those for dR_pen. This suggests higher in-plane scattering rates for the right array, which is consistent with the fact that a larger e-beam dose was used for this array, leading to larger hole sizes.

The fact that dpenhardly varied with input power, shows that scattering processes, rather than absorption, are the main cause of SP losses. In turn, this further supports the importance of R(λ) in explaining the spectral

-80 -60 -40 -20 0 20 40 60 80 In te n s it y ( a .u .) Position (µm) 50µm 2 4 6 8 10 12 0 0 2 4 6 8 10 e-1 e0 e1 1480 ± 12 nm (R) 1480 ± 12 nm (L) 1480 nm 1484.5 nm 3.26µm dpen=2.24µm 4.03µm 4.23µm 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 In te n s it y ( a rb . u n it s ) Left Array Right Array 1484.5 nm 1480 nm

Depth Into Array (µm)

-150 -100 -50 0 50 100 150 200 X Axis Title (a) (b) SPs (b) 0 1 2 3 4 5 6 7 0.1 1 1480 ± 12 nm (R) 1480 ± 12 nm (L) 1480 nm 1485 nm In te n s it y ( a .u .)

Depth into array (µm)

-80 -60 -40 -20 0 20 40 60 80 In te n s it y ( a .u .) Position (µm) 50µm 2 4 6 8 10 12 0 Perimeter of Arrays Excitation Spot (FWHM) (c)

Figure 4.2: (a) Direct image of emission observed by the CCD. (b) Intensity of emission (black) and excitation spot (red dashed) along the central green line in (a). (c) Log-plot of the intensity, indicating the exponential decay into the arrays.

16 A Surface plasmon laser with two arrays as cavity mirrors

selection of the system.

The reason for the lower penetration depths for the measurements with the scanning-fiber is as of yet unknown. A possibility might be the fact that for measurements obtained with a CCD behind a BP filter, the outcoupling of SPs that did not traverse the FP-cavity is more dominant and broadens the observed decay. When the measurement is spectrally filtered around a single lasing peak, this contribution can most likely be neglected.

A simple calculation, based on a 1D model, shows that the scattering rate γ (units s−1) can be determined for SP plane waves incident on the arrays (visualized as subsequent rows of holes) by

γ/ω₀ = a0

2πdpen =0.018−0.038 (4.2)

where γ/ω0 is the normalized dimensionless scattering rate. These values are larger than those (γ/ω₀ = 0.013 - 0.017) deduced from disper-sion of SPs in single metal hole arrays [28]. This difference can most likely be explained by considering the fact that this scattering rate doesn’t just take direct backscattering into account, but also contributions from suc-cessive sideward scattering in the array that result in SPs being scattered back into the cavity. The values quoted for the dispersion measurements only take direct backscattering into account.

4.4

Angle-Resolved Imaging

Results

The far-field pattern, as observed by the CCD, is depicted in figure 4.3. Co-herent lasing emission from the edges of two arrays, separated by 50 µm, naturally leads to a Young’s interference pattern. The interference pattern measured with the CCD can be fitted with a periodicity of 28 ± 2 mrad and a Lorentzian envelope of FWHM = 140 ± 5 mrad. The interference pattern obtained by the scanning-fiber showed the same periodicity, but larger envelope width of around 170 ±20 mrad for measurements spec-trally resolved both at 1480 and 1485 nm. Additionally, a 180◦phase shift is apparent in the interference pattern at λ = 1480 nm, as compared to the 1485 nm and CCD measurements. The visibility (V) of the oscillations increased significantly when a single-mode fiber was used, with V≈1 for a large part of the pattern (except for a slight discrepancy in the baseline on the left side of the pattern).

16

4.4 Angle-Resolved Imaging 17

The interference pattern is modulated in the vertical direction by a low-intensity dark band along θy ≈0. The presence of this band is further in-dicated by the projected profile on the right. The dark band was present in each measurement, but the profile itself varied periodically with wave-length. As the wavelength increased from 1480 to 1485 nm, the two high-intensity peaks gradually move inward and slowly lose high-intensity until at 1482.5 nm they both vanish at θy ≈ 0. Simultaneously, two new high in-tensity peaks move in from wider angles and arrive at the same θy at the next lasing peak of 1485 nm.

-250 -200 -150 -100 -50 0 50 100 150 200 250 0,10,4 0,71,0 A n g le ( m ra d ) Intensity (a.u.) -250 -200 -150 -100 -50 0 50 100 150 200 250 0,1 0,4 0,7 1,0

A

n

g

le

(

m

ra

d

)

Intensity (a.u.)

In

te

n

s

it

y

(

a

.u

.)

Angle (mrad)

Figure 4.3: Far-field (angular-resolved) emission from structure shown in figure 4.2. The intensity of the interference pattern along the horizontal green solid line is plotted in blue (baseline shifted by 0.6). Spectrally-resolved measurements at λ= 1480 nm (black) and 1485 nm (red) gave similar results. The righthand figure shows the integrated intensity of the pattern along the horizontal axis

18 A Surface plasmon laser with two arrays as cavity mirrors

Discussion

By Fourier transforming two exponentially decaying intensities, 50 µm apart, the effective spacing between the emission regions can be related to the periodicity of the far-field pattern by ∆θ = λ0/Le f f. Where the effective spacing is L_{e f f} = 50 µm + dR_pen +dL_pen. The observed period-icity ∆θ = 28(2) mrad corresponds to L_{e f f} = 53(3) µm. Similarly, the Lorentzian envelope can be used as an alternative estimate of the pene-tration depth via dpen = λ0/(π· FWHM) = 3.3(2) µm for FWHM = 140 ±5 mrad. This compares favorably with the dpen as observed with direct imaging with the CCD.

The dark band that modulates the radiation pattern probably origi-nates from the 2D nature of the reflection gratings, which differs from an ordinary 1D DBR model. The eigenmodes of these arrays (which are effectively 2D plasmonic crystals) can be divided into radiative and non-radiative modes. It has been shown that the non-non-radiative modes are re-sponsible for lasing in individual arrays and that the symmetry of these modes prevents emission along the surface normal, which results in a donut-shaped lasing profile [13]. Similar results might be expected for back-reflection from a 2D DBR structure.

The periodically varying intensity profile could be related to the fact that different transverse wavevectors correspond to different emission an-gles. As the longitudinal momentum is fixed by the roundtrip condition between the arrays, this will result in different total wavevectors. This cor-rectly predicts the periodicity and the larger transverse angles observed for blue-detuned emission.

4.5

Conclusion

Surface plasmon lasing has been observed in a Fabry-Perot type mirror-cavity where metal hole arrays were used as effective SP mirrors. Lasing in this geometry has been characterized extensively. By considering the DBR and Fabry-Perot type characteristics of this geometry, an effective scatter-ing rate for SPs impedscatter-ing on a metal hole array and the group velocity of SPs have been deduced. Several explanations for intriguing phenom-ena, such as the nature of the dark band in the far-field radiation pattern and the spectral selection of the lasing modes, remain to be verified. Nev-ertheless this technique promises to be a viable tool for visualizing and characterizing SP propagation on any metal-dielectric interface.

18

Chapter

5

Surface plasmon scattering in

elliptical-hole arrays

5.1

Introduction

Chapter 4 of this work has already alluded to the physics of SP action in structured arrays on metal-dielectric interfaces. Such arrays are usually composed of either sub-wavelength sized holes on the metallic side or of metallic nanoparticles placed on top of the dielectric [13, 17, 27, 29, 30]. A natural first step in describing the properties of SPs propagating in these structures, is to vary the array parameters (hole size, lattice spacing) or the nature (read: materials used) of the interface itself. In previous years the effect of varying the array parameters has been investigated extensively in this group and the work is documented in references [13, 15, 27, 28].

In this chapter, measurements on a hole array that is composed of el-liptical, rather than circular, sub-wavelength sized holes will be described. The motivation for this work lies in the understanding of SP propagation in such arrays, which is determined to a large extent by their scattering rates. The scattering of the SPs in such arrays can be directly related to the dispersion of the SP-mediated emission via a simple model described in reference [27]. This chapter extends that work to dispersion measure-ments of emission by an actively pumped elliptical-hole array on the gold-semiconductor interface that was described in chapter 3.

SP lasing has been observed in these arrays for sufficiently high excita-tion powers. This chapter, however, will put emphasis on measurements that were performed below lasing threshold. This choice was made be-cause laser emission tends to saturate the detectors, which complicates straightforward detection of dispersion characteristics.

20 Surface plasmon scattering in elliptical-hole arrays

5.2

Sample

Figure 5.1: Detailed SEM image of part of angled elliptical-hole array under in-vestigation (ad4).

The layer structure, array param-eters and hole shapes of the vice under investigation were de-scribed in the overview in

chap-ter 3 of this thesis. All

mea-surements described below have been performed on an array that was named ad4 on wafer RA2389-I. This (50×50 µm2_{) array was} cho-sen after inspection with a Scan-ning Electron Microscope (SEM) showed that it was printed most successfully. SEM imaging further showed the exact lattice spacing of this array to be a0 = 441 nm. The hole diameters were found to be

345 nm × 250 nm (long axis × short axis). The ellipses were tilted in-plane at an angle of 52◦ relative to the horizontal axis. A SEM image of part of this array is depicted in figure 5.1. The sample was excited by a circular pump spot with a FWHM diameter of approximately 30 µm. The excitation power was 150 mW, which was below lasing threshold for this particular array.

5.3

Dispersion Measurements

Possible in-plane scattering processes and corresponding scattering rates for SPs impeding on elliptical holes are depicted in figure 5.2b. In reference [27] similar scattering processes were experimentally and theoretically de-scribed for circular holes. This theory was based on a scattering matrix description of the time evolution of the out-of-plane electric field. As cir-cular holes don’t break left/right symmetry (from the point of view of the incoming SP), the sideward scattering rate in both directions was absorbed into a single parameter κ. For elliptical holes that are tilted with an angle θ 6= 0◦, 90◦ relative to the path of the incoming SP, this symmetry con-sideration breaks down. The breaking of this symmetry can theoretically be accounted for by describing left/right sideward scattering by different rates κ/µ respectively. Using this slight adaptation of the model from [27] the measurements depicted below can be used to determine the SP scat-tering rates in elliptical hole arrays.

20

5.3 Dispersion Measurements 21

Figure 5.2a shows the measured equi-energy planes of the dispersion relation. This data was obtained by making a spectrally-resolved 2D far-field image of the SP-mediated sam-ple emission and plotting this data at a fixed wavelength (i.e. frequency). The plots that are shown were chosen to show most clearly the evolution of the dispersion bands with a change in wavelength (indicated in white for each plot).

Even without employing the coupled mode model, the absence of four-fold rotation symmetry as well as inversion (±x,±y) sym-metry of the dispersion bands, that was present for circular holes, can be observed. Mirror sym-metry along the diagonals is still somewhat vis-ible, but not perfect. Rather than along 45◦, the mirror plane seemingly lies along 52◦, corre-sponding to the in-plane rotation of the ellipses (see figure 5.2, where the ori¨entation of the el-lipse relative to the dispersion plots is exactly as in the exp. setup). The diagonal symmetry is also altered by the increased intensity of the bands in the lower right corner.

Figure 5.2:(a) Angle-resolved false color 2D dispersion plots of elliptical-hole ar-ray emission at specific wavelengths. Equal x,y,z-scale for all plots. (b) Overview of in-plane scattering processes and rates (κ, γ, µ) for SPs impeding on an ellip-tical hole. Ori¨entation of ellipse compared to the plots in (a) as in experimental setup.

22 Surface plasmon scattering in elliptical-hole arrays Figure 5.3 shows false color plots of the intensity of the sample emission as a function of wavelength in the far-field along four directions θ. The directions that are shown correspond to the x and y axis (0◦and 90◦resp.) and the two diagonals (45◦, 135◦). The (vacuum) wavelength can be re-lated to the frequency by(λ=2πc/ω). This means that these graphs are effectively ω(k) dispersion diagrams (usually depicted rotated clockwise by 90 degrees).

From the same symmetry considerations as described above the ob-served mirror symmetry (or lack thereof) along the “angle = 0◦-axis” for all four plots in figure 5.3 is expected. The plots for θ = 0◦ and θ = 90◦ are expected to be identical for ellipses oriented at 45◦. The fact that these are very similar but not fully identical could be due to the orientation of 52◦ of the ellipses in array ad4.

The clearest example of the effect of the ellipticity of the holes can be seen in comparing the plots along the diagonals θ =45◦and θ=135◦. For circular holes, these plots are identical (see figure 2 in [27]). Elliptical holes, however, break this symmetry and lead to two intriguing structures. The cut along θ = 45◦ shows a bandgap in the centre, bordered by two cross-ings. The cut along θ =135◦, however, shows a ’graphene-like’ dispersion diagram, with a crossing in the centre between two (almost) linear disper-sion bands.

As a final disclaimer it is noted that the 2D dispersion measurements (depicted in figure 5.2) were not centered perfectly. By this it is meant that the (0,0) coordinate does not exactly align with the direction of the surface normal. The cuts that produced figure 5.3 do assume a perfectly centered measurement. This means that the diagonal cuts did not cut through the exact centre of the structures depicted in figure 5.2.

Figure 5.3: Angle-resolved false color 1D dispersion plots of elliptical-hole array emission along directions specified in white. Directions as depicted in figure 5.2. 22

5.3 Dispersion Measurements 23

5.4

Discussion

The data presented in this chapter has not yet been fully analyzed. How-ever, the symmetry breaking, that occurs when an array of elliptical rather than circular holes is excited, can be observed. All observed symmetries can be understood and derive from the in-plane orientation of the ellipses. Further fitting to the coupled-mode description should result in the de-termination of the appropriate scattering rates and the phase and group refractive indices of SPs propagating on such interfaces.

Analysis of previous measurements performed by Michel Hubert [31] on the same array have resulted in surprising conclusions, namely µ > κ ≈γ. This conflicts with what is expected for a single small elliptical hole. Emission from such holes can be modeled by emission from two induced dipoles: an out-of-plane electric dipole and an in-plane magnetic dipole [32]. By modeling an elliptical hole with an anistropy in the magnetic polarizability and assuming similar phase relations between the dipoles as for small circular holes it can be deduced that the dominant scatter-ing rates should be κ ≈ γ. On the other hand, µ is expected to be much smaller.

A possible explanation for this discrepancy could be a deviation from the small-hole approximation, as the elliptical holes in this array are rel-atively large. Another important consideration is that the dipole model only describes scattering on a single hole. The presence of neighboring holes and scattering of SPs from these holes leads to local-field effects that are not taken into account by the single-hole dipole model.

One important reason, however, for the work described in this chapter, which is basically repeating those earlier measurements, is that there exists some doubt about earlier conclusions [31]. Not only do earlier dispersion plots lack the clarity of the band structures depicted in this chapter, but the exact orientation of the ellipses compared to the other parts of the exper-imental setup was not recorded properly. This raises questions about the left/right assignments and especially about the assignments of the scatter-ing rates to µ and κ. It is therefore possible that the observation µ >κwas actually false and should be changed to κ > µ, as they might have been interchanged. Care has been taken to avoid such confusions in the work presented here.

5.5

Conclusion

In conclusion, measurements of the dispersion of SP mediated emission of a square array with tilted elliptical holes have been performed. The presence of elliptically shaped holes clearly affects the symmetry of the far-field radiation pattern and promises to hold more information about the scattering rates and phase and group index of SPs propagating in such arrays.

Chapter

6

Characterization of newly

fabricated device: RA3586

Over the course of the project described in this thesis, a new device was produced. This was done in order to improve on the measurements on elliptical-hole arrays described in chapter 5 and investigate SP-induced dispersion characteristics of light emanating from circular-hole arrays with a larger lattice spacing. In this chapter the corresponding production pro-cess as well as the transmission and emission characteristics of this device will be presented and discussed. The device will be referred to as ’RA3586’ throughout this chapter, in accordance with the naming by the manufac-turer.

6.1

Production

The layer structure of RA3586 was chosen identically to the structure de-picted in figure 3.1. Thus the InGaAs gain layer had a thickness of 130 nm, rather than the 150 nm for the device discussed in chapters 4 & 5.

The epitaxial growth of the InP/In0.53Ga0.47As/InP layer structure of RA3586 was performed by P.J. van Veldhoven in the cleanroom of the Eindhoven University of Technology (TU/e). Contrary to previous de-vices, this wafer was subsequently handed over to nanoPHAB, a micro,-and nanofabrication company, based at the TU/e.

NanoPHAB etched away part of the InP-cap on top of the InGaAs layer and produced a thin (∼5 nm) SiN separator, Chrome sticking layer and the patterned gold on top. The pattern was fabricated by depositing ap-propriate nanopillars (400nm HPR504 resist, capped by 80 nm HSQ) be-fore gold deposition. The pillars were washed away at the end of the pro-duction process. A recipe of the lithography is presented on the next page.

26 Characterization of newly fabricated device: RA3586

Recipe metal hole arrays on InGaAs gain layer

Wafer: InP substrate, InGaAs gain layer (150 nm), InP capping layer (20nm).

0) The wafer for the dummies has a capping layer of 20nm. I remove part of this by doing a O2 stripper (5 minutes, 300 W) and then a etch with diluted phosphoric acid

(1:10). This is repeated 4 times, depending on the available time. 1) Create 5 nm SiN

a. SiN PECVD, 50nm waveguide recipe, aborted after 17seconds. 2) Make resist layers

a. Spin HPR 504 on gyroset, 5500 rpm, 30s, 50 accl

(thinner layers might reduce problems with falling pillars) b. Bake 2 min 150 oC

c. Bake 2 min 220 o_C

d. Bake 2 min 250 oC (Oven Heraeus) e. Cool down, 10 minutes

f. Spin 6% HSQ convac, 4500 rpm, 60s

(thinner layers might reduce problems with falling pillars) g. Bake 2 min 150 o_C

h. Bake 2 min 220 oC

3) 7.5 nm gold deposition, small strip on the wafer, for positioning in lithography 4) E-beam lithography exposure

a. Make some small scratches on edge with gold, such that there is something to see with E-beam.

b. Dosefactor test on small edge of sample. Depends on resist age. c. 10 µm aperture, 30 kV,100 µA/cm2, DoseFactor =17.5

5) Etch gold: KI/KI2 30 seconds

6) Development:

a. 2 minutes MAD 531s at 60o_C

b. Wash carefully for 5 minutes in water (be careful, rolling droplets might displace pattern)

c. Dry softly

d. Reactive Ion Etch: 7 minutes PR 50W. Settings: O2 20 sccm, pressure 15

mTorr, power 50 W.

7) Wash in water and dry (be careful, rolling droplets might topple over pillars) 8) Evaporation on FC2000

a. 0.5 nm Cr, 100 nm au, 20 nm Cr: For sticking layer: close shutter by clicking close button ASAP it says open (fast reactiontime is required) Previously Ti was used instead of Cr, this has been changed for the latest wafers.

b. Simple test for sticking layer on test sample: After evaporation of gold layer put in ultrasonic bath for 5 min. If golds falls off: thicker sticking layer.

9) Lift off

a. Stripper 5 min 300 W

b. Acetone, in ultrasonic bath (20 minutes) c. Rinse Isopropanol

d. Rinse water

26

6.2 Characterization 27

The mask files that are used by the e-beam lithographer in fabricating the nanopillar arrays were created by a simple program, written in Python, that is described in more detail in appendix A2.

In order to improve on the results described in chapter 5, square arrays of elliptical holes that are collectively rotated in-plane by 0, 30, 45, 52 and 90 degrees were created. The choices for 0, 30, 45 and 90 degrees were made, because these choices allow for a more extensive inspection of the effect of symmetry breaking than was done in chapter 5. The choice of 52 degrees was made in order to accommodate direct comparison of the results obtained with this device with previous results. The lattice separa-tion was defined by q·a0, where q∈ {1, 2, 3}and a0≈SP wavelength. The separation was varied by varying both a0=450, 460 or 470 nm and q =1, 2 or 3. The short axis of the elliptical holes was chosen to be of length 100 nm and the long axis was 200 nm. A successfully printed elliptical-hole array can be seen in figure 6.1b. A map of a typical block of elliptical-hole arrays is depicted in figure B.1a in appendix B.

In addition to the elliptical-hole patterns, square circular-hole arrays with a lattice spacing q ≥1 were fabricated. The lattice spacing for these arrays was similarly varied as for the elliptical hole arrays (a0 =450, 460, 470 nm, q=1, 2, 3). This lattice spacing was equal for both the horizontal and the vertical direction. For the circular holes, the hole-radius was also var-ied from 50 nm to 100 nm, in steps of 10 nm. Therefore the arrays were grouped in three blocks of six (for a single choice of q), which makes them directly distinguishable from the elliptical hole arrays that were grouped in blocks of five. A successfully printed circular-hole array can be seen in figure 6.1a. A map of a typical group of circular-hole arrays on RA3586 is depicted in figure B.1b.

In order to improve the chance of success, the patterns were fabricated 27 times, each time with a different e-beam dose and bias. An overview of the location of these arrays on the wafer and the parameters that were used during fabrication can be found in figure B.2 in appendix B. A method of ordering and labeling the enormous amount of arrays, 2673, on this wafer is also presented in the same appendix.

6.2

Characterization

OM and SEM imaging

Before the newly fabricated device RA3586 was investigated with the ex-perimental setup described in chapter 3, it was inspected with a Scanning

28 Characterization of newly fabricated device: RA3586

Figure 6.1:Typical images of successfully printed hole arrays. (a) and (b) are SEM images of circular and elliptical hole arrays respectively. (c) is a 5x OM image of block F1. The location of the 100x OM image (d), which shows a hint of the array lines when magnified, is indicated by the purple circle.

Electron Microscope (SEM) by NanoPHAB and an Optical Microscope (OM) in the cleanroom at LION. Figure 6.1 depicts typical SEM and OM images of successfully printed elliptical and circular hole arrays. The lo-cation and exact nature of the arrays depicted in the SEM images (figure 6.1a,b) was not recorded. OM imaging indicated that q >1 arrays were printed successfully across the entire wafer. Figure 6.1c shows an OM im-age of block F1 (named according to the labeling system described in ap-pendix B). Block F1, together with neighboring blocks E1 and G1, provided the best printed groups of both circular and elliptical q=1 arrays. Therefore these arrays were used for further characterization of the device (see next sections). Figure 6.1d shows an OM image of the array Fe1 q1a450Rot30, one of the arrays with the least amount of defects on the entire wafer.

In figure 6.2 several badly printed q=1 arrays on the wafer are depicted for a wide variety of e-beam settings (and thus locations on the wafer). The exact nature of these ’misprints’ has been determined for arrays that look similar to the array framed in red (entire gold-layer is stripped) and in blue (too much leftover polymer). Although the variation in misprints is some-28

6.2 Characterization 29

Figure 6.2: Overview of several different groups of failed hole arrays and their location on the wafer RA3586.

what unpredictable, it has been observed that arrays that showed a layer of leftover polymer tended to be arrays of small hole size. Conversely, the gold layer was generally stripped completely for arrays of larger hole size that were printed with a relatively high e-beam dose and bias.

Transmission

As a method of investigating the performance of the wafer RA3586, trans-mission measurements were performed with the 1064 nm pump light. These measurements were compared to similar measurements on an ’old’ wafer RA2389-I (investigated in chapter 4 & 5). This wafer functioned as expected, contained an a0 = 450 nm, q=1, R = 100 nm circular hole array (sd4) and had a similar layer structure (except for the thickness of the gain layer). Because of these similarities RA2389-I serves as a good reference.

Figure 6.3: The scratch - located between block B3 and C3 - that was used for transmission mea-surements on RA3586. Neigh-boring arrays are stripped of the gold layer as well. Pump focus depicted in red.

The transmission measurements on RA3586 were performed on three different locations. First, the transmission of a suc-cessfully printed circular hole array (a0 = 450 nm, q=1, R= 100 nm), that was named Fc1 q1a450R100, was determined. This al-lows for direct inspection of the perfor-mance of a single array. Second, the laser was focused on the scratch depicted in fig-ure 6.3. This enables transmission mea-surements of the InP/InGaAs/InP-wafer, without the gold layer on top. As a third means of investigation, the transmission through all layers was determined by po-sitioning the laser focus on an unpatterned patch of gold.

de-30 Characterization of newly fabricated device: RA3586

Table 6.1: Results of transmission measurements of the cryostat(windows) and the wafers RA2389-I and RA3586. All transmission coefficients depicted are rela-tive to the input pump power.

picted in table 6.1. This table depicts the power ratio (Pout/Pin) of the out-going and incoming 1064 nm pump light. The transmission through an empty cryostat - and thus for light passing only through the two cryostat windows - was also recorded and found to be 0.77(4). No scratch was present on the gold layer of device RA2389-I and thus this transmission measurement could not be performed. The transmission of the identical arrays (sd4, Fc1 q1a450R100) on both wafers yielded similar results of Pout/Pin = 3−4×10−3. Measurements through an unpatterned patch of gold show that the transmission through the gold layer is a factor of 10 lower for RA3586.

Emission

As another method of characterizing the device, the emission intensity and spectrum of RA3586 has been investigated and compared to that of RA2389-I. Using the infrared CCD camera, the maximum count rate of the fluorescence was compared to that of the transmitted 1064 nm pump-light. This transmission could subsequently be related to the intensity of the incoming pump via table 6.1. During measurements of the emitted fluorescence, the 1064 nm pump light was filtered out by placing two 1100 nm long-pass filters between the sample and the detectors.

The fraction Pemission

Pin , where Pinis the incoming pump power, was

found to be 1.8×10−6 _{for RA3586 and 3.8×10}−4 _{for RA2389-I when the} laser was focused on identical arrays. This means that the emission of the device RA2389-I was approximately 200 times more intense than that of RA3586.

For RA3586, the fluorescence through an array for which the gold layer has been completely removed was 38 times more intense than that of a patterned array where the holes encompass 15.5% of the total area.

In order to confirm the apparent lack of intensity of the fluorescence of RA3586, its emission spectrum was inspected with a grating spectrome-ter. The results for a fixed-point near-field spectrum of emission from the 30

6.3 Discussion 31 (c) (a) 14000 1450 1500 1550 1600 1650 100 200 300 400 500 600 700 In te n s it y ( C o u n ts /s e c ) Wavelength (nm) 340 mW 200 mW 100 mW RA3586 (array) 14000 1450 1500 1550 1600 1650 10000 20000 30000 40000 In te n s it y ( C o u n ts /s e c ) Wavelength (nm) 300 mW 200 mW 100 mW 25 mW RA2389-I (array) 14000 1450 1500 1550 1600 1650 2000 4000 6000 8000 10000 In te n s it y ( C o u n ts /s e c ) Wavelength (nm) 340 mW 100 mW 25 mW RA3586 (scratch) (a) (b)

Figure 6.4: Fixed-point near-field spectrum of emission of (a) RA2389-I and (b,c) RA3586. Array sd4 was used in obtaining the spectrum of emission through an array in (a). ARray Fc1 q1a450R100 was used for (b). (c) depicts the emission spectrum of RA3586 through the scratch, depicted in figure 6.3.

scratch and a successfully printed array are depicted in figure 6.4b (array) and 6.4c (scratch). The location of the fixed-point was optimized as to give the clearest and highest intensity spectrum.

For comparison, the same fixed-point near-field measurement was per-formed on array sd4 on RA2389-I. The results of this measurement are shown in figure 6.4a.

6.3

Discussion

The results of the transmission measurements indicate that the gold layer on device RA3586 is thicker than the layer for RA2389-I. For an exponen-tial decay of the field intensity into the gold, the reduction of the transmis-sion by a factor of 10 can be related to an increase in thickness of around 25 nm (assuming a thickness of 100 nm for the gold on RA2389-I).

32 Characterization of newly fabricated device: RA3586

Both the transmission measurements of an empty cryostat as well as the measurements through the arrays on both wafers show no other sur-prising results or discrepancies.

Intensity measurements of the sample emission showed that the inten-sity of the emission from RA3586 is 200 times less than that for RA2389-I.

When comparing the spectra for both devices, three observations stand out the most. First, the decrease in intensity (counts/sec) of the fluores-cence of RA3586 is again apparent. By determining the peak intensity of the spectrum through Fc1 q1a450R100 on RA3586 we find that this is approximately 50 times lower than that of sd4 on RA2389-I at a similar power. Even the intensity of the spectrum through the scratch on RA3586 is still a factor of 2-4 smaller than the intensity through the array sd4. The presence of the lasing peak at 300 mW most likely adds to the discrepancy between the observation of 200 times less intense emission from RA3586 with the Pemission/Pinmeasurement and that of 50 times less intense fluo-rescence from the spectral measurements.

A second noticeable aspect for the emission from the array on RA3586 is the lack of spectral structure (see figure 6.4b). Although it is still spec-trally broad, the spectrum for RA2389-I shows several peaks (around 1450 and 1500 nm) and even a 1430 nm lasing peak at P_in =300 mW. This is an indicator of SP action and stems from the standing wave SP intensity pro-file in the array. The spectrum for the emission from the array on RA3586 is much broader and does not exhibit similar features. Spectrally resolved measurements of its far-field intensity profile further showed the lack of dispersion bands that have been observed and discussed for RA2389-I in chapter 5 of this thesis (see figure 4.1b). Furthermore, no lasing was ob-served at even the highest input powers possible with this experimental setup (∼340 mW).

Thirdly, we can observe a shift of the emission spectrum to higher wavelengths for RA3586, as compared to RA2389-I.

The absence of far-field dispersion bands, structure in the emission spectrum and lasing at high input power all point to a lack of SP action on the device RA3586. This observation holds for all arrays, including those that did not show any defects when inspected optically. SPs on the gold/InGaAs-interface of the device are excited by the fluorescence of the pumped InGaAs gain layer. This means that the observed low intensity of the gain layer fluorescence could contribute to the apparent absence of ”long-range” SP propagation and coherent SP scattering.

32

6.3 Discussion 33

Three possible explanations for the lack of observed gain layer fluores-cence might be that

• the Chromium sticking layer is too thick. If this layer is too thick, the EM-field intensity could be damped too much and SPs in the gold layer won’t be excited. However, the manufacturer (NanoPHAB) has estimated the maximum thickness of the Cr-layer to be 1.5 nm. This is much less than the calculated penetration depth of the EM-field profile. Therefore, it is unlikely that the EM-EM-field is damped too much for SPs to be excited in the gold layer.

• the InP/SiN-spacer between the gain layer and the metal (Cr/Au) is too thin. This could cause optical quenching of the fluorescence to occur at the metal-semiconductor interface. However, the recipe that was used to etch away part of this spacer was used successfully multiple times in the past. Furthermore, the fluorescence measured at the scratch where all gold and possibly chromium is removed -was also strongly reduced on device RA3586. The lack of metal on top means that no optical quenching can occur in this area. Thus it seems equally unlikely that optical quenching is the underlying cause for the observed low intensity fluorescence.

• the In0.53Ga0.47As gain layer is defective or too thin. A thinner or defective gain layer would naturally lead to less fluorescence. Al-though the gain layer thickness was decreased from 150 nm (for RA 2389-I) to 130 nm, a thickness of 130 nm has proven success-ful for identical devices in the past. It might be possible that the InGaAs-layer was damaged by the electron beam during the lithog-raphy steps, as the usage of a new e-beam lithographer might have led to a more destructive electron beam than expected. Van Veld-hoven (the producer of the underlying InP/InGaAs/InP-wafer) has experienced no problems in production around the time this wafer was produced. X-ray diffraction measurements, performed before the lithography, showed no apparent defects in the crystal structure of the wafer. Van Veldhoven suggested the possibility of Rapid Ther-mal Annealing (RTA) of the wafer as a means of improving its func-tionality. However, care has to be taken as to not deteriorate the metallic side of the device.

Further possible explanations could be related to the experimental setup and sample contamination. The setup that was used for RA2389-I and RA3586 was identical. However, it was noticed that on remounting RA2389-I it exhibited higher lasing thresholds and less structured spectral profiles

34 Characterization of newly fabricated device: RA3586

than before. This could suggest a change in the experimental setup and/or method of mounting the sample. However, aging of the device is also a vi-able explanation. Also, some apiezon drops on top of device RA3586 can be seen on optical microscope images. Although these were not present on the arrays that were investigated in this chapter, they might suggest the presence of a thin apiezon layer stretching and affecting emission across the sample.

6.4

Conclusion

In conclusion, a device has been fabricated with the goal of investigating SP scattering in elliptical-hole and q > 1 circular-hole arrays. The device has been inspected optically and showed numerous successfully printed arrays. However, after further characterization, no SP action has been ob-served. The exact cause of this has been investigated but is yet to be deter-mined. The observed strong reduction in fluorescence is likely to have the same cause.

34

Chapter

7

Summary & Outlook

The main focus of this thesis lies on the work described in chapters 4, 5 & 6. Chapters 4 and 5 contain information that could be useful for the entire field studying the properties of SPs. Chapter 6, in combination with the appendices, was written in order to serve as a useful reference guide for further studies and sample production within the research group.

In chapter 4 a new form of SP lasing was introduced. A SP mirror cavity was created by actively pumping the area between two metal hole arrays on a gold/In₅₃Ga₄₇As interface. Direct imaging of this structure showed lasing emission emanating from the edges of the arrays. The las-ing characteristics (laslas-ing threshold, wavelength spectrum) of this SP laser have been determined. Its spectral characteristics were explained by con-sidering the Fabry-Perot interferometer nature of this cavity in combina-tion with a DBR-type descripcombina-tion of the reflectivity of the metal hole ar-rays. Although speculation is made about the origin of the intriguing far-field radiation pattern (with a dark band at θy ≈0 for all wavelengths), a conclusive explanation is yet to be found.

Naturally, a future challenge lies in the understanding of the dark band in the ffield emission. This could be investigated by pumping both ar-rays with a more stretched elliptical pump spot that itself penetrates far into the arrays. Typical dimensions of such a pump spot for the geome-try of chapter 4 would be an elliptical spot of size FWHM = 100 x 50 µm. By increasing the pump spot size, however, the lasing threshold is also increased. This has to be accounted for by increasing the lasing pump power.

A further possibility to add to the experiments described in chapter 4 is to model the observed effective scattering rate for SPs impeding on a metal hole array. Decomposing the total reflection as an addition of individual

36 Summary & Outlook

scattering processes could result in a model that provides (estimations of) individual scattering rates. Further research might also benefit from more exact determination and imaging of the penetration depth.

In chapter 5 the dispersion relation for SPs propagating in a metal hole array of elliptically shaped holes was investigated. Intriguing dispersion bands were presented that clearly showed the effect of breaking the sym-metry of SP scattering. The dispersion bands have not yet been fitted to the coupled-mode model [27] that determines the in-plane scattering rates of SPs from this data.

A logical future continuation of this experiment is to actually fit the obtained data to the coupled-mode model. Hopefully this will result in a determination of the in-plane SP scattering rates. This would then lead to a final conclusion on the preferred direction of scattering of SPs on elliptical holes. Care was taken in this experiment to record the orientation of the el-lipses relative to their dispersion diagrams. This should remove previous confusion of the determination of the leftward and rightward scattering rates.

Additionally, arrays could be investigated that contain ellipses that are not only rotated 52 degrees, but rather 0, 30, 45 and 90 degrees. These orientations allow for further investigation of the breaking of SP scattering symmetry. These arrays were, of course, already printed on the device that was described in chapter 6. However, the poor performance of this device suggests that further sample fabrication might be necessary in the future.

In chapter 6 the production process as well as the characterization of a newly fabricated device is presented. An overview of all steps in pro-duction as well as a map of the sample is given. The characterization of the sample showed that it did not perform as desired. Possible causes are mentioned but no definitive conclusion is made.

Further consultation with the manufacturers of the device (P.J. van Veldhoven of the TU/e and Francesco Pagliano of NanoPHAB) will hope-fully result in more insight into the apparent defectiveness of device ’RA3586’. In turn, this will shed light on possible improvements to be made in future sample production.

All in all, the presented results in this thesis have shown many inter-esting aspects of the propagation of SPs on structured metal-dielectric in-terfaces. They showcase both an intriguing and - as of yet - unexplored form of SP lasing as well as interesting dispersion diagrams for SPs prop-agating in elliptical-hole arrays. Even though much has been learned and understood from these measurements, the results promise to hold even more information than that what has already been extracted in this thesis.

36