MSc Chemistry
Science for Energy and
Sustainability
Master Thesis
Electro-deposited indium for
nanophotonics
by
Merlinde D. Wobben
Student Nr. UvA: 12477885
Student Nr. VU: 2644946
48 Credits
November 2019 - June 2020
Supervisor/Examiner:
Examiner:
Dr. Esther Alarcon-Llado
Dr. Hong Zhang
Merlinde Wobben
3D photovoltaics, AMOLF June 22, 2020
Abstract
Indium-arsenide is a promising III-V semiconductor material for photovoltaic applica-tions, but the currently available growth methods use energy intensive high vacuum sys-tems and high temperatures, which prevents the rapid commercialization of InAs photo-voltaic devices. A new possible route of fabrication is through metallic indium which can be electrochemically converted to InAs at ambient conditions. To have a full electrochemical methodology to grow InAs nanostructures, first an indium nanoparticle array must be fabri-cated with the targeted morphology and arrangement. This thesis focuses on understanding and optimizing the fabrication of these indium nanoparticle arrays that have the potential to be converted into InAs. We propose a new indium nanoparticle growth methodology, which makes use of electrochemistry and soft-conformal imprint-lithography (SCIL). SCIL can provide an insulating mask to any substrate, such that the indium nanoparticles can be grown constrained to a predefined shape and size.
With this method we have successfully fabricated indium lattices with a strong optical re-sponse that is dominated by the surface lattice resonance phenomenon. SEM images clearly show that the shape and arrangement of the SCIL template is successfully translated into the indium lattice, giving polydispersity only in height. In chapter 3, the optical properties are studied ex-situ using reflection measurements and angle-dependent light transmission measurements. Despite the polydispersity in particle height, the transmission measurements of the grown lattice are well explained with a single particle size FDTD simulation of trans-mitted light. In chapter 4, electrochemical techniques, scanning electron microscopy, and in-situ optical measurements are used to study the growth and nucleation of indium par-ticles on patterned and unpatterned substrates. Additionally, these results were compared with numerical (FDTD) and analytical (Mie’s theory) predictions.
The first results of the combination of SCIL and electrochemistry are demonstrated in this thesis, which can be further used for III-V semiconductor fabrication. Optimization of the electrochemical steps shows potential for a high-throughput methodology at ambient conditions to produce nanopillar arrays for photovoltaic applications.
Table of Contents
1 Introduction 6
1.1 Outline of the thesis . . . 7
2 Electrochemistry for nanophotonics 8 2.1 Electrochemistry . . . 8
2.1.1 Electrode kinetics . . . 9
2.1.2 Electrochemical set-up . . . 9
2.1.3 Electrochemical methods . . . 10
2.1.4 Electrochemical properties of indium . . . 12
2.2 Nanophotonics . . . 13
2.2.1 LSPR described by Mie’s theory . . . 13
2.2.2 Surface lattice resonances . . . 17
2.2.3 Optical methods . . . 17
3 Surface lattice resonances in indium arrays 19 3.1 Fabrication of the lattice . . . 19
3.2 Characterization of the lattice . . . 20
3.3 Optical properties . . . 22
4 Insights into nucleation and growth with in-situ measurements 28 4.1 Studying nucleation with electrochemical techniques . . . 28
4.2 Self-assembly vs templated growth with in-situ reflection measurements . . . 34
4.2.1 Unconstrained growth . . . 34
4.2.2 Templated growth . . . 37
5 Conclusion & outlook 41 Acknowlegdements 43 References 44 Appendix 47 Appendix A: More on SLRs in indium arrays . . . 47
Appendix B: FWHM analysis . . . 49
Appendix C: Templated growth 20 seconds SEM . . . 50
Appendix D: Mie’s theory code in Python . . . 51
Appendix E: Dispersion code in Python . . . 54
Introduction
Due to the use of fossil fuels, the concentration of CO2 in the atmosphere has risen
sig-nificantly in the past century, which leads to global warming1. In order to keep the same
living standards we have nowadays, but with a much smaller carbon footprint, we need new sustainable energy sources, such as solar energy. While the efficiency of commercial silicon solar cells has increased over recent years, every material has a theoretical efficiency limit that cannot be surpassed. New materials have to be found in order to make solar cells with efficiencies that are higher than currently available commercials cells. It has been shown by Tavakoli et al. (2019) in a theoretical analysis that the incorporation of III-V semiconductor nanopillar arrays on top of silicon solar cells can increase the efficiency to out-perform the film-variant of a tandem cell with silicon and a III-V semiconductor2.
An example of a III-V semiconductor is indium-arsenide, which applications in photo-voltaics have been reported3. A method of growing InAs has been shown by Fahrenkrug et al. (2014)4; metallic indium can be converted into the semiconductor indium-arsenide with one electrochemical growth step. This means that if we can grow indium nanopillars with control over the size, shape and arrangement, these properties can be transferred to III-V structures upon conversion to indium-arsenide, resulting in structures with a maximized ef-ficiency. Yorick Bleiji has successfully used the technique proposed by Fahrenkrug et al. in his master thesis directly onto indium foils. However, upon templated thermal evaporation of indium in order to form nanoparticles, he found out that this technique has its limita-tions; the lift-off of evaporated indium particles on the insulating parts of the template is hindered by the softness and low melting point of indium5.
Therefore, a different technique is explored to make indium nanostructures at ambient conditions. The production consists of two main steps. First, an insulating mask is stamped on a conductive substrate, leaving a pattern of conductive holes, using substrate-conformal imprint-lithography (SCIL). This technique is scalable, stamps of any size can be used and the same stamp can be used many times, making it an inexpensive technique. In the sec-ond step, indium nanoparticles are electrochemically grown in the csec-onductive holes in the substrate. In contrast with thermal evaporation and electron-beam lithography, all steps are carried out at ambient conditions, reducing the environmental impact of fabrication. Furthermore, it is a bottom-up approach instead of a top-down approach, which reduces the amount of wasted material.
The nucleation and growth of indium on templated and unpatterned substrates is studied in this work using the electrochemical techniques, SEM images and optical properties. An interesting consequence of metallic nanoparticle arrays is the coherent scattering of light, known as surface lattice resonances. SLRs are very sensitive to the lattice conformation and the size, shape and quality of the metallic particles. Therefore, the optical response of nanoparticle arrays can be used as a probe to characterize the lattice. In this work, the SLRs of the produced lattices are used to quantify the height of the nanoparticles and the oxygen content in the metal.
CHAPTER 1. INTRODUCTION 7
1.1
Outline of the thesis
This work is organised in 5 chapters as follows:
• Chapter 2 discusses the physical principles and fabrication techniques used in this thesis. First the principles of electrochemistry are explained, and shown how they are used in this work. In section 2.2 the physics of LSPRs and SLRs are explained, together with the techniques used for characterizing the samples made in this work. • Chapter 3 shows the results of the fabrication of indium lattices and the
characteriza-tion of the sample and the surface lattice resonance.
• Chapter 4 discusses the efforts to gain an understanding in the nucleation and growth of nanoparticles using in-situ spectroscopy.
• Chapter 5 concludes the work of this thesis and gives an outlook to new efforts that can be made towards the application of the techniques used in this work.
Electrochemistry for
nanophotonics
This chapter discusses the theoretical foundations of the work of this thesis. First, the concepts of electrochemistry (section 2.1), including the techniques cyclic voltammetry and chronoamperometry are covered. A background on nanophotonics is given in section 2.2, which covers Mie’s theory, localized surface plasmon resonances and surface lattice reso-nances.
2.1
Electrochemistry
Electrochemistry is the science that studies chemical reactions arising from charge transfer across a (semi)conductor/electrolyte interface. It revolves around the combination of the transport of ions in the electrolyte and reduction or oxidation reactions at the interface.
As in any chemical reaction, enough energy should be available for the reactions to occur. In electrochemistry the reactions can either be driven by externally supplied energy in the form of a potential, or it can, under certain conditions, be a spontaneous reaction producing a potential. The potential required for a reaction is called the standard reduction potential E0, which is measured at 298.15 K and 1 atm, and it is assumed that all soluble species have an activity of 1.000 M. The Nernst equation uses the standard reduction potential to calculate the reduction potential of a reaction if the conditions are varying from the standard conditions: E = E0−RT nFln ared aox ≈ E0−RT nFln cred cox (2.1) where R is the gas constant, T the temperature, n the amount of electrons involved in the reaction, F the Faraday constant and aredand aoxare the activity of the species, which can
often be approximated as the molar concentrations credand cox6.
A potential is always relative to another potential, so an important convention in elec-trochemistry is the definition of a voltage reference. The reduction potential of protons at any temperature, this reaction:
2H++ 2e− −→ H2(g), (2.2)
is defined to be 0 V vs SHE, the so-called standard hydrogen electrode (sometimes referred to as NHE).
In this work, all potentials are reported vs a Ag/AgCl electrode, which is known to be +0.197 V vs SHE7.
CHAPTER 2. ELECTROCHEMISTRY FOR NANOPHOTONICS 9
2.1.1
Electrode kinetics
Electrochemical reactions take place at the interface between a solid and a liquid. The transfer of charge involved in the reactions requires a conductive environment. The con-ductive solution is the electrolyte and the concon-ductive solid are two metals called electrodes. At both of the electrodes one reaction takes place. By definition, the reduction, where an electron is transferred from the electrode to the reactant takes place at the cathode. The inverse reaction, the oxidation, takes place at the anode.
When a potential is applied to an electrochemical system, the current can be predicted using the Butler-Volmer equation. To reach this equation we shall consider a simple uni-molecular one-step reaction which is taking place at one of the electrodes:
Ox + ne− k−−−*)−−−red
kox
Red (2.3)
Where kredand koxare the reaction rate constants of the forward and backward reactions.
The total rate is the difference between these two:
ν = koxcSRed− kredcSOx (2.4)
where cS
Red/Ox is the concentration of the reactants at the electrode surface. The
Butler-Volmer equation assumes that the Gibbs free energy on which the rate constants depend is completely dependent on the overpotential η = E −Eeq(Eeqbeing the equilibrium potential)
applied. Using the Nernst equation and Faraday’s law j = F ν the Butler-Volmer equation follows: j = j0 exp αF η RT − exp (1 − α)F η RT (2.5) In this equation, j0 is the exchange current density and α is the anodic transfer coefficient.
Close to equilibrium, α is close to 0.5, but far from equilibrium it might deviate strongly. In the Butler-Volmer model, it is assumed that the reactants are always present every-where at the surface of the electrode. However, this is not always true. The electrochemical reaction can be limited by three types of ion kinetics; diffusion, migration and convection. Diffusion is caused by a gradient in the concentration, which causes a gradient in the chemi-cal potential. Migration is caused by a gradient in the electrichemi-cal potential, which causes the movement of ions. Finally, convection is caused by the movement of the solvent itself6.
2.1.2
Electrochemical set-up
Three-electrode configuration
A set-up for electrochemical reactions can be as simple as a beaker glass with two metal wires connected to a battery and submerged in an electrolyte. By convention, the electrode where the reaction of interest takes place is called the working electrode. This can be either the cathode or the anode. The other electrode is then called the counter electrode, which either supplies or subtracts the necessary electrons. The counter electrode is usually made of an inert material that does not corrode (platinum or graphite e.g.) to ensure the stability of the electrode. Furthermore, the counter electrode must have a high activity and surface area to make sure it provides the necessary current.
In a three electrode configuration, a third electrode, called the reference electrode is added to measure the potential between the solution and the working electrode. In order to serve as a reliable reference the electrode has to have a stable and well known potential. The Ag/AgCl electrode is nearly an ideal non-polarizable electrode, meaning the potential is not affected by the passage of current. The potential between the working and counter electrodes is regulated by a device called a potentiostat. This potential is automatically adjusted throughout the experiment in order to control the potential of interest (between the reference electrode and the working electrode) that drives the reaction at the working electrode. The potential difference applied to the system causes a current flow in the electrolyte, closing the circuit. A schematic of this set-up is shown in figure 2.17.
Figure 2.1: Schematic of a three electrode configuration.
Electrochemical cell
The electrochemical experiments performed in this work are all conducted in a custom made three-electrode electrochemical cell shown in figure 2.2 made of polyether ether ke-tone (PEEK). The cell was originally designed for the electrochemical growth of InAs which has a possibility of reducing solid arsenic to toxic arsine gas (AsH3). For this reason the
blue in- and outlets were added to remove any gasses produced. In this work, the in- and outlet are not used, but they remain on the cell.
The used counter electrode is a coiled platinum wire, the reference electrode is an ET072-1 micro Ag/AgCl electrode and the working electrode is the sample that can be mounted on the back. In this work it is always a piece of transparent ITO (indium-tin-oxide) coated glass cut to roughly 1 by 1.5 cm. The cell is ensured to be leak-tight by pressing the ITO with the O-ring shown in figure 2.2b and an electrical connection with the sample is ensured with a copper spring. The window on the front of the cell is a quartz window which allows observation inside the cell. The diameter of the hole at the back of the cell is 6 mm, which allows for a deposition area of 28.3 mm2.
Before every electrochemical experiment in this work, the potential of the micro refer-ence electrode was measured versus an analytical XR300 Ag/AgCl referrefer-ence electrode in a two electrode configuration with a supporting electrolyte of 1 M Na2SO4. The open circuit
potential (OCP) between the two electrodes was measured for 3 minutes, and the last ac-quired value was used. All experiments were conducted with a Biologic SP-300 potentiostat to control the potential in the set-up and measure the currents, EC-lab® software v11.33 was used to control the potentiostat.
2.1.3
Electrochemical methods
With a potentiostat, one controls the potential or the current of the electrochemical system under study. Many different techniques have been developed to study processes while using the potentiostat. Two important techniques used in this work are briefly discussed here; cyclic voltammetry and chronoamperometry.
CHAPTER 2. ELECTROCHEMISTRY FOR NANOPHOTONICS 11
Figure 2.2: The electrochemical cell used in this work. (A) Front view of the cell. (B) Back view of the cell where the sample is mounted. Figure adapted from the master’s thesis of Y. Bleiji (2019)5.
Cyclic voltammetry
Cyclic voltammetry (CV) can give a better understanding of the reactions taking place in the system. The potentiostat applies a triangle wave potential to the electrochemical system, it sweeps the potential in one direction between E1 and E2 and then reverses the
sweep. The rate of the potential variation is called the scan rate. The whole time during the potential sweep the potentiostat measures the current response to the potential. A typical CV is shown in figure 2.3. A chemical reaction starts taking place as soon as the potential is high enough (onset potential), e.g. when the activation energy for the reaction is reached. Depending on the sign, this reaction can either be a reduction or an oxidation.
Confusingly, in the field of electrochemistry there are two conventions used regarding the sign of the current. In this work, positive currents are anodic currents. Meaning that a positive, anodic current is caused by oxidation, electrons are transferred from the electrolyte to the electrode. The negative peak is therefore a reduction, electrons are transferred from the electrode to the electrolyte. In figure 2.3 the anodic peak decreases at higher potentials because the species concentration at the surface is depleted, meaning the current becomes diffusion limited. The scan rate influences when the current becomes diffusion limited, and therefore the height of the peak7.
Chronoamperometry
A very simple but nonetheless very useful technique is chronoamperometry (CA). A constant potential is applied, while the current response is measured. At the start when the potential is applied, the system is brought out of equilibrium. In the scope of this work, CA is mostly used as a fabrication technique. At negative potential, a reduction reaction takes place, causing a metal to deposit on the working electrode.
Besides fabrication, CA can also be used to study the properties of the system. During the first part of the measurement the system responds to restore its equilibrium. Therefore, CA can be used to study the kinetics of the reaction system. Furthermore, if the current measured during the experiment is integrated the total transferred charge Q is obtained, which in this work is used to estimate the amount of metal deposited. The relation is as
Figure 2.3: (A) Typical CV for a one electron reaction (B) The corresponding time vs potential.
follows:
nm=
Q
N F (2.6)
Where F is the Faraday constant, N is the valency of the metal under study, and nmis the
number of moles of the deposited metal. Chronoamperometry can also be used to study the nucleation mechanisms as derived by Scharifker and Hills8. Their model supplies two
limiting nucleation mechanisms, the instantaneous and progressive. In instantaneous nucle-ation, all nuclei are activated at the same time and continue to grow slowly. In progressive nucleation, nuclei continue to be activated during the course of electrochemical growth af-ter which they grow fast. The models for instantaneous (equation 2.7) and progressive nucleation (equation 2.8) are as follows:
i2 i2 max =1.9542t tmax 1 − exp −1.2564 t tmax 2 (2.7) i2 i2 max =1.2252t tmax ( 1 − exp " −2.3367 t tmax 2#)2 (2.8) Where imax is the maximum current and tmax is the time corresponding to the maximum
current9.
2.1.4
Electrochemical properties of indium
In this work, we study the electro-deposition of indium. Indium has applications in electron-ics, solar cells, batteries, transparent conducting materials, etc. Indium is a malleable metal in group 13 of the periodic table, which makes it a post-transition metal. Its atomic number is 49 and it has 3 valence electrons. Indium resembles tin, with its silvery-white color, and its density is 7.31 g/cm310. The most stable ion in water is In3+ , but it is genererally accepted that the three-electron charge-transfer for the deposition of indium takes place in two consecutive steps with an In+ ion as intermediate. The existence of the intermediate In2+ ion has been demonstrated to be unlikely9.
A Pourbaix diagram is an E vs. pH phase diagram which indicates which species are in a stable phase in a certain electrochemical system. In a Pourbaix diagram the reduction po-tential versus SHE is shown in the y-axis and the pH on the x-axis. The Pourbaix diagram of an indium-water system is shown in figure 2.4. The green region indicates the potential-pH region where solid indium is stable. The vertical lines indicate acid/base transitions and the horizontal line depict redox reactions. The orange dashed lines show the standard potential for water reduction and oxidation. Figure 2.4 tells us that deposition of indium from In3+
CHAPTER 2. ELECTROCHEMISTRY FOR NANOPHOTONICS 13
Figure 2.4: Pourbaix diagram for the system indium-water at 25°, 1 atm and indium concentration of 0.01 M. Figure adapted from the Materials project11.
2.2
Nanophotonics
This work combines the field of electrochemistry with the field of nanophotonics. All the samples are fabricated with electrochemistry, and subsequently studied using the principles of nanophotonics. Nanophotonics is the study of the interactions of light with matter of the nanometer scale; matter that is smaller than the wavelength of light itself. Light, in this case refers to ultraviolet, visible and near-infrared light with wavelengths from ca. 300 to 1200 nm. Nanophotonics is a multidisciplinary field combining physics and optics, and has a broad scope of applications in lasers, detectors, solar cells, spectroscopy etc.12.
In this section a few aspects of nanophotonics that are used in this work are explained. First we take a look at localized surface plasmon resonances (LSPR) and how they are described by Mie’s theory. In section 2.2.2 the coupling of LSPRs with Rayleigh anomalies to form surface lattice resonances will be discussed. Finally the optical set-ups used in this work will be described.
2.2.1
LSPR described by Mie’s theory
When light interacts with metallic nanoparticles smaller than its own wavelength, an optical phenomena called a localized surface plasmon resonance (LSPR) is generated. The light wave interacts with the free surface electrons in the metal, which produces coherent electron oscillations with a frequency dependent on size, geometry, composition and refractive index of the environment13,14. The LSPR phenomenon has a strong optical response, with light
extinction cross sections much larger than the particle geometrical cross section15. The
LSPR scattering and absorption of a nanoparticle can be calculated according to Mie’s theory16. Mie’s theory is an analytical solution to Maxwell’s equations17 which describes
the scattering and absorption of a plane wave by a spherical particle. Mie’s theory has been generalized by Sinzig and Quinten18 to a formalism to compute the scattering and
absorption cross section of nanoparticles when illuminated with an incident plane wave. In this work, these calculations are carried out by a Python script, that has been translated from a Matlab script as published by Baffou (2012)19. The Python script can be found in appendix D.
In the calculations, a particle of radius r0 with a complex electric permittivity ε = n2
embedded in a medium of real refractive index nmis considered. The particle is illuminated
by a plane wave, resulting in a scattering and absorption cross section.
The permittivity of a metal plays a vital role in the determination whether an LSPR mode can exist or not. The dipole polarizability, α of a spherical nanoparticle is given as:
α = 3ε0 4 3πr 3 ε − εm ε + 2εm (2.9) Where V is the volume of the particle, ε0 is the free space permittivity, εm is the
permit-tivity of the surrounding medium, and ε is the permitpermit-tivity of the nanoparticle. A dipolar LSPR can only exist when the denominator diminishes, thus the real permittivity of the nanoparticle must be negative, since the permittivity of the surrounding medium is always positive20.
Figure 2.5 shows the real and imaginary permittivity of gold, silver, aluminium and indium. Gold and silver are commonly used in LSPR technologies, as they exhibit strong resonances in the visible. However, due to their positive real permittivity in the UV range, they have limited applications. Indium and aluminium do support LSPRs in the UV making them good plasmonic materials21.
Figure 2.5: (A) The real and (B) imaginary permittivity of indium, gold, silver and aluminium. Data of silver and gold from Johnson and Christy (1972)22, aluminium from Raki´c (1995)23 and indium from Koyama et al. (1973)24.
Size dependent permittivity
The line-width (Γ) of an LSPR mode as described above is based on the shape, size and per-mittivity of the bulk material. However, it has been shown that for small sizes of
nanoparti-CHAPTER 2. ELECTROCHEMISTRY FOR NANOPHOTONICS 15
cles the LSPR broadens due to the effects of additional electron-surface scattering pathways. For larger sizes these effects are diminished, but radiation damping dominates, resulting in an increased line-width. The total line-width of a resonance is summarized as the sum of the bulk damping constant, a term for the electron-surface scattering and a term for the radiation damping proportional to the volume:
Γ = Γintrinsic+ Γsurf ace+ Γradiation (2.10)
For particles of roughly 20 nm in radius, the resonances are the narrowest, showing only the intrinsic damping pathways. For these particles the line-width is completely dependent on the permittivity25:
Γ = 2ε2
|(δε1/δω)|
(2.11) Where ε1 and ε2 are the real and imaginary components of the permittivity respectively.
For particles larger than 20 nm, the permittivity obtained from metal films can be used when calculating the LSPR according to Mie’s theory, but for smaller particles, a size correction due to the electron-surface scattering is needed:
εcorrected(ω) = εbulk(ω) + ω2 p ω2+ iωγ bulk − ω 2 p ω2+ iωγ R (2.12) Where γbulk is the electron collision frequency of the bulk, γR is the radius-dependent
electron collision frequency and ωp is the plasma frequency. γR is given as
γR= γbulk+ 2gs
Vf~
2R (2.13)
With ~ the reduced Planck constant, R the radius of the particle, Vf the Fermi velocity
of the metal and gs a proportionality factor between 0 for no scattering to 1 for isotropic
scattering26.
The plasma frequency is given as
ωp= s 4πe2N AρN meM (2.14)
Where e is the elementary charge, NAis Avogadro’s constant, ρ is the density of the metal,
N is the number of free electrons per atom, me is the mass of an electron and M is the
molar mass of the metal27.
The proportionality factor gs in equation 2.13 is a key parameter which determines the
amplitude of the size correction. It is given by: gs(ω) = 1 ~ωE2f Z ∞ 0 E3/2(E + ~ω)1/2f (E)(1 − f (E + ~ω))dE (2.15)
Where Ef = ~2(3π2ne)(2/3)/(2me) is the Fermi energy and f is the Fermi-Dirac
distribu-tion26:
f (E) = 1
e(E−Ef)/kBT + 1 (2.16)
The results of this analysis are shown in figure 2.6. The absorption cross section of indium nanoparticles in the sizes 1 to 10 nm are plotted both with and without a size correction. It shows that the size corrected resonances are lower in intensity and thus have a larger line-width.
The full-width at half maximum (FWHM) of the calculated resonances for sizes ranging from 1 to 40 nm are shown in figure 2.7 for both the size corrected resonances and the resonances without size correction. It shows that indeed for small particles the FWHM is larger than calculated with bulk permittivity, but from 20 nm the difference is very small,
Figure 2.6: Comparison of Mie’s theory absorption cross section using the bulk permittivity for indium nanoparticles ranging from 1 to 10 nm (blue to green) with absorption cross section calculated with size corrected permittivity (orange to red).
and from 30 nm the FHWM is the same.
Note that the FWHM analysis is only possible for first order resonances. This makes an experimental analysis only possible for particles up to 20 nm; as soon as the quadrupole emerges the FWHM is not reliable anymore. According to Mie’s theory the absorption peak starts to shift to higher wavelengths for particles larger than 20 nm and the quadrupole emerges at roughly 300 nm, which will also broaden and shift to larger wavelengths as the particle is growning.
Figure 2.7: Comparison of the FWHM of the absorption cross section of indium nanoparticles ranging in rize from 1 to 40 nm. The orange dots are from resonances calculated with size corrected permittivity. The green dots are from resonances calculated with bulk permittivity.
Mie’s theory can only be used for spherical particles. For different shapes, a numerical analysis technique called finite-difference time-domain (FDTD) is used. In this work these analyzes are computed in the software Lumerical by Dr. Marco Valenti, a post-doc in our research group.
CHAPTER 2. ELECTROCHEMISTRY FOR NANOPHOTONICS 17
2.2.2
Surface lattice resonances
As seen in figure 2.7 the FWHM of the LSPR of larger particles rises. When plasmonic nanoparticles are placed in a periodic array, the FWHM can become much narrower due to an effect called a surface lattice resonance (SLR). In periodic arrays diffractive states known as Rayleigh anomalies (RA) trap and funnel light in the plane of the lattice. These diffractive orders can couple to the LSPR of single particles, giving rise to SLRs28.
Figure 2.8 shows both the transmission spectra of a LSPR in a single gold particle and a SLR in a periodic 1D array of 1000 gold particles with a periodicity of 620 nm.
Figure 2.8: Comparison between a LSPR and SLR. (A) Schematic transmission spectrum of one gold nanoparticle and (B) schematic transmission spectrum of a periodic array of gold nanoparticles. Figure adapted from Kravets et al. (2018)29
The energy of the Rayleigh anomalies that the LSPR couples to are solutions to the equation:
E = ~c
n|k//+ (m1T1+ m2T2)| (2.17)
where m1 and m2 are the orders of diffraction, T1 and T2 are the reciprocal lattice vectors
of the array, k// is the wave vector parallel to the surface and n is the refractive index of
the surrounding medium.
If we look at the example in figure 2.8, where the periodicity a of 620 nm and the wave vector zero, we find a RA at E = ~c
n(m 2π
Λ) = 2.0eV = 620nm for the first order diffractive
mode with a surrounding medium with refractive index of 1. This is also where we see the minimum in transmission in figure 2.8b.
2.2.3
Optical methods
In-situ reflectance set-up
To study the optics of the samples made in this work, we use two different set-ups. The first set-up is incorporated in the EC-cell discussed in section 2.1.2 to measure the reflectance of the sample during the electrochemical growth. A reflectance probe was added to the front of the cell, directed at the quartz window and through the cell to the sample (see figure 2.9). The probe contains 6 illumination fibers and 1 read fiber. The read fiber is connected to an Ocean Optics maya 2000 pro spectrometer wiht a spectral range of 165-1100 nm. The illumination fibers are connected to an Ocean Optics DH-2000, 26 W deuterium and 20 W Tungsten halogen white light source with a wavelength range of 190-2500 nm. The custom software ENS monitor made by Sjoerd Wouda was used to control the spectrometer. In every experiment first the background is recorded and the reflectance before and during the electro deposition to have a good reference. Furthermore, for every day that experiments were conducted, a reference measurement of the reflectance of a piece of silicon was done to be able to calculate the effective reflectance of the sample.
Figure 2.9: Top view of the EC-cell with the reflection probe. Figure adapted from the master’s thesis of Y. Bleiji (2019)5.
Ex-situ reflectance and transmittance set-up
The reflectance and transmittance of the electrochemically grown indium samples were measured in an integrating sphere setup after the deposition. For the reflection the sample was mounted behind an integrating sphere and illuminated with polarized collimated white light from a SuperK EXTREME/FIANIUM super-continuum laser at an angle of 8°. The reflected light was collected by an integrating sphere and sent to a Spectra Pro 2300i spec-trometer equipped with a Pixis 400 CCD through a multi-mode fiber.
Angle dependent transmittance was measured when the sample was mounted on a ro-tating stage and the transmitted light was collected in the integrating sphere.
Chapter 3
Surface lattice resonances in
indium arrays
As introduced in section 2.2.2 arranging metallic nanoparticles in a periodic array gives rise to surface lattice resonances. In this section the results of the fabrication of indium lattices with a new scalable nano-fabrication technique at ambient conditions will be discussed. First, the fabrication of the sample will be explained, after which the physical properties of the sample will be characterized. Finally, the reflection and angle dependent transmission of the sample will be discussed and compared with finite-difference time-domain simulations.
3.1
Fabrication of the lattice
Figure 3.1 summarizes the scalable methodology we use to fabricate indium nanoparticle arrays. The fabrication consists of two steps. Using soft-conformal imprint-lithography (SCIL) an insulating SiO2 and PMMA mask is nano-imprinted on an indium tin oxide
(ITO) coated glass (Fig 3.1a-c). First a 32 nm SiO2 layer, a 250 nm PMMA layer and a
second 74 nm solgel layer is spincoated on ITO coated glass5. A PDMS stamp is pressed into the solgel and the resulting imprinted toplayer of SiO2 is etched down with CHF3and
Argon. Subsequently the PMMA is etched down with O2 and the bottom SiO2 layer one
more time with CHF3+ Ar until the conductive oxide is reached. The resulting substrate
is a hexagonal array with a pitch of 534 nm. The average diameter of the holes that expose the conducting substrate is 434 nm.
In the second step, the hexagonal template was exposed to an aqueous electrolyte con-taining 0.05 M InCl3, 0.2 M KCl, and 0.005 M HCl (pH 2.5) in the electrochemical cell
shown in figure 2.2, to electro-deposit indium at a constant potential of -1.3 V vs Ag/AgCl for 33.7 seconds (Fig 3.1d-f). In the first stage of electrochemical growth nuclei form and grow into spheres. Then, the balls grow into one particle per hole and periodicity is estab-lished. Further on, the particles grow unilateral into nanopillars. As a final step the PMMA layer between the particles and the SiO2layer on top was removed by sonicating the sample
in acetone at 30° C for 30 minutes. The bottom SiO2layer was not removed in this process.
During the growth, the current was measured and the resulting chronoamperometry is shown in figure 3.2. We see a very fast peak in the current which is explained by the non-faradaic processes when the potential is applied. After that, the current decreases to -3 mA/cm2, where the deposition starts. The current increases again as the surface area of indium is growing. After 5 seconds, the process becomes diffusion limited and the current decreases again. However, after a few seconds the current increases again until the end of the deposition. This phenomenon is difficult to explain, but may be an effect of the template, as mass transport in the holes may get easier as the holes are filling up and become less deep.
Figure 3.1: Electrochemically deposited indium lattice fabrication steps. (A)-(C) Summary of soft conformal imprint lithography (SCIL). (D)-(F) Electrochemical growth of indium nanopillars in three stages.
3.2
Characterization of the lattice
The resulting periodic indium nanoparticle array is shown in figure 3.3. Figure 3.3a reveals that the growth has a yield of 95.6%. This is calculated as the amount of grown pillars divided by the amount of holes (1720/1800). The non-grown pillars have partial or no depo-sition of indium (see appendix A, figure A3). We suggest the growth was inhibited by less chemically active impurities or an insulating layer remaining from the SCIL process. Figure 3.3b clearly shows that the shape and diameter of the nano-imprint was well translated to the grown nanopillar. Figure 3.3c shows a cross section of one indium pillar made by FIB, which reveals the curvature of the indium particle. Atomic force microscopy (AFM) was used to measure the maximum pillar height of 378 pillars (see figure 3.4), revealing a mean pillar height of ∼ 230nm with a standard deviation (σ) of ∼ 30nm. We see a rather broad particle height distribution, which depends on the homogeneity of the electrical growth.
The electrochemical growth process of metals can be separated in two different stages; nucleation and growth (ruled by the electrochemical reaction rate, diffusion and migration
CHAPTER 3. SURFACE LATTICE RESONANCES IN INDIUM ARRAYS 21
Figure 3.2: Chronoamperometry graph of electro-deposition in hexagonal imprinted substrate.
Figure 3.3: (A-B) Scanning electron microscope (SEM) images of electrochemically grown Indium pillars in an hexagonal lattice. (A) Large scale overview image at a 30° tilt with a magnification of 10 000 times. (B) A single pillar at a 35° tilt with a magnification of 500 000 times. (C) Focused ion beam (FIB) image of a cross section of a single pillar at a 52° tilt and a magnification of 200 000 times.
of ions). Electrochemistry offers a toolbox of techniques to control these processes with the applied potential. In this work the deposition was done by applying a constant potential, resulting in a polydisperse pillar height. This limits the applications of the lattices, but it can be optimized by exploring other electrochemical growth techniques such as applying the potential in short pulses.
Figure 3.4: (A) Particle height distribution of 378 pillars in a 100 µm2 region. The mean pillar height is ∼230 nm and the standard deviation σ is ∼ 30 nm. (B) 3D AFM topography map from where the statistics were made.
Figure 3.5: FDTD simulated extinction cross section of a single indium nanopillar with the exper-imental geometry found in SEM and AFM in air and in RI matching oil.
3.3
Optical properties
To study the optical properties of the lattice we first take a look at the finite-difference time-domain (FDTD) simulation of the extinction cross section of a single pillar with the dimensions found in the SEM and AFM analysis. Figure 3.5 shows the extinction cross section of the particle exposed to air and embedded in index matching oil. For the particle in air we mainly see one big peak at lower energies, but for the particle in RI matching oil there are multiple modes to which the Rayleigh anomalies can couple.
As explained in section 2.2.2 a coupling of the LSPR with Rayleigh anomalies (RA) in a lattice gives rise to surface lattice resonances. The energies of the RAs can be calculated based on the lattice geometry. The direct lattice vectors of the geometry of our hexagonal sample with a pitch of 534 are:
t1= Λ√3 2 ˆx + Λ 2y =ˆ Λ√3 2 Λ 2 (3.1)
CHAPTER 3. SURFACE LATTICE RESONANCES IN INDIUM ARRAYS 23
Figure 3.6: Reflection (A and C) and transmission (B and D) measurements at 8 degree incidence angle. For A and B the sample was exposed to air and for C and D the sample was embedded in index matching oil (RI=1.51).
t2= Λ√3 2 ˆx − Λ 2y =ˆ Λ√3 2 −Λ 2 (3.2) This results in the following reciprocal lattice vectors:
T1= 2π |t1× t2| t2,y −t2,x ] = 2π Λ√3x +ˆ 2π Λyˆ (3.3) T2= 2π |t1× t2| t1,y −t1,x ] = 2π Λ√3x −ˆ 2π Λyˆ (3.4)
The RA energies with these vectors are solutions of: E =~c
n|k//+ (m1T1+ m2T2)|. (3.5)
where where m1 and m2 are the diffraction orders and k// is the crystal momentum, which
is a function of the incidence angle of the light30 31 32: kx=
E ~c
· sin θ. (3.6)
Reflection and transmission measurements at an incidence angle of 8° were carried out on the periodic sample. Figure 3.6 shows the reflection and transmission with air (a and b) and RI matching oil (c and d) as surrounding refractive indices. The transmission of the indium array in air (Fig 3.6b) is dominated by a broad background extinction, which is expected for the LSPR of large indium particles (see figure 3.5). However, narrower features appear for s polarization at 1.76 eV and for p polarization at 1.93 eV and 1.62
eV. In figure 3.6b it can be seen that these features correspond well with the Rayleigh anomalies in SiO2. This is evidence of the influence of the lattice to the strong optical
response of the large indium nanopillars. Moreover, according to the polarization, different diffractive modes are modifying the spectrum. This proofs that the spectral features near the RAs are a coupling between polarizable metallic nanoparticles (LSPRs) and the in-plane diffractive modes, resulting in surface lattice resonances (SLRs)30. These polarization dependent features in transmission correspond well with relatively narrow reflection peaks, as assigned with arrows in figure 3.6.
Two main changes were seen when using a refractive index matching oil. First, the overall broadband transmission was decreased, which corresponds well with the expected single-particle LSPR extinction cross section increase due to the higher RI of the environment (see figure 3.5). Secondly, the SLRs in SiO2 are more pronounced (in transmission) when the
RI of the environment is symmetrical (i.e. the RI of the oil is at 1.51 similar to the RI of the SiO2 at 1.45, that partly surrounds the particle), in good agreement with previous
reports33.
Figure 3.7: Analysis of the experimental angle dependent transmission of a hexagonal array. (A) SEM image of the hexagonal nanopillar array with direct lattice vectors. (B) Irreducible Brillouin zone of a hexagonal array with the Γ, K and M point, the diffractive modes, and the reciprocal vectors indicated. (C) Schematic of the polarization of p and s polarized light. (D-E) The indium lattice dispersion obtained from transmission data for the Γ-M case with the sample exposed to air (D) and RI matching oil (E) for s polarized light with overlaid dashed lines for diffraction orders.
To further understand the influence of the diffractive modes on the optical response of the electrochemically grown indium nanopillars, we carried out angle dependent transmis-sion measurements with the sample exposed to air and with the sample embedded in RI matching oil, as shown in figure 3.7d-e. In the figure the Rayleigh anomaly energies are indicated with dashed lines. The calculations are based on the geometry of the array which has a pitch Λ of 534 nm, as shown in figure 3.7a.
We show the experimental s polarized (as illustrated in 3.7c) light transmission in figure 3.7d-e. In figure 3.7d the sample is exposed to air, and in figure 3.7e the sample is embedded
CHAPTER 3. SURFACE LATTICE RESONANCES IN INDIUM ARRAYS 25
in refractive index matching oil. The sample was rotated around the y-axis as depicted in fig-ure 3.7a from θ 0 to 46° changing kxfrom the Γ-point to the M-point while the transmission
was recorded for every 2°. The transmission is displayed in colour as a function of photon energy and kx, where kxis the wave vector component parallel to the lattice surface in the
x direction. For every data point corresponding to a certain energy and angle, kxwas
calcu-lated according to equation 3.6. Figure 3.7b also shows the irreducible Brillouin zone for the hexagonal lattice, and the first order diffractive modes. While kxis changed and ky is
zero, the (1,0) and (0,1) energies are degenerate, as well as (-1,0) and (0,-1). The first order RA energies are indicated in figure 3.7e-f as colored dashed lines. We show the first order in SiO2and the first order in air, calculated with a refractive index of 1.45 and 1.0 respectively.
Figure 3.8: (A) shows the experimental extinction (1-transmission) of the electrochemically grown polydispersed (230nm, SD=30nm) sample, together with a FDTD simulation using the mean size and geometry found by SEM and AFM. (B) shows simulations of pillars with the same bottom and top curvatures but of different heights. The arrows indicate that information on the particle size, morphology and composition can only be obtained in the region where the sea of LSPRs overlap with the RAs.
Interestingly, in the dispersion measured in air we do not only see the SiO2 diffractive
modes modifying the indium extinction across the spectrum, but we also see features in the transmission that correspond to the air modes. In appendix A, an analysis of the propagation length of the SLRs in air and SiO2 in the x-direction is included. It shows that SLRs in
that hides the spectral features of the SLRs in SiO2 at high k-values. The analysis in the
appendix also includes the out of plane decay length (see figure A2).
In contrast to the measurements in air, the dispersion of the lattice measured with RI matching oil as shown in figure 3.7e completely removes the parasitic broad-band absorption coming from the LSPRs in air. The p polarized dispersion of the SLR (shown in appendix A, figure A1) reveals two strong absorption regions at low and high energies. Which could be explained by the presence of two LSPRs in the dispersion.
To further understand the SLR dispersions, a FDTD analysis was carried out using the software Lumerical. Figure 3.8 shows the simulated extinction of hexagonal indium arrays with the experimentally found morphology. The s polarization shows that the RAs fall on the low energy LSPR mode, significantly modifying its extinction with SLRs. In contrast, the high energy LSPR mode has the same shape and location in the spectrum as that of the non periodic single-particle cross section due to the lack of spectral overlap between the RAs and this high-energy LSPR mode. This is also consistent with the fact that as the crystal momentum is increased in the p-polarized dispersion (see appendix A, figure A1), the RA shifts to higher energies, strongly modifying the extinction of the high-energy LSPR mode (at ∼ 2.3 eV).
Interestingly, the simulation for an indium lattice with the average experimental height (230nm) and dome geometry found by SEM, reproduces the SLR peak positions of the polydispersed electrochemically grown indium lattice well. In contrast, the simulated high energy LSPR does not match the experimental extinction. The experimental extinction of the lattice is flat in this high energy region, and therefore no information on the pillar height, morphology or composition can be extracted in this region. This flat extinction is explained in figure 3.8b, where it can be seen that the LSPR at high energies broadens and increases in intensity as the pillar height increases. In contrast, the SLRs of the mean experimental height found by AFM contributes more to the measured extinction than the SLRs of the larger particles.
Figure 3.9: FDTD simulated extinction spectra of indium lattices with varying oxide contents, compared with the experimental extinction at an incidence angle of 8 degrees.
Finally, the SLR is used as a probe for the indium-oxide content by using known optical constants of indium and indium-oxide in the Bruggeman model:
εI− εc εI+ 2εc VI− εO− εc εI+ 2εc VO= 0 (3.7)
where εc, εI and εOare the dielectric constants of the composite, indium, and indium oxide,
CHAPTER 3. SURFACE LATTICE RESONANCES IN INDIUM ARRAYS 27
extinction of three different FDTD simulated lattices with the dielectric constants calculated with the Bruggeman model with different oxide content, together with the experimental extinction at an incidence angle of 8 degrees. With increasing oxide volume fraction the SLR exhibit features in the spectrum that are not present in the experimental spectra. The peak at 1.85 eV shifts to lower energies with more oxygen, whereas in the experimental results the peak is at ca. 1.85 eV. Furthermore, for increasing oxide content a feature at 1.63 eV starts to appear that is not present in the experimental spectrum. This infers that there is a low oxygen content in the fabricated lattice, which is consistent with the electrochemical conditions at a low pH according to the Pourbaix diagram shown in figure 2.4.
Insights into nucleation and
growth with in-situ
measurements
In this chapter we want to gain an understanding in the growth of indium nanoparticles. To do so, we will first focus on the growth of unconstrained indium on an unpatterned ITO substrate. In section 4.1 the nucleation of indium is researched through a study of the electrochemistry of a series of samples grown at varying times are discussed. In section 4.2 in-situ reflection spectroscopy is used to compare the optical response of unconstrained (section 4.2.1) growth and the templated (section 4.2.2) samples as discussed in chapter 3 to discuss the growth mechanism. An important part of understanding the growth is the in-situ spectroscopy measurements as explained in section 2.2.3; during the electrochemical growth the reflection of the sample is recorded with ten spectra every second. The difference in reflection during the growth gives us insights about the growth.
4.1
Studying nucleation with electrochemical techniques
Figure 4.1: Cyclic voltammetric response of an ITO electrode with an electrolyte containing 0.05 M InCl3, 0.2 M KCl, and 0.005 M HCl (pH 2.5). The used scan rate is 0.01 V/s. Note that the
cycles plotted in this figure are not the first cycles done in this experiment.
In this section, the growth of indium on unpatterned ITO substrates will be discussed.
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 29
First, we look at the cyclic voltammetry, two CV cycles are shown in figure 4.1. Negative current due to indium deposition on an ITO substrate in a 0.05 M InCl3, 0.2 M KCl, and
0.005 M HCl (pH 2.5) solution started around -0.65 volt, as expected by the Pourbaix diagram in chapter 2, followed by a maximum current around -0.8 volt. The onset of hydrogen evolution was observed at -1.4 volt. In the first cycle we see a cross-over in the reverse scan, the current at potentials around -0.8 volt is stronger for the reverse scan than the forward scan. This phenomenon is generally due to the fact that it is favorable for the metal to grow on ’itself’ than on the substrate, and since there is already deposited indium on the substrate from the forward scan, the current can stay larger even at lower over-potentials34. At ca. -0.65 volt a large stripping peak at positive current is observed, that
continues in the next cycles forward scan. In the second cycle no cross-over was observed which is explained by the incomplete indium removal during the stripping.
Figure 4.2: (A) Chronoamperometry plot of five samples grown for 1, 2, 4, 8, and 15 seconds. (B) Dimensionless chronoamperometry plots and models for instantaneous and progressive nucleation.
Based on the information gained from the CV, a series of five indium depositions on ITO (indium tin oxide) substrates with different growth times were made at a potential of -1.0 volt vs Ag/AgCl. The current vs. time graphs of the depositions are shown in figure 4.2a. We see that in all cases the maximum current was reached after a little over one second, after which the system became diffusion limited which decreases the current. Figure 4.2b shows the dimensionless CA curves for all samples except for the sample grown for 1 second, because at 1 second the maximum current was not yet reached. In the same graph the models for instantaneous nucleation and progressive nucleation are shown, calculated according to the equations shown in section 2.1.3. This shows that the depositions most closely resemble the model of instantaneous nucleation, but it is not completely explained by either of the models. A more complex model is required to fully explain this CA.
The resulting samples of indium grown on ITO were characterized with scanning electron microscopy. The images and particle size distributions from these images are found in figure
Figure 4.3: SEM images of five samples with different growth time. Grown at a potential of -1.0 V vs Ag/AgCl.
4.3 and 4.4, respectively. In all samples we see round nanoparticles of varying sizes. With the naked eye we see from these images that the particles are densely packed, and the density seems to be decreasing for the last sample with the largest particles. We also see that the sizes keep on increasing with growth time, as expected. The particle size distributions of the growth times of one and two seconds look similar as they are narrow (i.e. ∼100 nm) and peak at small particles of less than 100 nm. For the other three samples the peak is shifting to larger particles from 100 nm for 4 seconds to 150 nm for 15 seconds, and we also see that the peaks are broadening.
Using the particle analysis from the SEM images the density of particles was calculated for every sample, and is shown in figure 4.5a. We see a peak in particle density at 4 seconds, after which it decreases. This is according to expectations, since particles will touch each other and merge as they are growing bigger, which decreases the total amount of density. The sample grown for 2 seconds breaks with the expected trend of first increasing and then
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 31
Figure 4.4: Particle size distribution of five samples with different growth time. Grown at a potential of -1.0 V vs Ag/AgCl. The mean (µ), standard deviation (σ), and number of particles (n) are indicated per sample.
decreasing density. This can be due to the fact that the SEM images of this sample were all made in just one area of the sample, which could have been at the edge of the sample, or in a part where the electric field was lower due to the asymmetry of the counter electrode. However, this was the case for all samples, so all of this data should be interpreted with care.
In figure 4.5b, the amount of deposited indium is shown as a function of growth time. This is calculated based on two different data types, the chronoamperometry and the SEM images. From the CA data, the total charge is the area under the curve of the current, with the charge, using equation 2.6, the total moles of indium deposited is calculated. A very linear trend is observed, which is expected, as longer growth time will lead to more charge, until the concentration of ions decreases significantly. Caution should be exercised
Figure 4.5: (A) Indium particle density deposited by chronoamperometry on ITO substrate as a function of growth time. Data retrieved from particle analysis of SEM images shown in figure 4.3. (B) Amount of indium deposited by chronoamperometry on ITO as a function of growth time calculated from two types of data. Green points show the moles calculated from the charge measured during CA, orange points are based on the acquired particle radii gained from SEM images, assuming all particles are perfect spheres. The density and molecular weight of indium used in these calculations are 7.31 g/cm3 and 114.82 g/mol, respectively35. The inset (C) shows the selectivity for indium deposition, calculated with the data of (B) as SEM data/CA data.
with these measurements since it is now assumed that all the observed charge was due to deposition of indium, which may not be true. More electrochemical phenomena can take place in the cell which will increase the observed charge.
The second trend, shown in orange in figure 4.5, shows the calculated deposited indium from the particle analysis of the SEM images. We estimate the indium deposited by assuming that the particles are all perfect spheres. Using the reported bulk density and molecular weight of indium the amount of moles is calculated. Again caution should be exercised. We do not know if the density of the electro-deposited indium nanoparticles is the same as for bulk indium. Furthermore, it is an approximation that all particles are perfect spheres, since we can see faceted particles on the SEM images. Also, the shape of the particles may change as they grow, which would result in either more or less perfect spheres as growth times increase.
In the data we see that the deposited indium calculated from the SEM is significantly lower than than for the CA data. This data gives the opportunity to quantify how selective the electrochemical system is towards the deposition of indium versus other processes. The selectivity is calculated as:
Selectivity =SEM data
CA data (4.1)
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 33
second of growth has a selectivity higher than 1, which cannot be the case, there cannot be more deposition than there is current. The second sample gives a very low selectivity. Since the SEM images of the first two samples are of low quality, these two data cannot be trusted. The other three data points at longer growth times show interestingly that the selectivity for the indium deposition does not change during the reaction. Selectivities of around 0.6 mean that 40% of the current did not lead to de deposition of indium, but was used towards other processes. The most likely candidate for this process is hydrogen evolution either via 2H++ 2e−−→ H
2. Or through use of the intermediate In+ species by
In++ 2H+ −→ In3++ H 29.
Figure 4.6: In-situ ∆reflection spectra of five samples with different growth time. Grown at a potential of -1.0 V vs Ag/AgCl. For each sample, the yellow flat line is at time zero, right before the deposition starts, and the red line is the last spectrum at the end of the deposition.
4.2
Self-assembly vs templated growth with in-situ
re-flection measurements
4.2.1
Unconstrained growth
As stated earlier, during the electrochemical growth, reflection spectra were taken, which can give us more insights into the growth. The difference in reflection compared to a spectrum taken right before the electro-deposition was started is plotted in figure 4.6 for all five samples. The delta reflection is calculated as:
∆R = Rsamplet
RSiref erence/RSiknown
− Rsamplet=0
RSiref erence/RSiknown
(4.2) where RSiref erenceis a reference reflection measured with a silicon slice in the electrochemical
cell which is filled with electrolyte, and RSiknown is the reported reflection of silicon with a
surrounding refractive index of water at 1.3336.
In figure 4.6 we see that for all samples within the first second of growth a strong peak in negative reflection appears in the UV at 300 nm, which we interpret as absorption. At the same time, a small scattering peak at 400 nm appears and soon disappears again. In the lower energies a broad scattering peak slowly arises after roughly 2 seconds, which peaks at 8 seconds and mostly disappears after 15 seconds of growth.
The change in reflection in this experiment can be simulated with FDTD by choosing
Figure 4.7: FDTD simulations of single indium particles of different sizes which simulate the spectra of growing particles.
a range of particle sizes. The results of these simulations for sizes of 40, 60, 80, and 100 nm are shown in figure 4.7. The particle size distribution of the grown samples show that at 4 seconds the mean radius is 92.8 nm, so the simulated spectrum for the biggest particle at 100 nm should be similar to the final spectrum at 4 seconds. The simulated particles are on a bare ITO substrate, of which the reflection is also simulated. The change in reflection arises from the subtraction of the reflection of ITO from the reflection of the particles on ITO. The spectrum clearly resembles the experimental data ,the same strong absorption peak at 300 is observed. We also see some scattering at 400 nm, but in contrary to experimental results it does not disappear for larger particles and seems to keep on growing. At higher wavelengths we see a broad scattering feature similar to the experimental results that keeps growing for larger particles and seems to shift to higher wavelengths.
In figure 2.6, we see a sharp absorption peak arising at ca. 300 nm for small indium particles according to Mie’s theory. This peak seams to correspond to the absorption peak
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 35
Figure 4.8: Full-width at half-maximum in energy of a peak in absorption at a wavelength of 300 nm in both Mie’s theory and in experimental in-situ reflection. The orange points and the top x-axis correspond to Mie’s theory, the green curve and bottom x-axis belong to the experimental data.
in the experimental reflection spectra. In figure 2.7 the full-width at half-maximum of this peak is shown using the size-corrected permittivity of indium. This revealed that the FWHM first decreases until it finds a minimum and then starts to increase again. The FWHM of the experimental peak was measured to compare these results. The results are shown in figure 4.8, which shows a comparison of the FWHM of Mie’s absorption versus particle size and the FWHM of the experimental data versus time. We see that the trends are very similar in the first stages of the growth, but the experimental data quickly reached a plateau whereas the theoretical data shows an increase. A more detailed analysis of this comparison can be found in the appendix B.
To gain more insights into this absorption peak, we take a look at the Mie absorption and scattering cross section of particles ranging from 1 to 200 nm in radius. The results are shown in figure 4.9. Again, we see an absorption starting to arise for small particles at 300 nm. However, in this analysis, the peak shifts to higher wavelengths and broadens until it covers almost the entire spectrum when the particles are 200 nm. Simultaneous to this absorption, a broad scattering band emerges which spans the whole spectrum and is much larger in intensity than the absorption. To conclude, Mie’s theory predicts that an absorption peak similar to the one we see in the experimental data exists, but only for small particles and when they grow the scattering will dominate. Since this is not what experimental reflection shows us, more experiments were conducted to investigate this absorption peak.
To see if the peak would disappear for particles growing even bigger than the roughly 200 nm we have after 15 seconds of growth, a sample was grown for 2 minutes, with other than growth time the same conditions. A SEM image and the in-situ reflection spectra are shown in figure 4.10. The SEM image shows very interesting structures that grew, as the particles were getting bigger they merged to form networks and lines can be observed which show where individual particles merged. The reflection shows the same pattern as for the other samples at shorter growth times. The absorption peak still remained at the same wavelength throughout the entire deposition. We do see that it decreased in intensity, but it seems that the reflection increased with a constant factor over the entire spectrum in the final stages of the deposition, which is most likely due to the fact that the coverage of the sample is getting higher, leading to more reflection.
Since the question about the strong absorption peak still is not solved, more FDTD simulations were done. A homogeneous indium film of various thicknesses on top of an ITO substrate was simulated and the reflection is plot in figures 4.11. It shows us that over the
Figure 4.9: Absorption (A) and scattering (B) cross section of indium nanoparticles calculated using Mie’s theory. The yellow line belongs to the smallest particle with radius 10 nm, the red line belongs to a particle with radius 200 nm. The lines in between are of particles with sizes in between 10 and 200 nm with increments of 10 nm.
entire spectrum, the reflection increases when a film is added to an ITO substrate. For thinner films, at 5 and 10 nm, we see a feature at ca. 300 nm wavelength that resembles the peak in absorption we see in experimental data, but it has a high positive reflection. For a film of 100 nm thick which is a thickness comparable to the radius of the experimental particles, we do not see this feature anymore.
To conclude this section, a lot of insights are made into the electrochemical growth of indium nanoparticles. There is evidence for an instantaneous nucleation model where most nuclei form at the start of the growth after which they continue to grow, and later on merge which decreases the density of particles. During the growth a broad scattering feature emerges in the visible and IR, and a sharp stable absorption peak in the UV arises in the first second of the growth. Even after two minutes of electrochemical deposition, this peak did not disappear. After thorough investigation of the phenomena unfortunately no explanation towards this peak can be given.
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 37
Figure 4.10: (A) Scanning electron microscope image of electrochemically deposited indium with a growth time of 2 minutes. (B) In-situ reflection spectroscopy of the same sample.
Figure 4.11: Reflection of FDTD simulated indium films of 5, 10 and 100 nm thick on ITO. The ∆Reflection is the difference between bare ITO and a film on top of ITO.
4.2.2
Templated growth
Figure 4.12: Scanning electron microscope images of a SCIL templated electrochemically grown indium lattice. (A) Image before the lift-off of a PMMA and SiO2 layer at a tilt of 40°. (B) Image
Using SCIL and electrochemistry, periodic arrays of indium nanopillars were grown and the properties of these lattices are discussed in chapter 3. With the insights from the previous section, we will now take a look at the in-situ growth of hexagonal indium lattices. Unfortunately, the in-situ reflection was not recorded for the sample analyzed in chapter 3, so a new sample was grown under the same conditions, except that the growth time first was decreased from 33 to 20 seconds for the new sample. After this first deposition on the sample, only some very small nuclei formed (see SEM image in appendix C, figure A4). Thus, a second deposition was done in this same sample, now for 45 seconds. SEM images of this sample after two depositions are shown in figure 4.12. Figure 4.12a shows the sample before the lift-off of PMMA, where the walls surrounding the particles are around 350 nm high. The PMMA and top layer of SiO2were removed by sonicating the sample in acetone
for 30 minutes at 30°C. A SEM image of the sample after lift-off is shown in 4.12b. We see that the lift-off was not completely successful, as in some parts the layers remained. In the parts where the lift-off was successful, the pillars are sticking out of the SiO2. This image
also shows that not all the holes are filled, and these empty holes are mostly in groups of multiple. Empty holes can be due either to the substrate itself, maybe they were filled with left-over SiO2 that was not completely etched away, making the surface not conductive.
Another possibility is that the electric field was not homogeneous, which lead to lower potential in some part, making deposition not possible. We can also see that, although the conditions were the same as for the sample of chapter 3, there are some differences. In this new sample, not all particles fill up the entire hole, they do not touch the walls all around the particle, whereas the old sample mostly filled up the holes perfectly. Also, the new sample has some particles with very narrow, pointy tips, an interesting phenomena we did not see in the old sample. It is difficult to say what has caused these differences. Maybe it has to do with the fact that the new sample was made in two depositions in stead of one or maybe there were some differences in the electric field due to the flexibility of the counter electrode.
Figure 4.13: Chronoamperometry (current vs time) graph of the electrochemical growth of a SCIL templated indium lattice.
The chronoamperometry graph of the new samples can be found in figure 4.13. The curve looks very different from the one shown in chapter 3, it has much less features and resembles the curves for unconstrained growth more. It also reaches a maximum in current in the very beginning after which the current decreases because a diffusion limited state has been established.
During the growth of this new sample, 10 reflection spectra per second were recorded, just as with the unconstrained growth. This gives the opportunity for a comparison of the reflection trends between the templated and unconstrained growth, which is shown in figure 4.14. In the beginning of the growth we see the same trends in both graphs. In the
CHAPTER 4. INSIGHTS INTO NUCLEATION AND GROWTH WITH IN-SITU
MEASUREMENTS 39
Figure 4.14: Comparison of in-situ reflection spectroscopy of a sample with unconstrained growth (A) and a sample of SCIL templated hexagonal lattice growth (B). In figure (A) the yellow line is at time 0, and the red line is after 15 seconds. In figure (B) the cyan line is at time 0 and the dark blue line is after 45 seconds.
templated growth, we see the same absorption peak at 300 nm again, but the intensity is a bit lower. We also see scattering peaks emerging in after a couple of seconds that look very similar to the unconstrained growth. However, we see an important difference between the two samples after the first part of the growth between 700 and 800 nm in wavelength. A rather sharp peak in scattering arises at these wavelengths which first is getting larger in intensity as it is broadening, after which the intensity is increasing as the peak is still broadening and shifting to higher wavelengths. We believe this peak appears at the moment that the particles growing in the holes of the template are starting to touch the SiO2walls
of the holes as depicted in the illustration in figure 3.1e. From this moment periodicity is established as the particles are set in place. Note that we see in the SEM images that not all particles touch the walls surrounding the entire particles which means there is no perfect periodicity. The periodicity of this lattice is 534 nm, the refractive index of the surrounding medium (water) is 1.33 and the angle at which this experiment was done is zero, so using equation 2.17 we can calculate the energy of the first order Rayleigh anomaly: E = ~c
n(m 2π
Λ) = 1.75eV = 710nm, which is very close to the wavelength where the peak
starts to emerge so this peak is a result of the lattice.
In figure 4.15 we take a closer look to the lattice peak found in the reflection. Figure 4.15a and b shows this peak in two different stages. A shows all spectra from time 0 to 10 seconds,where the maximum of the lattice peak is reached. B shows the spectra from time 10 to 17.5 seconds. These images more clearly show the trend of a narrow peak growing in intensity until it start decreasing in intensity and growing more broad. We observe the same trends in FDTD simulations shown in figure 4.15c and d, which simulate spectra of ’growing’ particles by showing 8 different heights of nanopillars ranging from 225 to 400 nm. A height of 400 nm seems large since these heights were not reached for the sample in 3, but for this sample the growth time was longer so it seems feasible the pillars will be higher. However, no AFM measurements have been conducted on this sample, so the height cannot be confirmed.
