Measuring heat at the nanoscale using the temperature
dependent Raman peaks of CuPc on nanodisk arrays
Master’s Thesis Erik de Rijke
VU:2539881 UVA:10685928
M Physics & Astronomy: Science for Energy and Sustainability
Supervisor: Elizabeth von Hauff Second corrector: Esther Alarcon Llado
June 27, 2019
Abstract
In this research it is investigated whether regularly structured plasmonic nanoarrays of nanodisks can be used for nanothermometry via measuring the temperature dependent Raman spectrum of Copper(II)-phthalocyanine (CuPc). For this it is first investigated which nanodisks give the highest Raman enhancement, after which the temperature-dependent Raman spectrum oc CuPc signal is investigated.
Golden nanodisks of sizes from 50 nm to 220 nm are fabricated on silicon, glass and glass with a 30 nm gold layer. The Raman enhancement is estimated using Rhodamine 6G (R6G), such that the nanoparticles with the highest enhancement factor can be found. The nanoparticles on glass were found to have the highest enhancement, with more than 7.5 and 2.5 times better than the silicon and glass with gold substrates. This difference is due to the dielectric properties of the substrates, in which the silicon and glass with gold don’t confine the electrons as well as the glass substrate. CuPc was applied to these nanoparticles, which showed that the enhancement is better for R6G than CuPc by 11.3 times. This can either be attributed to the difference in relative amount of enhanced CuPc, as it is a continuous thin film while the R6G is a monolayer directly adsorbed to the surface of the nanoparticles. The secondary option is a lower amount of resonance raman scattering, which is lower for CuPc.
The Raman spectrum of the CuPc was measured from 30 °C to 150 °C, which showed a linear temperature dependency of the positions of the 1339 cm−1, 1452 cm−1 and 1528 cm−1 peaks. The peak centers decrease in energy for higher temperatures, with varying amounts per peak. The peak width was also found to be directly dependent on the temperature, also showing a linear behaviour. Measuring the spectrum using higher laser powers showed that nanoparticles are effective heat sources, possibly heating up the substrate enough for evaporation of the CuPc (330 - 350°C). Lower laser powers were not found to have anomalous behaviour, and can be used for nanothermometry. At higher laser powers the peak center of the CuPc increases in energy in time, indicating cooling or a structural change in the CuPc. Using nanodisks together with CuPc for nanothermometry seems very feasible, and a further analysis of the behaviour of CuPc and the local temperature of the particles could extend the usability.
Contents
1 Introduction 2
2 Theory 3
2.1 Plasmonics . . . 3
2.1.1 LSPR resonance of gold cylinders . . . 4
2.2 Surface enhanced Raman spectroscopy (SERS) . . . 4
2.3 Nanothermometry with Copper(II)phthalocyanine . . . 6
3 Methods 8 3.1 Sample specifications . . . 8
3.2 E-Beam Lithography using lift-off . . . 10
3.2.1 Sample Cleaning . . . 10 3.2.2 Resist spincoating . . . 11 3.2.3 E-Beam exposure . . . 12 3.2.4 Development . . . 16 3.2.5 Evaporation . . . 18 3.2.6 Lift-off . . . 19
3.3 Scanning Electron Microscope (SEM) imaging . . . 21
3.4 Transmission measurements . . . 22
3.5 Rhodamine 6G application . . . 22
3.6 CuPc analyte layer . . . 22
3.7 Raman spectroscopy . . . 23
3.7.1 Enhancement quantification . . . 23
3.7.2 Controlled temperature measurements . . . 24
4 Results and discussion 25 4.1 Plasmonic enhancement . . . 25
4.1.1 Example analysis of R6G and glass . . . 25
4.1.2 R6G on silicon . . . 28
4.1.3 R6G on glass . . . 30
4.1.4 R6G on gold on Glass . . . 30
4.1.5 CuPc on Glass . . . 31
4.2 Nanothermometry . . . 32
4.2.1 Hotplate reference measurements . . . 32
4.2.2 Heating by laser . . . 36
5 Conclusion 41
6 Acknowledgement 42
A Supplementary Figures 45
1
Introduction
Determining the temperature on a small scale is becoming increasingly more important, for example in applications like medicine, nano-electronics or photovoltaics [1, 2]. Several tech-niques are available, but are often lacking in some way or another, e.g. resolution, speed or limited applicability [3]. Examples of current nanothermometry methods are: scanning thermal microscopy, fluorescence intensity, fluorescence time correlation and finally Raman spectroscopy.
In this research Raman spectroscopy is used in combination with the dye Copper(II)-phthalocyanine (CuPc) which has a highly temperature dependent Raman signal. Measuring the Raman spectrum of the CuPc allows for determination of the temperature of the CuPc by determining the peak positions of the CuPc. By depositing only several nanometers of CuPc, the temperature of the CuPc will mainly be determined by the underlying surface. Furthermore, the surface sensitivity of surface enhanced Raman spectroscopy can then be exploited to only probe the CuPc on the surface. This in turn allows Raman spectroscopy to determine the temperature of the surface on a scale only limited by the laser spot size. The main limiting factor for Raman spectroscopy is the signal intensity, which can be very low, especially for smaller volumes and laser powers.
To improve signal intensity, Raman spectroscopy can be combined with plasmonic nanoparti-cles (NPs), highly absorbing partinanoparti-cles smaller than the wavelength of light. The absorbed light can create plasmon resonances, oscillations of electron clouds inside the NPs, which greatly increase the local electric field compared to the incident light. This improves the yield of Raman spectroscopy, of which the intensity is dependent on the electric field to the power of four. This method is called Surface Enhanced Raman Spectroscopy (SERS), and can even be used to measure the Raman signal of single molecules. Such an enhancement makes it possible to reduce the measured area and dye amount to sub micrometer scale, and consequently measure the temperature at these scales. Furthermore, these nanoparticles can also be used as effective heat sources, by using the laser of the Raman spectroscope to heat the sample during a measurement [4][5].
This research focuses on making a substrate with NPs suitable for nanothermometry using Raman spectroscopy. For this nanolithography using lift-off is used, which allows for regularly structured NPs of predefined shapes and sizes. The initial design are nano-cylinders with a diameter of 110 nm and a period of 160 nm which is imitated from Yue et al. [6]. This design is then improved upon by testing different substrates and NP sizes, in order to get the highest enhancing nano-cylinders.
By making nano-cylinders it is possible to model the structures and simulate the temperature theoretically (not done in this thesis). This provides an additional layer of reliability, by allowing to compare the simulated temperature with the measured temperature.
This thesis will first explain the theory behind plasmons, surface enhanced Raman scattering (SERS) and nanothermometry with CuPc. Afterwards the methodology of the creation of the sample is explained, with an in depth overview on nanolithography. Next, analysis methods and measurement procedures are explained. This is followed by the results, in which first the most optimal enhancement for this application is found, after which the temperature dependent behaviour of CuPc on nanoparticles is shown. The thesis ends with the conclusion on the obtained results.
2
Theory
This section contains a short overview of the theory involved in this research. First theory on plasmonic is explained, about how they arise, and some of the resulting behaviours. After that topic proceeds to Raman spectroscopy, which is the measurement technique used in this thesis. Lastly some information on Copper(II)phthalocyanine is given, together with an example Raman signal obtained on the nanoparticles fabricated.
Theory on nanolithography and the underlying processes are given in the section on nano-lithography itself (section 3.2).
2.1 Plasmonics
Small nanoscale (noble) metal particles are able to interact with light in a unique and interesting way. They absorb light strongly, at wavelengths dependent on the particle size, shape and material. This can create vibrant colors and interesting optical effects, for example coloured stained glass, surface enhanced Raman spectroscopy, and LEDs.
NPs which are smaller than the wavelength of the incoming light feel the electric and magnetic field like it is (almost) homogeneous. This can cause the electrons inside the NP to collectively oscillate, creating the plasmon resonance [7]. The noble metal particles need to be larger than 5 nm to be able to have such a resonance, because otherwise the density of states will be too low [8]. The incident light can be scattered or absorbed efficiently by the nanoparticles [9]. See figure 1 for a sketch of a plasmonic nanoparticle which is excited with light.
Figure 1: Schematic representation of light interacting with a small metal sphere [10]. The electric field of the light can cause a collective oscillation of the electrons inside the NP, which is called the plasmon resonance.
The oscillation of the electron cloud inside the NP creates a strong electric field around the NP, which can be many times higher than the electric field of the incident light [11]. This field is often concentrated at a small location on the NP, called a hotspot. Where the E-field is concentrated is highly dependent on the shape, size and material of the electric field. A dipole oscillation in a spherical nanoparticle can have a hotspot which coves more than 10% of the surface, while a ellipsoid or rod can have the hotspot concentrated on the tips of the particle [7]. Confining the hotspot to a smaller area also results in a stronger electric field, which has many optical applications (e.g. SERS). Often a more pointy shape has a higher and more localized electric field [12].
The resonance originates from a special condition, where polarizability rapidly increases near the resonance causing a drastic increase in the electric field. In the quasistatic approximation the Maxwell equations can be solved for a dipole on a spherical particle, which results in the electric field of Eout= E0x − αEˆ 0 ˆx r3 − 3x r5(xˆx + yˆy + zˆz) (1) where E0 is the incident electric field vector, ˆx, ˆy, ˆz are the usual unit vectors [13]. The
polarizability α can be expressed as
α = εi− ε0 εi+ 2ε0
a3 (2)
with a the radius of the particle, εi the (wavelength dependent) dielectric constant of the NP
and ε0 the dielectric constant of the surrounding medium. Once εi = −2ε0 the polarizability
explodes, which in turn causes an extreme enhancement of the electric field around the NP (assuming no dampening forces). For the noble metals of gold and silver the size of the NPs can be tuned such that this condition is satisfied within the visible wavelengths of light. This is due to the quantum confinement, which changes the dielectric properties of the NP [14]. For higher order resonances the condition changes slightly, for example the quadrupole resonance has the polarizability of
β = εi− ε0 εi+ 3/2ε0
a5 (3)
For a more in-depth theoretical approach, see the paper of Kelly et al. [13]. Non-spherical particles and higher order modes are often difficult to analytically calculate, and need to be simulated to obtain the resonance conditions or the electric field.
Placing particles close together can combine the hotspot of both particles into a single hotspot. However, this is not a linear superposition, because the NPs also interact with each other at such close distances. Inter-particle distances of around 1 to 10 nm can quench the luminescence of the NPs [8].
2.1.1 LSPR resonance of gold cylinders
Of particular interest in this research are nano-cylinders. The NPs fabricated are imitated from Yue et al. [6], which have also simulated the electric field of a nano cylinder as shown in figure 2a. This shows that the light is concentrated at the edges of the disk, vaguely resembling how the electric field of a point-dipole would look. The E-field attenuates quickly farther away from the disk.
Kessentini et al. [15] has done several computations of gold cylinders, from which the position of the plasmonic peak was determined, shown in figure 2b. Increasing the diameter or decreasing the height of the cylinders red-shifts the resonance. It was found that a Chrome adhesion layer of less than 6 nm has a low effect on the shift of the resonance (maximum shift of 20 nm). In the calculations the interparticle distance is fixed to 200 nm, which is a lot larger than what is used in this research. The near-field effect between NPs red-shifts when particles are placed close together, where the peak can already shift by 60 nm when reducing spacing from 200 to 100 nm [14]. Hence the calculations have been used for initial guesses, instead of ready-to-go designs.
2.2 Surface enhanced Raman spectroscopy (SERS)
Raman spectroscopy is a versatile technique, capable of measuring the vibrational and rotational energy levels of a molecule or system. Light absorbed by a molecule generally re-emits at the original wavelength, which is called Rayleigh scattering. In a rare case it will scatter inelastically,
(a) (b)
Figure 2: a) The electric field of a 110 nm gold nanocylinder, excited with 532 nm light [6]. b) Calculated localized surface plasmon resonance (LSPR) wavelengths of gold nanodisks, with a chrome adhesion layer of 2 and 4 nm (averaged) for several disk heights and diameters [15]. The separation between the disks is 200 nm, and thus results in a low amount of interaction between the different NPs. The best cylinders for Raman enhancement have the LSPR in the middle of the excitation and emission frequencies of the system (e.g. for excitation at 632.8 nm and the Raman wavelength at 685 has the best enhancement with the LSPR at 659 nm).
which is called Raman scattering, shown in figure 3a [16, 17]. In most cases the Raman scattered light will be “Stokes” scattered light, in which a little energy is given to the molecule. In other cases some energy will be taken from the molecule, called “anti-Stokes” Raman scattering. The scattering is highly dependent on the available energy states, and is thus quantized according to the energy levels in the system. By using a single wavelength for excitation, all other wavelengths detected are either fluorescence or Raman scattering.
(a) (b)
Figure 3: a) Energy level diagram of the energy states relevant for Raman scattering [18]. b) Extinction spectra of silver and gold thin films, with the Raman spectrum shown over top [19]. The inset shows AFM measurements of the silver thin film. Optimal enhancement occurs when both the excitation wavelength and the emmission wavelength are within the plasmon peak.
per ten million). Measuring a Raman signal often requires high laser powers or long measurement times. The signal can be improved by using a high electric field, because the Raman intensity is dependent on the E-field to the fourth power;
IR= IRmolecule· |Eloc|4 (4)
where IRmolecule is the Raman intensity of the isolated molecule, and Eloc is the local electric field [20]. The ratio between the un-enhanced and enhanced signal is generally referred to as the enhancement factor (EF), however definitions may differ per author. High electric fields can conveniently be created using NPs, placed on a surface, and excited using the same light which is used for the Raman spectroscopy. Figure 3b shows the relative location of the Raman spectrum in an extinction spectra. For optimal enhancement both the excitation wavelength and the emission wavelength are within the plasmon resonance. Correct optimization even allows for enhancements of magnitudes high enough to detect single molecules, for which an enhancement of 107− 1015 is required [21][22][23].
A second option to increase the Raman signal is by exciting the analyte molecule with a wavelength in resonance with a molecular absorption band [23]. For example, Rhodamine 6G (R6G) has an absorption maximum at 530 nm, and shows a molecular resonance if irradiated with the 532 nm laser. This further increases the emitted Raman signal. Furthermore, these resonances can quench the large fluorescence which R6G has, by non-radiative interactions with the metal surface on which it is adsorbed. Resonance Raman Scattering (RRS) is generally on the order of 104− 106, and is on the order of 105 for R6G [23][24].
2.3 Nanothermometry with Copper(II)phthalocyanine
The well known dye Copper(II)phthalocyanine, has three temperature dependent Raman peaks, as shown in figure 4 [1][25][26]. This research uses these three peaks to determine the temperature of the CuPc within the laserspot. This allows temperature measurements on the nanoscale. Some of the vibrational modes of CuPc are shown in figure 5a, furthermore the molecular structure of CuPc is shown in figure 5b. The 1339 cm−1 peak is a combination of two Raman wavenumbers at 1336 and 1339 cm−1 which correspond to the A1g and B1g in-plane vibrational
modes respectively [27]. Aroca et al. [28] asigns this mode to the Isoindole stretch. The 1452 cm−1 Raman peak is assigned to the B1g and B2g modes by Basova et al. [27], and is a so called
pyrrole stretch [28]. The most highest intensity mode at 1528 cm−1 corresponds to the B1g
vibrational mode, and is assigned to the C==N aza stretch. For a more in-depth overview, see either the simulations of Basova et al. [27], or the measurements of Aroca et al. [28].
When CuPc is heated up, it expands slightly. Hamann, C. and Schenk [29] have measured using X-ray diffraction that heating up α-CuPc from 22 to 100 °C increases the lattice size by 0.5%. At 150°C the lattice is approximately 1% larger than it’s original room temperature size. The expansion differs per lattice plane, e.g. the 204 direction grows slightly slower than the 111 direction. These changes of the lattice have an effect on the Raman spectrum.
α and β CuPc There are several crystal structures for CuPc, of which the most interesting for this research are α and β CuPc. Evaporating CuPc on a substrate at room temperature results in α-CuPc. Heating the substrate during evaporation can result in β-CuPc if the temperature is high enough. Kolesov et al. [31] report to have obtained β-CuPc when CuPc is evaporated onto a substrate of 150°C. Annealing α-CuPc thin films also transforms it into β-CuPc. Ghorai et al. [25] measured that α-CuPc transitions into β-CuPc at 183 °C, much lower than other authors who have measured transition temperatures of 200 °C up to 250 °C [32][33][34]. The
Figure 4: Example Raman spectrum, measured on nanoparticles using a 532 nm laser. The three main peaks are indicated with a red dot, these peaks are temperature dependent and will shift to a lower energy with increasing temperature.
(a)
(b)
Figure 5: a) Vibrational modes of the central structure of CuPc as shown by Bovill et al. [30]. They indicate that the v3 mode corresponds to the 1530 cm−1 peak, the v28 mode corresponds to the 1452 cm−1
peak and the v4 mode corresponds to the 1339 cm−1 peak. b) Molecular structure of CuPc.
phase transition temperature is much higher for bulk-CuPc, where 300 °C is required for the transformation [35].
Figure 6 shows the Raman spectra of α and β CuPc, which are almost the same. Three small peaks at 771 cm−1, 1195 cm−1 and 1303 cm−1 are in the spectrum of β-CuPc, which are not in that of α-CuPc.
Figure 6: Raman spectra of α (a) and β (b) CuPc measured by Ghorai et al. [25]. Spectra are fitted with Lorentzian functions, and individual phonon modes are indicated with solid lines (red and orange). The phonon modes are indicated according to the theoretical calculations of Basova et al. [27]. The grey area indicates where there are significant changes in peak parameters due changes in structure by transition from α to β CuPc. β-CuPc has three additional Raman peaks, indicated in the green area.
3
Methods
This section describes which samples are fabricated, and how they are made. First a general overview is given on the fabricated samples, after which the lithography process is explained which is used to make the samples. The lithography section also contains some background information about the process itself. The last parts contain the measurement methods used to investigate the sample.
The fabricated samples are cylindrical nanoparticles on a substrate (e.g. silicon or glass), on top of which an analyte is applied for Raman spectroscopy. R6G is used as analyte to determine the enhancement, as it is applied easily and has a high intensity Raman signal. CuPc is used as analyte for nanothermometry, due to it’s temperature dependent raman shifts.
3.1 Sample specifications
The base design used are the nano-cylinders as made by Yue et al. [6], which are then modified slightly to find the optimum enhancement. In the paper of Yue et al. a silicon substrate is used, on top of which gold nanoparticles are fabricated. They use 100 nm PMMA resist, and
evaporate 5 nm titanium and 30 nm gold, where the titanium is used as a adhesion layer. They then compare lift-off with plasma etching to see which gives the highest enhancement. In their research the lifted-off samples had a higher enhancement, likely due to the sharper edges of the NPs. They fabricated 4 different samples of nanocylinders, a diameter of 110 nm and periods of 160, 210, 260 and 310 nm. These patterns were calculated to have an EF of 38, 28, 22, and 14 thousand for R6G respectively. In this research it was chosen to only reproduce the 160 nm period from this paper, as the goal is to get the highest EF. After that, various sizes and period were tested, which were not tested by Yue et al..
(a) (b)
Figure 7: a) SEM image of Design 14, dose 9 (260 µC/cm2), on a glass substrate with a gold layer on
top. Using the SEM, the NPs are measured to have a diameter of 85 nm. b) A zoomed out image in the SEM of “AuSquare”, a control area where a layer of gold is deposited, with which Raman signals can be compared. The top gray part is the AuSquare, in the middle is the identification, and the dark grey is the substrate. The characters have filled in holes (A,q,a,e,0), as these areas do not lift-off that easily.
Three different substrates are used to investigate the enhancement, <100> p-type boron silicon, borosilicate glass, and borosilicate glass with 5 nm chrome and 30 nm gold. These three substrates are then coated in 200 nm CSAR 62, and used for lithography (see section 3.2 for the methodology and appendix B, table B.1 for the sizes of the NPs fabricated). The resulting NPs are 35 nm high, of which 5 nm chrome and 30 nm gold. R6G is is applied by adsorption of a monolayer by immersion of the sample in a solution of R6G, the same method as used by Yue et al. [6]. A SEM image of one of the designs is shown in figure 7a.
For the Raman measurements of R6G, an extra square was printed on each sample of size 150 by 150 µm. This square (named “AuSquare”) can then be used as a reference measurement, because this location does not have any plasmonic effects, and has a consistent surface behaviour between the different samples. A SEM image of this reference location is shown in figure 7b.
The CuPc samples are fabricated on a glass substrate, with the same NPs of 5 nm chrome and 30 nm gold. The CuPc is thermally evaporated onto the sample after lift-off to a thickness of roughly 15 nm.
The height of the metal layer and it’s composition have not been varied. It is easy to vary the design in the two lateral dimensions, because a single sample can contain many designs. However varying height will require multiple substrates, which greatly increases the time required to make
the samples, because many steps are performed in series and not in batches.
3.2 E-Beam Lithography using lift-off
Samples are made using E-beam lithography (EBL), with lift-off as patterning procedure. Samples are cleaned beforehand using sonication, and by immersion into a base piranha solution. An electron-sensitive resist is spin-coated onto the sample, which is then exposed with electrons in a computer-programmed pattern. The sample is then developed by dissolving the exposed resist, leaving holes at these spots (a positive resist is used in this research). Afterwards metals are evaporated onto the sample, which either deposit into the holes, or on top of the resist. Lift-off is performed by dissolving the resist, leaving only the metals which were deposited in the holes. A schematic overview is shown in figure 8.
To ensure the process is successful the sample can be investigated between process steps using a SEM or an optical microscope. It is common to use the optical microscope directly after development. The SEM is often used after development, evaporation and liftoff.
Figure 8: A schematic overview of the liftoff process [36]. First a electrosensitive resist is spincoated onto the surface of a substrate. Then an electron beam is used to expose specific parts of the resist. The sample is “developed” by dissolving the exposed areas (A negative resist dissolves the not-exposed areas). Metals are deposited onto the surface by evaporation. The remaining resist is then dissolved, which also removes the metals deposited on top of the resist. The end result will be the final structure, consisting only of metals, in areas which were exposed to the electron beam.
When not in use, samples in the fabrication process are stored in a sample storage box (max 5% humidity, automatically regulated by injection of nitrogen gas). Completed samples are stored in a nitrogen glove box. This ensures that the sample does not degrade in quality due to water molecules or dirt particles.
3.2.1 Sample Cleaning
Cleaning the samples beforehand is very important, as it determines the sticky-ness of the resist and or metal, and also determines the end quality of the pattern. Furthermore, the supplied Si samples are coated with an organic protection layer, a leftover of the sawing of the original wafer which has been cut into 12x12 mm squares.
A sample from the box of new samples is first put into acetone, to dissolve most organics (like the organic protection layer). The sample is then blow dried with a nitrogen gun, and placed in a beaker of water (de-mineralized). This is repeated for as many samples are needed, after which the water beaker with the samples is moved into a sonicator.
While the samples are sonicated a highly agressive “Base Piranha” solution is prepared. This solution will remove any leftover organic residues. Note that it will also make the surface hydrophilic. The samples are put into the solution for a minimum of 15 minutes. Afterwards the samples are rinsed first in water, and subsequently in IPA (2-propanol). The rinse in IPA is because IPA dries cleaner and faster.
Base Piranha Recipe
Place glass beaker (ca. 100 ml) in fumehood on heating plate Add 50 ml demineralized water to beaker
Heat water to 75-80 °C
Add 10 ml NH4OH (30%) to the beaker
Add 10 ml H2O2 (30%) to the beaker
The mixture should bubble, and is ready for use
Piranha is highly dangerous, and should be used and disposed carefully. Protocol is to wait 24h before disposing the solution into the inorganic base waste container.
The samples are stored in a plastic sample box, which has been cleaned beforehand using a tissue with ethanol and subsequent blowing with a Nitrogen gun.
3.2.2 Resist spincoating
A electron-sensitive resist is used to “hold” a negative of the pattern, during the fabrication process. The resist consists of a polymer where the chain lengths are initially long. Exposure by an electron beam breaks the polymer into shorter chains, which are easily dissolvable in a developer. After further process steps the resist can be removed by a more powerful solvent, which removes the remaining part.
Delta 10 The “Delta 10” spincoater at the AMOLF is used for spinning resists and conductive layers onto the substrate, which is used for HDMS and the resist CSAR 62. The maximum rotation speed is 4000 rpm, the maximum acceleration is 1000 rpm/s. The machine has access to a “gearset” which reduces the layer thickness to approximately 50%, at the cost of a less homogeneous layer.
HMDS The toxic chemical Hexamethyldisilazane (C6H19NSi2) is used as an adhesion layer
for the resist (promotor/primer), and has to be applied before the resist is put on. A droplet is applied to the substrate, which forms a monolayer in 10 seconds. Quickly spinning (4000 rpm, 1000 rpm/s, 45 s, with gearset) will remove any excess HMDS by centrifugal force, leaving only the monolayer. The sample is then baked for 1 minute at 150°C. The gearset is used to make the layer as thin as possible. The surface becomes hydrophobic due to the HMDS monolayer. The actual spin recipe does not matter a lot, as long as it has spun hard and long enough to remove the excess HMDS. HMDS does not adhere to gold, as it attaches to OH groups on surfaces.
CSAR 62 The Positive resist CSAR 62 (AR-P 6200) from Allresist GmbH is used because it has a high resolution and sensitivity [37]. It’s main component is poly(α-methylstyrene-co-methyl chloroacrylate), it also contains an acid generator and is dissolved in anisole. It can both be used for lift-off and plasma etching. Lift-off is used as that provides higher enhancement as the sidewalls of the metal will be straighter [6]. The substrate does not require a post-exposure tempering, because the acid-generator is activated during exposure.
Layer thicknesses vary depending on CSAR dillution and spin rpm. The layer thickness is chosen to be at least 3 times the height of the target structure, lower thicknesses complicate liftoff and higher thicknesses reduce resolution. As the target structure is around 35 nm high, 200 nm of resist is used, roughly 6 times the feature height.
For smaller substrates (12x12 mm) 100 µL CSAR 62 solution is deposited onto the substrate. For larger substrates (24x24 mm) 250 µL is deposited onto the substrate. The sample is then spun at high rpm for 45 s with 1000 rpm/s acceleration. The rpm used determines the thickness of the resist. The gearset of the spincoater can halve the layer thickness if necessary. Layer thickness also varies depending on the viscosity of the CSAR, hence older/younger CSAR might give slightly varying results.
After spinning the sample is moved to a hotplate to be baked at 150 °C for 2 minutes. Profiler v8.2 AFM The thickness of the resist is measured after spincoating using the “Profiler v8.2 AFM” at the AMOLF. A scratch is made into the resist using a metal tweezers.
The AFM is then used to determine the height difference between the resist and the bottom of the scratch.
Elektra 92 Glass substrates are nonconductive, and thus need a conductive layer spincoated on top to provide a way for electrons to escape from the substrate and prevent charging. The substrate needs to be conductive for both electron-exposure and for SEM imaging. Available at the AMOLF is Elektra 92, a polythiopene based thin film, which is spincoated on top of the resist the sample at 2000 rpm with 1000 rpm/s acceleration for 60 s. The sample is baked afterwards for 2 minutes at 90°C. This recipe should result in a Elektra thickness of roughly 60 nm. Elektra 92 dissolves in water, and can thus be easily removed from the sample after exposure in the Voyager EBL in preparation of sample development. Elektra is also spincoated on top of finished glass samples, such that they can be viewed using the SEM.
Gold colloids Gold colloids (50-100 nm) are applied to each corner after spincoating, to help with focusing during the calibration of the E-beam exposure. Only 3 corners are required, however that does require handling the sample such that the orientation stays the same. The solution is applied onto the substrate using an adjustable pipette (0.2 µL is normal). The droplet will evaporate, leaving a small spot of gold colloids on the sample. The evaporation can be sped up by baking the substrate, e.g. 90°C for 1 minute (The CSAR is not affected by this). The spot is usually slightly “coffee-ring” shaped, where most of the gold colloids remain at the edge of the droplet. Similar to Elektra, the gold colloids can be washed off by immersion in water. 3.2.3 E-Beam exposure
Exposing the resist using electrons will change the chain length of the polymers inside the resist. These exposed parts can then be “developed”, which removes these areas. E-beam exposure is a very precise method of nanolithography, capable of making sub-light-wavelength structures on a substrate. In this research the nanolithography is limited to 2D structures (lateral dimensions), but it is possible to create 3D structures using multiple layers of resist.
Raith Voyager electron beam system The Voyager EBL system at the AMOLF is used for electron exposure. It is advertised by Raith to have a reliable and fast speed, and a high resolution. It is possible to do fixed beam exposures, but also fixed beam moving stage (FBMS) and moving beam moving stage (MBMS) exposures, which can create large fields with no stitch errors. It uses a 50 kV acceleration voltage, and has beam currents of 0.1 up to 20000 nA. The write field (fixed beam fixed stage) has a maximum size of 500 by 500 µm, in which it can create features as small as tens of nanometers.
Pattern design The locations which are to be exposed are designed in the software of the Voyager, using a 2D map (also called “Positionlist”). This map contains references to “Writefields”, sub-maps of size of 500 by 500 µm which contain the individual elements which are to be exposed (e.g. circles). The writefields are not larger, because it is limited to the area which can be exposed without moving the stage or beam (note that FBMS or MBMS can have larger writefields). The writefields can contain all elements needed for exposure (e.g. several thousands of circles), however that can slow down the pattern generator (the computer which operates the electron beam following the pattern). This slowdown can be reduced by dividing up the writefield in smaller “structure references”, smaller amounts of elements (e.g. a couple hundred of circles) in a single file, which is then referenced in the writefield many times to create the original envisioned design. Care should be taken such that there are no features which lie on the edge of the structure reference or writefield, as that may create stitching errors (incorrectly aligned features).
Proximity effect When the electrons hit the resist they do not dissipate all their energy in one spot, because they create secondary electrons on impact, as shown in figure 9a. These secondary electrons also expose the resist, causing these areas to also be removed during development. This has both a large scale effect (um scale) and a small scale effect (nm scale). The NP fields are not printed to be infinitely large, and thus have edges where the resist is not uniformly exposed as shown in figure 9b. Features closer to the edge are written with less electrons, causing them to be comparatively smaller. In some cases this means that the edges are underexposed, while the center of the field is good. In other cases the center is overexposed while the edges are properly shaped features. On a small scale the proximity effect changes the feature size depending on the feature spacing. Exposing features which are very close together will increase the size of the features, while features far apart will be smaller. The change in size was found to be on the order of 10 nm for features sized around 100 nm and spacings between 50 and 100 nm.
Substrate influence The conductivity of the substrate has a large influence on the lithography process, mainly in the form of conductivity. Electrons are repelled by other electrons, and thus low conductivity can increase feature size by deflecting incoming electrons. If the substrate is not conductive at all, then the surface charges quickly, preventing many electrons from reaching the resist. Furthermore the substrate also changes the amount of electrons scattered on the substrate surface back into the resist.
Dosetesting The proximity effect, the substrate conductivity, and the feature design have a large effect on the written features. This makes it very difficult to predict which amount of electrons should be used for exposing the design (µC/cm2). Thus it is necessary to do a “dose-test”, where the same design is exposed with multiple different doses. Afterwards the substrate can be investigated with the SEM, to see which dose properly creates the design, without under- or over-exposure, and with features of the correct size.
(a)
(b)
Figure 9: a) An incident electron (red) carries a lot of momentum, which can be transferred to secondary electrons (blue), after which it can backscatter (as shown) or stay in the substrate (yellow) [38]. The creation of secondary electrons causes exposure on areas not exposed to the electrons from the electron beam, resulting in a larger area of resist which is affected. b) SEM image of the bottom-left corner of an underexposed NP field which has been lifted off. Due to the proximity effect the NPs close to the edge of the field have a different size (and possibly shape) compared to the NPs more in the middle of the field. In this case the NPs closest to the edge are underexposed, resulting in no NPs remaining. The NPs slightly more inwards show a white color in the SEM which indicates sharp edges, also a result of underexposure.
For a 200 nm layer of CSAR 62 the working doses are generally between 100 µC/cm2and 300 µC/cm2. Densely packed patterns with low feature spacing have a small range of working doses (e.g. 120 to 140 µC/cm2). Sparse patterns with a lot of space between features are easier to print as they have a large range of doses which turn out ok (e.g. 160 to 240 or higher µC/cm2). Highly conductive substrates are less likely to be overexposed due to electron deflection by charging, which reduces electron spread due to charging. This also improves the range of doses which are not under or overexposed.
Underexposure can result in two situations. In the first situation the dose is much too low, resulting in a almost un-affected resist. Thus after development there will still be resist at the exposed areas. Liftoff will result in no NPs on the substrate, as the metals do not make contact with the substrate below the resist. The second situation is where the resists is exposed critically, just between complete underexposure and good exposure. In this case the resist is exposed from top to bottom, with the top part exposed slightly more. The result after development will be sloped side walls. Evaporating metals on these sloped side walls will result in the top metal sticking to the bottom layer attached to the substrate. This complicates liftoff, which requires both layers to be not be connected. The (partially) connected features might come off the substrate completely, or they might stick but keep part of the sidewalls. Figure 9b shows both situations, as part of the particles are removed completely, and others remain but have sharp edges as indicated by the white color. Figure 12 shows a more schematic overview of such situations.
Overexposure also results in bad NPs, as the NPs are either inter-connected or are not formed at all. A just-above-good dose will make the feature sizes just large enough that they start to overlap, causing nearby particles to be connected. An even higher dose will remove such large amounts of resist, that the remaining resist is likely to fall over or float away during
(a) (b)
Figure 10: a)SEM image of an underexposed area, taken at an angle of 50°. The “Ears” are due to the side walls of the resist to be sloped instead of straight. Subsequent evaporation of metal on the resist then results in metal on these side walls, creating non-flat NPs. The image also shows NPs which are damaged as their side walls were ripped off or because the exposed area was not perfect. b) Overexposed field after evaporation, where the resist has partly fallen over. These fields can give a high enhancement due to the sharp shapes (resist pillars), but are difficult to reproduce consistently.
development, see for example figure 10b.
Writing speed and Resolution The speed at which the features are written into the resist depends on several parameters, the dosage, the beam current, and the dwell time. The beam current can be chosen by selecting a specific “column mode” in the Voyager, of which there is only a limited selection. Furthermore the curve dwell time is dependent on both the dose and beam current, and can be calculated using the following equation;
Dose = Beam current · Dwell time
Step size · Line spacing (5) Here the step size and line spacing are parameters which dictate the x and y resolution respectively. The dwell time is restricted by a minimum value, limited by the physical speed at which the electron beam can be aimed. For the Voyager this is instructed to have a minimum of 60 ns dwell time per point. It is advisable to keep the dwell time as low as possible and the beam current as high as possible, as this will reduce the time needed to write all the features. Most fields printed in for this research use a resolution of 10 by 10 nm for step size and line spacing. Some fields were printed with a higher resolution (4 nm), to improve the corners of the shapes (squares, triangles and similar shapes).
Writing a single field of 120 by 120 µm) takes roughly 1 minute with a beam current of 0.5 nA, depending on the density of features. A full dose test can take between 9 minutes and 30 minutes, depending on the amount of doses which are to be tested. A field of 50 by 50 µm is found to be good enough for Raman spectroscopy, and thus the dosetesting can be sped up slightly.
Voyager problems It should also be noted that writing too many features (e.g. hundreds of millions of circles) is difficult for the Voyager as its memory will fill up during writing. This
slows down the writing speed, until it is so slow it does not write anymore. Increasing the writing resolution greatly enhances this slowdown as more exposure points will be loaded into the memory.
Writing large areas of the same pattern can be achieved by using the “batch-file exposure” procedure, where it saves every lens position from a single exposure and repeats that repeatedly. This procedure can only do a single dose and a single pattern (thus unavailable for dose-testing). It can be used to create samples for transmission measurements, which requires a larger effective area of NPs. Raman measurements don’t require large areas and thus does not need this procedure. Other users of the Voyager at the AMOLF generally don’t write millions of tiny 100 nm sized circles, and thus this procedure is only rarely used and should thus be used carefully. Note that the sample used must be very clean and flat on the backside, because the “height-sensing” procedure, used to level the sample by laser, is not working correctly for this
procedure (can be or possibly is repaired by Raith). 3.2.4 Development
After exposure the exposed resist has to be removed, which is done during the development step. This will result in a negative template of the design, where the exposed areas of the resist are transformed into holes.
First the sample is rinsed in water to remove any Elektra and gold colloids from the sample. Then the sample is submerged in a developer (pentyl acetate). The developer will remove the exposed resist, and slowly remove partially exposed areas. For lift-off it is necessary that the side walls have a small overhang, such that evaporation of metals on top of the sample will not result in connected top and bottom layers. For this the development time has to be adjusted, until the desired shape is achieved. If the goal is to use the resist for etching then a shorter development time is preferable (resulting in sloped or straight side walls). The recipe below was used for all samples, and has shown to give consistent results.
Development of sample
Rinse sample in demineralized water to remove Elektra and gold colloids Dry with Nitrogen gun (Optional as the surface is hydrophobic)
Immerse for 60 s in pentyl acetate Immerse for 5 s in O-xylene Immerse for 15 s in 9:1 MIBK:IPA Immerse in IPA
Dry with Nitrogen gun
The time spent in pentyl acetate can be varied, to change the shape of the sidewalls. 60 seconds was found to be good enough to form an under-cut for lift-off on the samples fabricated. Note that the sample should not be wobbled while developed, as the resist might come loose from the substrate or tear on edges of fields.
After the sample has been developed in pentyl acetate it is necessary to remove the developer with the dissolved resist which is present on the sample. If the sample is immediately transferred for rinsing, it might be possible that any dissolved resist in the developer might precipitate on the sample. For this reason a development cascade is used, where the sample is briefly immersed in the second and third (fresh) developers, to remove any leftover dissolved resist [39]. This procedure has the most effect when developing many samples, as the first developer will be consumed significantly, significantly polluting the solution. After the development cascade the sample is immersed in IPA, which will remove any leftover developer on the sample.
The developer used can be changed depending on availability, however the resist sensitivity is different per developer and thus a different development time is required. It is also possible to interrupt the development using a stopper, and then later continue later on (for up to several times)[40]. This allows for relatively easy testing of the required development time.
SEM imaging of developed samples After immersion the sample can be investigated using an optical microscope and/or an SEM. This shows whether the sample is under or overexposed and if the side walls are good. Care should be taken in the SEM, because the resist is sensitive to electrons. The resist can change shape and composition in the SEM, which influences both the evaporation step, and the lift off. Exposed areas might not give nice NPs as designed initially by using the Voyager. Carbon deposition might also contaminate the surface, and thus prolonged exposure should be avoided. Long exposure can give changes big enough to see visually during the SEM imaging.
Glass samples are nonconductive and should be respun with an Elektra layer if the sample is to be investigated in the SEM. Note that this Elektra layer will obscure part of the information available. For example, it is very difficult to make informative pictures taken at an angle, as the surface shown is the Elektra instead of the Resist. Top-down pictures can provide enough information to investigate under and over-exposure, and to make an estimate of feature sizes.
3.2.5 Evaporation
After development metals will be evaporated on the sample to create the NPs. Metal will deposit on the sample, where it will go either on the substrate in the holes, or on the resist next to the holes. A properly developed sample will have a small overhang on the side walls of the holes, which will result in no metals on the side walls themselves. This creates two separate layers on the sample, which is required for lift-off.
Nanoontje evaporator The device used to evaporate chrome and gold is “Nanoontje” a homebuilt evaporation system at the AMOLF. It uses a quartz film thickness monitor (FTM), and evaporates metals from tungsten boats via resistive heating of the boat. The tungsten boats can have a maximum of 350 A until the boats melt. The sample is mounted on a rotating plate, with a second rotating plate which serves as a shutter. The rotating plate can be moved such that the sample is directly above the evaporation source, which reduces the chance of metals evaporating on the side walls of the resist. The sample mount is very flexible and can be used to evaporate many samples simultaneously. The small sample plate used in this research can hold five 24 by 24 mm samples. It is possible to evaporate on to up to three different batches of samples differently, more batches will require re-mounting samples. Up to 4 materials can be simultaneously mounted in the vacuum chamber. The distance between the sample and the boat is roughly half a meter. The pressure before evaporation is usually around 5 · 10−7 mbar or lower, and drops to 1 · 10−6 mbar during evaporation.
This device is not allowed to be used for evaporating low vapour pressure materials, like CuPc (which has a boiling temperature of 330 °C). Evaporating materials at high temperatures will also heat up the surrounding walls, possibly evaporating other contaminants which would have a low vapour pressure.
Chrome adhesion layer To improve adhesion of the NPs to the surface of the substrate, a 5 nm Chrome layer is deposited first using Nanoontje. Chrome has a better adhesion to glass and silicon than gold, and greatly improves the handling of the sample. NPs with a Cr adhesion layer can survive megasonication, provided the connected surface of the NPs is big enough. It was found that cylinders with diameters of 50 nm and larger stay on the substrate after megasonication. Smaller sizes have not been tested.
Chrome is deposited for a total thickness of 5 nm, at a rate of 0.1 ˚A/s. The current used is around 280 A, depending on the amount of chrome in the boat. Warming up of the boat is done in roughly 5 minutes. Chrome chunks sublimate, and do not pose a risk of explosive boiling or dust explosions. After evaporation the current is tuned down over a period of roughly 5 minutes. Note that the boat stays hot for almost 30 minutes afterwards.
Gold nanoparticle layer Using Nanoontje, 30 nm of gold is deposited onto the sample after the chrome has been deposited. It is evaporated at a rate of roughly 1 ˚A/s, with a current of around 225 A. During warmup of the boat care should be taken to not cause explosive boiling. Warming up the boat in 5 minutes did not seem to cause any problems. There is no waiting period between the evaporation of the chrome and gold other than the cooling down period of the chrome and the warming up of the gold, so the surface of the sample might still be warm at the start of the evaporation of the gold. Less than 0.5 g of gold is needed to evaporate 30 nm onto the sample.
Using the Profiler v8.2 AFM the combined layer thickness of the chrome and gold was determined to be 35 nm.
After evaporation of both the chrome and gold, the sample will be partially transparent (10-20% estimated). The top part is gold in color, and the bottom is chrome (silver) coloured.
SEM imaging of samples after evaporation After evaporation it is possible to investigate the status of the sample. The most important parts to look at are the side walls, where it will be visible whether the top layer is connected to the bottom layer or not. Example SEM images are shown in figure 11. The sizes of the NPs can also be estimated, but care has to be taken not to measure the size of the hole instead of the NP. Exposing the resist to electrons in the SEM after evaporation has less effect than before evaporation, because the metal layer is already deposited. However, the resist is still changed and might prevent lift-off by failing to dissolve during the lift-off step. Non-conductive substrates do not need an Elektra layer, because the 5 nm chrome and 30 nm gold is conductive enough to produce good images. Only topologically disconnected parts might not image correctly.
(a) (b)
Figure 11: a) SEM image of a design which has evaporated correctly, the sidewalls (black) do not show any material on them. The top layer is entirely disconnected from the bottom layer which will make lift-off possible. b) This field is underexposed, causing evaporation on the side walls. This field did lift-off, but a lot of NPs (>10%) were missing and those left were NPs with “ears”.
3.2.6 Lift-off
After evaporation it is possible to do lift-off with the sample. The sample will be submerged in a strong solvent (“remover”), which will remove the resist. The top metal layer will come loose from the sample and float away into the solution. Lift-off can be very inconsistent, from two identical samples one might fail while the other will lift-off successfully. Increasing the resist thickness will increase the chance for proper lift-off. Larger nano-patterned features will also improve lift-off, because the solvent will reach the resist easier. Below is procedure shown to lift-off the fabricated samples.
Lift-off procedure
Heat anisole to 60-65 °C, using a water bath Immerse sample in heated anisole for 5 minutes Use a syringe to create turbulence and lift-off sample
Repeat 5 minutes immersion and syringe step if lift-off did not happen Softly mega-sonicate sample and syringe again if lift-off still did not happen Dip in acetone
Dip in IPA
Dry with nitrogen gun
Some samples lift-off very quickly (e.g. already after 5 minutes soaking), others need more soaking time, or might even some soft sonication. Do not use an ultra-sonicator, as that might damage the sample. Large areas of metals might look like they are “flaking off”, like bad or old paint, when it is ready to take off. Note that while the surrounding non-nanopatterned area might lift-off, the NP fields might not yet be lifted off. Inspection of the color of the fields might give a hint whether the area has been lifted off, because the optical properties change when the top layer comes off. Lift-off can not be interrupted and resumed later, once dry the sample will never lift-off again. Anisole, which is used as remover is heated up to 60-65 °C, which will improve the energy in the system. This will improve the dissolving of the resist, and will allow the solvent to more easily reach the resist. Once the resist is (partially) dissolved the sample will need to be agitated, which is done using a syringe. This will create enough force to remove the top metal layer from the sample, without removing the bottom layer which is adhered strongly to the substrate surface. If the sample does not lift immediately, it is an option to wait until more of the resist has been dissolved.
After lift-off the sample can be inspected in an optical microscope. Small fields are best viewed using a high magnification (50x or higher). The 30 nm gold will show a white, yellow or gold color (dependent on lighting and filter conditions). Dense fields with a lot of surface area of gold will show the whitest or most yellow. Areas with low metal coverage or highly absorbing areas can be gray-ish or even black. The NPs are too small to be seen by eye, thus any visible features (e.g. ridges or lines) are likely to be fabrication errors.
Megasonication In some cases the sample will not want to lift-off entirely. This might be due to the bottom layer which is attached to the top layer (e.g. bad dose, see figure 12), or it might be due to the resist failing to dissolve properly (e.g. old resist). Soft sonication might provide an outcome, as it will introduce enough energy into the system to damage the fragile connections between the top and bottom layer. It might also stir enough such that the remaining resist dissolves. Note that the syringe step is still needed, as sonication will not remove the top layer on itself. Sonication can not be performed in an ultra-sonicator, as firstly it will damage the sample as the energy is too high, and secondly anisole is flammable and is not safe to ultra-sonicate. Note that mega-sonication can also ignite the anisole though the chance is much lower, but
Figure 12: Problems which might occur during lift-off due to underexposure or bad development of the sample. If the sidewalls of the resist (orange) do not have an overhang, then the metal (black) might be evaporated onto the sidewalls. This results in a (semi-)continuous metal layer on the substrate (A). Lift-off on the sample (B) by dissolving the resist might fail and result in several situations (C). 1) The top metal layer might come off, but the sidewalls could still be standing. For the nano-cylinders this will resemble ears. 2&3) As part of the metal is connected, the top layer might lift-off partially and rotate around a connection point, or simply deposit nearby. 4) The top metal layer might not lift-off at all.
safety precautions still have to be taken.
Pollution Lifted metal flakes float around in the solution after lift-off, and might deposit onto the surface of the substrate. Lifting off multiple samples using the same solution was not found to be a problem, even though many metal flakes were suspended in the solution.
Direct redeposition of the metal foil shortly after lift-off is more likely, though still rare. Large metal pieces (order of tens of µm2 or smaller) can lift-off and directly deposit close by (<100µm). For the purpose of Raman spectroscopy this is not an issue, as the majority of the
NP field will be unaffected by such depositions if it happens at all.
Important to note is that the deposited pollution can not be removed, as it has adhered to the surface.
Note that the cleanliness needed during lift-off is lower than for the other steps, because the NPs have already been made, and are unlikely to be deformed due to a bit of dirt. It is still advisable to work as clean as possible, as this will improve the application of R6G or the evaporation of CuPc on the surface.
3.3 Scanning Electron Microscope (SEM) imaging
The AMOLF has several methods available to investigate the sample on a nano-scale. Available are several SEMs, and an atomic force microscope (AFM). The SEM can be used easily for all conductive substrates, and with an extra conductive layer like Elektra it can also be used for nonconductive substrates (though those might be imaged better with the AFM). One of the SEMs available at the AMOLF is the FEI Helios which was used for imaging in this research. The SEM is capable of imaging with a resolution of 0.9 nm with an acceleration voltage of 15 kV. Beam current can vary from 1.5 pA up to 20 nA, and is usually set to 86 pA in this research.
Acceleration voltage For this research an acceleration voltage of 5 kV was used, which is shows the features of interest clearly (edges of nano-patterned features). The acceleration voltage will determine the penetration depth of the electrons, and thus determines whether the surface or subsurface elements are imaged. A higher acceleration voltage might be used to image the substrate when it has been coated with Elektra, as the electrons will penetrate trough the conductive layer to show the lithography pattern underneath. A high voltage will reduce the yield of secondary electrons. A lower acceleration voltage will show more surface detail, but will reduce the electron yield from backscattered electrons.
Carbon deposition The vacuum in the SEM is never perfect, and as such there are hydro-carbons in the vacuum chamber. Furthermore, some hydrohydro-carbons might also be present on the surface of the sample itself. The high-energy electrons can crack these molecules, after which the carbons can diffuse to the surface and deposit [41]. Contamination can be reduced by reducing the exposure time. Contamination can also be reduced by improving the vacuum, which can be done waiting or by cleaning the microscope (e.g. cleaning the vacuum seals, or a bake-out of the e-beam column).
3.4 Transmission measurements
Transmission measurements are done at the VU University Amsterdam, with a spectrometer (PerkinElmer Instruments Lambda 900 spectrometer). This device has a spot size of around 5 by 5 mm (First release of the spectrometer in september 2000). Maneuvering the sample into the spot size is difficult, because the mounting area is limited and there is no direct view of the light spot on the sample (a mirror is required for proper inspection). A good signal requires an area of NPs of around the spot size, larger is preferable due to the mounting difficulty. The system is able to do reflection measurements (and thus measure e.g. silicon substrates), however this was found to be difficult to do due to the age and complexity of the system. Furthermore, this also requires a large enough sample. The signal is collected using an integrating sphere setup.
Creating a sample fit for transmission requires a fully completed dosetest in order to discover the correct dose, for the design in mind. Afterwards the sample for the transmission measurement can be made, where an area of 7 by 7 mm takes roughly 12h of writing time in the Voyager, for the designs in this research.
Due to time constraints only one sample was made for transmission measurements. The result is shown in figure 15 in the results.
3.5 Rhodamine 6G application
When necessary, Rhodamine 6G (R6G) is applied to lifted-off samples, using the method as described by Ye et al. [42]. An aqueous solution 1 · 10−3 M of R6G was prepared, which can be space-efficiently stored (dark-red transparent color). This solution is used to create a 20 times more dilluted solution (5 · 10−5 M, neon-orange transparent color) in which the sample is soaked for 3 hours. This will create a monolayer adsorbed to the gold surface on the sample (R6G also adsorbs easily to silver). Afterwards the sample is rinsed with ethanol. R6G dissolves easily in water and even more so in ethanol, so cleaning will remove any R6G not directly adsorbed to the surface, leaving only the monolayer. There is no visual change after adsorption of the R6G.
3.6 CuPc analyte layer
For samples which need a CuPc analyte layer (after lift-off), the thermal evaporator at the VU is used. This device is located in a Nitrogen filled glovebox, and is capable of evaporating CuPc.
For this a ceramic crucible is used, placed in within a heating element. The sample is mounted on a rotating sample holder, in front of which a shutter is placed. The sample holder rotates during evaporation to create a homogeneous layer, because the crucible is located not directly beneath the sample. It is likely that the CuPc will deposited on the sides of the NPs using this evaporation chamber.
The CuPc is heated at a rate of 15 °C/min up to 330 °C. The evaporation temperature is around 330 to 350 °C [43], at which point it sublimates. CuPc is evaporated at a rate of 0.01 ˚A/s, until 15 nm is reached as read on the monitor (set to density 1.62 g/cm3 and z-ratio 9.999). After evaporation the sample has a blue-ish color, indicating that the CuPc absorbs a non-negligible amount of light.
Evaporating CuPc on a room temperature substrate (as is done in this research) results in α-CuPc [1][33]. Heating up the substrate can change the crystal structure of the deposited CuPc and creates β-CuPc [33]. Both crystal forms are a herringbone pattern, where the difference is the angle of inclination of the molecular plane with respect to the crystal plane [1].
The original goal was to evaporate the CuPc before lift-off onto the sample, however due to difficulties with the evaporator at the AMOLF this step has moved to the VU. This resulted in the CuPc being evaporated after lift-off. Evaporating the CuPc before lift-off will result in CuPc only directly onto the nanoparticles. This simplifies the system, making it more predictable. Furthermore all the Raman signal will then come from CuPc directly on top of the NPs, which improves the temperature measurements directly of the NPs.
3.7 Raman spectroscopy
Raman measurements use the “Renishaw InVia Raman microscope”. The setup uses a microscope objective to focus the laser on the substrate. The measurements use a 63x objective. The image is focused using a camera in the device, using features present on the substrate (e.g. pattern edges). Individual cylinders are too small to be seen in an optical microscope and can thus not be used for focusing.
The Raman spectrum is fitted using Lorentzian peak functions, which show a better fit than Gaussian peaks for the measured spectra. Raman peaks are best described with Voigt peaks (convolution of a Gauss and Lorentzian peak), however these are computationally very expensive,
and the gained precision is negligible for the purpose of determining the peak center [44]. 532 nm Laser The setup has access to a 785 and 532 nm laser, from which the 532 nm laser is used. The 100mW 532 nm Laser has been installed by Renishaw in april 2019, and is relatively new. Most measurements (all except the Silicon and Glass R6G measurements) were measured using this laser. Earlier measurements also used a 532 nm laser, but that laser was at the end of it’s lifetime (very low power output). The laser is calibrated before each measurement using the Raman signal of a Si testing sample. The software or hardware limits the power output to a limited selection. The settings allow 100, 50, 10, 5, 1, 0.5, 0.01 percent laser power or lower. It is not possible to select for example 30% laser power. Laser power lower than 0.01% requires significant measurement times (>60 s) to get a signal with a good SNR on the tested samples. 3.7.1 Enhancement quantification
To quantify the enhancement a simple procedure is done. Measurements are taken on several spots on the same sample, including a reference location called “AuSquare” (see figure 7a for a SEM image of this location). AuSquare is a location on the sample which consists of an “infinite” square of smooth metal (same composition as the NPs, 5 nm Cr and 30 nm Au). Peak heights of Raman signals on the NPs are then compared (normalized) to the peak height of the Raman
signal of AuSquare. This allows for a simple quantification of the relative enhancement of the NPs, compared to a non-enhancing structure. Using the golden reference square also keeps the amount of R6G consistent in the reference location. E.g. R6G does not adsorb to glass, and as such using the sample substrate directly will mean that glass can’t be compared to e.g. silicon.
The equation used to calculate the relative enhancement is rEFωi =
px(ωi)
pAuSquare(ωi)
(6) where px(ωi) is the peak height of location x at wavenumber ωi, and similarly pAuSquare is the
peak height of AuSquare. For R6G the 612 cm−1 peak is used, as this peak is the biggest, and can be fitted easily with a low error. For similar reasons, the 1528 cm−1 peak is used for CuPc. True EF An important note to make is that the amount of adsorbed R6G depend on the amount of surface area of the gold. This means that there is comparatively less R6G on the NP locations, as the gold is not 100% surface filling. This means that the relative enhancement is an underestimate of the actual enhancement factor, because the signal of the NP fields is coming from less molecules of R6G compared to the AuSquare.
Actual quantification of this difference is difficult, as there are two contributions to the Raman signal; un-enhanced R6G and enhanced R6G. Both change in amount and the enhanced R6G also changes in signal intensity between different locations, which makes it difficult to separate the contributions to obtain the actual enhancement. For the CuPc it might be done, because the amount of CuPc is equal everywhere, however there is still a difference the amount of exposed glass behind the NPs which changes light reflection. A simple estimation for the CuPc would be to subtract the peak height of the AuSquare measurements from the NP measurements, which would result in only the enhanced signal. In other words, rEF−1. Such an assumption would not include re-absorption by the CuPc or light which is reflected or shadowed by the NPs. 3.7.2 Controlled temperature measurements
In order to quantify the temperature dependency of the peak shift of CuPc, the sample will be measured on a hotplate. This homemade device consists of a hotplate controller and a hotplate. The controller reads the temperature directly of the hotplate itself. When setting the temperature on the hotplate, the final temperature on the controller will likely show several degrees lower than what was set (e.g. setting T to 100 °C will result in a shown temperature of 95°C). Furthermore, the actual temperature of the hotplate is even lower (by about 5 °C). If the hotplate has already been heated up, then the temperature is likely to be above what was set on the controller. Due to this hysteresis and inaccuracy of the controller, a PT100 is used to determine the temperature of the hotplate. It was found that the temperature was stable within 1 °C, even though there is a large mismatch between the settings of the hotplate and the actual temperature.
4
Results and discussion
This section contains the results obtained, where the Raman signal was measured on several samples. The first part contains the search towards the best nano-cylinders, which will eventually be used for the nanothermometry measurements with CuPc. The second part contains these nanocylinders, with CuPc on top. The temperature dependency of CuPc is measured on a hotplate, and after that plasmonic heating is measured by using a high laser power on the plasmonic particles.
4.1 Plasmonic enhancement
To ensure that the substrate can be used for nanothermometry, a high EF is desired. Three substrates are compared, silicon, glass and glass with a metal layer (5 nm Cr, 30 nm Au). On top of these substrate various NPs are fabricated of different sizes and periods (50-220 nm diameter). Of these NP fields the relative EF is calculated using equation 6. This can then be used to determine the highest performing combinations of substrate and nanoparticles.
First an example is given how the enhancement is quantified, with an in-depth analysis of parameters which influence the result. After that the enhancement of R6G on silicon, glass, and gold on glass is compared. This shows that NPs directly on top of the glass are the best at enhancing, from the measured substrates. The relative enhancement of CuPc on glass is shown afterwards, for the substrate which will be used for nano thermometry.
4.1.1 Example analysis of R6G and glass
Rhodamine 6G is measured using the (now replaced) 532 nm laser, and can easily be seen due to the resonance Raman enhancement (order of 103). Measurements using the 785 nm laser only show noise, due to a lack of enhancement. The measurements are done at the lowest laser power possible, which still gives a good signal (0.1% laser power for Glass, 1% for Gold on Glass and 100% for the Si substrates). Measurements take 10 s on a single spot, after which a different spot is taken to check for consistency and to reduce signal decay. The following part is an example which shows how measurements of R6G look like, and what parameters have an influence on the signal.
Figure 13: Molecular structure of Rhodamine 6G (R6G). For the enhancement analysis the 612 cm−1
peak is investigated, which corresponds to the vibrational mode involving the xanthene ring and the phenyl ring. For an in depth theoretical analysis of the R6G Raman modes see Watanabe et al. [45].
