3 Design of Nano-scaled SRR
3.4 Material Used in SRR Design
For mm-scaled SRR, cupper is printed on the circuit board (so-called PCB) [13, 17, 19. 22. 26]. As scaling down, other metals have been utilized such as gold [4, 23, 38], silver [23, 36], aluminum [23, 35, 37], and nickel [38]. Aluminum is the most commonly chosen metal in silicon VLSI and has a higher bulk plasma frequency than gold or silver [23]. However, in this thesis work, gold is chosen to form the SRR array due to its low loss and acting closer to Drude model even at very high frequencies [34]. Moreover, it is widely utilized for its inertness, resistance to oxidation, capable of providing a platform for bio-molecular immobilization such as self-assembled monolayer (bio-compatible).
As mentioned before, LHMs were first fabricated on PCB. Later, silicon [23, 35, 37, 38] has been utilized as a common substrate. Recently, LHMs using cupper pattern on glass have been reported [39, 40]. In this thesis work, glass is chosen because it
enables transmission measurement in the interested frequency range, i.e. around 10 μm while silicon exhibits a strong absorption at 9 μm. Other ways to benefit transmission measurement include “releasing” pattern from the substrate [36, 38]. By choosing proper photo resist and sacrifice layer, SRRs can be embedded in a transparent sheet of PR.
Table 3-1 summarizes the design of the ring resonators in our experiment.
Table 3-1 The parameters of the square single-ring SRR designed for this thesis work. The metal is goal and the substrate is glass. For each SRR size, corresponding CRR and 2-cut SRR are also designed and studied.
unit cell size (nm) 600 900 1200 1500 1800
line width (nm) 120 180 240 300 360
gap size (nm) 120 180 240 300 360
lattice constant (nm) 1200 1800 2400 3000 3600
array size (μm) 90 90 90 90 90
array size (number) 75×75 50×50 38×38 30×30 25×25
Chapter 4
Fabrication and Results
4.1 Pre-patterning
The two sides polished 0.7 mm thick 4” Corning 1737 glass wafers were cut into 2×2 cm2 pieces as the substrates. Each piece of glass was then cleaned through the following standard process:
The positive photoresist (PMMA or ZEP520) was coated on the substrate by spin-coater. The glass was thereafter baked at 170-180° C for a while in order to enhance adhesion between the PR and the substrate. Because glass is nonconductor, a thin layer of E-spacer was coated on the PR after the bake for preventing the charging effect and thus enabled the successive e-beam exposure. The E-spacer is considered as a better choice for improving electrical conduction than the use of ITO because of its simple process and smooth surface achieved. Usually the aspect ratio of PR patterns should not exceed 5. If the aspect ratio value is too high, the pattern will collapse during the drying process after rinsing. For the line width of 120-360 nm, the resist thickness of 300-400 nm is reasonable. Three recipes were used for coating the PR and E-spacer layers, which are schemed as below:
(1) Resulted PR thickness: 400nm
(2) Resulted PR thickness: 300nm
(3) Resulted PR thickness: 300nm
The hot plate baking time in case (3) was extended to about 150sec for improving the adhesion and conformity of the PR layer.
4.2 Patterning with E-beam Lithography
Patterning was completed by e-beam lithography using an Elinoics ELS-7500EX high precision compact e-beam writing system. When the electron beam strikes a point on the resist that covers the specimen substrate, the spatial distribution g(r) of the electrical charge provided by the beam irradiation over the resist is approximated by the sum of the two Gaussian distribution functions as follows:
) ( ) ( )
(r g r g r
g ∝ f +η b (4.1) wheregf(r)=1/(πσ2f)exp(−r2/σ2f)is the forward scattering distribution mainly due
to the resist, gb(r)=1/(πσb2)exp(−r2/σb2)is the backward scatting distribution mainly due to the substrate, and η is the sensitivity ratio between the backward and forward scatting. For example, when a fine electron beam of Vacc=50keV strikes the
Si substrate covered with a 300nm thick resist, the mean square radii of the forward scattering and backward scattering will approximately be σf=13nm and σb=10μm respectively. This means that the term gf(r) changes sharply in relation to the distance r, whereas the term gb(r) constitutes a near-constant background in g(r). Accordingly, one only needs to consider σf when exposing nano-meter scaled patterns. Likewise, when exposing patterns of several micron-meters in size, one only considers σb. Due to this Gaussian beam feature, the designed pattern should not have the same width as the target pattern, or the resulted pattern would be wider than the target pattern (e.g.
by 13 nm for the Si substrate) at each side.
The field size for the ELS-7500EX can be varied between 75-600 μm, which should be chosen according to the position resolution requirement and the exposure time limit. Namely, a smaller field size constitutes a finer position resolution, while demands a linger exposure time. Given the pixel size and resist sensitivity, the electric charge required for sensitizing the pattern could be calculated. For example, for a 600 μm field size with 60,000×60,000 scan steps, the pixel size is 10 nm. And for ZEP520, the resist sensitivity is 30-60 μC/cm2 when being exposed with Vacc=50keV. It then needs 60μC/cm2×(1×10−6cm)2 =6×10−17C of charge to sensitize the pixel of PR.
The beam current and exposure time can then be assigned to achieve enough charges.
When using a larger beam current, a shorter exposure time is needed, while the electron forward scattering will also be larger, resulting in a larger beam diameter, i.e.
eventually a larger obtained line width. Test exposure should be done to find the suitable exposure time, which usually differs from the preliminary estimate. Factors that affect the optimum exposure time include: (1) substrate material, (2) background exposure distribution due to the backward scattering, and (3) pattern arrangement.
In order to optimize the exposure condition of this e-beam lithography system, exposure tests were made with several different conditions, i.e., two beam currents
(50/ 100 pA) and three field sizes (150/ 300/ 600 μm), while a scan steps of 60,000×60,000 was fixed. It was found that when the beam current and field size were both large, the position resolution was poor especially for a small line width, as shown in Figure 4-1 and Figure 4-2. There were certain shifts for both vertical and horizontal positions, which were more severe for smaller line widths and became less observable for larger line widths. This shift could be corrected by adjusting the pattern design. Figure 4-3(a) shows the modified layout for a SRR unit cell, and the improved exposure results are shown in Figure 4-3(b) and Figure 4-4. Notice that the sharper edges of large line width patterns (Figure 4-2 and Figure 4-4) compared with those of small ones (Figure 4-1 and Figure 4-3 (b)) revealed the resolution feature of e-beam lithography. Furthermore, overexposure was a serious problem for large beam current and field size. When the beam current and field size were small, the line width variation was small, as soon as reaching an enough dosage. This gives a better line width control of the target pattern. The chosen exposure recipe for our experiment is list in the following table:
Table 4-1 Exposure recipe accelerate voltage 50 (keV) field size 150 (μm)
scan step 60,000×60,000 (pixel) beam current 50 (pA)
dose time 0.85 (μsec/dot)
During the exposure experiments, strips of copper tape were attached from the edge of the glass to the metallic holder for the ground connection, to ensure the elimination of charge accumulation.
After e-beam exposure, the specimens were first rinsed by DI water for removing
the E-spacer, and dried by blowing nitrogen gas. The developer used depended on the PR. For ZEP520, the specimen was soaked in N50 for 45 seconds, but for PMMA, the specimen was soaked in an organic solution of MIBK:IPA=1:3 for 70 seconds. The IPA rinsing was then used, followed by nitrogen blow for drying.
Figure 4-1 The SEM image of a shifted CRR with line width=120 nm, beam current=100 pA, field size=600 μm, dosage=1 μsec/dot, PR coating recipe (2).
Figure 4-2 The SEM image of a shifted CRR with line width=360 nm, beam current=100 pA, field size=600 μm, dosage=1 μsec/dot, PR coating recipe (2).
(a) (b)
Figure 4-3 (a) Adjusted layout for a SRR with line width=120 nm. (b) The SEM image of the adjusted SRR after development. Line width=120 nm, beam current=100 pA, field size=600 μm, dosage=0.9 μsec/dot, PR coating recipe (2).
Figure 4-4 The SEM image of the adjusted SRR after development. Line width=360 nm, beam current=100 pA, field size=600 μm, dosage=0.9 μsec/dot, PR coating recipe (2).
4.3 Thermal Evaporation of Gold and Lift-off
A gold thin film was deposited on the resist-patterned glass substrate by thermal evaporation. Generally, the thickness ratio between the photo resist and the metal thin film (i.e. Au) must be larger than 10:1 to ensure the lift-off process. For the PR thickness of 300-400 nm, a metal thin film under 30 nm was required. Notice that the metal film can not be too thin, namely, it should be thicker than the skin depth. The bulk resistivity of Au is 2.44μΩ-cm, so that its skin depth at 30THz is about 10nm.
Therefore a gold thin film of 30nm thickness should be suitable for our devices. For the Au deposition, the background pressure in the vacuum chamber was pumped down to 2×10−6 Torr prior to the process, and would be slightly increased to2.6−2.7×10−6 Torr during evaporation. To physically remove the unwanted residual PR in the pattern region after development, oxygen plasma treatment was engaged prior to the Au thin film deposition. For PMMA, the removal rate was about 10 nm/min, and the duration of oxygen plasma process was about 1-2 minutes without excessively decreasing the PR thickness. Au evaporation was achieved by resistive heating of the target when passing 100A current, and the deposition rate was about 0.14-0.15 nm/sec. The deposition time was controlled by a coating thickness gage, and during the process the temperature of the chamber was about 100°C.
The lift-off process was performed by soaking the specimens in the ZEP520 PR remover, ZDMAC for about 10 minutes, and then rinsed by IPA and DI water. A square was drawn around the patterned area by a tweezers prior to dipping in ZDMAC, and the specimens were gently shaken by hand during the lift-off process to accelerate the peeling of the PR. The results of the lift-off process were examined by both optical microscope and SEM. Our experiment showed that the quality of lift-off result varied according to different PR coating recipes. Figure 4-5 shows the
successful results of a lift-off process for recipe (1), in which only PR in the patterned region was removed, thus the Au layer above was lifted off. The charging effect induced by the uncovered glass can be seen evidently, this is a useful indication for a completed lift-off process. Figure 4-6 shows the results for a failed lift-off process for recipe (2). A possible reason was the residual PR in the patterned region caused by an incomplete development process, such that the Au pattern was partly peeled off during the lift-off process because of poor adhesion. Figure 4-7 to Figure 4-21 show the fifteen samples contained Au patterns that were processed using the above mentioned lift-off technique from recipe (3). Finally, Figure 4-22 is the 3D SEM image of a SRR array. The overall results were pretty good. Among thousands of patterns in each array, less than five were damaged. Some specimens were immersed in acetone and sonicated, in order to shorten the process time. This process however resulted in poorer completeness of the arrays, i.e. up to ten unit cells came off, which is a more than twice higher damage rate compared to the best one. To summary, our fabrication experience and results showed that Au thin film has adequate adhesion on glass that could be survived through the lift-off process, which thus is a feasible way to produce nano-scaled devices for bio-optical studies.
Figure 4-5 The SEM image of a SRR array after lift-off. Beam current=100 pA, field size=600 μm, PR coating recipe (1).
Figure 4-6 The SEM image of a SRR array after lift-off. Beam current=100 pA, field size=600 μm, dosage=0.9 μsec/dot, PR coating recipe (2).
Figure 4-7 The SEM image of a CRR array after lift-off. Line width=120 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-8 The SEM image of a 2-cut SRR array after lift-off. Line width=120 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-9 The SEM image of a SRR array after lift-off. Line width=120 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-10 The SEM image of a CRR array after lift-off. Line width=180 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-11 The SEM image of a 2-cut SRR array after lift-off. Line width=180 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-12 The SEM image of a SRR array after lift-off. Line width=180 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-13 The SEM image of a CRR array after lift-off. Line width=240 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-14 The SEM image of a 2-cut SRR array after lift-off. Line width=240 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-15 The SEM image of a SRR array after lift-off. Line width=240 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-16 The SEM image of a CRR array after lift-off. Line width=300 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-17 The SEM image of a 2-cut SRR array after lift-off. Line width=300 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-18 The SEM image of a SRR array after lift-off. Line width=300 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-19 The SEM image of a CRR array after lift-off. Line width=360 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-20 The SEM image of a 2-cut SRR array after lift-off. Line width=360 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-21 The SEM image of a SRR array after lift-off. Line width=360 nm, beam current=50 pA, field size=150 μm, dosage=0.85 μsec/dot, PR coating recipe (3).
Figure 4-22 The 3D SEM image of a SRR array after lift-off. Line width=360 nm, beam current=50 pA, field size=150 μm, dosage=1.2 μsec/dot, PR coating recipe (3).
Chapter 5
Measurement and Results
5.1 Measurement
Most optic measurements of μm- or nm- scaled metamaterials were done by the FTIR (Fourier transform infrared spectrometer) [30, 36, 37, 38, 39]. In this thesis work, both reflection and transmission spectra of the specimens were measured carefully using the micro scoped-FTIR Hyperion 2000 manufactured by Bruker. The measurement range of this FTIR extends from k=370-25000 cm-1, which covers the middle-infrared (MIR), near-infrared (NIR) and visible (VIS) region by using different beam splitter and detector materials listed in Table 5-1. The wavelengths and wave numbers specifying the three spectral ranges are listed in Table 5-2. The microscope provides optical image of the sample that allows the light to impinge on the desired area when the devices area was very small, for example, 90×90 μm2 arrays in our experiment. Moreover, the Hyperion 2000 equips with a grazing incidence objective to allow the light impinged obliquely at a range of 52.2°-84.2°
which is useful and convenient for reflection study. The light path of the grazing incidence objective is illustrated in Figure 5-1.
In Chapter 2, we introduced the four frequently studied orientation and polarization combinations of the SRR (Figure 2-5 that is presented here in Figure 5-2 again for convenience). Measurement geometries were designed to match these four types of combinations in Figure 5-3. Due to the mechanical limitation of our equipment,
in-plane incident is impossible. Instead, we used the grazing incidence objective to achieve horizontal quantity of the incident light and vertical quantity of the magnetic field (Figure 5-3 (a) and (b)). Polarization was set either 0° or 90° to fulfill the required direction of the electric field. Owing to the weak signal achieved by the grazing incidence objective, the measurement conditions were set differently for
Table 5-1 Beam splitter and detector materials used in Hyperion 2000 FTIR.
beam splitter detector
MIR KBr MCT/DLaTGS
NIR CaF2 InGaAs diode
VIS quartz Si Diode
Table 5-2 The wavelengths and wave numbers specifying the NIR, MIR, and VIS spectral ranges. Most of the present spectral analysis are located in the range k=4000-400cm-1, i.e., the MIR region.
λ(μm) k(cm-1) MIR 2.5-50 4000-200 NIR 0.78-2.5 12800-4000 VIS 0.4-0.7 25000-14300
Figure 5-1 The light path of the grazing incidence objective in Hyperion 2000.
Table 5-3 Measurement conditions vertical incidence oblique incidence aperture 1.5 (mm) 8 (mm)
resolution 4 (cm-1) 8 (cm-1) scan time 32 (time) 64 (time)
Figure 5-2 SRR in the four nontrivial EM field propagation directions and polarizations. The figure is imaged from [36].
Figure 5-3 Measurement geometries that correspond to the four combinations in Figure 5-2, accordingly. Reflection was measured through (a)-(d), while transmission was measured only for (c) and (d).
oblique and vertical incidence as listed in Table 5-3. Despite the aperture set, the measurement window was narrowed down to approximately the array size, i.e. 90 μm, for both oblique and vertical incidence. Transmission measurements were done only for geometries (c) and (d) since it is not available for oblique incidence using our equipment.
5.2 Analysis procedure
Because the intensity of the incident light varies with the impinging and polarization angles, there is a need to calibrate and normalize the measured signal intensity, in order to facilitate a unified comparison among various spectra. This was done by drawing a slope from the lowest point to the highest point for each spectrum.
The lower intensity then can be compared to the higher intensity with a meaningful reference.
To highlight the spectral features that are caused by interactions between the incident light and the patterns, the measured spectrum were divided by the background spectrum. The background of the reflectance was a gold-coated reference specimen, while the background of the transmittance was the glass located outside the patterned region. The reflection/ transmission spectra that have been normalized and divided by the proper background spectra are called normalized reflectance/
transmittance.
Expected EM responses for the three types of ring resonator patterns under different measurement geometries are listed in Table 5-4. The main factor that influences the response of the resonators is whether the magnetic field is perpendicular or parallel to the pattern surface. Another factor is the orientation of the pattern, i.e. the gap-bearing side of the resonators, with respect to the electric field.
Table 5-4 Expected responses for the three types of ring resonators, i.e., SRR, CRR, and 2-cut SRR, under the four measurement geometries depicted in Figure 5-2 and Figure 5-3. E stands for the electric resonance, M stands for the magnetic resonance and C stands for the electric coupling effect.
(a) (b) (c) (d)
SRR E/M E/M/C E E/C
CRR E E E E
2-cut SRR E/M E/M E E
5.3 Results and Discussion
The measurements were essentially carried out twice for each of the three samples, and the obtained spectra are rather identical, indicating a good experimental reproducibility. Figure 5-4 shows the normalized transmittance of the SRR array for geometry (c). The transmission dips correspond to the peaks of the normalized reflectance for the same SRR array and measurement geometry shown in Figure 5-7, while strong absorption occurred for frequency below about 60 THz. This might be a transmission property of the Conning glass wafer, namely, the glass is opaque for part of the MIR region. The expected magnetic resonance wavelength 6-18 μm, i.e. 16-50 THz in frequency, was blocked by this substrate absorption. Thus, we will present only the normalized reflectance spectra in the following discussion.
0 5 0 100 1 50 200 2 50
Figure 5-4 Normalized transmittance of the SRR array for geometry (c).
Figure 5-5 to 5-8 show the normalized reflectance of the SRR array for the four measurement geometries. For all measurement geometries, there is a reflection band from 24 to 36 THz, while at higher frequency, another reflection peak can be clearly
Figure 5-5 to 5-8 show the normalized reflectance of the SRR array for the four measurement geometries. For all measurement geometries, there is a reflection band from 24 to 36 THz, while at higher frequency, another reflection peak can be clearly