323

P. Chandra et al. (eds.), Next Generation Point-of-care Biomedical Sensors

Technologies for Cancer Diagnosis, DOI10.1007/978-981-10-4726-8_14

X. Zhou (*) • T.I. Wong • J. Deng

Institute of Materials Research and Engineering, A*STAR (Agency for Science, Technology

and Research), 2 Fusionopolis Way, #08-03, Innovis, Singapore 138634, Singapore

e-mail: [email protected].edu.sg

L.L. Sun

Temasek Microelectronics Centre, School of Engineering, Temasek Polytechnic,

21 Tampines Avenue 1, Singapore 529757, Singapore

Point-of-Care Device withPlasmonic

Gold Nanoarray Sensing Chip

forBiomarker Detections

XiaodongZhou, TenItWong, LingLingSun, andJieDeng

14.1 Introduction

LSPR is generated on metal nanostructures upon the illumination of light whose

energy can be absorbed by the nanostructures and cause collective electron charge

oscillations on its surface. LSPR traps the light on the metal nanostructures surface

within tens of nanometers, where the light is strongly enhanced by tens to hundreds

of times. LSPR is particularly benecial for chemical or biological detections.

Because the sizes of the chemical or biological molecules are usually in a few nano-

meter range, they will be attached to the metal nanostructure surface roughly within

the coverage of the strong LSPR eld, thus the LSPR signal variation is relatively

specic to the molecule attachment, i.e., the molecules not bound to the metal nano-

structure will not much affect the detection signal, thus raw samples could possibly

be tested without wash or purications (Anker etal. 2008; Lee and El-Sayed 2006;

Sherry etal. 2005).

There are two methods to use LSPR for biosensors. One is to detect the shift of

the LSPR absorption due to the refractive index change upon molecule binding.

This is a direct assay that requires less time and cost; however, for some applica-

tions, the sensitivity might not be enough. The other is to utilize the LSPR to excite

the uorescent labels in a sandwich assay, because the labeled sandwich assay is

intrinsically more sensitive than direct assay, and LSPR eld is 10–100 times

stronger than the incident light, the LSPR amplies the uorescent labels and

324

achieves quite high sensitivity. Since both methods can be applied on the develop-

ment of a point-of-care device, the selection of detection methods is based on sen-

sitivity requirements for different applications. The development of a LSPR-based

point- of- care device has a ow chart illustrated in Fig. 14.1. The rst step is to

identify the nanoparticles and the related bioassay that can meet the requirement of

the specic application. After which, some plasmonic simulation is required to

identify the plasmonic peak and the LSPR eld distribution, thus the suitable light

source can be found and the size, shape, and inter-distance of the nanoparticles can

be optimized.

LSPR can be generated by metal nanoparticles in solution, or metal nanostruc-

tures (random or periodic nanoslits, nanobars, nanoholes, or nanopillars/islands) on

a substrate (Kuwata etal. 2003; Mayer and Hafner 2011; Murphy etal. 2005). We

established our point-of-care device with periodic gold nanoarray on glass sub-

strate, because the substrate-based LSPR chip can be washed with incorporated

microuidic channels to reduce nonspecic binding, the periodic nanostructures

have much narrower LSPR bandwidth that helps improving the detection sensitiv-

ity, and gold resists oxidation and extends the shelf-life of the fabricated sensing

chips. Furthermore, gold functions with different kinds of biomolecules and eases

the biofunctionalization of the nanostructures. However, periodic gold nanostruc-

tures are difcult to be fabricated, so we have developed a whole process of fabri-

cating the gold nanopillar or nanohole array on glass.

According to the simulated plasmonic eld as well as the simulated and charac-

terized LSPR wavelength of the nanochips (Ghaemi etal. 1998; Wu etal. 2012a;

Wu etal. 2012b), the bioassay is established on the top of the gold nanoarray, and a

microuidic channel with samples ow through the nanochip will be fabricated.

The microuidic chip can be actuated and detected by the automated point-of-care

device, and the point-of-care device will also quantify and report the concentration

of the biomarker in the specimen.

14.2 Nanochip Design, Fabrication, andCharacterizations

14.2.1 Design oftheGold Nanoarray

We rst designed the gold nanopillars and gold nanoholes with various pitches and

sizes, and the relation of the LSPR peak location to the size, pitch of the gold nano-

hole arrays are shown in Fig.14.2. The gold nanostructures can be designed by

plasmonic simulations according to the Maxwell equations, which are applicable

for the electromagnetic eld and transmission/refraction/absorption calculations for

metal nanoparticles larger than several nanometers. Commercial software available

Nanostructure

simulation

Nanochip

fabrication

Assay

decision

Microfluidic

chip

Biofunction

-alization

Point -of-care

development

Fig. 14.1 The development process of a LSPR-based point-of-care device

X. Zhou et al.

325

for such simulations includes COMSOL, CST Microwave, etc. For the ease of mass

productive nanofabrication of the sensing chips, we simulated the gold nanopillar

and nanohole arrays on glass substrate, by setting the frequency-dependent dielec-

tric constant of gold using the data from Palik handbook (Palik 1998), and the

refractive indices for quartz, air, and water were taken as 1.52, 1.0, and 1.33,

respectively.

1E6

R = 75

= 100

water

gold

glass

water

gold

glass

Hole @ 540 nm

Max = 11.815

Pillar @ 540 nm

Max = 25

(75,100)

(60,150)

(100, 90)

=(R, T)

(50, 400)

(125, 90)

(150, 90)

(175, 90)

(190,90)

050 100 150 200

Intensity (W/m

)

x (nm)

400 500600 700800 900

Wavelength (nm)

gold thin film

sensor structure

optical power absorption [%]

660

650

640

630

620

610

100150 20

diameter (nm)

1.0

0.8

0.6

0.4

0.2

0.0

400 500600 700800 900100

Transmission

Absorption

Reflection

Wavelength (nm)

mode α wavelength (nm)

Power (W)

670

660

650

640

630

620

50 100 150 200 250 300

thickness (nm)

720

680

640

600

560

300 350 400 450 500

pitch (nm)

mode α wavelength (nm)mode α wavelength (nm)

Fig. 14.2 Plasmonic peak and electromagnetic eld for water side light illumination. (a) is at

various gold nanohole dimensions with a xed pitch of 400 nm; (b) gold nanohole array

(p=400nm, R=75nm, T=100nm) showing the α and β modes; (c–e) the α peak shifts of gold

nanohole array in (b), when only varies the pitch, diameter, or gold thickness, respectively; (f) The

gold nanopillar array simulation with p=320nm, R=75nm, and T=50nm

14 Point-of-Care Device withPlasmonic Gold Nanoarray Sensing Chip

326

For biosensing, the gold nanoarray will be immersed in buffer with its reective

index similar to water. Plasmonics are generated at the top rims (at the water–gold

interface) and the bottom rims (at the gold–glass interface) of the gold nanostruc-

tures, and plasmonics on the top rims contributes more to the biosensing. The simu-

lated top rim plasmonic eld intensity is presented in Fig.14.2a for gold nanohole

arrays with a xed pitch of 400nm and various radii (R) and thicknesses (T). The

gold nanohole array with R=75nm and T=100nm yields the strongest plasmonic

eld of 9.538×10

W/m

(in our simulations, the incident power was 1nW within

a quarter of a single nanohole area, i.e., 200nm×200nm area). Figure14.2a also

indicates that the strongest plasmonic signal is at the rim of the gold nanoholes, and

this intensity drops to a half when X-axis distance is 6 nm away from the rim.

Figure14.2b further demonstrated that two plasmonic modes α (at around 647nm)

and β (at around 730nm) will be generated at the top and bottom rims of the gold

nanoholes, respectively. In comparison, in Fig.14.2b, 100nm gold lm on the glass

substrate presents no obvious plasmonic peaks.

The inuences of the pitch, diameter, and thickness to the wavelength of the α

mode in Fig.14.2c–e indicate that the pitch will inuence the absorption the most,

and 400nm is the most suitable pitch for having a plasmonic peak at around 647nm

which is suitable for red uorescent dye excitations. Figure 14.2c–e shows that

200nm of pitch variation (from 300 to 500nm) will cause a linear peak wavelength

shift of 180nm (from 555 to 735nm), 50nm of diameter variation (from 125 to

175nm) will cause a wavelength shift of 10nm (645–655nm), while 50 nm of

thickness variation (from 50 to 100nm) will cause a wavelength shift of 27nm

(from 620 to 647nm). It is also well known that engineering errors are unavoidable

in the gold nanostructure fabrications. So for different nanohole parameters, the

plasmonic peak absorption variation of the α mode can be roughly estimated by Δλ

=(∂λ/∂p)×Δp+(∂λ/∂D)×ΔD+(∂λ/∂T)×ΔT, with ∂λ/∂p=0.9, ∂λ/∂D=0.2 and

∂λ/∂T=0.54.

The gold nanopillar array with p=320nm, R=75nm, and T=50nm is also

used in our work, and its transmission, refraction, and absorption peaks in Fig.14.2f

present two plasmonic peaks at 540 and 690nm. For our work of using green light

to excite the quantum dot (QD) label in the sandwich assay, this kind of gold nano-

pillar array provides twice stronger plasmonic eld (by comparing the insets in

Fig.14.2b, f), thus outperforms the gold nanohole array of p=400nm, R=75nm,

and T=100nm.

14.2.2 Gold Nanoarray Fabrication

Based on our design, the gold nanohole array (with p = 400 nm, R = 75 nm,

T=100nm) and gold nanopillar array (with p=320nm, R=75nm, and T=50nm)

are our targets for nanofabrication. For point-of-care device, the gold nanoarray is

expected to be disposable, thus the mass fabrication of the gold nanochip is a key

enabling technology for point-of-care devices. We have developed a mass fabrica-

tion method as illustrated in Fig.14.3. It includes two parts, the rst is the nickel

X. Zhou et al.

327

GOLD NANOPATTERN

MASS FABRICATION

Nano patterns design for electron

beam lithography.

Electron beam resist coating on

silicon substrate.

Electron beam lithography process

and development.

Metal seeding layer, coating on the

sample.

Nickel plating process and separation

of the nickel mold and substrate.

Photoresist coating on glass wafer

and use the sample together with

the nickel mold for nanoimprinting.

Nanoimprinting and follow by UV

curing inside UV oven.

Gold depostion and then the lift-off

process.

RIE to remove the residual of

photoresist.

Fig. 14.3 Mass fabrication of metal nanostructures on glass wafer. The left side indicates the

nickel mold fabrication process, while the right is the gold nanostructure mass fabrication

process

mold fabrication, and the second part the gold nanoarray mass fabrication via nano-

imprinting (Wong etal. 2013).

Electron-beam (e-beam) lithography is a technology of using focused e-beam to

expose the resist which is sensitive to the e-beam (Altissimo 2010; Chen 2015). The

e-beam lithography can reach a resolution of a few nanometers, and it is a revolu-

tionary nanotechnology for obtaining precise and controllable nanostructures in any

shape. E-beam lithography has been heavily applied in industry and nanotechnol-

ogy research, for fabricating the accurately designed nanopatterns in plasmonics,

semiconductors, surface property modied materials such as for anti-fouling or

anti-fogging, and photonic waveguides. Although nanoimprinting can replace the

e-beam lithography in some cases, its master mold is produced by e-beam

lithography.

In our work, the e-beam lithography is used to fabricate the nanochips in two

ways: the rst is the fabrication of the gold nanostructure array directly by e-beam

lithography (Song etal. 2015a), the other is to make a nickel mold which is used to

get the gold nanostructure array through nanoimprinting (Wong etal. 2013; Song

etal. 2015a; Song etal. 2015b). In either case, for reducing the cost of nanochip

fabrication, e-beam lithography is conducted with 25 pieces of 1cm×1cm dies

arranged on one 4″ wafer. The chip arrangement is shown in Fig.14.4a. Each die

has its central 1.8mm×1.8mm area lled with fabricated 140nm×140nm sized

14 Point-of-Care Device withPlasmonic Gold Nanoarray Sensing Chip

328

nanohole or nanopillar array. We fabricated the nanoarray in square shape because

e-beam writes the squares slightly faster, and after nanofabrication, the gold

nanopatterns will have some round corners anyway.

The main caveat of the e-beam lithography, however, is its slow writing speed,

which substantially increases the cost. E-beam writing time includes the beam set-

tling time (i.e., the alignment time and beam shift time, as well as the time for beam

adjustment) and e-beam resist exposure time. According to the size of the nanopat-

terns, we optimized process and managed to reduce the e-beam resist exposure time

by using the writing voltage of 100keV, eld size of 300μm, lower dot map of 20k,

and a comparatively high current of 800pA on a single Gaussian beam (ELS-7000)

e-beam writing system. We did a series of experiments to compare, and found that

a minimum dose of 178μC/cm

for positive ZEP e-beam resist is required, in order

to get a reasonably high quality e-beam pattern for nanopatterns with 150nm size.

Compared with the normally used current of 200pA and dosage of 320μC/cm

for

e-beam writing on the ZEP, 800pA current and 178μC/cm

dose combination will

reduce the resist exposure time to 1/7 (Deng etal. 2016). This is very important,

because for the 4″ wafer e-beam writing shown in Fig.14.4a, even with our acceler-

ated e-beam writing recipe, it took 32h to complete. For the negative e-beam resist

NEB22, it can be exposed at 800pA current with 110μC/cm

dosage. Following the

140 nm

Pitch = 400 nm for

nanohole array,

Pitch = 320 nm for

nanopillar array

1.8 mm

………….

…………………

Diameter = 4 inch

1 cm

Fig. 14.4 (a) The arrangement of nanochips on a 4″ wafer with 25 dies, and scanning electron

microscope (SEM) images of the fabricated gold (b) nanopillar and (c) nanohole arrays under

10,000× magnication, where the scale bar represents 2μm

X. Zhou et al.

329

e-beam exposure, the e-beam resist was developed for a few minutes to get the

nanopatterns.

In order to reduce the cost while keeping the high precision of the designed gold

nanohole array, nanoimprinting technology was adopted for the nanochip fabrica-

tion, because nanoimprinting can achieve sub 5nm resolution with high throughput

for nanolithography (Austin etal. 2004).

After the nanoarray pattern was written by e-beam lithography on silicon wafer,

the wafer was used for electroplating to get a nickel mold. To fabricate the nickel

mold, a thin metal seed layer was coated on the patterns to initiate the electroplating

process. Two-step plating was adopted: at the beginning 12μm of nickel was slowly

electroplated on the silicon wafer at a low current density of 0.7A/dm

, followed by

a high current density of 12A/dm

to electroplate up to 300μm of nickel. After

which, the nickel mold was separated from the silicon wafer. Trion plasma etching

system with O

gas was used to clean the nickel mold for 3min, and the silicon

wafer was recycled (Wong etal. 2013).

To nanoimprint, UV curable photoresist was selected, taking its advantage of using

a lower pressure (10bar) and lower temperature (as low as room temperature) for the

imprinting process compared with thermal nanoimprinting (30–40 bar, >100 °C),

which greatly reduces the rate of wafer breakage in nanoimprinting. The nickel mold

was rst coated with an anti-sticky layer of (Heptadecauoro-1,1,2,2-tetra hydro-

decyl) trichlorosilane (FDTS), then it was nanoimprinted on the UV curable photore-

sist coated on a 4″ glass wafer with the photoresist cured by UV light source.

After nanoimprinting, the nickel mold was easily demolded, and was cleaned for

reuse. Reuse of the nickel mold is critical, because the mold is generally fabricated

through e-beam lithography which is the major cost for the process to fabricate the

nanochips. The cost saving for fabricating the nanochip is tremendous if a nickel

mold can be reused, and nickel mode is reported to be durable for 10,000 times of

imprinting without obvious damage (Rai-Choudhury 1997).

The residue of the UV curable photoresist on the glass wafer was removed by

etching in an Oxford reactive ion etching system. The glass wafer was place in a

Dexton Explorer e-beam deposition machine, to rst deposit a few nanometer

(<5nm) of chromium to enhance the adhesion of the gold on glass, then deposit the

gold lm on the top of the chromium layer. Finally, the resist was lifted off at room

temperature to obtain the gold nanoarray.

The fabricated gold nanopillar or nanohole arrays on the 4″ glass wafer were diced

into 1cm×1cm chips by diamond blade using a Disco dicing saw. The fabricated pat-

terns were viewed under scanning electron microscope (SEM) or atomic- force micros-

copy (AFM) for checking their morphology and quality. The fabricated gold nanopillar

and nanohole arrays are presented in Fig.14.4b and Fig.14.4c, respectively.

14.3 Point-of-Care Device Based onLSPR Wavelength Shift

A point-of-care device using the LSPR wavelength shift is developed for the mea-

surement of cardiac troponin I (cTnI) in serum, as illustrated in Fig. 14.5. In

Fig.14.5a, white light transmitted through an optical ber bundle was collimated

14 Point-of-Care Device withPlasmonic Gold Nanoarray Sensing Chip

330

and illuminated on the top of the gold nanohole array (p=400nm, R=75nm, and

T=100nm) placed inside a transparent ow cell made of thermoplastic elastomer

(Fig.14.5b). The transmitted light spectrum was collected by another ber bundle

and recorded by a UV-vis spectrometer HR2000+ (Ocean optic, Dunedin, FL, USA)

in the range of 300–1000nm. Each frame of transmission was collected using an

acquisition time of 20ms, and the nal spectrum was produced by averaging 100

frames within 2s. Spectra were analyzed and evaluated using OriginPro 9, and the

transmission peaks were identied using a Lorentz-based method (Ding etal. 2015).

SAM

Antibody

Glass

Analyte (cTnl)

Light source

Illumination fiber

Lens

Liquid cell

Sensor chip

Detection fiber

Spectrometer

020406080

Serum cTnl Concentration (ng/ml)

Peak Shift (nm)

1.5 mm

1.5 µm

15000

10000

5000

500 600 700 800

Wavelength (nm)

Transimission (arb. units)

644.07

654.85

620 630 640 650 660 670 680

Wavelength (nm)

Fig. 14.5 The point-of-care device based on the LSPR shift of a nanohole array (reproduced with

permission from reference Ding etal. (2015)). (a) Optical setup for the nanohole sensor detection.

(b) A liquid cell seals the gold nanohole array (with p=400nm, R=75nm, and T=100nm lling

a sensing area of 1.5mm×1.5mm) for the optical detection. (c) is the bioassay on the gold sur-

face, (d) shows the transmission spectrum of the nanochip upon cTnI (at 30ng/mL concentration)

binding, and (e) is the characterization curve of the detected cTnI concentration versus the LSPR

peak shift

X. Zhou et al.

331

The direct assay in Fig.14.5c was established by rst forming an amine-reactive

self-assembly monolayer (SAM) incubating the nanochips in ethanolic solution of

0.4mM of 10-carboxy-1-decanethiol and 1.6mM of 1-octanethiol for 12h at room

temperature, then washed thoroughly with pure ethanol and dried in room tempera-

ture. Then the chip was incubated in a mixture of sulfo-N-hydroxysuccinimide

(sulfo-NHS) and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) (Bio-

Rad, Hercules, CA, USA) to activate the carboxylic group of the SAM for 15min.

Next, 50μL of 200μg/mL anti-troponin antibody 560 (Hytest, Finland) in acetate

buffer at pH4.5 was spotted on the chip surface and incubated for 30min. Finally,

the sensor chip was immersed in 1 M of ethanolamine-HCl solution (Bio-Rad,

Hercules, CA, USA) for 15min to deactivate the unreacted esters, followed by a

rinse with deionized water and dried with nitrogen gas at room temperature. The

sensing chip is blocked by 1% bovine serum albumin (BSA) solution for 15min to

prevent the nonspecic binding, then 50 μL of human cTnI standard in serum

(Phoenix Pharmaceuticals, CA, USA) in the concentrations of 75, 30, 7.5, or 2.5ng/

mL was applied and incubated for 30min. The sensor chip was rinsed with PBS

solution three times after incubation, and inserted into the measurement cell con-

taining PBS buffer to record its transmission spectrum.

The binding of the cTnI on anti-troponin antibody on the sensor chip will shift

the plasmonic peak at the wavelength of 644.07–654.85nm (Fig.14.5d). The peak

at 644.07nm is the α peak in Fig.14.2b which has the plasmonic eld concentrated

on the gold–water interface where the bioassay is anchored. Thus the LSPR bio-

sensing for this plasmonic peak monitoring is relatively sensitive. Figure 14.5e

shows the characterization curve for cTnI detection, with error bars obtained by

measuring the same troponin concentrations with three individual nanochips.

The performance of this point-of-care device is found to have a full half- magnitude

width (FHMW) of 32.84nm, gure of merit (FOM) of 10.5, σ of signal at 0.017nm,

and signal-to-noise ratio equals to 256 (at 7.5ng/mL troponin in serum), and the limit

of detection (LOD) based on three times of σ is identied to be 0.55ng/mL.

14.4 Point-of-Care Device Based onFluorescent Label

Enhancement

We have developed several sandwich bioassays based on plasmonic-enhanced uo-

rescent labels to respectively detect the prostate specic antigen (PSA) for prostate

cancer, thrombin for blood clots, and procalcitonin (PCT) for sepsis. When estab-

lishing such a bioassay, a few factors are considered. The rst is that the distance of

the uorescent label should be around 10–20nm to the gold nanostructure surface,

so that the dyes are enhanced by the plasmons instead of being quenched by the

strong plasmons. The second is that the excitation wavelength of the dye must be

close or slightly shorter than the LSPR generated by the gold nanostructures, so as

to make sure that the light being used to excite the uorescent dyes are amplied by

the plasmon. Here we use quantum dot (QD) as labels, because compared with

organic uorescent dyes, the QDs are about 20 times brighter, and they have a

broadband of excitation wavelength range (Song etal. 2015a).

14 Point-of-Care Device withPlasmonic Gold Nanoarray Sensing Chip

332

These three kinds of biomarkers are all detected with a sandwich assay on the

gold nanoarray surface using the QD as uorescent label. However, they are still

slightly different. The rst type of assay demonstrated for PSA detection is to use

the tris(2-carboxyethyl) phosphine (TCEP) solution to cleave the PSA monoclone

capturing antibody (cAb) into a half and utilize the –SH link to bind the PSA captur-

ing antibody, which is an effective way to reduce the distance between the QD label

and the gold surface (Song etal. 2015a). The second is to use aptamer to capture the

thrombin, because compared with capturing antibodies, aptamers are relatively

cost-effective and with high repeatability, durability, reproducibility, and it is rela-

tively easy for the surface regeneration when the nanochip is planned to be reused

(Song etal. 2015b). The third is to use the whole PCT capturing antibody to capture

the PCT on the gold surface (Sun etal. 2016). All these three kinds of bioassays can

be detected by a microscope or a camera, and further the point-of-care device can

be fabricated based on imaging of the uorescent labels.

14.4.1 Bioassay forPSA Detection withCleft Antibody

PSA is a US FDA approved cancer biomarker indicating the risk and existence of

prostate cancer for males, with a cutoff value of 4ng/mL for prostate cancer screen-

ing. Since the antibody is about 14nm high, a standard sandwich assay will make

the QD labels probably more than 30nm above the gold nanostructure surface and

hinder the enhancement by the plasmons. In order to further reduce the distance

between the QDs and the gold surface, in our method, the whole capturing anti-

body was cleft into a half by TCEP solution (Song etal. 2014). This was achieved

by adding 5μL of TCEP into 50μL of 250μM cAb in PBS (1.26mg/mL, 8.4μM).

After incubating at room temperature for 30 min to allow the mixtures to fully

react, G25 column was used to purify the cleft cAb. The cleft cAb solution was

stored at 4°C, and diluted to 50μg/mL in 4-(2-hydroxyethyl)-1-piperazineethane-

sulfonic acid (HEPES) buffer (prepared by mixing 10mM HEPES with 150mM

NaCl) before use.

Compared with the whole capturing antibody, the cleft antibody is about half of

its size, and since the –SH link will be attached to the gold surface, the cleft cAb are

with their fragment antigen-binding sites facing upwards on the gold surface. The

cleft cAbs in an orientation controlled state have more binding sites compared with

the randomly orientated whole cAb, which will further increase the sensitivity for

biosensing. With the cleft cAb, a sandwich assay of “cleft cAb– PSA– biotinylated

dAb– QD with streptavidin” was immobilized on the gold surface (Fig.14.6a), and

the distance of the gold nanostructure to the QD is about 20–25nm, which is within

the distance for the QDs to be effectively excited by plasmons (Song etal. 2015a).

The QD-655 has an emission peak at the wavelength of 655nm. Practically, we

used green light at the wavelength of 540nm to illuminate the chip from the bioas-

say side and detect the QD emission from the same side of the chip. As the light

sheds from the top of the chip, the plasmonic elds for the gold nanohole array

(p=400nm, R=75nm, and T=100nm) and gold nanopillar array (p=320nm,

X. Zhou et al.