Background

Photonic crystal (PC) is defined as a relatively high refractive index layer of modulated periodic structure, and it is surrounded by two relatively low refractive index materials. When a collimated and polarized white light projects on the surface of the periodic structure, the PC can interact strongly with the specific optical wavelength, and the electric field of PC will be redistributed. Then, the specific wavelength is reflected back with a particular angle according to the light’s incident direction, while the remained wavelengths light transmit through the PC structure. The reflected optical wavelength is regarded as optical resonance wavelength, and the electric field of PC gratings can be enhanced many times than the initial incident light. Since Yablonovitch proposed the concept of photonic crystal [1], PC have been used extensively in various fields, including PC-based waveguide [2], fiber [3], back reflector [4], and light-emitting diode [5].

With the rapid development of life science, the methods for gene detection, protein-binding analysis, small molecules, and DNA detection are urgently needed. PC-based biosensors are studied and fabricated widely. The principle for PC-based biosensors is the interaction between the liquid sample and the enhanced electric field of PC. Basically, there are two kinds of PC-based detection methods: label-free detection [6, 7] and fluorescence-enhanced detection [810]. However, all the PC-related devices rely on the enhanced electrical field magnitude on the surface of the PC gratings, which is known as evanescent field. The functional range of the evanescent field is under 300 nm from the PC surface [11], which means that the tremendous space of the PC-based bulk device cannot make a contribution to the detection result. Moreover, for PC-based chemical, biological, and medical applications, the analytes are either required to flow through optical sensor or bind to the surface of PC gratings by diffusion [12, 13] so as to use the evanescent field on the PC surface. With traditional bulk sensors, it may take several hours for the analytes to reach the surface of optical flow sensor due to the laminar flow [14, 15], especially in low concentration detection [16, 17]. Thus, the PC-based bulk sensors that require large volume of analytes and need a long diffusion time may not be adaptable for future detection among proteins, genes, and small molecules. However, PC-based nanofluidic sensors paved a new way for efficient and high-performance detection [1820]. The nanofluidic channel can guarantee the detection area be confined around the evanescent region of PC structure, which can improve the sensor’s sensitivity and reduce sample volume. Also, due to the large surface to volume ratio, the analytes in the nanochannels can flow to the channel walls easily.

In this paper, a novel PC-based nanofluidic structure has been designed and fabricated with CMOS-compatible process [21]. PC-based nanofluidic structure is an important CMOS-compatible part, and it can be used to realize the lab on a Chip concept for parallel and high through-put applications. The PC-based nanofluidic structure was fabricated on a 4-in. Si wafer, with Si3N4 as the high refractive index layer and SiO2 as relatively low refractive index substrate. A PDMS stamp was used to directly seal the PC gratings and create the nanochannels. The PC gratings were patterned by PMMA with e-beam lithography, and then the grating patterns were transferred to a thin chrome layer. The chrome was used as hard mask to transfer the grating pattern precisely to the Si3N4 film. All the related parameters had been analyzed and calculated during the fabrication processes.

Methods

Principle of PC-Based Nanofluidic Structure

The proposed nanofluidic sensor structure is shown in Fig. 1a: the PC structure is configured by periodical Si3N4 layer as guided layer and a flat SiO2 layer on top of Si wafer as substrate, and PDMS is used as cladding layer to seal the PC gratings. As shown in Fig. 1b, the inlet and outlet of the nanofluidic structure were drilled from the back side of Si wafer. With this structure, the resonance wavelength energy can be largely coupled into the guided layer and reduce the leaked energy around resonance wavelength of the PC. The Si3N4 layer thickness is d0, and the grating depth is dg. The thickness of SiO2 layer is dsub, and it was used as relatively low refractive index for PC. As discussed in [22], there were several parameters influence the resonance wavelength of PC-based nanofluidic structure. The subwavelength pitch and the remained thickness of Si3N4 determine the peak wavelength position of the PC, while the grating depth dg can be used to tune the PC bandwidth. The resonance wavelength generated when the collimated and polarized incident light source matches with phase-matching condition [23], as shown in Eq. (1):

Fig. 1
figure 1

Principle of PC-based nanofluidic structure. a PC-based nanofluidic structure with related parameters: SiO2 thickness dsub, grating period Λ, channel depth dg, Si3N4 thickness d0. b Schematic diagram of PC-based nanofluidic structure

$$ \frac{2\pi}{\lambda}{n}_{\mathrm{eff}}=\frac{2\pi}{\lambda}{n}_{\mathrm{c}} sin\theta \pm m\left(\frac{2\pi}{\varLambda}\right) $$
(1)

where λ is the incident light which couples into the PC, \( {n}_{\mathrm{eff}} \) is the effective refractive index of photonic crystal, \( {n}_{\mathrm{c}} \) is the refractive index of the cover layer, θ is the incident angle, and \( m=\pm 0,1,2\dots \) is the diffractive order. In order to simplify the structure for fabrication and experiments, the incident angle is always vertical to the surface of the PC, and the fill factor of the grating is 0.5. So, the equation can be simplified as Eq. (2):

$$ \uplambda =\varLambda\ {n}_{\mathrm{eff}} $$
(2)

,which shows a concise relationship among resonance wavelength, PC grating pitch, and effective refractive index. The effective refractive index can be roughly regraded as weighting average refractive index of Si3N4, channel liquid, and PDMS.

Simulation Result and Analysis

As the shape of gratings were rectangular, the gap between PC gratings can be used to fabricate channels when the gratings’ top is sealed with PDMS. With this novelty design, the PC gratings were used as nanofluidic channel. FDTD Solutions from Lumerical Inc. had been used for electromagnetic field simulation. Based on our former design method that had been published in [24], the parameters of PC-based nanofluidic structure had been optimized [25]. The gratings period was set as 400 nm, which is a reasonable value for PC fabrication with resonance wavelength located at visible wavelength range. The reason for fixing PC’s grating period is that it is difficult to tune the pitch among fabrication processes, while it is relatively easier to vary the Si3N4 thickness d0 and grating depth dg. By tuning the Si3N4 thickness and grating depth, the PC resonance wavelength can be changed according to the requirements. Then, the remained parameters that need to be optimized in FDTD is the Si3N4 thickness d0 and grating depth dg. Particle swarm optimization algorithm in FDTD had been used to select the appropriate PC structure parameters.

During FDTD simulation, the incident light was a TE-polarized plane wave that ranges from 400 to 700 nm, and the direction of the incident light was perpendicular to the top surface of PC plane, as shown in Fig. 2. The fixed parameters of PC structure were grating period Λ = 400 nm, the thickness of SiO2 substrate dSiO2 = 2 μm, and the refractive indexes of Si3N4, PDMS, and channel liquid were set as 2.02, 1.45, and 1.33, respectively. Then, the thickness of Si3N4 d0 was varying as 500, 450, 400, 350, and 300 nm, while the grating depth dg was changing as d0, d0 − 50 nm, d0 − 100 nm, …, 100 nm, 50 nm. By using different Si3N4 thickness d0 and grating depth dg, a series of PC structure were obtained. The electric field distribution and enhancement factor of all these PC structures were obtained by running FDTD simulation software, and the PC structure which can yield an appropriate electric field distribution and enhancement factor was chosen as our sensor structure. The related PC based FDTD simulation process can be found in our former published papers [24, 26].

Fig. 2
figure 2

Illumination of photonic crystal-based nanofluidic structure simulation

The final optimized results were shown as follows: PC period Λ was 400 nm, Si3N4 thickness d0 was 450 nm, and grating depth dg was 350 nm. With the grating depth located at 350 nm, most of the channel area was inside the active region of evanescent field, and the performance of PC based nanofluidic structure can be improved with this structure.

As shown in Fig. 3a, when the light source is collimated and TE polarized, the PC’s spectrum has two peaks around visible range at normal incidence. The peak at 698 nm was the zero order mode TE00, and the peak at 595 nm is the first order mode TE01. In application, visible light is much easier to capture during the label-free detection. So, the first order peak was selected for label-free detection. As shown in Fig. 3b, the electric field intensity of peak at 595 nm has been obtained, and the red-dashed box represents the area in which it is used as nanochannel. It indicated that the enhanced E-field was almost confined in the channel area, which means that the electric field can interact with the sample in the channel sufficiently. The maximum electric field enhancement is more than 7 times larger than the initial incident light electric field. The former reported PC structures may have a much high maximum E-field-enhanced intensity. However, the enhanced area is usually located around the corners of the gratings or the inside part of the high refractive index layer [27, 28]. In the former structure, the enhanced E-field cannot interact with the sample thoroughly, which may degrade the performance of PC-based sensors.

Fig. 3
figure 3

Simulation result of PC-based nanofluidic structure, TE polarized, SiO2 thickness dsub 2 μm, grating period Λ 400 nm, channel depth dg 350 nm, Si3N4 thickness d0 450 nm. a Normalized reflection spectrum of PC, peak wavelength value (PWV) at TE00 698 nm, peak wavelength value at TE01 595 nm. b Electrical field enhancement of PC-based nanofluidic structure with PWV at 598 nm

Results and Discussion

Design Flow of PC-Based Nanofluidic Structure

All the experiments were accomplished on a 4-in. Si wafer, and CMOS-compatible process has been used for fabrication. As shown in Fig. 4, the designed flow of the nanofluidic sensor is as follows: (1) Si wafer clean, (2) 2 μm SiO2 deposited on Si wafer as PC substrate, (3) 450 nm Si3N4 deposited on SiO2 layer as PC high refractive index layer, (4) 30 nm Cr coated as hard mask, (5) 2600 nm PMMA spin coated on Cr as photoresist for e-beam grating writing, (6) etching Cr layer with patterned PMMA as mask, (7) etching Si3N4 layer with etched Cr layer as mask, (8) removing PMMA and Cr, (9) patterned inlets/outlets from backside of the wafer with optical lithography, (10) etching through Si layer with SPR 220 as photoresist, (11) etching SiO2 with Si as mask layer, (12) directed bonding between PDMS and patterned Si wafer.

Fig. 4
figure 4

Fabrication flow process of PC-based nanofluidic structure

SiO2 and Si3N4 Deposition Experiments

The double side polished Si wafer (100) was boron doped from Wafer World, Inc. A piranha process with 1:1 volumetric mixture of concentrated sulfuric acid (H2SO4) and hydrogen peroxide (wt. 30%) was used to clean the Si wafer at 60 °C for 10 min. Then, the wafer was rinsed by DI water and dried with N2 gas. For the deposition of SiO2, Trion PECVD (Plasma Enhanced Chemical Vapor Deposition) with high/low deposition rates were used. As shown in Fig. 5a, the high deposition rate is much faster while the film density of low deposition rate method is much higher and with less pin holes inside the SiO2 thin film. Since the SiO2 layer is 2 μm, the low deposition rate (320 Å/min, ±25 Å) was selected, with temperature at 350 °C, ICP power 102 W and gases of N2O, N2, and SiH4.

Fig. 5
figure 5

Deposition rate calibration. a Calibration of Trion PECVD deposition rate for SiO2. b Calibration of STS PECVD deposition rate for Si3N4

After the deposition of SiO2 film on Si wafer, a thickness of 450-nm Si3N4 film was deposited on the top of SiO2 film as guided mode layer. Since Si3N4 was used as the guided layer for photonic crystal, the thickness of Si3N4 has a critical influence to the resonance wavelength of photonic crystal-based nanofluidic structure according to Eq. (2). In order to deposit the Si3N4 layer matches with the designed Si3N4 thickness at 450 nm, the deposition rate of Si3N4 was carefully calibrated. The Si3N4 deposition process was processed in STS PECVD. The recipe for Si3N4 deposition was RF frequency 13.65 MHz, power 20 W, N2 flow rate (1960 sccm) 10%, SiH4 flow rate (80 sccm) 20%, NH3 flow rate (55 sccm) 10%, and chamber pressure 900 mTorr. In order to calibrate the deposition rate of Si3N4, a sequence of testing experiments had been done, as shown in Fig. 5b. The result from Fig. 5b shows that the deposition rate is almost constant, about 143 Å/min.

CMOS-Compatible Fabrication Process

As Si3N4 is dielectric and transparent upon visible wavelength region, it is difficult to find the nano patterns with microscope. Chrome had been selected as hard mask for Si3N4 layer, which is much easier to pattern with Cr under e-beam equipment. The coating of 30 nm chrome was done by CHA SEC-600 e-beam Evaporator, with a coating rate of 1 Å/s under vacuum condition. Then, e-beam lithography was used for patterning nano structures on the sample wafer with PMMA. Firstly, the sample wafer was cleaned by acetone, IPA, DI water, and IPA in consequence, and then dried with N2 gas. Then, PMMA was spun coated on the wafer as photoresist with Headway Spinner under 3000 rpm, and then preexposed on a 110 °C hot plate for 1 min. Then, the PMMA-coated wafer was patterned by the Zenith e-beam lithography equipment, and the designed grating patterns were written on the PMMA layer. Then, the patterned PMMA wafer sample was post exposure at 220 °C for 1 min and cleaned, as shown in Fig. 6a. With the PMMA patterns as mask for Cr etching, there are two etching methods: dry etching and wet etching. Dry etching was using reactive ions such as chloride ions to react with Cr as selectively etching, and also the ions physical bombardment was used to etch the surface of PMMA and Cr simultaneously. Wet etching was using Cr etchant to etch chrome, which is much faster but it would destroy the sidewall of nano patterns. As Cr was used as hard mask, dry etching method was much more stable and conformal than wet etching method, and the Cr was etched on a Plasma Thermal SLR 770 ICP RIE. As shown in Fig. 6b, c, dry etching is anisotropic, so the shape of grating is uniform, and the fidelity of grating structure is quite well, while wet etching is isotopically, and it can cause damage to the grating patterns.

Fig. 6
figure 6

SEM pictures of grating patterns. a E-beam-patterned grating of PMMA on top of Cr layer, b Wet etching Cr with Chrome etchant. c Dry etching Cr with STS thermal ICP RIE

The ion gas used for Cr etching was Cl2, Ar, and O2. The chemical reaction process for dry etching was as follows:

$$ \begin{array}{l}\mathrm{Cr} + 2{\mathrm{O}}^{*} + 2{\mathrm{Cl}}^{*}\to\ {\mathrm{CrO}}_2{\mathrm{Cl}}_2\\ {}{\mathrm{CrO}}_x + \left(2- x\right)\ {\mathrm{O}}^{*}\to\ {\mathrm{CrO}}_2{\mathrm{Cl}}_2\end{array} $$

Since argon is an inertial gas, the accelerated argon gas has high momentum which can be used to bombardment of the Cr surface as physical sputtering. The recipe for chrome etching was ICP 100 W, RIE 250 W, pressure 50 mTorr, Ar 4 sccm, and O2 6 sccm. The etch rate was about 4 nm/min. After Cr etching, the Cr pattern was used as hard mask for Si3N4 etching. As the etching rate between small features and buck samples was quite different, four 2 mm × 2 mm e-beam patterned samples were used for testing and calibrating. The dry etching result of Cr layer was shown in Fig. 7. The depth of the Cr layer is about 27.5 nm with grating pitch at 400 nm, and the profile of grating structures is quite uniform.

Fig. 7
figure 7

AFM picture of etched Cr layer profile

The equipment used for Si3N4 etching was STS Thermal Plasma ICP RIE. Since metal mask would contaminate the process chamber and it was hard to clean, this equipment was chosen specially for metal mask etching. The recipe for Si3N4 etching was Freon 14(CF4) 30 sccm, pressure 35 mTorr, RIE 95 W, ICP 300 W. The calibrated Si3N4 etching time was 4 min for 350 nm. In order to transfer the required feature to the Si3N4 layer, there were two different material layers that were acting as mask, PMMA layer, and Cr layer. As the grating feature was written on the surface of PMMA by e-beam lithography, the PMMA was used as mask for Cr etching. While the grating patterns transfer to the Cr layer, the etched Cr was used as mask for Si3N4 etching. Moreover, with the existence of undercutting effect, it was difficult to ensure the fill factor locates at 0.5. In order to obtain the desired fill factor and structure, a sequence of six different doses was used for e-beam lithography while writing on the surface of PMMA, as shown in Fig. 8a. Then, the dose-related samples were used for Cr etching and obtained SEM grating profiles, as shown in Fig. 8b. After the related etching of Cr and Si3N4 layer, six different Si3N4 grating patterns were sculptured, as shown in Fig. 8c.

Fig. 8
figure 8

SEM pictures of dose-related PC grating structures, six different e-beam lithography doses were used for fabrication. a PMMA gratings patterned by e-beam lithography. b Cr gratings with patterned PMMA as photoresist. c Si3N4 gratings as guided mode layer for PC, and etched with Cr as hard mask

According to the fill factor calculation results, dose 1 had been selected for e-beam lithography and final experiment, and the result is shown in Fig. 9a, b. The inspected result from AFM is shown in Fig. 9a. From the AFM profile, it can be known that the grating pitch is located at 400 nm and the grating depth is about 367 nm. As the grating pitch was 400 nm with a fill factor 0.5, it was difficult for the AFM tip to touch the bottom of the grating and have a rectangular shape, so the profile looks like V-shape. The grating shape was detected under SEM, as shown in Fig. 8b which indicates that the sidewall of the grating is good enough.

Fig. 9
figure 9

Inspection profile of Si3N4 gratings. a AFM image of Si3N4 grating depth and profile. b SEM image of Si3N4 gratings with periodic at 400 nm and fill factor 0.5

After the etching of Si3N4, the PMMA and Cr remained on the sample need to be removed. The removing process was as follows: the sample was immersed in the acetone, and ultrasonic was used to speed the dissolve, and then the sample was rinsed by acetone, IPA, DI water, and IPA in sequence. Chromium etchant from Sigma-Aldrich was used to remove remained Cr. The wet etching speed was about 40 Å per minute at room temperature. With the fabricated grating structures, the inlets and outlets of PC-based nanofluidic structure were sculptured with photolithography, backside alignment, and etching in sequence. A 5-in. ferric oxide photo mask was fabricated and been used as photomask for photolithography. The photolithography process was as follows: (1) patterned Si wafer were cleaned by piranha process and O2 plasma for 1 min at 300 W: (2) prebake at 110 °C for 2 min; (3) cooling 30 s on the cooling plate; (4) descum with O2 plasma at 250 W, 150 mTorr for 1 min, (5) spin coating HMDS and SPR-220 in sequence at 3000 rpm for 30 s respectively, and the thickness of the SPR-220 will be 5 um; (6) soft bake at 60 °C for 2 min and 110 °C for 1 min; (7) backside alignment and exposure with i-line photolithography from EVG 620 aligner. The photolithography dose was 180 mJ/cm2 with vacuum contact and 50 μm separation distance, (8) 1:5 mixture of AZ 400 K and DI water as developer for exposed SPR-220; (9) the wafer was cleaned by DI water and N2 gas. With the SPR-220 patterns on the back of the patterned wafer, the inlets and outlets were fabricated by etch through the Si and SiO2 from the SPR-220 patterned side of Si wafer.

The equipment used for Si etching was STS Pegasus ICP-DRIE, which has an etch rate of 20 μm/min with Bosch process. The whole Si layer etching process took 25 min, and the etching process would stop when the etching gases met the SiO2 layer. The etched Si pattern was used for SiO2 etching. Plasma Therm SLR-770 ICP RIE was used for the etching process, and the calibrated SiO2 etch rate was shown as in Fig. 10. The recipe used for SiO2 etching was CF4 30 sccm, Ar 4 sccm, O2 5 sccm, RIE 250 W, ICP 50 W, and pressure 30 mTorr. The RIE power was mainly used to break the chemical bond, while ICP power was aimed to enhance the plasma density.

Fig. 10
figure 10

Etching rate of SiO2 with RIE etching method

In order to fabricate the nanofluidic channel, a cladding layer was used to bond with the PC grating structure. Buck polydimethylsiloxane (PMDS) was made by mixing liquid PDMS and cross-linking agent. PDMS is a typical bonding method for polymer bonding, especially for rapid prototyping and used as plastic substrate. The buck PDMS and grating structure were treated with O2 plasma (300 W, 30 s) to make the surface hydrophilic. After the surface treatment, they were directly bonded with each other as shown in Fig. 11.

Fig. 11
figure 11

Prototype of PC-based nanofluidic structure

Discussion

As PC’s grating period Λ, grating depth dg, and fill factor are important parameters to the final performance of PC-based nanofluidic sensor, maintaining the fabricated results conform to the designed structure parameters is of key importance. In this work, in order to keep the fabricated PC structure conform with the designed parameters, Si3N4 and SiO2 deposition rates had been calibrated and deposition thicknesses were carefully controlled. The obtained Si3N4 grating thickness is 364 nm which is very close to the designed dg 350 nm, as shown in Fig. 9. Also, the measured PC grating period Λ 400 nm, which is equal to the designed grating period, as indicated from Fig. 7. To fabricate a uniform grating structure with fill factor 0.5, six different dose of e-beam process were used, and a uniform PC grating structure was yield according to the dose 2 process. A direct bonding method was used to realize nanofluidic channels between PC grating grooves and PDMS cladding layer. The bonding process is low cost and easy to implement, which is much convenient than traditional anodic bonding process.

This kind of structure is an improvement while contrast with PC-based microfluidic sensors, basically there are two kinds of applications: (1) Concentration detection: the reflected resonance wavelength of PC-based nanofluidic sensor will shift as the concentration of the liquid sample varies. There are a lot of applications, such as alcohol concentration detection, special protein concentration detection, small molecules, and even single molecule. It can be used for medical and industry applications. (2) Fluorescence enhancement: the enhanced electric field will improve the fluorescence efficiency when the incident light matches with the exciting wavelength range of the fluorescence dye. Since plenty of biological and medical applications are fluorescence based, the fluorescence enhancement can be used to improve the limit of detection for fluorescence-based experiments. The advantages of PC-based nanofluidic sensor have two basic aspects: (1) the volume of required sample is reduced to nanoscale, which can improve the efficiency of sample dose. (2) The nanofluidic channel can guarantee the detection area be confined around the evanescent region of PC structure, which can improve the sensor’s sensitivity. The function range of evanescent field is less than 400 nm, and it is the main region of enhanced electric field, which will interact with the fluid sample and obtain a resonance wavelength shift. With these unique performances, the PC-based nanofluidic structure sensor can be used for biomedical, chemical, and even clinic application applications in the near future.

Conclusions

CMOS-compatible process of PC-based nanofluidic structure had been proposed, designed, and fabricated. The nanochannels have been created by polymer bonding between periodic gratings in PC and PDMS substrate. A detailed fabrication flow process was designed to guide the top-down fabrication process, and CMOS-compatible techniques had been used to realize the structure. The PC-based nanofluidic structure, with Si3N4 as its guided mode layer and SiO2 as its substrate, has a grating period Λ 400 nm, Si3N4 thickness d0 450 nm, and grating depth dg 350 nm. The resonance wavelength at 598 nm for higher order TE mode was selected for potential label-free detection. It is the first time that a PC-based nanofluidic structure has been proposed and realized with CMOS-compatible process. The enhanced electric field is located in the center area of nanochannels, which can be further benefit for analytes interaction experiments.