All authors contributed to the writing of the manuscript. M.D.W, R.P.B., and F.Y. planned the experiments and analyzed the data. R.P.B. and W.C. fabricated the microcylinders. R.P.B and F.Y. performed the trap stiffness experiments. R.P.B and F.Y. wrote and analyzed the simulation programs. F.Y. fabricated the nSWAT devices used in this work.
Nanophotonic waveguides have enabled on-chip optical trap arrays for high-throughput manipulation and measurements. However, realization of the full potential of these devices requires trapping enhancement for applications that need large trapping force. Here, we demonstrate a solution via fabrication of high refractive index cylindrical trapping particles. Using two different fabrication processes, a cleaving method and a novel lift-off method, we produced cylindrical silicon nitride (Si 3 N 4 ) particles and characterized their trapping properties using the recently developed nanophotonic standing-wave array trap (nSWAT) platform. Relative to conventionally used polystyrene microspheres, the fabricated Si 3 N 4 microcylinders attain an approximately 3- to 6-fold trap stiffness enhancement. Furthermore, both fabrication processes permit tunable microcylinder geometry and the lift-off method also results in ultrasmooth surface termination of the ends of the microcylinders. These combined features make the Si 3 N 4 microcylinders uniquely suited for a broad range of high-throughput, high-force, nanophotonic waveguide-based optical trapping applications.
In this work, we take an alternative and complementary approach to trap stiffness optimization by creating high refractive index microcylindrical trapping particles. Trap enhancement from higher refractive index particles has not been explored in nanophotonic traps though it has been demonstrated in free space optical traps and laser fibers 22 – 24 . The new particles developed in this work are nanofabricated out of silicon nitride (Si 3 N 4 ) thin films, which offer a higher index of refraction (n = 2.0 at wavelength λ = 1064 nm) over conventional polystyrene (n = 1.57) or silica (n = 1.45) microspheres. In addition, as compared to spherical particles, their cylindrical shape allows for more overlap of a particle with the evanescent trapping field of a nanophotonic waveguide. These characteristics are congruent with stiffer traps. We characterized their trap stiffness using a nanophotonic standing-wave array trap (nSWAT) device capable of precise manipulation of a trap array 7 , 17 , 25 and found that the microcylinders provide significantly higher trap stiffness than conventionally used polystyrene microspheres. We anticipate that this approach for trapping enhancement should also be applicable to many other nanophotonic trapping functionalities in a broad range of microfluidic devices.
For biological applications, power enhancement is not a viable option when the waveguide materials cannot sustain high powers due to non-linear power absorption 21 or must be operated at sufficiently low power to minimize laser heating of the aqueous solution 4 . Thus while greater power in a device may be advantageous, a more desirable solution is power usage efficiency in applications involving biomolecules. This need requires a method that permits further nanophotonic trap stiffness enhancement.
While nanophotonic trapping potentially offers tremendous benefits over free space traps 14 , 15 , challenges remain in creating stiff traps for various manipulation applications. Thus far, significant effort has been devoted to the optimization of the coupling of a free-space laser to a given device 16 and to the minimization of waveguide propagation losses 17 , 18 . Several strategies have also been developed to enhance laser power at the trapping region and trapping efficiency. Slotted waveguides afford significant power enhancement but require trapping particles to have a size smaller than the slot 19 . Photonic crystal waveguides also provide substantial local power enhancement, but are limited in their flexibility of particle position manipulation 5 , 6 , 20 .
Over the past three decades, optical trapping has proven to be a powerful tool in the physical and biological sciences. Trapped particles, which serve as handles for biological molecules, can be manipulated with piconewton (pN) forces and nanometer (nm) precision 1 , 2 , making optical trapping especially valuable for biophysical, chemical, biochemical, and cellular studies. Recent advancements in a broad range of nanophotonic structures hold promise to enable optical trapping experiments at high-throughput on-chip 3 – 12 . These nanophotonic devices are compact and portable, and can be fully integrated with microelectronic and microfluidic components 13 , 14 .
For our previous nSWAT measurements7, 17, 25, we employed spherical trapping particles which are commercially available and widely used for optical trapping. However, microcylinders of desired dimensions are not commercially available and must be fabricated. We thus developed methods to fabricate Si3N4 microcylinders using two approaches: a cleaving method or a lift-off method. Here, we refer to the resulting microcylinders as “cleaved Si3N4 microcylinders” or “lift-off Si3N4 microcylinders” respectively. Both methods are based on high-throughput and cost-efficient deep ultraviolet (DUV) lithography to pattern at the sub-micron scale and are detailed below.
Microcylinders were cleaved from top-down fabricated micropillars composed of a Si3N4 core with a 50 nm thick hafnium dioxide (HfO2) inner shell and a ~5 nm silicon dioxide (SiO2) outer shell ( , Figure S1). This approach is inspired by our previously established protocol to fabricate quartz microcylinders26–32 and we adapted our protocol for Si3N4 microcylinders.
Open in a separate windowTo begin the fabrication process, a 4-inch standard silicon wafer was coated with 930 nm of low stress Plasma Enhanced Chemical Vapor Deposition (PECVD) Si3N4. We used DS-K101–312 for the underlayer antireflective coating (ARC) on the wafer, as this ARC is readily removed by standard 726 MIF developer. The negative-tone DUV photoresist UVN 2300–5 was spun on the ARC-coated wafer at 1300 rpm with a 70 s 110 C bake, for a final thickness of approximately 800 nm. The photoresist was then patterned in a 4x lithographic stepper (ASML 300C DUV Stepper) with 248 nm light, exposing a pattern of microylinders spaced in a hexagonal lattice pattern, with center to center distance of 1 μm between a cylinder and its 6 closest neighboring cylinders (5.1 billion cylinders total). The patterned Si3N4 was then etched by high power (3000 W) CH2F2/He plasma chemistry17, with etch rates of approximately 200 nm/min for Si3N4 and 80 nm/min for the photoresist.
After etching, we used isotropic, fluorine-based, Buffered Oxide Etch (BOE) wet etching of Si3N4, and SiO2 and HfO2 layer growth by Atomic Layer Deposition (ALD) to attain target cylinder dimensions guided by our simulations (see below). This allowed fabrication of smaller diameter cylinders (~350 nm) than possible with the DUV photolithography equipment used, but is not generally necessary if the photolithographic patterning tool achieves target dimensions at the start. Specifically, we coated a 50-nm layer of ALD HfO2 (n = 2.0) on the Si3N4 core after the BOE etch to reach our target dimensions, with core and post-ALD dimensions verified by scanning electron microscope (SEM). ALD HfO2 was chosen over ALD Si3N4 because oxide ALD processes have substantially shorter deposition times, and these two materials have nearly identical refractive indices. We then added a final 3–5 nm ALD SiO2 shell for potential chemical functionalization of the cylinders. The microcylinders were mechanically cleaved off with a razor blade and suspended in in 10 mM Tris-HCl pH 8.0 for measurements. See Figure S1 for a schematic of microparticle material layers.
Lift-off Si3N4 microcylinders were composed entirely of Si3N4, fabricated top-down, and were recovered by dissolving an Al2O3 sacrificial layer ( , Figure S1). To our knowledge, Al2O3 has not previously been used as a sacrificial layer for microparticle generation. In addition, this novel lift-off method results in a minimally toxic and biocompatible product compared to other microparticle lift-off methods used in optical trap applications which involve dissolving an entire gallium arsenide wafer to lift-off the microparticles33.
A 4-inch Si wafer was first coated with 60 nm of Al2O3 using plasma-based ALD before being coated with approximately 385 nm of low stress PECVD Si3N4. The wafer was then subjected to the same photolithography protocol for patterning and etching as detailed above for the cleaved microcylinders, except that the diameter shrinking of the pillars post-etching was not needed due to the larger ~500 nm target diameter being attainable via the DUV exposure. To perform the microcylinder lift-off, the alumina sacrificial layer was dissolved by placing the wafer in 726 MIF developer and sonicating at 60°C, for one hour and then heating for an additional 4 hours without sonication. The sample quality was then inspected by SEM and finally the microcylinder solution was placed in centrifuge tubes and centrifuged for 5 minutes at 8000 RCF and 20°C. The supernatant was decanted and the microcylinder pellet, easily visible by the naked eye, was washed, spun and resuspended twice with DI water before final suspension in 10 mM Tris-HCl pH 8.0.
Both methods produce 1 ml stock solutions of approximately 10 pM (close to 100% yield) at ~$500 (excluding labor) per 4” wafer using the Cornell NanoScale Science & Technology Facility. We anticipate significant cost reduction if these cylinders were to be mass-produced commercially. In our biophysics laboratory, multiple users can use one such stock solution for 10–12 months of experiments26.
We found the lift-off method to be especially advantageous over cleaving for fabricating shorter cylinders (<500 nm height) that would otherwise be difficult to cleave, producing cylinders with a smoother end termination on both the top and bottom surfaces, compared to the cleaved protocol where only the microcylinder top is smooth.
The carboxylated polystyrene microspheres were purchased from Polysciences Inc. (product #21753). The manufacturer measured the microspheres as 380 ± 10 nm diameter (CV 3%) and highly spherical geometry. These specifications were verified by SEM.
To quantitatively assess trap stiffness enhancement with the use of Si3N4 microcylinders, we first performed 3D full-field electromagnetic simulations of the microcylinders using the COMSOL Multiphysics finite element method software. These simulations for microcylinders used the same waveguide parameters as those previously published for polystyrene microspheres17, 25.
To experimentally determine the trap enhancement factor provided by the microcylinders, we monitored the Brownian fluctuations of a microcylinder trapped along the waveguide of an
nSWAT device and then analyzed these motion trajectories to determine trap stiffness using both power spectrum and variance analysis methods, including corrections for blurring and aliasing introduced by the camera17, 34–36.
Positions of microcylinders were tracked by a cross-correlation-based method37. The power spectrum of tracked microcylinder positions were resampled by box-car windowing in the log frequency space, and then fitted by a modified Lorentzian function with weighting factors proportional to the sample mean35–38.The modified Lorentzian function takes into account the camera blurring effect due to finite integration time, and the aliasing effect due to finite camera frame rate, and a white noise term35,36:
P(f)=∑n=−∞+∞kBT2π2βfc2+(f+nfs)2×(sin(π(f+nfs)W)π(f+nfs)W)2+ε2fs,
where kBT is the thermal energy, fc is the corner frequency, fs is the sampling frequency, W is the camera integration time, β is the drag coefficient, and ɛ2 is the white noise term. Fitting this modified Lorentzian function (truncating at n = 4) yields fc and β, and trap stiffness kpsd = 2πfcβ can be calculated. The variance analysis of the microcylinder traces was also modified for camera blurring and aliasing effects and white noise taken into account35:
Var=kBTkvar(1πWfc−12π2W2fc2(1−e−2πWfc))+ε2,
yielding the trap stiffness kvar. The above mentioned power spectrum analysis and variance analysis correspond to two methods of trap stiffness determination. When they are properly implemented and executed, kpsd and kvar are expected to show consistent values35,36.