Invited Paper

Nanoscale Fabrication of a Plasmonic Dimple Lens for Nano-focusing of Light Shantha Vedantam1, Hyojune Lee1, Japeck Tang1, Josh Conway1, Matteo Staffaroni1 Jesse Lu1, and Eli Yablonovitch1,2 1 Electrical Engineering Department, University of California, Los Angeles, California 90095 2 Electrical Engineering & Computer Science Department, University of California, Berkeley, California 94720

ABSTRACT The extent to which light can be focused with conventional optics is limited to λ/2 by the phenomenon of diffraction. Optical energy needs to be focused to less than 100nm to enable improvement and innovation in nanoscale applications. A novel lens structure to focus surface plasmons to a few tens of nanometer with high throughput is described here. This paper outlines the theoretical design and fabrication considerations of this novel plasmonic lens structure. Keywords: plasmonic, nanofabrication, nano focusing, polishing, dimple, surface plasmon, optical

1. INTRODUCTION Focusing optical energy to sub-wavelength dimensions has been an active area of research for the past two decades. Near field scanning techniques have been developed to probe the surfaces of single molecules and biological cells1. The semiconductor roadmap has been pushing the size of electronic devices to sub-100 nm dimensions, in turn pushing the limits of photolithography to the nanoscale2. The technological advances in magnetic storage media such Heat Assisted Magnetic Recording (HAMR) also require nanoscale focusing of light3,4. Sharp dielectric and metal tips with radius of curvature much smaller than the wavelength of light, have been shown to achieve sub-wavelength focusing5,6. Incident light flooding the space gets focused by this kind of tip, giving rise to field enhancement at its apex. Hence these tips are suitable for applications where absolute dark field ambience is not required. A better way to confine light in a dark field configuration is using a metalized aperture probe. Typically pulled or etched silica core fibers coated with metals like aluminum, are used as probes for near-field scanning applications. Here, the size of the optical spot is limited by the diameter of the silica core which is on the order of 50-100nm. Also the propagating waveguide mode in the silica core is evanescent beyond the cut-off diameter of the core and hence the throughput of this probe falls off the fourth power of the aperture diameter7,8. Several other structures like the C-shaped apertures, bow-tie and its complementary structure have been shown to confine light to a few tens of nanometers9,10. However the commonly employed focused ion beam (FIB) milling limits their critical dimension to ~30 nm. Optical free space photons can be efficiently coupled to the nanoscale via surface plasmons. Surface plasmons are electronic excitations of free electron gas of a metal at a metal-dielectric interface, oscillating at optical frequencies. Photons coupled via plasmons can be focused to a smaller spot size, thus beating the diffraction limit on optical focusing11. An optimal design and fabrication strategy of a novel plasmonic lens to focus surface plasmons is described in this paper.

2. THEORY AND DESIGN Figure 1 shows the dispersion relation12,13 of plasmonic modes in a metal-dielectric-metal geometry. Plasmonics: Metallic Nanostructures and Their Optical Properties V, edited by Mark I. Stockman, Proc. of SPIE Vol. 6641, 66411J, (2007) · 0277-786X/07/$18 · doi: 10.1117/12.735462

Proc. of SPIE Vol. 6641 66411J-1

hν

Plasmon Wavelength in nm 200

50

0.03141590.0628318

3.0

3.0

Plasmon Energy in eV

3.5

3.5

100

2.5

2.0

2.0

1.5

1.5

30

0.2094393

20

15

0.3141590

10

0.4188787

0.6283180

40

d=20nm

2.5

1.0

0.12566360.1570795

d=5nm

d=2nm d=1nm

Ag 1.0

d

d= thickness of SiO2

SiO2 Ag

0.5

0.5

0.0 0.0

0.1 0.1

0.2 0.2

0.3 0.3

0.4 0.4

0.5

0.6

0.5

0.6

k

Plasmon Wave-Vector (2π/wavelength in nm) Figure 1 Calculated dispersion relations for the Ag-SiO 2 -Ag stack for various thicknesses of the SiO 2 layer.

As the thickness of the dielectric decreases, double-sided plasmons in a metal-dielectric-metal geometry deviate from the light line at lower optical frequencies. In other words, a photon of a given energy assumes shorter plasmonic wavelengths as the thickness of the dielectric layer decreases. This observation forms the basis of our novel plasmonic lens design. Tapering down the dielectric layer sandwiched between two metals from 100 nm down to a few nanometers, couples photons into double-sided plasmons with wavelengths on the order of a few tens of nanometers (Figure 2a). This focuses the optical energy in the direction normal to the layers of the stack in the metal-dielectric-metal geometry. In order to focus in the direction parallel to the layers, the stack is rotated about the focal point of Figure 2a to form the structure shown in Figure 2b. Hereafter, this semicircular tapered lens is referred to as the plasmonic dimple lens. The spot size of the focused optical energy should be on the order of surface plasmon wavelength λsp, which is very small at the focal point of the structure.

fliLedi Lapei

far-field from lens dimple lens

LLL th-couplirig grating

(a)

point

(b)

Figure 2 (a) Cross-sectional view schematic of the plasmonic dimple lens. Thickness of SiO2 layer tapers in vertical direction (b) Perspective view of the plasmonic dimple lens

The optimal dimensions of the taper are determined primarily by the absorptive and reflective losses. Surface plasmons propagating on the metal dielectric interface decay in energy because metals have finite absorption in the optical regime.

Proc. of SPIE Vol. 6641 66411J-2

Double-sided plasmonic modes with larger wave-vectors suffer greater absorption than ones with smaller wave-vectors. In order to keep this loss low, the taper should be as short as possible. Another source of energy loss is plasmon reflection arising from tapering the dielectric layer too steeply. To keep the reflection losses low, the taper angle should be very small. To minimize the energy loss, an optimal angle should be used for tapering the dielectric. From Figure 3, it is found that 25°-35° is an optimal angle for the taper for an Ag-SiO2-Ag stack operating at a visible wavelength of 476 nm12. 10 10 9 9

Ag

8 8

50nm SiO2

7 7

Loss (dB)

6 6

θ°

1nm

Ag

5 5 4 4 3 3

2 2 1 1 0

00

5 5

10 10

15 15

20 20

25 25

30 30

35 35

Taper Angle θ (Degrees) A l (D )

40 40

45 45

50 50

Figure 3 A plot of energy loss of a double-sided plasmonic mode as it traverses across a SiO2 taper from 50 nm oxide down to 1 nm oxide thickness, with varying taper angles at λ = 476 nm.

Another aspect in the design of the plasmonic dimple lens is that it uses a circular grating to efficiently couple the light into single-sided surface plasmons. Since surface plasmons have a wave-vector larger than the photons of the dielectric, a grating coupler provides the essential momentum matching for efficient coupling. Apart from momentum matching, to improve the coupling from the incident gaussian laser beam to the evanescent field of the plasmon, the widths and the positions of each of the grating grooves is optimized to maximize the coupling efficiency14.

3. FABRICATION AND CHARACTERIZATION The established microfabrication techniques used in the semiconductor industry employ planar processes. A circularly symmetric nanoscopic feature poses a significant fabrication challenge. The focusing capability of the lens is determined by the thickness of the dielectric layer at the out-coupling facet as described in the previous section. Hence the foremost specification is that this thickness should be less than 10 nm for nanoscale focusing. Secondly, the dielectric layer needs to have a semi-circular dimple profile. The third requirement is, to minimize the scattering energy losses from arising from surface roughness, the interfaces between the metal and dielectric layers need to be extremely smooth. Figure 4 presents the fabrication process flow for the building a plasmonic dimple lens that meets all the above requirements. This section delves into the issues and considerations at each step that evolved this fabrication process flow. It is well known that the semiconductor industry developed processes to grow high quality amorphous thin films (<5 nm) of SiO2 and Si3N4 on single crystal Si-substrates15,16. These thin films serve well as the critical dimension of the plasmonic dimple lens. These amorphous films need to become the base of the tapered dielectric. A fortuitous process enables us to create the circular nanoscopic dimple shape in the polymeric resist. Polymers have a refractive index of about 1.46-1.5 that is close to that of SiO2. Thus the dielectric layer of the plasmonic dimple lens comprises of the dimple profile in an organic polymeric compound and a thin oxide/nitride layer. The dielectric layer needs to be sandwiched between two metal layers. This mandates the removal of the Si-substrate, on which the thin amorphous layer is grown.

Proc. of SPIE Vol. 6641 66411J-3

(a)

(b) Si N4

Ag

(<19nm)

Si3N4

epoxy

epoxy

glass

glass

(c)

(d)

Si N4—

Ag

(<1(9 nm)

I

Si N4—

(<19 nm)

Si3N4

I

PMMA

Ag Si3N4

_Ioxy glass

glass

(e)

(f)

glass

optical epoxy Si N4—

PMMA

(<1(9 nm)

SU-8 Ag Ag

I

< 10 nm

Si3N4

epoxy

glass

glass

Figure 4 (a) Ag islands patterned on 10 nm LPCVD Si3N4 on a Si-substrate. These islands are covered with ~ 300 nm of PECVD Si3N4 and the sample is bonded to a glass using silica-filled epoxy. (b) Si-substrate is completely etched away with HNO3+HF after thinning it down with DRIE process. (c) Semi-circular grating coupler and (d) dimple profile in PMMA are patterned with e-beam lithography (e) second layer of Ag is deposited and etched with SU-8 as a mask (f) The sample is bonded to glass using optically transparent epoxy and this stack is mechanically polished from one edge until the point when the circular dimple profile in PMMA is polished halfway through.

The removal of the silicon substrate required the testing of various methods of etching silicon with high etch rates. Among the explored methods were dry etch of silicon by XeF2, wet etch in KOH solution, wet etch is a mixture of HNA (HF+HNO3+CH3COOH). XeF2 gas is known to be highly selective17 in etching silicon versus SiO2. However presence of moisture in the etch chamber caused a 6 nm of thermal SiO2 to be etched away easily. On the other hand a 10 nm silicon nitride layer serves as a good etch mask while etching silicon away in both KOH and HNA mixture. KOH solution etches silicon anisotropically along selected etch planes of the silicon crystal18. In an attempt to remove the complete silicon substrate, it is found that KOH solution leaves behind mounds of silicon bounded by the extremely slow etching <111> planes. On the other hand, the HNA mixture etches silicon isotropically with the etch rate depending on the ratio of the three acids in the mixture18. A mixture of HF and HNO3 in equal proportions etches the entire silicon substrate of ~ 300 µm in less than 45 sec without damaging the Si3N4 layer (Figure 3(b)). Silicon nitride deposited by the low pressure chemical vapor deposition (LPCVD) process survives the HNA mixture better than the nitride formed by the plasma enhanced chemical vapor deposition (PECVD) process. Concentrated HF is used in the HNA mixture etches

Proc. of SPIE Vol. 6641 66411J-4

Si3N4 as well, hence it is critical to remove the sample from the etchant solution as soon as the entire silicon is etched away to minimize the etch of the nitride layer. Thinning the 300-500 µm thick silicon substrate to a thickness of about 50 µm requires a quick dip of ~ 10 seconds in a mixture of HF and HNO3. Silicon substrate is thinned down by a deep reactive ion etching (DRIE) process. In order to support the thin nitride layer before the removal of the silicon substrate, a piece of glass is bonded to this with some epoxy. Epoxy shrinks on curing and causes stress-induced rupture of the thin nitride layer when the entire silicon substrate is removed (Figure 5). In order to overcome this problem, about 300 nm thick layer of Si3N4 is deposited by PECVD process on the thin LPCVD nitride layer. Using a silica-filled epoxy for bonding glass to the nitride layer on silicon also helps to mitigate the abovementioned problem (Figure 4(a)). The fraction of epoxy in a silica-filled epoxy formulation is small enough to reduce the shrinkage stress of epoxy on the nitride layer.

'I (a)

(b)

Figure 5 (a) A large scale rupture of the thin nitride film after Si-substrate removal (b) Local stress in the LPCVD nitride causes the HNA mixture to penetrate into the PECVD nitride layer underneath.

After the removal of the silicon substrate, a few steps of electron-beam lithography are used to pattern a grating coupler and a dimple structure in PMMA on the thin nitride layer as shown in Figure 4(c) & 4(d). These e-beam lithography steps produce good alignment between the grating coupler and the dimple structure with a set of pre-patterned alignment marks. A three dimensional dimple shape in the e-beam resist on the nanometric scale is accomplished by a nonconventional single spot exposure. The primary electron beam focused at a point gives rise secondary electrons that are backscattered from the surface. These secondary electrons play a significant role in the scission of the PMMA polymeric chains into segments of varying lengths. The difference in the dissolution rates of these segments of varying lengths in the developer solutions gives rise to the 3D profile in the resist. Figure 6 depicts the AFM scan of this feature in PMMA.

50

50 1 150 20 250

0

Figure 6 Topographic AFM scan of the dimple in formed in PMMA resist and its cross-section

It is observed that thin films of metals deposited by e-beam evaporation on dielectric substrates have considerable surface roughness due to the formation of grains, with rms values on the order of at least 2-3 nm19,20. In our process

Proc. of SPIE Vol. 6641 66411J-5

flow, the silver is evaporated against a smooth nitride or PMMA surface as shown in Figure 4(a) & 4(e). The thin LPCVD nitride in Figure 4(a) inherits the roughness of the single crystal Si-substrate which is almost atomically flat. The rms surface roughness of PMMA after development is on the order of 1nm. This salient feature of our process flow meets the requirement of smooth metal-dielectric interfaces as described in the first paragraph of this section, thus making this approach superior to the one where the top surface of metal deposited on dielectric is used in plasmonics13. Finally, the sample is bonded with another piece of glass with an optically transparent epoxy. The two pieces of glass sandwiching the dimple lens and the grating coupler form a mechanically robust structure which can be clamped into a fixture that is to be mounted on an edge polishing machine. This sample is polished from the edge such that the circular dimple as shown in Figure 6 is cut into half to expose the focal point of the plasmonic dimple lens. This step is as depicted in Figure 4(f). The mechanical polishing process is monitored with the help of pre-patterned fiduciary patterns of metal that are viewed with an optical microscope. During the polishing, the material removal rate can be brought down such that the polishing can be stopped within a range of a few nanometers from the focal point of the dimple lens. Mechanical polishing of a stack of dissimilar materials with varying hardness proved to be a challenge. To overcome the problems of poor adhesion of metal to dielectric and the delamination of the softer metal and epoxy layers from the harder nitride and glass layers, careful tuning of the thickness of the layers of stack along with optimum selection of a few polishing process parameters like the polishing grit size, was required. A low material removal rate coupled with the continuous flow of water to wash away the polished debris from the pad, helped to improve the polishing results. Due to its high conductivity and low optical loss among metals, silver was the first choice for the plasmonic dimple lens. However gold seems to have better adhesion and post-polishing surface texture than silver as shown in Figure 7. Hence gold was chosen in the final plasmonic lens structure.

-'I

(a)

(b)

Figure 7 A perspective topographic image of the polished facet after mechanical polishing. The layers of the stack are outlined and labeled for (a) Ag-dielectric-Ag and (b) Au-dielectric-Au dimple lens. The polishing results for Au in (b) are better than for Ag in (a).

4. CONCLUSIONS In conclusion, we have designed and fabricated a novel plasmonic dimple lens structure in an Au-dielectric-Au geometry for focusing visible light to the nanoscale. The fabrication strategy is such that critical dimension of this plasmonic lens does not depend upon the lithographic limitations of the electron beam or ion beam processes/equipments. The experimental characterization of the sub-100nm focusing capability of the plasmonic dimple lens requires near-field measurement. Commercial NSOM equipment using metalized aperture fibers is limited to 100-150nm in resolution. Several other higher-resolution near-field techniques are being explored for the experimental characterization including scattering from a sharp AFM tip21. The mechanical edge polishing step in our fabrication sequence lends itself to be easily adopted by the process flow of the magnetic read-write head of a commercial hard drive. Thus, this structure could potentially be an excellent candidate

Proc. of SPIE Vol. 6641 66411J-6

for the Heat Assisted Magnetic Recording (HAMR) technology, where the plasmonic dimple lens would be used to focus light to provide the strong local heating required for the next generation higher density hard drives3,4.

ACKNOWLEDGEMENTS We gratefully acknowledge the support given by the NSF Nanoscale Science and Engineering Center for Scalable and Integrated Nanomanufacturing (SINAM) under the award number DMI-0327077 and the Defense Microelectronics Activity (DMEA) under the agreement number H94003-06-2-0607.

REFERENCES 1. N.F. van Hulst, J-A. Veerman, M. F. Garcia-Parajo, and L.K. Kuipers “Analysis of individual (macro)molecules and proteins using near-field optics,” J. Chem. Phys. 112, p7799 (2000) 2. ITRS roadmap report “ International Technology Roadmap for Semiconductors, 2006, Executive Summary,” http://www.itrs.net/Links/2006Update/FinalToPost/00_ExecSum2006Update.pdf 3. William A. Challener, Terry W. Mcdaniel, Christophe D. Mihalcea, Keith R. Mountfield, Kalman Pelhos and Ibrahim K. Sendur, “Light Delivery Techniques for Heat-Assisted Magnetic Recording,” Jpn. J. Appl. Phys. 42, p981988 (2003) 4. Terry W. Mcdaniel, William A. Challener and Kursat Sendur, “Issues in Heat-Assisted Perpendicular Recording,” IEEE Trans. on Magnetics 39, 1972-1979 (2003) 5. L. Novotny, D.W. Pohl, B. Hecht, “Scanning near-field optical probe with ultrasmall spot size,” Opt. Lett., 20, 970972 (1995) 6. A.Chimmalgi, T.Y.Choi, C.P. Grigorapoulos, K. Komvopoulos, “Femtosecond laser aperturless near-field nanomachining of metals assisted by scanning probe microscopy,” App. Phys. Lett. 82, 1146-1148 (2003) 7. H. Bethe, “Theory of diffraction by small holes,” Phys.Rev. 66, p163 (1944) 8. C.J. Bouwkamp, Philips Res. Rep. 5, p321 (1950) 9. X. Shi and L. Hesselink, “Design of a C aperture to achieve λ/10 resolution and resonant transmission,” J. Opt. Soc. Am. B 21, p1305 (2004) 10. E. Cubukcu, E. A. Kort, K. B. Crozier, and F. Capasso, “ Plasmonic laser antenna,” App. Phys. Lett. 89, p093120-1 (2006) 11. J.M. Steele, Z.W. Liu, Y. Wang, and X. Zhang, “Resonant and non-resonant generation and focusing of surface plasmons with circular gratings,” Opt. Exp. 14, 5664-5670 (2006) 12. J.A. Conway, “Efficient Optical Coupling to the Nanoscale,” Ph.D. thesis, University of California, Los Angeles (2006) 13. H.T. Miyazaki and Y. Kurokawa, “Squeezing visible light waves into a 3-nm-thick and 55-nm-long plasmon cavity,” Phys. Rev. Lett. 96, 097401 (2006) 14. J. Lu, C. Petre, J. Conway, and E. Yablonovitch, “Numerical Optimization of a Grating Coupler for the Efficient Excitation of Surface Plasmons at an Ag-SiO2 Interface,” J. Opt. Soc. Am. B 24, p 2268 (2007) 15. D.A. Muller, T. Sorsch, S. Moccio, F.H. Baumann, K. Evans-Lutterodt, and G. Timp, “The electronic structure at the atomic scale of ultrathin gate oxides,” Nature 399, 758 (1999) 16. X. Shi, M. Shriver, Z. Zhang, T. Higman, and S.A. Campbell, “Properties of high-k/ultrahigh purity silicon nitride stacks,” J. Vac. Sci. Technol. A 22, 1146 (2004) 17. K.R. Williams, K. Gupta, and M. Wasilik, “Etch Rates for Micromachining Processing –Part II,” J. Microelectromech. Syst. 12, p761 (2003) 18. M. Madou, Fundamentals of Microfabrication, Chapter 4, CRC Press, New York, 1997 19. R. Petrovic, S. Strbac, N. Bundaleski, and Z. Rakocevic, “Surface roughness minimum: Ag thin layer deposited on glass,” J. Serb. Chem. Soc. 66, 483 (2001) 20. A. Bendavid, P. J. Martin, L. Wieczorek, “Morphology and optical properties of gold thin films prepared by filtered arc deposition,” Thin Solid Films 354, 169 (1999) 21. T.J. Yang, G.A. Lessard, and S.R. Quake, “An apertureless near-field microscope for fluorescence imaging,” App. Phys. Lett. 76, 378 (2000)

Proc. of SPIE Vol. 6641 66411J-7