challenges in the fabrication of an optical frequency ...ecee.colorado.edu/~wpark/papers/j. vac....

Challenges in the fabrication of an optical frequency ground plane cloakconsisting of silicon nanorod arrays

J. Blaira�

School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332

D. BrownMicroelectronics Research Center, Georgia Institute of Technology, Atlanta, Georgia 30332

V. A. Tamma and W. ParkDepartment of Electrical, Computer and Energy Engineering, University of Colorado, Boulder, Colorado80309

C. SummersSchool of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332

�Received 27 April 2010; accepted 27 September 2010; published 28 October 2010�

The application of transformation optical techniques to photonic crystal-like dielectric structures hasfacilitated the creation of invisibility cloaks that operate at optical wavelengths. In this article, theauthors present the fabrication processes for an all-dielectric ground plane cloak structure thatconsists of multiple silicon nanorod arrays connected by input and output waveguides. An advantageof this particular design is that it does not use metals to obtain metamaterial-like device behavior.The structure consists only of dielectrics that can be processed by the use of electron beamnanofabrication technologies and is designed to operate in the 1400–1600 nm wavelength range.
© 2010 American Vacuum Society. �DOI: 10.1116/1.3504595�
I. INTRODUCTION

The idea behind invisibility cloaking was originally pub-lished by Pendry et al. in 2006.1 The concept involves cre-ating a closed space from which electromagnetic radiation isexcluded. In order for the cloak to perform correctly, theradiation must be steered around it. This steering can beaccomplished by the use of mathematical transformation onthe original space that creates a warped space, which in turnis implemented by spatially varying permittivity andpermeability.1,2 The creation of these cloaking structures re-quires the use of metamaterials to obtain the prescribed val-ues of permittivity and permeability necessary for correctdevice operation.3,4 The cloak designed for operation in themicrowave frequency range has been demonstrated by usingresonating metal elements.5 However, the ideal cloak re-quires extreme values of permittivity and permeability withstrong anisotropy, making the fabrication highly challengingeven in the microwave region. It is even more challenging tofabricate cloaks for optical frequencies because the metalstypically used in metamaterial structures are unable to main-tain low-loss performance characteristics at higher frequen-cies of operation.

Recently, a new cloaking design has been proposed thatsignificantly reduces the required range of material propertyvalues.6 This design, often dubbed as the “carpet” cloak,compresses a curved reflective surface �the cloak interfacearea� into a flat reflective surface, effectively shielding ob-jects behind the curved surface from outside observers. Fig-ure 1 shows the reflection of incoming radiation of a curvedreflecting surface with and without the carpet cloak. In the

a�
Electronic mail: [email protected]
1222 J. Vac. Sci. Technol. B 28„6…, Nov/Dec 2010 1071-1023/2010

Author complimentary copy. Redistribution subject to AIP lice

case of the noncloaked device shown in Fig. 1�b�, the incom-ing beam reflects off into two divergent beams. Compare thiswith the carpet cloak interface shown in Fig. 1�a�, where awell-defined single reflected beam of a flat mirror surface isrestored.

Figure 2 shows the required refractive index profile forthe carpet cloak. The quasiconformal mapping is performedsuch that the transformed cells of the structure reduce theoverall anisotropy of the device to a smaller range of isotro-pic medium indices. This approach eliminates the need forabsorptive metallic resonant elements, and thus, can be fab-ricated using only dielectric materials.

This cloak design was recently demonstrated to work inthe microwave7 and optical regimes.8,9 Our goal was to adoptthis cloaking design for operation at optical frequencies byscaling the device dimensions down to nanoscale dimen-sions. For device operation in the near infrared wavelengthspectrum, features in the range of 5–50 nm need to be cre-ated within high precision tolerances to prevent undesirableshifts in the frequency of operation of the cloak. To accom-modate the small feature size and range in the nanoscaleregime, specific design choices were made in the creation ofthis cloaking device to maximize performance and minimizethe complexity of fabrication. The basic design strategy is touse deep subwavelength scale silicon nanostructures to pro-duce a metamaterial structure with the desired effective re-fractive index values at different points within the device.When the features are much smaller than the wavelength oflight, the effective index of a composite structure is deter-mined by properly averaging the indices of the constituentmaterials. In our design, we chose our unit cell to be 150 nm,
which corresponds to � /10 for operation at �=1500 nm.
1222/28„6…/1222/9/$30.00 ©2010 American Vacuum Society

nse or copyright, see http://jvb.aip.org/jvb/copyright.jsp

1223 Blair et al.: Challenges in the fabrication of an optical frequency 1223

Each unit cell may contain an air hole or a silicon nanorod.Since the effective index of a given unit cell is given by thevolume average of the indices of silicon ��=12.0� and air��=1.0� within the unit cell, we can simply change the sizeof the air hole or silicon nanorod to realize a range of effec-tive index values required to implement the cloak structure.For a given unit cell size, air holes with varying diameterswere found to provide a smaller range of effective indicescompared with nanorods. Thus, we chose to use silicon na-norods of various diameters to implement the all-dielectriccloak structure. The sizes of the nanorods were then deter-mined using the effective medium theory. We first calculatedthe effective permittivity for the fundamental transverse-magnetic �TM, electric field perpendicular to the devicelayer� mode for the air-silicon-oxide slab waveguide, whichwas found to be �SOI-TM=7.55 at ��=1500 nm. If the nano-rods are small, the effective permittivity of the nanorod arrayis given by the simple volume average of the silicon slab andair, �eff=A�SOI-TM+ �1−A��air, where A is the total cross-sectional area encompassed by nanorods and �air=1. Therange of permittivity values required for the cloak can now

FIG. 1. �Color online� Radiation reflecting off a curved reflecting surface �a�with a carpet cloak and �b� without any cloak. In both cases, the incomingbeam is incident from the upper-right corner. Note the divergent reflectedbeams in case �b� compared to the single reflected beam in case �a� �Ref. 6�.

FIG. 2. �Color online� Quasiconformal dielectric map of the ground plane
cloak structure �Ref. 6�.
JVST B - Microelectronics and Nanometer Structures


be realized by progressively varying the nanorod diameterthroughout the structure. The validity of this simple averag-ing rule was confirmed by performing rigorous three-dimensional photonic band structure calculations as verifica-tion of the method.

Three different cloak designs were created, the first twohaving less challenging electron beam lithography require-ments to provide a design baseline for the third structure,which is more difficult to fabricate. All three designs re-quired permittivity values inside the cloak ranging from 1.5to 4.4, created using the volume averaging index method asdescribed above. In the base design for all the cloaks, thesmallest nanorods are on the order of 50 nm in diameter,while the largest nanorods required very small gaps betweenthem of approximately 20 nm. As a comparison, the Si de-vice layer is approximately 17 times the thickness of thesmallest gap size. Producing these small size features in asmall device area significantly increases the difficulty in fab-ricating these devices using current electron beam technolo-gies.

II. FABRICATION METHODS

The cloak structures under investigation were designedspecifically to be fabricated on a silicon-on-insulator �SOI�substrate in a microelectronics facility using readily availablecleanroom technologies. This strategy was used because theexisting knowledge and experience available in microelec-tronics fabrication is extensive, so the risks involved in mak-ing these new devices to correct specifications with minimaliterations was reduced. Reducing the number of iterationsrequired to successfully fabricate the devices was crucial inobtaining a successful outcome as well as in minimizing thedevice fabrication costs. The challenge of producing nanorodarrays that exhibit cloaking behavior at optical frequencies isthat the feature size begins to approach the limits of moderncleanroom processing and patterning techniques for high as-pect ratio structures. For example, this particular design re-quired narrow gaps between nanorod structures that are dif-ficult to keep from bridging in the fabrication process. Inorder to pattern the nanoscale devices successfully, state-of-the-art electron beam lithography �EBL� techniques havebeen used. The choice of resist used to pattern the device wasanother consideration, as the electron beam dose and otherprocessing parameters are critically dependent on obtaining ahigh-contrast pattern. An additional challenge to successfuldevice fabrication involves obtaining vertical sidewalls onnanoscale device features to reduce scattering. Meeting thischallenge involved the optimization of plasma-etching tech-niques and etch recipes, working within the limits of theprocessing capabilities of the equipment.

Several approaches to the design were investigated withthe goal of obtaining a reasonable trade-off between ease ofdevice fabrication and final performance. The choice of us-ing nanorods over holes allows a more advantageous choiceof electron beam resist, specifically the use of a negativeresist with a higher selectivity ratio. The higher selectivity
ratio enables deeper etching while keeping the mask intact,


thereby allowing more vertical sidewalls to be fabricatedduring the plasma-etching process. Also, the use of a nega-tive resist allows for a faster write time as the size of theresist mask areas is significantly reduced. Another advantageof the design is that the entire structure can be patternedusing the EBL system without the need for additional steps,such as focused ion beam milling and metallization.9 In ad-dition, the spatial variation within this device is accuratelyadjusted by exerting direct control over the nanorod pillarsize and spacing, as opposed to random placement of nano-rods in other designs that makes index control moredifficult.10 This feature size control advantage allowschanges to be made easily to reduce the minimum size writerequirements of the EBL patterning process, allowing formore reliable and consistent fabrication results.

The process flow for fabrication starts with the entry ofthe design structure into a computer-aided design �CAD� fileformat, followed by a conversion of the file into a.v30 formatreadable by the EBL system. Due to the complexity of draw-ing approximately 13 200 different sized circles using amanual command-line interface, the process was automatedthrough the use of Excel spreadsheets and AUTOCAD script-ing functions. The surrounding background photonic crystaldesign was then added through the use of array functions,and the entire structure was rotated by 45° to obtain thecorrect orientation for adding the waveguides.

Once the structure design was entered into the CAD sys-tem and converted to a.dxf format, a second conversion stepwas performed to convert the design to a.v30 file formatthrough the use of a proximity correction software. Proxim-ity correction was necessary during the file conversion pro-cess to allow for dose adjustments due to the Gaussian natureof the electron beam during the exposure of the structure,most importantly at the device edges and near the largewaveguide areas. The proximity correction software dividesthe structure into 64 different dose adjustment areas, eacharea being a varying percentage of the base dose to compen-sate for the Gaussian nature of the electron beam exposureprocess. The dose structure output data were then placed intothe EBL system control file to enable proximity correcteddose adjustments during the device patterning. However, thecurrent version of proximity correction software does nottake into account the structure size and shape when calculat-ing dose adjustments, and even an updated version is onlyable compensate for some adjustments to rectangular fea-tures. Since this design utilizes circular structures, manualadjustments were made to the proximity correction param-eters during file conversion. The exact proximity correctionparameters for the different designs are discussed further inSec. III.

The SOI stack used was a 340 nm thick Si layer over a1 �m SiO2 layer, which provides an adequate guiding thick-ness for light propagation in the 1400–1600 nm range. A 6%solution of hydrogen silsesquioxane �HSQ� was then appliedto the surface through a spin-coating process. HSQ is a poly-merlike negative-tone resist material that undergoes a transi-
tion to a more SiO2-like material upon e-beam exposure and
J. Vac. Sci. Technol. B, Vol. 28, No. 6, Nov/Dec 2010


has a demonstrated high resolution masking capability. Forthe high aspect ratio process, the spin coater was set to 5000and 2500 rpm/s spin up, followed by a wafer bake at 80 °Cfor 4 min to cure the resist and obtain a 100–110 nm layerthickness. Significant care was taken to keep the wafer cleanand free from contaminants prior to the application of theresist, as these could lead to defects in the final devices. Thewafer was then manually cleaved into smaller pieces for pro-cessing.

Electron beam resist patterning was performed using aJEOL JBX-9300FA EBL system running at an accelerationvoltage of 100 kV with a 5 nm spot pitch size. A piececassette was selected to allow for multiple device runs usinga single SOI wafer. In the design, the input and output wave-guide line lengths were set to 3 mm to allow enough roomfor manual line cleaving after the devices had been fabri-cated. Extensive experimentation was required to determinethe correct electron beam base exposure dose that would givethe best feature size match without either overexposing orunderexposing parts of the device pattern.

The best electron beam dose for all three device designswas found to be in the range of 2400–3000 �C /cm2 afterthe proximity correction adjustments had been made. Afterthe exposure step, the device pattern was developed using a25% solution of N-tetramethyl ammonium hydroxide�TMAH� heated at 80 °C. The wafer piece was developedfor 30 s, followed by a de-ionized water rinse for approxi-mately 2 min and sample drying using N2 gas. A visual in-spection was then performed to make sure the developmenthad been completed and that there were no residues or con-taminants left on the wafer. Before etching the device, mea-surements of the height of resist patterns were made using aTencor alpha-step profilometer to provide a reference pointfor calculating resist etch rates. The resist height patternswere identical to the original thickness of the resist layer�100–110 nm� after development, indicating that no resistmaterial had been removed in the process.

The etching of the devices was performed using an STSstandard oxide etcher using Cl2 as the enchant gas. The chlo-rine gas flow was set to 20 SCCM �SCCM denotes cubiccentimeter per minute at STP� and the plasma bias voltagewas set to 700 V. The objective was to etch completelythrough the 340 nm layer of Si without undercutting into theSiO2 layer, which would destabilize the pillar structures. Theetch rate of Si using this recipe was determined to be 2.6nm/s, while that of the HSQ was 0.55 nm/s, resulting in aselectivity of approximately 4.75: 1. The 30 nm of HSQremaining on the surface of the device was not removed dueto the negligible effect it would have on the device perfor-mance due to its low index over the frequencies of operation.After etching, the ends of both the input and outputwaveguides were cleaved to provide a clean edge for butt-coupling the laser light source to the wafer for device testing.The butt-couple allows for introduction of laser light into thedevice, allowing the near-field pattern of the cloaking struc-
ture to be examined in detail through measurements.


III. FABRICATION RESULTS AND DISCUSSION

The first cloak structure, which will be referred to as de-sign A, was designed for a background index of 1.5. Thisindex was chosen to test how a matched index between theunderlying silicon dioxide substrate and overlying cloakingstructure would affect light scattering in the device. For thisdesign, the cloak requires the nanorod diameters to varyfrom 0.35a to 0.87a or from 52 to 130 nm for a unit cell sizea=150 nm. This section of the cloak will be labeled as thephotonic crystal 1 �PC1� area. In design A, a small portion ofthe cloak structure nanorod array was replaced with largerpillar structures that have an increased lattice constant of a=300 nm. In this subarray, which will be notated as photoniccrystal 2 �PC2�, all feature sizes were doubled, and therefore,the smallest gap was increased to �40 nm, which is wellwithin the range of e-beam fabrication capabilities. Figure 3details the cloaking area for this design, showing both thePC1 and PC2 arrays.

A schematic for design A optical cloaking structure isshown in Fig. 4. The device region consists of two combined

FIG. 3. �Color online� Diagram of the nanorod structure in the cloak, show-ing the areas of smaller lattice spacing �PC1� and larger lattice spacing�PC2�.

FIG. 4. Block diagram of the design A nanorod optical cloaking structure.



areas of silicon rods in a 39.6�39.6 �m2 area, connectedby a 10 �m wide input and a 36.9 �m wide output Siwaveguide. In the optical experiments, light is fed into thesilicon input waveguide so that it would propagate throughthe nanorod array, reflect off the curved reflecting surface,and propagate on toward the output region. The size of theoutput Si waveguide was reduced in follow-on designs toreduce the fabrication expense of the device. The 32�12 �m2 cloaking region contains the main array of 150nm spaced pillars with diameters ranging from 90.75 to52.18 nm, and the secondary array of 300 nm spaced pillarswith larger diameters in the range from 184.05 to 256.18 nm.The modified spacing of the center pillars in the secondaryregion increased the minimum gap between adjacent struc-tures to around 40 nm, which reduced the spacing require-ments and, therefore, the difficulties in producing the rodsaccurately during fabrication.

The pillar diameters were arranged such that the averagerefractive index creates the quasiconformal dielectric indexmapping required for the device to exhibit cloaking behavior.The uniform background nanorod array was a square latticeturned 45° to the �−M direction to provide a well-matchedinterface to the cloak structure, so as to minimize the scat-tering at the interface. The input waveguide width is standardfor the wavelength of operation, while the output waveguidewas widened to allow some room to observe the direction ofthe output light, in case the reflected beam was not at thepredicted angular deviation.

For design A, the standard default forward scattering andback scattering correction for Si devices was used as a start-ing point. The � parameter describes the forward �shortrange� scattering range of the electrons and is importantwhen the aspect ratio of the device features is small. The �parameter describes the backward �long range� scattering ofthe electrons, which describes the distance the electronstravel into the substrate material. Both the � and � param-eters are set to values that are dependent upon the substrateand photoresist material used. The parameter describes theratio of the � Gaussian against the � Gaussian based on theirstored energies and determines the ratio of forward to back-ward scattering. The reduced spacing requirements in thelarge pillar subarray was expected to allow the smaller sidenanorods to be reproduced with the correct dimensions usingthe default proximity correction parameters for a Si sub-strate. These parameters were originally determined fromMonte Carlo simulations and extensive experimental runs ofSi devices through characterization work conducted overmany batches of different structures. The values that wereused were �=0.05, �=33.3, and =0.61.

Scanning electron microscope �SEM� images were takenof the fabricated device A structure to check their quality andfinal dimensions. For design A, Fig. 5 shows the overall fab-ricated device, Fig. 6 shows a close up of the cloaking area,and Fig. 7 is a high magnification image of the largest nano-rods in the cloaking area. The nanorod diameter matched theoriginal design within 5% experimental tolerances once the
best dose value was obtained. While the pillars were nearly


perfectly vertical, some nonuniformity in top and bottom di-ameters was observed in the sidewalls of the nanorods, espe-cially those near the device edges, where the bottom diam-eter was somewhat larger than the top. Close-up images ofthese nonuniformities are shown in Fig. 8 for a larger nano-rod �about an 8% difference� and Fig. 9 for a smaller nano-rod �a larger 24% difference�. This effect could be attributedto nonuniformities in the photoresist mask covering the na-norods, which caused sloping sidewalls due to the edges ofthe mask being thinner, and therefore, more susceptible toremoval during the etch process. Overall, the standard pa-rameters for the proximity correction of the device duringEBL exposure worked well for this structure fabrication.

The second cloak structure, named design B, is similar todesign A but reduced in size. In an effort to reduce the scat-tering losses in the device, design B cloak reduced the num-ber of pillars, and therefore, the number of scattering sourcesfor the incoming light beam. The smaller cloak consists of a

FIG. 5. SEM of the design A fabricated cloaking nanorod device.

FIG. 6. Higher magnification image of the Si pillars around the cloaking area
in design A.


143�53 nanorod array that reduces the device size to a26.4�26.4 �m2 area. The cloaking area size was also re-duced to 21.45�7.95 �m2, with an identical design A pillarspacing of a=300 nm. The layout of design B is identical todesign A except for the reduction in cloak and backgroundnanorod array areas. A block diagram for design B is shownin Fig. 10. Using the same proximity correction and fabrica-tion parameters as design A, these cloaks had similar non-uniformities to the design A devices due to the identical etchparameters being used. Figure 11 shows the smaller cloakingarea in device B postmeasurement, with the damage in theouter pillars having been caused during the measurementcharacterization work.

For the third design, called design C, the background in-dex was changed to implement 1.55 and 1.6 index structuresto see if changing the effective index would also reduce scat-tering, thereby improving the device performance. This in-

FIG. 7. High magnification SEM image of the largest nanorods in the cloak-ing area of the device in design A. Note the clear separation between nano-rods at the smallest gap point of �40 nm at the center of the structure.

FIG. 8. Side view of one of the larger nanorods in device A, showing thenonuniformity obtained in the sidewall thickness as an effect of the etch
process.


dex match change required slight adjustments in the pillarsizes throughout the device. In this design, the lattice con-stant was kept at a=150 nm for the cloaking section, recre-ating the original design for a smaller 20 nm pillar gap. Thechange in background index also resulted in slightly smallernanorod diameters ranging from 0.35a to 0.84a or from 52 to125 nm. In addition, a simple square lattice photonic band-gap �PBG� structure was added in design C to help minimizelight leakage at the back of the cloaking area. The cloak isdesigned to mimic a flat reflecting surface, and therefore, thecurved interface should be perfectly reflecting. In designs Aand B, we relied on the natural reflection at the interfacebetween the nanorod array and air. The imperfect reflection,however, compromises the cloak performance. In order toincrease the reflectivity along the curved interface, we placed

FIG. 9. Side view of one of the larger nanorods in device A, showing thenonuniformity obtained in the sidewall thickness as an effect of the etchprocess.

FIG. 10. Block diagram of the design B nanorod optical cloaking structure.



a PBG structure behind the curved interface in design C. ThePBG structure is composed of a square array of silicon na-norods with a diameter of 312 nm and a periodicity of 520nm. This structure exhibits a PBG for TM polarizationaround the wavelength of 1500 nm, where the measurementswere done. The PBG structure consisted of a 62�10 array of520 nm spaced nanorods, each having a diameter of 312 nm.The PBG present in the band structure of a square latticephotonic crystal is detailed in another work.11 A block dia-gram for design C is shown in Fig. 12.

The proximity correction required for the fabrication fordesign C proved more challenging to adjust due to smaller

FIG. 11. Close-up of the reduced cloaking area in design B; damage toseveral pillars caused during the measurement phase of the device.

FIG. 12. Block diagram of the design C nanorod optical cloaking structure.



gaps between nanorods as compared with design A. Initially,only the default forward scattering correction parameterswere used as in design A to fabricate the device. In this case,the central cloaking area nanorods are slightly overexposed,resulting in oversized structures, leading to rod bridging inthe central cloak area, while the side area nanorods wereunderexposed, leading to missing and defective rod struc-tures. This result is shown in Fig. 13.

In order to correct for the electron beam proximity effectsin design C, both forward and back scattering correctionswere required �adjustments to the � and � values� as well asa process blur adjustment. Both the � and � parameters wereadjusted to 0.08 and 38.9, respectively, to compensate for thesmaller feature sizes and gaps between nanorods. The pro-cess blur parameter �1 describes the midrange scatteringrange of the electrons and was used to compensate for pro-cess related effects. The �1 parameter describes the ratio ofthe midrange Gaussian against the � Gaussian based on theirstored energy. This correction required the process blur pa-rameter �1 to be set to 60 nm and, given a weight �1 of 0.20or 20%, to compensate for the additional scattering correc-tion. Figure 14 shows the results of adding this additionalproximity correction. Some bridging is evident in the largestpillar structures, where the gap was designed to be 20 nm,but this was found to be unavoidable due to the higher doserequired to correctly expose the smaller cloak edge pillars.This bridging was expected to shift the operating frequencyof the device slightly in the measurements.

IV. MEASUREMENT RESULTS AND DISCUSSION

For optical characterizations, three fiber-coupled laserstunable between 1400 and 1602 nm were used as lightsource. A polarization control paddle was used to set thecorrect polarization of the laser, and the light output from thefiber was butt-coupled into the silicon input waveguide. The

FIG. 13. SEM illustrating the overexposure of the central nanorods �causingbridging� and underexposure of the side nanorods �causing missing struc-tures� in the cloaking section of the design C cloak when only forwardproximity scattering correction is used.

light that comes out of the fiber within the critical angle was



captured by the input waveguide and fed into the nanorodarray. The light propagation through the nanorod structurewas then directly visualized by the near-field scanning opti-cal microscopy �NSOM�. Figure 15 shows a NSOM imagefor a device A cloaking structure at a wavelength of 1500nm. A well-defined input beam was observed propagatingvertically from the bottom of the figure into the cloakingstructure, producing a spot of intense scattering visible at thereflecting interface. The out-of-plane scattering at the reflect-ing interface significantly reduces the reflected beam inten-sity, but is unavoidable in a two-dimensional �2D� imple-mentation, in which the guiding condition is compromiseddue to the abrupt interface. Despite losses at the reflectinginterface, a clearly defined reflected beam was observed at areflection angle of 45° with respect to the reflecting interface.The reflected light beam does not reach the output wave-guide, however, due to the propagation loss within the nano-rod array as well as the scattering losses at the reflectinginterface.

A similar behavior was observed at an operating wave-length of 1460 nm in the device B cloak, as shown in Fig.16. The off-tuning of this device could be attributed to thenonuniform pillar profiles as well as a slight change in indexdue to pillars being off in exact design dimensions within

FIG. 14. SEM image of the cloaking area in design C, showing both thecloak and PC bandgap pillar structures. A small amount of bridging in thecentral cloak area can be seen, but the smallest side nanorods have beencorrectly patterned using both forward and back scattering proximitycorrections.

FIG. 15. �Color online� NSOM image for 1500 nm laser light propagating
off the cloaking structure area in design A.


experimental tolerances. This operating wavelength shiftcould also be attributed to the index matching of the design,which is close to the index of the SiO2 underneath the Siguiding layer. The reduction in scattering due to the reduc-tion in the background nanorod array area improved thestrength of the reflected beam at the same measurementwavelength as compared to design A. The output beamstrength was again subjected to propagation and scatteringlosses, but in this case, they have been reduced significantly.

Figure 17 shows the results of a NSOM measurement ona device C large cloak sample taken at a wavelength of 1420nm, showing a clearly defined output beam reflecting fromthe cloaking surface. The reduction in the spacing in thepillars in the cloaking area reduced the reflection from in-coming light significantly, resulting in a more uniform outputbeam. The shift in operating wavelength may be due to theunavoidable bridging of some of the larger pillars that leadsto a large index change than expected in that area. However,the strength and clarity of the output beam are much im-proved over designs A and B. The change in index match ofthe structure may have also helped improve the output, al-though it is hard to determine the exact effect based on theevidence from the experimental measurements.

All three cloak design structures are created from non-resonant dielectric elements and are expected to operate wellover a broad range of frequencies. Due to the high scatteringof the output beam, far-field effects, such as the coupling ofthe light into the output waveguide, were not observed. Theperformance of the devices could be improved through fur-ther reduction in the background PC size as well as correct-ing for bridging pillars and nonuniform pillar etching, whichincreases the index average of the cloak in the large pillarareas, shifting the operating wavelength of the device. Thechange in background index of the cloak did not appear to

FIG. 16. �Color online� NSOM image for 1460 nm laser light propagatingoff the cloaking area in design B.

FIG. 17. �Color online� NSOM image for 1420 nm laser light propagating
off the cloaking area in design C.


have a noticeable effect on the device performance, althoughfurther investigation is required for a complete conclusion.

V. CONCLUSIONS

In conclusion, successful fabrication and characterizationof an optical cloaking photonic crystal structure consisting ofnanorods has been demonstrated using an in-house devel-oped cleanroom process. Several fabrication parameters werekey in fabricating the nanosize device features correctly. Theselection of the best high-contrast resist exposure and devel-opment process for the electron beam lithography step wascritical in patterning the nanorod structures successfully. Theadjustment of proximity correction parameters through mul-tiple exposure runs was also key in replicating the difficult tofabricate nanorod structures, especially with each rod havinga different size in this particular cloak design. Careful con-sideration of both forward �� and backward �� scatteringprocesses based on the feature size and spacing has to betaken when determining the optimum proximity correctionparameters. In addition, further refinements to the midrangescattering processes using �1 and �1 are necessary to com-pensate for additional features sizes and process dependentcorrections. The choice of the plasma etch recipe is alsocritical in order to reduce scattering in the final device per-formance. Allowing for the creation of near-vertical rod side-walls while maintaining a high selectivity ratio with the re-spect to the resist is important to successful cloakingbehavior in the device.

Some refinements could be made to this set of processesto improve the device fabrication quality. Optimization of theplasma etch recipe to obtain more vertical sidewalls is onearea for improvement, as the nanorods in the current processshow some sidewall irregularities. Further adjustments to theproximity correction dose pattern and scattering correctionmay allow for improvement in the problem between overde-velopment and underdevelopment of the various sized nano-rods in the cloak arrays. For example, reducing the dose inthe area of the largest nanorods by manually changing theproximity correction dose pattern may allow the bridging inthis area to be eliminated.

This two-dimensional nanorod cloak design represents afirst step in an exciting and innovative field of study that willhave a significant impact on the field of transformation op-tics. While three-dimensional carpet cloak has recently beendemonstrated,12 2D waveguide based transformation opticaldevices will continue to remain an important class of de-vices. As has been demonstrated in the carpet cloak, severalslightly different fabrication approaches have been em-ployed. The e-beam lithography based approach presented inthis article provides precision control over the nanoscale ge-ometries while at the same time capable of fabricating largearea devices. We therefore believe it is one of the most effi-cient ways for manufacturing this new and exciting class ofdevices. Further work will explore adding passive and activetunability properties to the device by the use of optical and
electro-optical coatings and materials.


ACKNOWLEDGMENTS

The authors would like to thank the staff of the GT MIRCfor their assistance and support in the fabrication effort out-lined in this article. The work at the University of Coloradowas supported in part by the Army Research Office MURIContract No. 50432-PH-MUR.

1J. B. Pendry, D. Schurig, and D. R. Smith, Science 312, 1780 �2006�.2D. Schurig, J. B. Pendry, and D. R. Smith, Opt. Express 14, 9794 �2006�.3D. R. Smith, J. B. Pendry, and M. C. K. Wiltshire, Science 305, 788�2004�.

4D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S.Schultz, Phys. Rev. Lett. 84, 4184 �2000�.

5D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B. Pendry, A. F.



Starr, and D. R. Smith, Science 314, 977 �2006�.6J. Li and J. B. Pendry, Phys. Rev. Lett. 101, 203901 �2008�.7R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui, and D. R. Smith, Science

323, 366 �2009�.8J. H. Lee, J. Blair, V. A. Tamma, Q. Wu, S. J. Rhee, C. J. Summers, andW. Park, Opt. Express 17, 12922 �2009�.

9J. Valentine, J. Li, T. Zentgraf, G. Bartal, and X. Zhang, Nature Mater. 8,568 �2009�.

10L. H. Gabrielli, J. Cardenas, C. B. Poitras, and M. Lipson, Nat. Photonics3, 461 �2009�.

11John D. Joannopoulos, R. D. Meade, and Joshua N. Winn, Photonic Crys-tals: Molding the Flow of Light, 2nd ed. �Princeton University Press,Princeton, NJ, 1995�, p. 137.

12N. Liu, H. Guo, L. Fu, S. Kaiser, H. Schweizer, and H. Giessen, NatureMater. 7, 31 �2008�.


challenges in the fabrication of an optical frequency ...ecee.colorado.edu/~wpark/papers/j. vac....

Documents