Fabrication of One-Dimensional Programmable-Height Nanostructures via Dynamic Stencil Deposition
J. L. Wasserman, K. Lucas, S. H. Lee, A. Ashton, C. D. Crowl, N. Markovic
FFabrication of One-Dimensional Programmable-HeightNanostructures via Dynamic Stencil Deposition
October 24, 2018
J. L. Wasserman, K. Lucas, S. H. Lee, A. Ashton, C. D. Crowl, N. Markovi´cJohns Hopkins University, Baltimore, MD 21218
Abstract
Dynamic stencil deposition (DSD) techniques offer a variety of fabrication advan-tages not possible with traditional lithographic processing, such as the the ability todirectly deposit nanostructures with programmable height profiles. However, DSDsystems have not enjoyed widespread usage due to their complexity. We demon-strate a simple, low-profile, portable, one-dimensional nanotranslation system thatfacilitates access to nanoscale DSD abilities. Furthermore we show a variety of fabri-cated programmable-height nanostructures, including parallel arrays of such structures,and suggest other applications that exploit the unique capabilities of DSD fabricationmethods.
Fabrication of nanostructures by shadow mask deposition has been demonstrated in re-cent years as a means to work beyond lithographic limits. Metal can be evaporated througha nanopore in a thin membrane, leaving a nanodot as small as 10nm [1, 2]. Translationof the nanopore relative to the substrate during deposition, a technique known as dynamicstencil deposition (DSD), allows nanoscale features to be drawn directly onto the substrate.DSD has been recently demonstrated by a variety of methods such as using twin single-axispiezo actuators [3], uncontrolled thermal motion of the mask relative to the substrate [4],exploiting parallax through rotation of the angle of the evaporation source relative to themask [1], and physically translating the mask relative to the substrate [5]. More sophisti-cated systems have coupled the DSD mechanism to an AFM tip, allowing alignment of thedeposited nanostructures with existing sample features via surface probing techniques [2, 6]1 a r X i v : . [ c ond - m a t . m e s - h a ll ] F e b tencil mask deposition offers several advantages over traditional lithographic techniques.For example, employing a shadow mask stencil allows one to deposit patterned metal layerson surfaces or systems where the thermal and chemical processing of traditional lithographycannot be tolerated [7, 8].DSD techniques offer many additional advantages, a primary benefit being the abilityto fabricate a nanostructure with a controllable height profile. Modulation of the speed ofthe mask’s motion allows variation of the deposited feature height. Assuming a constantdeposition rate, the local height of the nanostructure is proportional to the integrated timethe nanopore mask uncovers it. In the static case, if a mask is held rigidly in place duringdeposition for an interval of time, the deposited structure will possess a height profile M ( x )where x is the longitudinal position variable. Translating the mask’s position along the ˆ x axisduring deposition therefore yields a structure with a height profile given by the convolution h ( x ) = c (cid:90) t ( x (cid:48) ) M ( x − x (cid:48) ) dx (cid:48) (1)where t ( x ) refers to the time the mask resides at longitudinal position x , and c scales as thematerial deposition rate [6].DSD systems tend to be fairly complicated and difficult to design and implement, therebyinhibiting the widespread incorporation of their techniques into nanoscale research. We haveconstructed a compact low-cost DSD device that easily allows creation of one-dimensionalprogrammable-height nanostructures. The device’s small size and simplicity allow it to beeasily adaptable to a variety of deposition chambers and systemsOur stencil masks consist of a suspended silicon nitride membrane 50nm thick withnanopore apertures[4]. The membranes are created from Double-Side Polished (cid:104) (cid:105) Siwafers, with 50 nm of low-stress Si N grown via LPCVD on each side. The nitride layer isselectively removed on one side through photolithography and reactive ion etch of CF / O .The substrate is anisotropically etched in a KOH bath, with both nitride layers acting asprotective masks, leaving a suspended nitride window on the back side. Nanopores of variousdiameters are fabricated in this membrane, which then acts as the shadow mask.We use two methods to create nanopores in the membrane. Pores can be milled directlythrough the nitride with a Focused Ion Beam (FIB). In this case we typically coat themembrane with a 10nm metallic (Cr or Pt) conductive layer to aid visualization in theSEM for alignment of the FIB. Pores as small as 50nm can be easily produced with theFIB. Likewise a layer of PMMA can be patterned with electron beam lithography, and themembrane subsequently etched in a CF / O plasma. The PMMA layer is best removed in apure O plasma than with an organic solvent, to preserve the membrane and pore integrity.In either case, the nanopores produced tend to be quite larger than desired, but even a porea few hundred nanometers in diameter can be shrunk down to a size of a few nanometers by2on-beam or electron-beam sculpting [9, 10, 11, 12].The DSD nanotranslation device ultimately needs the ability to press a mask firmlyagainst a substrate, while allowing translation of the mask as substrate remains stationary.The mask is translated relative to the substrate by means of a piezoelectric acuator, mountedin a mechanical assembly that allows translation in a single dimension. The piezo actuatoris used in an open-loop fashion, for added simplicity, whereby it extends fairly linearly ina single dimension by an amount proportional to an applied DC voltage. Our actuatortranslates a maximum of 6 microns at 100 volts. The actuator is firmly situated betweentwo PEEK holders, and pushes a carriage along four rods, kept under tension by four springs,as shown in Figure 1. The substrate is mounted to a substrate holder which moves alonga pair of vertical rods, also under spring-loaded tension. The spacing between the maskand substrate is controlled by three jacking screws, which keep the substrate firmly pressedagainst the mask during the translation process. The carriage and substrate holder rodsare lubricated with a high vacuum grease to allow smooth motion. The clearance of therods through the carriage and substrate holder is critical, there must be enough play toallow motion but enough constriction to provide the proper controllable one-dimensionalconstraint.The finite solid angle of the evaporation source causes spreading of the deposited featuresize beyond the size of the nanopore [4, 13]. Geometrical considerations reveal that the sizeof the deposited feature is approximately given by the relation w dot ≈ w hole + 2( d/L ) w source (2)The smallest possible features can be created if one minimizes the size of the evaporationsource w source and the separation gap between mask and substrate d . Usually L is constrainedphysically by the size of the deposition chamber. On some occasions, for example when mul-tiple angle evaporations of different materials are desired, it is useful to limit the separationgap d to a known value by means of a spacer layer of silica spheres [1]. Without using aspecific spacer layer, the gap is limited to the size of any dust or contaminants residing onthe substrate or mask surface, usually not much smaller than a few hundred nanometers. Inorder to minimize d , and hence our feature size, we opt not to use a spacer layer, and insteadwe push the mask directly against the substrate. Typically our cleanroom prevents d frombeing less than 2 - 3 µ m, limiting the minimal feature sizes to about 20 to 30 nm, althoughbetter cleanroom facilities should be able to easily improve on this number.The size of the evaporation source w source can be reduced to about a millimeter by em-ploying an aperature ‘plug’. Gold metal is evaporated from a conical wire crucible with anoxide coating, and the molten gold forms a bead of approximately 2 - 3 mm. To reducethis evaporation size a thin tungsten circular disc about 3 mm in diameter with a 1 mm3perature is placed inside the crucible above the gold. The tungsten plug gets hot enoughduring the evaporation process that the gold wets the bottom plug surface and evaporatesthrough the aperture, but does not significantly wet the top surface. Careful calibration ofthe plug height inside the crucible may be required to find an ideal position where the topsurface remains un-wetted. This method allows efficient usage of source material, as thegold remains molten on the bottom surface of the plug. Visual inspection of the plug duringdeposition reveals it is fairly close to the same temperature as the crucible, by the blackbodycolor. Other attempts to reduce source aperture size by employing a colder aperture plateatop the crucible did work, but significant quantities of source material were wasted and thedeposition rates were much slower. We expect this aperture size can be further reduced bya factor of two.A variety of nanostructures with programmable height profiles and feature sizes downto 50 nm have been created, and with further optimization we expect to produce featuresas small as 10 nm. A gold nanowire of width 100nm, with a narrow height constriction inthe middle, is shown in Figure 2, illustrating that the wire is substantially thicker at theends than in the middle. Depending on the deposited material, such a structure could be ananowire with a potential barrier in the middle, or an implementation of an S-c-S
JosephsonJunction. Controllable placement of peaks and valleys in the nanowires allows novel study ofinhomogeneous nanowires as repeatable wires can be fabricated, and ‘grains’ can be preciselypositioned in a controllable fashion. All of these systems would be measurable in situ , withoutthe need to break vacuum or perform any other processing [14]. Such capabilities create theopportunity to measure properties of highly-oxidizable complicated nanostructures, whichcould not be fabricated in any other way.Arrays of nanostructures can be simultaneously written by using a mask with a corre-sponding array of nanopores. An AFM scan of an array of uniform double-valley nanowiresis shown in Figure 3. Each nanowire has two intermediary shallow points, and the unifor-mity between the nanowires in the array is quite good. A parallel array of nanoramps isshown in Figure 4.a, with a cross-sectional height profile slice explicitly shown in Figure4.b. This sample used varying oblong shapes for the nanopores, which accounts for the non-uniformity between ramps. The arrays can be over an area as wide as the mask, allowingfairly large-scale nanostructure arrays to be created. The direct ability to create identicalprogrammable height profiles can have profound implications in the field of metamaterials[15]. Additionally the large-scale parallel array will allow Vibrating Sample Magnetometrymeasurement of magnetic nanostructure systems that were not previously possible.Interesting structures can be created by overlapping layers through discrete translationsteps. Figure 5 demonstrates a series of overlapping circles of various diameters, as fivestationary growth periods were separated by four equal-spaced translations. The four trans-lations were rapid enough that no appreciable material was deposited in the interim. Overlap-4ing deposition can be used with different materials to create novel devices [5, 16]. Howeverif the translation distance is larger than the nanopore size, one can create arrays of structureswith irregular spacing and of varying thickness.As with any technique, there are some limitations of the system. The motion of thepiezoelectric crystal in our device is limited to one dimension, but the ability of the deviceto translate in a straight line depends on the mechanical rigidity of the carriage movingon the rods. As observed in the figures, most fabricated structures do exhibit some lateralmovement perpendicular to the motion of the piezo, usually no greater than 200 nm for thefull range of motion. Of all nanostructures created, the largest lateral drift observed has been400 nm over 5 µ m length. We believe this is due to small amounts of slop from the rods andcarriage, as well as thermal expansion. To remedy the former one can employ tighter-fittingrods, and for the latter a method for liquid cooling of the assembly to maintain constanttemperature.For some one-dimensional applications, such lateral motion is not of great concern as themanufactured devices are still parametrically one-dimensional. If one is interested in makingmetallic or superconducting nanostructures, small curvature should not greatly affect theelectronic boundary conditions of the device. However this limitation may be of greaterconcern for magnetic nanostructures.We also envision using this nanotranslator assembly for other applications requiringcarefully-controlled motion techniques. By turning the device upside-down, droplets of fluidcan be placed in the pyramidal wells of the mask. This allows zeptoliter-scale quantities offluid to be dispensed through the nanopores, onto a subsurface. Control of the dispensingrate is achieved through the same velocity modulation of the piezo actuator used for pro-grammable height profiling. Such a dispensing mechanism has strong implications in thebiotechnological sciences.We wish to acknowledge Michael Fischbein from the University of Pennsylvania for usefuldiscussions, Huy Vo from the Johns Hopkins MBE cleeanroom for invaluable assistance, andSteve Patterson and Scott Spangler from the Johns Hopkins machine shop for precisionmachining. References [1] M. M. Deshmukh, D. C. Ralph, M. Thomas, and J. Silcox, Appl. Phys. Lett., , 1631(1999)[2] A. R. Champagne, A. J. Couture, F. Kuemmeth, and D. C. Ralph, Appl. Phys. Lett., , 1111 (2003) 53] K. Ono, H. Shimada, S. I. Kobayashi, and Y. Ootuka, Jpn. J. Appl. Phys. , 2369(1996)[4] J. K¨ohler, M. Albrecht, C. R. Musil, and E. Bucher, Physica E, , 196 (1999)[5] Z. Racz, J. L. He, S. Srinivasan, W. Zhao, A. Seabaugh, K. P. Han, P. Ruchhoeft, andJ. Wolfe, J. Vac. Sci. Technol. B, , 74 (2004)[6] S. Egger, A. Ilie, Y. T. Fu, J. Chongsathien, D. J. Kang, and M. E. Welland, NanoLett., , 15 (2005)[7] Y. X. Zhou and A. T. Johnson, Nano Lett., , 1371 (2003)[8] N. Takano, L. M. Doeswijk, M. A. F. van den Boogaart, J. Auerswald, H. F. Knapp, O.Dubochet, T. Hessler, and J. Brugger, J. of Micromech. and Microeng., , 1606 (2006)[9] J. Li, D. Stein, C. McMullan, D. Branton, M. J. Aziz, and J. A. Golovchenko, Nature, , 166 (2001)[10] D. Stein, J. L. Li, and J. A. Golovchenko, Phys. Rev. Lett., , 276106 (2002)[11] M.Y. Wu, D. Krapf, M. Zandbergen, H. Zandbergen, and P. E. Batson, Appl. Phys.Lett., , 113106 (2005)[12] H. Chang, S. M. Iqbal, E. A. Stach, A. H. King, N. J. Zaluzec, and R. Bashir, Appl.Phys. Lett., , 103109 (2006)[13] L. Gross, R. R. Schlittler, G. Meyer, A. Vanhaverbeke, and R. Allenspach, Appl. Phys.Lett., , 093121 (2007)[14] S. Egger, A. Ilie, S. Machida, and T. Nakayama, Nano Lett., , 3399 (2007)[15] R. A. Shelby, D. R. Smith, and S. Schultz, Science, , 77 (2001)[16] S. Egger, S. Higuchi, and T. Nakayama, J. of Comb. Chem., , 275 (2006)6 ask Front ViewSide View
SubstrateJacking ScrewSpringsPiezo
Figure 1: Diagram of the piezo assembly for mask translation. The rods are secured in anouter chassis (not shown), which also constrains the jacking screws.7 . ! m . ! m . ! m . ! m ab
200 nm 8.5 nm0 nm
Figure 2: AFM surface plot (a) and height map (b) of a single-valley gold nanowire of width100 nm. 8 ! m0 nm20.0 nm Figure 3: AFM height map of a uniform parallel array of double-valley gold nanowires.9
20 nm-20 nm0 nm 2 ! m 4 ! m 6 ! m 8 ! m ab ! m Figure 4: AFM scan of an array of chromium nanoramps shown as a surface plot (a). Thecross-sectional height profile of the black line is shown in (b). Height profiles vary betweennanoramps due to differing pore shapes. The triangle in upper image is a cursor.10 ! mm