Fabrication of Nanoring Arrays bySputter Redeposition Using PorousAlumina TemplatesKevin L. Hobbs, Preston R. Larson, Guoda D. Lian, Joel C. Keay, andMatthew B. Johnson*

Center for Semiconductor Physics in Nanostructures,Department of Physics and Astronomy, UniVersity of Oklahoma,Norman, Oklahoma 73019

Received September 26, 2003; Revised Manuscript Received October 17, 2003

ABSTRACT

Ordered arrays of Au, Ni, and Si nanorings have been fabricated using Ar+ sputter redeposition of material in a porous alumina mask. Typicalring dimensions are 50 nm inner diameter and 10−15 nm wall thickness with heights ranging from 50 to 200 nm. Ring composition wasconfirmed by electron microscopy. Ring diameter, height, and spacing are controllable by varying the process conditions. This process isscalable and parallel, so that highly ordered nanorings over millimeter-sized regions are possible.

Currently there is a great deal of interest in nanometer-sizedrings (nanorings) from theoretical, experimental, and deviceperspectives. Of interest are such properties as: persistentcurrents in metallic1 or superconducting rings,2 tunableoptical resonance in metal rings,3 novel magnetoopticalbehavior in semiconductor rings,4 magnetic response forapplication in patterned perpendicular recording media5,6 andnovel ring-shaped MRAM.7 Collective properties of orderedarrays, involving coupling between adjacent rings, mayexhibit new physical behavior. Such collective propertieshave received attention. In particular, collective excitationsin lattices have been theoretically investigated.8,9

In the past, nanorings have been fabricated using variousmethods including electron beam techniques1,10 and nano-sphere lithography.11 The alumina template methods, intro-duced here, have advantages over these two techniques; inparticular, the alumina template methods are scalable andparallel, providing a method to fabricate large, orderednanoring arrays. Moreover, we can control the nanoringheight independent of the ring diameter to produce highaspect ratio rings.

In this paper, ordered arrays of Au, Ni, and Si nanoringshave been fabricated using Ar+ sputtering and porous anodicaluminum oxide (AAO) templates. Two methods are pre-sented, both relying on the redeposition of sputtered materialinside the pores of an AAO mask. Typical nanoringdimensions are 50 nm inner diameter with 10 to 15 nm wallthickness and heights ranging from 50 to 200 nm. Ringcomposition and phase were confirmed by energy dispersive(X-ray) spectroscopy (EDS) and selected area electron

diffraction (SAED). Ring diameter, height, and spacing arecontrollable by varying the process conditions.

Figure 1 schematically shows the two methods used tofabricate nanorings. In method I, Figure 1a-c, a through-hole alumina template is used as a sputter mask on 20 nmof SiO2 on a thick layer of the desired ring material. In Figure1a the alumina template is on the sample. Figure 1b showsthe results of ion etching through the AAO pores leading tothe redeposition of ring material around the pore walls.Finally, the AAO template is removed (Figure 1c), leavingbehind the sputter redeposited rings on the SiO2 layer. Inmethod II, Figure 1d-f, the through-hole alumina templateis used as an evaporation shadow mask to define an arrayof dots as shown in Figure 1d. This is followed by an ion-etching step that leaves behind sputter redeposited dotmaterial around the pore walls as shown in Figure 1e. Afterremoval of the AAO template an array of nanorings or tubesremains as shown in Figure 1f. In the paragraphs below, wediscuss growth and transfer of the alumina template, andsubsequent evaporation, sputtering, and processing steps ingreater detail.

The AAO templates are fabricated using a two-stepanodization process described in detail elsewhere.12-14 Thistwo-step process starts with anodizing an aluminum foil fora long time (15 h) to grow a thick porous layer. This is thenchemically stripped off and followed by a second anodizationstep for a short time (5 min). Using this two-step technique,very good ordering over micron sized areas is obtained. Morerecently, a nanoimprinting technique designed to seed poreformation has achieved ordering over millimeter sized

NANOLETTERS

2004Vol. 4, No. 1

167-171

10.1021/nl034835u CCC: $27.50 © 2004 American Chemical SocietyPublished on Web 11/26/2003

areas.15,16 In addition to long range ordering, this nanoim-printing technique also creates remarkably regular, columnarpores. In the work reported here, the anodization was doneeither in 0.3 M oxalic acid at 40 V (producing 50 nmdiameter pores spaced 100 nm apart) or 0.3 M sulfuric acidat 27 V (producing 20 nm diameter pores spaced 60 nmapart). All anodizations were carried out at 1°C. After thesecond anodization, a through-hole mask about 300 nm thickwas prepared by separating the AAO film from the Al foilin a saturated HgCl2 solution and removing the remainingbottom alumina barrier layer in a solution of H3PO4 (5 wt%) at 30°C for 35 min. The resulting mask was then placedon the substrate. Typical mask dimensions were 3 mm× 3mm.

To evaporate appreciable material down the AAO tem-plate, it is crucial that the divergence of the evaporatedmaterial is lower than about 10° to accommodate the 6:1aspect ratio of the alumina mask. Our evaporations werecarried out using an Edwards E306A thermal evaporator witha source-to-sample distance of 18 cm and a typical sourcesize of 5 mm× 5 mm. Under these conditions evaporatinga 60 nm thick layer on the top surface of the AAO yields 40nm high nanodots and coats the top portion of the sidewallsof the AAO template, as shown schematically in Figure 1d.

This coated top portion is important as it contributes to thefinal nanoring structure, specifically the height, because thesputter step is highly directional and does not remove muchof this material.

As with the evaporation, it is important to sputter with alow divergence ion beam. In this work, a VG MicrotechEX05 ion source with a source-to-sample distance of 50 mmand an angular divergence of about 0.1° was used. We sputterwith Ar+ at an energy of 500 eV and current density of 0.15mA/cm2. The low incoming energy of the Ar+ results in anunder-cosine (flatter) distribution of emitted sputtered par-ticles.17 This flat distribution is better for the sputterredeposition leading to rings or tubes. The sputter time(typically 3 min) was chosen to remove the entire bulkevaporated film from the alumina template surface.

The AAO can be removed from the sample using anappropriate wet etch. In this work equal parts of chromic/phosphoric acid (1.8 wt %/6 wt %) solution was used at 60°C. Although removing the template is useful for charac-terization, it is not necessary, and rings left in the templatehave greater mechanical stability.

Samples were characterized with a JEOL 880 SEM; aJEOL 2000 FX TEM in TEM, secondary electron imaging(SEI), SAED, and EDS modes; and a JEOL 2010 F TEM inSTEM dark-field imaging and EDS modes. Samples preparedfor SEM analysis were fabricated on Si substrates with theAAO mask removed after the sputtering step. TEM sampleswere fabricated either on Si or directly on ultrathin carbonfilm TEM grids.18 TEM samples on Si were mechanicallypolished and then thinned with an Ar ion mill. Samplesfabricated on TEM grids required no further processing andwere examined with the AAO template intact.

Figure 2 shows Si rings on a SiO2/Si substrate fabricatedusing method I. Figure 2a is a top-down SEM image showingwell-ordered 50 nm inner diameter, 10 nm thick rings with100 nm center-to-center spacing. Size variations of thesedimensions were about 20% (typical for method I). The insetshows a 45° view from which we determine the ring height(Si plus SiO2) to be 50 nm. Figure 2b is a TEM image of athinned sample with an SAED pattern (inset). The sharpdiffraction peaks are from the crystalline substrate (Si (001)),while diffuse rings indicate that the nanorings are amorphousSi. The difference in SEM and TEM contrast is primarilythe result of reduced secondary electron yield from the etchedhole leading to the dark region in the SEM image. Finally,Figure 2c is a plot of a Si K-edge EDS line scan across asingle nanoring, shown on the STEM annular dark fieldimage (inset). This line scan clearly indicates reduced Si inthe center of the ring, as expected because the hole extendsinto the Si substrate (as seen directly in cross-sectionalviews). The line scan also shows reduced Si at either endoutside the ring, although to a lesser extent than in the center.Ni rings were also produced by this method. Figure 3 showsNi rings on a 20 nm thick SiO2 layer on Ni on Si substrate(inset shows a 45° view). These rings have similar dimen-sions to the Si rings described above.

Nickel rings have also been fabricated on Si using methodII. The electron microscopy results are summarized in Figure

Figure 1. Schematic views of method I and method II. For methodI: (a) shows the AAO template on 20 nm of SiO2 on the desiredring material, (b) shows sputter redeposited material around thepore walls after sputter etching, (c) shows the rings after AAO maskis removed (with a plan-view added to guide the eye). For methodII: (d) shows AAO mask on a silicon substrate after ring materialis evaporated down the mask, (e) shows sputter redeposited materialaround the pore walls after sputter etching, (f) shows the rings afterAAO mask is removed.

168 Nano Lett., Vol. 4, No. 1, 2004

4. Figure 4a is a large-scale view showing the typical micron-sized, well-ordered domains. Figure 4b is a close-up viewshowing a region of 50 nm ID, 10 nm thick rings separatedby 100 nm. Size variations for these dimensions were about10% (typical for method II).Top-right inset is a cross-sectional image of a related sample showing that the sputter-etched holes continue into the Si substrate. The top left insetshows a 55° view from which we determined a nanoringheight of 200 nm. The bottom left inset shows nanoringsfabricated with a sulfuric anodized AAO template. In thiscase, the rings have an inner diameter of about 35 nm, anouter diameter of 60 nm and a center-to-center spacing of

about 60 nm. Figure 4c shows an EDS spectrum from athinned sample prepared for scanning transmission electronmicroscopy. The SAED pattern (shown inset) indicates thepresence of Ni and NiO. Both results confirm the rings areNi, (the small Cu signal is from the TEM grid).

Similar results have been obtained with Au rings on a Sisubstrate and on an ultrathin carbon TEM grid using methodII. The electron microscopy results are shown in Figure 5.Figure 5a is a close-up view showing a well-ordered regionof 50 nm inner diameter, 10 nm thick rings separated by100 nm. Bottom-left inset shows a 45° view from which wedetermined the nanoring height to be approximately 150 nm.Figure 5b shows details from TEM analysis of the rings stillin the alumina template. The TEM image is shown inset (topcenter) and an SAED pattern (top-right inset). The EDSspectrum shows the presence of gold (again the Cu signal isfrom the TEM grid). The electron diffraction image (top-right inset) indicates that the rings are polycrystalline gold.

As described above, we have fabricated well-orderedarrays of rings with some variation in ring geometry. Theprospects for further controlling the ring and array geometrybased on our results with methods I and II are discussed inthe paragraphs below. For both methods, the center-to-centernearest neighbor spacing of the rings is the same as theinterpore spacing of the AAO template. This is controlledby details of the anodization conditions. The anodizationvoltage primarily controls the interpore spacing with theelectrolyte concentration playing a significant role in the porediameter and pore ordering. Interpore spacing of about 60nm (sulfuric), 100 nm (oxalic), and 500 nm (phosphoric)are typically used to obtain well-ordered pore arrays.However, using the proposed “10% porosity rule”,19 interporespacing between these values can be obtained while stillmaintaining a significant degree of ordering. Again, it isworth noting that long range order extending over millimeterscan be obtained using nanoindenting techniques to initiatepore formation.15,16

Figure 2. Electron microscopy results for Si nanorings fabricatedusing method I. (a) Top-down SEM micrograph of rings on a SiO2/Si substrate. Inset is a 45° oblique view showing the 50 nm tallrings (same mag.). (b) TEM image of the Si rings and SAED pattern(inset). (c) EDS profile of Si K-edge across the dotted line shownon the STEM annular dark field image of a single Si ring (inset).

Figure 3. Top-down SEM micrograph of Ni rings on SiO2/Sisubstrate fabricated using method I. Inset is a 45° oblique viewshowing the 50 nm tall rings (same mag.).

Nano Lett., Vol. 4, No. 1, 2004 169

We observe that the rings produced by method I are shorterthan those produced by method II. In method I, the ringheight is governed primarily by the incident ion energy. Forthe 500 eV Ar+ incident energy used in this work, the sputterdistribution is under-cosine resulting in short rings. Higherion energy with closer to cosine sputter distributions shouldresult in taller rings. One disadvantage of method I is that

the rings are above a bulk layer of the same material. Thiswould make measurement of many properties difficult, asthe underlying layer would dominate the signal from thenanorings. However, the SiO2 separation layer can, inprinciple, be selectively removed to allow the rings to betransferred to another substrate. With method II, nanoringheight is a combination of effects including the AAOtemplate thickness, details of the evaporation, and theincident ion energy. As shown schematically in Figure 1,the final ring is a combination of material redeposited fromthe dot and that coating the top-portion of the AAO pore.This has been confirmed by comparing dot (prior tosputtering) and ring (after sputtering) volumes. Dot volumewas ascertained from atomic force microscope measurements.The final ring height is slightly smaller than the thicknessof the template. Thus, reducing or increasing the templatethickness allows us to reduce or increase the ring height. Inthis regime, we expect the height of the shortest rings willbe close to the thickness of the thinnest AAO we can use,about 100 nm. On the other hand, much taller rings shouldresult from using thicker templates, increasing the amountof evaporated material, and decreasing the angular divergence

Figure 4. Electron microscopy results for Ni rings fabricated usingmethod II. (a) SEM image of a typical region. (b) SEM image ofa well-ordered 1µm × 1 µm region showing rings with innerdiameters roughly 50 nm with wall thickness around 15 nm. Allinsets are the same scale. Top right: a cross-sectional view showingthe sputtered hole in the Si substrate. Top left: a 55° oblique viewshowing the 200 nm tall rings. Bottom left: top down view of Nirings fabricated in sulfuric anodized alumina, period is halved. (c)EDS spectrum of Ni rings on a Si substrate. Indexed SAED pattern(inset) and spectrum confirms the rings are Ni and NiO (Cu peakis due to the TEM grid).

Figure 5. Electron microscopy results for Au rings fabricated usingmethod II. (a) Top-down SEM micrograph of Au rings on a Sisubstrate with 45° oblique view (inset) showing 150 nm tall rings.(b) EDS spectrum of Au rings still inside the AAO template. Thespectrum and indexed SAED pattern (inset) confirm the rings arepolycrystalline gold.

170 Nano Lett., Vol. 4, No. 1, 2004

of the evaporated material. Under these circumstances, weexpect to be able to fabricate rings with aspect ratios evenhigher than the 4:1 ratio demonstrated here. For sufficientlythick templates it should be possible to have the rings formedonly from the sputtered dot material, with the top portionmaterial not contributing. Under these circumstances weexpect that rings with heights less than 50 nm are possiblewith method II.

It is interesting to note that the outer diameters of free-standing rings often exceed the inner diameter of the as-grown AAO pore. For example, in Figure 5a, the outerdiameter is about 70 nm, while in the TEM image (inset ofFigure 5b) the rings in the AAO matrix have an outerdiameter of about 50 nm. A more dramatic example is shownin the SEM image of Ni rings fabricated using a sulfuricacid anodized template (lower-left inset Figure 4b). For asulfuric acid anodized AAO film, the pores are 20 nm indiameter with interpore spacing of 60 nm. The Ni rings haveinner diameters of about 35 nm and outer diameters near 60nm, while still maintaining the expected 60 nm period sothat the rings are nearly touching. The discrepancy isattributed to the widening of the pores during the barrier-layer opening wet etch step. Thus we have some control overthe outer diameter of the rings independent of the spacing.This is important for tailoring the coupling between neigh-boring rings. Currently we are exploring ways to improvethe control of the barrier-layer removal step. Recent develop-ments of a dry-etching step to more controllably remove thebarrier layer may improve this process.20

In summary, two methods for producing nanoring arraysthrough the sputter redeposition of Ni, Au, and Si in porousalumina templates have been presented. The composition andphase of the rings were verified by SAED and EDS. Ringsfabricated using oxalic acid AAO templates produced ringsof 50 nm inner diameter, 10 nm wall thickness and 100 nmcenter to center spacing. Deviations from these dimensionsare about 20% and 10% for method I and method II,respectively (using AAO films fabricated using nanoimprint-ing, the array and ring geometery uniformity will beimproved). Ring diameter, height, and spacing are control-lable by varying the process conditions. In particular, withmethod II a large variation in ring height can be achieved.

And for both methods, the outside ring diameter can beincreased until the rings touch. Finally these methods arescalable and parallel, which are important considerations forthe development of nanoscale arrays for applications.

Acknowledgment. This work was supported by the NSFunder Grants No. DMR-0080054, DMR-0209371, and ECS-0132534. The SEM and TEM analyses were conducted inthe Samuel Roberts Noble electron microscopy Laboratoryat University of Oklahoma. Finally, the authors thank Dr.Jinguo Wang at Penn State University for STEM, ADF andEDS analysis.

References

(1) Jariwala, E. M. Q.; Mohanty, P.; Ketchen, M. B.; Webb, R. A.Phys.ReV. Lett. 2001, 86, 1594.

(2) Matveev, K. A.; Larkin, A. I.; Glazman, L. I.Phys. ReV. Lett.2002,89, 096802.

(3) Aizpurua, J.; Hanarp, P.; Sutherland, D. S.; Ka¨ll, M.; Bryant, GarnettW.; Garcia de Abajo, F. J.Phys. ReV. Lett. 2003, 90, 057401.

(4) Lorke, Axel; Luyken, R. Johannes; Govorov, Alexander O.; Kotthaus,Jorg P.; Garcia, J. M.; Petroff, P. M.Phys. ReV. Lett.2000, 84, 2223.

(5) Nielsch, K.; Hertel R.; Wehrspohn, R. B.; Barthel, J.; Kirschner J.;Gosele, U.; Fischer, S. F.; Kronmu¨ller, H. IEEE Trans. Magnet.2002,38, 2571.

(6) Li, S. P.; Peyrade, D.; Natali, M.; Lebib, A.; Chen, Y.; Ebels, U.;Buda, L. D.; Ounadjela, K.Phys. ReV. Lett. 2001, 86, 1102.

(7) Zhu, J.; Zheng, Y.; Prinz, G.J. Appl. Phys.2000, 87, 6668.(8) Huang, D.; Gumbs, G.;Phys. ReV. B 1992, 46, 4147.(9) Roostaei, B.; Mullen, K., unpublished.

(10) Levy, L. P.; Dolan, G.; Dunsmuir, J.; Bouchiat, H.Phys. ReV. Lett.1990, 64, 2074.

(11) Winzer, M.; Kleiber, M.; Dix, N.; Wiesendanger, R.Appl. Phys. A1996, 63, 617.

(12) Masuda, H.; Fukuda, K.;Science1995, 268, 1466.(13) Li, A. P.; Muller, F.; Birner, A.; Nielsch, K.; Go¨sele, U.J. Appl.

Phys.1998, 84, 6023.(14) Jessensky, O.; Mu¨ller, F, Gosele, U.J. Electrochem. Soc.1998, 145,

3735.(15) Masuda, H.; Yamada, H.; Satoh, M.; Asoh, H.; Nakao, M.; Tama-

mura, T.Appl. Phys. Lett.1997, 71, 2770.(16) Asoh, H.; Nishio, K.; Nakao, M.; Tamamura, T.; Masuda, H.J.

Electrochem. Soc.2001, 148, B152.(17) Smith, R.; Makh, S. S.; Walls, J. M.Philos. Mag. A1983, 47, 453.(18) Ted Pella Type B carbon 300 TEM grid.(19) Nielsch, K.; Choi, J.; Schwirn, K.; Wehrspohn, R. B.; Go¨sele, U.

Nano Lett.2002, 2, 677.(20) Xu, Tao; Zangari, Giovanni; Metzger, Robert M.Nano Lett.2002,

2, 37.

NL034835U

Nano Lett., Vol. 4, No. 1, 2004 171