Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Nano Lett. Author manuscript; available in PMC 2006 April 7.
Published in final edited form as:
PMCID: PMC1432218



Nanostructures can be patterned with focused electron or ion beams in thin, stable, conformal films of water ice grown on silicon. We use these patterns to reliably fabricate sub-20 nm wide metal lines and exceptionally well-defined, sub-10 nanometer beam-induced chemical surface transformations. We argue more generally that solid phase condensed gasses of low sublimation energy are ideal materials for nanoscale patterning, and water, quite remarkably, may be amongst the most useful.

Electron beam (e-beam) lithography is a powerful established method used to pattern nanometer scale structures such as single electron transistors, single molecule detectors, and nanoelectromechanical devices14. E-beam lithography typically involves applying, chemically transforming, and chemically dissolving a polymer resist5. The desire to decrease device size, to enhance the role of quantum mechanical device characteristics, and to pattern increasingly complex substrates requires new lithographic approaches to define nanometer scale structures. Here we demonstrate a new approach to nanoscale e-beam patterning based on the condensation and beam stimulated sublimation of water ice. Electron and ion beam stimulated sublimation of solid condensed gasses has been studied in the context of cryo-electron microscopy6, plasma confinement7, and astrophysics8, but not, to our knowledge, for high resolution patterning. We have discovered that frozen water, a most common of substances, can be used to generate nanoscale patterns of metals and potentially very useful chemical transformations on a substrate surface.

We chose this approach because it is known that electrons and very high energy ions create electronic excitations, molecular dissociations and ionization in ice, but the consequences of these effects on nanoscale patterning have not been explored. In insulating surface layers with low sublimation energies, “Coulomb explosions” can lead to large localized agitation and ejection of surface atoms8. In addition, dissociation products of condensed molecular gasses can be ejected after diffusion to the surface from deeper regions within the beam exposure volume6,9. These processes, as well as others, have been discussed7,10,11, but we feel it fair to conclude that the fundamental physics responsible for these observations is still obscure.

In our experiments, we deposit very thin, conformal layers of ice from water vapor onto cryogenically cooled silicon substrates in the chamber of a combined scanning electron microscope (SEM) and focused ion beam (FIB) apparatus (FEI Co., Hillsboro, Oregon, USA). Subsequent exposure of the ice surface to focused energetic electron or gallium ion beams stimulates local removal of ice and ultimately exposes the underlying silicon substrate in whatever patterns the beams are programmed to produce. Afterwards, using lower beam doses, the exposed regions can be inspected nondestructively with the SEM or FIB operating in its standard scanning imaging mode. With the sample still cold, the exposed patterned regions of the substrate can then be metallized or modified by other techniques (ion implantation, reactive ion etching, sputter deposition, etc.) in the same vacuum chamber or an adjoining one. Finally, the ice can be removed either by in situ sublimation (eliminating liquid surface tension effects) or by rinsing.

Typically, we deposit ice at a rate of ~1 nm/sec using a leak valve controlled water vapor flow that is directed onto the cooled sample (Figure 1). Uniform ice coverage over ~1 mm2 is achieved with a single needle (inner diameter ~100 μm) water vapor source held ~5 mm above the sample surface. At operating sample temperatures of 128 K the deposited ice is amorphous and sublimes at a rate of only ~0.3 monolayers/hour with a sublimation energy of 0.45 eV12,13. In practice, no appreciable ice sublimation is observed over several hours when working at 128 K, even with ice films <10 nm thick. The ice is promptly removed by sublimation when the sample temperature is raised to ~180 K. We have successfully patterned ice with a focused ion beam (30 keV Ga+, 10 pA, diameter ~ 10 nm) or a focused electron beam (1–30 keV, 30–150 pA, diameter ~ 5 nm). All data presented here were patterned with an electron beam. After patterning, samples were transferred while cold to a connected chamber with a chromium sputter deposition source.

Figure 1
Schematic of the apparatus. Within the main vacuum chamber at <1X10−6 Torr, water from the vapor phase is directly condensed onto cryogenically cooled substrates, typically at 128 K. The water vapor pressure is controlled by leak valves ...

An in situ SEM image of a 75 nm thick layer of ice on a silicon substrate at 128 K immediately after 5 keV e-beam patterning is shown (Figure 2a). The e-beam dose used to create each of the 500 nm square patterns was increased from left to right. The sharp rise in contrast between the fourth and fifth square indicates that a dose above 8.8 ×105 μC/cm2 is required to remove the entire ice layer and expose the underlying silicon surface. An atomic force microscope (AFM) line scan and a SEM image of the same sample are shown (Figure 2b) after sputter deposition of 40 nm of Cr, removal from the vacuum chamber, and a rinse in isopropanol14 to remove the remaining ice and its overlayer. Only the Cr that comes in contact with the patterned ice-free regions of the substrate remains after complete ice removal. No aggressive methods such as ultrasonication or mechanical scrubbing were needed to assist the liftoff. Figure 2b confirms that 5 keV e-beam doses greater than 8.8 × 105 μC/cm2 are required to remove the 75 nm ice layer thickness and assure subsequent pattern transfer to the deposited Cr. This critical dose for water ice resist is roughly 3 orders of magnitude larger than that required for a typical exposure of the polymer resist polymethyl methacrylate (PMMA). Assuming the amorphous ice layer has a density13 0.91 gm/cm3, we calculate the sputter yield (i.e. H20 molecules ejected per incident electron) S = 0.03 for 5 keV electrons. S decreases by over an order of magnitude as the beam energy increases from 1 to 30 keV but it does not vary significantly with temperature between 128 and 158 K. In contrast to the relatively low electron sputter yield, we found S ≈ 12 for 30 keV gallium ions in an FIB15!

Figure 2
Patterning a silicon surface. (a) In situ cryogenic (sample T=128 K) SEM image of 500 nm × 500 nm squares formed in a 75 nm thick ice layer by a 5 keV, 112 pA electron beam. The electron beam patterning dose was increased from 2.2 (left side, ...

Figure 2c and 2d present results of varying the line dose under the same conditions as previously described for the area dose study, except with a thinner (20 nm) ice layer. Figure 2c is an in situ cryogenic SEM image of micron long single pixel lines of varying doses. Here the line dose is increased from 1.1 μC/cm (left) to 5.6 μC/cm (right). After 6 nm Cr deposition, an SEM image and an AFM linescan (Figure 2d) yield the critical line dose under these conditions of ~2.2 μC/cm. We note the patterned metal dot at the bottom end of each line allows for easy identification of underexposed regions.

The AFM scans in Figures 2b and 2d also reveal the growth of a layer of material ~ 1 nm thick in the region where the beam hit the ice even with doses too low to eject all of the ice down to the underlying substrate. We attribute this thin layer to a beam induced surface chemical transformation that involves both the ice layer and the underlying silicon substrate because (a) the effect is observed with no metallization step; (b) the effect is specific to silicon, i.e. we see no similar growth of material on SiO2 or Si3N4 substrates; (c) the material grows only to a self-terminating thickness above the silicon (111) surface; and (d) the final (self terminated) thickness of the material scales with the amount of ice initially deposited on the silicon surface. When 5 nm of ice was deposited, subsequent e-beam exposure produced a ~0.5 nm high chemical transformation; a 75 nm ice thickness produced a 3.0 nm high structure. Further e-beam exposure did not increase the material’s height. In contrast to contamination lithography16, observations (b) and (c) imply that we can rule out deposition of unspecified (e.g. hydrocarbon) vacuum contaminants in the chamber. Since the electron range in our experiments is larger than the ice layer thickness, observation (d) suggests that the stimulated surface chemistry required for the material growth persists only as long as the ice layer that is being removed still exists. Considering the possible species present (H20, its atomic or molecular fragments, silicon and hot electrons) on or near the silicon–ice interface when the beam impacts, we postulate that the thin layer of material is likely silicon oxide. This is further supported by the observation that the thin layers of material can be removed with hydrofluoric acid, leaving behind a depression in the silicon surface (Figure 4f). The electronic properties of the thin layer of material are currently under investigation.

Figure 4
AFM analysis of the local electron beam induced chemical transformation of a silicon substrate. All data except panel (d) are tapping mode AFM images with associated AFM linescans. All results in (a–c) and (e) were produced with beam doses too ...

The spatial resolution that we have achieved for ice patterned Cr lines on silicon using ~20 nm thick ice layers is shown in Figure 3 with beam exposures (a) at 30 keV and (b) at 20 keV. We note that this apparent difference in line width is potentially related to aperture alignment, astigmatism adjustment, and focus optimization steps necessary after changing the beam acceleration voltage. The narrowest dimensions of the material tentatively identified as silicon oxide (Figure 4) are, remarkably, much smaller than those obtained for Cr (Figure 3). A study of the lateral dimensions of these putative oxide structures as a function of beam energy illustrates the evident trend towards higher resolution with increasing electron beam energy. At 10 keV (Figure 4a), the structure’s full width at half maximum (FWHM) measures 18 nm. The FWHM is reduced to 14 nm at 25 keV and to 10 nm at 30 keV (Figures 4b and 4c). The measurements listed on Figure 4 are values obtained without subtracting the effective diameter of the AFM probe tip. When we account for the AFM tip effective diameter of ~5 nm (Figure 4d), the true width of these lines are of order 13 nm, 9 nm and 5 nm for the 10 keV, 25 keV and 30 keV exposures respectively. Figure 4e shows a 2 μm × 1.5 μm AFM image of two lines defined by a 30 keV electron beam. Here we see that the 3.0 μC/cm line measures ~1.2 nm in height above the Si, while the 1.5 μC/cm line on the left is only ~0.8 nm high.

Figure 3
SEM analysis of sub-30 nm wide Cr lines on a bulk silicon substrate. Both structures were defined on 20 nm ice resist layers and subsequently metallized with 8 nm Cr. (a) 30 keV line, 4.5 μC/cm dose. (b) 20 keV line, 4.4 μC/cm dose.

Although the “record” line widths of e-beam exposed PMMA are in the single digit nanometer range, such results were achieved with specialized higher energy beams (~100 keV) and attentive ultrasonication during resist development1719. A more typical minimum line width achieved with e-beam exposed PMMA on bulk silicon substrates with commercial e-beam lithography tools is of order 30 nm20. To reliably achieve thinner line widths, angled depositions of metals or other materials on exposed resist are common. Our demonstrated sub-20 nm metal lines are achievable with an ice resist on bulk silicon with e-beam energies <30 keV and without directional deposition of metal. We speculate that proximity effects21,22 are minimized with ice resist because the excitation of ice by backscattered electron exposure away from the point of beam incidence is not an additive process. Instead, excited ice that is not ejected from the surface can relax to its initial unexposed state.

Patterning with ices of any condensed gas is a straightforward and practical process. Ice resist does not require spinning or baking. All processing and patterning steps can occur in a single evacuated chamber and be monitored at high resolution. The final removal of unexposed resist leaves minimal residues. Environmentally harmful solvents are not required and complete dry removal of the ice layer can be performed by in situ sublimation. Although in our current apparatus dry resist removal by sublimation after Cr metallization causes large loose Cr flakes to settle on the substrate, an inverted configuration may avoid this problem. Completely dry resist processing will be particularly useful for preparing delicate micro and nano electromechanical devices on a variety of substrates, especially those exhibiting complex three dimensional geometries.

It will be important to discover the resolution limits and to minimize the critical dose requirements by testing a wide range of beam energies, beam diameters, and other condensed gases. To avoid induced interface chemistry, rare gasses are clearly called for. Alternatively, to yield desired chemical transformations, other gasses can be selected to react with the substrate and produce precisely defined thickness tunnel junctions and gate insulators needed for nanoscale electronic devices. The gate insulator needed for nanoscale field effect transistors is an important example for which the material grown with water may well be suited. We anticipate that the rich and varied chemical, electronic and mass transport properties of energetic beam stimulated solid condensed gasses will provide many opportunities for discovery and innovation in connection with nanoscale patterning.


This work was supported by awards from the NSF (#DMR-0073590), DOE (#DE-FG02-01ER45922), NIH (#RO1HG02338), and Agilent Technologies. We thank the Harvard Center for Imaging and Mesoscale Structures for facilities support as well as Damon Farmer and Trygve Ristroph for assistance and discussions.


1. Pashkin Y, Nakamura Y, Tsai J. Appl Phys Lett. 2000;76:2256.
2. Liu K, Avouris P, Bucchignano J, Martel R, Sun S, Michl J. Appl Phys Lett. 2002;80:865.
3. Craighead H. Science. 2000;290:1532. [PubMed]
4. Ito T, Okazaki S. Nature. 2000;406:1027. [PubMed]
5. Rai-Choudhury, P., Handbook of Microlithography, Mircomachining, and Microfabrication. ed.; SPIE Optical Engineering Press: London, UK, 1997.
6. Talmon Y, Davis HT, Scriven LE, Thomas LE. J Mircoscopy. 1979;117:321–332.
7. Erents SK, McCraken GM. J Appl Phys. 1973;44:3139–3145.
8. Brown WL, Lanzerotti LJ, Poate JM, Augustyniak WM. Phys Rev Lett. 1978;40:1027–1030.
9. Brown W, Augustyniak W, Lanzerotti L. Phys Rev Lett. 1980;45:1632.
10. Ellegaard O, Schou J, Sorensen H, Borgesen P. Surf Sci. 1986;167:474–492.
11. Watson CC, Tombrello TA. Rad Eff. 1985;89:263–283.
12. Sack N, Baragiola R. Phys Rev B. 1993;48:9973. [PubMed]
13. Echlin, P., Low-Temperature Microscopy and Analysis ed.; Plenum Press: New York, 1992.
14. Liquid water also works for this purpose. Techniques to remove the Cr overlayer in situ have not been explored.
15. Conventional organic polymer resists are 10–100 times more sensitive to ion irradiation than to electron exposure (Wolf, S. and Tauber, R., Silicon Processing for the VLSI Era; Lattice Press: Sunset Beach, California, 1986). Based on this measured ion sputter yield we estimate the effectiveness of ice resist as an ion implantation mask: ~ 4 nm of ice is required per monolayer of implanted ions (30 keV Ga).
16. Broers A, Hoole A, Ryan J. Microelectron Eng. 1996;32:131.
17. Vieu C, Carcenac F, Pepin A, Chen Y, Mejias M, Lebib A, Manin-Ferlazzo L, Couraud L, Launois H. Appl Surf Sci. 2000;164:111.
18. Chen W, Ahmed H. Appl Phys Lett. 1993;62:1499.
19. Craighead HG, Howard RE, Jackel LD, Mankiewich PM. Appl Phys Lett. 1983;42:38.
20. Raith 150 Ultra high resolution E-beam lithography and metrology tool, Product data sheet: minimum feature size = 30 nm. Raith GmbH, Dortmund (Germany).
21. Chang THP. J Vac Sci Technol. 1975;12:1271.
22. Chen W, Ahmed H. Adv Imag Elect Phys. 1998;102:87.