Silicon nanowire atomic force microscopy probes for high aspect ratiogeometriesBrian A. Bryce, B. Robert Ilic, Mark C. Reuter, and Sandip Tiwari

Citation: Appl. Phys. Lett. 100, 213106 (2012); doi: 10.1063/1.4720406 View online: http://dx.doi.org/10.1063/1.4720406 View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v100/i21 Published by the American Institute of Physics.

Related ArticlesRole of polymer matrix in large enhancement of dielectric constant in polymer-metal composites Appl. Phys. Lett. 99, 042905 (2011) Simulation and experimental study on compositional evolution of Li-Co in LiCoO2 thin films during sputterdeposition J. Appl. Phys. 109, 114910 (2011) Nano-engineered defect structures in Ce- and Ho-doped metal-organic chemical vapor deposited YBa2Cu3O6+films: Correlation of structure and chemistry with flux pinning performance J. Appl. Phys. 109, 113923 (2011) Hydrogenated amorphous silicon oxide containing a microcrystalline silicon phase and usage as an intermediatereflector in thin-film silicon solar cells J. Appl. Phys. 109, 113109 (2011) Preface to Special Topic: Plenary and Invited Papers from the 30th International Conference on the Physics ofSemiconductors, Seoul, South Korea, 2010 J. Appl. Phys. 109, 102301 (2011)

Additional information on Appl. Phys. Lett.Journal Homepage: http://apl.aip.org/ Journal Information: http://apl.aip.org/about/about_the_journal Top downloads: http://apl.aip.org/features/most_downloaded Information for Authors: http://apl.aip.org/authors

Downloaded 25 May 2012 to 140.254.87.101. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

Silicon nanowire atomic force microscopy probes for high aspect ratiogeometries

Brian A. Bryce,1 B. Robert Ilic,1 Mark C. Reuter,2 and Sandip Tiwari31School of Applied and Engineering Physics, Cornell University, Ithaca, New York 14853, USA2IBM Research Division, T. J. Watson Research Center, Yorktown Heights, New York 10598, USA3School of Electrical and Computer Engineering, Cornell University, Ithaca, New York 14853, USA

(Received 17 February 2012; accepted 7 May 2012; published online 22 May 2012)

Using site controlled growth of single vapor-liquid-solid silicon nanowires high aspect ratio atomic

force microscope probes are fabricated on a wafer scale. Nanowire probe aspect ratios as high as

90:1 are demonstrated. Probe performance and limitations are explored by imaging high aspect

ratio etched silicon structures using atomic force microscopy. Silicon nanowire probes are an ideal

platform for non-destructive topographic imaging of high aspect ratio features. VC 2012 AmericanInstitute of Physics. [http://dx.doi.org/10.1063/1.4720406]

In the 25 years since its introduction1 atomic force

microcopy (AFM) has gone from cutting edge research to an

indispensible and routine metrological tool. AFMs strength

remains in resolving small height variations on a nearly flat

surface. With the rise of nanofabrication it increasingly

plays, at the nanoscale, a role analogous to that of the stylus

profilometer at the micron scalenon-destructive probing of

lithographically defined features.

One of the defining traits of the nanoscale is an intrinsi-

cally large surface to volume ratio and often with its high as-

pect ratios. Conventional mass produced AFM probes such

as the tapping-mode etched silicon probe (TESP)2 are not

suitable for imaging high aspect ratio structures. To cope

with this difficulty AFM probes with higher aspect ratios

have been introduced, specifically focused ion beam (FIB)

milled tips and carbon nanotube/fiber tips.3,4 These exotic

AFM probes have clear advantages over their conventional

counterparts particularly in probing high aspect ratio fea-

tures. However wide spread use of such tips is limited by the

complex methods needed for their fabrication.

Bottom up high aspect ratio structures are not limited to

carbon nanotubes and fibers. Bottom up nanowires can be

grown from a variety of materials including silicon.5 Silicon

is a particularly attractive material to work with because of

its intrinsic chemical properties and the availability of a veri-

table arsenal of industrial tools designed to manipulate it.

This suggests the use of silicon nanowires as AFM probes.

Silicon nanowires can be grown using a variety of meth-

ods, the most popular of which is the vapor-liquid-solid

(VLS) method.6 In this method a metal catalyst is used along

with a vapor precursor such as SiH4 or SiCl4. In the VLS pro-

cess the catalyst particle remains at the tip of the wire as the

wire grows away from the growth substrate. Many different

metals may be used with the VLS or related vapor-solid-solid

(VSS) method. However, gold remains the most popular due

to its simple phase diagram and ease of handling.5

Silicon nanowire AFM tips have been demonstrated pre-

viously7 but not with single wire to catalyst site control. The

key to mass producing silicon nanowire AFM tips is the abil-

ity to grow a single isolated nanowire at a desired location

with a controlled orientation.

The growth direction of an epitaxial nanowire is thermo-

dynamically controlled on a statistical basis in the directions

which minimize the total interfacial free energy between the

metal catalyst and wire.8 This free energy has edge and sur-

face contributions. As the wire becomes larger the surface

interfacial free energy dominates over edge contributions. In

the Au-Si system this energy is minimized for a crystal

growing in the h111i direction. Thus, for wires of largeenough diameter, growth on a h111i surface will yield verti-cally oriented wires. This feature of VLS growth allows for

control of the wires orientation.

Nanowires can readily be grown at lithographically

defined catalyst sites in periodic arrays.9 Reducing the array

size to unity increases the edge site count of the array to its

maximal value. For gold catalyzed nanowires this is impor-

tant. When heated, gold in contact with a clean silicon sur-

face defuses both into the bulk crystal and out onto the

surface. For gold on a silicon h111i surface, the equilibriumstate is to cover the surface with a monolayer.10 With a sin-

gle growth site, surface diffusion is more important than in

the bulk array case due to the greater ratio of available sur-

face area to gold volume. Because of these effects, annealing

under vacuum (

utilizing a silicon-on-insulator (SOI) wafer with a 5lmh111i device layer and a 2 lm buried oxide (BOX) layer. Nolithographic features smaller than 1 lm are used. The fabri-cation process11 results in a suspended cantilever with Au

dot in a SiO2 well. The Au dot is nominally 1 lm in diameterand 50100 nm thick. Before wire growth the wafer is dipped

in hydrofluoric acid (HF) to remove the oxide layer leaving a

clean hydrophobic silicon surface. The wafer is then loaded

into a UHV CVD reactor. The wafer is annealed under UHV

conditions at 600 C for 5 min. With the reactor remaining at600 C, SiH4 flow begins pressurizing the reactor to a pres-sure of 250 mTorr, and growth is carried out. The growth

time determines the wire length. The wire diameter is deter-

mined by the volume of Au in the growth site minus any

losses to the surface/bulk diffusion during the annealing and

nucleation times. The result of these processes can be seen in

Fig. 1(a). Notable is the clear circle around the wire site

which is due to Au that has diffused onto the surface from

the growth site, thus changing the CVD morphology.

The probe yield after growth can readily be determined

via inspection in an optical microscope. By using the depth

of focus of the microscope, correctly oriented wires appear

simply as a dot, incorrectly oriented wires appear as a rod,

and failed growth sites appear blank. Using the parameters

described the site yield for the process is nominally 20%. It

is likely this can be improved by optimizing the Au thick-

ness, anneal profile, and the reactor parameters further. Each

probe body has 6 cantilevers defined (each with one site),

resulting in a high probe yield of nominally 74% with the

process described.

The nanowire AFM probes may be used with the Au left

in place. The Au tip offers a useful biological sensing plat-

form for fictionalization using thiol chemistry. Localized thi-

olates residing at the apex of the nanowire tip would allow

interrogation of molecular binding forces and binding events

in ambient and fluid environments.

Alternatively the Au can be removed and the wire

chemically thinned. Removal of Au is accomplished in a KI/

I2 based Au etch solution. After careful washing in HCl and

deionized water to remove residual iodine, the silicon can be

oxidized to a desired diameter. The oxide layer can then be

removed in a vapor HF chamber. This process can mass pro-

duce probes with very large aspect ratios (Fig. 1(b)). The

maximum aspect ratio attainable is determined by the ratio

of the vertical and sidewall growth rates in the VLS process.

This is determined by the catalyst material and reactor

parameters.

The diameter and length of the wire dictate the stiffness

of the probe. Assuming a cylindrical shape the maximum

force that may be applied to a sample via a nanowire probe

is given by the Euler buckling force FEuler p2EI=L2, whereE is Youngs modulus, I is the stress moment I p r4

4, and

L is the length of the nanowire. This force can be dramati-cally tuned via the thinning process. Taking Youngs modu-

lus to be 185 GPa for a 8 lm long wire FEuler ranges from9 nN at r 25 nm to 11 lN at r 150 nm.

To examine these effects and to assess the general per-

formance of the nanowire AFM probes we created a series of

trenches in a h100i silicon wafer using 248 nm DUV stepperlithography, a Cl based reactive ion etch (RIE) and an oxide

hard mask. The trenches created are nominally 2 lm deep.These trenches were scanned using a Bruker Icon AFM with

our experimental AFM probes. In this AFM the mounting

angle of the probe is nominally 13. Because the nanowiresare grown normal to the surface this results in the wire being

tilted 13 to the surface for scans without rotation. Perform-ance of a single AFM tip as grown and thinned is shown in

Fig. 2. The fabricated tip diameter combined with the 13 tiltlimit the trench width that can be probed to 900 nm, with the

tip just reaching the bottom of the trench before striking the

second sidewall. Thinning the wire via oxidation greatly

improves the imaging ability of the tip, allowing both the

probing of narrower trenches and the imaging of the flat bot-

tom of the trench. Thinning the wire still further decreases

the stiffness / r4 of the probe to the point that imagingbecomes difficult. The result is many ringing artifacts while

imaging (Fig. 2(c)). Thinning of a nanowire to this extreme

or beyond may still be desirable if X-Y resolution is impor-

tant, because the diameter of the nanowire tip is the limiting

factor of lateral resolution in images without artifacts.

The performance of these nanowire based probes greatly

exceeds traditional AFM probes such as the TESP tip for

imaging high aspect ratio features and approaches that of

FIG. 1. (a) A nanowire AFM tip as grown (scale bar: 10lm). The circularring around the wire is caused by Au migration onto the bulk surface before

the wire is fully catalyzed. This Au changes the morphology of the CVD

growth. Inset (scale bar: 1lm): detail of the nanowire tip as grown showingthe faceted single crystal wire and nearly hemispherical Au catalyst tip. (b)

An example of a nanowire AFM tip thinned via silicon oxidation (scale bar:

10lm). The aspect ratio of the tip is 90:1. Inset right (scale bar: 1 lm): detailof the wire after thinning, the slight taper in the wire is carried over from the

taper during wire synthesis. Inset top (scale bar: 100lm): shows the full can-tilever AFM probe, the cantilever is 5lm thick, 125lm long, and 30lmwide.

213106-2 Bryce et al. Appl. Phys. Lett. 100, 213106 (2012)

Downloaded 25 May 2012 to 140.254.87.101. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

commercially available FIB milled AFM tips. Direct com-

parison of AFM data, taken with the nanowire probes, with

SEM cross sections, of the same trenches, shows that the

AFM data provides a good representation of the trench struc-

ture (Fig. 3). When scanned at 90 to eliminate the tips tilt,the only notable differences between SEM and AFM data

are a narrowing of the trench due to the diameter of the tip

used and the tips failure to probe the very narrow ion chan-

neling indentions at the well edge again due to the tips

FIG. 2. Single nanowire AFM tip imaging a 900 nm wide, 1.95lm deep Cletched trench during repeated thinning of the tip. The long axis of the canti-

lever was orthogonal to the long axis of the trenches for the scans. The angle

on the negative X-axis is 13 and is caused by the mounting angle of theAFM probe. The depth traces plotted are the mean of column values in 2D

images shown. The insets show the nanowire probes at the time of scanning

(scale: 10 lm). The minimum wire diameters are (a) 464, (b) 160, and (c)40 nm. The thinning was performed via oxidation in a dry oxygen atmos-

phere followed by removal of the oxide in a vapor HF chamber. The as

grown diameter of the wire tip (a) when combined with the 13 tilt only justallows the wire to reach the bottom of the trench structure. Once the tip is

thinned (b) the bottom of the trench is clearly imaged. Thinning the tip fur-

ther (c) introduces artifacts in the image which cannot be eliminated by tun-

ing the feedback settings of the Bruker Icon AFM used. These artifacts

effectively reduce both lateral and vertical resolution.

FIG. 3. (a) The performance of the nanowire tip is compared to a conventional

TESP tip, tilt-corrected FIB milled tip, and an SEM cross section of the same

1.95lm deep, 900 nm wide trench. The symbols are plotted every 40th point forthe nanowire and TESP AFM data and every 20th point for the SEM data and

FIB milled AFM data. The nanowire tip significantly outperforms the TESP tip

and agrees very well with the SEM data. The nanowire tip is comparable to the

FIB milled AFM tip. The 90 image is scanned with the short axis of the cantile-ver orthogonal to the trenchs long axis and the 0 image is scanned with thelong axis of the cantilever orthogonal to the trenchs long axis. The depth traces

plotted are averaged by the same procedure as described for Fig. 2. The

encroachment seen at the left and right side of the 0 scan from the SEM dataindicate the tip diameter at the time of imaging. The nanowire diameter at the

time of imaging in (a) was 244 nm; (c) the diameter was thinned to 84 nm. The

commercial FIB milled tip had an 800 nm long chisel shaped tip terminated in a

nominally 25 nm point attached to a nominally 5lm long 200500 nm wideshaft. This finer point allowed the FIB milled tip to probe channel on the nega-

tive X-axis but not the positive X-axis as the tip chisel shape is asymmetrical.

On the positive X-axis the shaft diameter causes encroachment similar to the

nanowire probe. The SEM data shown (a) is calculated from edge of the cross

section shown in (b). (c) The probe as used during the scans shown in (a). It is

possible to reach the bottom of a 2.05lm deep, 300 nm wide trench by thinningthe nanowire (d), although it does contain ringing artifacts that cannot be sup-

pressed with the feedback settings on the Bruker Icon AFM used. The trenches

being imaged in (d) are shown in (e). The probe at the time of imaging (d) is

shown in (f). The scale bars for (b), (c), (e), and (f) are 2, 5, 1, and 5lm,respectively.

213106-3 Bryce et al. Appl. Phys. Lett. 100, 213106 (2012)

Downloaded 25 May 2012 to 140.254.87.101. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

diameter. The performance with rotation approaches that of

a commercially available tilt-corrected FIB milled tip. This

is due to the similar geometry of the two tips. Both the FIB

milled tip and the nanowire tip used in Fig. 3(a) consist of a

wire shaft in excess of 200 nm. This accounts for the shorten-

ing of the trench width on the positive X-axis in both scans.

The FIB milled tip is able to track the sidewall and measure

the ion channeling indention on the negative X-axis due to

its wedge sharpened end which the nanowire tip lacks.

The nanowire probe used in Fig. 3 was initially 244 nm

in diameter. With this diameter trenches as narrow as 500 nm

in width could be probed to their full 2lm depth.11 By thin-ning the probe to 84 nm in diameter we probed features as

narrow as 300 nm of depth 2.05 lm (Fig. 3). Although theseimages contained artifacts they remain useful in cases where

well depth is the primary concern and non-destructive prob-

ing is required. With an appropriately matched feedback sys-

tem or a less stiff cantilever it may be possible to reduce or

eliminate these stiffness related artifacts. Shortening the

nanowire tip to the minimum length needed to probe a

desired feature would likely also help remove such effects. It

is important to note that the minimum length of the wire

depends not only on the feature depth that is to be probed but

also on the offset from the front edge of the cantilever and

the mounting angle of the probe in the AFM. This offset is

required due to alignment tolerances between lithographic

layers; its minimum value is set by the tolerances of litho-

graphic tools used, and the acceptable reduction in yield due

to alignment variation across the wafer. We choose a con-

servative value to ensure layer to layer alignment was not an

issue, but in an optimized process the offset could be

reduced.

During imaging we noticed no appreciable degradation

in the tip performance. This is reasonable given the flat na-

ture of the tip, nearly uniform diameter of the wire, and rela-

tively low stiffness of the structure. The low stiffness of the

probes requires a relatively low tip velocity (