The influence of the substrate thermal conductivity on scanningthermochemical lithographyMarten Tolk, Oliver Fenwick, Sadi Ahmad, and Franco Cacialli Citation: J. Appl. Phys. 111, 124317 (2012); doi: 10.1063/1.4729809 View online: http://dx.doi.org/10.1063/1.4729809 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v111/i12 Published by the American Institute of Physics. Related ArticlesFacile large-area photolithography of periodic sub-micron structures using a self-formed polymer mask Appl. Phys. Lett. 100, 233503 (2012) Controlled addressing of quantum dots by nanowire plasmons Appl. Phys. Lett. 100, 231102 (2012) Effects of tip-substrate gap, deposition temperature, holding time, and pull-off velocity on dip-pen lithographyinvestigated using molecular dynamics simulation J. Appl. Phys. 111, 103521 (2012) Unbiased line width roughness measurements with critical dimension scanning electron microscopy and criticaldimension atomic force microscopy J. Appl. Phys. 111, 084318 (2012) Metallic nanomesh electrodes with controllable optical properties for organic solar cells Appl. Phys. Lett. 100, 143109 (2012) Additional information on J. Appl. Phys.Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors

The influence of the substrate thermal conductivity on scanningthermochemical lithography

Marten Tolk, Oliver Fenwick, Sadi Ahmad, and Franco Caciallia)

Department of Physics and Astronomy, and London Centre for Nanotechnology, University College London,London WC1E 6BT, United Kingdom

(Received 18 January 2012; accepted 19 May 2012; published online 22 June 2012)

We present a joint experimental and computational study of the role of the substrate thermal

conductivity on scanning thermochemical lithography (SThL) of thin organic films. We aim this

study at lithography of the luminescent conjugated polymer poly(p-phenylene vinylene) (PPV)

from its soluble precursor poly(p-xylene tetrahydrothiophenium chloride) (PXT), but our results

provide relevant insights into the SThL of thermosensitive polymers in general, and into a wide

range of nanoscale thermal and thermochemical processes in thin films. As high thermal

conductivity substrates we used gold films on silicon, and indium-tin oxide (ITO) films on glass,

successfully patterning PPV on both substrates. We find that a higher probe temperature (>300 �Cinstead of �250 �C) is necessary for lithography of PXT films on ITO compared to those on fused

silica (for the same scanning speed and comparable precursor thickness). Surprisingly, however,

our experiments show that minimum feature sizes are nearly independent of the underlying

substrate. While a lateral resolution (full width at half maximum, FWHM) of 37 nm was achieved

previously on fused silica for a 40 nm thick PXT film, we obtain here a FWHM of 36 nm for a

35 nm thick PXT layer on ITO. We compare our experiments with finite element simulations and

gain further insight into the possibilities of thermochemical lithography, the necessary minimum

probe temperature and the highest attainable resolutions. The model shows that for high thermal

conductivity substrates there should be a region of unconverted polymer near the polymer-substrate

interface. Our experiments demonstrate that patterned features are able to adhere to the substrate

despite this unconverted layer, thus allowing SThL to work on very high thermal conductivity

substrates such as gold. Our model builds on this experimental finding and accounts for the

experimental lack of dependence of lateral size with substrate conductivity, i.e. it predicts that

the minimum feature size increases only slightly for increasing thermal conductivities of the

substrates. VC 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4729809]

I. INTRODUCTION

Scanning thermochemical lithography (SThL)1–5 is a

versatile nanopatterning technique where a hot probe is

scanned across a surface to induce a local chemical reaction

in a thin film to generate the desired pattern.

In general, nanolithography is of crucial importance in

today’s science and industry, the most prominent application

of nanopatterning techniques being in electronics, for

fabrication of nm-sized transistors. In addition, novel and

emerging areas of application include the development of

nanosensors and of nanoelectromechanical systems (NEMS).

To fabricate devices incorporating components with dimen-

sions of only few tens of nm a variety of techniques are

available, including “conventional” ones such as optical

far-field,6,7 electron-beam or focused-ion beam (FIB) lithog-

raphy,8,9 and less conventional techniques such as scanning

near-field optical,10–13 thermal,14 thermomechanical,15,16

and thermochemical lithography.1–5

SThL is particularly appealing because it does not suffer

from the resolution limitations imposed by the Abbe diffrac-

tion limit in far-field photolithography, or from the irradiation

damage caused by e-beam lithography.17,18 High resolution

SThL probes are readily machined in silicon, and are more

robust than probes used in scanning near-field optical lithog-

raphy (SNOL).11,13 Although all scanning probe techniques

are typically “serial” in nature, their throughput can be

“upscaled” by using an array of probes, as demonstrated by

IBM.15 SThL is remarkably versatile and has already been

used to crosslink commercially available photoresists,2 to

convert a tetrahydropyranyl analogue from hydrophilic to

hydrophobic,3 to reduce graphene oxide,4 and to create three-

dimensional structures within a thin layer of a molecular

glass,5 with resolutions down to 15 nm (Ref. 5) and a through-

put5 of 5� 104 lm2/h or scanning speeds up to 1.4 mm/s.3

Previously, we have used SThL to prepare nanostruc-

tures of the prototypical electroluminescent polymer,

poly(p-phenylene vinylene), PPV, from its soluble precursor,

poly(p-xylene tetrahydrothiophenium chloride), PXT, depos-

ited on fused silica substrates. In this process, we scan our

heated probe, a Wollaston wire, across the polymer film

selectively converting the precursor PXT to insoluble PPV.

A subsequent rinsing step removes the unconverted PXT

leaving isolated lines of polymer on the surface. A final ther-

mal annealing ensures complete conversion of the remaining

material to PPV. Surprisingly, we could achieve minimuma)E-mail: f.cacialli@ucl.ac.uk.

0021-8979/2012/111(12)/124317/8/$30.00 VC 2012 American Institute of Physics111, 124317-1

JOURNAL OF APPLIED PHYSICS 111, 124317 (2012)

resolutions of better than 28 nm,1 even though the radius of

curvature of our probes was several orders of magnitude big-

ger (�2.5 lm). This was due to the combined effects of a

careful control of scanning speed and probe contact pressure,

to the dissolution of partially converted precursor during de-

velopment (rinsing), and also to the shrinking of the polymer

precursor in its conversion to PPV.

The question arises as to whether a significant change in

the substrate thermal conductivity would influence the

details of the conversion process and therefore the minimum

resolution attainable.

To a first order approximation, if the tip temperature

remained constant we would expect a decrease of the feature

size for an increase of the substrate thermal conductivity kbecause the increased rate of heat removal from the bottom

of the film would provide a better “vertical” confinement of

the thermal fields and avoid their lateral spread. However, a

higher thermal conductivity of the substrates would also

impose lower temperatures near the film-substrate interface,

and thus require a higher tip temperature to ensure film con-

version and anchoring at such interface, hence resulting in a

lateral heat spread and deteriorated resolution. Making pre-

dictions on which of these two antagonistic effects will dom-

inate and by which factor the lateral resolution changes is far

from trivial, due to both the complexity of the mechanical

contact between the tip and the sample, and also to an

incomplete understanding of the effects of development and

post-baking on the feature size.

Here, we gain further insight into these issues by

replacing the fused silica substrate with indium-tin oxide

(ITO)-coated glass, one of the most common transparent

electrodes/substrates in organic electronics. The thermal

conductivity of ITO ranges from 3.1 W m�1 K�1 for very

thin films19 to bulk conductivities20 of 14 W m�1 K�1,

which is significantly higher than that of fused silica21

(k¼ 1.4 W m�1 K�1). Surprisingly, we find that the lateral

resolution, here defined as the full width at half maximum

(FWHM) of the lithographed features, is essentially inde-

pendent of the thermal conductivity of the layer underneath

the precursor (which we will also term the “interlayer” from

now on). We corroborate these observations with experi-

ments on gold-coated Si wafers (kgold¼ 317 W m�1 K�1).22

We then model the conversion process by using the simu-

lated temperature profile and show how the final resolution

is affected by the competition of the two opposing effects

discussed above.

II. METHODS

A. Experimental methods

Our thermochemical lithography setup has been

described in detail in Ref. 1 and utilizes as the hot probe a

so-called Wollaston wire23,24 (Bruker) of the type employed

in micro-thermal analysis.25,26 It is a 75 lm diameter silver

wire which is etched, exposing a�5 lm diameter platinum-

rhodium (9:1) core that is bent around to from a probe (sche-

matically illustrated in Fig. 1(a)) that can be mounted onto

the head of an atomic force microscope, AFM.

The probe was scanned in constant-force contact mode

across the surface of a�35 nm thick PXT film which was

spin-coated from a water solution onto an oxygen plasma-

treated27 ITO-coated glass substrate. The chemical structure

of PXT (purchased as a 0.25 wt. % water solution from

Aldrich) and PPV is illustrated in Fig. 1(b). The ITO film is

120–160 nm thick as specified by the supplier. Various tem-

peratures, writing speeds, and line densities were used while

the force and feedback parameters were constant throughout

the experiments. We have previously noted1 that high resolu-

tions with this technique rely on avoiding mechanical defor-

mation of the polymer film surface by the probe, and

therefore use the minimum necessary contact force to keep

the probe in contact with the surface. This force was found

to be� 2 lN. After patterning, the samples were rinsed in

methanol for 10 s to remove the unconverted precursor and

annealed in a vacuum oven at 200 �C and <10�3 mbar for

2 h to ensure complete conversion of the precursor to fully

conjugated PPV.28

B. Finite element model: Equations and parameters

We use finite element modeling (Comsol Multiphysics)

to calculate the temporal evolution of the temperature within

the polymer and the interlayer, whose geometry is illustrated

in Fig. 1(c).

The model uses the following heat equation which

describes the temperature evolution at each point in the vol-

ume under study:29

q c@T

@tþr � ð�k rTÞ ¼ 0: (1)

Here, T is the temperature, t is the time, q is the mass den-

sity, and c is the specific heat capacity (we assume k is iso-

tropic). At thermal equilibrium @T@t ¼ 0� �

this becomes

FIG. 1. (a) Schematic representation of the experiment

set-up: The hot Wollaston wire probe is scanning across

a sample. (b) Chemical structure of PXT and PPV. (c)

Cross-section through the layer and the coordinate sys-

tem that is used later for the simulations. Point A marks

r¼ z¼ 0 and point B is at the polymer-substrate inter-

face. The PXT layer is 35 nm thick and the interlayer

140 nm thick. The point where the probe touches the

air-polymer interface it at r0¼ 122.5 nm.

124317-2 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

kDT ¼ 0; (2)

where D indicates the Laplace operator ðr2Þ.We assumed a perfectly round tip of constant tempera-

ture Ttip with a radius of 2.5 lm equal to that of the exposed

core of the Wollaston wire. We used a typical experimental

scan speed of 20 lm/s and 3 nm vertical penetration of the

polymer layer by the probe (as measured by atomic force mi-

croscopy after scanning the hot probe over the precursor

film), resulting in a contact radius of r0¼ 122.5 nm. Ttip was

set to 350 �C unless specified otherwise. Point A in Fig. 1(c)

marks the point of origin and point B is at the polymer-

substrate interface. The thicknesses of PXT, substrate inter-

layer (fused silica, ITO or gold), and glass substrate are

35 nm, 140 nm, and 1 mm, respectively.

We used the following physical parameters: (1) For

fused silica and glass: k¼ 1.4 W m�1 K�1, q¼ 2203 kg/m3,

c¼ 703 J kg�1 K�1; (2) for ITO: k¼ 8.7 W m�1 K�1,

q¼ 7100 kg/m3, c¼ 380 J kg�1 K�1; (3) for gold: k¼ 317

W m�1 K�1, q¼ 19300 kg/m3, c¼ 129 J kg�1 K�1. (4) Since

for PXT the values are not well known, as a first order

approximation we took the values of the well characterized

polymer poly(methyl methacrylate) (PMMA) that is often

used in e-beam lithography:8,9 k¼ 0.19 W m�1 K�1,

q¼ 1190 kg/m3, c¼ 1420 J kg�1 K�1.

The heat radiated from the probe to the polymer was

found to be negligible (less than 2 �C difference in tempera-

ture at the polymer surface for a 350 �C hot probe) and thus

neglected in the subsequent simulations. The complex

shrinking of the polymer upon conversion is not included in

the model.11,13

To make predictions about the adhesion of the polymer

and the minimum feature size, we need to calculate the con-

version ratio a, which is the proportion of monomer units on

the precursor polymer that are converted into PPV by the

heat from the probe. The Arrhenius equation, which is

known to describe the conversion of PXT to PPV,30 shows

the influence of exposure time t and temperature T on a:

a ¼ 1� exp �A e�EaR T t

� �; (3)

in which R is the molar gas constant. The activation energy

Ea and the pre-exponential factor A are taken from the litera-

ture (128 kJ/mol and 1019/min, respectively).30 A visualiza-

tion of this equation over time and temperature space can be

found in the supplementary information.37

III. RESULTS

A. Experimental results

Fig. 2(a) shows a tapping-mode AFM image of straight

and reproducible PPV lines (after the post-baking process)

written at 350 �C and 10 lm/s, demonstrating the feasibility

of nanoscale-resolution thermochemical lithography on ITO.

Tests conducted at 300 �C with a minimum scanning speed

of 10 lm/s did not result in patterns stably anchored to the

interlayer/substrate, so we conclude that they were washed

away during the rinsing step. Interestingly, for comparable

film thicknesses of PXT on fused silica, a lower temperature

of �250 �C ensured good adhesion of the lithographed PPV

features (writing speed of 20 lm/s).

The width (FWHM) of the lines in Fig. 2(a) is of the

order of 200 nm, but control of the scanning speed and tem-

perature allowed us to obtain much thinner lines, as indicated

in Fig. 2(b), which shows structures of lines written at

400 �C and at the maximum scanning speed of our setup of

150 lm/s. Here, the FWHM is of the order of 60 nm. Espe-

cially at these high speeds, trace and retrace of the probe do

not always overlap (see also supplementary material37) and

we achieve a trace-retrace gap of 320 nm with clearly sepa-

rated lines.

Interestingly, Fig. 2(c) shows lines with a minimum fea-

ture size down to 36 nm (380 �C, 20 lm/s), although these are

not as straight as those in Figs. 2(a) and 2(b), suggesting that

anchoring to the substrate is significantly weaker in this case,

possibly because of small inhomogeneities in the film thick-

ness, in the substrate surface roughness, or in the distribution

of surface adsorbates. Since the ITO surface roughness here

(�3.2 nm after oxygen plasma) is a factor of approx. six higher

FIG. 2. Image (a) shows a tapping-mode AFM (TM-AFM) image of lines

written at 350 �C and 10 lm/s demonstrating the ability of SThL to create

straight and reproducible lines. The initial precursor film for all experiments

on ITO was �35 nm thick. (b) Tapping mode AFM image of features pat-

terned at 400 �C and 150 lm/s and vertical profile along the white line. The

line profile is the result of the probe scanning twice across the surface (trace

and retrace), leading to a line spacing of 320 nm. (c) TM-AFM image of fea-

tures patterned at 380 �C and 20 lm/s. The resolution is 36 nm (FWHM),

albeit the lines are no longer straight and well-connected to the substrate. (d)

TM-AFM image of PPV lines written across a silicon oxide (SiO2)–gold

interface with an initially 20 nm thick precursor layer. The evaporated gold

layer is �200 nm thick. The lines were written at 400 �C at 20 lm/s across

the interface and are only hardly visible in the image due to the small height

of the features (�15 nm) compared to the interface step. The interruption of

the line near the SiO2 interface is an artifact that results from the large radius

of curvature along the patterning direction, causing the tip to touch the step

before the patterned line reaches it. A gradient image (i.e., an image of the

height gradient at each point) of (d) for better visualization of the lines is

shown in the supplementary information.

124317-3 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

than that of fused silica (as measured by atomic force micros-

copy), we suggest that a tighter control over the surface prop-

erties and possibly a higher degree of conversion is desirable

or even necessary when working on ITO. Nevertheless, these

results show that the maximum obtainable resolution is sub-

stantially independent of the thermal conductivity of the

underlying substrate, as a similar resolution of 37 nm was pre-

viously achieved on fused silica on a slightly thicker (40 nm)

precursor film.1

Experiments on gold show that SThL is also possible on

very high thermal conductivity interlayers as demonstrated

in Fig. 2(d). This is a positively surprising result as one

might have expected that the high thermal conductivity of

gold would have led to insufficient conversion near the

polymer-gold interface and hence to insufficient anchoring.

As we show in the following, finite element simulations give

us further insight into this and other aspects.

B. Simulation results

First, we studied the temporal evolution of the tempera-

ture TB at point B, at the polymer-substrate interface (illus-

trated in Fig. 1(c)), and report this evolution in Fig. 3(a). As

intuitively expected, TB vs. t curves are different depending

on k, and reach significantly different plateau values at ther-

mal equilibrium (namely 141 �C for fused silica, compared

to 60.4 �C and 21.8 �C for ITO and gold, respectively). We

also see that the time needed for the polymer temperature to

respond is of the order of 10�6 s. This is much shorter than

the “pixel exposure time” to the hot probe (12 ms), which we

can derive by using the writing speed of 20 lm/s and a

probe-sample contact width of 2 r0¼ 245 nm. Therefore,

from here on only steady-state temperature distributions will

be considered.

Figure 3(b) shows the (steady-state) temperature profile

along the z-axis. The curve starts at point A at the probe-

polymer interface with Ttip¼ 350 �C. Up to B, i.e., within the

polymer, the temperature gradient is in good approximation

constant and dependent on k. The constant temperature gra-

dient is a result of the large probe-sample contact width

(245 nm) as compared to the film thickness (35 nm), resulting

in a heat transfer in this region that is essentially one-

dimensional (in one dimension, Eq. (2) becomes @2

@z2 T ¼ 0

and hence describes a constant temperature gradient).

Figure 3(c) illustrates instead the conversion ratio aalong the z-axis, which follows from the temperature distri-

bution in Fig. 3(b). We have noted in our previous work1

that the non-linear dependence of the reaction rate on poly-

mer temperature can lead to sharp boundaries between con-

verted and unconverted material. Indeed, we observe that adrops from 95% to 5% within a layer of thickness Dz� 5 nm

independent of k.

In the case of ITO and gold, we also find that a drops to

essentially zero well before the base of the film (z¼ 35 nm),

thus signaling the presence of a region of unconverted mate-

rial at the base of the film. Although one might expect that

lack of conversion in this region should result in features

being washed away during rinsing, we do not observe this

effect in experiments on ITO or gold. We will discuss this

further in the following section, but we note in the meantime

that discrepancy with the experiments (demonstrating the

features to be retained) can be resolved for example by tak-

ing into account the significant ITO surface roughness, the

electrostatic interaction of the positively charged precursor

polymer chains with the ITO interlayer, the volume reduc-

tion of polymer chains during conversion and polymer chain

FIG. 3. Finite element modeling. (a) Simulated temporal evolution of the

temperature at point B as defined in Fig. 1. (b) Simulated temperature distri-

bution along the z-axis (r¼ 0) at steady state. (c) Conversion ratio a along

r¼ 0 belonging to the temperature distribution shown in (b). (d) Vertical dis-

tance dz of the conversion boundary from the substrate for different tip tem-

peratures Ttip. The red line marks an estimate of the largest possible dz

(dzmax) which still ensures that the structure will not be washed away during

the rinsing step. (e) Plot of a along the air-polymer interface. (f) Plot of

dr¼FWHM�2 r0 for different Ttip. The red circles indicate the expected

smallest dr which follow from dz¼ dzmax¼ 11.7 nm for the different sub-

strates. Note the overall increase of the minimum feature size as a function

of k. (g) Surface plots of the temperature and conversion ratio in both spatial

dimensions in the case of an ITO substrate and a 350 �C hot tip.

124317-4 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

entanglement in the region between the converted feature

and the base of the films.

To quantify the location of the conversion boundary,

we arbitrarily define it as the surface where a¼ 50% (given

the steep variation of a, the exact fraction is not crucial).

We find that the vertical distance dz between point B at the

base of the film and the conversion boundary increases

with k, and is 2.7, 11.0, and 13.2 nm for fused silica, ITO

and gold, respectively, in the case of a probe kept at

350 �C.

In Fig. 3(d), dz is plotted as a function of Ttip ranging

from 200 to 450 �C for the various substrates, and the hori-

zontal line marks dz¼ 11.7 nm. The significance of the cho-

sen value of 11.7 nm will be discussed more extensively in

Sec. IV, but we note already that if we assume that pattern

anchoring will occur for values of dz below a certain thresh-

old (to be determined, e.g., from the minimum temperature

required for such adhesion on silica substrates), we should

then be able to use Fig. 3(d) to help determine the minimum

temperature Ttipmin of the probe, necessary to ensure adhe-

sion of the lithographed patterns.

In Fig. 3(e), we plot a along the polymer top surface for

Ttip¼ 350 �C. The conversion ratio here drops from 95% to

5% within a layer of Dr� 10 nm for gold or 16 nm for fused

silica.

We now define dr as the FWHM of the region delimited

by the conversion boundary minus the probe contact width

(2 r0) and we plot it as a function of Ttip in Fig. 3(f). The rea-

son for this definition of dr is to serve as an indicator for the

expected minimum feature size as the probe-sample contact

width is reduced towards zero. (Because the width of the

converted volume decreases upon going deeper into the

polymer layer, dr< 0 is possible for small Ttip. However,

structures are not attached to the substrate for such small val-

ues of Ttip.) The circles in Fig. 3(f) indicate dr at the mini-

mum tip temperature Ttipmin which we obtained by assuming

that the distance between the converted region and substrate

may not exceed dz¼ 11.7 nm.

Figure 3(g) illustrates the temperature T and conversion

ratio a in both spatial dimensions to get a better impression

of the shape of the converted volume.

IV. DISCUSSION

We start our discussion from the temporal evolution

in Fig. 3(a). As already mentioned in the previous section,

this indicates that heat transfer occurs essentially on a

ls-timescale. For “pixel exposure times” that are signifi-

cantly larger than this, e.g., in the ms regime (as in our case),

we can then use the simplified form of Eq. (1) reported as

the time-independent Eq. (2), and it also follows that the spe-

cific heat capacity c and material density q should not influ-

ence size and shape of the final lithographed pattern.

Therefore, we concentrate solely on the thermal conductivity

k of the materials. Furthermore, the fact that point B (at the

base of the film) is approximately at room temperature in the

case of gold means that a further increase in k will not

change the curve significantly. Hence, gold can be consid-

ered as a limit case representing k-values from �kgold to1.

A. Adhesion to the substrate

Our experiments show that the lithographed patterns are

not rinsed away during development, even for those process

parameters for which the model predicts dz> 0, i.e., a finite

vertical distance between substrate and conversion boundary

(indicating an unconverted region at the base of the film), as

a consequence of the relatively low temperature of the inter-

layers with high thermal conductivity (e.g., for Au TB

� room temperature). This is surprising and in contrast to a

model of SNOL of the PPV precursor, for which reasonable

agreement was found when assuming that the UV-dose at

point B determines the resolution.13

1. Thermal conductivity of the polymer

The first parameter we scrutinize to analyze this discrep-

ancy is the value assumed in the model for PXT’s thermal

conductivity, kPXT. This is not known accurately and it is far

from trivial to measure because the material significantly

changes its thermal properties by chemical reaction during

the very experiment used to measure it. If the true kPXT was

greater than the value in our model, it would increase heat

flow from the tip, hence raise the polymer temperature in

thermal equilibrium and lead to a smaller dz. However, con-

jugated polymers are known for their relatively low thermal

conductivities and the chosen value of 0.19 W m�1 K�1 for

PXT is slightly larger than reported values for undoped and

unstretched conjugated polymers such as polyaniline,31 poly-

thiophenes,32 and other PPV derivatives.33 Therefore, we

consider that values of dz as high as 10 nm, which are pre-

dicted by the model, are not an artifact of a poorly chosen

kPXT. Corroborating evidence for this is that we calculate

that the thermal conductivity of the polymer would need to

be greater by a factor of five for dz to reduce to zero in case

of an ITO interlayer, or by an even larger factor for Au inter-

layers (for Ttip¼ 350 �C; scanning speed¼ 20 lm/s).

We are thus confident that a “gap” between the con-

verted material and the substrate does indeed exist, and must

therefore consider which factors might help the features

adhere to the surface even when dz> 0.

2. Factors affecting feature adhesion

Such factors (which we also already mentioned briefly

in the results section) are: (a) the interlayer/substrate surface

roughness, (b) entanglement between polymer chains in the

converted and unconverted region, and (c) electrostatic inter-

action of the precursor polymers with the interlayer ITO.

Surface roughness in the range of a few nm can already

give a sizable contribution to the reduction of the “effective

dz” because non-conformal coverage of the rough surface (so

as to form a flat top surface, as it is commonly observed)

would result in a spatially non-uniform film thickness, with

local minima a few nm smaller than the thickness assumed

in the model. Furthermore, it is conceivable that the precur-

sor polymer chains would more easily get entangled and

adsorbed via a range of physico-chemical interactions with

the nanocavities offered by the rough surface.

124317-5 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

In point (b), we need to consider the effect of polymer

chain entanglement at the boundary between the converted

and unconverted region. While it might be argued that this

should be of the order of the polymer gyration radius

(3–4 nm for some soluble PPVs, which should provide a rea-

sonable model for PXT),1 we cannot rule out a value a few

nanometers larger than that (up to an extreme value of

�10 nm for a fully elongated strand).1 This higher value

could in fact be induced by the significant uncoiling and

straightening of the chains taking place near and on both

sides of the conversion boundary, as a result of the conver-

sion of the single to double bonds. There is concomitant vol-

ume reduction associated with the conversion28 (due to

elimination of the tetrahydrothiophenium) that (by defini-

tion) implies local mass transport either vertically or laterally

and may affect the true value of dz.

In point (c), we note that we have observed that a mono-

layer of PXT can be electrostatically bound to the ITO sur-

face which has been treated by an oxygen plasma. The

oxygen plasma leads to the formation of a dipole layer

on ITO via the oxidation of surface SnIV–OH to surface

SnIV– O.27,34 PXT acts as a polyelectrolyte in water where

the chains are positively charged and compensated by Cl�

counterions, and the positively charged chains can adhere to

SnIV–O surface groups. As a final point, we note that conver-

sion of PXT to PPV releases HCl,35 which has in the past

been proposed to etch metallic or ITO substrates with forma-

tion of the relevant salts. Although HCl should be evolved

mainly towards the top of the film, at distances of >10 nm or

so from the boundary with ITO (or Au), diffusion down-

wards may enable HCl to reach the interlayer. These salts

may locally increase surface roughness or introduce polar

interactions that would aid adhesion of the features.

None of these effects question the finite value of dz, but

instead confirm that the predicted dz> 0 actually captures an

important aspect of the physics of this process, and ultimately

also explain why we can anchor nanopatterns on gold, despite

its very low surface temperature (TB � room temperature).

With this in mind, we can now look at the implications that

the scenario of adhesion in the dz> 0 regime has for the ulti-

mate performance of SThL on a range of substrates.

3. Minimum probe temperature

To do this, we first extract the maximum dz which

should ensure the patterns are anchored to the surface (dzmax)

by combining the results of this model with our previous

experimental results. To this end, we plot dz as a function

of Ttip in Fig. 3(d), as predicted by the model. In our previous

experiments,1 we found that the lowest tip temperature

(Ttipmin) necessary for anchoring was 250 �C, on fused silica

and for a writing speed of 20 lm/s and a similar film thick-

ness. Putting these parameters into our model, we obtain

dzmax¼ 11.7 nm and plot this as a thick (red) horizontal line

in Fig. 3(d). The intersection of this line with the ITO curve

gives Ttipmin¼ 335 �C, which is in good agreement with the

experimental observation that 300 �C was too low a tempera-

ture for writing speeds �10 lm/s, whereas we could write

very faint structures at 350 �C for speeds up to 80 lm/s.

This remarkable agreement between experiment and

model strengthens our assertion that we do not need to convert

to the base of the film to achieve high resolution SThL. Under

the same conditions, the model predicts Ttipmin� 393 �C for

gold, and furthermore our modeling shows that increasing the

thermal conductivity beyond that of gold has almost no effect

on dz. We can therefore conclude that SThL of this material

system is never limited by the thermal conductivity of the

substrate.

At this point, we note that the thermal conductivity of a

thin film is usually smaller than that of the same material in

the bulk. For 200 nm gold films, k may be reduced by a fac-

tor of about 0.5 compared to the bulk value,22 which, accord-

ing to our model, results in changes of dz of less than 0.1 nm

and changes in dr of no more than 0.2 nm for Ttip between

250 and 450 �C.

B. Minimum lateral feature size

1. Minimum feature size for a fixed probe temperature

In Fig. 3(e), we see that the lateral distance r0 from the

edge of the probe-polymer contact to the conversion bound-

ary is 10.4, 12.8, and 18.9 nm for gold, ITO and fused silica,

respectively, for a probe temperature of Ttip¼ 350 �C. Hence,

a higher k actually leads to a smaller width of converted

material if Ttip remains constant. dr, which was defined as

FWHM �2 r0, will serve as an indicator for the expected

minimum feature size f that can be estimated by f¼ drþw0,

where w0 is the experimental probe-polymer contact width.

2. Minimum feature size for the substrate dependentminimum probe temperature

The question now remains if the need for a higher Ttip

for substrates with a higher k will overcompensate the

improvement of the resolution that we would expect for a

constant Ttip. To examine this further we plotted dr as a func-

tion of Ttip in Fig. 3(f). The (red) circles mark the intersec-

tions between the relevant curves for dr and the smallest

necessary tip temperatures Ttipmin for feature adhesion

(which we obtained from dzmax¼ 11.7 nm from Fig. 3(d)),

and therefore provide us with the corresponding, substrate

dependent drmin. Since dr

min increases with k, the model pre-

dicts that the resolution will deteriorate for higher thermal

conductivity substrates.

Quantitatively, f is expected to increase with respect to

fused silica by 10.3 nm and 12.6 nm for ITO and gold,

respectively, for a 35 nm precursor layer. Experimentally, we

found a minimum FWHM of 37 nm on a 40 nm thick precur-

sor film on fused silica and 36 nm on a 35 nm thick precursor

film on ITO, but with evident anchoring problems (Fig.

2(c)). If we therefore consider that the FWHM of the small-

est well-anchored feature on ITO lies somewhere between

the 36 nm from Fig. 2(c) and the 65 nm from Fig. 2(b), we

find a reasonable agreement with the model, which predicted

�47 nm (37 nm on fused silicaþ 10.3 nm upon switching to

ITO).

We note that our model required only the input of

the experimental value of the Ttip necessary to obtain the

124317-6 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

minimum FWHM of adhered, lithographed features on a cer-

tain substrate interlayer (which we denote fopt, with the inter-

layer being silica in our case), to enable determination of the

missing parameters dzmax (e.g., from Fig. 3(d)) and the “true”

contact width w0 (as fopt� 2 dr, determined with the help of

Fig. 3(f)) for that particular type of substrate interlayer.

These two parameters are then sufficient to predict the mini-

mum feature sizes on arbitrary substrate interlayers (such as

ITO and Au, as in the present work).

3. Limitations of the model

We have shown that the lateral resolution achieved in

the experiment can be reproduced with the model after

accounting for the difference between the modeled contact

width (2 r0) and the “real” contact width w0, a parameter that

can be extracted from experiments. This adjustment is neces-

sary due to several effects causing a change in feature sizes

that have not been introduced in the model yet, most impor-

tantly: (a) the writing speed dependent penetration depth and

hence contact width that is caused by the viscoelasticity of

the polymer, (b) the observed difference between the probe

indentation width before development and the post-

development feature size (see supplementary information of

Ref. 1), and (c) how the shrinking of the polymer (due to the

elimination of the tetrahydrothiophene group during conver-

sion) influences the structure during the writing process. The

effect of post-baking on the other hand has been investigated

for PPV structures1,13 and it was shown1 that whereas the

height of the features written by SThL shrinks by about 30%,

the width (FWHM) is almost unchanged. In the supplemen-

tary information, we give further data on the influence of the

modeled contact width (2 r0) on values of dz and dr.

(d) Another aspect that has not been considered so far is

that the final structures seem to be wider at the bottom than at

the top (even after accounting for AFM tip convolution

effects), which is in contrast to the shape of the converted vol-

ume that follows from the simulated steady state temperature

distribution (Fig. 3(g)). A model for this apparent collapse of

the converted polymer onto the substrate has previously been

fitted to experimental data in the case of SNOL.36

The effect of film thickness has not been investigated

here in detail. General trends, however, are that a higher film

thickness requires a larger Ttip to obtain the necessary dzmax.

This will in turn increases the lateral heat spread and there-

fore increases the minimum feature size.

(e) We also investigated the influence of thermal contact

resistances at the various interfaces by modeling a thin, ther-

mally resistive air-layer in between the respective interface.

The results are shown in full in the supplementary informa-

tion, but we summarize here that thermal contact resistance

at the glass-interlayer interface and at the polymer-interlayer

interface is negligible as long as it does not exceed values

equivalent to a �20 nm air gap. The probe-polymer interface

is more sensitive to a thermal contact resistance but even

then we find that an effective 10 nm thick air-gap changes dz

only by up to 1 nm and dr by up to 3 nm.

Despite these margins for improvement, our model cap-

tures the fundamental physics of SThL and, most importantly,

provides quantitative explanations for both the almost sub-

strate independent minimum resolution as a result of two

counteracting effects, and the ability to retain lithographed

patterns on very high k interlayer/substrates, such as gold, for

which the temperature at the base of the film is both intui-

tively expected and quantitatively predicted to be close to

room temperature.

V. CONCLUSIONS

In summary, we have shown that thermochemical lithog-

raphy of the precursor of the conjugated polymer PPV is pos-

sible on substrates spanning three orders of magnitude in

thermal conductivity. We achieve a maximum resolution

(FWHM) of 36 nm on a �35 nm thick precursor layer on

ITO, which is almost identical to the published FWHM of

37 nm for a 40 nm precursor film achieved when using fused

silica substrates.1 Finite element simulations predict a

“conversion boundary” that is several nm away from the film

base (dz> 0) in the experimental conditions for high-

resolution features. Taken together with the experiments con-

firming that the patterns can still stick to the substrate during

development, this result provides clear evidence of additional

adhesion mechanisms between films and interlayer/substrate

surfaces. We propose that adhesion through a thin layer of the

“unconverted” precursor polymer is made possible by a com-

bination of non-covalent secondary interactions such as polar

interactions of the polycationic chain strands with the sub-

strates, chains entanglement, and enhanced roughness effects.

This explains why SThL is also possible on gold, which

features such a high thermal conductivity that the temperature

near its polymer interface is not high enough to convert the

polymer in that region. The model further predicts that

although a higher thermal conductivity substrate is, every-

thing else being constant, expected to lead to smaller feature

sizes, the achievable lateral resolution is expected to deterio-

rate slightly upon increasing k due to the higher probe temper-

atures required. Nevertheless, the difference of the FWHM

between fused silica and gold is predicted to be only 12 nm.

From a technological point of view, we showed that SThL is

an interesting technique that can achieve nanoscale resolu-

tions on a range of substrates and therefore applications.

ACKNOWLEDGMENTS

We thank the RS and the EC for funding of the RTN

THREADMILL (EU-Contract: MRTN-CT-2006-036040), of

the ITN SUPERIOR (PITN-CT-2009-238177), as well as the

EC Seventh Framework Programme (FP7/2007-2013) under

Grant Agreement No. 212311 (ONE-P).

1O. Fenwick, L. Bozec, D. Credgington, A. Hammiche, G. M. Lazzerini, Y.

Silberburg, and F. Cacialli, Nat. Nanotechnol. 4, 664 (2009).2A. S. Basu, S. McNamara, and Y. B. Gianchandani, J. Vac. Sci. Technol.

B 22, 3217 (2004).3R. Szoszkiewicz, T. Okada, S. C. Jones, T. D. Li, W. P. King, S. R.

Marder, and E. Riedo, Nano Lett. 7, 1064 (2007).4Z. Wei, D. Wang, S. Kim, S.-Y. Kim, Y. Hu, M. K. Yakes, A. R.

Laracuente, Z. Dai, S. R. Marder, C. Berger, W. P. King, W. A. de Heer,

P. E. Sheehan, and E. Riedo, Science 328, 1373 (2010).

124317-7 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

5D. Pires, J. L. Hedrick, A. D. Silva, J. Frommer, B. Gotsmann, H. Wolf,

M. Despont, U. Durig, and A. W. Knoll, Science 328, 732 (2010).6A. K.-K. Wong, Resolution Enhancement Techniques in Optical Lithogra-phy (SPIE—The International Society for Optical Engineering, Washing-

ton, 2001).7G. Gigli, R. Rinaldi, C. Turco, P. Visconti, R. Cingolani, and F. Cacialli,

Appl. Phys. Lett. 73, 3926 (1998).8C. Vieu, F. Carcenac, A. Pepin, Y. Chen, M. Mejias, A. Lebib, L. Manin-

Ferlazzo, and H. L. L. Couraud, Appl. Surf. Sci. 164, 111 (2000).9A. E. Grigorescu and C. W. Hagen, Nanotechnology 20, 292001 (2009).

10E. Betzig and J. K. Trautman, Science 257, 189 (1992).11R. Riehn, A. Charas, J. Morgado, and F. Cacialli, Appl. Phys. Lett. 82(4),

526 (2003).12L. P. Ghislain, V. B. Elings, K. B. Crozier, S. R. Manalis, S. C. Minne, K.

Wilder, G. S. Kino, and C. F. Quate, Appl. Phys. Lett. 74, 501 (1999).13D. Credgington, O. Fenwick, A. Charas, J. Morgado, K. Suhling, and F.

Cacialli, Adv. Funct. Mater. 20, 2842 (2010).14E. Schaffer, S. Harkema, M. Roerdink, R. Blossey, and U. Steiner, Adv.

Mater. 15, 514 (2003).15U. Durig, G. Cross, M. Despont, U. Drechsler, W. Haberle, M. I.

Lutwyche, H. Rothuizen, R. Stutz, R. Widmer, P. Vettiger, G. K. Binnig,

W. P. King, and K. E. Goodson, Tribol. Lett. 9, 25 (2000).16B. Gotsmann, U. Durig, J. Frommer, and C. J. Hawker, Adv. Funct. Mater.

16, 1499 (2006).17Y. Doi, A. Saeki, Y. Koizumi, S. Seki, K. Okamoto, T. Kozawa, and S.

Tagawa, J. Vac. Sci. Technol. B 23, 2051 (2005).18A. Tilke, M. Vogel, F. Simmel, A. Kriele, R. H. Blick, H. Lorenz, D. A.

Wharam, and J. P. Kotthaus, J. Vac. Sci. Technol. B 17, 1594 (1999).19Y. Takashi, T. Kimiaki, S. Yasushi, T. Naoyuki, B. Tetsuya, and S. Yuzo,

J. Vac. Sci. Technol. A 23(4), 1180 (2005).

20M. Schlott, W. Dauth, M. Kutzner, B. Gehman, and S. Vahlstrom, U.S.

patent 6,187,253 (2001).21D. H. Damon, Phys. Rev. B 8, 5860 (1973).22G. Chen and P. Hui, Appl. Phys. Lett. 74, 2942 (1999).23M. Reading, D. M. Price, D. B. Grandy, R. M. Smith, L. Bozec, M. Conroy,

A. Hammiche, and H. M. Pollock, Macromol. Symp. 167, 45 (2001).24V. V. Gorbunov, N. Fuchigami, J. L. Hazeland, and V. V. Tsukruk, Lang-

muir 15, 8340 (1999).25E. Gmelin, R. Fischer, and R. Stitzinger, Thermochim. Acta 310, 1 (1998).26A. Majumdar, Annu. Rev. Mater. Sci. 29, 505 (1999).27J. S. Kim, F. Cacialli, and R. Friend, Thin Solid Films 445(2), 358

(2003).28J. Morgado, F. Cacialli, J. Gruner, N. C. Greenham, and R. H. Friend, J.

Appl. Phys. 85, 1784 (1999).29D. V. Widder, The Heat Equation (Academic, New York, London, 1976).30H. V. Shah and G. A. Arbuckle, Macromolecules 32(5), 1413 (1999).31J. Jin, M. P. Manoharan, Q. Wang, and M. A. Haque, Appl. Phys. Lett. 95,

033113 (2009).32B.-Y. Lu, C.-C. Liu, S. Lu, J.-K. Xu, F.-X. Jiang, Y.-Z. Li, and Z. Zhang,

Chin. Phys. Lett. 27, 057201 (2010).33Y. Hiroshige, M. Ookawa, and N. Toshimab, Synth. Met. 157, 467

(2007).34D. J. Milliron, I. G. Hill, C. Shen, A. Kahn, and J. Schwartz, J. Appl. Phys.

87, 572 (2000).35J. Morgado, D. S. Thomas, R. H. Friend, and F. Cacialli, Synth. Met. 111,

549 (2000).36D. V. Cotton, C. J. Fell, W. J. Belcher, and P. C. Dastoor, J. Phys. D:

Appl. Phys. 41, 195107 (2008).37See supplementary material at http://dx.doi.org/10.1063/1.4729809 for

further simulation data and further lithography examples.

124317-8 Tolk et al. J. Appl. Phys. 111, 124317 (2012)

1

Supplementary  Information  

Paper title: The influence of the substrate thermal conductivity on scanning

thermochemical lithography

Authors: Marten Tolk, Oliver Fenwick, Sadi Ahmad and Franco Caciallia)

Affiliations: Department of Physics and Astronomy, and

London Centre for Nanotechnology, University College London,

London WC1E 6BT, United Kingdom

Email: a) f.cacialli@ucl.ac.uk

I.  Conversion  ratio    

SUPPLEMENTARY FIG. 1 shows the conversion ratio α , i.e. the ratio of converted to

initially unconverted precursor monomers, as a function of temperature T and exposure time t

as calculated from the Arrhenius equation (Eq. 1). The pre-exponential factor (A = 1019/min )

and the activation energy (Ea = 128 kJ/mol) for this conversion reaction was taken from the

literature1.

x ,1 e paE

RT tAeα−⎛ ⎞

⎜ ⎟⎜ ⎟⎝ ⎠

= − − (1)

2

SUPPLEMENTARY FIG. 1. Visualization of the Arrhenius equation (Eq. 1) applied to the precursor

poly(p-xylene tetrahydrothiophenium chloride).

II.  Probe  motion  

SUPPLEMENTARY FIG. 2 shows the path of a probe while scanning over a surface at

450°C and 80 µm/s. The hot probe approaches the surface at position A, moves up (trace) and

down (retrace) along the fast scan axis, then moves right along the slow scan axis and again

up and down along the fast scan axis, etc. At the end of the scanning process, the probe

returns to its initial/final position A along the path L. Due to the longer exposure time of the

precursor in the proximity of A, a considerable amount of polymer has been converted in that

region. The images shows that especially at high writing speeds of 80 µm/s and faster, there

can be a gap between trace and retrace of the probe. We see that the trace-retrace gap

becomes smaller with further distance from the trajectory along the slow scan axis.

3

SUPPLEMENTARY FIG. 2. AFM image of a patterned ITO substrate at 450°C and 80 µm/s showing

the motion of the probe while scanning across a surface. In particular one can see a trace-retrace gap

which becomes smaller towards the top of the image.

III.  Thermal  lithography  on  gold  

SUPPLEMENTARY FIG. 3 shows PPV lines written across a SiO2/gold interface. Because the

lines are flat compared to the interface step (200 nm), a gradient image of (a) is shown in (b)

to better distinguish the lines from the substrate. One can also see how the PPV line is broken

just above the interface, which is a result of the rather large radius of curvature of the probe

in direction of the fast scan axis.

4

SUPPLEMENTARY FIG. 3. (a) AFM image of PPV lines written across a silicon oxide (SiO2) –

gold interface at 400°C at 20 µm/s with an initially 20 nm thick precursor layer. The evaporated gold

layer is ≈ 200 nm thick. (b) Gradient image (i.e. an image of the height gradient at each point) of (a)

for better visualization of the lines.

IV.  Influence  of  thermal  contact  resistance  

We investigated the influence of thermal contact resistance at the various interfaces

(substrate/interlayer, polymer/interlayer and probe/polymer) on the results of the model by

changing the boundary condition of the respective interface from 'continuity' to a 'thin

thermally resistive layer' with the thermal conductivity of air (kair) and thickness tair. Note that

the geometry does not change and dz and dr keep their original meanings, as the contact

resistance effectively originates from an infinitely thin layer of finite thermal contact

resistance, leading to a temperature drop across the interface.

kair is set to the function given by the Comsol material library: kair = (-2.2758·10-3 + (T/K) ·

1.1548·10-4 + (T/K)² · (-7.90253)·10-8 + (T/K)³ · 4.11703·10-11 + (T/K)4 · (-7.4386)·10-15)

W m-1 K-1.

5

In SUPPLEMENTARY FIG. 4, we plot dz and dr as a function of tair for a tip temperature

(Ttip) of 350 °C. We find that an imperfect contact between the interlayer and the underlying

glass substrate (see SUPPLEMENTARY FIG. 4(a)) has no influence as long as the effective

air-gap is thinner than ~20 nm. Note that fused silica substrates do not feature a

substrate/interlayer interface as the material is the same for substrate and interlayer.

Nevertheless, for completeness the additional thermal contact resistance was modeled also in

this case.

A thin air-layer between polymer and interlayer (see SUPPLEMENTARY FIG. 4(b)) has a

similar influence on dz as a thin air-layer between glass and interlayer, but its influence on dr

is larger. Note, however, that we expect the effective air-gap to be small for all interlayers

because first, surface roughnesses are in the nm range and second, the precursor polymer

(PXT) solution is liquid during its deposition (via spin-coating) and is hence expected to

effectively fill any gaps.

Although we also find that a thermal resistance between the tip of the probe and the polymer

(see SUPPLEMENTARY FIG. 4(c)), has, by far, the strongest influence on the lateral

resolution and on dz., this is still relatively small in absolute terms. I.e. for an air-gap of

10 nm, dz changes by up to 1 nm and dr changes by up to 3 nm. At tair larger than ~ 300 nm,

the probe cannot convert material inside the polymer film anymore, so that dz approaches

32 nm, i.e. the polymer film thickness (35 nm) subtracted by the probe penetration

depth (3 nm at r = 0). At this point, dr approaches an interlayer-independent value of -39 nm.

6

SUPPLEMENTARY FIG. 4. Diagrams showing the influence of the addition of an air-layer between

(a) the glass/interlayer interface, (b) the polymer/interlayer interface, and (c) the tip/polymer interface

for a tip temperature Ttip of 350°C. Values of dz (black, filled symbols) and values of dr ( red, open

symbols) are given as a function of the thickness of the thin thermally resistive air-layer (tair).

V.  Influence  of  contact  width  

Another uncertainty in the model is the contact width between the tip of the probe and the

polymer (2 r0). The uncertainty stems from several effects, such as the writing-speed

7

dependent penetration depth and hence contact width, the observed reduction in feature size

upon development,2 and the lateral shrinking during the post-baking step2. Most of the

modeling is done with a penetration depth of 3 nm and hence a contact width of 245 nm. A

smaller contact width will reduce the temperature in the polymer at thermal equilibrium and

thus increase dz and reduce dr. This is shown in the case of a 350 °C hot probe in 5(a). We

also see that at very small contact widths (smaller than the polymer film thickness, here

35 nm), lateral heat diffusion in the polymer starts to dominate and dz becomes almost

independent of the type of interlayer.

We further show the effect of the tip temperature (Ttip) on dz and dr for different contact

widths in case of indium-tin oxide (ITO) as the interlayer in 5(b). We see again that dz grows

for smaller contact widths, further supporting the notion that we expect a region of

unconverted polymer near the interlayer.

SUPPLEMENTARY FIG. 5. (a) dz and dr as a function of the probe-polymer contact width for a

constant tip-temperature (Ttip) of 350 °C. (b) dz and dr as a function of Ttip for different contact widths

(30, 60, 120 and 245 nm) in case of indium-tin oxide (ITO) as the interlayer.

8

References

1 H. V. Shah and G. A. Arbuckle, Macromolecules 32 (5), 1413 (1999). 2 O. Fenwick, L. Bozec, D. Credgington, A. Hammiche, G. M. Lazzerini, Y. Silberburg, and F.

Cacialli, Nat. Nanotechnol. 4, 664 (2009).