SUPPLEMENTARY INFORMATION

doi: 10.1038/nphoton.2011.222

Breaking the diffusion limit with super hydrophobic

delivery of few molecules to plasmonic nanofocusing
structures

F. De Angelis1,2, F. Gentile1,2, F. Mecarini1, G. Das1, M. Moretti1, P. Candeloro2,

M.L. Coluccio1,2, G. Cojoc2, A. Accardo1,2, C. Liberale1, R.P. Zaccaria1, G.
Perozziello1, L. Tirinato2, A. Toma1, G. Cuda2, R. Cingolani1, and E. Di
Fabrizio1,2

1
Department of Nanostructrures, Istituto Italiano di Tecnologia (IIT), via Morego 30, I16163
Genova, Italy.
2
BIONEM Lab, University of Magna Graecia, Campus S. Venuta, Germaneto, viale Europa,
I88100 Catanzaro, Italy,

Supplementary Information

SI1. DESIGN OF SUPERHYDROPHOBIC SURFACES

SI2. FABRICATION OF SUPERHYDROPHOBIC SURFACES

SI3. SAMPLE PREPARATION ANDCHARACTERIZATION TECHNIQUES

SI4.-DNA DEPOSITION AT DIFFERENT DILUTION

SI5. FDTD simulation

nature photonics | www.nature.com/naturephotonics 1

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

SUPPLEMENTARY INFORMATION #1: DESIGN OF SUPERHYDROPHOBIC SURFACES

#1.A Introduction

It is well known that a drop post upon a solid surface makes a contact with the solid
described by the sole parameter θe (Fig.S1A,B), that is, the equilibrium contact angle at the
interface between the liquid and the solid. θe obeys the Young equation [1]:
γ SV − γ SL
cos θ e = (1)
γ LV
where γij is the surface tension between the phase i and j, and the letters S, L, V stand for
the solid, liquid and vapour. The equation (1) can be regarded as a simple balance of forces.
For sufficiently small droplets the dominant force becomes the liquid-vapor surface tension
and gravitation may be neglected [2]. The dimensionless Bond number can be consequently
introduced as Bo=ρ×g×R2/ γLV, where ρ is the density of the liquid, R is the radius of the
spherical drop prior the deposition upon the surface, and g is the acceleration due to
gravity. When Bo<<1 gravitational effects vanish and the shape of the droplet may be
assumed spherical everywhere [3-5]. For a drop of water with γLV=72.9 mJ/m2, ρ=1000
kg/m3, and diameter d=2×R=1 mm, it follows that Bo~0.035, and thus the physics of
micrometric or submillimetric drops is correctly governed by surface tension solely1.

Fig.S1.1

Flat surfaces may have, via chemical modifications, a contact angle that is 120° at most [6].
Artificial super hydrophobic states are possible for surfaces comprising an underlying
pattern or texture, in fact it is widely observed and very well understood that roughness
increases hydrophobicity [7]. Two distinct mechanisms exist that yield superhydrophobic
states, and these are well described by the models of Wenzel [8] and Cassie Baxter [9].
The first mechanism (Wenzel model) is based on the increased contact area between the
surface and the liquid as due to roughness, and thus relies upon geometric effects solely.
The modified angle θe* is related to the unmodified contact angle θe as:
cos θ e∗ = r cos θ e (2)

1
Another way of looking at the problem is considering the capillary length k-1=(γLV/ρg)1/2 that for water assumes
the value k-1(water)=2.73 mm. Drops smaller than k-1(water) are then small to the extent that tension dominates
over gravitation.

2 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information
r is the solid roughness defined as the ratio between the real surface and the projected one
[10], and thus the effect of surface roughness is to amplify the wetting. Noticeably, when
θe<π/2, r increases the hydrophilicity of the surface, whereas for θe>π/2 roughness
promotes hydrophobicity.
In the second mechanism (Cassie model), air may remain trapped below the drop thus
enhancing hydrophobicity. The increased contact angle θec>θe can theoretically be written as
cos θ ec = −1 + φs (cos θ e + 1) (3)
where φs is the fraction of solid in contact with the drop. The less φs the larger the apparent
contact angle. The Cassie model is intuitive and predicts that a drop upon a patterned
surface 'sees' a contact angle that is proportional to the fraction of air in contact with the
drop. A flat surface would have φs=1 and accordingly θec=θe.
In spite of the fact that they do both induce hydrophobicity, these situations are very
different when considering their adhesive properties. In Wenzel state drops are found to be
highly pinned (Fig.S1C). On the contrary, in the Cassie state the drop sits mainly upon air,
and this increases the contact angle (Fig.1SD). Rigorously, Wenzel and the Cassie states are
distinct, meaning that they cannot coexist. In other terms, a drop upon a micro textured
surface would undergo the Wenzel or the Cassie state, not both simultaneously. A drop
would assume either one configuration or the other depending on the geometry of the
substrate (thus on φs and r) and on the surface chemistry (thus on θe).
A transition from Cassie to Wenzel occurs when cosθe=cosθeT= (φs−1)/(r−φs) or, in terms of
the aspect ratio τ (the ratio between the height and the base of a structure):

 φs − 1 
θ eT = arccos  (4)
 φs (4τ − 1) + 1 
This means that a Cassie state holds for surfaces whose chemical induced hydrophobicity is
sufficiently large, and for a fixed τ, the smaller the value of φs, the larger θeT [7, 11]. Since
small values of θeT are preferable, one should design surfaces having a texture with the
largest ratio of the solid to the total projected area (large φs), and this would guarantee
broad working areas where Cassie dominates over Wenzel [12, 13]. Notice though that
large φs would induce small contact angles, and thus the choice for the best parameters in
(4) is not trivial, and it is indeed a matter of optimization, as exposed in paragraph 1C.
In the Cassie state the drop would roll upon the surface and, most importantly for biological
applications, it would progressively reduce its contact area during a process of evaporation.
The drop would then maintain the Cassie state over time, thus avoiding collapse and any
irreversible transition to Wenzel until a critical radius of impalement is achieved. The
mechanisms inducing collapse in a slowly evaporating droplet are two and, namely, (i) the
drop could either touch the surface below the posts, or (ii) the surface free energy gained as
the drop collapses wins over the surface free energy lost by increased contact with the
hydrophobic posts [14]. While the first mechanism regards surfaces decorated with short
posts or pillars, the second relies upon an energetic argument and is independent on the
pillars' height. In many practical situations one should consider the latter criterion solely. In
particular, both analytical calculations and numerical simulations show that the critical
radius of impalement depends upon the distance between the pillars δ and θe as
δ
rmin = (5)
cos θ e

and thus the closer the pillars, or the larger θe, the smaller the final area of contact [14-
17]. Experimental evidences confirm this theoretical finding [18].

nature photonics | www.nature.com/naturephotonics 3

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

#1.B Adhesive Properties of Superhydrophobic Surfaces

In this paragraph we show that super hydrophobic surfaces can exhibit vanishingly small
friction coefficients.
A drop sitting upon a surface experiences an equilibrium contact angle θe that follows
equation (1). Considering a drop sufficiently small, as to neglect gravitational effects
(Bo<1), the radius r of the solid liquid contact can be written as
1

 4 3
β = (1 − cos θ e ) (2 + cos θ e )
2
r =   R sin θ e , (6)
β
R is radius of the spherical drop prior the deposition on the surface. For symmetry solely
half a drop is considered as in Fig.S1.2A.

Fig.S1.2

The net adhesive force Fa acting along x may be calculated [19, 20] as

Fa = ∫ γ lv (cos θ r − cos θ a ) i ⋅ d l (7)

Γ

where θr and θa denote the receding and advancing contact angles, respectively, i denotes
the unit vector along the x axis and dl a differential length of boundary Γ. From the
geometry of the problem, Eq.(7) may be expressed as
π
Fa = ∫2
π γ lv (cos θ r − cos θ a ) r cos ϕ dϕ = 2γ lv (cos θ r − cos θ a ) r. (8)
2
−

Thus the adhesion force Fa would depend upon the length of the contact line (2r), the liquid-
vapor surface tension (γlv) and the term (cosθr−cosθa) that is directly related to the contact
angle hysteresis (CAH): θa−θr2.
We consider a drop upon a micro textured surface with a periodic hexagonal lattice of
cylindrical pillars (Fig.S1.2B), where d is the diameter and δ the distance (gap) between the

2
The larger the hysteresis, the more the drop is stuck to the substrate. The CAH is due to heterogeneities (in
topography and chemical composition) which are always present at a solid surface, inducing fluctuations of the
surface tensions between the solid-vapor and solid-liquid phases [21, 22]. Notice that the equilibrium contact
angle that may be experimentally observed is not unique in that it can range between the lower and upper limits θr
and θa. An ideally smooth surface has no hysteresis effects (θa=θr=θe) and a drop would roll or slip upon the
application of any infinitesimal external force.

4 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information
pillars. The fraction φs of solid remaining in contact with the drop is φs=πd2/4(d+δ)2, and the
fraction of air is φa=1−φs. We calculate the adhesive force considering the textured
substrate. Equation (8) still holds true, but now (i) the radius r of the solid liquid contact is
reduced due to the increased contact angle (Eq.(6)) and (ii) the CAH the drop sees is not
unique due to the heterogeneous contact surface (the CAH is maximum at the solid fraction,
zero otherwise). Relation (8) can be thus rewritten as

Fac = 2r c γ lv (cosθ r − cosθ a ) = 2r cγ lv (cosθr − cosθa )φs (9)

where <Ω> is the spatial average taken on the surface of contact of the drop (it is here
assumed that the radius of contact is large compared to the pillars dimension). From this, it
follows that tiling the surface (φs<1) the hydrophobicity increases and drastically reduces
adhesion. Using relations (6) and (3) it is possible to recast equation (9) in terms of θr, θa,
θe, R and φs:
1/ 3
 4 
F = 2 R γ lv  1 − (φ s (cosθ e + 1) − 1) (cosθ r − cosθ a )φ s
c 2
 (10)
a
(2 − φ cos (1 + θ ))2
(1 + φ cos (1 + θ )) 
 s e s e 

When the drop and the material are given, the quantities θr, θa, θe, R are constant and the
sole variable in (10) is φs.
The considerations above help to draw a comparison between ideally flat and textured
surfaces. The adhesion force is:
4
Fa = ν π R 3 ρ g cos α (11)
3
Whereby Fa is written in terms of the critical angle α that is the tilting angle above which the
drop would begin to roll off the plane of contact. The friction coefficient ν may be then
conveniently determined using the well known equation
ν = tanα (12)
For the configurations at study (say pillars 10 µm wide with a gap of 20 µm) tilting angles as
low as 1° were measured, and correspondingly Eq.(12) reads ν~0.017.
Evaporation of a drop upon the surface. Initially the drop is pinned and the contact line
is fixed. The solvent evaporates over time thus decreasing the volume of the drop. The
spherical shape of the liquid has to be preserved and consequently also the contact angle
decreases. This unbalances the forces acting upon the drop and would generate a radial
force that tends to recall the contact line towards the centre of the drop, thus recovering the
initial equilibrium contact angle θec (Fig.S1.3).

nature photonics | www.nature.com/naturephotonics 5

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

Fig.S1.3

Following the reasoning above, the total force in the x direction at a generic angle θe<θec
can be evaluated as

Fp = 2r cγ lv cos θ − cos θ ec
( ) (13)

and accordingly the condition for depinning is Fp>Fac or, equivalently, Fp/2rcγlv>Fpc/2rcγlv.
When Fp§Fac, the drop is stuck on the surface. (This condition is valid, strictly speaking, in
close proximity of θe). This model is effective in that allows predicting with sufficient
accuracy the tendency of a drop to reduce its contact area while evaporating, thus
concentrating the solute therein initially dispersed.
Consider for instance a sessile water droplet post upon a continuous, smooth Teflon surface,
having the parameters φs=1, θe=114°, θr=100°, θa=125°. Conveniently using these values
within equations (13) and (9) it is found that Fac~0.8Rγlv and Fp~2Rγlv(0.4067+cosθ), and
thus Fp would be smaller than Fac for a large range of θ. In other words, during the
evaporation process, the drop will predominantly be pinned on the surface.
Consider instead the same Teflon surface decorated with a regular array of pillars as to have
φs~0.087. It then follows that Fac~0.022Rγlv and Fp~2Rγlv (0.94+cosθ). These values suggest
that a small deviation from the equilibrium position (in other words, a small decrease θeØθ)
would induce Fp>Fac, as a consequence the contact line would jump inward from a thread of
pillars to another while the drop evaporates.
Let`s now define a new parameter Υ as the ratio Υ=Fp/Fac, useful for evaluating the
tendency of a drop to slide upon a surface: when Υ>1 the depinning dominates over the
friction force, and thus the drop slides toward its centre upon evaporation. For the
configuration above (φs~0.087, that is 10 µm wide pillars with a gap of 20 µm) Υ may be
plotted against the angle ∆θ=θe−θ, that is the deviation from equilibrium. Initially, at t=0,
Υ<1, meaning that the drop would be sticky. As evaporation starts, the contact angle
decreases, and accordingly ∆θ increases. The depinning force rises with ∆θ, and so does Υ.
When Υ is sufficiently large, the contact line of the drop would recede, the system would
assume a state that is energetically favourable and the contact angle would match again the
expected equilibrium contact angle θe. Under these conditions, ∆θ=Υ=0. The process starts
over again and the same mechanism is repeated. In Fig.S1.4 the calculations are
accomplished for 10 cycles, and, at each cycle, the contact radius is reduced by a quantity
that is the distance between two lines of pillars. When r<rmin, the drop gets pinned, and the
process ends.

6 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

Fig.S1.4

For sufficiently high pillars the minimum contact radius achievable is limited by energetic
considerations and depends upon the distance of the pillars δ and θe as described by Eq.(5).
When rc§rmin the drop is impaled and an irreversible transition to the Wenzel state occurs. In
this situation the drop is definitely pinned and no further contact area shrinkage is possible.
Fig 1.5 illustrates the same process following the reduction of the contact radius of the drop
with time. The diagram is a discontinuous step function, whereby the radius is constant
while the ratio Υ is lower or equal to one. When Υ>1, (when the angle of contact becomes
sufficiently low to yield a depinning force significantly larger than the adhesion force), a
transition occurs, and the line of contact would jump inward from a line of pillars to another.
Beyond the lower bound given by Eq.(5) the contact radius is hold constant.

nature photonics | www.nature.com/naturephotonics 7

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

Fig.S1.5

#1.C Considerations for an optimal design

A closer glance to equations (1) and (5) evidences how the apparent contact angle θec
decreases monotonically with increasing φs and, similarly, for a given pillars' diameter d the
smallest contact radius rmin would be attained for large values of φs. In short, φs has
coincidental effects on both θec and rmin. Let the microtextured surface be described by the
parameters d and δ (see above for the significance of these terms). If d is fixed, then the
system would be described by the sole δ and thus φs. The aim here is to determine the
optimal subset (d0, δ0), or equivalently φs0, that would induce the higher θec still retaining the
advantages of a small rmin.
To do this, consider the radius r of the solid liquid contact as in Eq.(6), where it is intended
that the contact angle θe is actually the apparent, Cassie's θec: r would depend upon the sole
independent variable θec and thus φs (Eq.(3)). As φs gets smaller even r does it. Let's then
divide r by R, and call the output function Ψ1 (Ψ1=r/R). Ψ1 is normalized to one and accounts
for the effects of the micro structure of the surface on the contact line. Consider now the
minimum contact radius rmin before the drop collapse as in equation (5). rmin is a function of
δ
 1 − φs 
rmin = rmin (δ ) = rmin (f1 (φs ) × d ) = rmin  × d (14)
 φs 
 
Notice that rmin is not a function of the sole φs, and may be instead cast into a form that is
consistent with Ψ1 if divided by d: rmin*= rmin/d. rmin* is consequently normalized to one
upon dividing by the value of rmin* that is obtained at φs: θec(φs)=ξ, where ξ is an angle of
contact very high and still physically achievable (170°<ξ<180°). Let's term the latter
quantity Ψ2

Ψ2 = rmin
∗
/ rmin
∗
(15)
θ ec = ξ

Ψ2 accounts for the effects of the micro structure of the surface upon the smallest radius of

8 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information
contact prior the drop collapse. Let's then introduce Ψ as the sum
Ψ = Ψ1 + Ψ2 (16)

For a given d, Ψ varies with δ. Most importantly, a value of δ exists where Ψ attains a
minimum: this, for a given hexagonal lattice, would be the best value of δ, δ=δ0, providing a
sufficiently small contact line (or equivalently, high contact angle) still assuring a small
impalement radius. δ0 was computed for different d’s (0§d§30 µm) and for ξ ranging
between 162° and 170°. The results are shown in Fig.S1.6.

Fig.S1.6

Fig.S1.6 is a contour plot of the contact angle θec as a function of the pillars' diameter (d)
and distance (δ). Each point of the diagram would thus represent a lattice configuration. The
radial straight lines stemming from the origin of the axes represent the sites at constant
contact angle or constant φs. The region of the diagram in light turquoise (region of interest)
comprises all the possible couples (d0, δ0) of optimal design according to the reasoning
above. The points in the diagram are the actual parameters used for nano-patterning the
surface and mostly fall in the region of interest. Generally, d should be as small to consent a
sufficiently large number of pillars to interact with the drop. The lower boundary for d
depends upon limitations intrinsic to the process of fabrication.

#1.D Symmetry Breaking and Spontaneous Impalement

Here we consider the geometry of the device as in Fig.2E of the main text, where a cone is
enclosed in a regular pattern of cylindrical pillars. A rationale is given that would explicate

nature photonics | www.nature.com/naturephotonics 9

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

how a slowly evaporating droplet would preferentially collapse in close proximity of the
cone. An argument is introduced that relies on the mean field theory approach proposed by
Moulinet and colleagues in [15, 16].
On the application of a drop of water upon a regular hexagonal array of pillars, the portion
of the water exposed to the air experiences a force due to the internal Laplace pressure,
which may be written as
F p = PL A (1 − φ ) (17)

where A is the projected area of the surface of interest, φ=4/3π(d/δ2)2 has the meaning
given above, and PL
2γ
PL = (18)
R (t )
with R(t) the radius of the drop that is, in general, function of time. FP would push the drop
towards the substrate and thus the larger FP the more the transition to the Wenzel state is
likely to take place. Fp would be balanced by the capillary force Npf where Np= φA/π(d/2)2 is
the number of pillars in the area of interest, and f is the elementary force per pillar given by
d
f = 2πγ cos θ . (19)
2
The drop would penetrate inside the pillars and collapse if Fp>Npf or, equivalently,
combining all the quantities in Eq.s from (17) to (19), if
2
2γ  2π  d   4π d
1−   > γ cos θ 2 (20)
R (t ) 
 3  2δ   3 δ

That can be rearranged as

B d B 4π
= Ξ < 1, B= γ cos θ (21)
PL δ − π 2 3 d
2 2
PL 3
And thus the criterion for the stability of the Fakir state would be the left hand part of
Eq.(21) to be larger than one. Eq.(21) is true on average (mean field theory approach), and
here we make the assumption that the above results may be extended locally to a lattice
where the distribution of pillars in x (horizontal coordinate) and y (vertical coordinate) is
described by the sole function d(x,y).
Fig.S1.7 reproduces Ξ(d) for a network of micro pillars with a central defect (that is, a pillar
is substituted by a cone, with radius of curvature at the apex of few nm, thus severely
smaller than the pillars` diameter) where the parameters d=10 µm and d=30 µm were
used.
Notice that at `defect` Ξ(d) attains a minimum, meaning that, over the area of interest,
this is the position where the probability of impalement is the highest. This result is valid for
a generic drop`s radius, that is, when the drop is sufficiently small and energetically
favourable for impalement, the collapse would occur, most likely, at the cone.

10 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

Fig.S1.7

nature photonics | www.nature.com/naturephotonics 11

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

1. Young, Phil. Trans., 1805; (84):1; edit by Peacock.

2. McHale G., Shirtcliffe N. J. and Newton M. I., Analyst, 2004; (129):184-187.
3. Mahadevan L., Pomeau Y., Phys. Fluids, 1999; (11):2449.
4. Aussillous P. and Quéré D., Nature, 2001 ; (411):924.
5. Blossey R., Nature Materials, (2):301-306, 2003.
6. Shafrin, E. G. & Zisman, W. A. in Contact Angle, Wettability and Adhesion Advances in Chemistry series,
Vol. 43, ed. Fowkes, F. M., 145-167, American Chemical Society, Washington D.C., 1964.
7. Lafuma A. and Quéré D., Nature Materials (2):457-460, 2003.
8. Wenzel model - Wenzel, R. N., Ind. Eng. Chem. 28, 988-994, 1936.
9. Cassie model - Cassie, A. B. D. & Baxter, S., Trans. Faraday Soc. 40, 546-551, 1944.
10. J. Bico et al, Pearl Drops, Europhys. Lett., 47 (2), pp. 220-226, 1999.
11. Bico J., Thiele U. & Quéré D., Colloids Surf.A 206, 41-46, 2002.
12. Neelesh A. Patankar, Langmuir 2004, 20, 7097-7102.
13. Michael Nosonovsky and Bharat Bhushan, J. Phys.: Condens. Matter 20, 2008, 395005 -6pp.
14. H. Kusumaatmaja, M. L. Blow, A. Dupuis and J. M. Yeomans, EPL, 81, 2008, 36003.
15. S. Moulinet and D. Bartolo, Eur. Phys. J. E 24, 251-260, 2007.
16. D. Bartolo, F. Bouamrirene, E. Verneuil, A. Buguin, P. Silberzan and S. Moulinet, Europhys. Lett., 74 (2),
pp. 299-305, 2006.
17. M. Reyssat, A. Pepin, F. Marty, Y. Chen and D Quéré, Europhys. Lett., 74 (2), pp. 306-312, 2006.
18. M. Reyssat, J. M. Yeomans and D. Quéré, EPL, 81, 2008, 26006.
19. Dussan E. B. and Chow R. Tao-Ping, J. Fluid Mech. (1983), 137, 1-29.
20. Quéré D., Lafuma A. and Bico J., Nanotechnology 14 (2003), 1109-1112. De Gennes P G, 1985, Rev. Mod.
Phys. (57):827.
21. De Gennes P G, 1985, Rev. Mod. Phys. (57):827.
22. G. McHale, S. Aqil, N. J. Shirtcliffe, M. I. Newton and H. Y. Erbil, Langmuir 2005, 21, 11053-11060.

12 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information
SUPPLEMENTARY INFORMATION #2: FABRICATION OF SUPERHYDROPHOBIC
SURFACES

#2.A Fabrication process o micropillars covered with electroless

silver nanograin
(100) silicon wafers (from Jocam, Milan, Italy) were cleaned with acetone and isopropanol
to remove possible contaminant and then etched with a 4% wet HF solution. The wafers
were then rinsed with DI water and dried with N2. Standard optical lithography technique
was employed to realize regular patterns of disks within a layer of positive resist (S1813,
from Rohm and Haas) that was spin-coated onto clean silicon wafers. Electroless deposition
techniques were employed to grow silver in form of grains within the holes.
The electroless deposition is based on an autocatalytic or a chemical reduction of aqueous
metal ions. This process consists of an electron exchange between metal ions and a
reducing agent. In this work, Si substrate was used itself as reducing agent. A fluoridric acid
(HF) solution containing silver nitrate (AgNO3) was used, where Ag was reduced to metal
form by the Si substrate oxidation. In particular, the patterned silicon wafer was dipped in a
0.15 M HF solution containing 1 mM silver nitrate for 60 s at a constant temperature T=313
K. After the growth process the silicon wafer was rinsed with water and dried under nitrogen
flux. The driving force in this process is the difference between redox potentials of the two
half-reactions, which depends on solution temperature, concentration and pH.
Consequently, these parameters influence the particles size and density.
Reactive Ion etching. After residual resist removal with acetone, a Bosch Reactive Ion
Etching (MESC Multiplex ICP, STS, Imperial Park, Newport, UK) process was utilized to
obtain the final structures that are cylindrical pillars having an aspect ratio greater than 5,
diameters ranging from 5 to 15 µm and a heights larger than 20 µm, in accordance with the
model introduced in Suppl. Info #1. The electroless grown Ag layer served as mask during
the RIE process. The substrates, as a whole, were then covered with a thin (few nm) film of
a Teflon like (C4F8) polymer to assure hydrophobicity. The masks necessary for optical
lithography were fabricated using standard Electron Beam Lithography (CRESTEC CABL-
9000C Electron beam lithography system) methods. Fig.S2.1 shows SEM images of
microstructured surfaces. The cartoon in Fig.S2.2 briefly recapitulates the fabrication
process.

Fig.S2.1 Micropillars covered with silver nanograin assemblies

nature photonics | www.nature.com/naturephotonics 13

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

Fig.S2.2 Electroless and RIE Fabrication process of micropillars

#2B fabrication of micropillars decorated with regular arrays of

silver nanodots
Optical lithography, Electron beam lithography, RIE and electroless techniques were used
for the fabrication of the devices as in Fig.2c in the main text. (100) silicon wafers were
cleaned with acetone and isopropanol to remove possible contaminant and then etched with
a 4% wet HF solution. The wafers were then rinsed with DI water and dried with N2;
thereafter they were baked on a covered hot plate for 30 min at 170 ◦C to remove residual
humidity. A high resolution positive electron resist (PMMA 950k A2) was spin–coated for 60
s at 4000rpm to obtain a ~50 nm thick layer of resist. Prior to the EBL exposure, the sample
was pre-baked at 170 ◦C for 2 min to evaporate the solvent from the resist. A regular
pattern of dots spanning above an area of 500×500 µm2 was written on the sample using
an EBL system, employing a 50 keV acceleration voltage. Disc-shaped structures were
obtained using the spot-scan modality, with a dose of 150 C/cm2. The sample was then
immersed for 30 s in PMMA 950k developer (MIBK:IPA=1:4 v/v) to selectively remove the
exposed resist. Si was then processed using reactive ion etching (RIE) in oxygen for 40 s to
remove any residual resist from the discs. Arrays of empty discs with 80nm diameter and an
interspatial gap ranging from 10 to 30nm were finally obtained. The successive Ag growth
was accomplished using an electroless growth as described above. Due to resist patterning,
silver grains could grow solely inside the holes. The remaining PMMA resist was removed
with solvent. Grains assemblies have a 100nm diameter and 80nm thickness with inter-
spatial gap of 10–30 nm.
AZ 5214 image reversal photoresist (MicroChemicals GmbH, Germany) was spin-coated
onto the nano structured samples. Standard optical lithography techniques were employed
to realize regular micro-patterns of AZ 5214 squares. A Bosch RIE process was utilized to
etch the samples thus yielding, upon removal of the residual AZ 5214, prismatic Si micro-
pillars decorated with regular arrays of Ag nanodots. Please notice that these regular
patterns or nano structures induce, to different extents, a giant, local electromagnetic field
and a consequent enhancement of the Raman signal. In Fig.S2.3 is reported the flow chart
description of the process. Several sample were produced and SEM micrographs were taken
to assess uniformity and reproducibility (Fig.S2.4).

14 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

Fig.S2.3

Fig.S2.4

#2.C Fabrication micropillars combined with plasmonic nanocones

Silicon micropillars were obtained as described in supplementary information #2.A. Ag
nanograins were removed using binary mixture of HF:H20=1:4 v/v. The circular grating was

nature photonics | www.nature.com/naturephotonics 15

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

milled on the surface of the silicon micropillar by focused ion beam milling (DUAL BEAM
(SEM-FIB) - FEI Nova 600 NanoLab, acceleration voltage 30 kV, ion beam current 500pA,
polar coordinates, grating pitch 360 nm, depth 200 nm). Afterwards, a nanocone was
growth on the top of the silicon tapered pillar by employing electron beam induced
deposition (EBID) from a Platinum-based gas precursor. Finally, a thin layer of silver (40
nm) was deposited of the device by means of thermal evaporation. The apex of the silver tip
has a radius of curvature of less than 10 nm.

Fig.S2.5

#2.D FABRICATION OF Micropillars surrounding a plasmonic nanocone

The body of the nanocone was obtained by tapering a silicon micro-pillar with Focused Ion
Beam (acceleration voltage 30 kV, ion beam current 1nA, polar coordinates). A grating was
milled on the side of the cone to allow an efficient laser coupling, and surface plasmon
generation through the grating order -1: grating pitch 490 nm, depth 200 nm, number of
lines 12. Afterwards, a nanocone was growth on the top of the silicon tapered pillar by
employing electron beam induced deposition (EBID) from a Platinum-based gas precursor.
The body of the cone is about 10 µm height, whereas the base is 2 µm large. Finally, a thin
layer of silver (40 nm) was deposited of the device by means of thermal evaporation. The
apex of the silver tip has a radius of curvature of less than 10 nm, and the vertex angle
(about 7°) was adjusted to be compatible with adiabatic compression of plasmons.

16 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

Fig.S2.6

nature photonics | www.nature.com/naturephotonics 17

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

SUPPLEMENTARY INFORMATION #3: SAMPLE PREPARATION AND

CHARACTERIZATION TECHNIQUES

#3.A Materials

Rhodamine R6G, myoglobin, ribonuclease B, lysozyme and sodium chloride were purchased
from Sigma; Lambda DNA was purchased from Fermentas (Genbank/EMBL accession
number J02459). De-ionized (D.I.) water (Milli-Q Direct 3, Millipore) was used for all
experiments. All chemicals, unless mentioned otherwise, were of analytical grade and were
used as received.
Rhodamine6G is an organic compound and is used extensively in biotechnology applications.
It is a dye which can be observed very clearly by fluorescence microscopy. Its absorption
and emission wavelength are 530 and 556 nm, respectively.
Lysozyme is an enzyme able to damage bacterial cell walls by hydrolysis of the glycosidic
bond in peptidoglycans and chitodextrins. In humans, it is part of the immune system and is
present in a number of secretions such as tears, milk, saliva and mucus. A lack of the
enzyme is related to bronchopulmonary displasia in newborns [1] and disease in infant fed
with enzyme lacking formula [2].
Lambda DNA is the double stranded DNA of the Enterobacteria Phage Lambda. It is a virus
bacteriophage that infects Escherichia coli. It is a 49490 base pair long flanked by 12 base
single stranded sequence that make up the cos site matching 5’ with 3’ end. It is usually
used as a model organism and as a useful tool in molecular biology laboratory. In particular
it is used as a cloning vector, for the shuffling of the cloned DNAs in the Gateway method
and in the method called "recombeenering".

#3.B Methods

Small drops (V<10 µl, R<1.35 mm) of D.I. water containing infinitesimal amounts of
analytes were gently posted upon the surfaces and the entire process of evaporation
followed over time. An automatic contact angle meter (KSV CAM 101, KSV INSTRUMENTS
LTD, Helsinki, Finland) was used at room temperature. Notice that the energy of adhesion γ
per unit area at the gas/water interface is ~72.8 mJ/m2 at 20 °C. The process enabled to
concentrate very tiny amounts of agents over micrometric areas. The evaporation processes
were performed in a clean room to reduce the presence of external contaminants and lasted
approximately 20 min. The residual solute was observed using scanning electron microscope
(SEM), fluorescent microscopy and Raman spectroscopy techniques.

Protocol for Lambda DNA

Lambda DNA (Fermentas. Genbank/EMBL accession number J02459) comes with a
concentration of 0.3 ug/ul (10 nM). A 1:100 dilution of the stock solution is necessary to
make DNA combing on microscopy glass to clearly distinguish one molecule from the other.
More concentrated solutions yield a “green carpet” of DNA covering the entire surface.
Stained Lambda. For the single binding event experiments, 10 ul of Lambda (dilution at
10-17 M) were stained with 10 ul of a 0.1 uM freshly prepared YOYO solution (in PBS) (see
protocol Invitrogen MP03600) for 1 hour at room T in the dark. (For the multiple binding

18 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information
event experiments, as in supplementary information #4, starting concentration of 10-15 M
was used). The solution was taken with the suitable syringe and put on the super-
hydrophobic pillars. The drop volume was about 2-3 ul. The drop volume is variable
depending on the pillars structure. The drop was let de-wet in the dark for 20 min.
Unstained Lambda. Following the same protocol as before, a drop of Lambda at same
dilution was put on the pillars without staining. Background control. PBS solution. PBS YOYO
solution. The sample was gold sputtered at 40 mA for 20” before SEM imaging.

Fluorescence microscopy Characterization

Fluorescence microscopy measurements were performed using an inverted microscope, with
infinity-corrected optics (Nikon-ECLIPSE TE 2000-U). The microscope objectives used were
a Plan Fluor 40x and a Plan Apo 60x, with 0.75 and 0.95 numerical apertures. For probe
excitation a violet diode laser source emitting at 408 nm, an Argon source emitting at
488nm, Helium Neon laser source emitting at 543 nm were used. A Nikon D - Eclipse C1
scanning head with three channels was utilized for the measurements.

Raman Characterization of Rhodamine deposits

Micro-probed Raman spectra were obtained using Renishaw inVia Raman microscope at
room temperature through 20x objective of a Leica microscope. The Raman spectra were
excited by the 514.0 nm line of an Ar+ laser in backscattering geometry. The laser power
was 0.18 mW with an integration time of 50 s. Mapping Raman measurements were carried
out with the step size 19.50 µm and 34.60 µm in x and y-axis direction, respectively.

Raman Characterization of Lysozyme on cone

Lysozyme 10-15 M solution was deposited through a glass micropipette (World Precision
Instruments) on top of the superhydrophobic device. The micropipette has a tip with an
outer diameter of 1µm and it is driven by means of micrometer screws (microinjector from
Eppendorf GmbH), thus allowing an accurate positioning over the plasmonic antenna. The
whole operation was carried out under an optical microscope in order to have visual
inspection of the deposition process. A microinjector with resolution in the 10 femtoliter
regime was used to release the drops of solution on the super hydrophobic chip.
Evaporation of the drop occurred in reduced time lapse (that is, few seconds) because of
the small liquid volume. After evaporation, Raman measurements were performed over the
plasmonic antenna, and also in the surrounding areas (both on top and at the bottom of the
pillars) as control measurements. For all Raman spectra the laser light had wavelength of
514nm with 100 X objective, the power at the sample level was about 0.1 mW, and the
accumulation time was 50 s. Measurements on the plasmonic antenna were recorded by
focusing the laser light on the tip of the cone. Several measurements were also recorded
along the cone axis at different heights below the tip.

1. Revenis ME, Kaliner MA, August 1992. "Lactoferrin and lysozyme deficiency in airway secretions: association
with the development of bronchopulmonary dysplasia". J. Pediatr. 121 (2): 262–70.
2. Lönnerdal B, June 2003. "Nutritional and physiologic significance of human milk proteins". Am. J. Clin. Nutr. 77
(6): 1537S–1543S.
3. JC Kendrew, G Bodo, HM Dintzis, RG Parrish, H Wyckoff, and DC Phillips, 1958. "A Three-Dimensional Model of
the Myoglobin Molecule Obtained by X-Ray Analysis". Nature 181 (4610): 662–666.

nature photonics | www.nature.com/naturephotonics 19

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

SUPPLEMENTARY INFORMATION #4: λ-DNA DEPOSITION AT DIFFERENT DILUTION

#4.A single binding event detection

The experiments heretofore described would prove that superhydrophobic devices deliver
the ability to concentrate solutions with extremely high efficiency. Nevertheless, the
recognition of one single binding event would ultimately establish the novelty and
adequacy/efficiency of these devices. In this context, lambda DNA solutions were analysed.
When the concentration is sufficiently low (10-18 M), single molecules may be detected:
single double stranded Lambda DNA molecule (48502 bp) keeps stretched on top of two
pillars. Fig.4 in the main article, and a collection of SEM images reported in Fig.S4.1 show
this.

Fig.S4.1

#4.B multiple binding event detection

When the initial concentration of the lambda DNA solution gets higher, more filaments are
likely to be detected upon evaporation. Fig. S4.2 shows two DNA filaments connecting 3
pillars as to form a row. For the present configuration the initial concentration is 10-17 M.
Increasing the concentration (10-16 M) yields an arrangement as in Fig.S5.3. Here, multiple
filaments thread the pillars over a large area as to form a dense λ-DNA network (Fig.45.3).

Fig.S4.2

20 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

Fig.S4.3

In some cases, intricate network of DNA complexes are created: DNA filaments combine
and overlap yielding multifaceted complexes where different diameters are observed. This is
evidenced in Fig.S4.4, where three or more filaments cross link simultaneously in the same
point (concentration 1fM).

Fig.S4.4

nature photonics | www.nature.com/naturephotonics 21

© 2011 Macmillan Publishers Limited. All rights reserved.
supplementary information doi: 10.1038/nphoton.2011.222

Supplementary information #5: FDTD simulation

Three-dimensional Finite difference time domain (FDTD) calculations were performed to

optimize surface plasmons generation, propagation and accumulation at the tip end.
Simulations were carried out employing commercial software specifically developed to FDTD
calculation of photonic structure (www.lumerical.com). A sketch of devices cross section is
reported in the figure below, where the main geometrical parameters are also indicated. As
showed, the overall simulation volume was divided in two regions in order to save
simulation time, required memory, and to improve accuracy. The bottom region
comprehends the body of the cone and the metallic grating, and it takes the most part of
the simulation volume (3x3x8 µm3, minimum mesh size 4 nm). The top region comprehends
the final part of the nanocone, and being much smaller allows a fine mesh refinement
(volume 0.5x0.5x0.2 µm3, minimum mesh size 0.5 nm). First, a simulation was run in the
bottom region to calculate the generation of SPPs through the coupling between the
incoming laser and the metallic grating. SPPs travelling toward the tip were stored in a
“monitor” placed at the boundary of the two regions. The stored electric field was then used
as a source to run a second simulation involving the top region. When the radius of
curvature of the tip is 10 nm, the calculated electric field enhancement is in the order of 30
with the respect of the incident laser amplitude.
Simulation details.
Source: 532 nm laser, continuous pulse, focused on the metallic grating through a 0.9
Numerical Aperture objective.
Simulation time 100 fs (time step 0.01 fs).
Simulation volume: bottom region 3x3x8 µm3, minimum mesh size 4 nm; top region
0.5x0.5x0.2 µm3, minimum mesh size 0.5 nm
Index of refraction of the materials: Palik database.
Grating details: pitch=480 nm; depth= 200 nm; order=-1; Ag= 40 nm thick.

22 nature photonics | www.nature.com/naturephotonics

© 2011 Macmillan Publishers Limited. All rights reserved.
doi: 10.1038/nphoton.2011.222
supplementary information

nature photonics | www.nature.com/naturephotonics 23

© 2011 Macmillan Publishers Limited. All rights reserved.