Content uploaded by Lothar Wondraczek
Author content
All content in this area was uploaded by Lothar Wondraczek on Dec 10, 2015
Content may be subject to copyright.
1
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
www.nature.com/scientificreports
Quantitative image analysis for
evaluating the abrasion resistance
of nanoporous silica lms on glass
Karsten H. Nielsen1, Stefan Karlsson1,2, Rene Limbach1 & Lothar Wondraczek1,3
The abrasion resistance of coated glass surfaces is an important parameter for judging lifetime
performance, but practical testing procedures remain overly simplistic and do often not allow for
direct conclusions on real-world degradation. Here, we combine quantitative two-dimensional
image analysis and mechanical abrasion into a facile tool for probing the abrasion resistance of
anti-reective (AR) coatings. We determine variations in the average coated area, during and after
controlled abrasion. Through comparison with other experimental techniques, we show that this
method provides a practical, rapid and versatile tool for the evaluation of the abrasion resistance
of sol-gel-derived thin lms on glass. The method yields informative data, which correlates with
measurements of diuse reectance and is further supported by qualitative investigations through
scanning electron microscopy. In particular, the method directly addresses degradation of coating
performance, i.e., the gradual areal loss of antireective functionality. As an exemplary subject, we
studied the abrasion resistance of state-of-the-art nanoporous SiO2 thin lms which were derived
from 5–6 wt% aqueous solutions of potassium silicates, or from colloidal suspensions of SiO2
nanoparticles. It is shown how abrasion resistance is governed by coating density and lm adhesion,
dening the trade-o between optimal AR performance and acceptable mechanical performance.
in lms are omnipresent on at glass substrates for applications in almost any area of daily life. Most
prominently, they are used to improve or impose specic functionality on glass sheet in architecture,
automotive engineering and solar energy harvesting, for example, to generate wavelength-specic reec-
tivity or anti-reectivity (AR), electrical conductivity, photocatalytic activity and/or self-cleaning abil-
ity1. To increase light transmission in photovoltaic modules, the cover glasses are oen equipped with
nanoporous SiO2 AR coatings, which can be synthesized by vapor deposition processes2 or by sol-gel
routes from potassium silicates3, silica nanoparticles4 or silicon alkoxides, such as tetraethyl orthosili-
cate (TEOS)5,6. For photovoltaic cover glasses in outdoor applications which are designed for a lifetime
of up to 40 years7, chemical and mechanical stability are crucial arguments and have therefore been
addressed continuously, with strategies ranging from increasing the precursor reactivity5, controlling the
gel morphology6,8, partial crystallization9, enhancement of interfacial reactions for improving lm adhe-
sion10 to post-deposition treatments with gases or solutions4,10–12. Accurately judging and quantifying the
mechanical stability (usually expressed as the abrasion resistance) of such coatings, however, remains a
standing issue. In particular, today’s typical evaluation procedures, which either rely on instrumented
indentation or on more simplistic but standardized macroscopic tests, do not provide direct insights into
the degradation of actual performance (which is, for AR coatings, the ability to improve optical transpar-
ency over a certain area of glass sheet). e former approaches for evaluating the mechanical properties
of thin lms on glass involve advanced and precise but very local testing through nanoscratching or
nanoindentation13–17. e latter, macroscopic techniques range from wipe tests12,18,19, pencil hardness
1Otto Schott Institute of Materials Research, University of Jena, Fraunhoferstrasse 6, D-07743 Jena, Germany.
2Glafo–the Glass Research Institute, PG Vejdes väg 15, SE-351 96 Växjö, Sweden. 3Center of Energy and
Environmental Chemistry (CEEC), University of Jena, Max-Wien-Platz 1, D-07743 Jena, Germany. Correspondence
and requests for materials should be addressed to L.W. (email: lothar.wondraczek@uni-jena.de)
received: 26 August 2015
Accepted: 04 November 2015
Published: 10 December 2015
OPEN
www.nature.com/scientificreports/
2
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
tests9,20 or adhesive tape tests10,21 to controlled abrasion such as done with the widely-employed rotating
abrader (Taber® Abraser), or with linear abraders5. A major objective of these tools is to mimic, in one
way or the other, real-world contact damage. Laboratory abrasion is then typically followed by optical
inspection, which usually addresses local defects and aims to determine a critical threshold load or
parameter, which causes visible scratching of the coating9 or coating detachment22. A rotating abrader is
schematically shown in Fig.1. It comprises rotating abrasive wheels which are equipped with a selectable
surface material for mimicking specic contact situations. Further adjustable parameters are the contact
load, the rotation speed, and the number of abrasion cycles. is setup complies with several stand-
ards for transparent materials and glazed materials23,24. For example, it is applied regularly for empiri-
cally testing of coated glass sheet and the mechanical resistance of sol-gel5,25,26 as well as physical vapor
deposition (PVD)-derived coatings27. e abrasive wheels typically consist of either felt5 or a polymer
matrix with embedded corundum (Al2O3)24. Aer abrasion, the response of the coated surface is oen
judged through mass loss26, or by recording optical parameters such as the direct or diuse transmis-
sion8,27, the solar-weighted photon spectrum5 or haze25. e mechanical test is then oen followed by
microscopic investigations for a qualitative evaluation of the underlying abrasion process22. However,
any such approach remains largely phenomenological, and the potential for drawing quantitative con-
clusions is limited. In particular, the comparison of dierent materials and dierent abrasion situations
is complicated by the present inability to directly provide a facile numerical evaluation of areal surface
degradation.
In the present study, we target this issue through combining quantitative image analysis with standard
abrasion tests. We show that using mature tools of computational image analysis on scanned micro-
graphs enables rapid determination of areal variations in the optical properties of the coated substrate
area aer abrasion. e sensitivity of this approach is limited only by the quality of the image collection,
enabling to resolve even small variations that occur between comparably similar coatings, or between
small steps of mild abrasion. For this, we consider two kinds of SiO2 sol-gel coatings as model systems:
lms based on colloidal SiO2 with adjustable colloid size and comparably weak adhesion28, and more
durable lms which are derived from the deposition of aqueous potassium silicate solutions3,29 with
varying lm density.
Materials and Methods
Sample synthesis. Nanoparticle-derived layers were produced from commercial, alkaline suspen-
sions of colloidal SiO2 (Köstrosol® , CWK Bad Köstritz, Bad Köstritz, Germany). Dierent suspensions
with particle diameters of 7, 20, 35 and 45 nm, respectively, were used. e suspensions were adjusted
to pH = 7 with 0.1 M HCl, subsequently diluted to a solid fraction of 6 wt%, and immediately applied
to 2 mm low-iron oat glass (Tg = 566 °C) by dip-coating (RDC 15, Bungard Elektronik, Windeck,
Germany) to generate homogeneous coatings such as previously described by Cook28. For the solu-
tion-derived lms, aqueous potassium silicates (BASF, Düsseldorf, Germany) with a nominal silicate
concentration of ~6 wt% were employed (for details, see Refs. 3,29). Here as well, coatings were depos-
ited by dipping under ambient atmosphere, followed by subsequent washing in demineralized water so
as to remove residual alkaline carbonates3. Aer drying, both types of samples were thermally annealed
for 20 minutes at 500 °C on graphite plates in a mue furnace (LM 312, Linn High erm GmbH,
Eschenfelden, Germany) with the tin-side of the glass substrate facing upwards. Both types of precursor
Figure 1. Principle of the test method. in lm coated glass is abraded with the Taber® abraser (a)
above the critical load to cause delamination. An optical microscope is used to collect images from the
middle of the abraded surface area (b). For anti-reective coatings, the coating and substrate can easily be
distinguished. Finally image analysis (c) is used to determine the average coated area. is procedure is
repeated several times for each sample.
www.nature.com/scientificreports/
3
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
yield thin SiO2 lms with AR functionality. Film thickness d and refractive index np of the coatings were
estimated through ellipsometric analysis (EP3, Accurion GmbH, Göttingen Germany). For that, meas-
urements were performed at several angles and a xed wavelength of λ = 514 nm. Data on these coatings
are summarized in Table 1.
Microindentation. An empirical indication of the mechanical stability of thin lms can be derived
from the probability of radial crack formation aer Vickers indentation, the so-called crack propensity
index (CPI)30,31. is comprises two separate eects: (i) the eect of reduced contact-stress during inden-
tation through a coating which is so, when compared with the substrate, and (ii) the eect of tensile
(or compressive) Eigen-stress in the coating as generated through strain dierences which occur relative
to the substrate upon coating consolidation. Both eects lead to variations in CPI, which are caused by
the coating but occur in the substrate. Vickers indents (Duramin-1, Struers, Ballerup, Denmark) were
made on the coated tin-side of the glasses, and the number of radial cracks formed at the corners of the
residual Vickers hardness imprints was determined by optical microscopy 20 seconds aer indentation.
e measurements were conducted under ambient atmosphere at 5 dierent loads between 50 and 500 g
and were repeated 20 times for each load.
Abrasion analysis. e coated glass samples were prepared for abrasion testing by washing under tap
and distilled water and subsequent drying with pressurized air. e prepared samples were kept upright
in closed plastic boxes before the abrasion experiment in order to ensure uniform conditioning of the
surface and to protect them from dust. All abrasion tests were conducted on the coated tin-side of the
glass to minimize the inuence of any eventual glass substrate corrosion1.
Abrasion experiments were performed in a rotating-wheel abrader (Fig. 1a, Abraser 5135, Taber
Industries, North Tonawanda, USA). In the following, one complete 360° rotation of the sample is called
an abrasion cycle. Initially, a rotation speed of 60 rmp under no additional load with Cs-10f abrasive
wheels was chosen as reference condition. ese abrasion wheels consisted of Al2O3 particles in a poly-
mer matrix24, and represent a typical benchmark for thin-lm testing23,26,27. All particle-derived coatings
where abraded with this set-up. Following initial reference testing, more violent abrasion conditions
were employed for further testing of the potassium silicate-derived coatings, i.e., using a coarser Cs-10
abrasive wheel at 72 rpm and with an additional load of 750 g on each wheel. Aer a certain number of
cumulative abrasion cycles, samples were demounted, rinsed with ethanol and drying again with pres-
surized air. Digital images of the thus-abraded samples were taken with an optical microscope (Axiolab,
Zeiss, Oberkochen, Germany, 20× lens CP-Achromat, Zeiss, Oberkochen, Germany) in reectance
mode, equipped with a simple camera (Microcam 3 M, Bresser, Borken, Germany). With this micro-
scope, images were collected from three arbitrary positions on the abraded surface, while microscope
settings were optimized for maximum contrast in each picture (shown exemplarily in Fig. 1b). Aer
Potassium Silicate-Derived in Films
SiO2/K2O 4 5 6
d (nm) 110 ± 20 120 ± 20 110 ± 10
np1.46 ± 0.02 1.46 ± 0.02 1.43 ± 0.01
P (%) 0–5 0–5 5-10
Tincrease (%point) 3.4 3.9 3.9
H40 nm (GPa) 4.58 3.26 3.28
E40 nm (GPa) 55.6 40.3 36.2
CPI50 (N) 2.0 1.8 1.9
SiO2 Nanoparticle-Derived in Films
Particle size (nm) 7 20 35 45
d (nm) 93 ± 7 110 ± 12 70 ± 12 101 ± 5
np1.38 ± 0.02 1.36 ± 0.04 1.34 ± 0.04 1.36 ± 0.03
P (%) 20 25 30 25
Tincrease (%point) 5.2 ± 0.3 6.5 6.5 ± 0.2 6.8
H40 nm (GPa) 2.14 2.24 1.42 1.65
E4sm (GPa) 18.5 21.3 12.0 17.7
Table 1. Properties of the tested coatings on low iron oat glass. ickness (d), refractive index (np),
porosity (P) and maximum transmission increase (Tincrease) of a 2 mm low iron oat glass (T = 91.4%).
Hardness (H) and elastic modulus (E) as evaluated through nanoindentation. e force needed to obtain
50% crack propensity index (CPI50), this is 1.3 N for the uncoated glass substrate.
www.nature.com/scientificreports/
4
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
image recording, abrasion was continued as described above, repositioning the sample into the abrader.
Each test was repeated 2–3 times for any type of sample. Between dierent samples, the abrasive wheels
were refreshed according to standard prescriptions. Image analysis was conducted with the Axiovision
soware (Zeiss, Oberkochen, Germany) with the goal to determine the extent of areal abrasion as a func-
tion of the number of abrasion cycles (Fig.1c). ereby, it was assumed that image recording provides
direct access to the areal increase in surface reectivity (or loss in sample transmission) which follows
local removal of the AR layer upon abrasion. As a prerequisite, abrasion treatment was done so as to
not damage the underlying glass substrate, which would lead to an increase in diuse reection and
scattering, and would compromise the degree of information which can be extracted from the data for
performance of the coating alone.
Reference testing. For reference, further experiments were conducted to judge the abrasion resist-
ance by conventional means17, and to verify the sensitivity and selectivity of image analyses. ese exper-
iments involved collecting diuse reection (DR) spectroscopy (using a Cary 5000 spectrophotometer,
Agilent, Santa Clara, US, equipped with a 110 mm integration sphere and a goniometer for angle-sensitive
analyses) and instrumented indentation testing, using a Nanoindenter G200 (Agilent, Santa Clara, US,
equipped with a Berkovich diamond tip with a nominal tip radius of 50 nm). In order to increase the
instrumental resolution at very shallow indentation depths the indenter tip was calibrated based on its
equivalent contact radius32. Depth proles of the hardness H and elastic modulus E were obtained by
operating in the continuous stiness measurement mode (CSM)33. Values of H and E were extracted at
an indentation depth of ~40 nm, where substrate-independent values of E were achieved through the
method proposed by Hay and Crawford34.
Furthermore, the abrasion wear resistance was investigated using the same nanoindenter as described
above (but equipped with a conical diamond tip with a nominal tip radius of 5 μ m). In this test the
indenter tip was scratched multiple times across the sample surface along a distance of 200 μ m at a con-
stant load of 10 mN and a velocity of 50 μ m/s. Aer each ve consecutive wear cycles the topography of
the residual scratch groove was scanned with the indenter tip at a constant load of 50 μ N along as well
as perpendicular to the wear path.
Finally, a scanning electron microscope (SEM) was used to investigate selected samples aer abra-
sion (JSF7001F, Jeol Ltd., Tokyo, Japan). For this, samples were coated with a thin carbon layer prior to
micrograph collection (Auto 306, Edwards, Crawley, United Kingdom).
Results and Discussion
Coating microstructure. e microstructure of the employed coatings is taken as the predominant
factor for leveraging a broad variability of abrasion resistance. is is to judge the applicability and
selectivity of digital image analyses in the quantication of areal abrasion damage. e most versatile
design parameters are the coating porosity (which directly determines its eective refractive index) and
interfacial adhesion. en, the two employed types of coatings represent the extremes of low porosity
and strong interfacial bonding (solution-derived lms) and high porosity/weak bonding (lms which are
derived from colloidal suspensions). In the latter, interfacial adhesion is primarily governed by the degree
of thermal curing and sintering, which again results in lower porosity. Beyond these extreme variations,
ne variations are obtained within the individual classes of coatings through adjusting either the solution
concentration and, in particular, the ratio of SiO2/K2O, or the colloid size.
As summarized in Table 1, the applied synthesis procedures yield lms with thickness ranging from
70 to 120 nm, for both kinds of precursors. e potassium silicate-derived lms exhibit higher refractive
indices (nP = 1.43–1.46) than the lms derived from SiO2 nanoparticles (nP = 1.34–1.38), what indicates
lower lm density and, eventually, lower abrasion resistance. e porosity of the obtained lms was
estimated from equation(1):35
−
−
=−
()
n
n
P
1
1
1
100 1
p
2
2
where n is the refractive index of the constituting material (for SiO2, n = 1.46), np the eective
refractive index of the porous layer, and P the volume fraction of pores (%). e calculated poros-
ity of the potassium silicate-derived thin lms is comparably low, i.e., ~0–8%, while that of the SiO2
nanoparticle-derived lms is much higher, i.e., ~20–30%, within the geometrical optimum of a system of
close-packed, monodisperse balls (26%). Noteworthy, the absolute porosity value may be inuenced by
capillary condensation of water which occurs under ambient conditions29. en, the value of np is strictly
not a convolution of silica and air, but of silica and water with nH2O ~ 1.33. In this case, the actual pore
fraction would be higher than the here-employed estimate.
e results of the nanoindentation testing, shown in Fig. 2a and summarized in Table 1, conrm
the expectation, that the coatings based on potassium silicates in general have a higher hardness
(H = 3.28–4.58 GPa) and elastic modulus (E = 36.2–55.6 GPa) than the more porous coatings based on
SiO2 nanoparticles (H = 1.42–2.24 GPa; E = 12.0–21.3 GPa). e hardness values of the porous SiO2 nan-
oparticle coatings are in a range comparable to acid-catalysed TEOS-derived coatings with P = 35 and
www.nature.com/scientificreports/
5
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
37%, respectively, and tempered at 450 °C (H = 1.3 and 1.5 GPa)17. For the dierent types of coatings of
this study, however, the nanoindentation measurements do not enable a clear ranking of the mechanical
resistance, although the expected higher stability with lower SiO2/K2O ratio and lower particle size is
indicated by the results. at is, these experiments do not readily distinguish the properties of the present
coatings.
Using lateral-force control and/or measurement, instrumented nanoindentation enables an alterna-
tive method for testing the abrasion resistance of coatings, demonstrated exemplarily in Fig.2b. With
increasing number of wear cycles the indenter tip progressively penetrates deeper into the coating, if the
applied load is above the critical load for the coating to fail. However, such wear experiments, are very
time consuming, limited to relatively small observation length-scale and 1D-analyses, and still depend
on the availability of highly accurate models for data evaluation. Figure3 exemplarily shows the variation
of CPI versus indentation load for the uncoated substrate and glass coated with potassium silicates of
dierent SiO2/K2O molar ratios, that is SiO2/K2O = 4, 5 and 6, respectively. In general, for all coatings
which are derived from potassium silicates, the crack resistance, in terms of the force which is needed
to obtain a 50% probability (CPI50) for radial crack initiation in the substrate, is seen to increase from
1.3 N for the uncoated substrate to 1.8–2.0 N for the coated glasses, as summarized in Tab. 1. is is a
signicant improvement of surface defect resistance upon sharp contact loading, but does not necessarily
let expect a notable variation also in abrasion resistance (blunt loading and lateral damage). e general
phenomenon has been described earlier as a result of tensile stresses in the coatings, which, upon su-
ciently strong lm adhesion, act negatively on the crack-opening probability at the underlying substrate.
Beyond the scope of the present study, this can be tailored through precursor dilution (coating density),
lm thickness, thermal treatment and the initial state of the glass surface30. Besides the generation of
Figure 2. Nanoindentation and nano-wear experiments on the model coatings. (a) Hardness as a
function of indentation depth for coatings based on potassium silicate solutions with varying SiO2/K2O
molar ratios and solutions of SiO2 nanoparticles with dierent particle diameters, respectively. e inset
shows the elastic modulus as a function of indentation depth. (b) Average penetration depth as a function of
wear cycles for nano-wear experiments on a thin lm derived from a potassium silicate solution with SiO2/
K2O = 4. e inset illustrates the principle of the experiment.
Figure 3. Crack propensity index for uncoated and coated glass. Crack propensity index as a function of
indentation load for uncoated glass and potassium silicate coated glasses. Glasses are less prone to cracking
upon coating. e inset illustrates radial crack formation upon Vickers indentation.
www.nature.com/scientificreports/
6
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
Eigen-stresses, also the sealing of defects has been noted as another factor in the improvement of surface
damage resistance14,21,36. On the other hand, the data in Fig. 3 demonstrate that CPI does not provide
sucient selectivity to dierentiate between the eects of the dierent types of coatings (with varying
SiO2/K2O ratios), but only reveals a general reduction of surface defect sensitivity.
Abrasion testing. As shown in Fig. 4a–c, picture analysis enables a clear distinction between the
areas where the coating is removed (light) and the area with intact coating, which appears darker due
to the AR properties of the coating. e observation length scale (Fig. 4b,c) largely depends on the
applied imaging method and, in particular, its numerical aperture (NA). For rapid sample throughput
and intermediate optical resolution, commercial at-bed scanners can be employed with typically pro-
vide low NA. Shown here, however, are images taken with an optical microscope, as mentioned in the
experimental section.
Solution-derived coatings. e abrasion resistance of the solution-derived coatings was character-
ized by means of average coated area aer a given number of abrasion cycles as shown in Fig.5a. In con-
trast to the CPI analyses, image analyses enables a clear dierentiation of the coatings with signicantly
increasing abrasion resistance for decreasing SiO2/K2O ratio. As summarized in Table1, decreasing SiO2/
K2O ratio leads to decreasing lm porosity, probably due to the presence of more depolymerized precur-
sor anions3. Interestingly, for coatings with SiO2/K2O ≥ 5, we observe removal from the surface through
gradual delamination, whereas the coating with SiO2/K2O = 4 is suciently adherent to the surface so
that the coating itself is abraded. Complementary microscope investigations conrm this observation
(to be explained later, Fig.6), where this most dense coating remains visually unaected even aer 90
cycles of rough abrasion, and notable scratches and signs of delamination are seen only aer > 200 cycles.
e corresponding evolution of diuse reectance (DR), at a wavelength of 400 nm, is shown in
Fig.5b. As noted previously, measurement of DR is a common tool for evaluating the progress of abra-
sion on coated or uncoated glass surfaces. Here, with progressing abrasion, all samples initially exhibit
an increase in DR, which corresponds to the roughening of the surface. is eect is particularly pro-
nounced for the samples with SiO2/K2O ≥ 5, where abrasion is primarily governed by coating delamina-
tion. However, aer reaching a maximum in DR the DR starts to decrease and only when the coating is
virtually completely removed, the value of the bare (uncoated) glass is reached. For the more dense coat-
ing with SiO2/K2O = 4 and for the bare glass, a similar but less pronounced increase in DR is seen, and a
plateau is reached aer prolonged abrasion. Hence, when DR is taken as a measure of abrasion-induced
surface damage, the coating with SiO2/K2O = 4 apparently protects the surface against scratches. is is
in contrast to the information, which is provided by CPI analyses, where all three solution-derived coat-
ings were found to prevent the formation of radial cracks to approximately the same extent (Fig.3), at
least upon sharp contact loading. As intermediate conclusions, while analyses of DR may provide some
information on the mechanism of abrasion (delamination versus gradual material removal), it is not an
unambiguous way to judge the gradual and areal loss of AR functionality, in particular, as is obtained
by digital image analysis.
Figure 4. Image analysis of the abraded surface on dierent length scales. For each length scale, the
segmented picture, grey scale original and the corresponding grey scale histogram are given as a basis for
segmentation. (b,c) are obtained by cropping from (a), as marked on the insets. Sizes: (a) 588 × 441 μ m
(b) 163 × 122 μ m (c) 65 × 49 μ m.
www.nature.com/scientificreports/
7
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
Qualitative SEM analysis conrms this picture. in lm failure occurs in two principle ways, i.e.,
due to failure of lm adhesion or lm cohesion22, exemplarily illustrated in Fig.6a,b for solution-derived
samples with SiO2/K2O = 5 and 4, respectively. e SEM micrograph in Fig.6a was obtained from the
edge of the former aer 360 abrasion cycles, revealing lm delamination without characteristic scratches
on the intact parts of the remaining lm fraction. e removal of the coating from the surface clearly
Figure 5. Abrasion testing of potassium silicate-derived lms. (a) e average coated area as a function of
abrasion cycles. (b) Diuse reectance (DR) evaluated at 400 nm as a function of number of abrasion cycles.
e inset displays the development of DR for a coating with SiO2/K2O = 4 at higher numbers of abrasion
cycles.
Figure 6. Scanning electron microscopy of abraded surfaces. e micrographs visualize the inuence of
the SiO2/K2O molar ratio on the wear behavior of potassium silicate coatings. (a) in lm (SiO2/K2O = 5)
aer 360 abrasion cycles. e arrows exemplarily mark the regions where the coating is peeled o. (b) in
lm (SiO2/K2O = 4) aer 1360 abrasion cycles. Remaining pieces of coating, marked by arrows, indicate
higher interfacial strength.
www.nature.com/scientificreports/
8
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
point to adhesion failure, where the cracked or deformed edges indicate wedge spallation22. For com-
parison, the latter, more-dense coating (Fig.6b) also starts to delaminate aer extended abrasion (1360
cycles), with an increase in DR aer 180 cycles (shown in the inset, Fig. 5b). However, pronounced
individual scratches are also evident on the parts of coating which did not delaminate. ese scratches
are the reason for the comparably weak increase in DR in the early abrasion stages, as shown in Fig.5b.
e observations conrm that the latter lm abrades through cohesive failure, and that adhesive failure
comes into play only in the later stages of abrasion. However, also aer prolonged abrasion, even very
small islands of coating remain adherent to the surface (Fig.6b, indicated with arrows).
Colloid-derived coatings. As shown in the exemplary study of colloid-derived coatings, digital
image analysis can also be employed to lms with comparably low mechanical resistance, where other
techniques, such as nanoindentation, fail to yield selective data (Fig.2). Figure7a clearly shows dieren-
tiation between as-deposited and thermally set layers, exemplarily for a coating, which was derived from
a suspension of SiO2 spheres with a diameter of 7 nm. As expected, the thermal treatment increases the
abrasion resistance of the coating, which agrees with previous observations on similar types of coatings6.
In Fig.7b, the average coated area aer 15 abrasion cycles is shown as a function of the particle size,
which was employed in the precursor suspensions, aer thermal treatment. Here, the average coated area
decrease, i.e., the abrasion resistance decreases, with increasing particle size, and a clear dierentiation is
possible between all samples. e type of precursor particle does not only inuence abrasion resistance,
but also coating morphology as illustrated in the insets in Fig.4b. e SEM micrographs show coatings
derived from SiO2 particles with diameters of 35 and 45 nm, respectively. A complete sintering of SiO2
nanoparticles is size-dependent under ~40 nm and is not expected for these coatings due to the short
duration and relatively low applied temperatures37. However, the increased abrasion resistance upon
thermal treatment (Fig.4a) indicates that solidication takes place up to a certain extend. e observed
increase in abrasion resistance with decreasing particle size may thus derive from size size-dependent
reactivity37, but decreased roughness must also be considered, as suggested for PVD-derived thin lms13.
Despite sintering and similar properties in terms of thickness and porosity, abrasion resistance for
both kinds of thin lms is greatly inuenced by the employed starting solutions. Similarly, coatings
derived from methyl trimethoxy silane25 were found to show a decrease in their abrasion resistance with
Figure 7. Investigations on SiO2 colloidal coatings. (a) e average coated area as a function of abrasion
cycles for 7 nm SiO2 nanoparticle coatings. e line is a guide for the eye. (b) Coated area as a function of
particle size for SiO2 nanoparticle coatings evaluated aer 15 abrasion cycles. e line is a guide for the eye.
e insets show SEM micrographs of two coatings, prepared from 35 nm and 45 nm particles, respectively.
For both (a,b), two or more data points corresponds to two or more repeats of the same abrasion
experiment.
www.nature.com/scientificreports/
9
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
increasing degree of polymerization and those ndings correlate quite well with both types of thin lms
investigated in the present study. For TEOS-derived coatings it was noticed, that sols with polymeric
chain structures give more stable coatings than particles, due to a better contact between gel-forming
particles6. is explanation may also be valid for the two investigated thin lm types in this study. For
potassium silicate coatings, coating adhesion increases with increasing potassium content. is may cor-
relate with potassium ions being enriched in the interface between coating and substrate3, where they
could enhance an interfacial reaction by diusion38 or by reducing the melting point of the SiO2 gel10
and thereby increase the adhesion of the coating to the glass.
In general, the two tested types of thin lms are very dierent in terms of their mechanical stability,
but due to the convolution of physical reactions which form the phenomenological behavior of abrasion
resistance, this dierence can oen not readily be assessed, even by in-depth studies through, e.g., instru-
mented indentation. As shown above, simple image analyses might sometimes be much more suitable,
where samples in each of the studied series could be clearly be distinguished in terms of their abrasion
resistance.
Conclusions
Controlled abrasion followed by image analysis for determination of the average coated area is suggested
as a fast and practical tool for evaluating thin lm abrasion resistance. is evaluation method provides
informative and highly selective data which correlates with measurements of diuse reection and qual-
itative evaluation by scanning electron microscopy. e method was applied on two exemplary types
of SiO2 coatings derived from two dierent sol-gel precursors, for which even slight dierences in the
coating chemistry and morphology could be traced successfully by the present high-throughput method.
at is, abrasion resistance of the potassium silicate thin lms was found to increase with decreasing
SiO2/K2O molar ratio due to better coating adhesion, and that of SiO2 particle-derived coatings increases
upon thermal treatment, and also with smaller particle size.
References
1. Gläser, H. J. Dünnlmtechnologie au Flachglass. (Verlag arl Hofmann, Schorndorf, 1999).
2. Nagel, H., Metz, A. & Hezel, . Porous SiO2 lms prepared by remote plasma-enhanced chemical vapour deposition - a novel
antireection coating technology for photovoltaic modules. Solar Energy Materials and Solar Cells 65, 71–77 (2001).
3. Nielsen, . H. et al. Large area, low cost anti-reective coating for solar glasses. Solar Energy Materials and Solar Cells 128,
283–288, doi: http://dx.doi.org/10.1016/j.solmat.2014.05.034 (2014).
4. Cathro, ., Constable, D. & Solaga, T. Silica low-reection coatings for collector covers, by a dip-coating process. Solar Energy
32, 573–579, doi: http://dx.doi.org/10.1016/0038-092X(84)90131-2 (1984).
5. Pop, S. C. et al. in 40th Photovoltaic Specialist Conference (PVSC), Denver (CO), IEEE, 2715-2719. (doi: http://dx.doi.org/ 10.1109/
PVSC.2014.6925490)(8th-14th, June, 2014).
6. Bautista, M. C. & Morales, A. Silica antireective lms on glass produced by the sol–gel method. Solar Energy Materials and
Solar Cells 80, 217–225, doi: http://dx.doi.org/10.1016/j.solmat.2003.06.004 (2003).
7. Deubener, J., Helsch, G., Moiseev, A. & Bornhö, H. Glasses for solar energy conversion systems. Journal of the European
Ceramic Society 29, 1203–1210 (2009).
8. Ye, H. et al. Preparation of antireective coatings with high transmittance and enhanced abrasion-resistance by a base/acid two-
step catalyzed sol-gel process. Solar Energy Materials and Solar Cells 95, 2347–2351 (2011).
9. Chen, C.-H., Li, S.-Y., Chiang, A. S. T., Wu, A. T. & Sun, Y. S. Scratch-resistant zeolite anti-reective coating on glass for solar
applications. Solar Energy Materials and Solar Cells 95, 1694–1700, doi: http://dx.doi.org/10.1016/j.solmat.2011.01.032 (2011).
10. Helsch, G., rzyza, M., Heide, G. & Frischat, G. H. Adherent antireection coatings on borosilicate glass for solar collectors.
Glass Technology - European Journal of Glass Science and Technology Part A 47, 153–156 (2006).
11. Belleville, P. F. & Floch, H. G. Ammonia hardening of porous silica antireective coatings. Proc. SPIE 2288, 25–32 (1994).
12. Cai, C. et al. obust anti-reective silica nanocoatings: abrasion resistance enhanced via capillary condensation of APTES.
Journal of Materials Chemistry C 3, 4254–4259, doi: 10.1039/C5TC00357A (2015).
13. Bull, S. J. Size eects in the mechanical response of nanoscale multilayer coatings on glass. in Solid Films 571, Part 2, 290–295,
doi: http://dx.doi.org/10.1016/j.tsf.2014.04.014 (2014).
14. Meszaros, ., Wild, M., Merle, B. & Wondracze, L. Flexural strength of PVD coated oat glass for architectural applications.
Glass Technology - European Journal of Glass Science and Technology Part A 52, 190–196 (2011).
15. Meszaros, . et al. Eect of thermal annealing on the mechanical properties of low-emissivity physical vapor deposited multilayer-
coatings for architectural applications. in Solid Films 520, 7130–7135, doi: http://dx.doi.org/10.1016/j.tsf.2012.07.086 (2012).
16. Meszaros, ., Wild, M. & Wondracze, L. Eects of substrate and long term corrosion on PVD-multilayer coatings for
architectural glazing. Glass Technology - European Journal of Glass Science and Technology Part A 54, 177–184 (2013).
17. Chemin, N., lotz, M., ouessac, V., Ayral, A. & Barthel, E. Mechanical properties of mesoporous silica thin lms: Eect of the
surfactant removal processes. in Solid Films 495, 210–213, doi: http://dx.doi.org/10.1016/j.tsf.2005.08.260 (2006).
18. Floch, H. G. & Belleville, P. F. A scratch-resistant single-layer antireective coating by a low temperature sol-gel route. Journal
of Sol-Gel Science and Technology 1, 293–304, doi: 10.1007/BF00486172 (1994).
19. Aegerter, M. A. et al. Coatings made by sol–gel and chemical nanotechnology. Journal of Sol-Gel Science and Technology 47,
203–236, doi: 10.1007/s10971-008-1761-9 (2008).
20. Xu, Y., Peng, C., Xin, C. & Wu, J. Preparation of silica antireective lms for solar energy application. Materials Letters 94, 89–91,
doi: http://dx.doi.org/10.1016/j.matlet.2012.12.013 (2013).
21. Ershad-Langroudi, A., Mai, C., Vigier, G. & Vassoille, . Hydrophobic hybrid inorganic-organic thin lm prepared by sol-gel
process for glass protection and strengthening applications. Journal of Applied Polymer Science 65, 2387–2393 (1997).
22. Bull, S. J. Failure mode maps in the thin lm scratch adhesion test. Tribology International 30, 491–498 (1997).
23. DIN Standard 52347, Testing of glass and plastics; abrasion test; method using abrasion wheels and measurement of scattered
light, Berlin: Deutsche Institut für Normierung e.V., (1987).
24. Sun, J., Muamal, H., Liu, Z. & Shen, W. Analysis of the Taber Test in Characterization of Automotive Side Windows. Tribology
Letters 13, 49–54, doi: 10.1023/A:1016536500591 (2002).
www.nature.com/scientificreports/
10
Scientific RepoRts | 5:17708 | DOI: 10.1038/srep17708
25. ondow, T. & Ninomiya, . Application of ZnO lms to glass substrates by the dipping-pyrolysis method using organic acid salt.
Proc. SPIE 1758, 150–159 (1992).
26. Cairns, D. ., essman, A. J., ichter, P. J., Bottari, F. J. & andall, N. X. Mechanical and tribological investigations of sol–gel
derived SiO2 optical coatings. Wea r 265, 411–416, doi: http://dx.doi.org/10.1016/j.wear.2007.11.028 (2008).
27. Suzui, S. & Ando, E. Abrasion of thin lms deposited onto glass by the Taber test. in Solid Films 340, 194–200 (1999).
28. Coo, . T., Tettey, . E., Bunch, . M., Lee, D. & Nolte, A. J. One-Step Index-Tunable Antireection Coatings from Aggregated
Silica Nanoparticles. ACS Applied Materials & Interfaces 4, 6426–6431, doi: 10.1021/am3020586 (2012).
29. Nielsen, . H., ittel, T., Wondracze, . & Wondracze, L. Optical breathing of nano-porous antireective coatings through
adsorption and desorption of water. Sci. ep. 4, doi: 10.1038/srep06595 (2014).
30. Maschio, . D., Pegoretti, A., izzo, C., Soraru, G. D. & Carturan, G. Vicers Crac Nucleation of Glass Sheets Coated by in
Silica Gel Layers. Journal of the American Ceramic Society 72, 2388–2390, doi: 10.1111/j.1151-2916.1989.tb06098.x (1989).
31. Sorarù, G. D., Guglielmi, M. & Dal Maschio, . Inuence of sol-gel coatings on crac initiation by vicers indentation in soda-
lime glass. Journal of Non-Crystalline Solids 100, 440–446, doi: http://dx.doi.org/10.1016/0022-3093(88)90061-0 (1988).
32. Tadjiev, D. ., Hand, . J. & Hayes, S. A. Calibrating a nanoindenter for very shallow depth indentation using equivalent contact
radius. Philosophical Magazine 90, 1819–1832, doi: 10.1080/14786430903571420 (2010).
33. Limbach, ., odrigues, B. P. & Wondracze, L. Strain-rate sensitivity of glasses. Journal of Non-Crystalline Solids 404, 124–134,
doi: http://dx.doi.org/10.1016/j.jnoncrysol.2014.08.023 (2014).
34. Hay, J. & Crawford, B. Measuring substrate-independent modulus of thin lms. Journal of Materials esearch 26, 727–738, doi:
10.1557/jmr.2011.8 (2011).
35. Yoldas, B. E. Investigations of porous oxides as an antireective coating for glass surfaces. Appl. Opt. 19, 1425–1429 (1980).
36. Wondracze, L. et al. Towards Ultrastrong Glasses. Advanced Mater ials 23, 4578–4586, doi: 10.1002/adma.201102795 (2011).
37. Buscarino, G., Ardizzone, V., Vaccaro, G., Agnello, S. & Gelardi, F. M. Atomic force microscopy and aman investigation on the
sintering process of amorphous SiO2 nanoparticles. Journal of Applied Physics 108, 074314, doi: http://dx.doi.org/10.1063/1.3481670
(2010).
38. Homann, U., Heide, G. & Frischat, G. H. eactions between sol-gel coatings and some technical glass substrates during
consolidation. Journal of Non-Crystalline Solids 351, 3562–3569 (2005).
Acknowledgements
We gratefully acknowledge Linde (Munich, Germany), BASF (Düsseldorf, Germany) and CWK Bad
Köstritz (Bad Köstritz, Germany) for providing substrate and coating materials. We further thank
our colleagues G. Gao, D. K. Orzol, T. Kittel, S. Zankovych and M. Arras, for experimental assistance
and valuable discussions. KHN gratefully acknowledges the uringian State Ministry of Science and
Education for nancial support through its graduate fellowship program.
Author Contributions
K.H.N. and L.W. perceived of the experiments and designed model coatings. K.H.N. synthesized the
coatings. K.H.N. and R.L. undertook the analysis. K.H.N., R.L., S.K. and L.W. interpreted the results and
wrote the manuscript. All authors were involved in the discussions and manuscript revisions.
Additional Information
Competing nancial interests: e authors declare no competing nancial interests.
How to cite this article: Nielsen, K. H. et al. Quantitative image analysis for evaluating the abrasion
resistance of nanoporous silica lms on glass. Sci. Rep. 5, 17708; doi: 10.1038/srep17708 (2015).
is work is licensed under a Creative Commons Attribution 4.0 International License. e
images or other third party material in this article are included in the article’s Creative Com-
mons license, unless indicated otherwise in the credit line; if the material is not included under the
Creative Commons license, users will need to obtain permission from the license holder to reproduce
the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
