You are on page 1of 12

Applied Surface Science 280 (2013) 578–589

Contents lists available at SciVerse ScienceDirect

Applied Surface Science


journal homepage: www.elsevier.com/locate/apsusc

Physical vapor deposited titanium thin films for biomedical


applications: Reproducibility of nanoscale surface roughness and
microbial adhesion properties
Claudia Lüdecke a,b,d , Jörg Bossert a,d , Martin Roth b,d , Klaus D. Jandt a,c,d,∗
a
Chair of Materials Science (CMS), Faculty of Physics and Astronomy, Friedrich Schiller University Jena, Löbdergraben 32, D-07743, Jena, Germany
b
Leibniz Institute for Natural Product Research and Infection Biology, Hans Knöll Institute (HKI), Bio Pilot Plant, Beutenbergstraße 11a, D-07745, Jena,
Germany
c
Jena Center for Soft Mater, Friedrich Schiller University Jena, Jena, Germany
d
Jena School for Microbial Communication (JSMC), Friedrich Schiller University Jena, Jena, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The surface topography is of great importance for the biological performance of titanium based implants
Received 24 July 2012 since it may influence the initial adsorption of proteins, cell response, as well as microbial adhesion.
Received in revised form 6 May 2013 A recently described technique for the preparation of titanium thin films with an adjustable surface
Accepted 10 May 2013
roughness on the nanometer scale is the physical vapor deposition (PVD). The aims of this study were
Available online 17 May 2013
to statistically evaluate the reproducibility of nanorough titanium thin films prepared by PVD using an
atomic force microscopy (AFM) based approach, to test the microbial adhesion in dependence of the
Keywords:
nanoscale surface roughness and to critically discuss the parameters used for the characterization of the
PVD
Titanium thin films
titanium surfaces with respect to AFM microscope settings. No statistically significant differences were
Nanoroughness found between the surface nanoroughnesses of the PVD prepared titanium thin films. With increasing
Escherichia coli surface nanoroughness, the coverage by Escherichia coli decreased and the microbial cells were increas-
AFM scan size ingly patchy distributed. The calculated roughness values significantly increased with increasing AFM
Biomaterials scan size, while image resolution and pixel density had no influence on this effect. Our study shows that
PVD is a suitable tool to reproducibly prepare titanium thin films with a well-defined surface topography
on the nanometer scale. These surfaces are, thus, a suitable 2D model system for studies addressing the
interaction between surface nanoroughness and the biological system. First results show that surface
roughness even on the very low nanometer scale has an influence on bacterial adhesion behavior. These
findings give new momentum to biomaterials research and will support the development of biomaterials
surfaces with anti-infectious surface properties.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction implant surface roughness strongly influences the initial protein


adsorption [4], the cell response [5], the tissue adhesion [6] as well
There is a wide range of biomaterials that can be used as implant as the adhesion of microorganisms [7].
materials. Some of the commonly used materials for implants are Several techniques, such as grinding and polishing, blasting,
metals, such as titanium or titanium alloys [1]. Titanium is a bio- plasma spraying or chemical etching can be used for the prepa-
compatible material, largely due to its inherently stable surface ration of titanium surfaces with different roughnesses [8,9]. Most
oxide layer. It is extensively used in dental, orthopedic and cardio- of these methods may be accompanied by not only modifying the
vascular fields in the form of dental implants, hip-joint replacement surface roughness, but also applying scratches or superordinate
devices or heart-valves and cardiac pacemakers [2]. For the bio- structures and, particularly, changing the surface chemistry [10].
logical performance of titanium based biomaterials, the texture For the investigation of the influence of a variation of the bio-
and roughness of their surfaces is of great importance [3]. The materials surface roughness on the biological environment, these
preparation techniques might, thus, be disadvantageous.
Previously, we demonstrated that physical vapor deposition
(PVD) is a suitable and straight forward technique to prepare tita-
∗ Corresponding author at: Chair of Materials Science (CMS), Faculty of Physics
nium samples with a surface roughness on the nanometer scale
and Astronomy, Friedrich Schiller University Jena, Löbdergraben 32, D-07743, Jena,
Germany. Tel.: +49 3641 947730; fax: +49 3641 947732. without changing the surface chemistry [11]. These titanium thin
E-mail address: k.jandt@uni-jena.de (K.D. Jandt). films (TiO2 -TFs) might be suitable surfaces to study the influence

0169-4332/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.apsusc.2013.05.030
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 579

of titanium surface roughness in the nanometer range on the bio- 2.2. Atomic force microscopy (AFM)
logical surrounding. Reproducibility of the preparation method is
a mandatory requirement for model surfaces to assure the repro- An atomic force microscope (Dimension 3100, Digital Instru-
ducibility and comparability of the results within a study and ments, Santa Barbara, CA, USA), equipped with a standard silicon
between different studies [12]. The statistical evaluation of the tip (tip diameter <10 nm) was used to investigate the topography
PVD-TiO2 -TFs for reproducibility was, to our best knowledge, not of the TiO2 -TF surfaces. Prior to the measurements, the accuracy
part of previous studies. of the AFM was tested by the measurement of a standard AFM
To characterize surface structures on the nanometer scale, calibration sample (TGQ1; NT-MDT Co., Moscow, Russia). For
atomic force microscopy (AFM) is an appropriate and established TiO2 -TF surface characterization, the AFM was operated in tapping
method [13]. Investigations of the interaction between a biomate- mode with a scan rate of 2 Hz, and a scan size of 1 ␮m × 1 ␮m. For
rial and its biological environment often require different scan sizes the titanium samples of sample group B, in addition, scan sizes of
for AFM measurements, due to differences in the size of proteins, 25 ␮m × 25 ␮m and 50 ␮m × 50 ␮m were used. All AFM data were
cells, microorganisms or biofilms [14]. Calculated roughness values acquired with a resolution of 512 × 512 points. The point-to-point
based on AFM measurements can vary with the used AFM scan size distance was, thus, 1.95 nm (1 ␮m × 1 ␮m scan size), 48.83 nm
or the image resolution [14–16] which negatively affects the com- (25 ␮m × 25 ␮m scan size) and 97.66 nm (50 ␮m × 50 ␮m scan
parability of the results between different studies. For uniform and size). Additional images were obtained from three titanium
smooth PVD-TiO2 -TFs with roughnesses in the lower nanometer samples (sample group B) prepared during the fifth deposition
range this was not shown before. run with a resolution of 128, 256, 512, 1024 and 2048 points
The first aim of our study was, therefore, to evaluate the repeat- per line with a scan size of 12.5 ␮m × 12.5 ␮m, 25 ␮m × 25 ␮m
ability and reproducibility of the PVD technique for the preparation and 50 ␮m × 50 ␮m, to keep the pixel density of the AFM images
of TiO2 -TFs with respect to commonly used surface characteriza- constant with increasing scan size.
tion parameters using an AFM based approach. The suitability of the
PVD-TiO2 -TFs as model surfaces for the investigation of the influ- 2.3. Bacterial adhesion test and model organism
ence of nanoroughness on the interaction between the biological
system and the biomaterials surface was discussed. In addition, the Bacterial adhesion was investigated on the physical vapor
influence of an increasing surface nanoroughness on the adhesion deposited nanorough titanium surfaces of the sample groups A–D.
of Escherichia coli was investigated. The second aim of our study was E. coli EC081 was used as model organism for the bacterial adhe-
to critically analyze and discuss the calculated surface parameters sion tests. EC081 was obtained by transformation of the E. coli K-12
of the TiO2 -TFs with respect to AFM scan size, image resolution and strain RV308 (=ATCC 31608) carrying the plasmid pMK3c2GFP. The
pixel density. green fluorescent protein (GFP) is constitutively produced in EC081
which ensures a completely fluorescent cell population suitable for
2. Materials and methods easy monitoring of the cell attachment on the titanium surfaces
with confocal laser scanning microscopy. In addition, the plasmid
2.1. Titanium thin film preparation contains a kanamycin resistance gene causing post-segregational
killing of plasmid-free cells in the presence of kanamycin.
Glass slides (diameter 15 mm; Borofloat® B33; Jena 4H Engi- E. coli EC081 was cultivated as a batch culture in LB (lysogeny
neering GmbH, Jena, Germany) were used as substrates for the broth, Luria Bertani) medium with kanamycin (50 ␮g/mL) for 3 h, in
TiO2 -TF deposition. The glass slides were cleaned by different an orbital shaker at 200 rpm and 30 ◦ C. In the exponential growth
sonification steps: 30 min in Deconex® 11 Universal Cleaner (1% phase, the culture was diluted with LB to an optical density OD
v/v solution in double distilled water; Borer Chemie, Zuchwil, ( = 600 nm) of 0.5, measured with a spectrophotometer (Spekol
Switzerland) (80 ◦ C), 10 min in double distilled water (20 ◦ C), 1100, Analytic Jena, Jena, Germany). 3 mL of the diluted culture
10 min in ethanol (>99.9%), 10 min in acetone (20 ◦ C) and 10 min were used to inoculate each titanium sample (group A–D, n = 3,
in double distilled water (20 ◦ C). Subsequently, the glass slides sterilized) placed in a standard 12-well culture plate. Samples were
were dried with filtered, compressed air and immediately used for taken after 6 h of incubation at room temperature.
the deposition of the TiO2 -TFs. For the preparation of the TiO2 -
TFs, the PVD method was used as described before [11]. With 2.4. Confocal laser scanning microscopy and scanning electron
this technique, the titanium surface roughness can be adjusted microscopy
by the deposition rate and the film thickness. For evaluation of
the PVD technique, titanium (purity of 99.99%; Mateck-Material- Titanium samples with adhered bacteria were fixed with
Technologie & Kristalle GmbH, Jülich, Germany) was evaporated phosphate-buffered formaldehyde solution (4%, Roti® -Histofix,
with an electron-beam evaporator (Univex 350, Leybold, Germany) Carl Roth GmbH, Karlsruhe, Germany) for 12 h at 4 ◦ C and air-dried
with a deposition rate of 0.5 nm/s and a film thickness of 200 nm at room temperature in the dark, afterwards. During preparation,
at a vacuum of 6 × 10−6 Torr. The film thickness was monitored samples were handled with special care to prevent a detachment
by a quartz balance installed inside the evaporation chamber. The and loss of adhered bacteria e.g. due to shear forces during rins-
deposition rate was kept constant and controlled by a deposi- ing. Dried samples were mounted on microscope slides, embedded
tion monitor with an error range of 0.01 nm/s. 150 glass slides using the ProLong® Antifade Kit (P7481, Molecular Probes® , Darm-
were titanium coated in overall 6 deposition runs. From each run, stadt, Germany) according to the manufacturer’s instruction and
three coated glass slides were investigated with AFM at three ran- covered with cover slips to prevent photobleaching effects dur-
domly chosen locations on each sample. Due to the contact of the ing fluorescence microscopy and to enable long-time storage (for
evaporated titanium thin films with atmospheric oxygen, a tita- months) of the samples at −20 ◦ C.
nium dioxide (TiO2 ) layer with an approximate thickness of 5 nm A confocal laser scanning microscope (CLSM; Zeiss LSM 510
is immediately formed [11]. Meta, Carl Zeiss MicroImaging, Jena, Germany) equipped with an
For bacterial adhesion tests, in addition, TiO2 -TFs were prepared Argon/2 laser unit (488 nm) and a 63× NA 1.3 oil immersion lens
with deposition rates and film thicknesses of 0.1 nm/s and 100 nm objective (Zeiss PLANAPOCHROMAT® ) was used for fluorescence
(sample group A), 0.5 nm/s and 200 nm (group B), 0.5 nm/s and imaging of the bacteria. With a 1.5-fold digital zoom, the basic field
500 nm (group C), and 1.0 nm/s and 500 nm (group D), respectively. of view for each image was 71.4 ␮m × 71.4 ␮m. Images were taken
580 C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589

at five different randomly chosen points on the sample groups 2.7. Statistical analysis
A–C and at 10 different randomly chosen points on the samples of
group D. Using the ANOVA analysis and a Scheffé t-test (95% confi-
For scanning electron microscopy (SEM), biofilm samples were dence interval, p ≤ 0.05), the deposition runs and the roughness
fixed with glutardialdehyde solution (2.5% in PBS) at room tem- values were statistically analyzed with Statgraphics Centurion XV
perature for 12 h. After fixation, samples were washed for 10 min (StatPoint Inc., Warrenton, USA). Results are given as p-values and
in PBS buffer solution and twice in distilled water. Subsequently, F-ratio. The F-ratio calculated by the software is the quotient of
samples were dehydrated using an ascending ethanol series from the variances of two groups of samples. The critical value for the
10 to >99.9% with 10 min incubation per step and two times 30 min F-ratio is 4.43 based on the F-distribution: F21 (p-value), whereas
for the last step. The dehydrated samples were critical point dried 1 and 2 are the degrees of freedom from the sample groups, and
(EMITECH K850, Quorum Technologies Ltd., East Grinstead, UK) and p is 0.05. If the F-ratio is >4.43, the variances of the sample groups
gold sputter coated (∼5 nm) (S150B, Edwards Ltd., Crawley, UK). are statistically significantly different. Using the ANOVA analysis
For SEM imaging, an AURIGA 60 CrossBeam® FIB-SEM scanning and a Scheffé t-test (95% confidence interval, p ≤ 0.05) the bacterial
electron microscope (Carl Zeiss AG, Oberkochen, Germany) was adhesion on the titanium samples (estimated surface coverages)
used at a magnification of ∼6000× operated at 5 kV and a working was statistically analyzed.
distance of about 4 mm.
2.8. Image processing and analysis

2.5. Propagation of uncertainty The AFM image processing and the calculation of the surface
parameters were performed with Gwyddion 2.25 free SPM data
The AFM measurements of the TiO2 -TFs and respective calcula- analysis software (Brno, Czech Republic, http://gwyddion.net/).
tions of the surface characterizing values were carried out on three The AFM image processing was applied based on the guidelines
different randomly chosen points on each sample with overall n = 3 for raw AFM micrographs [18]. A plane leveling was used, whereas
titanium samples, thus, giving 9 different locations, overall. CLSM the plane was computed from all the image points and was sub-
images were acquired at 5 (sample group A–C) or 10 (group D) tracted from the data [19]. In addition, a line by line fit was applied
different randomly chosen points on each titanium sample with and scanning errors parallel to the scanning axis were removed.
overall n = 3 samples. For all mathematical operations which were Roughness parameters were calculated according to the following
performed on the measured quantities, a propagation of uncer- equations (Standard: ISO 4287-1997) with the number of points N
tainty was applied according to [17] since the quantity of interest (intersections of the surface profile with the mean line), rj the height
for statistical analysis was not obtained by measuring this quantity of the profile at point j and Rq as the RMS roughness parameter [19]:
directly. Arithmetic average roughness (Ra )

1
N
Ra = |rj | (1)
2.6. Accuracy and precision N
j=1

The experimental error was determined by its accuracy and its Root mean square roughness (RMS)
precision [12]. The accuracy describes how close a measured value 
 N
is to the true or accepted value. The accuracy of the AFM measure- 1
ments was determined as: RMS =  r2 j
(2)
N
j=1
|E − A|
%error = (i) Skewness (Rsk )
A

1 
N
where E (experimental value) is the measured average surface Rsk = rj3 (3)
profile height of a standard sample for AFM calibration and A NRq3
j=1
(true/accepted value) is the height of the calibration sample given
by the manufacturer (see Section 2.2). Kurtosis (Rku )
For the calculated surface roughnesses of the TiO2 -TFs, the pre-
cision was determined. The precision describes how closely two 1 
N

Rku = rj4 (4)


or more measurements agree with each other and is in this study NRq4
j=1
referred to as repeatability and reproducibility [12]. Repeatability is
the degree of agreement of measurements on replicate specimens In addition, the measured surface area (MSA) and the fractal
by the same researcher in the same laboratory. The repeatability dimension (FD) were calculated. The FD characterizes the rough-
was tested by the statistical comparison (analysis of variance, see ness of a self-similar surface and varies from 2 for a totally smooth
below at Section 2.7) of the mean roughness values and according surface to 3 for an infinitely rough surface. For the fractal analysis,
to standard deviations of the titanium surfaces produced during the cube counting method was used [19]. The algorithm is based
the 6 different deposition runs. on the following steps: a cubic lattice with a lattice constant h is
Reproducibility is the degree of agreement between measure- superimposed on the z-expanded surface. Initially, h is set at X = 2
ments on replicate specimens in different locations by different (where X is the length of the edge of the surface), resulting in a
people and at different time points. The reproducibility was, there- lattice of 2 × 2 × 2 = 8 cubes, with N(h) being the total number of
fore, determined by the comparison of the experimental results cubes. The lattice constant h is then reduced stepwise by a factor of
of the current study with the results of the study from Cai et al. 2 and the process is repeated until h equals the distance between
[11]. The mean roughness values of the TiO2 -TFs produced during two adjacent pixels. The slope of a plot of log N(h) versus log (1/h)
the 6 deposition runs calculated for the scan size 50 ␮m × 50 ␮m gives the fractal dimension directly [19].
were statistically compared to the surface roughness values of the For the calculation of the average peak-to-peak distances of
TiO2 -TFs investigated in Cai’s study. the nanorough titanium surfaces, the coordinates of the local
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 581

Table 1
Roughness values, fractal dimension, measured surface area (MSA) and the deviation of MSA from the scanned surface area (SSA) of the titanium thin films (deposition rate
0.5 nm/s, film thickness 200 nm) dependent on scan size and deposition run (mean ± standard deviation; with n = 3 for the titanium thin films (with 3 images/sample and
per scan size) and n = 1 for industrial purchased Borofloat glass slides.

Scan size Runa Ra (nm) RMS (nm) Skewness Kurtosis Fractal dimension MSA (␮m2 ) Deviation MSA
from SSA (%)

1 ␮m × 1 ␮m 1 2.41 ± 0.46 3.00 ± 0.54 −0.13 ± 0.12 −0.14 ± 0.22 2.42 ± 0.03 1.10 ± 0.04 8.74 ± 4.37
2 2.07 ± 0.35 2.58 ± 0.44 0.13 ± 0.09 −0.09 ± 0.14 2.41 ± 0.04 1.10 ± 0.07 8.42 ± 5.69
3 2.15 ± 0.09 2.70 ± 0.11 0.21 ± 0.10 −0.01 ± 0.08 2.43 ± 0.01 1.10 ± 0.02 8.84 ± 1.35
4 2.15 ± 0.17 2.70 ± 0.23 0.35 ± 0.05 0.20 ± 0.49 2.45 ± 0.02 1.18 ± 0.02 9.16 ± 1.12
5 1.89 ± 0.12 2.37 ± 0.14 0.40 ± 0.06 0.06 ± 0.14 2.42 ± 0.02 1.07 ± 0.01 6.28 ± 1.13
6 1.98 ± 0.15 2.47 ± 0.19 0.31 ± 0.05 −0.11 ± 0.06 2.42 ± 0.02 1.07 ± 0.03 6.37 ± 2.18
Glass 0.21 0.27 0.38 0.98 2.37 1.002 0.179

25 ␮m × 25 ␮m 1 2.80 ± 0.24 3.46 ± 0.31 0.29 ± 0.09 0.07 ± 0.33 2.58 ± 0.05 626.84 ± 0.81 0.29 ± 0.13
2 2.62 ± 0.61 3.29 ± 0.77 0.27 ± 0.13 0.35 ± 0.41 2.61 ± 0.04 628.09 ± 1.80 0.49 ± 0.28
3 2.73 ± 0.08 3.40 ± 0.10 0.27 ± 0.06 0.01 ± 0.13 2.61 ± 0.02 627.51 ± 0.24 0.40 ± 0.04
4 2.11 ± 0.16 2.45 ± 0.48 0.25 ± 0.04 −0.02 ± 0.31 2.61 ± 0.04 626.20 ± 0.31 0.19 ± 0.05
5 1.86 ± 0.44 2.41 ± 0.74 0.47 ± 0.23 −0.24 ± 0.10 2.55 ± 0.05 625.46 ± 0.07 0.07 ± 0.01
6 2.81 ± 0.51 3.50 ± 0.64 0.31 ± 0.05 0.04 ± 0.16 2.64 ± 0.03 629.12 ± 1.81 0.65 ± 0.28
Glass 2.52 2.96 0.67 0.01 2.17 625.04 0.006

50 ␮m × 50 ␮m 1 6.42 ± 0.39 7.71 ± 0.56 0.56 ± 0.07 −0.10 ± 0.63 2.41 ± 0.08 2501.95 ± 2.31 0.08 ± 0.09
2 4.98 ± 0.96 6.08 ± 1.11 0.49 ± 0.09 0.01 ± 0.37 2.55 ± 0.04 2504.97 ± 0.62 0.20 ± 0.04
3 5.66 ± 0.43 6.72 ± 0.49 0.41 ± 0.07 −0.49 ± 0.22 2.45 ± 0.04 2501.83 ± 0.40 0.07 ± 0.02
4 4.66 ± 0.77 5.52 ± 0.89 0.41 ± 0.06 −0.45 ± 0.59 2.49 ± 0.01 2500.91 ± 0.43 0.04 ± 0.02
5 4.41 ± 1.19 5.20 ± 1.30 0.35 ± 0.29 −0.19 ± 0.89 2.40 ± 0.05 2500.31 ± 0.08 0.01 ± 0.00
6 4.95 ± 0.79 6.12 ± 1.02 0.59 ± 0.08 0.03 ± 0.33 2.63 ± 0.05 2509.05 ± 4.41 0.36 ± 0.18
Glass 5.22 6.19 0.78 −0.23 2.15 2500.05 0.002

AFM measurements: images were obtained in tapping mode in air with a resolution of 512 × 512 pixel; scan speed, scan angle, gains and setpoint were kept constant. AFM
image processing: raw images were subjected to a plane leveling, a 1st order line-by-line fit and a scar removal parallel to the scanning axis.
a
Number of deposition run during sample preparation of the titanium thin films; deposition rate 0.5 nm/s; film thickness 200 nm; in comparison roughness values, fractal
dimension, MSA, and deviation of MSA from SSA are given for the uncoated Borofloat® B33 glass slides.

maxima of the AFM height images were estimated using ImageJ standard deviations. Thus, no further calibration of the AFM was
(National Institutes of Health NIH, Bethesda, Maryland, USA). Based necessary.
on a Delaunay triangulation algorithm the distances between Table 1 summarizes the roughness values, the FD, the measured
the neighboring maxima were calculated using MATLAB (Math- surface area (MSA) and the deviation of the MSA and the scanned
Works, Natick, Massachusetts, USA). Surface coverages of the CLSM surface area (SSA) depending on the deposition run (run 1–6) and
images were calculated using the free software bioImage L v.2.1 the used scan sizes for the TiO2 -TFs prepared with a deposition
[20]. For image analysis a factor of 0.03 for noise reduction was rate of 5 nm/s and a film thickness of 200 nm. Table 2 shows the
applied. results of the statistical analysis (Analysis of Variance, ANOVA) of
the different deposition runs for the scan size 1 ␮m × 1 ␮m. There
were no statistically significant differences between the roughness
3. Results values, the FD and the MSA of the six deposition runs (p ≥ 0.05, F-
ratio <4.43). For skewness (p ≤ 0.05, F-ratio >4.43), we found the
3.1. Evaluation of titanium thin films prepared with PVD titanium deposition run 1 statistically significantly different from
the other five deposition runs.
The accuracy of the AFM was tested by measuring an AFM To test the reproducibility of the surface roughness of the TiO2 -
calibration standard sample and revealed an experimental value TFs and, thus, the reproducibility of the PVD technique for the
of 21.18 ± 0.86 nm height compared to the true, respectively, preparation of these surfaces, the calculated RMS mean values
accepted value of 20.5 ± 1.5 nm given by the manufacturer. There- of the titanium samples (deposition rate 5 nm/s, film thickness
fore, the accuracy (%error) of the AFM measurements calculated 200 nm, AFM scan size 50 ␮m) produced during the 6 deposition
according to Eq. (1) was 0.0332 which corresponds to a deviation runs were compared to the surface roughness values of the TiO2 -
of 3.32% from the measured value to the true value. Furthermore, TFs (prepared with the same parameters) investigated by Cai et al.
the calculated mean value and the true value given by the company [11]. The statistical analysis (Scheffé t-test) showed that there was
are in the range of each other which is shown by their overlapping no statistically significant difference between the roughness values

Table 2
Results of the statistical analysis (Analysis of Variance) of the surface roughness values, fractal dimension and measured surface area (MSA) of the titanium thin films
(deposition rate 0.5 nm/s, film thickness 200 nm) for scan size 1 ␮m × 1 ␮m dependent on the deposition run (n = 9 per run); confidence interval 95% with p ≤ 0.05, critical
value for F-ratio is 4.43.

Ra RMS Skewness Kurtosis Fractal dimension MSA

Run 1 × × × × × ×
2 × × × × × × ×
3 × × × × × × ×
4 × × × × × ×
5 × × × × × ×
6 × × × × × ×
p-Value 0.18 0.18 0.00 0.08 0.25 0.20
F-ratio 1.82 1.82 8.13 2.58 1.55 1.76

Homogeneous groups are displayed by crosses (×); if the crosses are beneath each other, there is no statistical significance of difference. If the crosses are moved towards
the right, there is a statistically significant difference.
582 C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589

Table 3
Characterization of the PVD-TiO2 -TF surfaces prepared with different deposition rates and film thicknesses (sample group A–D) used for the test of the bacterial adhesion:
surface roughness values Ra and RMS, the maximum vertical peak-to-valley distance and the average horizontal peak-to-peak distance. In addition, bacterial surface coverage
after 6 h of incubation is given. AFM scan size 1 ␮m × 1 ␮m, resolution 512 × 512 pixel; n = 3 for AFM images and n = 3 for CLSM images used for analysis.

Sample group Rate (nm/s)/thickness (nm) Ra ± SD (nm) RMS ± SD (nm) Peak-to-valley ± SD (nm) Peak-to-peak ± SD (nm)a Bacterial surface
coverage ± SD (%)

A 1/100 1.58 ± 0.10 2.00 ± 0.12 16.94 ± 1.72 13.12 ± 6.43 23.12 ± 1.60
B 5/200 1.81 ± 0.01 2.28 ± 0.01 19.93 ± 1.21 17.87 ± 8.05 22.78 ± 5.67
C 5/500 2.35 ± 0.14 2.96 ± 0.19 24.66 ± 6.49 26.40 ± 11.33 20.54 ± 3.28
D 10/500 4.82 ± 0.18 6.13 ± 0.19 45.17 ± 2.84 42.17 ± 23.20 19.11 ± 7.09
a
Coordinates of the local maxima of the AFM height images were estimated and based on a Delaunay triangulation algorithm the distances between the neighboring
maxima were calculated.

of the TiO2 -TFs prepared in this study compared to that prepared MSA proportionally decreased with increasing scan size with up
in the study of Cai et al. (data not shown). The calculated mean to 12% deviation from the measured surface area to the scanned
surface roughness (RMS) of all TiO2 -TFs produced during the six surface area for 1 ␮m × 1 ␮m scan size and below 0.4% deviation
deposition runs was 6.2 ± 1.6 nm. The mean RMS value of the TiO2 - from the MSA to the SSA for 50 ␮m × 50 ␮m scan size.
TFs prepared and investigated by Cai et al. was 4.9 ± 0.2 nm. Both The effect of increasing calculated Ra and RMS values with
values were within the deviation range of each other. increasing scan size used for AFM measurements was described
in the literature before but no comprehensive explanation for this
3.2. Bacterial adhesion on nanorough titanium thin films was found in these studies [14,15,21]. This effect was, therefore, for
the most often used roughness values investigated in more detail.
TiO2 -TFs were prepared with different deposition rates and film The roughness values Ra and RMS were calculated with respect
thicknesses resulting in different surface nanoroughnesses (sample to pixel resolution and scan size (Fig. 6a) and in relation to scan
groups A–D; Table 3, Fig. 1). The calculated average surfaces rough- size and pixel density (Fig. 6b). Our results show that an increas-
ness (Ra ) and root mean square roughness (RMS) of the TiO2 -TFs ing resolution (256, 512 and 1024 pixel/line) had no influence on
based on AFM images with a scan size of 1 ␮m × 1 ␮m increased the measured Ra and RMS values, nevertheless, with increasing
from 1.58 ± 0.10 nm, respectively, 2.00 ± 0.12 nm for the samples scan size the calculated roughness values increased. The effect
of group A to 4.82 ± 0.18 nm, respectively, 6.13 ± 0.19 nm for the of increasing measured roughness with increasing scan size was
samples of group D. The maximum peak-to-valley distance (verti- identical for all measured pixel resolutions (Fig. 6a) and pixel den-
cal) and the average peak-to-peak distance (horizontal) increased sities (Fig. 6b), respectively. Measured Ra and RMS values were, as
from 16.94 ± 1.72 nm and 13.12 ± 6.43 nm to 45.17 ± 2.84 nm and expected for the uniform titanium layers, not statistically signif-
42.17 ± 23.20 nm. icantly different from each other within the same scan size, but
The bacterial coverage decreased from 23.12 ± 1.60% to they became significantly higher with increasing AFM scan size
19.11 ± 7.09% with increasing surface roughness (Table 3, Fig. 2). independent of resolution or pixel density of the AFM image.
Although this decrease is an almost one fifth reduction of the sur- As additional or alternative parameter for the surface character-
face coverage, there were no statistically significant differences ization of the TiO2 -TFs next to the roughness, the fractal dimension
between the average coverage of the TiO2 -TFs of groups A and D, was calculated with respect to the deposition run and the scan size
since the standard deviation of group D is notably high. The bacte- with a constant resolution of 512 × 512 pixel per image (Table 1)
rial adhesion on the nanorough titanium surfaces after 6 hours as and with respect to different resolutions between 256 and 1024
well as histograms of the distribution of the data are shown in Fig. 3. pixel/line for the three scan sizes. Fig. 7 shows that the FD, given by
It has been observed that the distribution of the bacteria attached to the slope of the linear fitting, was the same for the different reso-
the thin films of group A was very homogeneous resulting in a low lutions (a–c) and differed only slightly between the different scan
standard deviation of the mean coverage value. On the contrary, sizes (d).
the distribution of the bacteria attached to the thin films of group
D was inhomogeneous and patchy, as shown by the CLSM image
Fig. 3 and visualized by the according histogram, which results in 4. Discussion
a relatively high standard deviation of the bacterial coverage on
the titanium surfaces of sample group D. A beginning production 4.1. Evaluation of the PVD technique
of extracellular polymeric substances (EPS) after 6 h of incubation
of the bacteria on the surfaces was observed (Fig. 4). The PVD technique allows for the preparation of TiO2 -TFs with
desired surface properties of structure and composition [11]. By
3.3. Titanium surface characterizing parameters varying the deposition rate and the film thickness it is easily and
controllably possible to prepare titanium samples with various
AFM micrographs of a titanium surface prepared with the PVD surface roughnesses in the nanometer range between 1 nm and
technique (deposition rate 5 nm/s, film thickness 200 nm) obtained 21 nm [11,22]. Since surface roughness is one of the most impor-
with different scan sizes are shown in Fig. 5b–d. The TiO2 -TFs were tant factors influencing the biological performance of an implant,
deposited as uniform layers on the glass substrates. Fig. 5a shows decisively affecting the adsorption of proteins and biomolecules
the uncoated borofloat® glass substrate. [23], the attachment of cells [3] or the adhesion of microorganisms
The calculated values of Ra , RMS and skewness statistically sig- [24], the PVD technique provides a powerful tool to significantly
nificantly increased with increasing AFM scan size, whereas the influence the interaction of an implant surface with its surrounding
values of kurtosis and FD did not statistically significantly increase environment. For the qualitative and quantitative characterization
with increasing scan size (Table 1). With increasing scan size the of biomaterials surfaces and interfaces in general and rough tita-
calculated Ra mean value increased from ∼1.9 nm to ∼6.4 nm, the nium surfaces in particular, atomic force microscopy is a suitable
calculated RMS mean value from ∼2.4 nm to ∼7.7 nm and the calcu- and commonly used tool which provides excellent results on the
lated mean value of skewness from ∼−0.1 to ∼0.6. Kurtosis values nanometer scale [13] and was, therefore, the method of choice to
varied between −0.5 and 0.3 and the FD between 2.4 and 2.6. The be used in our study.
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 583

Fig. 2. Surface coverage of Escherichia coli cells (plotted as empty bars with standard
deviation; n = 3 with 5 measured points on each sample) adhered to PVD-TiO2 -TF
prepared with different deposition rates and film thicknesses resulting in differ-
ent surface nanoroughnesses (plotted as filled bars; n = 3): 0.1 nm/s and 100 nm,
2.00 ± 0.12 nm (sample group A), 0.5 nm/s and 200 nm, 2.28 ± 0.01 nm (sample
group B), 0.5 nm/s and 500 nm, 2.96 ± 0.19 nm (sample group C), and 1 nm/s and
500 nm, 6.13 ± 0.19 nm (sample group D).

size 1 ␮m × 1 ␮m. The statistical analysis (Table 2) revealed that


there were no significant differences between the calculated rough-
ness parameters Ra , RMS and kurtosis of the TiO2 -TFs produced
during the 6 deposition runs. PVD is, therefore, a suitable method
to repeatably prepare TiO2 -TFs with a surface roughness on the
nanometer scale. The mean arithmetic average surface roughness
Ra calculated for 1 ␮m × 1 ␮m scan size of all deposition runs was
2.10 nm with a standard deviation of 0.30 nm. This result showed
that within an error range of 14.3%, including the error of tita-
nium sample preparation and AFM measurements, it is possible
with the PVD technique to prepare TiO2 -TFs with desired surface
roughnesses. Nevertheless, since the surface roughnesses are on
the nanometer scale, this error range means a deviation of much
less than 1 nm. The fractal analysis of the surfaces and the calcula-
tion of the measured surface area and the deviation between the
measured and the scanned surface area supported these findings.
However, it remains unclear why the mean skewness value of the
first run statistically differs from the mean values calculated for the
titanium surfaces prepared during the other five runs.
Reproducibility is the degree of agreement between measure-
ments on replicate specimens in different locations by different
Fig. 1. AFM 3D height images of the PVD-TiO2 -TF surfaces prepared with different people and at different time points [12]. The reproducibility was,
deposition rates and film thicknesses used to test of the bacterial adhesion: 0.1 nm/s therefore, determined by the comparison of the experimental
and 100 nm (a), 0.5 nm/s and 200 nm (b), 0.5 nm/s and 500 nm (c) and 1 nm/s and results of the current study with the results of the study from Cai
500 nm (d); scan size 1 ␮m × 1 ␮m.
et al. [11] which was performed at a different location, by a dif-
ferent researcher and at a different time point. It was shown that
To evaluate the PVD method, the repeatability and reproducibil- there was no statistically significant difference between the rough-
ity of this technique was investigated. Repeatability is the degree ness values of the TiO2 -TFs prepared in this study compared to that
of agreement of measurements on replicate specimens by the same prepared in the study of Cai et al. Furthermore, the mean surface
researcher in the same laboratory [12]. Six independent titanium roughness (RMS) for the TiO2 -TFs investigated in the current study
deposition runs were carried out. Several surface characterizing and in the study of Cai et al. were within the standard deviation
parameters were calculated and analyzed with respect to scan size range of each other. Thus, the PVD method provides next to repeat-
and deposition runs (Table 1). To test statistically for variations able also reproducible results since both compared measurements
between the runs, thus, for the repeatability of the method, an anal- agree well with each other. Accordingly, the PVD technique can
ysis of variance (ANOVA) was performed on the results of the scan be described as a precise method referring to the definition given
584 C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589

Fig. 3. CLSM images of Escherichia coli cells adhered to PVD-TiO2 -TF prepared with different deposition rates and film thicknesses resulting in different surface nanorough-
nesses (RMS): 0.1 nm/s and 100 nm, 2.00 ± 0.12 nm (Group A), 0.5 nm/s and 200 nm, 2.28 ± 0.01 nm (Group B), 0.5 nm/s and 500 nm, 2.96 ± 0.19 nm (Group C) and 1 nm/s
and 500 nm, 6.13 ± 0.19 nm (Group D). In addition, the histograms of the estimated surface coverages of E. coli cells on the different nanorough titanium surfaces are shown,
giving the distribution of the data.

above. The higher mean and the relatively high standard deviation and low costs [25]. Our results showed that with the PVD technique
of the surface roughness in our study could be explained by the it is possible to prepare titanium surfaces with controlled surface
lower number of measured replicates but could be also due to the nanotopographies which match these requirements, as the thin
different software used for the calculations or was even caused by a films provide uniform and reproducible two-dimensional surfaces
higher order fit of the measured values during the data processing with well-defined nanoroughnesses. The thin films were deposited
in the study of Cai et al. which could have led to a loss of height on glass surfaces with an average surface roughness of 0.21 nm
data and correspondingly decreasing roughness values [18].
We demonstrated that the PVD technique provides a powerful
tool for the repeatable and reproducible preparation of TiO2 -
TFs with a well-defined nanoscale surface topography. The PVD
method, thus, expands the pool of available techniques for the
preparation of nanorough titanium surfaces.

4.2. PVD deposited titanium thin films as 2D nanorough model


surfaces

The surface topography of biomaterials is well known as one of


the most important factors influencing biological reaction [3] and,
thus, contributing to the biocompatibility of the material. But, much
more work is necessary to further advance a detailed understand-
ing of the very complex interactions between biological systems
and nanostructured biomaterials surfaces. For this, first, simple
two-dimensional (2D) model systems are required with well-
characterized surfaces and carefully designed nanotopographies
[3]. To fabricate 2D surfaces with well-defined nanotopographies
for cellular and microbial studies or protein adsorption, it is impor- Fig. 4. Scanning electron micrograph of the model organism Escherichia coli EC081
tant to focus on preparation techniques which produce highly used in this study. The image shows microbial cells which are adhered to a PVD-
TiO2 -TF (sample group B; deposition rate 5 nm/s, film thickness 200 nm) after 6 h
reproducible nanoscale features, uniform and coherent structures
of incubation and produced extracellular polymeric substances (EPS) as matrix for
over large areas in the range of cm2 and which are quick with an biofilm formation. Samples were critical-point dried and gold sputter coated before
appropriate number of produced replicas for large scale screenings imaging (average gold thickness of ∼5 nm).
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 585

Fig. 5. AFM height images of the uncoated Borofloat® B33 glass substrate (a) and the titanium thin films (deposition rate 0.5 nm/s, film thickness 200 nm) with a scan size
of 1 ␮m × 1 ␮m, 25 ␮m × 25 ␮m and 50 ␮m × 50 ␮m (b–d).

Fig. 6. Surface roughness values of the titanium thin films (n = 3; deposition rate 0.5 nm/s, film thickness 200 nm) depending on resolution and scan size (a) and depending
on scan size and pixel density (b); results of the statistical analysis (multi-way ANOVA) are given with a confidence interval of 95% and p ≤ 0.05; *denotes a statistically
significant difference between the values (a) or the groups (b), respectively.
586 C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589

Fig. 7. Fractal dimension FD (calculated with cube counting method) of the titanium thin films (deposition rate 0.5 nm/s, film thickness 200 nm) depending on resolution (pixel
per line) for the different scan sizes: 1 ␮m × 1 ␮m (a), 25 ␮m × 25 ␮m (b) and 50 ␮m × 50 ␮m (c) and depending on the scan size with a constant resolution of 512 pixel/line
(d); FD is given by the slope of the linear fitting.

(Table 1) in comparison to that of the thin films between 1.58 and biological performance of biomaterials. The use of these model
2.82 nm (Table 3). Thus, there is no influence of the underlying sub- surfaces for relevant studies, thus, provides a high comparability
strate surface topography on the topography of the titanium thin within one study and also between different studies.
films and no surface structures on the micrometer scale are present
on the sample surfaces. In addition, titanium samples in a number 4.3. The interaction between nanorough surfaces and the
of tenth to some hundreds can be fabricated within relatively short biological system
time and with moderate costs.
As any differences, even small, in the surface chemistry may On the macro- and microscale it has been shown that sur-
influence cellular, microbial or protein adhesion behavior and face topography has strong effects on human cells [28], since it
can, thus, mask the effect of topography, model substrates with can influence cell adhesion, morphology, migration and orienta-
an absolutely uniform surface chemistry even while varying the tion as well as cytoskeleton development and cell differentiation
topography are required [25]. It has been shown that in most [25]. Much less is known about how cells react to nanoscale struc-
cases it is difficult to decouple topographical from chemical effects tures. Cell dimensions are about 10–20 ␮m [29], however, some
[26] since topographical variations of the materials surfaces due to functional domains present on the cell membrane have dimensions
e.g. grinding, polishing or blasting are most often accompanied by on the nanoscale. In addition, in vivo cells live inside an extracel-
chemical heterogeneities [9,22,27]. Cai et al. observed no statisti- lular matrix containing nanoscale collagen fibrils, thus, cells are
cally significant changes in the chemical surface properties of the likely to be able to respond not only to surface topography on
PVD-TiO2 -TFs during the variation of the surface nanoroughness. the micrometer range but also on the nanometer range [3]. Cell
Qualitative and quantitative X-ray photoelectron spectroscopy as filopodia, for instance, have a tip diameter of 50–100 nm, while
well as X-ray diffraction measurements showed no differences in integrin-mediated adhesion is in the order of 15 nm, hence, are of
the chemical composition or the crystal structure of the TiO2 -TFs sufficient resolution to determine changes in the substrate nanoto-
prepared with different deposition parameters. pography. Nevertheless, the mechanisms by which cells respond
In summary, the findings of Cai et al. and our statistical evalu- to nanoscale surface topography and the subsequent signal trans-
ation of the PVD method and the deposited titanium thin films in duction pathways have to date not been resolved in detail [25]. The
the current study showed that these surfaces are due to their high PVD-TiO2 -TFs as 2D model surfaces provide a suitable option for
repeatability and reproducibility and constant chemical composi- the investigation of these interactions, which should be addressed
tion during the variation of the surface nanoroughness an excellent in further studies.
simple 2D model system required for studies addressing the influ- It is well established that biomaterials are immediately coated
ence of titanium surface roughness on the nanometer scale on the with biomolecules and proteins from the blood or interstitial
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 587

fluid when they are implanted or inserted into the human body. A more detailed investigation and discussion of the bacterial
Therefore, cells may be in some cases essentially controlled by adhesion on the nanorough PVD-TiO2 -TFs goes beyond the scope
these surface adsorbed biomolecules and proteins rather than by of this study and will be published elsewhere.
the nanoscale surface topography [30]. It is, thus, of fundamen-
tal interest to investigate how nanostructured materials influence 4.4. Critical analysis of the titanium surface characterizing
the adhesion behavior, structure and stability of biomolecules parameters
and proteins to understand the actual contribution of the sur-
face topography of a biomaterial to its biocompatibility. Recently, One of the most commonly used parameters for the quantita-
we investigated the adsorption of bovine serum albumin and tive characterization of biomaterials surfaces is the roughness [34].
human plasma fibrinogen on PVD-TiO2 -TFs [22]. A significantly Comparing the results of the different scan sizes to each other (for
larger amount of adsorbed fibrinogen has been observed on the TiO2 -TFs sample group B, deposition rate of 0.5 nm/s, film thick-
nanorough titanium surfaces compared to albumin. No significant ness of 200 nm), it was shown that the calculated roughnesses
differences have been found regarding the different surface nanor- values Ra , RMS and skewness statistically significantly increased
oughnesses of the TiO2 -TFs. In contrast, Singh et al. [31] found a with increasing scan size (Table 1). For Ra and RMS of grinded and
significant increase in the amount of adsorbed proteins originated polished titanium surfaces, this effect was described in the liter-
from fetal bovine serum on nanorough titanium surfaces prepared ature before but no comprehensive explanation for this has been
with supersonic cluster beam deposition. The conflicting results of found in these studies [14,15,21]. To our knowledge, for skewness
these two studies show that much more work has to be done to fully this effect was not shown before.
understand the mechanisms behind protein adsorption in general The surface roughness depends on the lateral and vertical
and the influence of surface nanoroughness on protein adsorption dimensions of the surface features and only if all these features
in particular. are included in the scanned and analyzed image frame, the rough-
Biomaterials-related infections are most frequently associated ness is unaffected of the scanning area [15]. With decreasing AFM
with microbial adhesion and biofilm formation on the indwelling scan size, some surface features, e.g., some height information may
implants. Far less is known about the mechanisms of surface get lost and the calculated roughness becomes smaller. Vice versa,
sensing by microorganisms than by eukaryotic cells and the influ- with increasing scan sizes the calculated surface roughness might
ence of surface topography on the nanometer scale on the microbial increase [15,35]. Since the TiO2 -TFs investigated in our study had
adhesion [3]. In the current study, the bacterial adhesion of E. coli a very coherent and uniform structure with surface features in the
on TiO2 -TFs with different nanoroughnesses was investigated. The nanometer range [11], we rule out this effect for our samples.
bacterial coverage decreased from approximately 23% to 19% with The resolution of the AFM image, more precisely the number of
increasing surface nanoroughness from ∼2 nm (sample group A) measured points per line, might be another possible reason hypoth-
to ∼6 nm (sample group D) (Fig. 2). The influence of surface struc- esized in the literature for an increased measured, respectively,
tures size on the microbial attachment is controversially discussed calculated surface roughness with increasing scan size [11,15,16].
in the literature. Some studies consider that microorganisms pre- In the studies mentioned above, the resolution was not increased
fer the attachment to surfaces which provide surface features, such proportionally with increasing scan size, thus, the point-to-point
as roughness, corresponding in size to the size of the bacterial distance increased and, accordingly, the pixel density of the images
cells [32]. It was hypothesized that the microorganisms are in that decreased. Using a mathematically constructed surface profile, it
way protected against shear stress and abrasion. Other studies has been shown previously that with an increasing number of
showed that microorganisms attach less to surfaces with features points per line the measured roughness decreased until a critical
corresponding in size to the size of the cells [3]. It is believed value of approximately 105 points per line and roughness values
that these structures cause mechanical stress in the cytoskele- remained constant above this value [15], although this value is only
ton of the microorganisms which make them uncomfortable in a theoretical one since in reality with AFM it is just possible to
this sessile state. Since the maximum peak-to-valley distance and measure with resolutions some order of magnitudes below. To test
the average peak-to-peak distance for the roughest titanium sur- these reported findings for the TiO2 -TFs investigated in our study,
faces (group D) were approximately 45 nm and 42 nm, respectively, we calculated Ra and RMS in dependence of the image resolution
these effects can be excluded as influencing factors on the bacte- and the scan size (Fig. 6a) and in relation to scan size and pixel den-
rial adhesion on the nanorough TiO2 -TFs. Moreover, the average sity (Fig. 6b). The effect of the increasing measured roughness with
size of the thin film surface structures of the sample groups A increasing scan size occurred independently from the resolution or
and B (in the range of ∼15 nm) which were correlated with the the pixel density. Our results, therefore, demonstrate that the res-
highest measured bacterial surface coverages, corresponds to some olution as well the pixel density had no influence on the calculated
extent to the size of the EPS (mediate the bacterial adhesion to sur- roughness parameters of the PVD-TiO2 -TFs.
faces and provide the mechanical stability of biofilms) produced All images were obtained and processed by a plane leveling, a 1st
by the bacteria. Scanning electron micrographs of the adhered order line-by-line fit and a scar removal parallel to the scanning axis
bacteria revealed a diameter of an EPS strand of 5–10 nm (Fig. 4). to ensure that the calculated roughness parameters for the TiO2 -
This size roughly corresponds to the structure size of the tita- TFs were valid and comparable. However, one more reason for the
nium surfaces of the samples of group A and B and might, thus, effect of the increasing measured roughness with increasing scan
promote bacterial adhesion on the one hand and increase mechan- size that we found may be a still insufficient image processing. A
ical stability of the adhered bacteria against abrasion on the other common appearance in AFM images is the scanner bow. It is caused
hand. by a swinging motion of the free end of the scanner. This leads to
The distribution of the bacteria attached to the thin films became a curve in the image plane especially with increasing scan sizes.
increasingly inhomogeneous and patchy with increasing surface Possibly, this effect was still present in the AFM images. Neverthe-
nanoroughness (Fig. 3). This observed effect might be due to bacte- less, it has to be considered that the way of image processing can
rial co-adhesion [33] and correlates with the generally decreasing directly affect the calculated roughness values, since a high order
bacterial surface coverage. On surfaces less attractive for bacteria, image processing, which can be useful to erase the AFM scanner
they might prefer to attach to areas on the materials surface, where bow, may lead to the loss of height data, thus, the roughness val-
other bacteria have previously attached and already produced EPS. ues decrease with a progressive image processing [18]. AFM image
This EPS may then promote bacterial adhesion. processing should be applied without the elimination of important
588 C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589

superordinate structures, such as waviness or structuring features. scan sizes, e.g. for the investigation of the interaction of biomateri-
Thus, the processing of AFM images is mostly a balance act between als surfaces with their biological environment. The used AFM scan
a sufficient image processing and height data loss. size might be larger for the investigation of the interaction of the
The percentage increase of a projected surface area relative to biomaterial with tissues, cells or biofilms and might be smaller for
the scanned image area gives considerable information about the the investigation of the interaction of the biomaterials surface with
topography of a surface, too, and was used in numerous studies microorganisms, proteins or other biomolecules. Then, the effects
for the investigation of a variety of materials surfaces and inter- discussed above have to be taken into account during data analysis,
faces [14]. This percentage increase is a suitable parameter in interpretation and discussion. As additional or alternative parame-
particular for biomedical applications since the deviation of the ter for the characterization of rough surfaces next to Ra and RMS, the
measured surface area (MSA) from the scanned surface area (SSA) fractal dimension seems to be well suited, since it is not influenced
is positively correlated with the specific surface area available for by the AFM scan size, pixel density or image resolution.
the initial adsorption of proteins or biomolecules, tissue adhesion
or the attachment of microorganisms. The surface area that can 5. Summary and conclusions
participate in any exchange between an implant surface with its
surrounding environment is, thus, much bigger for surfaces with a PVD is a suitable method to reproducibly prepare TiO2 -TFs with
higher deviation of the MSA from the SSA. a surface roughness on the lower nanometer scale. Within an error
Our results show that the deviation of the measured from range of 14.3% which corresponds to a deviation of much less than
the scanned surface area was, contrary to the roughness values 1 nm it is possible to prepare TiO2 -TFs with desired surface rough-
discussed above, negatively correlated with the scan size. With nesses.
increasing scan size, the percentage deviation between the mea- We demonstrated that surface roughnesses even on the low
sured and the scanned area decreased significantly. Furthermore, nanometer scale influence bacterial adhesion behavior. With
we found that the deviations mean value (of run 1–6, Table 1) for increasing surface roughness bacterial surface coverage decreased
the scan length of 50 ␮m (0.12 ± 0.14%) represented approximately and became increasingly inhomogeneous.
1/50 of the mean value of the 1 ␮m scan length (9.17 ± 2.01%) and PVD-TiO2 -TFs are an excellent 2D model system for studies
1/2 of the mean value of the 25 ␮m scan length (0.32 ± 0.16%). Thus, addressing the influence of surface roughness on the nanometer
the differences between the deviations mean values of the differ- scale on the biological performance of biomaterials. Our findings
ent scan sizes directly depended on the magnifications used for the give, thus, new momentum to biomaterials research and will sup-
AFM images. This observed negative correlation between the devi- port the development of biomaterials surfaces with anti-infectious
ation of the measured from the scanned surface area and the AFM surface properties.
scan size seems to be a mathematical effect since with increasing The critical analysis of the surface characterizing parameters
scan size the AFM image resolution remained constant and the den- used in our study showed that the calculated values of Ra , RMS and
sity of measured points per line decreased accordingly. This led to skewness were significantly influenced by the AFM scan size. This
the relative decrease of the measured surface area with increas- effect has, in general, to be taken into account for studies using dif-
ing scan size since surface features, especially for such nano rough ferent scan sizes or for comparing results between different studies.
surfaces, were missed due to the increasing distance between the
measured points. Thus, for the characterization of surface prop-
Acknowledgements
erties in the low nanometer range, higher AFM scan sizes, such
as 25 ␮m or 50 ␮m accompanied by a relatively low resolution of
The authors thank M.M.L. Arras for help with ImageJ and MAT-
512 × 512 pixels are not sufficient to display the real surface topog-
LAB, M. Beyer for introduction to the statistical analysis of the data,
raphy since the pixel size is bigger than the smallest surface feature.
T. Keller and J. Reichert for helpful discussions and the Graduate
Nevertheless, some surface characterization parameters used in
School “Jena School for Microbial Communication” (JSMC MikroIn-
our study were not affected in the way described above. The calcu-
ter supported by the TMBWK) which is funded by the German
lated values for fractal dimension and kurtosis have not increased
Excellence Initiative for financial support. Furthermore, KDJ grate-
or decreased with increasing scan size. The fractal dimension (FD) is
fully acknowledge the partial financial support of the Deutsche
a frequently used parameter for the characterization and compari-
Forschungsgemeinschaft (DFG), grant reference INST 275/241-1
son of rough surfaces in natural sciences, e.g. biology [36], medicine
FUGG, and the TMBWK, grant reference 62-4264 925/1/10/1/01.
[37] or geo sciences [38]. The FD characterizes the roughness of a
self-similar surface and varies from 2 for a smooth surface to 3
for an infinitely rough surface. From the literature it is known that References
the fractal dimension is not influenced by an underlying surface
[1] M. Geetha, A.K. Singh, R. Asokamani, A.K. Gogia, Ti based biomaterials, the ulti-
topography [39] which was confirmed by the results of our study.
mate choice for orthopaedic implants – a review, Progress in Materials Science
The small difference of the FD values between the scan length of 54 (2009) 397–425.
1 ␮m and 50 ␮m and the scan length of 25 ␮m remains unclear [2] S.A. Brown, J.E. Lemons, Medical Applications of Titanium and its Alloy: The
Material and Biological Issue, American Society for Testing and Materials,
and should be addressed in further studies. However, the fractal
Philadelphia, 1996.
dimension as well as the kurtosis seem to be suitable parameters [3] K. Anselme, P. Davidson, A.M. Popa, M. Giazzon, M. Liley, L. Ploux, The interac-
for the characterization of rough surfaces in general since they are tion of cells and bacteria with surfaces structured at the nanometre scale, Acta
not influenced by an increasing scan size, by the image resolution or Biomaterialia 6 (2010) 3824–3846.
[4] Y. Huang, X.Y. Lue, J.W. Ma, N. Huang, In vitro investigation of protein adsorp-
the pixel density. Furthermore, these surface characterizing param- tion and platelet adhesion on inorganic biomaterial surfaces, Applied Surface
eters seem not to be influenced by the commonly occurring AFM Science 255 (2008) 257–259.
scanner bow. Thus, it is not absolutely necessary to apply a high [5] L. Altomare, S. Fare, Cells response to topographic and chemical micropatterns,
Journal of Applied Biomaterials & Biomechanics 6 (2008) 132–143.
order processing to the AFM raw images which may be accompa- [6] R. Zimmerman, D. Ila, C. Muntele, M. Rodrigues, D.B. Poker, D. Hensley,
nied by a possible height data loss. Enhanced tissue adhesion by increased porosity and surface roughness of car-
For AFM based investigations, an identical scan size and image bon based biomaterials, Nuclear Instruments & Methods in Physics Research
Section B – Beam Interactions with Materials and Atoms 191 (2002) 825–829.
processing are particularly important as well as to use the smallest [7] S.D. Puckett, E. Taylor, T. Raimondo, T.J. Webster, The relationship between the
possible scan sizes which include all important surface features. nanostructure of titanium surfaces and bacterial attachment, Biomaterials 31
Nevertheless, sometimes it may also be required to use different (2010) 706–713.
C. Lüdecke et al. / Applied Surface Science 280 (2013) 578–589 589

[8] M.E. Barbour, D.J. O’Sullivan, H.F. Jenkinson, D.C. Jagger, The effects of polish- [25] M.S. Lord, M. Foss, F. Besenbacher, Influence of nanoscale surface topog-
ing methods on surface morphology, roughness and bacterial colonisation of raphy on protein adsorption and cellular response, Nano Today 5 (2010)
titanium abutments, Journal of Materials Science – Materials in Medicine 18 66–78.
(2007) 1439–1447. [26] M.H.P. da Silva, G.D.A. Soares, C.N. Elias, S.M. Best, I.R. Gibson, L. Disilvio,
[9] P. Cacciafesta, K.R. Hallam, A.C. Watkinson, G.C. Allen, M.J. Miles, K.D. Jandt, et al., In vitro cellular response to titanium electrochemically coated with
Visualisation of human plasma fibrinogen adsorbed on titanium implant sur- hydroxyapatite compared to titanium with three different levels of surface
faces with different roughness, Surface Science 491 (2001) 405–420. roughness, Journal of Materials Science – Materials in Medicine 14 (2003)
[10] A. Curtis, C. Wilkinson, Topographical control of cells, Biomaterials 18 (1997) 511–519.
1573–1583. [27] A. Diener, B. Nebe, F. Luthen, P. Becker, U. Beck, H.G. Neumann, et al., Control
[11] K.Y. Cai, M. Muller, J. Bossert, A. Rechtenbach, K.D. Jandt, Surface structure of focal adhesion dynamics by material surface characteristics, Biomaterials 26
and composition of flat titanium thin films as a function of film thickness and (2005) 383–392.
evaporation rate, Applied Surface Science 250 (2005) 252–267. [28] K. Anselme, Osteoblast adhesion on biomaterials, Biomaterials 21 (2000)
[12] ISO, International Vocabulary of Basic and General Terms in Metrology, Geneva, 667–681.
1993. [29] R.G. Flemming, C.J. Murphy, G.A. Abrams, S.L. Goodman, P.F. Nealey, Effects of
[13] K.D. Jandt, Atomic force microscopy of biomaterials surfaces and interfaces, synthetic micro- and nano-structured surfaces on cell behavior, Biomaterials
Surface Science 491 (2001) 303–332. 20 (1999) 573–588.
[14] A. Mendez-Vilas, M.G. Donoso, J.L. Gonzalez-Carrasco, M.L. Gonzalez-Martin, [30] B. Elmengaard, J.E. Bechtold, K. Soballe, In vivo study of the effect of RGD treat-
Looking at the micro-topography of polished and blasted Ti-based biomateri- ment on bone ongrowth on press-fit titanium alloy implants, Biomaterials 26
als using atomic force microscopy and contact angle goniometry, Colloids and (2005) 3521–3526.
Surfaces B – Biointerfaces 52 (2006) 157–166. [31] A.V. Singh, V. Vyas, R. Patil, V. Sharma, P.E. Scopelliti, G. Bongiorno, et al.,
[15] A. Mendez-Vilas, J.M. Bruque, M.L. Gonzalez-Martin, Sensitivity of surface Quantitative characterization of the influence of the nanoscale morphology
roughness parameters to changes in the density of scanning points in multi- of nanostructured surfaces on bacterial adhesion and biofilm formation, PloS
scale AFM studies. Application to a biomaterial surface, Ultramicroscopy 107 ONE 6 (2011) 12.
(2007) 617–625. [32] M. Fletcher, Bacterial attachment in aquatic environments: a diver-
[16] P.J. Ramon-Torregrosa, M.A. Rodriguez-Valverde, A. Amirfazli, M.A. Cabrerizo- sity of surfaces and adhesion strategies, in: M. Fletcher (Ed.), Bacterial
Vilchez, Factors affecting the measurement of roughness factor of surfaces and Adhesion: Molecular and Ecological Diversity, Wiley, New York, 1996,
its implications for wetting studies, Colloids and Surfaces A – Physicochemical pp. 1–24.
and Engineering Aspects 323 (2008) 83–93. [33] R. Bos, H.C. Vandermei, J.M. Meinders, H.J. Busscher, A quantitative method
[17] J.R. Taylor, An Introduction to Error Analysis: The Studies Of Uncertainties in to study co-adhesion of microorganisms in a parallel-plate flow chember –
Physical Measurements, second ed., University Science Books, Sausalito, 1997. basic principles of the analysis, Journal of Microbiological Methods 20 (1994)
[18] P. Eaton, P. West, Atomic Force Microscopy, first ed., Oxford University Press 289–305.
Inc, New York, 2010. [34] E.S. Gadelmawla, M.M. Koura, T.M.A. Maksoud, I.M. Elewa, H.H. Soliman,
[19] D. Necas, P. Klapetek, Gwyddion: an open-source software for SPM data anal- Roughness parameters, Journal of Materials Processing Technology 123 (2002)
ysis, Central European Journal of Physics 10 (2012) 181–188. 133–145.
[20] L.C. de Paz, Image analysis software based on color segmentation for char- [35] D.V. Kilpadi, J.J. Weimer, J.E. Lemons, Effect of passivation and dry heat-
acterization of viability and physiological activity of biofilms, Applied and sterilization on surface energy and topography of unalloyed titanium implants,
Environmental Microbiology 75 (2009) 1734–1739. Colloids and Surfaces A – Physicochemical and Engineering Aspects 135 (1998)
[21] G.J. Simpson, D.L. Sedin, K.L. Rowlen, Surface roughness by contact versus tapp- 89–101.
ing mode atomic force microscopy, Langmuir 15 (1999) 1429–1434. [36] S.K. Rice, D. Collins, A.M. Anderson, Functional significance of variation
[22] K.Y. Cai, J. Bossert, K.D. Jandt, Does the nanometre scale topography of titanium in bryophyte canopy structure, American Journal of Botany 88 (2001)
influence protein adsorption and cell proliferation? Colloids and Surfaces B – 1568–1576.
Biointerfaces 49 (2006) 136–144. [37] B.R. Masters, Fractal analysis of the vascular tree in the human retina, Annual
[23] P. Cacciafesta, A.D.L. Humphris, K.D. Jandt, M.J. Miles, Human plasma fibrino- Review of Biomedical Engineering 6 (2004) 427–452.
gen adsorption on ultraflat titanium oxide surfaces studied with atomic force [38] P.S. Dodds, D.H. Rothman, Scaling, universality, and geomorphology, Annual
microscopy, Langmuir 16 (2000) 8167–8175. Review of Earth and Planetary Sciences 28 (2000) 571–610.
[24] E.P. Ivanova, V.K. Truong, J.Y. Wang, C.C. Berndt, R.T. Jones, I.I. Yusuf, et al., [39] P.F. Chauvy, C. Madore, D. Landolt, Variable length scale analysis of surface
Impact of nanoscale roughness of titanium thin film surfaces on bacterial reten- topography: characterization of titanium surfaces for biomedical applications,
tion, Langmuir 26 (2010) 1973–1982. Surface & Coatings Technology 110 (1998) 48–56.

You might also like