- Email: [email protected]
Depth profiling of galvanoaluminium–nickel coatings on steel by UV- and VIS-LIBS
Accepted Manuscript Title: Depth Profiling of Galvanoaluminium–Nickel Coatings on Steel by UV- and VIS-LIBS Author: T.O. Nagy U. Pacher A. Giesriegl M.J.J. Weimerskirch W. Kautek PII: DOI: Reference:
S0169-4332(16)32776-3 http://dx.doi.org/doi:10.1016/j.apsusc.2016.12.059 APSUSC 34606
To appear in:
APSUSC
Received date: Revised date: Accepted date:
30-6-2016 15-11-2016 8-12-2016
Please cite this article as: T.O. Nagy, U. Pacher, A. Giesriegl, M.J.J. Weimerskirch, W. Kautek, Depth Profiling of GalvanoaluminiumndashNickel Coatings on Steel by UV- and VIS-LIBS, (2016), http://dx.doi.org/10.1016/j.apsusc.2016.12.059 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t
Depth Profiling of Galvanoaluminium–Nickel Coatings on Steel by UV- and VIS-LIBS Nagy, T. O.1 , Pacher, U., Giesriegl, A., Weimerskirch, M. J. J., Kautek, W.
us
cr
Universit¨ at Wien, Institut f¨ ur Physikalische Chemie W¨ ahringer Str. 42, 1090 Wien
Abstract
an
Laser-induced depth profiling was applied to the investigation of galvanised steel sheets as a typical modern multi-layer coating system for environmental corrosion protection. The samples were ablated stepwise by the use of two different
M
wavelengths of a frequency-converted Nd:YAG-laser, 266 nm and 532 nm, with a pulse duration of τ = 4 ns at fluences ranging from F = 50 – 250 J·cm−2 . The emission light of the resulting plasma was analysed as a function of both pen-
d
etration depth and elemental spectrum in terms of linear correlation analysis.
te
Elemental depth profiles were calculated and compared to EDX-cross sections of the cut sample. A proven mathematical algorithm designed for the reconstruc-
Ac ce p
tion of layer structures from distorted emission traces caused by the Gaussian ablation profile can even resolve thin intermediate layers in terms of depth and thickness. The obtained results were compared to a purely thermally controlled ablation model. Thereby light-plasma coupling is suggested to be a possible cause of deviations in the ablation behaviour of Al. The average ablation rate h as a function of fluence F for Ni ranges from 1–3.5 µm/pulse for λ = 266 nm as well as for λ = 532 nm. In contrast, the range of h for Al differs from 2–4 µm/pulse for λ = 532 nm and 4–8 µm/pulse for λ = 266 nm in the exact same
fluence range on the exact same sample. Keywords: LIBS stratigraphy; galvanic coatings; ablation rate; plasma shielding; light-plasma-interaction; depth profiling 1 Correspondence:
[email protected]
Preprint submitted to Applied Surface Science
December 19, 2016
Page 1 of 27
ip t
Contents 1 Introduction
4
Galvanic aluminium coatings . . . . . . . . . . . . . . . . . . . .
1.2
Layer thickness and composition analysis . . . . . . . . . . . . .
4
cr
1.1
us
2 Theory and Experimental Setup
5
5
Galvanoaluminium samples . . . . . . . . . . . . . . . . . . . . .
5
2.2
Laser and spectroscopic setup . . . . . . . . . . . . . . . . . . . .
5
2.3
Correlation stratigraphy . . . . . . . . . . . . . . . . . . . . . . .
7
2.3.1
Data acquistion and processing . . . . . . . . . . . . . . .
7
2.3.2
Data analysis . . . . . . . . . . . . . . . . . . . . . . . . .
7
M
an
2.1
2.4
Stratigraphic modelling . . . . . . . . . . . . . . . . . . . . . . .
2.5
Surface morphology
. . . . . . . . . . . . . . . . . . . . . . . . .
te
Spectral information and depth profiles . . . . . . . . . . . . . .
10 10
Spectral analysis . . . . . . . . . . . . . . . . . . . . . . .
10
3.1.2
Shape of depth profiles and fitting procedure . . . . . . .
11
3.1.3
3.2
10
3.1.1
Ac ce p
3.1
d
3 Results and Discussion
8
Change of correlation stratigrams with the laser fluence .
14
Evaluation of the stratigraphic measurements . . . . . . . . . . .
15
3.2.1
Comparison of LIBS-stratigrams to EDX cross sections .
15
3.2.2
Fluence dependency of the ablation rate . . . . . . . . . .
16
3.2.3
Wavelength dependency of the ablation rate . . . . . . . .
16
4 Interpretation of the results
17
5 Acknowledgements
19
6 References
20
2
Page 2 of 27
List of Figures SEM-pictures of the Al surface and cross-sections . . . . . . . . .
6
2
Confocal 266 nm LIBS-setup for stratigraphy. . . . . . . . . . . .
6
3
Standard spectra for the Al/Ni/Fe depth-profiles . . . . . . . . .
24
4
3-layer correlation stratigrams and schematics . . . . . . . . . . .
25
5
532 nm Ni and Fe correlation stratigrams . . . . . . . . . . . . .
25
cr
ip t
1
−2
EDX / 532 nm LIBS / 266 nm LIBS / SEM (60 J·cm
) . . . .
26
7
EDX / 532 nm LIBS / 266 nm LIBS / SEM (170 J·cm−2 ) . . . .
26
8
Fluence dependency of ablation rates . . . . . . . . . . . . . . . .
26
9
Comparison of αeff and
. . . . . . . . . . . . . . . . . . . . .
an
1 Lth
us
6
M
List of Tables
26
Parameters used in Equation 2 and Equation 3. . . . . . . . . . .
12
2
Fitting parameters for the stratigrams in Figure 4. . . . . . . . .
14
Ac ce p
te
d
1
3
Page 3 of 27
1. Introduction
ip t
Stratigraphic LIBS-investigations of multi-layer systems offer a minimal invasive non-vacuum method for the determination of the thickness [1, 2] and chemical composition [3, 4, 5] of films and coatings.
The fast accessibility of multi-layer stratigraphic information is of major im-
cr
5
portance for online quality control in industrial production. In this context,
us
the in-depth elemental analysis of multilayer samples in the micrometer thickness range via Laser-Induced Breakdown Spectroscopy (LIBS) offers a fast and non-contact way of thickness and lateral homogeneity-control in quality management. For this, a stratigraphic method using total spectral correlation [6] was
an
10
used to analyse a modern galvanic coating system consisting of aluminium on
M
steel with an intermediate nickel layer. The main focus in this study was set on the investigation of laser parameters, in particular the fluence and wavelength of the ablating nanosecond-laser on the ablation rate and correlation signal as a measure of stratigraphic performance.
te
d
15
1.1. Galvanic aluminium coatings
Ac ce p
Coatings with an aluminium protective agent have advantages: an ability to adhere well to a multitude of metal subtrates and to be passivated and/or
20
converted into durable PEO-layers2 , as well as a re-healing ability in case of damage. Full corrosion protection is provided by layers of about 6 microns and there is no danger of hydrogen incorporation during the plating process. Common layers with 10 to 40 µm are produced at current densities of 1 to 5 A·dm−2 , which can be anodised for excellent corrosion protection, coloured,
25
deformed, ultrasonic-welded, polished and stained [8]. 2 PEO. . . plasma-electrolytically
oxidised, see [7]
4
Page 4 of 27
1.2. Layer thickness and composition analysis
ip t
Since modern electrochemical surface refinement uses a multi-step plating approach in most cases, the first aspect to have an eye on is the accurate and fast
online control of not only one but several chemically-different deposited layers on the respective bulk materials. For this, cross-section preparation, combined
cr
30
with electron microscopy and energy-dispersive X-ray spectroscopy (EDX) or
us
impedance measurements [9, 10] as well as several ion beam methods (mainly for very thin layers in the nm-regime) like ion-beam/TOF-MS or Rutherford backscattering spectrometry (RBS) [11, 12, 13], and (micro-)X-ray fluorescence (µ−XRF) [14] are state-of-the-art methods in industry. The drawbacks of these
an
35
techniques are time-consuming sample preparation and/or the need for highvacuum conditions (EDX) or insufficient thickness-resolution and accuracy for
M
multi-layers (XRF), particularly on structured samples. The introduced method using Laser-induced Breakdown Spectroscopy (LIBS) combined with total spec40
tral correlation avoids these main disadvantages. The only requirement for LIBS
te
d
is a free line of sight to the test-workpiece.
2. Theory and Experimental Setup
Ac ce p
2.1. Galvanoaluminium samples
The samples investigated in this study were standardised carbon-steel test
45
sheets (Rasant Alcotec Beschichtungstechnik GmbH, Germany) with a total sample thickness of about one millimetre, which were coated with 10 µm of
nickel and topped with 30 µm of aluminium (Figure 1a and Figure 1b). 2.2. Laser and spectroscopic setup The sample sheets were cut to 15×15 mm2 pieces and mounted on a mi-
50
crometer positioning stage for precise and reproducible lateral adjustment (x and y direction). A micrometer screw in a precise cage assembly was used to adjust the focus position of the 2” focusing lens (focal length: 92 mm @ 266 nm, 96 mm @ 532 nm, respectively. Figure 2). The same lens (fused silica) was used
5
Page 5 of 27
(b) on the edge
(c) array of laser ablated spots
(d) one ablation site
ip t
(a) at the surface
cr
Figure 1: SEM-pictures of polished cross-sections of the untreated sample (a,b) and of the laser treated surface of the galvanoaluminium sample (c,d). Picture (a), from the top: Carbonepoxy (investment compound, black), galvanoaluminium (dark-grey), nickel (thin, smooth
Scale bars: 20 µm (a), 10 µm (b), 100 µm (c), 10 µm (d).
us
light grey) and carbon steel (porous light-grey at the bottom).
55
an
to collect the emitted light from the plasma tube, following an analogue lens for refocussing the light into a broadband fiber guide to the spectrograph/camerasystem. Due to the short laser wavelength of 266 nm and the fact that most of
M
the emission light is in the same wavelength regime (compare emission spectra, Figure 3), the laser was coupled collinearly into the detection beam path via a small (1/2”) dichroic mirror between the two lenses and adjusted to a Gaussian beam radius (1/e2 ) of ω = 100 µm for both wavelengths, respectively.
d
60
te
Figure 2: Confocal 266 nm LIBS-setup for stratigraphy. ‘2ω’ and ‘4ω’ indicate the frequency duplication and quadruplication, ‘pol’ is a polarizing beamsplitter for energy adjustment,
Ac ce p
respectively.
The 266 nm laser pulses with a pulse duration of 4 ns were generated by
frequency-quadrupling (β-barium oxide, BBO) of the 1064 nm fundamental wavelength of a Nd:YAG laser (Quantel, model Brilliant EaZy), the 532 nm pulses were generated by an analogue setup without second frequency duplica-
65
tion module. Energy-adjustment was performed via a rotatable polarizing beam splitter positioned directly behind the frequency conversion modules. The emitted light was analysed via an ´echelle-spectrograph (LTB, Lasertechnik Berlin GmbH, model Aryelle 200) and detected via an intensified and gated multichannel plate charge coupled device (ICCD) camera (Andor, model ICCD
70
734 84 Gen II). Data acquisition was performed with the standard SophiTM -
6
Page 6 of 27
software from LTB, any further data processing was carried out via special
ip t
FortranTM -routines. 2.3. Correlation stratigraphy 2.3.1. Data acquistion and processing
In order to achieve better signal-to-noise ratio at low fluences, which are
cr
75
necessary to obtain low ablation rates (and therefore high depth-resolution [1]),
us
each analysed spectrum is the result of 25 ablation events (compare [15]). Spectra were recorded at a laser pulse frequency of 0.5 Hz, limited by the camera’s acquisition speed. The gate delay was set to 50 ns for 266 nm and 400 ns for 532 nm, with the gate width set to 1000 ns in both cases. The difference in
an
80
delay times can be attributed mostly to differences in the respective conversion
M
module’s electronics.
Each position in a 5 by 5 grid was ablated at least 40 times, with every single spectrum saved. In order to compensate for possible calibration drift, all spectra were oversampled to a resolution of 5 pm by performing a linear interpolation
d
85
and calculating the resulting I(λ). Spectra corresponding to the same pulse
te
number N at different spots were combined by calculating the mean intensity for each wavelength. This procedure yields 40 spectra for every fluence, which
Ac ce p
were used for correlation analysis.
90
2.3.2. Data analysis
Contrary to single- or multi-line tracing in depth profiling the measured peak
intensities were replaced by a linear correlation coefficient r (Pearson productmoment correlation coefficient, Equation 1) of each spectrum with independently acquired high-resolution spectra of pure aluminium (Figure 3a), nickel
95
(Figure 3c) and iron (Figure 3e). P (Iλ,M − I¯M )(Iλ,S − I¯S ) λ rP r = rP (Iλ,M − I¯M )2 (Iλ,S − I¯S )2 λ
(1)
λ
The correlation coefficient is calculated from the spectral intensities of the measurement Iλ,M and the standard Iλ,S according to Equation 1 for all data 7
Page 7 of 27
points in the spectrum. Here the term in the numerator is the statistical covariance of the two correlated spectral intensities Iλ,M and Iλ,S while the two factors written in the denominator are the standard deviations of the measured spec-
ip t
100
tral intensity Iλ,M of each laser pulse and of the stored standard spectrum Iλ,S ,
cr
respectively (I¯M and I¯S being the arithmetic mean of the spectral intensities).
This method, which is described in detail in [6] and was proved to be a suit-
105
us
able method for metal thin-film analysis in [1, 15] has the advantage of eliminating experimental noise e.g. fluence fluctuations, pulse-to-pulse fluctuations in plasma density and temperature and signals resulting from point impurities
an
in a certain metal layer from the resulting depth profile.
A linear direct relation of the correlation coefficient r and corresponding signal intensity (of a certain element as an ensemble of emission lines) can be assumed, even if the standard spectra’s signal-to-noise-ratio is higher than that
M
110
of the (averaged) single-pulse measurement at a certain N . This method does not require explicit atomic emission line identification and
d
can therefore be used for depth- and ablation analysis of complex layer compo-
115
te
sitions like alloys or other microcrystalline, homogeneous mixtures, as long as a reference spectrum of a certain material is available [6]. The method theoret-
Ac ce p
ically gives values of r ranging from -1 (perfect anticorrelation) to +1 (perfect correlation). Since the majority of the data points in every spectrum is white noise, values of r far below zero are very unlikely (compare Figure 4).
120
2.4. Stratigraphic modelling There are multiple ways to obtain stratigraphic information from LIBS raw-
data. Many of them have been developed for analytical and industrial contexts [5, 16, 17]. The easiest way to map elements through (thick and homogeneous)
layers is to compare well known line intensities determined from pure standard 125
materials while consecutively ablating material pulse-by-pulse. This approach can be stretched to its limits, when elements with a multitude of low-intensity lines should be traced in one single experiment. 8
Page 8 of 27
Statistical approaches were taken e.g. for industrial quality control, where the amount of sample material is not a limiting factor. Balzer and coworkers presented a method, where bursts of laser-pulses are used to control the film-
ip t
130
thickness of galvanised (Zn) steel by simply comparing the ratio of zinc over
cr
iron emission, using many known lines of each element, respectively.
This method averages out one of the biggest problems in LIBS-stratigraphy:
135
us
The pulse-to-pulse fluctuations in laser fluence and plasma emission behaviour [17]. However, an exact depth profile can not be obtained this way, only the yes/no information of penetration and the layer-thickness in a certain confidence
an
interval can be gained by calibrating the system with standard samples of known layer thickness.
The third way of obtaining depth information from LIBS-emission data uses total spectral information instead of single emission line intensities by calculat-
M
140
ing the linear correlation coefficient of a single-pulse spectrum with a standard spectrum of a pure element or alloy to be traced through the layers as described
d
in the last section. This method, initially introduced for LIBS by Mateo et al in
145
te
2001 [1] results in comparatively smooth correlation stratigrams, which represent the similarity of a certain spectrum to its standard at each pulse number N
Ac ce p
and offers the possibility of developing fitting functions and models of ablation behaviour in a novel way. Since this technique was optimized and taken further in this study, results and achievements should be discussed in detail later on. A major issue in stratigraphy is, besides recording a reliable element-vs.-
150
pulse number (or depth) signal with high accuracy, the description of elemental in-depth traces by means of suitable mathematical functions which allow the quantification or prediction of characteristics in the signal with simple calculations (e.g. surface contaminations by debris [15]). Compositional depth profiling was treated empirically with respect to the
155
irradiance dependence [1, 6] and was theoretically modelled in detail [18]. Both the empirical and the theoretical approach have been successfully applied to layers with thicknesses much higher compared to the average ablation rates of the constituent materials. However, for thinner layers the proposed models 9
Page 9 of 27
either yield no good fit to obtained experimental data (in the case of the method 160
proposed in [6]) or require substantial manual intervention (smoothing) in the
ip t
fitting process [18], which is not desirable for a model intended for simple,
automated, widespread, everyday application. Therefore, the development of
following the first approach for a two-layer system in [15]. 2.5. Surface morphology
us
165
cr
a generalised simple mathematical description was another aim of this work,
The shape, diameter and depth of laser ablated craters was estimated via op-
an
tical (Olympus, 93 model STM-MJS) and confocal laser scanning (Zeiss, model LSM 800) as well as scanning electron microscopy (Zeiss, model 94 Supra 55 VP), see Figure 1c and Figure 1d. Energy dispersive X-ray spectroscopic line scans were performed with the Zeiss electron microscope as well (30 s per spot,
M
170
5 keV).
d
3. Results and Discussion
te
3.1. Spectral information and depth profiles 3.1.1. Spectral analysis
Figure 3 shows the standard spectra of the three main constituent elements
Ac ce p
175
of the respective layers (left) which were used for the correlation analysis in all subsequent stratigrams. The spectra were recorded using metal foils of the highest available purity (Al, Ni, Aldrich >99.99 %, respectively) and polished
bulk material (carbon steel, “Fe”).
180
On the right side of Figure 3 three spectra at different pulse numbers N , and therefore different depths are shown. It can clearly be seen that the spectrum recorded on top of the sample (N = 3) shows tremendous similarity to the aluminium standard, at N = 10 a mixed Al/Ni-spectrum is obtained and at a pulse number of N = 18 it is mainly the spectrum of the bulk material.
185
10
Page 10 of 27
3.1.2. Shape of depth profiles and fitting procedure When the three-layer sample is ablated by consecutive laser pulses, the col-
ip t
lected emission spectra contain all information necessary for estimating the breach of layer interfaces in terms of pulse number N . Because of the non-
rectangular cross-section of ablation craters, the calculated elemental correla-
cr
190
tion coefficient does not change abruptly when the centre of the ablation crater
us
reaches an interface. Due to the Gaussian ablation profile, where a significant portion of the ablation and plasma formation takes place on the flanks of the Gaussian fluence distribution, the stratigrams show a gradual rising and decaying behaviour rather than sharp changes. Figure 4 shows a stratigram of
an
195
the galvanoaluminium-nickel-steel system calculated with the standard spectra in Figure 3 and 50 separately averaged (25 spots) spectra, each recorded at a
M
certain N . The three plotted datasets stem from correlation of the single N spectra with Al-, Ni- and Fe-standard spectra, respectively. 200
Several characteristics can be noticed in the stratigrams which should be dis-
d
cussed in the following paragraphs. All parameters are described in Table 1.
Ac ce p
te
i rji (N ) = Ai 1 − exp −kji (N − Nj−1,j ) − aN 2 + cij " i #2 i k + N − N j (j−1,j) rji (N ) = Ai exp 1 − − aN 2 + cij kji
(2)
(3)
Equation 2 shows an empirical fitting function for the case of increase of correlation coefficient rji (N ) of of the acquired spectra at pulse number N with
205
the standard spectrum of substance i in the layer j (counting from the surface), while Equation 3 is used for the description of a decay of ri (N ) with N (usually in another layer j). Generally, an increase (and, for thick layers, a constantly high r following it), is observed in a layer which actually contains the analysed compound, while a decay can be observed after such a layer has been passed
210
through by the ablation. The scaling factor Ai is used to describe the relative intensity of rji (N ) (for all layers j, with respect to the noise-baseline level ci of the first baseline). The term in curly brackets describes the exponential rise/decay by the use of a decay constant kji and the number of pulses N . The 11
Page 11 of 27
Table 1: Parameters used in Equation 2 and Equation 3.
meaning
ip t
parameter
cr
i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . element or compound
j . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . number of layer (from outermost)
us
Ai . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . amplitude of fit/linear scaling factor
kji . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . rise- or decay-constant
an
i Nj,j+1 . . . . . . . . . . Position of the interface between layers j and j + 1
cj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . cumulative baseline shifts for fit
M
a . . . . . . . . square decay parameter (not observed in these experiments)
in terms of N . The term −aN 2 is used as a general damping term, describ-
te
215
d
i parameter Nj−1,j finally gives the layer boundary between layer j − 1 and j
ing the overall decay of all signals with high N which may be attributed to
Ac ce p
plasma and vapour shielding effects [15, 19]. This effect was not observed in this study due to the low overall experimental profiling depth/diameter-ratio (aspect ratio) of the produced cavities, but it should be mentioned as a part of
220
our routine for reasons of completeness. For a three-layer system consisting of three different homogeneous substances (pure element or alloy), there is exactly one characteristic amplitude of the correlation coefficient r, representing the highest achievable correlation of that substance with the corresponding standard. It is described by Ai , with superscript i being the element (Al, Ni, Fe),
225
see also Table 1. Furthermore, each substance i has at least one rise and one decay constant (kji , which quantifies the signal behaviour of the correlation coefficient of the element i in the layer j, counting from top). A certain decay of correlation coefficient r
12
Page 12 of 27
(Fe)
(Ni)
with pulse number N can happen before (iron: k1 of cross-contamination [15]) or after (aluminium: nickel, the rise
(Ni) (k2 )
(Al) k2 )
(Ni)
. The initial rise in Al-
is a direct cause of the cross-contamination too, compare [15]
cr
correlation
the rise. In the case of
is followed by a second decay due to its intermediate-layer
character, described by the second decay constant k3 (Al) k1
- because
ip t
230
and nickel: k1
(Fe)
and the rise of iron-correlation when laser-drilling into the bulk material k3 is obvious.
us
235
N12 and N23 indicate the calculated layer-breach in terms of N for the fitting functions of each substance between the layers 1 and 2 (Al and Ni) and 2 and
an
3 (Ni and Fe), respectively. The value of c1 is the baseline-offset in correlation coefficient. For Ni, c3 describes an additional baseline increase (c = c1 + c3 ) 240
observed after breaching the layer.
M
Note that the fitting models applied here will not yield all parameters of interest for all elements. For example, it is not possible to determine N12 from the Fe
Ac ce p
te
d
curve. These cases are indicated by a “-” in Table 2.
13
Page 13 of 27
Al
Ni
Fe
Ai
0.80 ± 0.011
0.46 ± 0.018
0.693 ± 0.0053
k1i
1.04 ± 0.043
1.51 ± 0.058
2.17 ± 0.085
k2i
34 ± 1.1
0.20 ± 0.016
-
k3i
-
39 ± 2.4
i N12
15.8 ±0.18
13.93 ± 0.062
i N23
-
22.8 ± 0.19
22.6 ± 0.93
c1
0.11 ± 0.010
0.027 ± 0.0017
0.047 ± 0.0031
c3
-
0.05 ± 0.011
-
us
cr
i
ip t
Table 2: Fitting parameters for the stratigrams in Figure 4.
0.183 ± 0.0059
d
M
an
-
245
te
3.1.3. Change of correlation stratigrams with the laser fluence The correlation coefficients as a function of N are plotted in Figure 5. The
Ac ce p
positions of the layer breaches, defined by the increase of the Ni and Fe signal respectively, are clearly visible. Two effects of increasing fluence should be mentioned. First, the number of pulses required to breach the top layer decreases with increasing fluence. This corresponds to an increase in the ablation rate
250
(which we will hereinafter refer to as h), as would be expected. Second, the
maximum r increases with the fluence as well, signifying an improved signalto-noise ratio in the individual spectra. For efficient stratigraphy, therefore a balance of ablation rate and signal-to-noise ratio has to be found for practical routine deph profiling.
14
Page 14 of 27
255
3.2. Evaluation of the stratigraphic measurements
ip t
3.2.1. Comparison of LIBS-stratigrams to EDX cross sections Figure 6 and Figure 7 show a comparison of the depth profiles gathered using
LIBS at both laser wavelengths and two different fluence regimes. In order to
260
cr
obtain a reference for the layer structure, a cross-section of the sample was prepared by embedding a part of the sample sheet and analysing the polished
us
cross-section in an electron microscope.
The EDX scan (top) is the result of 5 consecutive line scans, performed at different lateral positions across the layers. Due to their similarity, the resulting
265
an
data points were combined into one graph. It should be noted that, since the EDX-system was not calibrated for quantitative analysis of the respective elements, the exact values obtained for the atomic weight fraction of each element
M
are not trustworthy. Nevertheless, the qualitative presence of each element as a function of position (and thus depth d under the original surface) could be easily determined by fitting sigmoid switch functions (based on the error function erf(d)) to the values. The width of these steps and, as a consequence, the width
d
270
te
of the sigmoid fit yield the uncertainty of the layer thickness determination, mainly resulting from the lateral resolution of the EDX measurement.
Ac ce p
Considering the LIBS depth profiles, shown as r(d), several effects can be observed: First, both wavelengths show an increase in ablation rate with increasing
275
fluence, as can be seen from the increasing distance between the points. Second, evidence of surface contamination with iron and nickel can be seen at the very beginning of the pulse-by-pulse ablation at the left side in each LIBS-stratigram [15]. This results from the re-deposition of material on the surface during the drilling of one point, which produces an additional very thin top layer observed
280
during the measurement of subsequent spots. In the results, it is visible as a lower r(Al) at the first and second laser pulse, accompanied by higher r(Ni) and r(Fe) values. For both fluence regimes, the UV laser yielded higher ablation rates than the 532 nm in the Al layer, resulting in fewer pulses needed to reach the onset
15
Page 15 of 27
285
of the Ni signal. Conversely, this effect could not be observed in the Ni layer itself, where the ablation rate was found to depend only on the fluence. An
determining the depth in this layer could be applied.
290
cr
A graph of all observed ablation rates h is shown in Figure 8.
ip t
analysis of the Fe ablation rate was not performed, since no reliable means of
3.2.2. Fluence dependency of the ablation rate
us
The ablation rate h was determined as the thickness of a given layer, divided Ni Fe by the number of pulses required to drill through that layer (N1,2 or N2,3 −
an
Ni N1,2 , respectively). As expected, it was found to increase with the logarithmic
fluence, as predicted in [20]. The plots of h vs ln(F ) shown in Figure 8 allow 295
for the calculation of the effective absorption coefficients αeff , which are shown
M
in Figure 9 (for calculation procedure see section 4 or [20]). 3.2.3. Wavelength dependency of the ablation rate
d
Figure 8 and Figure 9 show that h(F ) for Ni appears to be independent of the wavelength used, whereas h(F ) for Al was substantially higher when using the UV laser. The two independent measurement series yielded consistent results
te
300
within the measurement precision. The average ablation rate h as a function of
Ac ce p
fluence F for Ni ranges from 1–3.5 µm/pulse for λ = 266 nm as well as for λ = 532 nm. In contrast, the range of h for Al differs from 2–4 µm/pulse for λ =
532 nm and 4–8 µm/pulse for λ = 266 nm in the exact same fluence range on
305
the same sample.
Comparison of αeff with the calculated values for the inverse thermal diffu-
sion length 1/Lth shows that the latter does serve as a very rough prediction for all the Ni values of αeff , as well as for those of Al at 266 nm. The deviation Al of the Al ablation rates and therefore αeff with changing wavelength can not be
310
explained by this simple model.
16
Page 16 of 27
4. Interpretation of the results
ip t
Laser-material interaction in the nanosecond pulse regime has long been discussed by a multitude of different physical, but also more or less empirical approaches, describing the involved processes from different points of view [20, 21, 22, 23, 24, 25].
cr
315
Let us compare laser ablation of metals as a simple thermal process with
us
no attenuation of the laser beam in the gas phase i.e. when all the energy is deposited and absorbed at the very surface of the metal: The ablation rate h(F )
(4)
whenever the ablation of a solid takes place with a gaussian laser beam [20].
M
320
an
(ablated depth per pulse) is expected to follow the logarithmic law 1 F h(F ) = ln αeff Fth
Here αeff is the effective absorption coefficient. While this does follow the same mathematical form as the Lambert-Beer absorption coefficient α, it has both
d
different physical background and different numerical values. For nanosecond lasers on metal, these values are usually considerably lower, implying deeper penetration into the material than mere optical absorption would account for.
te
325
The reason for this behaviour is expected to be thermal diffusion; consequently,
Ac ce p
one approximate prediction for αeff is the thermal diffusion length Lth , resulting from solving the linear heat equation for the given system. Its calculation and relation to αeff are as follows:
330
√ 1 ∝ Lth = 2κτ αeff
(5)
where κ is the thermal diffusivity of the material, and τ is the laser pulse length [26].
Differences in the behaviour of Ni and Al can be observed by comparing
the effective absorption coefficients of Ni and Al (datapoints, calculated with Equation 4 from LIBS-stratigrams) to the calculated inverse of the thermal dif335
fusion length (dashed lines, calculated from laser- and material-parameters with Equation 5) as plotted in Figure 9.
17
Page 17 of 27
While the effective absorption coefficient correlates well for Ni in all four experimental series, two performed separately with green light and two with UV laser
340
ip t
pulses, the ablation behaviour of aluminium only shows pure thermal energy transfer in the experiments with UV light. At 532 nm the effective absorption
cr
coefficient strongly deviates from the values predicted by the thermal diffusion theory.
us
Since there are no significant differences in optical absorptivity, reflectivity, or other optical properties of solid Ni and Al in both wavelength regimes, nor any 345
observable changes in surface roughness or morphology, the influence of surface
an
phenomena on energy transfer from light into the material can be excluded. The cause for the deviations is likely to be looked for in a combination of effects: On the one hand, it may be necessary to apply a more complex
350
M
thermodynamic model capable of dealing with the high-energy environment the sample is exposed to. Mechanisms such as explosive boiling [27, 28] are likely to influence the ablation rate. Based on the results collected by Miotello et
d
al., it is to be assumed that a phase explosion occurs throughout the applied
te
fluence range. The lack of discontinuities in the h vs. ln(F ) curves indicates the presence of one constant mechanism of ablation. However, this thermodynamic model does not directly imply wavelength-dependent effects.
Ac ce p
355
On the other hand, one has to consider the possible influence of plasma
shielding, which would have to be element-specific in the case presented: Hot
expanding plasma can physically or chemically modify or ablate bigger surface areas, than those confined by a Gaussian laser spot, laterally and in-depth.
360
Apart from that, it can also scatter and absorb the incoming laser light during the nanosecond-pulse [29, 30, 31]. This pressure-dependent effect describes the partial absorption of the nanosecond laser-pulse by the expanding plasma and vapour during the pulse, which obviously is not observed in femtosecond-laser
ablation where laser-material interaction takes place before the plasma forma365
tion process. Furthermore, and this is an especially big issue in spectroscopy, the characteristic light, emitted by the hot plasma in the confinement of the drilled crater, is partially self-absorbed and scattered in the outer parts of the plasma 18
Page 18 of 27
plume by the expanding plasma itself (characteristic self-absorption of Dopplerbroadened emission lines), cold vapour and re-condensed particles [6, 19]. Nevertheless, more detailed studies of these processes, such as time-resolved
ip t
370
imaging of the evolving plasma plume and variation of the pulse duration, as
cr
shown in [31] have to be performed to shed light upon this wavelength-dependent and element-specific effect. Experiments including both ablation and trans-
375
us
mission measurements of the laser-induced plasma itself by frequency-tunable (OPO) nanosecond laser beams will offer the opportunity to investigate the temporal and spatial (perpendicularly delayed double-pulse measurements) trans-
an
parency of the evolving plasma as a function of wavelength for a wide range of materials including pure metals and alloys.
380
M
Summary
Laser ablation behaviour of the Al-Ni-Fe layer system was evaluated by empirical fitting functions of elemental emission traces. Thus, consecutive pulse-by-
d
pulse laser ablation was recorded with different fluences and wavelengths. The
te
thermal diffusion length, as presented in this study, gives a good first approximation for the ablation rate as a function of F . In particular, the logarithmic 385
description of h(F ) has been found reliable. The experimental data of nickel
Ac ce p
ablation with both UV and visible laser light agree well with a simple heat diffusion model. The ablation behaviour of aluminium at visible wavelength deviates strongly from this model.
5. Acknowledgements
390
The authors would like to thank Stephan Puchegger (Faculty Centre for
Nanostructure Research, University of Vienna) for his invaluable assistance with the scanning electron microscopy and EDX-spectroscopy and James Bedillion for fruitful discussions and corrections, as well as Rasant Alcotec GmbH for the generous supply of galvanoaluminium samples.
19
Page 19 of 27
395
6. References
ip t
[1] M. P. Mateo, J. M. Vadillo, J. J. Laserna, Irradiance-dependent depth profiling of layered materials using laser-induced plasma spectrometry, Journal
cr
of Analytical Atomic Spectrometry 16 (11) (2001) 1317–1321.
[2] A. Jurado-Lopez, M. D. Luque de Castro, Laser-induced breakdown spectrometry in the jewellery industry. part i. determination of the layer thick-
us
400
ness and composition of gold-plated pieces, Journal of Analytical Atomic Spectrometry 17 (5) (2002) 544–547.
an
[3] L. M. Cabalin, J. J. Laserna, Surface stoichiometry of manganin coatings prepared by pulsed laser deposition as described by laser-induced breakdown spectrometry, Analytical Chemistry 73 (6) (2001) 1120–1125.
M
405
[4] A. Jurado-Lopez, M. D. Luque de Castro, Laser-induced breakdown spectrometry in jewellery industry. part ii: Quantitative characterisation of
d
goldfilled interface, Talanta 59 (2) (2003) 409–415.
410
te
[5] D. G. Papazoglou, V. Papadakis, D. Anglos, In situ interferometric depth and topography monitoring in libs elemental profiling of multi-layer struc-
Ac ce p
tures, Journal of Analytical Atomic Spectrometry 19 (4) (2004) 483–488. [6] M. P. Mateo, G. G. Nicolas, V. Pi˜ non, A. Ya˜ nez, Improvements in depthprofiling of thick samples by laser-induced breakdown spectroscopy using linear correlation, Surface and Interface Analysis 38 (5) (2006) 941–948.
415
[7] T. O. Nagy, U. Pacher, A. Giesriegl, L. Soyka, G. Trettenhahn, W. Kautek, Laser-induced electrochemical de- and repassivation investigations on plasma-oxidized aluminium alloys, Applied Surface Science 302 (2014) 184– 188. [8] R. D¨ otzer, Galvano-aluminium und seine anodische oxidation galvano-al-
420
eloxal-schichten - eine neue oberfl¨ache, Chemie Ingenieur Technik 45 (9-10) (1973) 653–658.
20
Page 20 of 27
[9] A. P. Abbott, S. Nandhra, S. Postlethwaite, E. L. Smith, K. S. Ryder, Electroless deposition of metallic silver from a choline chloride-based ionic
425
ip t
liquid: A study using acoustic impedance spectroscopy, SEM and atomic force microscopy, Phys. Chem. Chem. Phys. 9 (2007) 3735–3743.
cr
[10] E. P. Gusev, M. Copel, E. Cartier, I. J. R. Baumvol, C. Krug, M. A. Gribelyuk, High-resolution depth profiling in ultrathin Al2O3 films on Si,
us
Applied Physics Letters 76 (2) (2000) 176–178.
[11] C. Jeynes, N. P. Barradas, P. K. Marriott, G. Boudreault, M. Jenkin, E. Wendler, R. P. Webb, Elemental thin film depth profiles by ion beam
an
430
analysis using simulated annealing - a new tool, Journal of Physics D: Applied Physics 36 (7) (2003) R97.
M
[12] J.-C. Dran, J. Salomon, T. Calligaro, P. Walter, Ion beam analysis of art works: 14 years of use in the louvre, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms
d
435
219-220 (0) (2004) 7 – 15, proceedings of the Sixteenth International Con-
te
ference on Ion Beam Analysis.
[13] C. Neelmeijer, W. Wagner, H. Schramm, Depth resolved ion beam analysis
Ac ce p
of objects of art, Nuclear Instruments and Methods in Physics Research
440
Section B: Beam Interactions with Materials and Atoms 118 (1-4) (1996) 338 – 345, ion Beam Analysis.
[14] R. Dargel, M. Azeroual, B. Mogwitz, J. Janek, C. Vogt, Non-destructive analysis of silver selenide films obtained by pulsed laser deposition (pld) with micro-xrf, Journal of Materials Science 42 (17) (2007) 7375–7380.
445
[15] T. O. Nagy, U. Pacher, H. P¨ohl, W. Kautek, Atomic emission stratigraphy by laser-induced plasma spectroscopy: Quantitative depth profiling of metal thin film systems, Applied Surface Science 302 (2014) 189–193.
21
Page 21 of 27
[16] I. Bozs´ oki, B. Balogh, P. Gordon, 355 nm nanosecond pulsed nd:yag laser profile measurement, metal thin film ablation and thermal simulation, Optics & Laser Technology 43 (7) (2010) 1212–1218.
ip t
450
[17] H. Balzer, S. H¨ olters, V. Sturm, R. Noll, Systematic line selection for on-
cr
line coating thickness measurements of galvanised sheet steel using libs, Analytical and Bioanalytical Chemistry 385 (2) (2006) 234–239.
455
us
[18] L. St-Onge, A mathematical framework for modeling the compositional depth profiles obtained by pulsed laser ablation, Journal of Analytical
an
Atomic Spectrometry 17 (9) (2002) 1083–1089.
[19] R. E. Russo, X. Mao, S. S. Mao, The physics of laser ablation in micro-
M
chemical analysis, Analytical Chemistry 74 (3) (2002) 70A–77A. [20] J. Kr¨ uger, W. Kautek, Ultrashort pulse laser interaction with dielectrics 460
and polymers, Adv. Polym. Sci. 168 (2004) 247–289.
d
[21] Z. Y. Wang, J. T. Liu, D. M. Hirak, D. C. Weckman, H. W. Kerr, Deter-
te
mining the spot size and gaussian distribution coefficient of pulsed laser beams using kapton films, Journal of Laser Applications 5 (1) (1993) 5–12.
Ac ce p
[22] Y. Jee, M. F. Becker, R. M. Walser, Laser-induced damage on single-crystal
465
metal surfaces, J. Opt. Soc. Am. B 5 (3) (1988) 648–659.
[23] A. Bogaerts, Z. Chen, R. Gijbels, A. Vertes, Laser ablation for analytical sampling: What can we learn from modeling?, Spectrochimica Acta - Part B Atomic Spectroscopy 58 (11) (2003) 1867–1893.
[24] H. Howard, A. J. Conneely, G. M. O’Connor, T. J. Glynn, Investigation of
470
a method for the determination of the focused spot size of industrial laser beams based on the drilling of holes in mylar film (2003) 541–552.
[25] C. Arag´ on, J. A. Aguilera, Characterization of laser induced plasmas by optical emission spectroscopy: A review of experiments and methods, Spectrochimica Acta Part B: Atomic Spectroscopy 63 (9) (2008) 893–916. 22
Page 22 of 27
475
[26] E. Matthias, M. Reichling, J. Siegel, O. W. Kading, S. Petzoldt, H. Skurk, P. Bizenberger, E. Neske, The influence of thermal-diffusion on laser-
ip t
ablation of metal-films, Applied Physics a-Materials Science & Processing 58 (2) (1994) 129–136.
480
cr
[27] A. Miotello, R. Kelly, Critical assessment of thermal models for laser sputtering at high fluences, Applied Physics Letters 67 (24) (1995) 3535–3537.
us
[28] A. Miotello, R. Kelly, Laser-induced phase explosion: new physical problems when a condensed phase approaches the thermodynamic critical tem-
an
perature, Applied Physics A 69 (1) (1999) S67–S73.
[29] J. M. Vadillo, J. M. Fern´andez Romero, C. Rodr´ıguez, J. J. Laserna, Depth485
resolved analysis by laser-induced breakdown spectrometry at reduced pres-
M
sure, Surface and Interface Analysis 26 (13) (1998) 995–1000. [30] J. M. Vadillo, J. M. Fern´andez Romero, C. Rodr´ıguez, J. J. Laserna, Effect
d
of plasma shielding on laser ablation rate of pure metals at reduced pressure,
490
te
Surface and Interface Analysis 27 (11) (1999) 1009–1015. [31] A. Semerok, C. Chal´eard, C.ard, V. Detalle, J. L. Lacour, P. Mauchien,
Ac ce p
P. Meynadier, C. Nouvellon, B. Sall´e, B., P. Palianov, M. Perdrix, G. Petite, Experimental investigations of laser ablation efficiency of pure metals with femto, pico and nanosecond pulses, Applied Surface Science 138-139 (0) (1999) 311–314.
23
Page 23 of 27
ip t cr us an
(b) N =3
M
(a) standard spectrum of aluminium
d
(c) standard spectrum of nickel
(f) N =18
te
(e) standard spectrum of iron
(d) N =10
Figure 3: Standard spectra (left) used for the Al/Ni/Fe correlation stratigrams, and LIBS-
Ac ce p
spectra (right), recorded on the sample at different pulse numbers (average of 25 neighbouring experiments). F = 220 J·cm−2 , spot distance 0.5 mm, λ = 532 nm.
24
Page 24 of 27
ip t cr us an M d te
Ac ce p
Figure 4: Top: 3-layer correlation stratigrams calculated from averaged (25 spots) single-pulse spectra and standard spectra (Figure 3, left). The fit–parameters for the stratigraphic model functions (general rise- Equation 2 and decay-function Equation 3) are explained in the text and listed in Table 1 and Table 2.
: aluminium, 4: nickel, 3: iron. F = 60 J·cm−2 , λ = 532 nm.
Bottom: Schematic illustration of layers (Al: yellow, Ni: green) and bulk (Fe: grey) material with a Gaussian nanosecond-laser ablation profile. i,ii: only top layer material is ablated and only its correlation signal is high. iii,iv: the spectra show a finite correlation for both, layer and bulk material, which changes with depth (pulse number N ).
Figure 5: Calculated 532 nm Ni (top) and Fe (bottom) correlation stratigrams, measured at different fluences: Nickel: ◦ : 80 J · cm−2 , 2 : 110 J · cm−2 , 3 : 170 J · cm−2 , 4 : 220 J · cm−2 . Iron: ◦ : 80 J · cm−2 , 2 : 110 J · cm−2 , 3 : 170 J · cm−2 , 4 : 220 J · cm−2 .
25
Page 25 of 27
ip t
Figure 6: Comparison of EDX cross-section line scan data (top) to LIBS-correlation strati-
grams recorded with 532 (middle) and 266 nm (bottom) at a maximum fluence of 60 J·cm−2 .
◦: aluminium, 2: nickel, 3: iron, 4: carbon. A backscatter image is plotted below for
an
us
cr
comparison.
Figure 7: Comparison of EDX cross-section line scan data (top) to LIBS-correlation stratigrams recorded with 532 (middle) and 266 nm (bottom) at a maximum fluence of 170 J·cm−2 .
M
◦: aluminium, 2: nickel, 3: iron, 4: carbon. A backscatter image is plotted below for com-
te
d
parison.
Ac ce p
Figure 8: Fluence dependency of ablation rates for two different experimental series (left and right). Top: 532 nm, Bottom: 266 nm. The slope in these plots gives the effective absorption coefficient αeff in Figure 9. ◦: Al, 3: Ni.
Figure 9: Effective absorption coefficient αeff , ◦: Al, 3: Ni, versus inverse thermal diffusion length 1/Lth (dashed lines, Ni: purple, Al: blue) for the two experimental series in Figure 8.
26
Page 26 of 27
Highlights:
ip t cr us an
iv)
M
iii)
d
ii)
Depth profiling via nanosecond-LIBS was performed on a galvanic 3-layer coating system via automated fitting routines with green (532 nm) and ultraviolet (266 nm) light. The developed stratigraphic model holds for both wavelengths and all involved layer materials. The change of the ablation wavelength from VIS (532 nm) to UV (266 nm) affects the ablation rate (= in-depth resolution) of the Al top layer but not of the Ni intermediate layer. A pure heat diffusion model can not explain the ablation process in Al. Laser-plasma interaction needs to be considered.
Ac ce pt e
i)
Page 27 of 27
S0169-4332(16)32776-3 http://dx.doi.org/doi:10.1016/j.apsusc.2016.12.059 APSUSC 34606
To appear in:
APSUSC
Received date: Revised date: Accepted date:
30-6-2016 15-11-2016 8-12-2016
Please cite this article as: T.O. Nagy, U. Pacher, A. Giesriegl, M.J.J. Weimerskirch, W. Kautek, Depth Profiling of GalvanoaluminiumndashNickel Coatings on Steel by UV- and VIS-LIBS, (2016), http://dx.doi.org/10.1016/j.apsusc.2016.12.059 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t
Depth Profiling of Galvanoaluminium–Nickel Coatings on Steel by UV- and VIS-LIBS Nagy, T. O.1 , Pacher, U., Giesriegl, A., Weimerskirch, M. J. J., Kautek, W.
us
cr
Universit¨ at Wien, Institut f¨ ur Physikalische Chemie W¨ ahringer Str. 42, 1090 Wien
Abstract
an
Laser-induced depth profiling was applied to the investigation of galvanised steel sheets as a typical modern multi-layer coating system for environmental corrosion protection. The samples were ablated stepwise by the use of two different
M
wavelengths of a frequency-converted Nd:YAG-laser, 266 nm and 532 nm, with a pulse duration of τ = 4 ns at fluences ranging from F = 50 – 250 J·cm−2 . The emission light of the resulting plasma was analysed as a function of both pen-
d
etration depth and elemental spectrum in terms of linear correlation analysis.
te
Elemental depth profiles were calculated and compared to EDX-cross sections of the cut sample. A proven mathematical algorithm designed for the reconstruc-
Ac ce p
tion of layer structures from distorted emission traces caused by the Gaussian ablation profile can even resolve thin intermediate layers in terms of depth and thickness. The obtained results were compared to a purely thermally controlled ablation model. Thereby light-plasma coupling is suggested to be a possible cause of deviations in the ablation behaviour of Al. The average ablation rate h as a function of fluence F for Ni ranges from 1–3.5 µm/pulse for λ = 266 nm as well as for λ = 532 nm. In contrast, the range of h for Al differs from 2–4 µm/pulse for λ = 532 nm and 4–8 µm/pulse for λ = 266 nm in the exact same
fluence range on the exact same sample. Keywords: LIBS stratigraphy; galvanic coatings; ablation rate; plasma shielding; light-plasma-interaction; depth profiling 1 Correspondence:
[email protected]
Preprint submitted to Applied Surface Science
December 19, 2016
Page 1 of 27
ip t
Contents 1 Introduction
4
Galvanic aluminium coatings . . . . . . . . . . . . . . . . . . . .
1.2
Layer thickness and composition analysis . . . . . . . . . . . . .
4
cr
1.1
us
2 Theory and Experimental Setup
5
5
Galvanoaluminium samples . . . . . . . . . . . . . . . . . . . . .
5
2.2
Laser and spectroscopic setup . . . . . . . . . . . . . . . . . . . .
5
2.3
Correlation stratigraphy . . . . . . . . . . . . . . . . . . . . . . .
7
2.3.1
Data acquistion and processing . . . . . . . . . . . . . . .
7
2.3.2
Data analysis . . . . . . . . . . . . . . . . . . . . . . . . .
7
M
an
2.1
2.4
Stratigraphic modelling . . . . . . . . . . . . . . . . . . . . . . .
2.5
Surface morphology
. . . . . . . . . . . . . . . . . . . . . . . . .
te
Spectral information and depth profiles . . . . . . . . . . . . . .
10 10
Spectral analysis . . . . . . . . . . . . . . . . . . . . . . .
10
3.1.2
Shape of depth profiles and fitting procedure . . . . . . .
11
3.1.3
3.2
10
3.1.1
Ac ce p
3.1
d
3 Results and Discussion
8
Change of correlation stratigrams with the laser fluence .
14
Evaluation of the stratigraphic measurements . . . . . . . . . . .
15
3.2.1
Comparison of LIBS-stratigrams to EDX cross sections .
15
3.2.2
Fluence dependency of the ablation rate . . . . . . . . . .
16
3.2.3
Wavelength dependency of the ablation rate . . . . . . . .
16
4 Interpretation of the results
17
5 Acknowledgements
19
6 References
20
2
Page 2 of 27
List of Figures SEM-pictures of the Al surface and cross-sections . . . . . . . . .
6
2
Confocal 266 nm LIBS-setup for stratigraphy. . . . . . . . . . . .
6
3
Standard spectra for the Al/Ni/Fe depth-profiles . . . . . . . . .
24
4
3-layer correlation stratigrams and schematics . . . . . . . . . . .
25
5
532 nm Ni and Fe correlation stratigrams . . . . . . . . . . . . .
25
cr
ip t
1
−2
EDX / 532 nm LIBS / 266 nm LIBS / SEM (60 J·cm
) . . . .
26
7
EDX / 532 nm LIBS / 266 nm LIBS / SEM (170 J·cm−2 ) . . . .
26
8
Fluence dependency of ablation rates . . . . . . . . . . . . . . . .
26
9
Comparison of αeff and
. . . . . . . . . . . . . . . . . . . . .
an
1 Lth
us
6
M
List of Tables
26
Parameters used in Equation 2 and Equation 3. . . . . . . . . . .
12
2
Fitting parameters for the stratigrams in Figure 4. . . . . . . . .
14
Ac ce p
te
d
1
3
Page 3 of 27
1. Introduction
ip t
Stratigraphic LIBS-investigations of multi-layer systems offer a minimal invasive non-vacuum method for the determination of the thickness [1, 2] and chemical composition [3, 4, 5] of films and coatings.
The fast accessibility of multi-layer stratigraphic information is of major im-
cr
5
portance for online quality control in industrial production. In this context,
us
the in-depth elemental analysis of multilayer samples in the micrometer thickness range via Laser-Induced Breakdown Spectroscopy (LIBS) offers a fast and non-contact way of thickness and lateral homogeneity-control in quality management. For this, a stratigraphic method using total spectral correlation [6] was
an
10
used to analyse a modern galvanic coating system consisting of aluminium on
M
steel with an intermediate nickel layer. The main focus in this study was set on the investigation of laser parameters, in particular the fluence and wavelength of the ablating nanosecond-laser on the ablation rate and correlation signal as a measure of stratigraphic performance.
te
d
15
1.1. Galvanic aluminium coatings
Ac ce p
Coatings with an aluminium protective agent have advantages: an ability to adhere well to a multitude of metal subtrates and to be passivated and/or
20
converted into durable PEO-layers2 , as well as a re-healing ability in case of damage. Full corrosion protection is provided by layers of about 6 microns and there is no danger of hydrogen incorporation during the plating process. Common layers with 10 to 40 µm are produced at current densities of 1 to 5 A·dm−2 , which can be anodised for excellent corrosion protection, coloured,
25
deformed, ultrasonic-welded, polished and stained [8]. 2 PEO. . . plasma-electrolytically
oxidised, see [7]
4
Page 4 of 27
1.2. Layer thickness and composition analysis
ip t
Since modern electrochemical surface refinement uses a multi-step plating approach in most cases, the first aspect to have an eye on is the accurate and fast
online control of not only one but several chemically-different deposited layers on the respective bulk materials. For this, cross-section preparation, combined
cr
30
with electron microscopy and energy-dispersive X-ray spectroscopy (EDX) or
us
impedance measurements [9, 10] as well as several ion beam methods (mainly for very thin layers in the nm-regime) like ion-beam/TOF-MS or Rutherford backscattering spectrometry (RBS) [11, 12, 13], and (micro-)X-ray fluorescence (µ−XRF) [14] are state-of-the-art methods in industry. The drawbacks of these
an
35
techniques are time-consuming sample preparation and/or the need for highvacuum conditions (EDX) or insufficient thickness-resolution and accuracy for
M
multi-layers (XRF), particularly on structured samples. The introduced method using Laser-induced Breakdown Spectroscopy (LIBS) combined with total spec40
tral correlation avoids these main disadvantages. The only requirement for LIBS
te
d
is a free line of sight to the test-workpiece.
2. Theory and Experimental Setup
Ac ce p
2.1. Galvanoaluminium samples
The samples investigated in this study were standardised carbon-steel test
45
sheets (Rasant Alcotec Beschichtungstechnik GmbH, Germany) with a total sample thickness of about one millimetre, which were coated with 10 µm of
nickel and topped with 30 µm of aluminium (Figure 1a and Figure 1b). 2.2. Laser and spectroscopic setup The sample sheets were cut to 15×15 mm2 pieces and mounted on a mi-
50
crometer positioning stage for precise and reproducible lateral adjustment (x and y direction). A micrometer screw in a precise cage assembly was used to adjust the focus position of the 2” focusing lens (focal length: 92 mm @ 266 nm, 96 mm @ 532 nm, respectively. Figure 2). The same lens (fused silica) was used
5
Page 5 of 27
(b) on the edge
(c) array of laser ablated spots
(d) one ablation site
ip t
(a) at the surface
cr
Figure 1: SEM-pictures of polished cross-sections of the untreated sample (a,b) and of the laser treated surface of the galvanoaluminium sample (c,d). Picture (a), from the top: Carbonepoxy (investment compound, black), galvanoaluminium (dark-grey), nickel (thin, smooth
Scale bars: 20 µm (a), 10 µm (b), 100 µm (c), 10 µm (d).
us
light grey) and carbon steel (porous light-grey at the bottom).
55
an
to collect the emitted light from the plasma tube, following an analogue lens for refocussing the light into a broadband fiber guide to the spectrograph/camerasystem. Due to the short laser wavelength of 266 nm and the fact that most of
M
the emission light is in the same wavelength regime (compare emission spectra, Figure 3), the laser was coupled collinearly into the detection beam path via a small (1/2”) dichroic mirror between the two lenses and adjusted to a Gaussian beam radius (1/e2 ) of ω = 100 µm for both wavelengths, respectively.
d
60
te
Figure 2: Confocal 266 nm LIBS-setup for stratigraphy. ‘2ω’ and ‘4ω’ indicate the frequency duplication and quadruplication, ‘pol’ is a polarizing beamsplitter for energy adjustment,
Ac ce p
respectively.
The 266 nm laser pulses with a pulse duration of 4 ns were generated by
frequency-quadrupling (β-barium oxide, BBO) of the 1064 nm fundamental wavelength of a Nd:YAG laser (Quantel, model Brilliant EaZy), the 532 nm pulses were generated by an analogue setup without second frequency duplica-
65
tion module. Energy-adjustment was performed via a rotatable polarizing beam splitter positioned directly behind the frequency conversion modules. The emitted light was analysed via an ´echelle-spectrograph (LTB, Lasertechnik Berlin GmbH, model Aryelle 200) and detected via an intensified and gated multichannel plate charge coupled device (ICCD) camera (Andor, model ICCD
70
734 84 Gen II). Data acquisition was performed with the standard SophiTM -
6
Page 6 of 27
software from LTB, any further data processing was carried out via special
ip t
FortranTM -routines. 2.3. Correlation stratigraphy 2.3.1. Data acquistion and processing
In order to achieve better signal-to-noise ratio at low fluences, which are
cr
75
necessary to obtain low ablation rates (and therefore high depth-resolution [1]),
us
each analysed spectrum is the result of 25 ablation events (compare [15]). Spectra were recorded at a laser pulse frequency of 0.5 Hz, limited by the camera’s acquisition speed. The gate delay was set to 50 ns for 266 nm and 400 ns for 532 nm, with the gate width set to 1000 ns in both cases. The difference in
an
80
delay times can be attributed mostly to differences in the respective conversion
M
module’s electronics.
Each position in a 5 by 5 grid was ablated at least 40 times, with every single spectrum saved. In order to compensate for possible calibration drift, all spectra were oversampled to a resolution of 5 pm by performing a linear interpolation
d
85
and calculating the resulting I(λ). Spectra corresponding to the same pulse
te
number N at different spots were combined by calculating the mean intensity for each wavelength. This procedure yields 40 spectra for every fluence, which
Ac ce p
were used for correlation analysis.
90
2.3.2. Data analysis
Contrary to single- or multi-line tracing in depth profiling the measured peak
intensities were replaced by a linear correlation coefficient r (Pearson productmoment correlation coefficient, Equation 1) of each spectrum with independently acquired high-resolution spectra of pure aluminium (Figure 3a), nickel
95
(Figure 3c) and iron (Figure 3e). P (Iλ,M − I¯M )(Iλ,S − I¯S ) λ rP r = rP (Iλ,M − I¯M )2 (Iλ,S − I¯S )2 λ
(1)
λ
The correlation coefficient is calculated from the spectral intensities of the measurement Iλ,M and the standard Iλ,S according to Equation 1 for all data 7
Page 7 of 27
points in the spectrum. Here the term in the numerator is the statistical covariance of the two correlated spectral intensities Iλ,M and Iλ,S while the two factors written in the denominator are the standard deviations of the measured spec-
ip t
100
tral intensity Iλ,M of each laser pulse and of the stored standard spectrum Iλ,S ,
cr
respectively (I¯M and I¯S being the arithmetic mean of the spectral intensities).
This method, which is described in detail in [6] and was proved to be a suit-
105
us
able method for metal thin-film analysis in [1, 15] has the advantage of eliminating experimental noise e.g. fluence fluctuations, pulse-to-pulse fluctuations in plasma density and temperature and signals resulting from point impurities
an
in a certain metal layer from the resulting depth profile.
A linear direct relation of the correlation coefficient r and corresponding signal intensity (of a certain element as an ensemble of emission lines) can be assumed, even if the standard spectra’s signal-to-noise-ratio is higher than that
M
110
of the (averaged) single-pulse measurement at a certain N . This method does not require explicit atomic emission line identification and
d
can therefore be used for depth- and ablation analysis of complex layer compo-
115
te
sitions like alloys or other microcrystalline, homogeneous mixtures, as long as a reference spectrum of a certain material is available [6]. The method theoret-
Ac ce p
ically gives values of r ranging from -1 (perfect anticorrelation) to +1 (perfect correlation). Since the majority of the data points in every spectrum is white noise, values of r far below zero are very unlikely (compare Figure 4).
120
2.4. Stratigraphic modelling There are multiple ways to obtain stratigraphic information from LIBS raw-
data. Many of them have been developed for analytical and industrial contexts [5, 16, 17]. The easiest way to map elements through (thick and homogeneous)
layers is to compare well known line intensities determined from pure standard 125
materials while consecutively ablating material pulse-by-pulse. This approach can be stretched to its limits, when elements with a multitude of low-intensity lines should be traced in one single experiment. 8
Page 8 of 27
Statistical approaches were taken e.g. for industrial quality control, where the amount of sample material is not a limiting factor. Balzer and coworkers presented a method, where bursts of laser-pulses are used to control the film-
ip t
130
thickness of galvanised (Zn) steel by simply comparing the ratio of zinc over
cr
iron emission, using many known lines of each element, respectively.
This method averages out one of the biggest problems in LIBS-stratigraphy:
135
us
The pulse-to-pulse fluctuations in laser fluence and plasma emission behaviour [17]. However, an exact depth profile can not be obtained this way, only the yes/no information of penetration and the layer-thickness in a certain confidence
an
interval can be gained by calibrating the system with standard samples of known layer thickness.
The third way of obtaining depth information from LIBS-emission data uses total spectral information instead of single emission line intensities by calculat-
M
140
ing the linear correlation coefficient of a single-pulse spectrum with a standard spectrum of a pure element or alloy to be traced through the layers as described
d
in the last section. This method, initially introduced for LIBS by Mateo et al in
145
te
2001 [1] results in comparatively smooth correlation stratigrams, which represent the similarity of a certain spectrum to its standard at each pulse number N
Ac ce p
and offers the possibility of developing fitting functions and models of ablation behaviour in a novel way. Since this technique was optimized and taken further in this study, results and achievements should be discussed in detail later on. A major issue in stratigraphy is, besides recording a reliable element-vs.-
150
pulse number (or depth) signal with high accuracy, the description of elemental in-depth traces by means of suitable mathematical functions which allow the quantification or prediction of characteristics in the signal with simple calculations (e.g. surface contaminations by debris [15]). Compositional depth profiling was treated empirically with respect to the
155
irradiance dependence [1, 6] and was theoretically modelled in detail [18]. Both the empirical and the theoretical approach have been successfully applied to layers with thicknesses much higher compared to the average ablation rates of the constituent materials. However, for thinner layers the proposed models 9
Page 9 of 27
either yield no good fit to obtained experimental data (in the case of the method 160
proposed in [6]) or require substantial manual intervention (smoothing) in the
ip t
fitting process [18], which is not desirable for a model intended for simple,
automated, widespread, everyday application. Therefore, the development of
following the first approach for a two-layer system in [15]. 2.5. Surface morphology
us
165
cr
a generalised simple mathematical description was another aim of this work,
The shape, diameter and depth of laser ablated craters was estimated via op-
an
tical (Olympus, 93 model STM-MJS) and confocal laser scanning (Zeiss, model LSM 800) as well as scanning electron microscopy (Zeiss, model 94 Supra 55 VP), see Figure 1c and Figure 1d. Energy dispersive X-ray spectroscopic line scans were performed with the Zeiss electron microscope as well (30 s per spot,
M
170
5 keV).
d
3. Results and Discussion
te
3.1. Spectral information and depth profiles 3.1.1. Spectral analysis
Figure 3 shows the standard spectra of the three main constituent elements
Ac ce p
175
of the respective layers (left) which were used for the correlation analysis in all subsequent stratigrams. The spectra were recorded using metal foils of the highest available purity (Al, Ni, Aldrich >99.99 %, respectively) and polished
bulk material (carbon steel, “Fe”).
180
On the right side of Figure 3 three spectra at different pulse numbers N , and therefore different depths are shown. It can clearly be seen that the spectrum recorded on top of the sample (N = 3) shows tremendous similarity to the aluminium standard, at N = 10 a mixed Al/Ni-spectrum is obtained and at a pulse number of N = 18 it is mainly the spectrum of the bulk material.
185
10
Page 10 of 27
3.1.2. Shape of depth profiles and fitting procedure When the three-layer sample is ablated by consecutive laser pulses, the col-
ip t
lected emission spectra contain all information necessary for estimating the breach of layer interfaces in terms of pulse number N . Because of the non-
rectangular cross-section of ablation craters, the calculated elemental correla-
cr
190
tion coefficient does not change abruptly when the centre of the ablation crater
us
reaches an interface. Due to the Gaussian ablation profile, where a significant portion of the ablation and plasma formation takes place on the flanks of the Gaussian fluence distribution, the stratigrams show a gradual rising and decaying behaviour rather than sharp changes. Figure 4 shows a stratigram of
an
195
the galvanoaluminium-nickel-steel system calculated with the standard spectra in Figure 3 and 50 separately averaged (25 spots) spectra, each recorded at a
M
certain N . The three plotted datasets stem from correlation of the single N spectra with Al-, Ni- and Fe-standard spectra, respectively. 200
Several characteristics can be noticed in the stratigrams which should be dis-
d
cussed in the following paragraphs. All parameters are described in Table 1.
Ac ce p
te
i rji (N ) = Ai 1 − exp −kji (N − Nj−1,j ) − aN 2 + cij " i #2 i k + N − N j (j−1,j) rji (N ) = Ai exp 1 − − aN 2 + cij kji
(2)
(3)
Equation 2 shows an empirical fitting function for the case of increase of correlation coefficient rji (N ) of of the acquired spectra at pulse number N with
205
the standard spectrum of substance i in the layer j (counting from the surface), while Equation 3 is used for the description of a decay of ri (N ) with N (usually in another layer j). Generally, an increase (and, for thick layers, a constantly high r following it), is observed in a layer which actually contains the analysed compound, while a decay can be observed after such a layer has been passed
210
through by the ablation. The scaling factor Ai is used to describe the relative intensity of rji (N ) (for all layers j, with respect to the noise-baseline level ci of the first baseline). The term in curly brackets describes the exponential rise/decay by the use of a decay constant kji and the number of pulses N . The 11
Page 11 of 27
Table 1: Parameters used in Equation 2 and Equation 3.
meaning
ip t
parameter
cr
i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . element or compound
j . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . number of layer (from outermost)
us
Ai . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . amplitude of fit/linear scaling factor
kji . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . rise- or decay-constant
an
i Nj,j+1 . . . . . . . . . . Position of the interface between layers j and j + 1
cj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . cumulative baseline shifts for fit
M
a . . . . . . . . square decay parameter (not observed in these experiments)
in terms of N . The term −aN 2 is used as a general damping term, describ-
te
215
d
i parameter Nj−1,j finally gives the layer boundary between layer j − 1 and j
ing the overall decay of all signals with high N which may be attributed to
Ac ce p
plasma and vapour shielding effects [15, 19]. This effect was not observed in this study due to the low overall experimental profiling depth/diameter-ratio (aspect ratio) of the produced cavities, but it should be mentioned as a part of
220
our routine for reasons of completeness. For a three-layer system consisting of three different homogeneous substances (pure element or alloy), there is exactly one characteristic amplitude of the correlation coefficient r, representing the highest achievable correlation of that substance with the corresponding standard. It is described by Ai , with superscript i being the element (Al, Ni, Fe),
225
see also Table 1. Furthermore, each substance i has at least one rise and one decay constant (kji , which quantifies the signal behaviour of the correlation coefficient of the element i in the layer j, counting from top). A certain decay of correlation coefficient r
12
Page 12 of 27
(Fe)
(Ni)
with pulse number N can happen before (iron: k1 of cross-contamination [15]) or after (aluminium: nickel, the rise
(Ni) (k2 )
(Al) k2 )
(Ni)
. The initial rise in Al-
is a direct cause of the cross-contamination too, compare [15]
cr
correlation
the rise. In the case of
is followed by a second decay due to its intermediate-layer
character, described by the second decay constant k3 (Al) k1
- because
ip t
230
and nickel: k1
(Fe)
and the rise of iron-correlation when laser-drilling into the bulk material k3 is obvious.
us
235
N12 and N23 indicate the calculated layer-breach in terms of N for the fitting functions of each substance between the layers 1 and 2 (Al and Ni) and 2 and
an
3 (Ni and Fe), respectively. The value of c1 is the baseline-offset in correlation coefficient. For Ni, c3 describes an additional baseline increase (c = c1 + c3 ) 240
observed after breaching the layer.
M
Note that the fitting models applied here will not yield all parameters of interest for all elements. For example, it is not possible to determine N12 from the Fe
Ac ce p
te
d
curve. These cases are indicated by a “-” in Table 2.
13
Page 13 of 27
Al
Ni
Fe
Ai
0.80 ± 0.011
0.46 ± 0.018
0.693 ± 0.0053
k1i
1.04 ± 0.043
1.51 ± 0.058
2.17 ± 0.085
k2i
34 ± 1.1
0.20 ± 0.016
-
k3i
-
39 ± 2.4
i N12
15.8 ±0.18
13.93 ± 0.062
i N23
-
22.8 ± 0.19
22.6 ± 0.93
c1
0.11 ± 0.010
0.027 ± 0.0017
0.047 ± 0.0031
c3
-
0.05 ± 0.011
-
us
cr
i
ip t
Table 2: Fitting parameters for the stratigrams in Figure 4.
0.183 ± 0.0059
d
M
an
-
245
te
3.1.3. Change of correlation stratigrams with the laser fluence The correlation coefficients as a function of N are plotted in Figure 5. The
Ac ce p
positions of the layer breaches, defined by the increase of the Ni and Fe signal respectively, are clearly visible. Two effects of increasing fluence should be mentioned. First, the number of pulses required to breach the top layer decreases with increasing fluence. This corresponds to an increase in the ablation rate
250
(which we will hereinafter refer to as h), as would be expected. Second, the
maximum r increases with the fluence as well, signifying an improved signalto-noise ratio in the individual spectra. For efficient stratigraphy, therefore a balance of ablation rate and signal-to-noise ratio has to be found for practical routine deph profiling.
14
Page 14 of 27
255
3.2. Evaluation of the stratigraphic measurements
ip t
3.2.1. Comparison of LIBS-stratigrams to EDX cross sections Figure 6 and Figure 7 show a comparison of the depth profiles gathered using
LIBS at both laser wavelengths and two different fluence regimes. In order to
260
cr
obtain a reference for the layer structure, a cross-section of the sample was prepared by embedding a part of the sample sheet and analysing the polished
us
cross-section in an electron microscope.
The EDX scan (top) is the result of 5 consecutive line scans, performed at different lateral positions across the layers. Due to their similarity, the resulting
265
an
data points were combined into one graph. It should be noted that, since the EDX-system was not calibrated for quantitative analysis of the respective elements, the exact values obtained for the atomic weight fraction of each element
M
are not trustworthy. Nevertheless, the qualitative presence of each element as a function of position (and thus depth d under the original surface) could be easily determined by fitting sigmoid switch functions (based on the error function erf(d)) to the values. The width of these steps and, as a consequence, the width
d
270
te
of the sigmoid fit yield the uncertainty of the layer thickness determination, mainly resulting from the lateral resolution of the EDX measurement.
Ac ce p
Considering the LIBS depth profiles, shown as r(d), several effects can be observed: First, both wavelengths show an increase in ablation rate with increasing
275
fluence, as can be seen from the increasing distance between the points. Second, evidence of surface contamination with iron and nickel can be seen at the very beginning of the pulse-by-pulse ablation at the left side in each LIBS-stratigram [15]. This results from the re-deposition of material on the surface during the drilling of one point, which produces an additional very thin top layer observed
280
during the measurement of subsequent spots. In the results, it is visible as a lower r(Al) at the first and second laser pulse, accompanied by higher r(Ni) and r(Fe) values. For both fluence regimes, the UV laser yielded higher ablation rates than the 532 nm in the Al layer, resulting in fewer pulses needed to reach the onset
15
Page 15 of 27
285
of the Ni signal. Conversely, this effect could not be observed in the Ni layer itself, where the ablation rate was found to depend only on the fluence. An
determining the depth in this layer could be applied.
290
cr
A graph of all observed ablation rates h is shown in Figure 8.
ip t
analysis of the Fe ablation rate was not performed, since no reliable means of
3.2.2. Fluence dependency of the ablation rate
us
The ablation rate h was determined as the thickness of a given layer, divided Ni Fe by the number of pulses required to drill through that layer (N1,2 or N2,3 −
an
Ni N1,2 , respectively). As expected, it was found to increase with the logarithmic
fluence, as predicted in [20]. The plots of h vs ln(F ) shown in Figure 8 allow 295
for the calculation of the effective absorption coefficients αeff , which are shown
M
in Figure 9 (for calculation procedure see section 4 or [20]). 3.2.3. Wavelength dependency of the ablation rate
d
Figure 8 and Figure 9 show that h(F ) for Ni appears to be independent of the wavelength used, whereas h(F ) for Al was substantially higher when using the UV laser. The two independent measurement series yielded consistent results
te
300
within the measurement precision. The average ablation rate h as a function of
Ac ce p
fluence F for Ni ranges from 1–3.5 µm/pulse for λ = 266 nm as well as for λ = 532 nm. In contrast, the range of h for Al differs from 2–4 µm/pulse for λ =
532 nm and 4–8 µm/pulse for λ = 266 nm in the exact same fluence range on
305
the same sample.
Comparison of αeff with the calculated values for the inverse thermal diffu-
sion length 1/Lth shows that the latter does serve as a very rough prediction for all the Ni values of αeff , as well as for those of Al at 266 nm. The deviation Al of the Al ablation rates and therefore αeff with changing wavelength can not be
310
explained by this simple model.
16
Page 16 of 27
4. Interpretation of the results
ip t
Laser-material interaction in the nanosecond pulse regime has long been discussed by a multitude of different physical, but also more or less empirical approaches, describing the involved processes from different points of view [20, 21, 22, 23, 24, 25].
cr
315
Let us compare laser ablation of metals as a simple thermal process with
us
no attenuation of the laser beam in the gas phase i.e. when all the energy is deposited and absorbed at the very surface of the metal: The ablation rate h(F )
(4)
whenever the ablation of a solid takes place with a gaussian laser beam [20].
M
320
an
(ablated depth per pulse) is expected to follow the logarithmic law 1 F h(F ) = ln αeff Fth
Here αeff is the effective absorption coefficient. While this does follow the same mathematical form as the Lambert-Beer absorption coefficient α, it has both
d
different physical background and different numerical values. For nanosecond lasers on metal, these values are usually considerably lower, implying deeper penetration into the material than mere optical absorption would account for.
te
325
The reason for this behaviour is expected to be thermal diffusion; consequently,
Ac ce p
one approximate prediction for αeff is the thermal diffusion length Lth , resulting from solving the linear heat equation for the given system. Its calculation and relation to αeff are as follows:
330
√ 1 ∝ Lth = 2κτ αeff
(5)
where κ is the thermal diffusivity of the material, and τ is the laser pulse length [26].
Differences in the behaviour of Ni and Al can be observed by comparing
the effective absorption coefficients of Ni and Al (datapoints, calculated with Equation 4 from LIBS-stratigrams) to the calculated inverse of the thermal dif335
fusion length (dashed lines, calculated from laser- and material-parameters with Equation 5) as plotted in Figure 9.
17
Page 17 of 27
While the effective absorption coefficient correlates well for Ni in all four experimental series, two performed separately with green light and two with UV laser
340
ip t
pulses, the ablation behaviour of aluminium only shows pure thermal energy transfer in the experiments with UV light. At 532 nm the effective absorption
cr
coefficient strongly deviates from the values predicted by the thermal diffusion theory.
us
Since there are no significant differences in optical absorptivity, reflectivity, or other optical properties of solid Ni and Al in both wavelength regimes, nor any 345
observable changes in surface roughness or morphology, the influence of surface
an
phenomena on energy transfer from light into the material can be excluded. The cause for the deviations is likely to be looked for in a combination of effects: On the one hand, it may be necessary to apply a more complex
350
M
thermodynamic model capable of dealing with the high-energy environment the sample is exposed to. Mechanisms such as explosive boiling [27, 28] are likely to influence the ablation rate. Based on the results collected by Miotello et
d
al., it is to be assumed that a phase explosion occurs throughout the applied
te
fluence range. The lack of discontinuities in the h vs. ln(F ) curves indicates the presence of one constant mechanism of ablation. However, this thermodynamic model does not directly imply wavelength-dependent effects.
Ac ce p
355
On the other hand, one has to consider the possible influence of plasma
shielding, which would have to be element-specific in the case presented: Hot
expanding plasma can physically or chemically modify or ablate bigger surface areas, than those confined by a Gaussian laser spot, laterally and in-depth.
360
Apart from that, it can also scatter and absorb the incoming laser light during the nanosecond-pulse [29, 30, 31]. This pressure-dependent effect describes the partial absorption of the nanosecond laser-pulse by the expanding plasma and vapour during the pulse, which obviously is not observed in femtosecond-laser
ablation where laser-material interaction takes place before the plasma forma365
tion process. Furthermore, and this is an especially big issue in spectroscopy, the characteristic light, emitted by the hot plasma in the confinement of the drilled crater, is partially self-absorbed and scattered in the outer parts of the plasma 18
Page 18 of 27
plume by the expanding plasma itself (characteristic self-absorption of Dopplerbroadened emission lines), cold vapour and re-condensed particles [6, 19]. Nevertheless, more detailed studies of these processes, such as time-resolved
ip t
370
imaging of the evolving plasma plume and variation of the pulse duration, as
cr
shown in [31] have to be performed to shed light upon this wavelength-dependent and element-specific effect. Experiments including both ablation and trans-
375
us
mission measurements of the laser-induced plasma itself by frequency-tunable (OPO) nanosecond laser beams will offer the opportunity to investigate the temporal and spatial (perpendicularly delayed double-pulse measurements) trans-
an
parency of the evolving plasma as a function of wavelength for a wide range of materials including pure metals and alloys.
380
M
Summary
Laser ablation behaviour of the Al-Ni-Fe layer system was evaluated by empirical fitting functions of elemental emission traces. Thus, consecutive pulse-by-
d
pulse laser ablation was recorded with different fluences and wavelengths. The
te
thermal diffusion length, as presented in this study, gives a good first approximation for the ablation rate as a function of F . In particular, the logarithmic 385
description of h(F ) has been found reliable. The experimental data of nickel
Ac ce p
ablation with both UV and visible laser light agree well with a simple heat diffusion model. The ablation behaviour of aluminium at visible wavelength deviates strongly from this model.
5. Acknowledgements
390
The authors would like to thank Stephan Puchegger (Faculty Centre for
Nanostructure Research, University of Vienna) for his invaluable assistance with the scanning electron microscopy and EDX-spectroscopy and James Bedillion for fruitful discussions and corrections, as well as Rasant Alcotec GmbH for the generous supply of galvanoaluminium samples.
19
Page 19 of 27
395
6. References
ip t
[1] M. P. Mateo, J. M. Vadillo, J. J. Laserna, Irradiance-dependent depth profiling of layered materials using laser-induced plasma spectrometry, Journal
cr
of Analytical Atomic Spectrometry 16 (11) (2001) 1317–1321.
[2] A. Jurado-Lopez, M. D. Luque de Castro, Laser-induced breakdown spectrometry in the jewellery industry. part i. determination of the layer thick-
us
400
ness and composition of gold-plated pieces, Journal of Analytical Atomic Spectrometry 17 (5) (2002) 544–547.
an
[3] L. M. Cabalin, J. J. Laserna, Surface stoichiometry of manganin coatings prepared by pulsed laser deposition as described by laser-induced breakdown spectrometry, Analytical Chemistry 73 (6) (2001) 1120–1125.
M
405
[4] A. Jurado-Lopez, M. D. Luque de Castro, Laser-induced breakdown spectrometry in jewellery industry. part ii: Quantitative characterisation of
d
goldfilled interface, Talanta 59 (2) (2003) 409–415.
410
te
[5] D. G. Papazoglou, V. Papadakis, D. Anglos, In situ interferometric depth and topography monitoring in libs elemental profiling of multi-layer struc-
Ac ce p
tures, Journal of Analytical Atomic Spectrometry 19 (4) (2004) 483–488. [6] M. P. Mateo, G. G. Nicolas, V. Pi˜ non, A. Ya˜ nez, Improvements in depthprofiling of thick samples by laser-induced breakdown spectroscopy using linear correlation, Surface and Interface Analysis 38 (5) (2006) 941–948.
415
[7] T. O. Nagy, U. Pacher, A. Giesriegl, L. Soyka, G. Trettenhahn, W. Kautek, Laser-induced electrochemical de- and repassivation investigations on plasma-oxidized aluminium alloys, Applied Surface Science 302 (2014) 184– 188. [8] R. D¨ otzer, Galvano-aluminium und seine anodische oxidation galvano-al-
420
eloxal-schichten - eine neue oberfl¨ache, Chemie Ingenieur Technik 45 (9-10) (1973) 653–658.
20
Page 20 of 27
[9] A. P. Abbott, S. Nandhra, S. Postlethwaite, E. L. Smith, K. S. Ryder, Electroless deposition of metallic silver from a choline chloride-based ionic
425
ip t
liquid: A study using acoustic impedance spectroscopy, SEM and atomic force microscopy, Phys. Chem. Chem. Phys. 9 (2007) 3735–3743.
cr
[10] E. P. Gusev, M. Copel, E. Cartier, I. J. R. Baumvol, C. Krug, M. A. Gribelyuk, High-resolution depth profiling in ultrathin Al2O3 films on Si,
us
Applied Physics Letters 76 (2) (2000) 176–178.
[11] C. Jeynes, N. P. Barradas, P. K. Marriott, G. Boudreault, M. Jenkin, E. Wendler, R. P. Webb, Elemental thin film depth profiles by ion beam
an
430
analysis using simulated annealing - a new tool, Journal of Physics D: Applied Physics 36 (7) (2003) R97.
M
[12] J.-C. Dran, J. Salomon, T. Calligaro, P. Walter, Ion beam analysis of art works: 14 years of use in the louvre, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms
d
435
219-220 (0) (2004) 7 – 15, proceedings of the Sixteenth International Con-
te
ference on Ion Beam Analysis.
[13] C. Neelmeijer, W. Wagner, H. Schramm, Depth resolved ion beam analysis
Ac ce p
of objects of art, Nuclear Instruments and Methods in Physics Research
440
Section B: Beam Interactions with Materials and Atoms 118 (1-4) (1996) 338 – 345, ion Beam Analysis.
[14] R. Dargel, M. Azeroual, B. Mogwitz, J. Janek, C. Vogt, Non-destructive analysis of silver selenide films obtained by pulsed laser deposition (pld) with micro-xrf, Journal of Materials Science 42 (17) (2007) 7375–7380.
445
[15] T. O. Nagy, U. Pacher, H. P¨ohl, W. Kautek, Atomic emission stratigraphy by laser-induced plasma spectroscopy: Quantitative depth profiling of metal thin film systems, Applied Surface Science 302 (2014) 189–193.
21
Page 21 of 27
[16] I. Bozs´ oki, B. Balogh, P. Gordon, 355 nm nanosecond pulsed nd:yag laser profile measurement, metal thin film ablation and thermal simulation, Optics & Laser Technology 43 (7) (2010) 1212–1218.
ip t
450
[17] H. Balzer, S. H¨ olters, V. Sturm, R. Noll, Systematic line selection for on-
cr
line coating thickness measurements of galvanised sheet steel using libs, Analytical and Bioanalytical Chemistry 385 (2) (2006) 234–239.
455
us
[18] L. St-Onge, A mathematical framework for modeling the compositional depth profiles obtained by pulsed laser ablation, Journal of Analytical
an
Atomic Spectrometry 17 (9) (2002) 1083–1089.
[19] R. E. Russo, X. Mao, S. S. Mao, The physics of laser ablation in micro-
M
chemical analysis, Analytical Chemistry 74 (3) (2002) 70A–77A. [20] J. Kr¨ uger, W. Kautek, Ultrashort pulse laser interaction with dielectrics 460
and polymers, Adv. Polym. Sci. 168 (2004) 247–289.
d
[21] Z. Y. Wang, J. T. Liu, D. M. Hirak, D. C. Weckman, H. W. Kerr, Deter-
te
mining the spot size and gaussian distribution coefficient of pulsed laser beams using kapton films, Journal of Laser Applications 5 (1) (1993) 5–12.
Ac ce p
[22] Y. Jee, M. F. Becker, R. M. Walser, Laser-induced damage on single-crystal
465
metal surfaces, J. Opt. Soc. Am. B 5 (3) (1988) 648–659.
[23] A. Bogaerts, Z. Chen, R. Gijbels, A. Vertes, Laser ablation for analytical sampling: What can we learn from modeling?, Spectrochimica Acta - Part B Atomic Spectroscopy 58 (11) (2003) 1867–1893.
[24] H. Howard, A. J. Conneely, G. M. O’Connor, T. J. Glynn, Investigation of
470
a method for the determination of the focused spot size of industrial laser beams based on the drilling of holes in mylar film (2003) 541–552.
[25] C. Arag´ on, J. A. Aguilera, Characterization of laser induced plasmas by optical emission spectroscopy: A review of experiments and methods, Spectrochimica Acta Part B: Atomic Spectroscopy 63 (9) (2008) 893–916. 22
Page 22 of 27
475
[26] E. Matthias, M. Reichling, J. Siegel, O. W. Kading, S. Petzoldt, H. Skurk, P. Bizenberger, E. Neske, The influence of thermal-diffusion on laser-
ip t
ablation of metal-films, Applied Physics a-Materials Science & Processing 58 (2) (1994) 129–136.
480
cr
[27] A. Miotello, R. Kelly, Critical assessment of thermal models for laser sputtering at high fluences, Applied Physics Letters 67 (24) (1995) 3535–3537.
us
[28] A. Miotello, R. Kelly, Laser-induced phase explosion: new physical problems when a condensed phase approaches the thermodynamic critical tem-
an
perature, Applied Physics A 69 (1) (1999) S67–S73.
[29] J. M. Vadillo, J. M. Fern´andez Romero, C. Rodr´ıguez, J. J. Laserna, Depth485
resolved analysis by laser-induced breakdown spectrometry at reduced pres-
M
sure, Surface and Interface Analysis 26 (13) (1998) 995–1000. [30] J. M. Vadillo, J. M. Fern´andez Romero, C. Rodr´ıguez, J. J. Laserna, Effect
d
of plasma shielding on laser ablation rate of pure metals at reduced pressure,
490
te
Surface and Interface Analysis 27 (11) (1999) 1009–1015. [31] A. Semerok, C. Chal´eard, C.ard, V. Detalle, J. L. Lacour, P. Mauchien,
Ac ce p
P. Meynadier, C. Nouvellon, B. Sall´e, B., P. Palianov, M. Perdrix, G. Petite, Experimental investigations of laser ablation efficiency of pure metals with femto, pico and nanosecond pulses, Applied Surface Science 138-139 (0) (1999) 311–314.
23
Page 23 of 27
ip t cr us an
(b) N =3
M
(a) standard spectrum of aluminium
d
(c) standard spectrum of nickel
(f) N =18
te
(e) standard spectrum of iron
(d) N =10
Figure 3: Standard spectra (left) used for the Al/Ni/Fe correlation stratigrams, and LIBS-
Ac ce p
spectra (right), recorded on the sample at different pulse numbers (average of 25 neighbouring experiments). F = 220 J·cm−2 , spot distance 0.5 mm, λ = 532 nm.
24
Page 24 of 27
ip t cr us an M d te
Ac ce p
Figure 4: Top: 3-layer correlation stratigrams calculated from averaged (25 spots) single-pulse spectra and standard spectra (Figure 3, left). The fit–parameters for the stratigraphic model functions (general rise- Equation 2 and decay-function Equation 3) are explained in the text and listed in Table 1 and Table 2.
: aluminium, 4: nickel, 3: iron. F = 60 J·cm−2 , λ = 532 nm.
Bottom: Schematic illustration of layers (Al: yellow, Ni: green) and bulk (Fe: grey) material with a Gaussian nanosecond-laser ablation profile. i,ii: only top layer material is ablated and only its correlation signal is high. iii,iv: the spectra show a finite correlation for both, layer and bulk material, which changes with depth (pulse number N ).
Figure 5: Calculated 532 nm Ni (top) and Fe (bottom) correlation stratigrams, measured at different fluences: Nickel: ◦ : 80 J · cm−2 , 2 : 110 J · cm−2 , 3 : 170 J · cm−2 , 4 : 220 J · cm−2 . Iron: ◦ : 80 J · cm−2 , 2 : 110 J · cm−2 , 3 : 170 J · cm−2 , 4 : 220 J · cm−2 .
25
Page 25 of 27
ip t
Figure 6: Comparison of EDX cross-section line scan data (top) to LIBS-correlation strati-
grams recorded with 532 (middle) and 266 nm (bottom) at a maximum fluence of 60 J·cm−2 .
◦: aluminium, 2: nickel, 3: iron, 4: carbon. A backscatter image is plotted below for
an
us
cr
comparison.
Figure 7: Comparison of EDX cross-section line scan data (top) to LIBS-correlation stratigrams recorded with 532 (middle) and 266 nm (bottom) at a maximum fluence of 170 J·cm−2 .
M
◦: aluminium, 2: nickel, 3: iron, 4: carbon. A backscatter image is plotted below for com-
te
d
parison.
Ac ce p
Figure 8: Fluence dependency of ablation rates for two different experimental series (left and right). Top: 532 nm, Bottom: 266 nm. The slope in these plots gives the effective absorption coefficient αeff in Figure 9. ◦: Al, 3: Ni.
Figure 9: Effective absorption coefficient αeff , ◦: Al, 3: Ni, versus inverse thermal diffusion length 1/Lth (dashed lines, Ni: purple, Al: blue) for the two experimental series in Figure 8.
26
Page 26 of 27
Highlights:
ip t cr us an
iv)
M
iii)
d
ii)
Depth profiling via nanosecond-LIBS was performed on a galvanic 3-layer coating system via automated fitting routines with green (532 nm) and ultraviolet (266 nm) light. The developed stratigraphic model holds for both wavelengths and all involved layer materials. The change of the ablation wavelength from VIS (532 nm) to UV (266 nm) affects the ablation rate (= in-depth resolution) of the Al top layer but not of the Ni intermediate layer. A pure heat diffusion model can not explain the ablation process in Al. Laser-plasma interaction needs to be considered.
Ac ce pt e
i)
Page 27 of 27