Content uploaded by Junliang Dong
Author content
All content in this area was uploaded by Junliang Dong on Oct 17, 2017
Content may be subject to copyright.
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 1
Terahertz Quantitative Nondestructive Evaluation of
Failure Modes in Polymer-Coated Steel
Junliang Dong, Student Member, IEEE, Alexandre Locquet, D. S. Citrin, Senior Member, IEEE
Abstract—Terahertz reflective imaging is applied to charac-
terize the failure modes in a polymer coating on a steel plate.
The coating was initially scratched, then after accelerated aging,
several types of failure have occurred. In order to resolve
the thin coating (∼50 µm), terahertz frequency-wavelet domain
deconvolution is implemented. With the deconvolved signals, the
temporally overlapping echoes of the incident, roughly single-
cycle terahertz pulse are clearly resolved, and three important
failure modes, viz. corrosion, delamination, and blistering, are
characterized quantitatively. Terahertz images in three dimen-
sions clearly exhibit the coating thickness distribution across
the entire damaged coating, highlighting the terahertz features
associated with different failure modes, thus demonstrating that
terahertz imaging can be considered as an effective modality
for characterizing damage mechanisms in polymer coatings on
metals.
Index Terms—Terahertz radiation, Imaging, Protective coat-
ings, Deconvolution, Nondestructive testing
I. INT ROD UC TI ON
NONDESTRUCTIVE evaluation (NDE) techniques for
monitoring and characterizing coatings on metals are
essential to verify protection of the metal substrate from
corrosion during service. Coating failure can have many
causes and manifestations. Especially, when a region of a
coating system becomes detached from its substrate, the term
“adhesion failure” is commonly used. Delamination and b-
listering are two important types of coating failure in which
compromised adhesion is strongly implicated [1]. Since the
exact cause and nature of coating failure is still in dispute [2],
various NDE techniques have been explored to study damage
mechanisms in coatings, such as electrochemical impedance
spectroscopy (EIS) [3], scanning electrochemical microscopy
[4], acoustic emission [5], and thermography [6][7], to name
a few. However, not all of the NDE techniques mentioned
above can provide quantitative information in depth, nor may
they all be capable of monitoring the condition of adhesion.
Scanning acoustic microscopy (SAM) [2][8][9] and laser-
ultrasonics [10] can provide depth-specific information with
enough resolution for characterizing coating systems; however,
high attenuation in polymer materials limits the penetration
depth of the ultrasonic waves [8]. Therefore, alternative NDE
techniques with relatively high resolution are still sought for
quantitative evaluation of polymer coatings.
J. Dong, A. Locquet, and D. S. Citrin are with the School
of Electrical and Computer Engineering, Georgia Institute of Tech-
nology, Atlanta, GA, 30332-0250 USA, and also with UMI 2958
Georgia Tech-CNRS, Georgia Tech Lorraine, 2 Rue Marconi, 57070
Metz, France (e-mail: junliang.dong@gatech.edu; alexandre@gatech.edu;
david.citrin@ece.gatech.edu).
Manuscript received XXXX XX, 2016; revised XXXX XX, 2016.
Terahertz (THz) imaging, as a relatively new and promising
NDE technique, has attracted considerable interest as a non-
invasive, noncontact, and nonionizing method to characterize
various non-metallic materials. The THz portion of the elec-
tromagnetic spectrum extends from approximately 100 GHz
to 10 THz, and lies between the microwaves and infrared; the
wavelength range in this region is 3 mm down to 30 µm.
THz-frequency electromagnetic waves are known to penetrate
numerous nonmetallic materials that may be opaque in the
range of visible and infrared light. Moreover, as nonionizing
radiation, THz waves present minimal known health risks.
Due to these remarkable properties, THz waves have al-
ready been successfully used for characterizing materials with
multi-layered structures, such as fiber-reinforced composite
laminates [11][12], coated pharmaceutical tablets [13][14],
dental tissues [15], and art paintings [16]. The application
of THz imaging to metals, however, is problematic, as the
conductivities of metals preclude significant penetration of
the THz radiation below the surface. As a result, metals are
typically highly reflective in the THz-frequency range. This
reflectivity can be turned to advantage in the NDE of coatings
on metals. To date, THz imaging has been demonstrated on
automobile paints [17] and marine paints [18], thermal barrier
coatings of gas turbine blade [19], and marine protective
coatings [20]. Various efforts have also been made to enhance
the depth resolution of THz imaging, for example using the
time-domain numerical parameter fitting method based on
multiple regression analysis [21], fitting the theoretical model
with the experiment transfer function in the frequency domain
[22], and multivariate analysis [23].
In this study, THz reflective imaging is employed to charac-
terize different failure modes in a one-layered polymer coating
on a steel plate. The coating was initially scratched, then after
accelerated corrosion over a few months, various types of
failure, such as corrosion, delamination, and blistering, are
observed to have occurred. Because the thickness of this coat-
ing is optically thin in the THz regime, multiple THz echoes
partially or totally overlap in the time-domain waveforms,
which are particularly noticeable due to the high reflectivity of
the metallic substrate. THz frequency-wavelet domain decon-
volution is implemented to resolve the overlapping echoes and
distinguish the features of various failure modes, especially
for revealing the features associated with adhesion failure.
Based on the THz deconvolved signals, several failure modes
are characterized quantitatively in three dimensions, and the
surface topology is successfully reconstructed. Our work thus
supports effort to apply THz imaging to investigate the damage
mechanisms in polymer coatings on metals.
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 2
II. PR IN CI PL E
For samples with layered structure, time-domain THz imag-
ing can provide information in depth by analyzing the reflect-
ed THz signals with an incident approximately single-cycle
THz pulse. Due to dielectric variations with depth, reflected
temporal THz echoes associated with the Fresnel coefficients
between various layers are recorded as a function of transverse
position in amplitude and time delay. The echo amplitudes
provide information on the various refractive indices, while the
time delays between THz echoes provide the optical thickness
of successive layers. With a knowledge of the refractive indices
of the corresponding materials, the physical layer thickness
can be extracted. When the layer thickness is optically thin,
THz echoes will partially or totally overlap; therefore, the
amplitude and time delay cannot be directly extracted from the
time-domain waveform. In this case, THz deconvolution can
be utilized to resolve the overlapping echoes and reconstruct
the intrinsic impulse-response function, and hence the physical
structure of the sample studied.
In the time domain, the THz reflected signal (electric field)
r(t)is the convolution of the incident THz pulse i(t)with
the impulse-response function h(t), which corresponds to
the structure and properties of the sample at a given two-
dimensional position,
r(t) = i(t)⊗h(t).(1)
Deconvolution retrieves the impulse response function h(t)by
applying the inverse Fourier transform based on the convolu-
tion theorem,
h(t) = F F T −1[F F T (r(t))
F F T (i(t))],(2)
where F F T denotes the Fourier transform and F F T −1the
inverse Fourier transform. Frequently, successful deconvolu-
tion cannot be expected by directly applying Eq. (2), since
division by small numbers will give rise to large spikes in the
reconstructed impulse-response function, especially in the high
frequency region, leading to severe ringing in the time domain.
Therefore, deconvolution process is usually further augmented
by frequency-domain filtering to suppress the high-frequency
noise, which can be expressed as,
h′(t) = F F T −1[F F T (f(t))×F F T (r(t))
F F T (i(t))],(3)
with f(t)the filter function in the time domain. In order to
obtain a successful reconstruction, the temporal duration of
f(t)should be short enough to resolve the time intervals be-
tween featured echoes, and f(t)should not contain extra signal
cycles before or after the main peak, which will obscure the
real featured echoes in the reconstructed signal; however, the
selection of f(t)is also a compromise between time resolution
and frequency-domain filtering [24]. If the duration of f(t)is
too short, its frequency spectrum will include large spikes at
high frequencies, which will degrade the reconstructed signal
in the time domain.
A double Gaussian filter or Wiener filtering can be selected
to serve as the frequency-domain filtering [25], and a tapered
8 10 12 14 16 18 20 22 24
0
0.005
0.01
0.015
0.02
0.025
Optical Delay (ps)
Amplitude (a.u.)
0 1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
1
Frequency (THz)
Power (a.u.)
fc= 4 THz
t0= 10 ps
Fig. 1. Hanning window function with typical values, t0=10 ps and fc=4
THz, in the time domain and its Fourier transform (power spectrum) in the
inset.
cosine apodisation function has also been found to work
well [26]. Considering the complexity and effectiveness, a
Hanning window function is chosen as the filter function
f(t)in this study, and its frequency spectrum F(ω)can be
expressed as
F(ω) = {eiωt0cos2(ω
4fc)|ω| ≤2πfc,
0|ω|>2πfc,(4)
where tocorresponds to the arrival time of the main peak in the
time domain and fcis the cutoff frequency. This frequency-
domain filtering is easy to manipulate just by changing the
cutoff frequency fc. An example of f(t)and its frequency
spectrum F(ω)for typical parameters (t0= 10 ps and fc= 4
THz) is shown in Fig. 1.
Quite often, deconvolution only with frequency-domain
filtering cannot guarantee a satisfactory signal-to-noise ratio
when a relatively high value of fcis selected. Stationary
wavelet shrinkage is applied to further attenuate the resid-
ual noise. This technique decomposes a 1D signal into the
approximation coefficients vector and detail coefficients by
convolving with a low-pass filter and a high-pass filter along
the temporal axis at each level. Wavelet coefficients with
small absolute values can be considered as noise, and wavelet
coefficients with large absolute values are regarded as the
main featured information of the signal [25][27]. Removing
the small absolute-value coefficients by thresholding and then
reconstructing the signal is expected to produce a signal in
which the contribution of noise has been reduced. Sometimes,
the signal after frequency-wavelet deconvolution contains slow
fluctuations corresponding to the low frequency noise due to
the THz source being deficient in the low THz frequency
region. This kind of low-frequency noise can be canceled by
subtracting the baseline of the deconvolved signal.
III. EXP ER IM EN T
The THz time-domain reflection imaging system (Teraview
TPS Spectra 3000), which is employed in this study, is shown
schematically in Fig. 2. The GaAs photoconductive antenna is
excited by an ultrafast (femtosecond) laser to produce roughly
single-cycle THz pulses with bandwidth extending from 60
GHz to 3 THz. The ultrafast laser used here is an Er-doped
fiber laser that emits 780 nm pulses with sub-100 femtosecond
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 3
Ultrafast Laser
Optical Delay Line
Probe Pulse
Pump Pulse
Lock-in
Amplifier
Current
Pre-Amplifier
SAMPLE
Photoconductive
Emitter
Photoconductive
Receiver
Fig. 2. Schematic diagram of THz time-domain reflection imaging system.
optical pulse duration at a repetition rate of 100 MHz and
has an average output power in excess of 65 mW. Coherent
detection of the THz radiation is performed in a similar photo-
conductive antenna circuit. By gating the photoconductive gap
with a femtosecond pulse synchronized to the THz emission,
a current proportional to the THz electric field is measured.
By varying the optical path length, the THz time domain
signal can be sampled, resulting in both amplitude and phase
information on a sub-picosecond timescale.
In this study, THz imaging is performed in reflection at
almost normal incidence. Before imaging the sample, a THz
reference signal, shown in Fig. 3, was recorded by setting
a bare metal plate (i.e., an excellent THz reflector) at the
sample position. The tested sample, shown in Fig. 4(a), is a
one-layered polymer (polyester) coating on a zinc-coated steel
substrate. Prior to the application of the polymer coating, a
thin (∼5µm) proprietary primer coating was applied. Both the
primer and polymer were roll coated onto the substrate. The
coating was initially scratched through the coating in the cen-
ter, and after a multi-month accelerated-corrosion exposure,
various types of failure, including corrosion, delamination,
and blistering, are visually evident in the coating. Optical
microscopy was employed to estimate the thickness of the
coating, shown in Fig. 4(b). (The primer coating was not
evident in the optical micrographs.) The thickness of the
0 5 10 15 20 25 30 35 40
−0.5
0
0.5
1
1.5
Optical Delay (ps)
THz Field (a.u.)
0 1 2 3 4
−5
0
5
Frequency (THz)
Log(THz Power) (a.u.)
Fig. 3. THz reference signal with its frequency spectrum in the inset.
Fig. 4. (a) Visible photograph of the surface of the coating sample with
four typical positions highlighted to represent different failure modes. This
photograph is also labeled with the pixel number which is the same as the
THz C-scans in the following section. (b) Optical micrograph from the edge
of the coated sample to roughly estimate the thicknesses of the coating and
the steel plate.
coating is about 0.05 mm based on observation from the edge
of the sample. We then proceeded with the THz imaging.
This sample was raster-scanned by a set of motorized stages
moving in xand ydirections in 0.2 mm steps over a 33 mm
×55 mm region of the sample plane, corresponding to 165 ×
275 pixels. Each recorded reflected temporal THz waveform
contains 4096 data points, and the signal is averaged over 10
shots per pixel. After completing the scanning, a 3D volume
raw data set was acquired.
IV. RES ULT S AN D DIS CU SSION
Based on the observation of the coating surface fea-
tures, four typical positions, which exemplify different failure
modes, are selected in Fig. 4(a). The received THz signal at
position (50, 60), where no evidently visible damages exist,
is shown in Fig. 5(a1). Overlapping THz echoes at this pixel
are observed due to the optically thin coating. In order to
resolve the overlapping echoes and reveal the structure, THz
frequency-wavelet domain deconvolution introduced previous-
ly is applied to the raw reflected THz signal at each pixel. In
order to make the time resolution as high as possible, the
cutoff frequency fcchosen for all the waveforms is 4 THz. In
the wavelet denoising procedure, the symlet (sym4) wavelets
are selected with a maximum level of 7 for the wavelet
decomposition, as no significant improvement can be achieved
for higher levels to justify the extra computational expense.
After deconvolution, a 3D deconvolved data set is obtained.
Peak detection is performed on the deconvolved signals for
identifying the existence of an echo. A threshold value for
peak detection is set for all pixels, above which we consider
a feature as a valid peak.
The deconvolved THz signal at position (50, 60) is shown
in Fig. 5(a2). Two positive peaks and one negative peak
are clearly identified, which are illustrated in Fig. 5(a3).
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 4
80 100 120 140 160
-1
-0.5
0
0.5
1
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-0.5
0
0.5
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-5
0
5
10
15
x 10-3
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-4
-2
0
2
4
6
8
x 10-3
Optical Delay (ps)
Amplitude (a.u.)
Metal Substrate
Coating
1st echo
2nd echo
3rd echo
Incidence
1st echo 2nd echo
3rd echo
Metal Substrate
Coating
1st echo
2nd echo
3rd echo
Incidence
Delamination
4th echo
1st echo
2nd echo
3rd echo
4th echo
(a1)
(a2)
(a3)
(b1)
(b2)
(b3)
Pixel: (50,60) Pixel: (70,190)
Fig. 5. The THz raw signals [(a1) and (b1)] and deconvolved signals
[(a2) and (b2)] at positions (50, 60) and (70, 190), with the corresponding
representations of round-trip echoes in [(a3) and (b3)].
The first two positive peaks correspond to the echoes from
the air/coating interface and the coating/substrate interface
respectively. The third peak, corresponding to the second
round echo, is negative due to the phase shift at the coating/air
interface. We consider pixels with this kind of deconvolved
signal as normal i.e., undamaged, since no features related with
damages can be identified. The optical thickness of the coating
can be acquired by calculating the optical delay between the
first and the second peaks, which is ∼0.5 ps. The refractive
index of polyester in the THz range is ∼1.7 [28], therefore, the
physical thickness of the coating is ∼44 µm, which is close
to the value we estimated with the optical micrograph in Fig.
4(b).
The THz raw signal and deconvolved signal at position (70,
190) are shown in Fig. 5(b1) and (b2). Compared with the nor-
Pixel along X
Pixel along Y
20 40 60 80 100 120 140
50
100
150
200
250
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pixel along X
Pixel along Y
20 40 60 80 100 120 140
50
100
150
200
250
20
30
40
50
60
70
80
90
100(7m)
(a) (b)
Fig. 6. (a) THz C-scan based on the raw signals with the contrast mechanism
as the maximum amplitude of the received signal, which is similar to the
optical photograph in Fig. 4(a); (b) THz C-scan based on the deconvolved
signals associated with the delamination, which indicates the regions with
delamination and the physical thicknesses of the delamination across the
coating plane.
mal THz deconvolved signal in Fig. 5(a2), three positive peaks
and one negative peak are identified. In this case, the second
peak is negative, which indicates the separation between the
coating and the substrate, corresponding to the existence of
delamination (air gap). The round trip of the identified echoes
are illustrated in Fig. 5(b3). Compared with the raw THz signal
at this pixel, the deconvolved THz signal clearly reveals the
prominent feature associated with the delamination, which is
the most important type of adhesive failure. Moreover, the
severity of delamination can be evaluated by calculating its
physical thickness. With the knowledge of refractive index
of air and the measurement of the optical delay between
the first peak (positive) and the second peak (negative), the
delamination can be characterized quantitatively.
THz C-scans can be plotted to provide a two dimensional
presentation of data as a top or planar view of features in the
coating system. The THz C-scan based on the THz raw data
with contrast mechanism as the maximum amplitude value
is firstly plotted in Fig. 6(a), which is similar to the optical
photograph in Fig. 4(a). Based on the expected features in
the deconvolved signal associated with delamination, we can
extract all the pixels with this feature present and calculate
the physical thickness of the delamination as a function of
position. The THz C-scan in Fig. 6(b) clearly exhibits the
delamination areas where the existence of air gap is indicated
by the THz deconvolved signals. We also obtain the physical
thickness of the delamination across the coating plane, which
demonstrates that the delamination has been successfully char-
acterized in three dimensions quantitatively by THz imaging.
80 100 120 140 160
-0.5
0
0.5
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-1
-0.5
0
0.5
1
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-2
0
2
4
6x 10-3
Optical Delay (ps)
Amplitude (a.u.)
80 100 120 140 160
-5
0
5
10
15
x 10-3
Optical Delay (ps)
Amplitude (a.u.)
1st echo
2nd echo
3rd echo
1st echo
2nd echo
(a1)
(a2)
(b1)
(b2)
Pixel: (79,115) Pixel: (21,121)
3rd echo
Fig. 7. THz raw signals [(a1) and (b1)] with the corresponding deconvolved
signals [(a2) and (b2)] at positions (79, 115) and (21, 121) where a blister is
visually present.
The THz raw signal and deconvolved signal at position (79,
115), where the scratch was initially located and the metal
oxide first developed upon aging, are shown in Fig. 7(a1) and
(a2). The amplitude of the raw and deconvolved signals are
relatively small because of the diffuse reflection at the rough
surface of the corrosion area. In addition, compared with the
signals at the undamaged area, a larger time delay between
the first and the second echoes can be observed both in the
raw and deconvolved signals. The optical delay in this case
corresponds to the thickness of the grown metal oxide due to
corrosion at the scratch area.
Next, we discuss the signal features at the blisters. The THz
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 5
(a) (b)
Pixel along X
Pixel along Y
50 100 150
50
100
150
200
250
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
(ps)
Pixel along X
Pixel along Y
50 100 150
50
100
150
200
250
2
4
6
8
10
12
14
x 10-3
Fig. 8. (a) THz C-scan based on the optical delay between the first positive
peak and second positive peak in the deconvolved signals, which represents
the optical distance between the coating surface and the top surface of the
metal substrate; (b) THz C-scan based on the amplitude of the second positive
peak in the deconvolved signals, which indicates the adhesive condition.
raw signal and deconvolved signal at position (21, 121), where
a blister is clearly observed, are shown in Fig. 7(b1) and (b2).
Both THz raw and deconvolved signals present quite similar
features compared with the signals at the undamaged area.
However, by careful comparison with the THz deconvolved
signal at the undamaged area in Fig. 5(a2), we observe: (1)
the optical delay between the first and the second echoes
is larger in blisters; (2) the amplitude of the second echo
is smaller in blisters. Based on these two features, THz C-
scans are plotted in Fig. 8 to reveal all the blisters in the
coating. In Fig. 8(a), THz C-scan is plotted based on the
optical delay between the first two positive peaks of the
deconvolved signals, which represents the optical distance
between the coating surface and the top surface of the metal
substrate. The blistering areas, represented as red spots in the
image, are clearly revealed. Compared with the undamaged
areas, the blisters are associated with a slightly larger optical
thickness, and the delamination and corrosion areas around
the scratch exhibit much larger optical distance due to the
existence of the air gap under the delaminated coating or the
metal oxide, respectively. In Fig. 8(b), THz C-scan is plotted
based on the amplitude of the second positive peak, which
indicates the condition of adhesion, because the second peak
corresponds to the echo bouncing back after reflecting off the
substrate. Compared with the undamaged areas, the blisters
show smaller amplitude which indicates a weaker adhesion
with the substrate, with the delamination areas exhibit much
weaker adhesion, viz. adhesion failure.
The physical origin of the contrast between blisters and the
surrounding undamaged regions are now discussed. Compared
with the undamaged area, the relatively large optical delay
between the first two positive peaks at the locations of the
blisters is due to the separation between the coating and the
substrate and the existence of the air gap. However, the air
gap under the blisters is too small to be resolved even in
the deconvolved signals. In this case, the negative peak cor-
responding to the echo from the coating/air interface and the
positive peak corresponding to the echo from the air/substrate
(a) (b)
(c) (d)
Oxidized
Steel
Normal
Steel
Fig. 9. Optical micrographs associated with different failure modes after
peeling off the coating. (a) Oxidized substrate in the blistering area; (b)
Oxidized substrate in the delamination area; (c) Comparison between the steel
substrate at the undamaged area and the blistering area by peeling off the
coating at one blister area and the adjacent undamaged area; (d) Grown metal
oxide due to corrosion along the scratch.
are not well-separated and largely cancel with each other. The
result of this overlap and cancellation is only one observable
positive peak with a later arrival time and smaller amplitude
compared with the peak corresponding to the echo from
the coating/substrate interface in undamaged areas [12][29].
Another reason, which is also responsible for the smaller
amplitude of the second positive peak, is the presence of
oxidization of the metal substrate in blisters locations induced
by the accelerated aging. The oxidized substrate under the
blisters provides a smaller reflection compared with the normal
steel under the undamaged coating, where the coating and the
substrate adhere tightly and an almost total reflection can be
expected.
Optical microscopy was next employed to verify the phys-
ical explanations above. In order to observe the conditions of
adhesion associated with different failure modes, the coating
at representative positions is carefully removed. Compared
with the undamaged areas, the coating at delamination and
Fig. 10. THz image of the thickness distribution, as well as the surface topolo-
gy, of the coating. Different failure modes, including corrosion, delamination,
and blistering, have been characterized quantitatively in three dimensions.
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 6
blistering areas is easier to peel off due to the existence
of an air gap. Optical micrographs in areas with blistering,
delamination and corrosion along the scratech are shown in
Fig. 9(a), (b), (d), respectively. Oxidized substrate is observed
under both the blister and delamination. Comparison between
the substrate in the undamaged area and under the blister is
shown in Fig. 9(c) by peeling off the coating at one blister and
the adjacent undamaged area. The substrate in the undamaged
area is visually shiny; on the contrary, the oxidized substrate
under the blister appears black in the optical micrograph. The
optical micrographs support our conclusions above based on
the THz images concerning the compromised adhesion of the
coating in these areas.
Based on the analysis above, the characteristics of various
failure modes, including corrosion, delamination, and blister-
ing, have been successfully revealed using time-domain THz
reflective imaging. The thickness distribution of this damaged
coating can be estimated, as shown in Fig. 10. To do this,
we assume that the refractive indices of the corrosion (metal
oxide) is similar to that of the coating [30]. We note in the
image the metal oxide is grown in the scratch area, on both
sides of which are raised areas of delaminations. Beyond this
is a relatively flat area with isolated raised areas due to blisters.
With this 3D image, the thicknesses associated with various
failure modes, as well as the surface topology of the coating
system are clearly reconstructed, exhibiting the capability of
THz imaging for the quantitative NDE of polymer coatings on
metals.
V. CO NC LU SI ON
In this study, THz reflective imaging was demonstrated to
characterize different failure modes in a polyester-coated zinc-
coated steel plate. THz frequency-wavelet domain deconvolu-
tion was adapted to resolve the optically thin coating. Based on
the deconvolved signals, the characteristics of various failure
modes, including corrosion, delamination, and blistering, have
been successfully identified. The THz deconvolved signals also
enable us to evaluate the condition of adhesion, especially for
the delamination and blisters, which are related with adhesion
failure. The thickness distribution across the entire damaged
coating, as well as its surface topology is also obtained. These
interpretations were supported by optical microscopy. Based
on these results, we conclude that THz imaging, which can
provide a noninvasive, noncontact, and nonionizing modality
for characterizing coatings quantitatively in three dimensions,
can be utilized as an effective tool for investigating the damage
mechanisms and monitoring the corrosion process in polymer-
coated metals.
ACK NOW LE DG ME NT
The authors gratefully acknowledge the financial support of
the Conseil R´
egional de Lorraine and the Fonds Europ´
een de
D´
eveloppement R´
egional (FEDER). The authors thank Sophie
Berveiller and Denis Bouscaud for their help in the analysis
of the samples.
REF ER EN CE S
[1] J. Graystone and R. Kennedy, “Non-destructive measurement of coating
adhesion,” Surface Coatings International, vol. 83, no. 8, pp. 389–398,
2000.
[2] I. A. Koch, M. Bargmann, H. Oehler, D. Lellinger, M. Wanner, and
D, “Investigation of delamination mechanisms in polymer coatings by
scanning acoustic microscopy,” Journal of Physics D: Applied Physics,
vol. 44, no. 3, p. 34009, 2011.
[3] K. Darowicki, P. ´
Slepski, and M. Szoci´
nski, “Application of the dynamic
EIS to investigation of transport within organic coatings,” Progress in
Organic Coatings, vol. 52, no. 4, pp. 306–310, 2005.
[4] R. M. Souto, Y. Gonz´
alez-Garc´
ıa, J. Izquierdo, and S. Gonz´
alez,
“Examination of organic coatings on metallic substrates by scanning
electrochemical microscopy in feedback mode: Revealing the early
stages of coating breakdown in corrosive environments,” Corrosion
Science, vol. 52, no. 3, pp. 748–753, 2010.
[5] A. Gallego, J. F. Gil, E. Castro, and R. Piotrkowski, “Identification
of coating damage processes in corroded galvanized steel by acoustic
emission wavelet analysis,” Surface and Coatings Technology, vol. 201,
no. 8, pp. 4743–4756, 2007.
[6] Y. He, G. Y. Tian, M. Pan, D. Chen, and H. Zhang, “An investigation
into eddy current pulsed thermography for detection of corrosion blister,”
Corrosion Science, vol. 78, pp. 1–6, 2014.
[7] A. Sch ¨
onberger, S. Virtanen, V. Giese, C. Spießberger, H. Schr¨
ottner,
J. Rattenberger, and J. Wagner, “Non-destructive evaluation of stone-
impact damages using pulsed phase thermography,” Corrosion Science,
vol. 56, pp. 168–175, 2012.
[8] I. Alig, S. Tadjbach, P. Kr¨
uger, H. Oehler, and D. Lellinger, “Characteri-
zation of coating systems by scanning acoustic microscopy: Debonding,
blistering and surface topology,” Progress in Organic Coatings, vol. 64,
no. 23, pp. 112–119, 2009.
[9] H. Oehler, I. Alig, D. Lellinger, and M. Bargmann, “Failure modes in
organic coatings studied by scanning acoustic microscopy,” Progress in
Organic Coatings, vol. 74, no. 4, pp. 719–725, 2012.
[10] Y. Watanabe, S. Fujisawa, A. Yonezu, and X. Chen, “Quantitative
evaluation of adhesion quality of surface coating by using pulse laser-
induced ultrasonic waves,” Surface and Coatings Technology, vol. 286,
pp. 231–238, 2016.
[11] J. Dong, B. Kim, A. Locquet, P. McKeon, N. Declercq, and D. S.
Citrin, “Nondestructive evaluation of forced delamination in glass fiber-
reinforced composites by terahertz and ultrasonic waves,” Composites
Part B: Engineering, vol. 79, no. 0, pp. 667–675, 2015.
[12] J. Dong, A. Locquet, N. F. Declercq, and D. S. Citrin, “Polarization-
resolved terahertz imaging of intra- and inter-laminar damages in hybrid
fiber-reinforced composite laminate subject to low-velocity impact,”
Composites Part B: Engineering, vol. 92, pp. 167–174, 2016.
[13] Y. C. Shen and P. F. Taday, “Development and application of terahertz
pulsed imaging for nondestructive inspection of pharmaceutical tablet,”
Selected Topics in Quantum Electronics, IEEE Journal of, vol. 14, no. 2,
pp. 407–415, 2008.
[14] H. Lin, Y. Dong, Y. C. Shen, and J. Axel Zeitler, “Quantifying pharma-
ceutical film coating with optical coherence tomography and terahertz
pulsed imaging: An evaluation,” Journal of Pharmaceutical Sciences,
vol. 104, no. 10, pp. 3377–3385, jun 2016.
[15] D. Crawley, C. Longbottom, V. P. Wallace, B. Cole, D. Arnone, and
M. Pepper, “Three-dimensional terahertz pulse imaging of dental tissue,”
Journal of Biomedical Optics, vol. 8, no. 2, pp. 303–307, 2003.
[16] M. Picollo, K. Fukunaga, and J. Labaune, “Obtaining noninvasive
stratigraphic details of panel paintings using terahertz time domain
spectroscopy imaging system,” Journal of Cultural Heritage, vol. 16,
no. 1, pp. 73–80, 2015.
[17] K. Su, Y. C. Shen, and J. Axel Zeitler, “Terahertz sensor for non-contact
thickness and quality measurement of automobile paints of varying
complexity,” Terahertz Science and Technology, IEEE Transactions on,
vol. 4, no. 4, pp. 432–439, 2014.
[18] J. Choi, W. S. Kwon, K.-S. Kim, and S. Kim, “Nondestructive evaluation
of multilayered paint films in ambient atmosphere using terahertz
reflection spectroscopy,” NDT & E International, vol. 80, pp. 71–76,
2016.
[19] T. Fukuchi, N. Fuse, M. Okada, T. Ozeki, T. Fujii, M. Mizuno, and
K. Fukunaga, “Topcoat Thickness Measurement of Thermal Barrier
Coating of Gas Turbine Blade Using Terahertz Wave,” Electrical Engi-
neering in Japan, vol. 189, no. 1, pp. 1–8, 2014.
[20] W. Tu, S. Zhong, Y. C. Shen, and A. Incecik, “Nondestructive testing of
marine protective coatings using terahertz waves with stationary wavelet
transform,” Ocean Engineering, vol. 111, pp. 582–592, 2016.
1077-260X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSTQE.2016.2611592, IEEE Journal
of Selected Topics in Quantum Electronics
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. XX, NO. X, XXXX 2016 7
[21] T. Yasuda, T. Iwata, T. Araki, and T. Yasui, “Improvement of mini-
mum paint film thickness for THz paint meters by multiple-regression
analysis,” Applied Optics, vol. 46, no. 30, pp. 7518–7526, 2007.
[22] M. Scheller and M. Koch, “Fast and accurate thickness determination of
unknown materials using terahertz time domain spectroscopy,” Journal
of Infrared, Millimeter, and Terahertz Waves, vol. 30, no. 7, pp. 762–769,
2009.
[23] T. Iwata, S. Yoshioka, S. Nakamura, Y. Mizutani, and T. Yasui, “Predic-
tion of the thickness of a thin paint film by applying a modified partial-
least-squares-1 method to data obtained in terahertz reflectometry,”
Journal of Infrared, Millimeter, and Terahertz Waves, vol. 34, no. 10,
pp. 646–659, 2013.
[24] J. R. Fletcher, G. P. Swift, D. Dai, J. M. Chamberlain, and P. C. Upadhya,
“Pulsed terahertz signal reconstruction,” Journal of Applied Physics, vol.
102, no. 11, p. 113105, 2007.
[25] Y. Chen, S. Huang, and E. Pickwell-MacPherson, “Frequency-wavelet
domain deconvolution for terahertz reflection imaging and spectroscopy,”
Optics Express, vol. 18, no. 2, pp. 1177–1190, 2010.
[26] R. K. H. Galv˜
ao, S. Hadjiloucas, A. Zafiropoulos, G. C. Walker,
J. W. Bowen, and R. Dudley, “Optimization of apodization functions
in terahertz transient spectrometry,” Optics Letters, vol. 32, no. 20, pp.
3008–3010, 2007.
[27] J. Dong, A. Locquet, and D. S. Citrin, “Enhanced Terahertz Imaging of
Small Forced Delamination in Woven Glass Fibre-reinforced Composites
with Wavelet De-noising,” Journal of Infrared, Millimeter, and Terahertz
Waves, vol. 37, no. 3, pp. 289–301, 2015.
[28] J. Yun-Sik, K. Geun-Ju, and J. Seok-Gy, “Terahertz Dielectric Properties
of Polymers,” Journal of Korean Physical Society, vol. 49, no. 2, p. S.
513, 2006.
[29] M. Schwerdtfeger, E. Castro-Camus, K. Kr¨
ugener, W. Vi¨
ol, and
M. Koch, “Beating the wavelength limit: three-dimensional imaging
of buried subwavelength fractures in sculpture and construction mate-
rials by terahertz time-domain reflection spectroscopy,” Applied Optics,
vol. 52, no. 3, pp. 375–380, 2013.
[30] N. Fuse, T. Fukuchi, T. Takahashi, M. Mizuno, and K. Fukunaga,
“Evaluation of applicability of noncontact analysis methods to detect rust
regions in coated steel plates,” IEEE Transactions on Terahertz Science
and Technology, vol. 2, no. 2, pp. 242–249, 2012.
Junliang Dong (S’16) received the M.S. degree
in control science and engineering from Tsinghua
University, China, in 2011. He is currently working
toward the Ph.D. degree in Georgia Institute of
Technology. His research interests include terahertz
imaging and spectroscopy, terahertz optics, and non-
destructive testing.
Alexandre Locquet received the M.S. degree in
electrical engineering from the Facult´
e Polytech-
nique de Mons, Mons, Belgium, the Ph.D. de-
gree (doctorat) in engineering science, and electrical
and computer engineering from the Universit´
e de
Franche-Comt´
e, Besanc¸on, France, and the Ph.D.
degree in electrical and computer engineering from
the Georgia Institute of Technology (Georgia Tech),
Atlanta, GA, USA. He is currently a Researcher
with the Unit´
e Mixte Internationale, Georgia Tech-
CNRS Laboratory, Georgia Tech Lorraine, Metz,
France, and an adjunct professor with the School of Electrical and Computer
Engineering, Georgia Tech. His research interests are in semiconductor laser
dynamics and chaos, nonlinear time series analysis, physical-layer security,
and terahertz imaging. He has authored or co-authored over 40 journal
publications and conference presentations, and one book chapter. He is a
member of Eta Kappa Nu, the Optical Society of America (OSA), and the
IEEE Photonics Society.
D. S. Citrin (M’93-SM’03) received the B.A. degree
from Williams College, Williamstown, MA, USA,
in 1985, and the M.S. and Ph.D. degrees from the
University of Illinois Champaign, IL, USA, in 1987
and 1991, respectively, all in Physics. From 1992 to
1993, he was a Post-Doctoral Research Fellow with
the Max Planck Institute for Solid State Research, S-
tuttgart, Germany, where he was involved in exciton
radiative decay in low-dimensional semiconductor
structures. Subsequently, from 1993 to 1995, he was
a Center Fellow with the Center for Ultrafast Optical
Science, University of Michigan, Ann Arbor, MI, USA, where he studied
ultrafast phenomena in quantum wells. He was then an Assistant Professor
of Physics with Washington State University, Pullman, WA, USA from 1995-
2001. In 2001, he joined the faculty of the Georgia Institute of Technology,
where he is currently a Professor of Electrical and Computer Engineering. In
addition, he coordinates the research effort on Nonlinear Optics and Dynamics
with the Unit´
e Mixte Internationale, Georgia Tech-CNRS UMI 2958 at
Georgia Tech Lorraine, Metz, France. His research interests include terahertz
science and technology, nonlinear dynamics in external-cavity semiconductor
lasers, and nanophotonics. He has served as an Associate Editor of the IEEE
JOURNAL OF QUANTUM ELECTRONICS. He was a recipient of the
Presidential Early Career Award for Scientists and Engineers and the Friedrich
Bessel Prize from the Alexandre von Humboldt Stiftung.
