Over the last several decades, broadband pulses of terahertz (THz) radiation have sparked many new forms of research due to their non-destructive and non-ionizing characteristics [1
R. A. Cheville, M. T. Reiten, J. O'Hara, and D. R. Grischkowsky, “THz time domain sensing and imaging,” R. J. Hwu, and L. W. Dwight, eds. (SPIE, 2004), pp. 196–206.
L. Duvillaret, F. Garet, and J.-L. Coutaz, “Highly precise determination of optical constants and sample thickness in terahertz time-domain spectroscopy,” Appl. Opt.
38(2), 409–415 (1999). [CrossRef]
]. For example, terahertz time-domain spectroscopy has been widely used to investigate the physical properties of many nonpolar and nonmetallic materials such as ceramics, plastic explosives, and paintings, because they all exhibit spectral features in the terahertz region [4
T. Yasui, T. Yasuda, K.-i. Sawanaka, and T. Araki, “Terahertz paintmeter for noncontact monitoring of thickness and drying progress in paint film,” Appl. Opt.
44(32), 6849–6856 (2005). [CrossRef]
]. If a material is measured in a reflection geometry, only a portion of the THz-pulse energy will be transmitted through the boundaries of different materials due to the differences in refractive index. The resulting reflections can be exploited to obtain spectral and spatial information about layers of materials or even elements within a system. This reflection-geometry experimental embodiment is known as terahertz time-domain reflectometry (THz-TDR), and it can be effectively used when performing spectroscopy, imaging, and ranging.
Material thickness and refractive index are the two most common properties measured by THz-TDR. Using a generic two-layer system as an example, where layer 1 is a transparent dielectric and layer 2 is electrically conductive, Fig. 1
shows the principal of the measurement of either thickness or refractive index of the dielectric. When the THz pulse is incident on the insulator’s open surface, it is partially reflected at the air/dielectric boundary – due to the discontinuity in the group refractive index – and then totally reflected at the dielectric/metal interface. The time delay Δt between two pulses is given by
is the group refractive index of the dielectric in the THz region, d is the thickness of the dielectric layer, c is the velocity of light in vacuum and φ is the reflection angle at the metallic layer. According to Eq. (1)
, the time delay can be used to determine the insulator-coating refractive index, if its thickness is known, or the layer thickness if one has knowledge of its refractive index.
Fig. 1 Principal of the thickness and group refractive index measurement by using THz-TDR.
2. Layered material system under study
Many organic, high explosive, polymer, and other materials interact with air, moisture, or heat during fabrication, storage, use, or shipping. This potentially leads to hardening, weakening, corrosion, or even failure. One critical material system that happens to degrade over time due to operation is the thermally insulating Thermal Barrier Coating (TBC), a heat-protection layer that coats metal blades and other turbine components found in aircraft engines and power-generation equipment. This coating enhances the temperature differential between the air and the alloy, thus reducing the oxidation and corrosion of the alloy and protecting the physical integrity of the turbine blade, as well as increasing the lifetime, efficiency and reliability of turbine engines. Common TBC structures usually contain three layers: a ceramic oxide, such as Yttria-stabilized Zirconia (YSZ), a bond coat (BC), and a Ni-alloy bulk substrate. The BC deposited at the ceramic/Ni-alloy interface, typically containing Al and other elements, is used to promote YSZ adhesion to the Ni-alloy turbine parts. During operation, the YSZ layer, intentionally created to be slightly porous to maintain its integrity during thermal expansion and contraction, is penetrated by air, creating a thermally-grown oxide (TGO) of alumina, and consequently internal stress, between the BC and the YSZ. Stress levels are further increased during thermal cycling of the TBCs due to thermal mismatches among the bond coat, TGO, and YSZ. This mismatch creates voids in the YSZ coating and would eventually lead to delamination of the YSZ and failure of the coating and part [5
O. Trunova, T. Beck, R. Herzog, R. W. Steinbrech, and L. Singheiser, “Damage mechanisms and lifetime behavior of plasma sprayed thermal barrier coating systems for gas turbines–Part I: Experiments,” Surf. Coat. Tech.
202(20), 5027–5032 (2008). [CrossRef]
M. Y. He, J. W. Hutchinson, and A. G. Evans, “Simulation of stresses and delamination in a plasma-sprayed thermal barrier system upon thermal cycling,” Mater. Sci. Eng. A
345(1-2), 172–178 (2003). [CrossRef]
], if the part was not taken out of service and re-coated. The failure of thermal-barrier coatings is taken very seriously due to the potential for performance degradation and the catastrophic threat to safe operating conditions. To address this potentially dangerous situation, reliable diagnostic tools to identify the severity and location of degradations within thermal-barrier coatings are vigorously pursued [9
W. A. Ellingson, R. J. Visher, R. S. Lipanovich, and C. M. Deemer, “Optical NDT Methods for Ceramic Thermal Barrier Coatings,” Mater. Eval.
64, 45–51 (2006).
K. W. Schlichting, K. Vaidyanathan, Y. H. Sohn, E. H. Jordan, M. Gell, and N. P. Padture, “Application of Cr3+ photoluminescence piezo-spectroscopy to plasma-sprayed thermal barrier coatings for residual stress measurement,” Mater. Sci. Eng. A
291(1-2), 68–77 (2000). [CrossRef]
Given that TBCs are relatively transparent to far-infrared radiation, THz-TDR was chosen to investigate the non-destructive inspection of the condition of the ceramic/metal interface during a variety of simulated service times at an appropriate elevated temperature [12
L. Hess, and R. A. Cheville, “Nondestructive evaluation of ceramic bearings using THz impulse ranging,” in Conference on Lasers and Electro-Optics (2004), p. 2, vol.1.
J. I. Eldridge and T. J. Bencic, “Monitoring delamination of plasma-sprayed thermal barrier coatings by reflectance-enhanced luminescence,” Surf. Coat. Tech.
201(7), 3926–3930 (2006). [CrossRef]
]. Compared to other nondestructive evaluation methods, THz-TDR utilizes pulses that penetrate better and are scattered less within typical TBC layers, such as the one investigated here that was air-plasma-sprayed onto a rough surface of grit-blasted bond coat. In this paper, time-of-flight terahertz measurements have been found to reveal the presence and growth of the thermally-grown oxide layer with a resolution on the order of a single micrometer. This sensitivity to intermediate layers of various refractive index can then be used to monitor the progression of the buried air voids that ultimately lead to TBC failure. In Fig. 1
, which depicts a material system that has yet to undergo thermal cycling, layer 1 would be the YSZ ceramic, while layer 2 would be the metallic bond coat. The alumina TGO and the voids would be additional intermediate layers appearing between layers 1 and 2.
The interface of the TBC sample is shown before oxidation in a scanning-electron-microscope (SEM) image in Fig. 2(a)
. The air plasma sprayed TBC has a thickness of 300 ± 50 μm, and the BC layer is 50 ± 25 μm, where the thickness variations are local and frequent – around an average value within the area of the THz beam spot on the sample – and are not due to measurement uncertainty. Given the relatively large spot size of the THz beam, it is believed that the measurements yield information that is averaged over the thickness variations of the various material components. The YSZ/BC interface is very rough due to the fact that air-plasma-sprayed TBC samples require such a surface to reduce strain and premature delamination.
Fig. 2 SEM images of the YSZ interface (a) before thermal cycling and oxidation, (b) after 100 hours, (c) after 348 hours, (d) after 790 hours, (e) after 1100 hours and (f) after 1350 hours at 1100 °C. Each SEM image is taken at a different location along the cross-section of the multi-layer-sample material system.
In order to monitor the oxidation process, the sample was heated in a tube furnace at 1100°C for 1350 hours to simulate operational service time. Since spallation is the continuous evolution of damage in the TBC, with the BC oxidation process playing a key role in the formation and growth of microcracks, the sample was removed periodically to conduct the THz measurements and to cut a small piece from the edge of the TBC sample for the SEM interface inspection. After continuous thermal exposure for 100 hours, the TGO, which is the dark grey layer between the YSZ and BC, was formed as shown in Fig. 2(b)
. The average TGO thickness, as suggested by Fig. 2(c)
, rapidly increased with thermalization time to about 5 μm within the first 348 hours of exposure, before the rate of growth slowed. Infrequent small voids were initially observed in the sample that was thermally exposed for 348 hours and were typically found at positions where the TGO was non-planar, causing the YSZ to experience larger internal stress. The voids increased in number and size with the thermal exposure, as seen in particular in Fig. 2(e)
. The air voids coalesced to form more uniform, nearly continuous gaps in the YSZ layer after 1100 hours in the tube furnace. After 1350 hours, the edge of the YSZ layer seen in Fig. 2(f)
delaminated (where the delamination was not in view of the SEM). Due to polishing artifacts such as “pull-out” and chipping, and also as a result of epoxy-mount shrinkage that weakens interfaces, the void and interface gap sizes shown particularly in Figs. 2(e)
appear exaggerated beyond what would actually be present in the specimen before cross-section preparation.
3. Multilayer simulation
The SEM images show that the average thickness of the TGO and the air gap are on the order of several micrometers. According to Eq. (1)
, the time delay Δt between reflected THz pulses for a few-micrometer thickness should be less than 1 picosecond at the YSZ/Air, Air/TGO, and TGO/BC interfaces. Compared to the original THz pulse width of ~3 ps, the time-domain signal is thus expected to be a single pulse, comprised of a superposition of multiple reflected pulses, as compared to a series of distinct, consecutive pulses in the time-domain.
In order to better interpret the reflected THz-pulse behavior, a time-domain simulation of the multilayer conditions observed in the SEM images was conducted using a Fresnel-reflection analysis with lossless, non-dispersive layers and a realistic THz incident transient, which had a frequency content ranging from 0.2 to 1.5 THz [14
D.J. Lee and J.F. Whitaker, “Spectro-temporal and echo-response analysis of arbitrary optical multilayer systems,” submitted to Appl. Optics, Nov. 2009.
]. A digitized version of a measured THz input pulse was used as the input to the Fresnel-reflection algorithm. With the initial reflected THz pulses from the air/YSZ interface aligned so that they were always coincident in time, the positions and shapes of the second reflected THz pulse from the internal multilayers for different interface-layer-thickness conditions were computed (Fig. 3
). The simulation assumed idealized dielectric layers of uniform thickness. The red line in Fig. 3
represents the THz signal with no interface oxidation (or additional interface layers), the black line is the THz signal reflected from the YSZ/TGO and TGO/BC interface system, and the blue and green lines simulate the interface with additional air gaps between the YSZ and TGO. Multiple reflections within the relevant surfaces were observed to delay and either narrow or broaden the THz interface reflection, due to the superposition of numerous round-trip THz signals, some which experience π-phase shifts upon reflection and some that do not. Inclusion of the internal reflections from the larger air-filled voids also appears to damp out the last noticeable oscillation in the interface reflection of the THz signal.
Fig. 3 Simulation results of the reflection of a THz-pulse from the TBC/BC interface for different interface conditions
The delay between the YSZ front-face reflection and the internal-interface compound reflection was quantified by considering both the negative-valley-to-negative-valley time difference for the two reflections, as well as their positive-peak-to-positive-peak time difference. The time delays of these components from the second THz-pulse reflections, simulated from the idealized interfaces in Fig. 3
, were thus reduced and plotted against the hypothesized accumulated sample heating times in Fig. 4
. Four data points for both the pulses’ negative and positive peak shifts are given for four different sample conditions: YSZ + BC, YSZ + TGO (5µm) + BC, YSZ + Air Gap (1µm) + TGO (5µm) + BC, and YSZ + Air Gap (8 µm) + TGO (5µm) + BC, corresponding to increasingly long exposure times. The delays of the negative and positive components both increase together as the TGO becomes thicker, up until the voids begin to appear at the 348 hour mark. This is plausible, considering the fact that the refractive indices of the YSZ and the TGO are similar, so that the largest reflections should arise at the YSZ front face and at the TGO/BC interface. (The refractive index for YSZ is 4.5, as shown in the next section, and the refractive index for TGO is expected to be 3.5 for 0-3 THz [15
N. Matsumoto, T. Nakagawa, A. Ando, Y. Sakabe, S. Kirihara, and Y. Miyamoto, “Study of Multilayer Ceramic Photonic Crystals in THz Region,” Jpn. J. Appl. Phys.
44(No. 9B), 7111–7114 (2005). [CrossRef]
].) Thus, as the TGO grows, the two dielectric layers of YSZ and TGO act basically as one continuously thickening layer, and the entire THz-pulse reflection is delayed with respect to the reflection from the front-face.
Fig. 4 The delay time of the second reflected THz pulse from simulation versus the heating time in furnace
The TGO growth rate slows as the voids begin to form, although it mostly still increases with thermal exposure as observed in the SEM images. In this range, despite the increasing thickness of the TGO/void layers and the eventual formation of uniform air gaps, the delays of the THz-pulse negative and positive peaks essentially stop increasing. It appears that this happens due to a trade-off that develops between competing mechanisms. One is an increasing THz-pulse delay from the increasing thickness of the total interface-layer. The other is that multiple reflections at the multi-dielectric interface, some of which are π-phase-shifted with respect to each other (due to the refractive indices present), become superimposed when they contribute to the overall THz-pulse reflection from the internal-interface. That is, when all the THz-pulse reflections, phase shifts, and time delays are taken into account by the simulation algorithm, both the negative and positive peaks exhibit very small shifts with respect to each other, up until the point where the air gaps become very large (and delamination occurs). The simulation results also show that the multilayer system with the large, delamination-like air gap broadens the pulse width dramatically (with the positive part of the pulse delayed from the incident pulse more than the negative part), and the contrast ratio between the peak and valley of the combined reflected THz pulse is reduced because of the deconstructive interference. As the air gap increases in thickness, the THz-pulse reflection from the YSZ/air-gap interface is separated from the other multiple reflections, and the increasing separation of the multiple reflections causes the time-domain tail of the THz pulse to rise (as seen in Fig. 3
4. Experimental results
For the THz-TDR experiments, a conventional THz-imaging system as shown in Fig. 5
was employed, utilizing low-temperature-grown GaAs emitter and receiver elements driven by 80 fs laser pulses (800 nm wavelength) at an 80-MHz repetition rate. The TBC sample was placed on a motorized translation stage to perform line scans and investigate sample uniformity using THz pulses with bandwidth between 0.2 and 1.5 THz.
Fig. 5 THz-TDR experimental Setup
The TGO and void formation were detected by comparing the delay of the second THz-pulse peak, reflected from the internal interface, with the initial THz reflection from the TBC top surface, as shown in Fig. 6
. The red line represents the THz signal before the BC on the TBC sample was oxidized, with the delay of the second THz pulse being caused only by the extra optical path inside the YSZ layer. Since the YSZ thickness is known, the refractive index of YSZ was calculated to be 4.5 ± 0.4 from the delay of 9.37 picoseconds in the 0.2-1.5 THz range. Up until the voids were observed at 348 hours, the interface-reflection transition (i.e.,
falling) edges starting at about the 13 ps point in time of Fig. 6
are all virtually aligned, as one would expect for a TBC-layer thickness that did not change with thermal exposure. The virtually aligned falling edge was observed in multiple samples for as many as five occasions of thermal cycling up to the 348-hour exposure point, although only one thermal cycling, 130 hours, is shown in Fig. 6
. However, when the THz pulse encounters the TGO “etalon,” the remaining elements of the THz transient, starting at about 14 ps, are delayed. Furthermore, the additional etalon formed by the voids for heating times in excess of 348 hours leads to a delay even in the transition edge, as shown in Fig. 6
, for the longer thermal-exposure times. The rate of the pulse delay-time increase began to slow after 620 hours, corresponding to the average enhancement of the TGO thickness also exhibiting a decreased rate.
Fig. 6 Experimental reflected THz transients for a sample heated at 1100 °C for five different accumulated thermal-exposure times between 0 and 1350 hours
The THz-TDR experimental results indicate a larger pulse delay in time compared to the simulations in Fig. 3
. This is likely due to the assumptions inherent in the simulation of ideal conditions, i.e.,
the simulation considered neither scattering effects inside the porous ceramic oxide nor the roughness of the interfaces across the whole layer. However, the experimental results qualitatively follow the simulation in its time-delay trends and the changes in pulse shape. Comparing the purple and green lines of the experimental results in Fig. 6
with the green curve of the simulation data in Fig. 3
, the gradually damped oscillation becomes more obvious with the increasing time spent by the sample in the furnace. Based on the simulation results, the damped oscillation originates from the separation of the multiply-reflected THz pulses, the results indicating that the air gap or the TGO thickness is large enough to cause a separation in the internal reflection at the YSZ/air-gap and air-gap/ TGO interfaces.
The delays in the negative valleys and positive peaks of the second THz pulse after the round-trip through the TBC sample of the THz pulse are plotted against the accumulated sample heating times and presented in Fig. 7
. The differences in the valley-to-valley and peak-to-peak delay times demonstrate that the second reflected THz pulse appears to narrow after 348 hours. Additional pulse narrowing, which can also be confirmed from the change in the slope of the negative-to-positive transition in the THz transient as shown in Fig. 6
, occurs with the increasing thermalization time up to 1100 hours. The pulse narrowing effect is hinted at by the simulations, but not at all to the degree seen in the experiments. However, the uniform THz-pulse delays observed during the air-void and then air-gap formation did appear to change dramatically, as suggested by the simulation, after 1100 hours exposure time, due to the air gap being large enough to separate the reflected pulse at the YSZ/air-gap boundary from the other internally reflected pulses. This pulse separation can also be confirmed from the experimental THz time-domain signal of Fig. 6
, where the damped-out oscillation of the green curve has only two small peaks at the end of the trace. The increasing delay time with thermal exposure and TGO/void growth was confirmed for multiple TBC samples.
Fig. 7 The delay time of the second reflected THz pulse versus the heating time in furnace
Finally, Fig. 8
compares the average TGO thickness obtained by SEM measurements with the delay time of the reflected THz pulse (negative-peak-to-negative-peak). The delay time of the THz pulse rapidly increased for thermal-exposure times up to 348 hours, and the SEM images confirmed that the TGO thickness was increased from 0 to 5 micrometers within the first 348 hours. The average TGO thickness had increased to 9 micrometers by 1300 hours of exposure, although, the variation of the TGO thickness was much greater than during the first 348 hours. Overall, the trend in the delay time of the THz pulse matches the average TGO thickness obtained from the SEM measurements.
Fig. 8 Comparison between average TGO thickness from SEM measurements and the delay time of THz pulse.
Overall, three unique features in the experimental THz time-domain signal can be used to monitor the TBC health condition. First, the increasing delay time of the second reflected THz pulse with respect to the YSZ front-surface reflection corresponds to the TGO growth. Second, the constant delay time of the second reflected THz pulse indicates that air voids were formed inside the YSZ layer. Third, the broadening of the THz internal-interface reflection and the damped-out oscillation in the tail of this THz pulse suggests a warning that delamination is likely imminent.
The use of THz-TDR to nondestructively study multilayered thermal-protection systems is novel, sensitive, and readily conceivable, although it is still a process that is under development. In this paper, THz-TDR experimental results have demonstrated the ability to monitor the evolution of thermally-induced oxide layers and voids – embedded at a ceramic/metal interface – that are on the order of a single micrometer in thickness. This was accomplished through the observation of THz pulse time delays and changes in the width and shape of the THz pulses. The experimental data also qualitatively follow Fresnel-reflection-based time-domain simulations of THz pulses in etalons. A comparison with SEM images further shows that the average TGO thickness matches the trend in the delay time of the initial and interface THz pulse reflections. Refinement of the technique presented here could lead to diagnostics that predict the failure of turbine-blade coatings. This would allow TBC replacement to be based on THz assessment rather than on a fixed timetable, although since changes in delay times are analyzed rather than the delay times themselves, the technique will only be effective if a history of pulsed reflectometry measurements is maintained.