Open Access
Add to BibSonomy
Abstract
Recent developments and commercial availability of low-noise and bright infrared (IR) supercontinuum sources initiated intensive applied research in the last few years. Covering a significant part of near- and mid-infrared spectral ranges, supercontinuum radiation opened up unique possibilities and alternatives for the well-established imaging technique of optical coherence tomography (OCT). In this contribution, we demonstrate the development, performance, and maturity of a cost-efficient dual-band Fourier-domain IR OCT system (2 µm and 4 µm central wavelengths). The proposed OCT setup is elegantly employing a single supercontinuum source and a pyroelectric linear array. We discuss adapted application-oriented approaches to signal acquisition and post-processing when thermal detectors are applied in interferometers. In the experimental part, the efficiency of the dual-band detection is evaluated. Practical results and direct comparisons of the OCT system operating within the employed sub-bands are exhibited and discussed. Furthermore, we introduce the 2 µm OCT sub-system as an affordable alternative for art diagnosis; therefore, high resolution and sensitive measurements of the painting mock-ups are presented. Finally, potentials of the dual-band detection are demonstrated for lithography-based manufactured industrial ceramics.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Since first reports on supercontinuum generation in the mid-infrared (MIR) spectral range [1–3], the unique parameters of the emission [4–6] as well as its stability have been significantly improved and optimized [7–11] making these sources technologically mature and attractive for applied research, especially in spectroscopy, imaging and material characterization fields [12–19]. Due to high output powers [20], particular coherence properties and spectral broadening towards longer wavelengths [21,22], nowadays, MIR supercontinuum lasers are becoming a game-changing factor in optical coherence tomography (OCT) raising close attention within the community.
OCT, as a well-established imaging technique, is widely used in bio-medicine [23,24] and non-destructive testing (NDT) [25] allowing non-invasive inspection of structures, sub-interfaces or hidden defects at a high spatial resolution. State-of-the-art OCT systems are commonly targeted to biological tissues and operate in visible (VIS) or near-infrared (NIR) spectral ranges. However, they suffer from inherited strong scattering, the magnitude of which is inversely proportional to the wavelength of light and has been a physical constraint limiting the effective penetration depth of industrial OCT systems. Recently predicted performances of long-wavelength MIR OCT [26] have been experimentally verified by several groups [27–29], despite the challenging detection solutions. Thereby, the expedient sensitivity and enhanced penetration depth (at reasonable measurement times) have been successfully demonstrated for the inspection of normally opaque samples. These results have obviously initiated deeper material testing research and promise to have a significant impact on, for instance, in-line monitoring of defects in lithography-based ceramic manufacturing (LCM) [30,31], artworks preservation and restoration [32,33] or in the pharmaceutical industry. However, maintaining a high (e.g. comparable to NIR OCT systems) axial resolution, which is proportional to the square of the central wavelength, is a complex task and possible only for a significantly extended bandwidth, which in turn causes a sharp drop in the signal roll-off [34,35]. In such case, a multi-band OCT system exhibiting the strengths of different sub-bands that complement each other, i.e. penetration and resolution, could be beneficial. The current extension of the OCT technology to new spectral ranges is expected to generate original technical solutions and to reveal novel potentials of OCT for the investigation of formerly problematic materials, such as ceramics, polymers, paintings, pharmacy etc.
In our contribution, we demonstrate and assess a dual-band NIR and MIR Fourier-domain (FD) OCT. The system preserves the high axial resolution within the short-wavelength 2 µm sub-band, while the MIR OCT (4 µm range) enables to increase the probing depth in highly scattering materials. The developed setup employs a Czerny-Turner-type spectrometer that is built on a single pyroelectric detector array. Due to the inherent long-term integration of this thermal detector, the pulse-to-pulse noise, which is usually an issue in supercontinuum-based OCT [36], is not detectable. The multi-band measurements are achieved using the detection of overlapping diffraction orders to be mechanically selected by spectral filtering. Furthermore, a purely OCT-oriented detection approach with an advanced post-processing is implemented and considered in detail. In summary, an elegant and affordable instrument that is based on a single IR supercontinuum source is introduced and demonstrated.
2. Dual-band Fourier-domain IR OCT system
2.1 Supercontinuum source
The dual-band sensing approach being presented in our study is achieved by means of a single commercially available supercontinuum source (NKT Photonics, SuperK MIR). The spectrum spanning from 1.1 µm to 4.4 µm is generated within a step-index ZrF$_4$-BaF$_2$-LaF$_3$-AlF$_3$-NaF (ZBLAN) fiber (two zero-dispersion wavelengths) that is pumped by an all-fiber picosecond laser with two-stages, i.e. erbium-ytterbium and thulium, amplifier. Through the soliton related dynamics, non-linear and dispersive properties of the fiber [37,38], ultra-broadband emission is delivered. The emission spectrum of the supercontinuum source recorded with a Fourier transform IR spectrometer (FTIR, Vertex 70, Bruker Optics) is depicted in Fig. 1, the OCT sub-bands are indicated. In order to avoid the oversaturation of the FTIR’s HgCdTe (MCT) detector, a neutral density filter (5% transmission) was applied; the sensitivity curve of the detector was used to compensate for the possible spectral distortion. NIR spectral components, for instance, a strong peak of the seed laser at 1.55 µm, were suppressed using an edge-pass filter.
Fig. 1. Emission spectrum of the supercontinuum source measured by an FTIR spectrometer; OCT spectral sub-bands (spectral interferograms recorded for the flat mirror using the OCT spectrometer) are indicated; emission power levels are denoted.
The source emits sub-nanosecond pulses at 2.5 MHz repetition rate, which is high enough with respect to the employed detector (operates at frequencies of tens of Hz) to consider the emission as quasi-continuous wave. A mean pulse-to-pulse relative noise of around 6% was specified over the spectrum using a monochromator (Gilden Photonics, GM500) and an MCT detector (time constant $\leq$0.7 ns). A detailed report on the source characteristics is given in [39]. Due to the inherent properties of the pyroelectric array, this noise is averaged out naturally and not detectable in OCT images. The average power of the supercontinuum is around 496 mW of which 200 mW are radiated in the spectral range beyond 2.4 µm (measured with a corresponding edge-pass filter). Figure 1 indicates the power levels within the sub-bands (measured using LabMax TOP Power Meter, Coherent, LM-10 HTD detector head). The outgoing fundamental Gaussian beam is collimated by a parabolic mirror and can be characterized by the beam quality parameter M2=1.1. The wavelength dependent diameter of the beam is around 3 mm for the 2 µm region and 6 mm for the 4 µm range, as observed using a bolometer focal-plane array (FLIR, Boson) and applying appropriate band-pass filters (center wavelengths CWL, $\lambda _c$, 2 µm and 4 µm respectively).
2.2 Optical system
The collimated beam of the supercontinuum radiation passes through a spectral filter (SF1, edge-pass, 1.65 µm cut-on wavelength) and enters the free-space Michelson interferometer (Fig. 2).
Fig. 2. Layout of the experimental setup, PBS - pellicle beamsplitter, SF1 and SF2 are the spectral filters utilized to suppress the 1.55 µm seed laser line (edge-pass filter, 1.65 µm cut-on wavelength), and to select the operational OCT spectral band; the inset displays en-face OCT images of an 1951 USAF resolution test target (resolved line widths are 39.37 µm and 12.4 µm for the NIR and MIR sub-systems respectively).
A pellicle beamsplitter (PBS, nitrocellulose membrane, 2 µm thick) was employed to arrange the reference (protected gold mirror) and sample arms of the interferometer. In addition, the PBS eliminates double reflections causing parasitic interferences, common for any thick-plate beamsplitter [27]. To focus the beam onto the sample, a BaF$_2$ lens (50 mm focal length) was utilized. It exhibits a higher refractive index than e.g. CaF$_2$, but has a weaker total dispersion over the wavelength range [40,41]. Thus, it causes a reduced focal-spot shift induced by chromatic aberrations. The scanning of the sample was achieved using an XYZ stage system. The lateral resolution of the system was characterized using the 1951 USAF resolution test target (see the inset in Fig. 2) yielding a resolution of around 39 µm and 12.5 µm for the 4 µm and the 2 µm OCT system respectively. From a practical point of view, focusing optics is challenging in the MIR spectral range and particularly for the broadband radiation: commercial achromatic doublets are available only for relatively narrow spectral bands, while parabolic mirrors are difficult for precise alignment.
The grating-based OCT spectrometer employs a pyroelectric linear array requiring light modulation for operation. In contrast to our previous report [27], the chopper is positioned in the sample arm of the interferometer. Thereby, we introduce a smart approach to detection which is adapted for OCT imaging (discussed in detail in section 2.4). The novel modulation concept allows to automatically exclude a background and extend the effective dynamic range.
2.3 Operation principle of the dual-band OCT
The dual-band measurements (see details in Fig. 3) are enabled by an optimized grating-based spectrometer [27] designed and adapted for a maximum performance in both OCT sub-bands (NIR band from 1875 nm to 2125 nm and MIR from 3750 nm to 4250 nm correspondingly).
Fig. 3. (a-c) Details on the dual-band detection enabled by a single pyroelectric array, (d) Characterization of the axial resolution of the system (17.5 µm and 37 µm at full-width at half-maximum (FWHM) for NIR and MIR OCT sub-systems correspondingly)
A blazed diffraction grating (300 lines/mm, Thorlabs), a concave gold mirror (100 mm focal length, Thorlabs, CM254-100-M01) and a pyroelectric linear array (DIAS infrared, 510 px, $20\times 500$ µm$^2$ pixel size) are employed to form the system that is schematically shown in Fig. 3(a). The optimal rectangular shape of the pixels eliminates the effects of astigmatism for the mirror-based spectrometer with a tilted spherical mirror [27]. The ability to switch between the NIR and MIR ranges using a single detector array is unique and provided by the ultra-broadband coverage of the MIR supercontinuum source, since the emission spectrum exhibits doubled wavelengths. Therefore, an overlap of the first (4 µm sub-band) and second (2 µm sub-band) diffraction order is exploited to sequentially record the interferograms applying a suitable band-pass filter.
Figure 3(b) illustrates a typical raw spectral interferogram recorded for a flat mirror surface, while only the neutral density (optical density, OD=1) and edge-pass (cut-on wavelength 1650 nm) filters were inserted during the measurements. The signal exhibits both NIR and MIR wavelength ranges simultaneously so that it carries two distinct frequencies. Thereby, the band-pass filter was applied to split the ranges and select the desired spectral window.
According to the Nyquist-Shannon-Kotelnikov theorem, the maximum imaging depth of the FDOCT is inversely proportional to the bandwidth sampling rate [34,35]. Therefore, considering the utilized detector array and the actual OCT spectral bandwidths, the effective probing depth limits of approximately 4.2 mm and 2.1 mm are derived for the MIR and NIR OCT sub-systems respectively [34]. The visual interpretation of the consequences of the sampling theorem, such as stretching and the axial resolution disparity, can be observed for the b-scan of a pharmaceutical coating shown in Fig. 3(c); since the OCT image shows a simultaneous output of both sub-systems, the scale of the Y-axis is not displayed.
The axial resolution of the OCT sub-system was characterized using a-scans of a gold mirror; during the measurements, a neutral density (optical density of 1) filter was inserted into the sample arm to avoid oversaturation of the array. The corresponding point spread functions [Fig. 3(d)] yield the resolutions (full-width at half-maximum) of 17.5 µm and 37 µm for the NIR and MIR FDOCT systems respectively. The a-scan rate of the OCT system is limited by the light modulation frequency, which, in practice, determines the heat accumulation time and has an effect on the total amplitude of the signal. For samples with strong surface and bulk scattering, the optimal range is 40–50 Hz. A chopping frequency of up to 80-100 Hz is used for highly reflective specimens to avoid oversaturation of the detector. Considering the post-processing applied to the raw signal, the maximum a-scan rate is around 40 scan/s.
2.4 OCT-adapted signal acquisition applying the pyroelectric detector array
The employed LiTaO$_3$ pyroelectric detectors are thermal detectors in essence [42]. This fact allows us to implement the measurements in a wide spectral range (surface blackening is commonly used to enhance absorption), and, more important, to conform the signal acquisition scheme to the specific requirements of OCT applications. The linear array that was used in our study comprises 510 px (characteristic dynamic range >75 dB), the responsitivity of the detector (dependence of the signal voltage on the incident radiation flux) is 680000 V/W; the dimensions of the rectangular pixels are $20\times 500$ µm$^2$. The thickness of the pyroelectric layer is reduced to 5 µm by means of the ion-etching technique in order to optimize the noise equivalent power (NEP=1.3 nW), while the pixel-to pixel cross-talk is only a minor effect due to the minimal gap (5 µm) between the elements [43].
Considering a single pixel of the pyroelectric array, the individual surface charge being amplified and detected is defined as [42,44]:
where $\gamma$ is a material specific pyroelectric coefficient that indicates the magnitude of the polarization evolution induced by temperature changes, i.e. $d\mathrm {P}/d\mathrm {T}$ (230 $\mathrm {\mu cm^{-2} K^{-1}}$ for LiTaO$_3$), $A$ is the area of the pixel. Here, $\overline {\mathrm {T}}$ is the spatial average of the temperature $\mathrm {T}$ of the crystal over its volume V. The average temperature of the sensing element is directly related to the absorbed light intensity $I$, i.e. $\overline {\mathrm {T}}\propto I$, thus, emanating the operation principle of this type of detectors.Due to thermal relaxation and the leakage current by free carriers, the generated surface charge $Q$ rapidly disappears, i.e. the response of the detector on a step function of the incident radiation drops down over time [45]. Hence, the pyroelectric detectors basically operate only with alternating or modulated illumination. Thereby, in the Michelson interferometer [27] that employs a supercontinuum emission totally modulated at the angular frequency $\omega$, the intensity of light $I$ absorbed by the individual element of the array can be expressed in the one-dimensional case as follows [32]:
Obviously, the light reflected from the reference mirror is dominant in comparison to the weak but essential component $\sqrt {I_{\mathrm {r}} I_{\mathrm {s}}}\cos {\Delta \phi }$ that carries the information about the structure. In practice, this means that the reference field acts first as an amplification factor and then as a constant intensity floor in the signal, which is ordinarily subtracted during the post-processing.
As discussed earlier [32], we propose and implement the OCT-adapted approach to the detection that implies the modulation of the relevant object field only, i.e. the chopper is inserted in the sample arm. Therefore, the strong pedestal of the signal that limits the actual dynamic range of the system is eliminated. Consequently, the moment pixel charge to be transformed into the signal voltage is proportional to the intensities as:
Since the amplified term normally exhibits significantly higher intensity than a part reflected from a scattering sample, i.e. $\sqrt {I_{\mathrm {r}} I_{\mathrm {s}}}\cos {\Delta \phi }\;>>\;I_{\mathrm {s}}$, the visibility of interference fringes tends to one allowing to simplify the traditional for OCT post-processing algorithms and increase integration time avoiding oversaturation.In Fig. 4 the advantages of the described solution are demonstrated. It depicts raw interferograms that were recorded for a simple multilayer ceramic sample using the conventional method of total emission modulation (reported previously [27]) as well as the proposed detection approach.
Fig. 4. Comparative measurements: raw interferograms recorded for the multilayer ceramic sample depending on the chopper position (system input versus the sample arm) in OCT interferometer; in the case of total emission modulation, a neutral density filter (OD=0.3, 50% transmittance) was additionally inserted to avoid oversaturation; approximately 2-times higher visibility advantage is achieved (neutral density filter is taken into account).
The negative values are real since the pyroelectric array does not register intensity, but temperature changes (AC sensing mode) that could have the opposite sign. It should be noted, however, that the modulation depth of the signal in the standard (first) configuration is affected stronger, since a neutral density filter (50% transmission) was applied to avoid oversaturation of the array (limiting level is 2.5 a.u. with respect to the axis in Fig. 4). Other parameters (lateral position, 95 Hz modulation frequency, no averaging, and 1 ms trigger delay) remained unchanged for both measurements.
2.5 Roll-off steering and advanced post-processing
Figure 5(a) depicts a characterized sensitivity roll-off for both OCT sub-bands; a total system sensitivity of around 82 dB was measured at 100 Hz modulation frequency (i.e. shortest integration time) for a gold mirror (98% reflectivity) positioned near the zero delay plane [46]. Low modulation frequencies provide a longer integration time, which increases the signal-to-noise ratio. During the sensitivity evaluation, a neutral density filter (10% transmittance, 1% to be detected due to the double pass) was inserted in the sample arm of the interferometer in order to avoid oversaturation of the pyroelectric array. The specified roll-off rates are determined by the spectral resolution of the spectrometer [34,35,47] that is, in practice, various in the exploited ranges. The main reason limiting the resolution in the NIR sub-band (the sharper drop-off) is not optical but introduced by the detector array, i.e. the fixed number of pixels and their geometry. The imaging depth ranges of the 2 µm system defined at the 6 dB roll-off length could still be optimized applying a higher resolution detector array. A bolometer focal plane array (FLIR, Boson, 12 µm pixel size, 640$\times$480 pixels) was used for precise alignment of the characterized system. Figure 5(b) illustrates the resulting beam profile with an interference pattern recorded at the output of the interferometer (mirror is slightly out of focus).
Fig. 5. Sensitivity roll-off evolution on the axial range (a); and (b) interference pattern over the beam cross-section recorded with a band-pass filter (500 nm, 4 µm CWL).
Another consequence of using the thermal detector, which was first observed experimentally, makes the maximum sensitivity position of the system flexible (see Fig. 6) allowing us to extend the post-processing tools commonly applied in OCT imaging. The read-out cycle of the pyroelectric array starts with the first stage of charge-to-voltage (Q/V) conversion (sensing element operates in a current mode) that requires a lowpass settling time to stabilize. The actual integration time that is defined by two clock parameters and triggering delay starts when the amplifier transfers the signal from the Q/V converter to the sample-hold stage to be acquired using a multiplexer-amplifier circuit. Since the signal level is a function of time and temperature, both parameters should be taken into account with respect to the sampling.
Fig. 6. Roll-off steering: (a,b,c) MIR OCT b-scans of the multilayer ceramics (tilted by 45$^{\circ }$) recorded for different chopper trigger delay; modulation frequency 50 Hz.
The responsive elements of the pyroelectric array are thermally isolated between each other, however, they are fixed on the supporting frame that acts as an uniform heat sink. Therefore, for the linear, i.e. one-dimensional, array that could be approximated as a a thin rod, the heat flow and thermal dissipation are described by Fourier’s law of heat conduction. Consequently, as known, the heat flux density $\Pi$ is proportional to the negative local gradient in the temperature expressed as follows:
where $\kappa$ is a conductivity coefficient of the material.Obviously, for the steady-state condition (4), the magnitude of temperature gradients is dependent on the spatial frequency $f_{\mathrm {f}}$ of the interference pattern (correspond to depth components in FDOCT imaging). Since a slope, i.e. the maximum gradient at the peak temperature $\mathrm {T_{max}}$ of the fringe with respect to the surrounding temperature floor $\mathrm {T_{e}}$, is expressed as:
Therefore, the heat evolution and diffusion $\mathrm {T}(x,t)$ within the uniform rod would vary over time for high and low frequency interference patterns (i.e. spatially distributed thermal peaks of the different width) as described by a one-dimensional heat equation:
where $\alpha$ is a thermal diffusivity coefficient that represents the thermal inertia of the media and assumed to be constant for our case.The practical meaning of the above reasoning reflects the features of the time sampling and integration for the signal that is periodically generated over the array due to the light modulation. Hence, high-frequency interference components that are related to deeper structures in OCT image would be detectable prior to the appearance of low-frequency fringes of equal amplitude (due to high energy density and quick heating). However, the former rapidly disappear due to the thermal conductivity and relaxation, while the latter would produce the charges with a delay (due to the low energy density), but the low-frequency signal will be detectable over a longer time. In view of this aspect, the b-scans of the multilayer ceramics sample (similar to one used in [26–28]) were obtained for different trigger delays (Fig. 6), i.e. the integration was shifted relative to the modulation phase.
Therefore, by tuning the signal read-out delay, which might be performed with microseconds steps, the roll-off of the system can be steered enhancing either low- or high-frequency components of the interference patterns; this effect is slightly visible in Fig. 5(a), since the maximum sensitivity is shifted from zero depth position. Thereby, the scan acquisition applying the varying integration parameters enables a fully-electronic OCT image compounding as depicted in Fig. 6(d). Besides the common post-processing of the data (remapping into $k$-space, filtering, and discrete Fourier transform [27]), the a-scan averaging within the modulation-trigger-delay space is implemented to enhance the visibility of deeper sub-interfaces.
2.6 Samples
2.6.1 Multilayer ceramic sample
For the direct comparison of the OCT system operating in both sub-bands, the multilayer ceramic stack (Al$_2$O$_3$, porosity 1%, mean air pore diameter 0.4 µm), which has practically turned out to be a benchmark sample in recent studies [26–28], was utilized. The specimen is composed of two plates with the thicknesses of 300 µm and 450 µm (top and bottom stratums). A series of microchannels (65 µm height, variable width) is milled in the top interface of the rear plate.
2.6.2 Painting test samples
The spectral window of around 2 µm is a well-known range optimally suitable for OCT imaging of art objects [48]. In order to introduce the developed 2 µm OCT sub-system as a reasonable and affordable alternative for non-invasive imaging in art conservation (the state-of-the-art systems [33,49] are based on the expensive MCT or InSb cameras), the multilayer mock-up that is made up of different commercial paints was prepared. As the first layer, a gouache madder lake red (Nevskaya palitra, PR63:1 and PR187 constituent pigments) paint was applied. Further, two oil paints were deposited: titanium white (Nevskaya palitra, PW6, PW4 pigments), and naples yellow light (Nevskaya palitra,PY83, PY3, PW6 pigments). The paint test sample was covered by a varnish layer resulting in a total thickness of around 400 µm.
For the complex analysis, a pure oil painting mock-up was prepared. The sample consisted of the already listed naples yellow and titanium white mixed with cadmium red (Nevskaya palitra, PR108) and yellow green (Vista artista studio, PG7, PY1, PW6) to imitate the brush-style painting. The mock-up was covered with varnish.
2.6.3 LCM sintered alumina monolith
In order to show the potentials of the system for at- and on-line process control for additive ceramic manufacturing, an especially designed ceramic monolith (Al$_2$O$_3$) was fabricated by means of the LCM technology [50], a revolutionary method developed during the last years. Figure 7(a) demonstrates a 3D design model of the sample with denoted dimensions.
The additive manufacturing system, which was used to produce the test sample, is characterized by a lateral resolution of 40 µm and axial resolution (i.e. thickness of the layer) of around 25 µm. The component was printed vertically (perpendicular to the following OCT measurements), so that the layers were grown starting from the back face. Details of the printed specimen are shown in Fig. 7(b). The ceramic material is quite inhomogeneous and has a high porosity of 1.5%. A strong surface roughness, which depends on orientation, is characterized by the average value 1 µm < $R_a$ < 1.5 µm and makes the sample opaque for e.g. state-of-the-art 1.3 µm systems due to the surface scattering. The typical refractive index of the alumina is 1.768 and was utilized for the calculation of the depth axis; the relative density of the material is around 98.5% - 99%, theoretical density is 3.985 g/cm3.
3. Results and discussions
3.1 Direct comparison of the near- and mid-infrared OCT system
Figure 8 demonstrates a b-scan that is obtained within both NIR and MIR OCT sub-bands and presented as a single RGB-compounded image.
Fig. 8. Comparative measurements of the highly scattering ceramics; obtained by the dual-band OCT system operating in the near- and mid-infrared spectral ranges (central wavelengths of 2 µm and 4 µm correspondingly) and presented as an RGB-compounded b-scan; the left part of the image demonstrates a weighted overlay of both OCT measurements.
During the measurements, the multilayer ceramic sample was tilted by 45$^{\circ }$ in order to avoid artifacts (ghost surfaces and vertical stripes) caused by the strong reflection from the polished top interface, thereby, the total geometrical thickness of the inclined specimen to be imaged was around 1.1 mm. The lateral as well as longitudinal positions of the sample were fixed; the same data acquisition (no averaging) and post-processing algorithms were applied.
The left part of Fig. 8 exhibits the depth-weighted overlay of both scans, the OCT images were stretched and stitched numerically; the Y-axis, i.e. the effective depth, is calculated as an optical path difference divided by the effective refractive index. Hence, the combined measurements reflect the benefits of the dual-band IR system, i.e. resolution and penetration depth, for inspection of highly scattering materials. With respect to the probing depth of the system, the visibility of the top ceramic layer (420 µm thick while tilted) and a specific signal decay caused by multiple scattering are precisely consistent with the prominent results of simulations that predict the performance of OCT at longer wavelengths [26]. The bottom ceramic layer is fully accessible for MIR OCT, while only the top surface, i.e. the microchannels sub-interface, was partly revealed by the 2 µm system. Since the measurements were performed employing the same optical system and detector, this section supplements the results reported previously [27], verifies the theoretical evaluations, and indicates the potentials of a dual-band OCT detection. The limiting factor for NIR OCT is scattering. However, we suppose that the penetration depth of the NIR system could be slightly enhanced applying the detector array with a higher number of pixels (sensitivity roll-off optimization), or if the first diffraction order is used to record the interferogram.
3.2 Performance of 2 µm OCT for inspection of the painting mock-up
High axial resolution (in comparison to MIR OCT) enabled by the 2 µm OCT as well as a reasonable penetration (in contrast to commercially available OCT systems) allowed us to introduce a prospect cost-efficient solution for the investigation of the artificial painting test samples. Figure 9 depicts the typical b-scan of the mock-up obtained by the experimental NIR OCT, the distribution of the constituted paints is indicated.
Fig. 9. Near-infrared (spectral region around 2 µm) OCT b-scan of the painting mock-up; top-view photo roughly indicates the scanning position.
The following measurements were carried out in the normal incidence regime at the modulation frequency of 40 Hz. In order to avoid possible damage of the sample, both band- (500 nm bandwith, 2 µm CWL) and edge-pass (1.65 µm cut-on wavelength) filters were applied.
The analysis of the obtained scan shows that the varying grain size and structure of the pigments cause different bulk scattering, which, together with a speckle pattern and signal decay, makes the types of the paints clearly distinguishable. The penetration depth of the system is dependent on the material properties. The interface of the substrate wafer is detected on the left and weakly on the right side of the scan, however, it is invisible under the layer of gouache. This could be caused by the irregular thickness of the sample, intense scattering, or by the absorption that is inherited from the residual water content. In practice, since OCT systems that are operating in the IR region would suffer from the water due to the the high absorption, the sample preparation is crucial in this wavelength range.
OCT measurements of the complex and irregular oil-paintings mock-up are depicted in Fig. 10 as a volume scan (not in 1:1 aspect ratio) and a cross-section, the dimensions of the measured area are 8$\times$8 mm2. The top view photo of the part of the sample is shown in the inset in the figure. Due to the contrast that is induced by different scattering, the prospect NIR OCT system allows us to differentiate between pure colours (cadmium red and yellow green) that were deposited as ground layers and a brush-style mixture of the paints. The apparent transparency of the normally highly scattering titanium white was investigated. The invasive mechanical removal of the layer demonstrated that varnish was embedded underneath the thin paint layer during the drying.
Fig. 10. C-scan of the complex varnished oil-paintings mock-up (with the varnish inclusions under the thin titanium white layer); the position of the cross-section is displayed in the inset.
Figure 10 confirms that non-invasive inspection of artworks and detection of the sub-interfaces and structures (e.g. underlying varnish under the thin and highly scattering layer of titanium white) are effectively achievable by means of the proposed NIR OCT system. The total measurement time at an optimal (with respect to the surface reflectivity) modulation frequency (40 Hz) was about 5.5 hours, which is not crucial for non-biological samples.
3.3 Dual-band IR OCT for inspection of lithography-based manufactured ceramics
The diverse measurements of the LCM sintered alumina sample were performed. A typical dual-band output of the system presented as a multi-band compounded b-scan, c-scan and en-face image is shown in Fig. 11. It should be noted, the thickness of the sample was designed to be in the range, where a drop-off in the sensitivity for the NIR and MIR OCT systems is not a prevailing effect on the penetration depth. Therefore, the comparison of the OCT within different sub-bands could be effectively performed for these types of samples.
Fig. 11. Performance and benefits of the dual-band IR OCT for investigation of the additively manufactured high-performance ceramics, positions in the images are linked by numbering, pixelated structure caused by the limited resolution of the 3D printer is detected; the refractive index of the ceramic sample is estimated by fitting a triangle to the b-scan (OPD effect is exploited), a ratio between the slopes results in the refractive index of 1.80; the insert in (a) displays a light microscopic image of the facet, the dimensions of the structures are measured and denoted (included for reference)
The sample was fixed on paper and a glass slide, the underlying structure is visualized on the right side of the b-scan [Fig. 11(a)] using MIR OCT. A chopping frequency of 40 Hz was set for the measurements; the same post-processing approach was utilized for both sub-systems.
The scans yield that the LCM sintered ceramic is a highly scattering microstructured material. Due to the rough surface and high porosity, the monolith is considerably more turbid (high surface and bulk scattering) than the benchmark sample investigated in section 3.1. For this reason, the sample was positioned normally with respect to the probing beam during the measurements, the axial position was fixed. Despite the lower signal, the geometry of the sample is completely accessed by the 4 µm OCT system. The en-face image of the bottom interface of the channels is shown in Fig. 11(c). The high-resolution NIR OCT measurements allowed to clearly see only the second sub-interface, meanwhile, providing a better compromise for the close-to-surface details.
The design of the sample is an ideal example that demonstrates a possible misleading of OCT imaging. Since OCT systems measure the optical path difference (OPD) and not the exact geometry, the flat rear interface of the ceramic is disturbed. Taking into account the refractive index of the ceramics, this effect results in an almost symmetric shape, which is, however, not real. Meanwhile, the formed angle [Fig. 11(a)] can be used to estimate the refractive index of the material, a rough assessment of the slopes results in $n \approx 1.80$.
As shown in Fig. 11(a,b), the layered structure of the sample is detectable, since the axial resolution of the system is in the same range as the resolution of the 3D printer. Hence, the dual-band 2 µm and 4 µm OCT could be successfully applied for the quality control of the manufacturing process: to track possible layer displacements, melting, thickness variation, inclusions etc. An at-line non-invasive inspection of the manufactured parts is feasible, while a possible integration into existing systems could potentially deliver a fully-integrated system with closed-loop feedback on the printing process.
4. Conclusion and outlook
Exploiting the distinctive properties of supercontinua generated in ZBLAN fibers, we have developed and introduced a dual-band near- and mid-infrared (NIR and MIR, 2 µm and 4 µm central wavelengths correspondingly) Fourier-domain OCT system that is based on a single supercontinuum source and a single inexpensive pyroelectric array. Operating in both spectral windows, the OCT system, improved and optimized since the last report [27], opens up new possibilities and enables enhanced penetration depth into strongly scattering materials while maintaining a high spatial resolution when operating in the NIR regime. Considering the dual-band sensing (six-dimensional data as an output) reported in our work, we have effectively utilized a significant part of the emission spectra of the employed supercontinuum source.
The design of the prospective system is technically original. We consider in detail the integration of the pyroelectric linear array into OCT interferometers and discuss unique aspects of the OCT-adapted signal acquisition scheme. We examine and present a sensitivity roll-off for both sub-systems, features of sensitivity steering, and report an advanced and uncommon approach to the detection. In the experimental part, by the direct comparison of the OCT system operating in both NIR and MIR range, well-known [26] predictions have been confirmed.
The gained results demonstrate the capabilities and maturity of the system for art inspection and study of scattering ceramics. Artificial painting mock-ups have been investigated using the developed and affordable 2 µm NIR OCT sub-system, internal sub-interfaces and inhomogeneities that appeared during the drying were successfully detected.
Being an active field, IR imaging is currently of great interest. Therefore, the quality and availability of optics and detectors are expected to be improved, while the price of components is decreasing. Besides enhancing the sensitivity of focal plane arrays, we predict a tenable increase in the number of pixels. Therefore, as a far-reaching goal, developments in the field of IR detectors should lead state-of-the-art techniques to so-called spectroscopic OCT that operates in several sub-bands providing depth-resolved spectral information. Operating in the IR range, such systems would certainly be a further game-changing factor in non-destructive testing.
Due to the economic efficiency of the presented system, we expect to attract attention in the art preserving community and in industry, especially in additive ceramic manufacturing, where in- and on-line process control and monitoring could be a promising application scenario.
Funding
Horizon 2020 Framework Programme (722380); Strategic economic- and research program of the province of Upper Austria (Innovative Upper Austria 2020); EFRE Urban Innovative Actions (IWB2020).
Acknowledgments
The authors thank Rainhard Köhler and Uwe Hoffmann from DIAS Infrared GmbH for the fruitful and detailed discussions about the pyroelectric linear arrays; Etienne Genier from NKT Photonics for discussions on supercontinuum generation in ZBLAN fibers; Rong Su for providing the samples and his valuable comments on our work. We are also immensely grateful to Martin Schwentenwein and Dominik Brouczek from Lithoz GmbH for valuable information on additive ceramic manufacturing systems and for producing the test samples according to our design.
Disclosures
The authors declare no conflicts of interest.
References
1. L. Shaw, V. Nguyen, J. Sanghera, I. Aggarwal, P. Thielen, and F. Kung, “Ir supercontinuum generation in as-se photonic crystal fiber,” in Advanced Solid-State Photonics (TOPS), (Optical Society of America, 2005), p. 864.
2. C. Xia, M. Kumar, O. P. Kulkarni, M. N. Islam, J. Fred L. Terry, M. J. Freeman, M. Poulain, and G. Mazé, “Mid-infrared supercontinuum generation to 4.5 µm in zblan fluoride fibers by nanosecond diode pumping,” Opt. Lett. 31(17), 2553–2555 (2006). [CrossRef]
3. P. Domachuk, N. A. Wolchover, M. Cronin-Golomb, A. Wang, A. K. George, C. Cordeiro, J. Knight, and F. G. Omenetto, “Over 4000 nm bandwidth of mid-ir supercontinuum generation in sub-centimeter segments of highly nonlinear tellurite pcfs,” Opt. Express 16(10), 7161–7168 (2008). [CrossRef]
4. P. M. Moselund, C. Petersen, S. Dupont, C. Agger, O. Bang, and S. R. Keiding, “Supercontinuum: broad as a lamp, bright as a laser, now in the mid-infrared,” in Laser Technology for Defense and Security VIII, vol. 8381M. Dubinskii and S. G. Post, eds., International Society for Optics and Photonics (SPIE, 2012), pp. 265–270.
5. C. R. Petersen, P. M. Moselund, L. Huot, L. Hooper, and O. Bang, “Towards a table-top synchrotron based on supercontinuum generation,” Infrared Phys. Technol. 91, 182–186 (2018). [CrossRef]
6. S. Dai, Y. Wang, X. Peng, P. Zhang, X. Wang, and Y. Xu, “A review of mid-infrared supercontinuum generation in chalcogenide glass fibers,” Appl. Sci. 8(5), 707 (2018). [CrossRef]
7. A. M. Heidt, J. H. V. Price, C. Baskiotis, J. S. Feehan, Z. Li, S. U. Alam, and D. J. Richardson, “Mid-infrared zblan fiber supercontinuum source using picosecond diode-pumping at 2 µm,” Opt. Express 21(20), 24281–24287 (2013). [CrossRef]
8. P. M. Moselund, C. Petersen, L. Leick, J. S. Dam, P. Tidemand-Lichtenberg, and C. Pedersen, “Highly stable, all-fiber, high power zblan supercontinuum source reaching 4.75 µm used for nanosecond mid-ir spectroscopy,” in Advanced Solid-State Lasers Congress, (Optical Society of America, 2013), p. JTh5A.9.
9. J. Swiderski and M. Michalska, “High-power supercontinuum generation in a zblan fiber with very efficient power distribution toward the mid-infrared,” Opt. Lett. 39(4), 910–913 (2014). [CrossRef]
10. W. Yang, B. Zhang, G. Xue, K. Yin, and J. Hou, “Thirteen watt all-fiber mid-infrared supercontinuum generation in a single mode zblan fiber pumped by a 2µm mopa system,” Opt. Lett. 39(7), 1849–1852 (2014). [CrossRef]
11. A. N. Ghosh, M. Meneghetti, C. R. Petersen, O. Bang, L. Brilland, S. Venck, J. Troles, J. M. Dudley, and T. Sylvestre, “Chalcogenide-glass polarization-maintaining photonic crystal fiber for mid-infrared supercontinuum generation,” JPhys Photonics 1(4), 044003 (2019). [CrossRef]
12. K. Ke, C. Xia, M. N. Islam, M. J. Welsh, and M. J. Freeman, “Mid-infrared absorption spectroscopy and differential damage in vitro between lipids and proteins by an all-fiber-integrated supercontinuum laser,” Opt. Express 17(15), 12627–12640 (2009). [CrossRef]
13. M. Kumar, M. N. Islam, F. L. Terry, M. J. Freeman, A. Chan, M. Neelakandan, and T. Manzur, “Stand-off detection of solid targets with diffuse reflection spectroscopy using a high-power mid-infrared supercontinuum source,” Appl. Opt. 51(15), 2794–2807 (2012). [CrossRef]
14. S. Dupont, C. Petersen, J. Thøgersen, C. Agger, O. Bang, and S. R. Keiding, “Ir microscopy utilizing intense supercontinuum light source,” Opt. Express 20(5), 4887–4892 (2012). [CrossRef]
15. C. R. Petersen, N. Prtljaga, M. Farries, J. Ward, B. Napier, G. R. Lloyd, J. Nallala, N. Stone, and O. Bang, “Mid-infrared multispectral tissue imaging using a chalcogenide fiber supercontinuum source,” Opt. Lett. 43(5), 999–1002 (2018). [CrossRef]
16. J. Kilgus, K. Duswald, G. Langer, and M. Brandstetter, “Mid-infrared standoff spectroscopy using a supercontinuum laser with compact fabry-pérot filter spectrometers,” Appl. Spectrosc. 72(4), 634–642 (2018). [CrossRef]
17. J. Kilgus, G. Langer, K. Duswald, R. Zimmerleiter, I. Zorin, T. Berer, and M. Brandstetter, “Diffraction limited mid-infrared reflectance microspectroscopy with a supercontinuum laser,” Opt. Express 26(23), 30644–30654 (2018). [CrossRef]
18. A. Saleh, A. Aalto, P. Ryczkowski, G. Genty, and J. Toivonen, “Short-range supercontinuum-based lidar for temperature profiling,” Opt. Lett. 44(17), 4223–4226 (2019). [CrossRef]
19. M. K. Dasa, G. Nteroli, P. Bowen, G. Messa, Y. Feng, C. R. Petersen, S. Koutsikou, M. Bondu, P. M. Moselund, A. Podoleanu, A. Bradu, C. Markos, and O. Bang, “All-fibre supercontinuum laser for in vivo multispectral photoacoustic microscopy of lipids in the extended near-infrared region,” Photoacoustics 18, 100163 (2020). [CrossRef]
20. K. Liu, J. Liu, H. Shi, F. Tan, and P. Wang, “High power mid-infrared supercontinuum generation in a single-mode zblan fiber with up to 21.8 w average output power,” Opt. Express 22(20), 24384–24391 (2014). [CrossRef]
21. Z. Zhao, B. Wu, X. Wang, Z. Pan, Z. Liu, P. Zhang, X. Shen, Q. Nie, S. Dai, and R. Wang, “Mid-infrared supercontinuum covering 2.0-16 µm in a low-loss telluride single-mode fiber,” Laser Photonics Rev. 11(2), 1700005 (2017). [CrossRef]
22. R. A. Martinez, G. Plant, K. Guo, B. Janiszewski, M. J. Freeman, R. L. Maynard, M. N. Islam, F. L. Terry, O. Alvarez, F. Chenard, R. Bedford, R. Gibson, and A. I. Ifarraguerri, “Mid-infrared supercontinuum generation from 1.6 to >11µm using concatenated step-index fluoride and chalcogenide fibers,” Opt. Lett. 43(2), 296–299 (2018). [CrossRef]
23. A. M. Zysk, F. T. Nguyen, A. L. Oldenburg, D. L. Marks, and S. A. Boppart, “Optical coherence tomography: a review of clinical development from bench to bedside,” J. Biomed. Opt. 12(5), 051403 (2007). [CrossRef]
24. N. D. Shemonski, F. A. South, Y.-Z. Liu, S. G. Adie, S. P. Carney, and S. A. Boppart, “Computational high-resolution optical imaging of the living human retina,” Nat. Photonics 9(7), 440–443 (2015). [CrossRef]
25. D. Stifter, “Beyond biomedicine: a review of alternative applications and developments for optical coherence tomography,” Appl. Phys. B 88(3), 337–357 (2007). [CrossRef]
26. R. Su, M. Kirillin, E. W. Chang, E. Sergeeva, S. H. Yun, and L. Mattsson, “Perspectives of mid-infrared optical coherence tomography for inspection and micrometrology of industrial ceramics,” Opt. Express 22(13), 15804–15819 (2014). [CrossRef]
27. I. Zorin, R. Su, A. Prylepa, J. Kilgus, M. Brandstetter, and B. Heise, “Mid-infrared fourier-domain optical coherence tomography with a pyroelectric linear array,” Opt. Express 26(25), 33428–33439 (2018). [CrossRef]
28. N. M. Israelsen, C. R. Petersen, A. Barh, D. Jain, M. Jensen, G. Hannesschläger, P. Tidemand-Lichtenberg, C. Pedersen, A. Podoleanu, and O. Bang, “Real-time high-resolution mid-infrared optical coherence tomography,” Light: Sci. Appl. 8(1), 11–7538 (2019). [CrossRef]
29. A. Vanselow, P. Kaufmann, I. Zorin, B. Heise, H. Chrzanowski, and S. Ramelow, “Mid-infrared frequency-domain optical coherence tomography with undetected photons,” in Quantum Information and Measurement (QIM) V: Quantum Technologies, (Optical Society of America, 2019), p. T5A.86.
30. U. Scheithauer, E. Schwarzer, T. Moritz, and A. Michaelis, “Additive manufacturing of ceramic heat exchanger: Opportunities and limits of the lithography-based ceramic manufacturing (lcm),” J. Mater. Eng. Perform. 27(1), 14–20 (2018). [CrossRef]
31. A. Zocca, P. Colombo, C. M. Gomes, and J. Günster, “Additive manufacturing of ceramics: Issues, potentialities, and opportunities,” J. Am. Ceram. Soc. 98(7), 1983–2001 (2015). [CrossRef]
32. I. Zorin, J. Kilgus, R. Su, B. Lendl, M. Brandstetter, and B. Heise, “Multimodal mid-infrared optical coherence tomography and spectroscopy for non-destructive testing and art diagnosis,” in Optics for Arts, Architecture, and Archaeology VII, vol. 11058H. Liang, R. Groves, and P. Targowski, eds., International Society for Optics and Photonics (SPIE, 2019), pp. 74–88.
33. C. S. Cheung, J. M. O. Daniel, M. Tokurakawa, W. A. Clarkson, and H. Liang, “High resolution fourier domain optical coherence tomography in the 2 µm wavelength range using a broadband supercontinuum source,” Opt. Express 23(3), 1992–2001 (2015). [CrossRef]
34. W. Drexler and J. G. Fujimoto, Optical Coherence Tomography, Technology and Applications (Springer International Publishing, 2008).
35. S. W. Lee, H. W. Jeong, B. M. Kim, Y. C. Ahn, W. Jung, and Z. Chen, “Optimization for axial resolution, depth range, and sensitivity of spectral domain optical coherence tomography at 1.3 um,” J. Korean Phys. Soc. 55(6), 2354–2360 (2009). [CrossRef]
36. M. Jensen, I. B. Gonzalo, R. D. Engelsholm, M. Maria, N. M. Israelsen, A. Podoleanu, and O. Bang, “Noise of supercontinuum sources in spectral domain optical coherence tomography,” J. Opt. Soc. Am. B 36(2), A154–A160 (2019). [CrossRef]
37. C. Agger, C. Petersen, S. Dupont, H. Steffensen, J. K. Lyngsø, C. L. Thomsen, J. Thøgersen, S. R. Keiding, and O. Bang, “Supercontinuum generation in zblan fibers—detailed comparison between measurement and simulation,” J. Opt. Soc. Am. B 29(4), 635–645 (2012). [CrossRef]
38. G. Genty, S. Coen, and J. M. Dudley, “Fiber supercontinuum sources (invited),” J. Opt. Soc. Am. B 24(8), 1771–1785 (2007). [CrossRef]
39. I. Zorin, J. Kilgus, K. Duswald, B. Lendl, B. Heise, and M. Brandstetter, “Sensitivity-enhanced fourier transform mid-infrared spectroscopy using a supercontinuum laser source,” Appl. Spectrosc, doc. ID ASP-893364 (posted 17 November 2019, in press).
40. H. H. Li, “Refractive index of alkaline earth halides and its wavelength and temperature derivatives,” J. Phys. Chem. Ref. Data 9(1), 161–290 (1980). [CrossRef]
41. I. H. Malitson, “A redetermination of some optical properties of calcium fluoride,” Appl. Opt. 2(11), 1103–1107 (1963). [CrossRef]
42. A. Rogalski, Infrared Detectors 2nd Edition (CRC, 2011).
43. R. Köhler, D. Wassilew, V. Norkus, M. Schossig, and G. Hofmann, “Enhanced pyroelectric linear arrays for infrared spectroscopy,” in Proceedings IRS2 2017, IR Sensors and Arrays, vol. I1 (2017).
44. D. E. Marshall, “A review of pyroelectric detector technology,” in Utilization of Infrared Detectors, vol. 0132 (1978).
45. A. Hossain and M. H. Rashid, “Pyroelectric detectors and their applications,” IEEE Trans. Ind. Appl. 27(5), 824–829 (1991). [CrossRef]
46. A. Agrawal, T. J. Pfefer, P. D. Woolliams, P. H. Tomlins, and G. Nehmetallah, “Methods to assess sensitivity of optical coherence tomography systems,” Biomed. Opt. Express 8(2), 902–917 (2017). [CrossRef]
47. S. H. Yun, G. J. Tearney, B. E. Bouma, B. H. Park, and J. F. de Boer, “High-speed spectral-domain optical coherence tomography at 1.3 µm wavelength,” Opt. Express 11(26), 3598–3604 (2003). [CrossRef]
48. H. Liang, R. Lange, B. Peric, and M. Spring, “Optimum spectral window for imaging of art with optical coherence tomography,” Appl. Phys. B 111(4), 589–602 (2013). [CrossRef]
49. C. S. Cheung, J. M. O. Daniel, M. Tokurakawa, W. A. Clarkson, and H. Liang, “Optical coherence tomography in the 2 µm wavelength regime for paint and other high opacity materials,” Opt. Lett. 39(22), 6509–6512 (2014). [CrossRef]
50. M. Schwentenwein and J. Homa, “Additive manufacturing of dense alumina ceramics,” Int. J. Appl. Ceram. Technol. 12(1), 1–7 (2015). [CrossRef]
