1. Introduction

Metal corrosion occurs when environmental factors such as humidity, temperature, and pH lead to the deterioration, reduced lifespan, and functional damage of metals [1,2]. This is primarily because of the reactivity and ionizing properties of metals, which tend to return to thermodynamically stable states through corrosion [1]. The economic costs associated with corrosion are substantial and include facility shutdowns, maintenance expenses, and additional equipment for corrosion protection [1,2]. Therefore, research on corrosion-monitoring techniques is crucial to mitigate these losses.

Various methods have been developed to monitor corrosion, including coupon tests, electrochemical techniques, electrical resistance methods, and non-destructive tests [3–11]. Coupon tests involve using experimental specimens placed in the same environment and analyzing them later; however, their reliability depends on factors such as location and recovery time [3]. Electrochemical techniques provide detailed information but require expertise for data analysis, and the estimated corrosion rates may not always match the actual rates [4,5]. Electrical resistance methods can measure corrosion but may complicate the detection of local corrosion [6,7]. Nondestructive tests, such as ultrasonic waves, eddy currents, sound emission, and infrared thermal imaging, can detect corrosion without causing damage but have limitations in terms of sensitivity and response speed [8–11].

Optical coherence tomography (OCT) is a noninvasive technique that enables the acquisition of high-resolution three-dimensional images with micron-level accuracy [12]. Although primarily used in medical imaging [12,13], OCT has also been applied in previous studies to measure anti-corrosion coatings or paint layers using a broadband superluminescent light-emitting diode or swept laser with a 1300 nm wavelength range [14–16]. In these studies on metal corrosion using OCT, depth profiles generated by corrosion were observed; however, the analysis of the oxide layer caused by corrosion has not been discussed in detail because of its complexity. The depth profile is deeper than the actual thickness of the oxide layer. In [17], it was confirmed that this was due to multiple scatterings of the incident light caused by effects similar to those of a thin oxide layer formed on a metal surface. To mitigate the effects of multiple scattering from the oxide layer, we used optical coherence microscopy (OCM) with a high numerical aperture (NA) optical system. When light with a large NA is incident on the oxide layer, the light that returns owing to multiple reflections is not reacquired through the optical system, allowing for a clearer distinction of the oxide layer.

Recently, a longer-wavelength light source (approximately 1700nm) has been shown to reduce light scattering in strong turbid media and enhance OCT image contrast [18]. It is well known that the transmittance characteristics of copper oxide under the tests in this study are higher at longer wavelengths [19]. In our study, we propose a novel nondestructive corrosion measurement method using OCM for enhanced lateral resolution and utilize a longer near-infrared wavelength range (1700nm) swept laser source for the optical imaging system to accurately measure the thickness of a copper oxide layer. This approach reduces light scattering and enhances the dynamic range of OCM signals, especially when dealing with turbid media such as metallic oxide layers, and the OCM schematic was selected for metal corrosion analysis to capitalize on its capability to acquire surface depth images of oxidized specimens at a higher resolution.

The proposed method offers improved sensitivity and faster response times, enabling accurate assessment of local corrosion within the optical measurement area.

2. Methods

2.1 Optical properties of copper oxide

An optical method was employed to measure the thickness of the copper oxide layer resulting from corrosion. Typically, metals reflect most of the incident light on their surfaces. However, metallic oxides exhibit only partial light absorption and transmission. Metallic oxides, such as Cu₂O and CuO, are commonly formed during the corrosion process, with Cu₂O being particularly abundant. The refractive indices of Cu₂O and CuO are approximately 2.7 at longer wavelengths [17]. Cu₂O exhibits high transmittance at long wavelengths [19]. This property has been exploited in solar heat systems where Cu₂O serves as an effective absorber of sunlight because of its transmission characteristics [17].

Figure 1 illustrates the generation mechanism of Cu₂O through corrosion. Karlsson [17] utilized a metallic oxide layer to enhance solar heat absorption. The artificial oxide layer acts as a thin-film coating, reducing the reflectivity of the metal surface and improving solar heat absorption. The film exploits offset interference based on the wavelength of light reflected from the oxide layer surface after forming Cu₂O on the Cu surface, as well as the light reflected from the copper interface after passing through the oxide layer. This suggests the feasibility of measuring and analyzing the metallic oxide layer using an optical measurement method. Consequently, the thickness of the metallic oxide can be determined by analyzing the light that scatters and returns inside the specimen using OCT.

 

Fig. 1. Generation mechanism of copper oxide (I).

Download Full Size | PDF

2.2 Fabrication and composition analysis of corrosion specimens

To produce thin copper oxide layers as corroded metal samples for testing, the copper specimens were subjected to high-temperature heating in a vacuum environment with an extremely low oxygen concentration. Initially, square copper specimens with dimensions of 30 mm × 30 mm were prepared. The surfaces of the specimens were precisely polished using SiC abrasive papers of varying roughness (up to 4000 grit) to ensure the formation of uniform oxide layers. Further polishing was performed using diamond suspension, achieving a surface finish of up to 1 µm.

The formation of oxide layers was accomplished by heating the polished specimens in an alumina tube furnace at a temperature of 900 °C. The furnace (DVF1600, DaeHeung Science Inc.) was maintained at a base pressure of approximately 1.33 × 10⁻⁴Pa throughout the process. The oxidation times were varied from 1 to 8 h to examine the impact of the heat treatment duration after reaching the target temperature. The vacuum heat treatment followed a controlled heating rate of 5 °C per minute until the specimens reached 900 °C. Subsequently, the specimens were adequately cooled in the equipment before being removed for further analysis.

The compositions of the copper oxidation specimens were analyzed using X-Ray Diffraction (XRD; XPERT PRO, PANalytical B.V.). The position and relative intensity of the peaks appearing in the XRD graph are unique characteristics determined by the crystal structure and size of the sample being analyzed. For Cu₂O, peaks appear at 30, 35, and 60 degrees in the XRD graph. The compositional analysis data obtained through XRD are presented in Fig. 2, indicating the presence of peaks corresponding to Cu₂O and Cu in all three specimens

 

Fig. 2. XRD composition analysis data for copper oxidation specimens (sample images in inset) manufactured by vacuum heat treatment at 900°C for 1, 2, 4, and 8 h, respectively.

Download Full Size | PDF

2.3 1700nm OCM for metal corrosion measurement

The optics for the measurement system was an optimized setup, depicted in Fig. 3(a), which was used to analyze the prepared copper oxidation specimens discussed in Section 2.2. The light source employed was a swept source (HSL-1.7-90-B, Santec Inc.) with a wavelength range of 1600–1790nm and sweeping rate of 90 kHz. The optical spectrum of the swept source is obtained from a previous study [20]. The interference signal was acquired using a high-speed digitizer with a sampling rate of 1.8 GS/s and a data length of 4096 (ATS9360, AlazarTech Inc.).

 

Fig. 3. (a) Schematic diagram of 1700nm swept-source OCM system, (b) comparison of the measurement area between SEM and OCM in a sample. Images and schematic diagram of the sample arms of (c) OCM and (d) OCT system. (e) Specifications such as numerical aperture (NA), axial resolution, lateral resolution, and DOF of both OCM and OCT.

Download Full Size | PDF

All systems were constructed using an optical fiber device, allowing the interchange of the sample and reference arms to enable selective measurement using a conventional doublet lens and an objective lens. To address the phase variation of the swept source with a changing wavelength [21], an additional auxiliary interferometer for phase correction, as shown in Fig. 3(a), was designed and integrated into the setup. Figure 3(b) shows an oxide sample with the size of 30 × 30 mm. The scanned area of the OCM was 4 mm × 4 mm as shown in the dotted box in Fig. 3(b). For comparison with OCM measurement in the Section 3.2, SEM provides a single-line scan for oxide layers up to 1.3 mm, whereas OCM offers oxide layer information for 1,000 lines in a 4 mm × 4 mm area. This demonstrates the advantage of the OCM in providing a larger amount of data for analysis.

Figure 3(c) and (d) show OCM operates on the same principle as OCT but utilizes a different objective lens (OL), a tube lens at the sample arm, and additional free space on the reference arm to compensate for the optical path length difference [22–25]. The back focal plane of the objective lens (PAN-5-NIR, Sigmakoki Inc.) was placed on the mounting thread, hindering the definition of the laser beam scanning design without additional tubes and scan lenses. The the focal lengths of the tube lens (AC508-075-C-ML, Thorlabs Inc.) and scan lens (AC254-045-C-ML, Thorlabs Inc.) were 75 and 45 mm, respectively. By employing an objective lens with a higher NA value, it is possible to enhance the lateral resolution of OCM images. However, the drawback of using a high NA value is that it results in a shallow depth of focus (DOF).

In this system, the axial resolution was determined experimentally by capturing an image of a glass slide with a known thickness of 1 mm and a refractive index of 1.5 at 1700nm. Based on the OCT images, when the distance between the top and bottom surfaces of the glass slide was measured in pixels, the pixel resolution was approximately 8.3 µm/pixel, considering the refractive index and slide glass thickness.

Conversely, the lateral resolution corresponded to the horizontal resolution. This is determined by the objective lens of the sample stage in the optical imaging systems. The lateral resolution ($\delta x$) was calculated as follows [12]:

The DOF of the optical imaging system depend on the NA value of the objective lens. The DOF can be calculated using Eq.2:

Based on the characteristics of the developed OCM system, it was determined that the measurable depth range for metallic oxidation analysis is approximately 8 to 170 µm considering the pixel resolution of 8.3 µm and the DOF of 169 µm. This means that our optical system allows for the nondestructive and fast monitoring of the early corrosion process. Using our OCM system, it is possible to observe and analyze the initial stages of corrosion with high precision without damaging the samples.

3. Results

3.1 Thickness measurement of the copper oxide layers using OCM

The OCM images of the copper oxidation specimens were compared with the OCT images in Fig. 4(a)-(d). Phase calibration techniques have been applied to enhance image quality and reduce noise [21]. Figure 4(c) provides deeper tomographic images compared to OCM, benefiting from its larger DOF of 806 µm. In optical tomogram images, the reflected light signal may appear to pass through a copper layer. However, this effect was caused by multiple scattered light beams occurring within the copper oxide layer. In Fig. 4(b) and (d), the A-line of two tomogram images in Fig. 4(a) and (c), respectively, for a copper specimen after 1 h can distinguish the boundaries between the air and oxidation layers and the metal boundary of the copper oxidation layer. Owing to the deeper DOF of the OCT, multiple scatterings were detected, resulting in the appearance of a signal passing through the copper oxide layer, as shown in Fig. 4(d). However, in the OCM image, the effect of multiple scattered lights was reduced, enabling a clear distinction of the copper oxide layer, as shown in Fig. 4(b). Even with OCM, the NA of the objective lens was not high enough to eliminate the effects of multiple scattering or to prevent back-coupling of multiple scattering signals. As seen in the A-line graph in Fig. 4(b), after the two main peaks, smaller peaks were observed, which is believed to be a residual effect of multiple scattering. However, when compared to the A-line graph of OCT, it is evident that the multiple scattering signals have been reduced and suppressed. The analysis of the metallic oxides validated the effectiveness of the OCM technique in this study.

 

Fig. 4. Tomographic B-scan images of copper oxidation specimens measured by (a) OCM and (b) OCT. A-line OCM (c) and OCT (d) signals in depth direction of a copper specimen., (e) Oxidation thickness histograms obtained from OCM images in Fig. 4(a).

Download Full Size | PDF

In the OCM image, it is possible to distinguish the boundary between “air” and “ oxidation layer, “ as well as the boundary between “oxidation layer” and “copper layer”, sequentially. The A-line OCM signal clearly indicates the thickness of the copper oxidation layer. The red arrows on the A-line graphs indicate two points: the left arrow represents the reflection position on the sample surface and the right arrow represents the reflection at the interface between the oxide layer and copper.

The distance between the two peaks in the A-line graph corresponds to the thickness of the oxidation layer formed on the metal surface. By considering the axial resolution of OCM as 8.3 µm and the group refractive index of Cu₂O as 2.7 [17], the thickness profile of the oxidation layer can be determined from the tomographic OCM images. In the obtained A-line scan data, 10 addition data points were added through the interpolation algorithm to find a more accurate peak. Considering the pixel resolution of OCM as 8.3 µm and the group refractive index of Cu₂O as 2.7, the effective pixel resolution was calculated as 3.1 µm/pixel in the A-line graph. This enhanced resolution enables accurate determination of the thickness of the oxide layers in the A-line graph.

Figure 4(e) shows normalized histograms of the OCM thickness distributions of the specimens. Thickness histograms were divided by the overall count of data points. Because of this normalization, the area under the histogram curve, computed as an integral, sums to one. As the oxidation time increased, the thickness of the oxide layer increased, whereas its uniformity decreased. This indicates that, with longer oxidation times, the oxidation layer becomes thicker and exhibits more variation in thickness across the surface. Due to the optical depth resolution being limited to 8.3 µm, it becomes difficult to distinguish between the surface reflection at the boundary of air and the oxide layer and the reflection at the boundary between the oxide layer and copper for oxide layer thicknesses less than 8 µm. The values below 8 µm in all graphs are believed to be errors from the peak finding processing. Therefore, when using the OCM technique to measure the thickness of oxide layers, it is crucial to consider this limitation and focus on analyzing the oxide layers above the resolution threshold to ensure more accurate and reliable results.

3.2 Thickness measurement of the copper oxide layers using Scanning Electron Microscope

Figure 5 shows the scanning electron microscope (SEM) images and composition analysis results of the copper oxidation specimens after heat treatment for 1, 2, 4, and 8 h, as determined by scanning electron microscope-energy dispersive (SEM-EDS) X-ray spectrometry (SU5000, Schottky Inc.). The atomic percentage ratio of copper to oxygen was approximately 7:3, again indicating that the predominant copper oxide in all specimens was Cu₂O.To obtain cross-sectional images of the Cu₂O specimens, the samples were precisely cut using a diamond cutter and then thoroughly polished. The thickness profiles of the oxide layers were determined from these SEM images of the Cu₂O specimens. For comparison with the measurements obtained using OCM, SEM images were acquired near the center of the OCM images of the specimens.

 

Fig. 5. SEM-EDS composition analysis data for copper oxidation specimens for 1, 2, 4, and 8 h.

Download Full Size | PDF

To validate the reliability of the oxide layer thickness measurement technique, oxide specimens were compared. The thickness of the oxide layer was measured by capturing several SEM images similar to those in Fig. 5 while moving the image capturing frame sideways at a magnification of 2000X. Approximately 25 images were captured for each specimen and stitched together to create a panoramic image using PhotoShop.

Figures 6(a) show the SEM panoramic image and extracted oxide layer image of copper oxidized for 1 h, respectively. Through the SEM-EDS compositional analysis process, the oxide layer was identified and then the oxide layer was extracted from the obtained SEM images using a graphic tool, and the oxide layer thickness profile and histogram were obtained using MATLAB. An example of the obtained thickness profile is shown in Fig. 6(b). The thickness of the copper oxide layer ranged from 4 to 10 µm. By comparing the results obtained from the OCM techniques with the measurements from the SEM images, the reliability of the oxide layer thickness measurement technology was verified.

 

Fig. 6. (a) Extracted oxide layer image from SEM panoramic image and (b) thickness profile of the oxide layer, (c) SEM panorama images and extracted oxide layers of the copper oxidation specimens for 1, 2, 4, and 8 h., (d) oxidation thickness histograms obtained from SEM images

Download Full Size | PDF

Figure 6(c) presents the SEM panoramic images of the specimens shown in Fig. 2 with a scanning range of up to 1.3 mm. A uniform oxide layer was observed when the oxidation time was short, indicating a consistent thickness across the surface. However, as the oxidation time increased, the thickness of the oxide layer became nonuniform. This can be observed in the variations in color and intensity across the oxide layer in the SEM images. The non-uniform thickness of the oxide layer indicates the presence of localized variations in the oxidation process, possibly owing to factors such as surface conditions or reaction kinetics.

The histograms in Fig. 6(d) depict the thicknesses of the oxidation layers extracted from the SEM images in Fig. 6(c). A noticeable difference in the thickness and uniformity of the oxidation layers is observed when comparing the histograms of the samples after 1 and 8 h of oxidation. As the oxidation time increases, the thickness distribution of the oxidation layer becomes more pronounced.

Typically, the thickness of the oxide layer is significantly affected by the concentration of oxygen around the metal during oxidation. Even in the presence of ample oxygen, the already formed oxide layer can act as a barrier to further oxidation, inhibiting the process even under oxidative conditions. In this study, we oxidized copper in a low-oxygen environment, resulting in the formation of a copper oxide layer of approximately 9 µm after one and two hours of oxidation, as shown in Fig. 6(d). We also observed that the variation in the oxide layer thickness increased with the duration of oxidation. Even after four and eight hours of oxidation, most of the oxide layer formed is over 10 µm, resulting in a more uneven oxide layer.

4. Analysis and discussions

Owing to the challenge of precisely matching the measurement positions between OCM and SEM, histograms were generated from the thickness profiles obtained using both methods and compared. The oxide layer formation can vary depending on the surface conditions of the metal sample, even under identical oxidation conditions, and different locations on the sample surface can exhibit different oxide layers. Therefore, the histograms in Fig. 4(e) and Fig. 6(d), which represent the oxide layer thickness distributions for each oxidation time, do not exactly match. The histogram obtained from SEM exhibits a high resolution but can measure the oxide layer length up to a maximum of only 1.3 mm. In contrast, OCM with a lateral resolution of 4.1 µm can acquire an oxide layer thickness over an area of 4 mm × 4 mm. Furthermore, this significant difference in the absolute dimensions of the measurable oxide layer thickness is the reason for the two histograms do not perfectly coinciding. However, from the histograms in Fig. 4(e) and Fig. 6(d), we confirmed that most of the oxide layers were formed within approximately 10 µm, and as the duration of oxidation increased, the inhomogeneity of the oxide layer also increased.

Table 1 delineates the correlation between the thickness measurements procured from OCM and those ascertained from SEM images. The widths of the oxide thickness distributions, as illustrated in Fig. 4(e) and Fig. 6(d), were examined in correlation with oxidation time. The findings reveal that the thickness distribution of the oxide layer intensifies concomitantly with the increase in oxidation time, indicating a progressive augmentation in the extent of localized corrosion on the metal surface.

Table 1. Correlation of thickness measurements between OCM and SEM images: Mean and standard deviation of thickness

View Table

For the statistical analysis to obtain agreement between OCM and SEM result, we obtained the Pearson correlation coefficient (PCC) [26] that measures linear correlation between two data sets of thickness measurement results. It is the ratio between the covariance of two variables and the product of their standard deviations. The PCC (γ_xy) is expressed by [26]

It is important to note that while the analysis of the oxide layer via SEM yields precise information regarding thickness, it is concomitant with several disadvantages including intricate sample preparation, potential sample damage, and the necessity for costly equipment. Conversely, the proposed method of oxide layer analysis based on 1700nm OCM proffers the benefits of swift and non-destructive characterization of the oxide layer utilizing comparatively economical equipment. This method of analysis possesses extensive applicability in the early detection of corrosion, measurement of corrosion rate, and analysis of metal artifacts across diverse domains, encompassing semiconductor processing, fine chemical processing, and metal artifact analysis. Figure 7 shows the en-face OCM mapping images of the copper oxide specimens. Regions with thick oxide layers are depicted as bright areas, whereas those with thin oxide layers appear as dark areas. During the initial oxidation stage, Cu typically undergoes uniform oxidation. However, as oxidation progressed, localized corrosion became evident. This observation indicates that the extent of oxidation is influenced by the conditions of the metal surface, even under similar oxidation conditions. This explains the common occurrence of severe corrosion once metal oxidation is initiated.

 

Fig. 7. En-face OCM mapping images of oxidation layer thickness for 1, 2, 4, and 8 h.

Download Full Size | PDF

The OCM analysis presented in this study offers a new approach to metal corrosion analysis, enabling nondestructive acquisition of metallic oxide thickness information within seconds. This technique holds promise for various applications, including ensuring industrial stability, safeguarding cultural assets, and conducting research on metallic materials. Its speed, nondestructive nature, and ability to provide oxide thickness information make it a valuable tool for metal corrosion analysis.

5. Conclusion

This study proposes a nondestructive method for analyzing metal corrosion using optical coherence imaging. By exploiting the transmission characteristics of the metallic oxide layers, the thickness of the copper oxide layer was accurately determined using a 1700nm swept laser in the SS-OCM system. A comparative analysis of the OCM and OCT images confirmed the superior efficiency of the OCM technique for metallic oxide layer analysis to avoid multiple scattered lights from the copper oxide layer. The results obtained from the OCM analysis were further validated by comparison with those of the SEM analysis.

Unlike SEM, which involves time-consuming sample preparation and image acquisition, 1700nm OCM, which has the axial and lateral resolution of 8.3 µm and 4.1 µm, respectively, provided rapid acquisition of 3D depth images covering a 4 mm x 4 mm area within approximately 10 s. This nondestructive approach allows for the precise evaluation of local corrosion on metal surfaces, making OCM a promising tool for precise corrosion analysis. Future research should focus on exploring the applicability of 1700nm OCM to metals other than copper and investigating its potential for commercialization in practical industrial settings.

Furthermore, developing algorithms that improve the accuracy and efficiency of oxide layer thickness determination from OCM images should be prioritized. This study has academic significance as it expands the application of OCT, traditionally used in medical and biological research, to the field of metallic oxide layer analysis. The 1700nm OCM-based nondestructive metal corrosion analysis technology presented here has the potential to significantly contribute to industrial development through future research and commercialization efforts.