Introduction

Corrosion management and mitigation remain a major concern that directly impacts the lifetime of large metallic structures at the core of many industrial processes. The costs of dealing with corrosion are large and have been consistently rising with economic growth1. For context, the economic impacts of corrosion globally2 and in the United States3,4 are between 3 and 4% of GDP. Steam-based electric power generation plants are no exception, with corrosion management being a major operating expense5. Their boiler tubes, in particular, have a finite lifespan and are susceptible to many different types of corrosion-related failures with intimate links to water treatment and monitoring practices6,7. In recent years, the power grid experienced higher penetration of low-cost renewable and emerging power sources that resulted in conventional base load plants changing into peaking plants8, which is a critical challenge for aging plants9,10 that are more susceptible to tubing failures. The cycling of these old plants designed for constant base load operation has correlated to increased instances of localized corrosion events that cause boiler tube failures, especially due to condenser cooling water in-leakage, which is the biggest source of chloride and sulfate in-leakage for power plants11,12. Thereby, with investment towards plant restoration and refurbishing unlikely, considering capital cost13, the looming chances of plant closures due to corrosion are high. Given that corrosion-related damages are estimated to account for nearly 40% of boiler tube failures14,15, with a substantial portion attributed to localized damage16,17, understanding localized corrosion in these systems is essential. Therefore, accurate estimation of the rate of localized corrosion possible for common operating conditions to prevent failures and determine the residual lifetime of the parts are crucial for reliable monitoring practices.

Under normal operating conditions, waterwall tubes experience annual corrosion rates significantly lower than the ASME guideline allowance of 2 mm18 for carbon steel in steam systems. This allows them to maintain their integrity for many years without the need for replacement, with a design life spanning decades. However, the introduction of chloride and sulfate species due to make-up water contamination or feedwater contamination via condenser tube leaks19 have led to severe localized corrosion events that can cause tubing failures in a matter of days20,21,22,23. While the occurrence of these events can be managed by adherence to strict industry guidelines based on the type and level of contamination (called action level limits), the examination of the underlying localized corrosion phenomena in these extreme environments remains difficult. Standard corrosion quantification techniques are destructive and/or lack the resolution needed to capture specific localized events within the response timeframes (24–60 h) recommended by the industry. Weight-loss methods using corrosion coupons require long testing durations on the order of a few days to months in order to obtain reliable data and can only provide information on the localized damage if the oxide layer is removed, which introduces additional uncertainties24,25. Scanning electron microscopy can provide high-resolution data of individual corrosion features, but it is destructive and requires meticulous sample preparation to acquire a single cross-section image of a feature26. Although electrochemical noise techniques have been used for identifying localized corrosion initiation, it requires complex data analysis and need further development to overcome the limitations in terms of models and methods used to accurately quantify localized corrosion and pitting26,27,28. Optical profilometry is able to examine individual events and penetration depth of corrosion events, like X-ray computed tomography (X-ray CT), but the studies focus only on flat surfaces and also fail to capture the structure or shape of the pit beneath an oxide surface as they cannot penetrate the corrosion layer29,30. Ultrasonic detection is preferred as a technique to detect localized corrosion in situ, but similar to profilometry it cannot provide detailed structures of pit propagation underneath the surface31. Additionally, quantification is affected by surface roughness and surface defects that can render the technique ineffective. Immersion methods used to decrease the interference from bulk defects to improve detection limits the usage to low temperature conditions32. Therefore, developing a technique that addresses the limitations of the various methods previously discussed can greatly advance the accuracy of localized corrosion quantification.

X-ray CT microscopes can now reach submicron voxel resolutions without the need for a synchrotron to provide thousands of cross-sectional images non-destructively33. Although their use in corrosion degradation studies is increasing, most studies focus on large objects typically embedded in a media that is difficult to remove, like reinforced concrete34,35,36. The size of scan samples used ranged from a few millimeters to a few centimeters34,36, with resolutions as high as a few micrometers37. Volume loss is calculated from data processing, a method of corrosion quantification analogous to weight-loss methods38. In most studies with this technique, test durations are usually multiple days with large samples to calculate uniform corrosion rates34. While a few studies have identified the potential of this technique to detect much smaller corrosion features with lesser time durations, these advantages have not been demonstrated35,38. Örnek et al performed a study on steel wire exposed to atmospheric conditions over a duration of two years to both visualize the effects of localized corrosion, and also to quantify penetration and calculate localized corrosion rate39. Herein, we examine how high-resolution X-ray CT (submicron) microscopy can non-destructively assess the early stages of localized corrosion events with 24–60-h exposure durations on millimeter-sized samples, to provide new insights into the influence of chloride and sulfate contaminants on high-temperature, high-pressure steam cycle corrosion processes. We found that X-ray CT was able to provide measurements of the size, depth, and frequency of pits formed while exposed to different contamination levels in standard AVT steam cycle chemistries.

Results

Localized corrosion surface coverage

Three-dimensional imaging captured clear differences in degradation patterns from sulfate and chloride-contaminated solutions. Control scans (uncontaminated AVT solutions) showed an almost smooth surface largely missing localized corrosion events (See Fig. 1a), with only a few small features found on the surface (See Supplementary Figs. 1, 2). With contaminated solutions, several smaller pits were formed from sulfate contamination, whereas a few larger pits were formed from chloride contamination. Sulfate-contaminated solutions exhibited around 10 or more pits per mm2, which were fairly evenly distributed along the surface of the sample (See Fig. 1b). In contrast, chloride-contaminated solutions yielded just a few pits (1–3 pits per mm2), but they covered significant portion of the scanned region, in some cases as much as 0.45 mm2 in surface coverage (See Fig. 1c).

Fig. 1: 3D renderings from CT data of corrosion coupons after 60 h of exposure at 300 °C and 124 MPa.
figure 1

a uncontaminated AVT solutions, b AVT with 3 ppm hydrogen sulfate, and c AVT with 0.6 ppm hydrogen chloride.

Full size image

In some cases, chloride-exposed samples had localized corrosion events that extended beyond the initial scan region, suggesting larger scan sizes are beneficial given the feature sizes of chloride-induce localized corrosion (See Supplementary Section 4.2). These findings appear to align with existing literature40,41, albeit their samples were exposed for multiple days, wherein concentrated chloride environments resulted in pits averaging 0.05 mm in diameter. The findings affirm chloride’s nature as an aggressive pitting agent with a much faster growth rate compared to sulfate, which resulted in pits that were much slower to propagate40,42.

Localized corrosion penetration depth

Cross-sectional images of the corrosion coupons showed that chloride contamination resulted in much greater pit propagation and penetration rates when compared to sulfate contamination. Even with control scans (coupons exposed to uncontaminated AVT solutions), localized corrosion rates were noticeably faster than uniform corrosion rates (See Fig. 2a). In 168 h, pits penetrated 0.003 to 0.01 mm into the sample surface, and the localized corrosion rates varied from 0.15 to 0.50 mm y−1 depending on the pit observed (See Supplementary Section 1). For coupons exposed to action level 3 chloride-contaminated solutions, the deepest pit observed penetrated 0.054 mm into the sample after 60 h of exposure, equating to a corrosion rate of 7.78 mm y−1. Other notable pits ranged from 0.021 to 0.048 mm, equating to a corrosion rate of 3.06 to 7 mm y−1. In contrast, sulfate contamination resulted in relatively shallow pits, ranging from 0.014 to 0.022 mm in depth, correlating to corrosion rates of 2.07 to 3.32 mm y−1. Unlike chloride, sulfate contamination led to the formation of multiple pits with smaller depths and lower surface footprint, distributed roughly evenly around the sample, consistent with the observations from the 3D analysis (See Fig. 2b).

Fig. 2: 2D cross-sectional views from CT data of corrosion coupons showing depth of penetration after exposure.
figure 2

a uncontaminated AVT b AVT with 3 ppm hydrogen sulfate, and c AVT with 0.6 ppm hydrogen chloride.

Full size image

Major pits formed like the ones observed in Fig. 2c displayed oxide domes above the surface of the pit, showing that the under-deposit corrosion resulted in an inflated oxide layer from the excess corroded iron from the base metal (See Supplementary Section 2). Oxide layer thickness resulting from uniform corrosion were found to be around 0.005 mm or smaller, but the inflated oxide domes over localized features were much bigger in comparison—for example, the two surface features observed in Fig. 2c measures 0.029 (left) and 0.033 mm (bottom). 2D cross-sectional images also depict the differences between the pits formed by chloride and sulfate contamination—multiple small pits all along the circumference for sulfate and a few big pits focused on a specific region for chloride.

Further increasing the chloride concentration did not noticeably increase the pit frequency to a significant extent but increased pit depths and subsequently elevated localized corrosion rates (See Supplementary Section 4). As a specific example, a 10-ppm chloride exposure test resulted in a pit depth of 0.078 mm after 24 h, which equates to a corrosion rate of nearly 30 mm y−1 (See Fig. 3). The complex pit structure observed also highlights how X-ray CT can be used to examine many three-dimensional aspects of pit formation.

Fig. 3: Severe pitting corrosion features formed from 10-ppm chloride exposures.
figure 3

a 2D cross-sectional view, b Processed 3D rendering of the segmented pit showing penetration into the surface (Gray = metallic outer surface, Red = metal oxide due to inward pit growth).

Full size image

Uniform corrosion rates

While uniform corrosion rates were comparable between electrochemical techniques and X-ray CT methods, neither captured the trend or severity of the localized corrosion events. Uniform corrosion rates were often in an order of magnitude smaller than localized corrosion values. Likewise, uniform corrosion rate measurements, although easier to monitor in real-time, did not capture the increased risk of failure from larger localized corrosion rates. The uniform corrosion rates calculated from electrochemical methods showed that uniform rates from sulfate fell within the range observed in control samples, with chloride values being a little larger. However, they failed to accurately depict the severity of the increasing localized corrosion rates, which increased by more than an order of magnitude with contamination, thereby limiting their reliability as indicators of risk. Given that the localized corrosion rates varied considerably from pit to pit, we averaged the localized corrosion rates from the ten deepest pit sites for each sample to assess how uniform and localized corrosion values differed for the cases examined. The averaged localized corrosion rate from sulfate contamination was found to be 2.69 mm y−1, approximately 17 times higher than the measured uniform corrosion rate of 0.16 mm y−1. Similarly, chloride contamination led to an averaged localized corrosion rate of 5.55 mm y−1, nearly 14 times higher than the uniform corrosion rate of 0.40 mm per year obtained from electrochemical methods (See Fig. 4).

Fig. 4
figure 4

Comparison of corrosion rates obtained from X-ray CT scan and electrochemical methods.

Full size image

The stark difference between the corrosion rates obtained through the two methods emphasizes the importance of addressing localized corrosion and highlights the advantages of X-ray CT scans in enhancing corrosion rate quantification alongside established electrochemical methods. The uniform corrosion rates calculated from X-ray CT scan data using the volume loss method were found to be comparable to those calculated from the electrochemical method and displayed a similar trend in terms of the impacts of the contaminants. Although the precision of volume loss measurement is lower at the micrometer scale, the uniform corrosion rates from volume loss calculations were within ±0.3 mm y−1 of rates from the electrochemical method. The calculated uniform corrosion rate using X-ray CT scan technique from sulfate contamination was 0.38 mm y−1 and chloride was 0.35 mm y−1, showing the same trend as that of the corrosion rates from electrochemical method—increased uniform corrosion rates in the presence of chloride or sulfate contaminants. The sample region for the scan is almost one-tenth of the area considered in the electrochemical measurements, and the accuracy of the volume loss method can be improved significantly by scanning a larger region. This correlation reinforces the credibility of the X-ray CT scan method of corrosion quantification at this scale, demonstrating the capability to assess both uniform and localized corrosion rates with micrometer resolutions within a short timeframe of exposure to the contaminants.

Discussion

We used high-resolution X-ray CT to examine carbon steel degradation due to localized corrosion events while exposed to contaminated boiler water treatment solutions. For action level 3 contamination levels, chloride resulted in severe pitting in localized regions with a maximum corrosion rate of 7.78 mm y−1 observed. Despite the higher concentrations in sulfate tests, the pitting was less severe with a maximum corrosion rate of 3.32 mm y−1 resulted in significantly less penetration. However, the localized corrosion features from sulfate were distributed all along the sample. Experiments with higher concentrations of chloride resulted in faster penetration rates, with one pit sustaining an average corrosion rate of 30 mm y−1 over a 24-h period.

This study provided new insights into localized corrosion behavior, showing how localized corrosion rates deviate from uniform corrosion trends for sulfate and chloride contamination. In addition to confirming the widely accepted notion that conventional electrochemical techniques (and the uniform corrosion rates they measure) do not fully capture the extent of corrosion, the findings underscore the significant difference in magnitude between localized and uniform corrosion rates. Furthermore, it demonstrates the distinct impact of sulfate and chloride contaminants on localized corrosion behavior. This study, more importantly, validates the ability of the technique to estimate localized corrosion at submicron resolutions within 24 h, demonstrating that X-ray CT can be a vital tool in the field of corrosion, and with further refinement, holds promise for accurately quantifying both uniform and localized corrosion rates across various applications.

Methods

Materials

Corrosion studies were performed in three environments: standard all-volatile treatment (AVT)21,43, AVT with chloride contamination, and AVT with sulfate contamination. Solutions were prepared from commercial stock solutions of 1 M hydrochloric acid (HCl, Reagents), 1 M sulfuric acid (H2SO4, Reagents), and 1 M ammonium hydroxide (NH4OH, Reagents). All dilutions were made with deionized water to reach the target pH and ion concentrations. To obtain a reducing environment, the deionized water used in solution preparation was purged for 24 h with argon which reached dissolved oxygen levels less than 10 ppb44. Ammonium hydroxide was used to set the solution pH at 9.4 at 25 °C, prior to the addition of any contaminants. In this study, different concentrations of chloride and sulfate contaminants were used to examine their impact on the corrosion behavior. Prior to testing, all samples are subjected to pre-oxidation for 168 h with AVT solutions without contaminants, to aid in oxide layer formation. Unless stated otherwise, contamination tests were performed for 60 h each with a concentration of 3 ppm for sulfate contamination studies and 0.6 ppm for chloride contamination studies. These contamination levels were selected to match EPRI action level 3 limits of the recommended cycle chemistry guidelines for feedwater treatment—the highest permissible contaminant level requiring immediate action if exceeded45. All corrosion coupons were black annealed carbon steel wires with a 0.9 mm (1/28”) diameter. The exposed length of the wires for X-ray CT scanning was 4 cm, and the exposed length for the electrochemical working electrode was 2 cm. A minimum of two tests were conducted for both sulfate and chloride contamination conditions, with additional tests performed for the control conditions. Representative results are presented in the main manuscript, while the remaining data are available in the supplementary information.

Experimental setup

The laboratory test system used for corrosion experiments was published previously by Raman et al.44. The same test procedure was followed to replicate the high-temperature and high-pressure (HTHP) condensed phases within power plant steam cycles. Briefly, ports at the top of the autoclave were used to include electrochemical sensors, a thermocouple, a corrosion coupon holder, and fluid flow tubing (See Fig. 5). Electrochemical measurements were performed via a three-electrode assembly wherein all three electrodes were carbon steel wires connected to a potentiostat through one of the autoclave ports (See Fig. 6).

Fig. 5
figure 5

A diagram of the internal components of the high-pressure, high-temperature autoclave system.

Full size image
Fig. 6: Diagram of the experimental setup and supporting equipment used for the corrosion tests.
figure 6

Primary components are: (1) potentiostat, (2) autoclave, (3) heater, (4) HPLC pump, (5) test solution vessel with sparger, (6) inert gas cylinder, (7) outlet to release excess fluid and pressure.

Full size image

Electrochemical techniques

All electrochemical assessments were conducted via a Gamry 600+ potentiostat. Two electrochemical tests were used to acquire in situ corrosion data: electrochemical impedance spectroscopy (EIS) and linear sweep voltammetry (LSV). EIS and LSV were performed continuously over the test duration, a test sequence was performed every 5 h with the final 12 results averaged. For EIS, data points were systematically gathered across a frequency range from 10 mHz to 10 kHz, with intervals set at five points per decade at a potential of 0 V vs open-circuit potential (OCP) and a perturbation potential of 30 mV. In this study, the solution resistance (Rsol) of the corrosion sensor obtained from EIS was quantified as the real impedance value at high frequencies. For LSV, data points were acquired using a scan rate of 1 mV s−1 within the potential range of −0.7 V to 0.7 V vs OCP. Although the expected OCP was 0 V due to the use of carbon steel electrodes for all three components, the measured OCP fluctuated within a ±10 mV range. To account for these variations, the applied potential was referenced against the OCP. Collected LSV data were corrected for their open-circuit potential and Rsol values46 to generate overpotential-current curves.

The obtained overpotential-current curves, and the Tafel slopes of the anodic and cathodic overpotential curves were used to determine the corrosion current, Icorr, which was then used in the corrosion rate calculation using the following equation24:

$${CR}=\frac{{I}_{{corr}}M}{A{\rho }{zF}}$$
(1)

Here, we used the molar mass of iron (55.85 g mol−1) for M, A was the exposed surface area of the working electrode, ρ was the density of the iron (7.87 g cm−3), z was the number of electrons released in the corrosion reaction (z = 2) and F is Faraday’s constant47.

X-ray computerized tomography

X-ray CT scans were used to obtain three-dimensional images of corrosion samples to observe localized corrosion effects. The imaging process utilized a Zeiss Xradia Versa 620 instrument to scan cylindrical wire samples, achieving a voxel resolution of 710 nm and an optical magnification of 4X unless otherwise stated. An estimated length of 1.4 mm of the sample was scanned to obtain approximately 2000 cross-sectional images. Each cross-sectional image exhibited dimensions within the range of 1300 to 1500 voxels for both length and width. The grayscale representation of the images ranged from 0 to 255 for each voxel, where 255 denotes the densest region, shown as white, while 0 represents the least dense region, displayed as black.

Corrosion products had a lower density than the base steel, resulting in lower grayscale values, typically below 200. This distinction visibly aided in identifying corroded regions and is frequently observed as a thin gray ring surrounding the base metal (See Fig. 7a). The quantification of pit depths along the radius of the sample was determined from 2D cross-sectional images using ImageJ software. A grayscale range of 100–200 was set as the threshold to segregate the corroded layer from the base metal (See Fig. 7a). After processing, the selected region was converted to a black (corroded region) and white (base steel) image to maximize the contrast of the region impacted by the corrosion experiment (See Fig. 7b). For each steel sample analyzed, the deepest pits were identified, and their pit depths were measured to obtain their local corrosion rates. The total penetration into the base metal from each pit was then converted to a localized corrosion rate (mm y−1) using the following equation:

$${CR}=\frac{D* 24* 365}{t}$$
(2)

Here, D is the penetration depth (mm) observed during the experiment, t is the duration of the experiment (hours), which are converted to the corrosion rate per year (24 h × 365 days). Assessing whether the corrosion rate remained constant throughout the duration of the test and if it would propagate further at a uniform rate is beyond the scope of this study. Consequently, an assumption was made to extrapolate the corrosion rate to units of mm per year, enabling comparison with values typically reported in the literature.

Fig. 7: 2D cross-sectional CT image depicting segregation of the corroded layer.
figure 7

a One raw image from around 2000 cross-sectional images within a scan data stack and b the same image after refining the data in ImageJ to segregrate the corroded regions from base metal.

Full size image

Amira-Avizo software was used for the generation and analysis of the 3D images of corrosion samples. Segmentation was also performed on the 3D images to facilitate the visualization of localized corrosion features under the oxide layer of the sample (See Fig. 8).

Fig. 8: 3D data analysis showing segmentation of the base layer from raw data.
figure 8

a Preliminary volume rendered sample image from a sample scan data. b Final rendering after segregation of the data into layers to only extract base metal portions.

Full size image

The three-dimensional surface view offered insights into the distribution and frequency of pitting events, enhancing the understanding of the corrosion behavior in the studied material. Additionally, the extraction of the base steel provides an opportunity to measure the volume of the sample, which could be used as an alternative method of uniform corrosion rate measurement using volume loss calculations. For this study, pits were classified as small if the area covered, in terms of length and width were smaller than 0.1 × 0.1 mm, and they were considered shallow if the penetration was less than 0.02 mm (which equated to <3 mm y−1).