Anthony E. Hughes

In an interview in 1982, which was 1 year after he shared the Nobel Prize, Kai Siegbahn was asked about his opinion regarding the name of the technique he had developed. Siegbahn had named it “electron spectroscopy for chemical analysis” (ESCA), but the community was choosing to call it “X‐ray photoelectron spectroscopy” (XPS). Now, more than 40 years later, 20 XPS experts have given their opinions on Siegbahn's response and the name of the technique. Some of these participants have been doing XPS for many years and have provided a historical perspective on this issue. While there is no call in these comments for the community to return to “ESCA”—“XPS” is regarded as a more than an adequate name, and insisting on a name change at this point in time would probably only create confusion. However, some of the participants of this study still consider “ESCA” to be an acceptable way to refer to the technique, especially when it is used in a chemical context.

This study considers how poor x-ray photoelectron spectroscopy (XPS) peak fitting in the scientific literature is both affected by previous precedent and affects future published work. It focuses on a highly cited paper (the “Subject” paper) from a respected journal that contains incorrect S 2p peak fits. This paper was studied in a genealogical fashion vis-à-vis the XPS peak fitting in its “child,” “parent,” “grandparent,” and “great-grandparent” papers. Interestingly, precedents were not followed to a high degree between parent and child papers. However, in many cases, even when the authors of a study did not follow the incorrect precedent that they cited, they still incorrectly fit their data. Thus, not necessarily for good reasons, the effects of poor XPS peak fits on future generations of papers may be less than some experts had expected or feared. In many cases, older papers appear to contain better XPS peak fitting than newer ones.

Computed tomography (CT) is used in many fields from medicine to materials science. CT covers a range of different techniques with most types falling into X-ray or electron generated. Generally, X-ray CT is nondestructively, electron techniques are a mixture: for nontransmission electron techniques (e.g., backscatter electrons) the sample needs to be sectioned, whereas transmission (such as TEM) can be done nondestructively. X-ray CT has been used to address a range of problems in corrosion from determining pit volumes in metals to determining the transport paths of inhibitor molecules within paint systems. Electron techniques have also been used with serial sectioning being the most common. Atom Probe tomography is another CT technique which recreates 3D volumes of a sample at the nanoscale and is based on ion position detected after ablation or evaporation. This overview highlights some of the work that has been performed in these areas.

In this paper, a review of the analysis of Fe 2p3/2 peak and other transition metals in the austenitic stainless steel literature is presented. It reveals the significant shortcomings of the most widely used approaches, based on the principle of “chemistry fitting,” where single symmetric peaks are used to represent either individual oxidation states or specific compounds. No meaningful conclusions can be drawn from these commonly employed two- or three-component peak fitting (2C and 3C) approaches; the implication being that a large portion of the literature that relies on this approach is flawed. As a significantly more accurate and reliable alternative to “chemistry fitting,” we also assess “envelope fitting” (using empirical multiplet structures) and examine its limitations when applying the approach to austenitic stainless steel data. A detailed comparison of these two fitting approaches is described in Part I. For other elements such as Cr 2p, the problems associated with using single components to represent oxidation states or compounds are not as severe. It was found that it does not impact binding energy measurements, but does influence relative intensities, which will have a flow-on effect for oxide thickness calculations and obtaining a correct understanding of the surface more broadly.

Analyzing transition metal XPS peaks is widely used to determine surface composition and chemistry. However, these peaks have a complex structure, which is still the subject of investigation. Fe 2p analysis is a case in point where the multiplet structure and many-electron-effects lead to peak shapes that cannot be analyzed using standard approaches. Examination of the literature reveals that one of the most widely used approaches to data reduction when processing Fe 2p3/2 spectra involves using symmetric two- or three-component peak fitting with each peak effectively acting to capture a single chemical species (chemistry fit) in the complex spectra. Herein, this approach is compared to an envelope fit approach using Biesinger multiplet components of known iron oxides to determine how effective these methods are in reproducing iron oxide composition. Mixed oxide and metal XPS Fe 2p spectra were synthesized using reference spectra collected experimentally. For the first time, the accuracy and differences between the two approaches are reported. It is demonstrated that no meaningful conclusions can be drawn using single symmetric peaks to analyze complex Fe 2p3/2 spectra, implying that a large portion of the literature is flawed. The envelope fit approach, however, is shown to provide useful information regarding oxide ratios in mixed iron oxide materials, though limitations do exist. A methodology for evaluating the quality of XPS analysis of Fe 2p3/2 spectra is proposed for benchmarking new submissions so that reviewers, authors, and editors can assess these submissions.

This work demonstrates an approach towards the understanding of multi-scale and open-circuit localised electrochemical processes of AA2024-T3 in the presence and absence of an environmentally friendly rare-earth inhibitor; cerium diphenyl phosphate (Ce(dpp)₃). At high temporal resolution, a wire bean electrode (WBE) made from 100 identical AA2024-T3 wires revealed sudden increases in galvanic anodic and cathodic activities immediately after dosing of 50 and 100 ppm of the inhibitor and an overall suppression of macro-scale activities by increasing the inhibitor concentration to 200 ppm, suggesting it as a fast-screening tool for inhibitors and measuring inhibition efficiency. At high spatial resolutions, scanning probe electrochemical techniques confirmed local activation of corroding microstructures on individual AA2024-T3 wires similarly by dosing the inhibitor up to 100 ppm. In agreement with WBE findings, the effective shutdown of both anodic and cathodic activities occurred after increasing the inhibitor concentration to 200 ppm confirming the optimal concentration of the Ce(dpp)₃ and the mixed mode inhibition mechanism of this selected inhibitor on AA2024-T3.

The protection of ferrous metals in acidic environments is important in many industries. Extending the pH range of organic inhibitors to low pH has been achieved with the addition of iodide ions, which facilitate adsorption. It was of interest to see whether similar outcomes could be achieved with inorganic inhibitors. To this end, this paper examines the influence of potassium iodide addition on the level of corrosion protection provided by Ce(NO₃)₃ in 3.5% NaCl electrolytes over a pH range of 2.5 to 8. Potentiodynamic polarization was used to assess percentage inhibitor efficiency (IE%), and scanning electron microscopy, energy dispersive X-ray spectrometry, and X-ray photoelectron spectroscopy were used to characterize the corrosion product. It was found that KI alone provided only poor corrosion inhibition except at pH 2.5, where nearly 85IE% was achieved. Its addition to the cerium electrolytes was generally in excess of 90% and over 97% for the optimum concentration. The addition of KI seemed to change the mechanism of formation of corrosion products from predominantly Fe₂O₃ to a mixture of FeOOH, Fe₃O₄, and Fe₂O₃, which were more adherent. Corrosion protection was extended to pH 4, but under the conditions explored here, no additional protection was evident at pH 2.5.

The inhibition of localised corrosion is known to involve multiple processes taking place over a range of time and length scales that are difficult to study by conventional electrochemical methods. This work demonstrates an approach to probing complex localised electrochemical processes by combining an electrochemically integrated multi-electrode array (also known as wire beam electrode (WBE)), scanning vibrating electrode technique, scanning electrochemical microscopy and surface analytical techniques. Each technique reveals certain aspects of the dynamically changing, multi-scale, inhibitor-microstructure interactions and local chemistry over a heterogeneous AA2024-T3 alloy surface in the presence of an environmentally friendly inhibitor; cerium diphenyl phosphate (Ce(dpp)3).

Machine learning (ML) is providing a new design paradigm for many areas of technology, including corrosion inhibition. However, ML models require relatively large and diverse training sets to be most effective. This paper provides an overview of developments in corrosion inhibitor research, focussing on how corrosion performance data can be incorporated into machine learning and how large sets of inhibitor performance data that are suitable for training robust ML models can be developed through various corrosion inhibition testing approaches, especially high-throughput performance testing. It examines different types of environments where corrosion by-products and electrolytes operate, with a view to understanding how conventional inhibitor testing methods may be better designed, chosen, and applied to obtain the most useful performance data for inhibitors. The authors explore the role of modern characterisation techniques in defining corrosion chemistry in occluded structures (e.g., lap joints) and examine how corrosion inhibition databases generated by these techniques can be exemplified by recent developments. Finally, the authors briefly discuss how the effects of specific structures, alloy microstructures, leaching structures, and kinetics in paint films may be incorporated into machine learning strategies.