Abstract
In-situ and operando techniques in heterogeneous electrocatalysis are a powerful tool used to elucidate reaction mechanisms. Ultimately, they are key in determining concrete links between a catalyst’s physical/electronic structure and its activity en route to designing next-generation systems. To this end, the exact execution and interpretation of these lines of experiments is critical as this determines the strength of conclusions that can be drawn and what uncertainties remain. Instead of focusing on how techniques were used to understand systems, as is the case with most reviews on the topic, this work instead initiates a nuanced discussion of 1) how to best carry out each technique and 2) initiate a nuanced analysis of which level of insights can be drawn from the set of in-situ or operando experiments/controls carried out. We focus on several commonly used techniques, including vibrational (IR, Raman) spectroscopy, X-ray absorption spectroscopy and electrochemical mass spectrometry. In addition to this, we include sections of reactor design and the link with theoretical modelling that are applicable across all techniques. While we focus on heterogeneous electrocatalysis, we make links when appropriate to the areas of photo- and thermo-catalytic systems. We highlight common pitfalls in the field, how to avoid them, and what sets of complementary experiments may be used to strengthen the analysis. We end with an overview of what gaps remain in in-situ and operando techniques and what innovations must be made to overcome them.
Introduction
Catalysts are instrumental in carrying out chemical transformations towards a desired product with high rates and selectivity. In the context of sustainability, the design of effective catalysts helps minimize the carbon footprint of the fuels and chemicals industries as their chemical production routes require less energy and minimize undesired by-products1. Catalyst innovation stands to also aid in the maturation of renewable-based electro and photochemical technologies to the point where they are technoeconomically competitive with conventional thermochemical systems2. These aforementioned thrusts are inherently linked to the United Nations (UN) Sustainable Development Goals (SDGs) of affordable and clean energy (SDG 7), industry, innovation, and infrastructure (SDG 10), responsible consumption and production patterns (SDG 12) and climate action (SDG 13).
Underpinning the development of next-generation heterogeneous catalysts is, of course, a thorough mechanistic understanding of how they function, which precise reaction pathways are taking place on their surfaces, and the determination of the structure of this surface under reaction conditions. Within this context, characterization techniques that probe the catalyst structure and the reaction as it is occurring (e.g. in-situ and operando) are exceptionally useful as they can elucidate catalyst transformations, reaction intermediates, and reaction environment to help piece together a more comprehensive picture of the reaction mechanism(s) at play3. In this perspective, we define in-situ techniques as those being performed on a catalytic system as it is under simulated reaction conditions (e.g. elevated temperature, applied voltage, immersed in solvent, presence of reactants) while operando techniques are those that probe the catalyst under the same, or as close as possible, conditions and when its activity is being simultaneously measured. These include considerations of mass transport, gas/liquid/solid interfaces, quantifying product formation, and more. Each in-situ/operando technique provides different pieces of the puzzle. Alongside various performance measurements, ex-situ characterization and theoretical modelling, the information from in-situ/operando experiments can be used to put together a plausible mechanism and to construct links between key characteristics of a catalyst and its macroscopically measured activity4,5,6. To this end, some techniques like X-ray diffraction (XRD) are geared towards measuring the crystalline structure of a catalyst while others like X-ray absorption spectroscopy (XAS) are more suited for measuring its local electronic and geometric structure under reaction conditions (Fig. 1). Complementary to this, techniques used to measure reactants, intermediates, and products include infrared (IR) and Raman spectroscopy (though the two can be useful for material analysis), as well as electrochemical mass spectrometry (ECMS). These primary techniques, while not being the only ones used, are the principal ones covered in this perspective. In addition, while this analysis will mainly focus on the techniques applied to heterogeneous electrocatalysis, we will also select key examples when the discussion is relevant to thermo- and photo-catalytic systems. Techniques of UV–Vis absorption, X-ray photoelectron spectroscopy, atomic force microscopy, transmission electron microscopy and others also provide important information but fall outside of the scope of this perspective.
Fig. 1: Overview of analytical techniques for mechanistic studies of electrocatalytic systems.
figure 1
Simplified illustration of the role of in-situ*, operando* and analytical techniques, as well as individual information provided by them.
Full size image
As the use of in-situ and operando techniques rapidly grows, there is correspondingly a growing number of reviews in the literature that discuss what the insights derived from the techniques were4. In contrast to this, there is not yet an adequate analysis dedicated to how to best carry out such experiments and what degree of conclusions (e.g. interpretation) can be drawn regarding the reaction mechanisms from the specific experiments carried out. Thus, the overall aim is to help the field’s development by proposing technique-dependent frameworks that can be followed to draw valid, translatable interpretations of in-situ and operando data while minimizing often-encountered pitfalls such as false positives and mechanistic overreach. This perspective is structured differently than reviews that discuss how such techniques have been used for mechanism discovery and is rather concentrated on the process of carrying out the techniques and interpretation of resultant data. Further, as mechanisms are difficult to conclusively prove, we do not seek to have an absolute ‘yes or no’ checklist but will detail layers of supporting measurements that strengthen interpretation. We first discuss key criteria and limitations translatable across all in-situ and operando measurements. These include reactor design and experimental cross-referencing of catalytic data. Next, we focus on three main classes of techniques used in heterogeneous electrocatalysts: XAS, vibrational (IR + Raman) spectroscopy, and ECMS. With each technique, we underline a base set of minimal experiments necessary and a progressive set of complementary experiments that can be used to further strengthen claims. This former includes the need to perform standard control experiments that lack the reactant/catalyst and properly correlating to previous literature examples, while the latter entails endeavors like isotope labeling and product measurement. We further detail how maximizing connections with multi-modal analysis, catalytic data and theoretical modelling can bolster findings, as well as how to avoid over-interpretation in this direction. Outlooks on big data and the role of machine learning in this direction are also discussed in a forward-looking conclusion. While this perspective focuses on heterogeneous electrocatalysis, we anticipate the general framework to be translatable to fields such as molecular catalysis to help accelerate the fields’ progression.
Reactor design
A crucial component of in-situ and operando measurements is the use of a reactor that enables a researcher to incorporate or simulate reaction conditions while simultaneously using the characterization technique of choice. This often entails the implementation of optical windows to allow the portion of the electromagnetic spectrum to be incident on the catalyst that can be integrated into the analytical instrument as well as a modification of reactor dimensions. These alterations may consequently lead to differences in the catalyst environment between that in the “regular” reactor and in the in-situ/operando cell. Against this backdrop, this section delves into the intricacies of reactor design for in-situ and operando measurements in heterogeneous electrocatalysis, aiming to provide an overview of gaps and strategic recommendations for optimizing this critical aspect of catalysis research. While each technique has its inherent requirements for reactors, we use this section to identify generally translatable aspects in reactor design.
Mismatch between the characterization and real-world experimental conditions
In-situ*/operando* reactors are typically designed per the specifications required by the instruments for characterization. This introduces a significant difference in the transport of the species in benchmarking reactors vs. in-situ reactors. While electrolyte flow and gas diffusion electrodes are typically leveraged within benchmarking reactors to control convective and diffusive transport of species, most in-situ reactors are designed for batch operation and employ planar electrodes4,5,6. Specific problems arising due to this change include poor mass transport of reactant species to the catalyst surface and a more drastic change in electrolyte composition for a batch system (e.g. pH gradients), leading to a change in microenvironment. This increases the likelihood of misinterpreting insights from such techniques. For example, Watkins et al. show that the reactor hydrodynamics controls Tafel slopes for CO2 reduction (CO2R) by altering the microenvironment at the catalyst surface7. Additionally, while an in-situ XAS batch reactor with planar oxide-derived Cu electrode showed that undercoordinated Cu sites promote binding of CO and enhance electrochemical activity8, another study in a vapour-fed device for CO reduction showed no correlation between Cu oxides and high electrochemical activity9. These examples emphasize the importance of species transport and analysis under benchmarking conditions as in-situ characterization work in batch reactors can likely have convoluted mass transport effects, limiting the ability to attribute mechanistic conclusions to intrinsic reaction kinetics.
Furthermore, sub-optimal operando reactor designs can significantly impact the response time and signal-to-noise ratio of measurements, leading to increased data acquisition time and potentially obscuring short-timescale reaction events. The path through which a spectroscopic beam travels and the path length between a reaction event and the spectroscopic probe is pivotal for rapid and precise data collection. Poor design can result in delayed response time and increased residence time of species, decreasing the probability of observing short-lived reaction intermediates. For instance, in differential electrochemical mass spectrometry (DEMS), the reactor configuration controls the proximity of the pervaporation membrane to the electrode, tuning the response time of the mass spectrometer to reactive species. With regards to best practices in eliminating long response times, Clark and Bell addressed this by depositing a CO2R catalyst directly onto the pervaporation membrane in a DEMS electrochemical cell thereby eliminating a long path length between CO2R intermediates generating at the catalyst surface and the mass spectrometry probe10. This enabled them to detect much larger concentrations of acetaldehyde and propionaldehyde (intermediates for ethanol and n-propanol, respectively) at the surface of the catalyst compared to their concentrations in the bulk.
Similarly, in grazing incidence X-ray diffraction (GIXRD), co-optimizing X-ray transmission through the liquid electrolyte and the beam’s interaction area at the catalyst surface is crucial for a good signal-to-noise ratio. Farmand et al. showed that in-situ GIXRD requires careful consideration of both the path and path length for the incident beam to minimize contact with the aqueous electrolyte, preventing signal attenuation while ensuring that the beam interacts with a sufficient surface area of the catalyst to generate useful signals rapidly11. Therefore, co-designing reactors with spectroscopic probes is necessary to bridge the gap between characterization and real-world experimental conditions.
Additionally, direct in-situ spectroscopic characterization within zero-gap reactors has been challenging as typical cell components can be opaque to measurement probes such as infrared, Raman, and X-rays. Many measurements for reactions such as CO2R, oxygen evolution reaction (OER), and hydrogen evolution reaction (HER) fall short of matching the complexities of zero-gap configurations and current densities of high-performance operation, diminishing the industrial relevance of mechanistic conclusions. Literature from more mature electrocatalytic processes such as fuel cells and some recent reports on CO2R recommend modification of the end plates of zero-gap reactors with beam-transparent windows to enable operando XAS to circumvent the challenges associated with opaque components12,13. Hence, to bridge the gap of the reactor transport discrepancy, optimizing electrochemical reactors for operando measurements requires simultaneous consideration of design criteria for benchmarking and in-situ characterization.
Standardization challenges in reactor design for in-situ and Operando techniques
Variability in reactor geometry and operation poses a challenge when comparing the results of in-situ*/operando* experiments between different research groups. Each group may have collected the data in different operating regimes, where their system could be either kinetically limited, diffusion-limited, or in an intermediate regime. This often leads to different interpretations for product selectivity and reaction mechanisms and has been a frequent occurrence in the more explored electrocatalytic reactions such as CO2R14,15. Different research groups often design reactors with varying electrode configurations, flow cell designs, and electrolyte management systems. This results in different local pH levels, CO2 concentration gradients, and product distribution even under macroscopically similar experimental conditions16,17. These discrepancies often stem from differences in mass transport and local reaction environments, influenced by the specific reactor designs. This highlights the broader challenge within the community of standardizing reactor designs to ensure that the mechanistic insights derived from in-situ experiments are directly comparable across different studies. Alternatively, another source of variability can arise from separately conducting benchmarking measurements and in-situ experiments and then correlating the results from the two by claiming it as operando measurements. This is a common practice in CO2R and nitrogen reduction studies where online product analysis is possible18,19. Researchers often use one reactor for in-situ characterization and one for benchmarking reactivity measurements, claiming their findings to be operando, however, these reactors for in-situ experiments and benchmarking may differ significantly in their transport properties, leading to misinterpretation of mechanistic insights when correlating the results from the two reactors. Even if the benchmarking is done in the same reactor used for in-situ characterization, there may still be statistical uncertainties or other variations such as quality control of the reactor fabrication or the catalyst under observation. Therefore, measuring and reporting the reactivity of the system in the operando reactor (current density, product selectivity) when operando data is acquired is the best practice to reduce uncertainty in claiming operando measurements.
Considering these major gaps impeding the approachability and reliability of the in-situ and operando characterization techniques, there is a strong need to co-design and co-optimize the benchmarking and in-situ*/operand*o reactors. The following recommendations are provided that aim to rapidly bridge these gaps, ensuring robust and reproducible experiments.
Advancements in manufacturing techniques for reactor design
The advent of additive manufacturing (AM) has introduced a positive development in the landscape of reactor design across the domain of electrocatalysis20. Specifically, incorporating AM in the design and fabrication of the reactors can help bridge the gaps noted with reactor design. AM can assist in tuning the mass transport of the reactant species by optimizing flow channels in the reactor or by introducing flow pathways to a stagnant-electrolyte cell. Proper design and optimization of the reactor, potentially guided by machine learning21, can also reduce the beam transmission path, resulting in a cleaner signal and faster data acquisition, thereby reducing temporal misalignment between the reaction events and the in-situ measurements. Moreover, rapid prototyping through AM can address the challenges with reactor variability by accelerating the co-design of benchmarking and in-situ characterization reactors to claim such reactors as operando reactors with greater certainty. Achilli et al. demonstrated this co-design aspect of benchmarking and in-situ reactors for water splitting using AM22. By designing and 3D printing an in-situ XAS reactor they showed a near-identical cyclic voltammogram (CV) compared to their benchmarking reactor, highlighting the conformity between the electrochemical behaviours of the two reactors. Efforts like these in the integration of benchmarking and in-situ characterization reactors for simultaneous measurements will be necessary to build a fundamental understanding of electrocatalytic processes and to develop new electrochemical technologies.
Evaluating transport phenomena: mimicking practical catalytic reactors
It is unrealistic to expect the entire field to adopt the same reactors due to resource availability and budget constraints. Additionally, it is not always realistic for every in-situ measurement to have ideal transport properties. Regardless of the limitations of a reactor, it is crucial to measure and disseminate its transport properties to evaluate the strength of extrapolating conclusions between in-situ characterization and benchmarking. For instance, in-situ near-ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) might necessitate gas-phase measurements or operate under low pressures (<30 mbar), which may not precisely mimic benchmarking conditions23,24. Without accommodating these factors, conducting the measurement may not be feasible. However, this does not invalidate the measurements; it simply indicates increased uncertainty in aligning in-situ characterization with benchmarking conditions. Alternatively, co-designing a spectroscopic probe with the in-situ reactor may yield an optimized reactor suitable for the characterization technique, but it may still not have the desirable mass transport. Consequently, if researchers could evaluate the transport properties of their reactors, it would significantly enhance the comparability of their data across different research groups. This entails analyses of fluid flow patterns and species diffusion rates, using redox couple single-electron transfer to measure mass transport, and electrode–surface interactions. For example, Avilés Acosta et al. conducted an evaluation of mass transport effects in operando IR measurements for CO2R25. Their findings directly influenced the design principles of the operando reactor, enabling them to construct an operando reactor that performed similarly to their benchmarking reactor. Additionally, there are several works in the literature that describe protocols for measuring transport properties using ferricyanide/ferrocyanide redox couple26,27. By prioritizing transport phenomena evaluation in reactor design, this recommendation specifically addresses the present challenges associated with reactor variability and comparison of data across different in-situ reactor configurations and data between benchmarking and in-situ reactor for operando measurements. Thus, achieving robust operando measurements requires multidisciplinary expertise to co-design the reactor and prioritize complex design tradeoffs.
Overcoming mass transport and bubble management challenges
Operating in-situ and operando cells under practical conditions requires careful attention to mass transport and bubble management. Transitioning from batch to flow systems is particularly effective for improving mass transport. Flow cells inherently employ forced convection through continuous electrolyte circulation, ensuring a steady supply of reactants and the efficient removal of products, thereby minimizing concentration gradients and maintaining a stable reaction environment. This forced convection not only enhances reactant delivery but also prevents the accumulation of gas bubbles by physically displacing them from the electrode surface. Additional approaches, such as sparging inert gas into the electrolyte, can further promote mixing and improve transport at the catalyst interface further contributing to enhanced mass transport. Strategies such as incorporating microfluidic channels enhance flow control and uniform reactant distribution, while gas diffusion electrodes (GDEs) and three-dimensional electrode architectures improve reactant delivery and provide efficient transport pathways. Specific design modifications, such as positioning electrodes closer to pervaporation membranes—as demonstrated in DEMS—can significantly reduce diffusion path lengths, enhancing response times and measurement reliability10.
Bubble formation, a common issue in XAS cells, can be mitigated through several approaches. Vertical cell designs facilitate bubble rise and removal through natural buoyancy, while hydrophobic coatings prevent bubble adhesion and promote detachment. Additional measures, such as integrating separators or optimizing flow rates and pressure, ensure that bubbles are efficiently removed before they interfere with measurements. A notable advancement in this area is the development of a recent XAS cell, which addresses gas evolution and consumption at the electrolyte/catalyst interface28. By employing a thin, porous membrane and efficient electrolyte and gas circulation loops, this design establishes an improved three-phase interface, enabling high-quality XAS data under operational conditions. The cell’s performance was demonstrated through in-situ studies of amorphous Ir oxide during oxygen evolution (OER), copper oxide during CO2 reduction (CO2R), and a highly dispersed platinum catalyst during oxygen reduction (ORR), all of which achieved high-resolution spectroscopic insights even at high overpotentials.
By integrating these forced convection strategies with targeted reactor design innovations, researchers can address mass transport limitations and bubble formation challenges, enhancing the reliability and standardization of in-situ and operando measurements.
Transparent information sharing: fostering collaboration through openness
Transparency is a cornerstone of advancing reactor design for in-situ/operando electrochemical techniques. By providing clear documentation of reactor geometries, CAD files, materials, fabrication methods, assembly instructions, and operational parameters, researchers can expedite the replication and refinement of successful designs. Many research groups are adopting this in practice by providing detailed descriptions and transport analyses of their custom-fabricated devices29,30,31. Lopez-Astacio et al. have made their cell design for in-situ XAS open source32. This open exchange of information culminates in the establishment of a shared repository of optimized configurations. This accelerates innovation while ensuring adherence to standards of reliability and reproducibility, effectively limiting the uncertainties linked to reactor variability. Furthermore, forming multidisciplinary teams is crucial for addressing challenges associated with reactor co-design. Such collaboration will require partnerships between experts in benchmarking reactor design, advanced characterization scientists, and in-situ*/operando* reactor design experts, recognizing the limitations of relying solely on expertise in one area. This can foster the development of co-optimized reactor configurations and facilitate information exchange, providing newcomers to the field with valuable insights while offering opportunities for fresh contributions. By implementing such practices, developing multidisciplinary collaboration and sharing reactor designs, this recommendation addresses all challenges associated with reactor design discussed in this perspective. In all, the synergy between reactor standardization, AM, and information sharing is illustrated in Fig. 2.
Fig. 2: Reactor design limitations and best practices.
figure 2
Schematics of gaps and recommendations to design reliable reactors to ensure best practices for in-situ and operando characterization.
Full size image
X-ray absorption spectroscopy
X-ray absorption spectroscopy (XAS) is a technique that is used to determine the element-specific local geometric and electronic structure around the absorbing element. For convenience, mainly due to the interpretation of the data, XAS is divided into two main regions: the X-ray absorption near edge structure (XANES) and the extended X-ray absorption fine structure (EXAFS) regions. We keep our discussion focused on the use of hard X-Rays (above 3000 eV) which are more often used in in-situ*/operando* experiments in electrocatalysis.
The XANES region is loosely defined as the region from ~50 eV below to ~100 eV above the absorption edge. The intensity and shape of XANES are directly correlated to the density of unoccupied and delocalized states resulting from the excitation of a core electron to higher energy levels19. Additionally, localized electronic states can also contribute prominently to XANES, giving rise, e.g., to intense pre-edge features. XANES provides insights into the oxidation states and electronic configurations of the chemical absorbing species, where a change in oxidation states can cause a shift of several electron volts in the absorption edge position. The EXAFS region resulting from the modulation of an atom’s probability, due to a quantum interference effect, of absorbing an X-ray due to the chemical and physical state of the atom extends to about one thousand eV from the absorption edge and offers information about interatomic distances and the coordination of neighbouring atoms.
Thus, XAS is a probe of the local structure around the absorbing atom due to the mean free path of the photoelectron. It thus does not require long-range order to gain information on the oxidation states, coordination environment, and interatomic distances. This is helpful for characterizing amorphous samples, which can often be the active phase in electrochemical reactions. Moreover, as the core levels that are studied are in the hard x-ray range and it is a photon-in/photon-out technique it can be used to probe the structure of a working catalyst in the presence of the reactants and products.
Detection mode and sample preparation
There are two primary modes of data collection in XAS: transmission and fluorescence. The most utilized mode is transmission, where x-ray energy before and after passing through the sample is measured, following Beer’s law. However, this mode presents challenges in electrochemical experiments, as the cell design must allow x-rays to pass through the material and the cell without influencing the outgoing x-rays, and often the electrolyte absorbs much of the x-ray intensity. In fluorescence yield mode, the fluorescence signal emitted by the absorber element is collected. This is by far the most prevalent mode for collecting XAS data of electrocatalysts. The fluorescent signal is collected at a 45° incidence angle, with the detector placed perpendicular to the incoming beam. Using this mode coupled with the design of the electrochemistry cell, discussed below, entails the x-ray beam only traversing a thin layer of electrolyte.
When employing fluorescence yield mode, several considerations must be assessed when preparing the sample on the working electrode to be studied. Fluorescence measurements require more dilute samples to ensure that the signal is not attenuated by self-absorption effects which results in a distortion of the spectrum. However, even a dilute sample for XAS might require a higher loading than the amount the researcher typically uses for electrochemistry in their home lab. The electrode preparation for XAS might involve using multiple layers of the catalyst ink on the electrode, which can alter the electrochemical activity of the sample. For example, thick layers of catalyst can lead to gas being trapped in catalyst layers, blocking catalyst sites, and causing catalyst delamination and loss of electrochemical activity. Studies have indicated that reducing the flux and catalyst layer thickness can result in a better XAS signal while maintaining electrochemical activity similar to lab settings33. Additionally, the catalyst layer should be uniform to produce an ideal low-noise signal and to ensure that the whole electrode is at the same potential. Therefore, careful attention to sample preparation and experimental conditions is crucial when using fluorescence mode for XAS in electrochemical studies, and it has been shown that even changes of 1 nm thickness of catalyst layers can dampen XAS features34. To demonstrate that the catalyst is performing similarly (e.g. similar rates, selectivity) during the XAS measurement and its activity is not altered through deposition in a thicker layer, a best practice here would be, when possible, to carry out product measurements during/post experiment and to compare this data to benchmark catalytic data on the material.
XAS cell design
Choosing the appropriate electrochemical cell design is critical to ensuring that the maximum information content is obtained with the highest quality XAS spectrum. There is often a compromise that needs to be made between a cell design that optimizes the XAS data collection and one that optimizes the electrocatalysis. There are several classes of in-situ/operando XAS cells described in the literature including batch cell, flow cell, and H-cell designs. The first two are often set up in a half-cell mode while the third is a full-cell. Batch cell34,35,36 designs for gas evolving electrochemical reactions are simple and all components are static. However, it is often the case that longer in-situ experiments can change the local environments of the cell, and fresh electrolyte is needed. Moreover, this design can be susceptible to bubbles which interfere with XAS data quality during collection. For flow cell designs37,38,39, they ensure that fresh electrolytes can keep a constant environment, and these cells aid in minimizing bubble formation on the catalyst. Third, there is the use of H-Cell14,40,41 where the design is similar to a reactor. H-Cell designs have two chambers, where the working and counter electrodes are separated from one another by a membrane. This is useful if one wants their working electrode environment to be different than that of the counter electrode, such as in CO2 reduction, or if one wishes to keep products formed separated. Finally, there is a grazing incidence cell which allows for surface sensitivity and has been used to study the surface of continuous film electrodes11,42. Grazing incidence cells can often be modified for bulk or surface sensitive measurements such as in Fig. 3.
Fig. 3: In-situ XAS spectroelectrochemical cell and resultant data.
figure 3
Schematics of the in-situ GI-cell in (A) full and (B) in cross-section view. The Ag/AgCl reference, platinum wire, and sample are labeled RE, CE, and WE, respectively. The configuration of the beam and detector is indicated with arrows showing the paths of the X-rays and fluorescence from the sample. Total normalized absorption (top) and its first derivative (bottom) for the MoN film in O2-bubbled 0.1 M HClO4 at 0.5, 0.7, and 0.8 V vs. RHE are displayed for the (C) surface (0.1°) and the (D) bulk (5.0°) with EXAFS and fits for the (E) surface and (F) bulk for the different potentials in-situ with data indicated with circles and fits noted with lines. Shaded regions highlight the pre-edge and edge positions. Reprinted from ref. 42. Copyright 2020 American Chemical Society.
Full size image
Limitations in surface sensitivity
A key technical issue is that in electrocatalysis solid–liquid surface boundary is where the catalyst is active and where the catalyst may restructure with applied potential. XAS is a bulk averaging technique and, therefore, it is not in itself surface sensitive. If the surface atoms are not a significant fraction of the total amount of the element under study, then XAS may not be sufficiently sensitive to observe small changes. In this case, one must be careful in stating that minimal changes occur during catalysis when, in fact, the surface may be changing substantially while the bulk of the material remains the same. To avoid false interpretations and incorrectly determine that minimal surface changes are occurring, one can carry out measurements that specifically increase surface sensitivity.
The easiest way to improve the surface sensitivity of an XAS measurement is to use smaller diameter nanoparticles or ultrathin films of the electrocatalyst. Nanoparticle samples of a diameter less than 3 nm with low loading can allow for >50% of your total sample to be at the surface43. In the work of Alesker et al.44 highly active 2 nm Pd/γ-NiOOH cubes for hydrogen oxidation reaction (HOR) were investigated under operando conditions. Due to the small dimension of the particles small changes at the surface were able to be observed, where it was discovered that it was not the interaction of Pd and γ-NiOOH that made the catalyst active, but rather adsorption of hydroxyl groups onto the y-NiOOH surface causing the oxidative removal of H from the Pd surface. In another example, Thakur et al.45 used PtxCoy alloy nanoparticles with a diameter of <2.5 nm to show that increased cobalt at the surface increased the oxidation of platinum and that this lead to a degradation of the particle and decrease in electrocatalytic activity. Li et al.34 noted that oxygen within a layer of IrOx of 2 nm was dynamic at OER potentials. When the layer of IrOx was increased from 2 to 3 nm no changes in the O spectra were noted because the lattice oxygen sites became inaccessible for catalysis.
For catalyst particles larger than ~3 nm spectral subtraction methods can be used. Here, a reference material spectrum is subtracted from a spectrum at a potential of interest46 to enhance subtle spectral changes. Cao et al.47 using the difference of the EXAFS for Ni(OH)2 with sulfur before and after OER it was shown that rearrangement of sulfur ions occurred during the OER process. This inference was not clear through the XANES measurement which only showed a slight increase in absorption edge position after OER catalysis.
Another method of distinguishing the surface from the bulk during an operando process is through the use of modulation excitation spectroscopy (MES), where temperature, pH, or potential is modulated. MES is an analytical technique utilized to enhance the sensitivity and selectivity of spectroscopic measurements, This technique involves the application of a modulating signal, typically in the form of an oscillatory perturbation, to the system under investigation while simultaneously monitoring the response signal. By employing phase-sensitive detection methods, MES effectively distinguishes the desired signal from background noise and interference, allowing for the analysis of weak spectral features. Its ability to resolve subtle changes in spectral changes enables researchers to understand the parts of the catalyst that change with the stimulus, and thus are likely to be active. However, the formation of such species should be correlated to the reaction rate to help determine that it is part of the reaction mechanism of interest instead of a competing reaction pathway. Czioska et al.37 used MES by varying the potential for OER and in doing so, the researchers were able to demodulate their spectrum to see clear changes in their active surface species at different potentials, suggesting a mechanism change. However, one should keep in mind that this technique requires that the reaction should be reversible.
While a comprehensive discussion on the use of soft X-rays (around 100–3000 eV) is outside the scope of this work, their use can help circumvent some inherent challenges associated with hard X-Rays (above 3000 eV). The use of lower energy X-Rays often prohibits EXAFS analysis and also requires a high-vacuum setup because of their lower penetration depth. However, one can perform measurements in total electron yield (TEY) mode with soft X-Rays because of higher electron yields when using lower energy X-Rays48. In this setup, the emission of Auger electrons and photoelectrons is quantified as a sample is irradiated with X-Rays and this emission is proportional to the X-Ray absorption coefficient of the sample. This measurement is inherently surface sensitive (several nm) because only the Auger/photoelectrons generated near the sample surface contribute to the signal. As such, measurements in TEY mode can reveal critical information like the catalyst’s surface electronic structure, nature of adsorbed reactants49, and structure of solvent molecules at the electrode/electrolyte interface50. While technically challenging, experiments using soft X-Rays and TEY mode can provide complementary information to that from hard X-Rays. Beyond surface analysis, there is a wealth of information that can be derived from soft X-Ray analysis and we point interested readers to several recent reviews on this topic48,51.
Limitations in temporal resolution
The structure of an electrochemical catalyst may change dynamically52, and these changes would be averaged out by a standard XAS measurement which can take from 1 to 15 min to record. Therefore, as a best practice, time resolved XAS data are needed to more comprehensively observe these dynamic changes occurring to the catalyst. In a recent study using continuous scans, which allowed for scans of 90 s, the change from nickel hydroxide to oxyhydroxide was resolved as a voltage jump from 0.7 to 0.8 V vs Ag/AgCl was performed53. Even faster scans are being able to be made with quick X-ray absorption spectroscopy (QXAS)37 where hundreds of scans can be taken in minutes. This allows for temporal resolution and, in some cases, where bubble formation is a problem having hundreds of scans can average out “noisy” data. Lin et al. 14 probed the change in the chemical composition of CuOx after discovering that using redox shuttling (e.g. pulsed potentials) was more selective for ethanol products from CO2 reduction than the conventional chronoamperometry (CA) methods. Using QXAS under operando conditions it was found that redox shuttling led to the CuOx being half Cu and half Cu(I) for the entire experiment, whereas CA treatment led to a complete conversion of CuOx to Cu.
Limiting the energy range of the scan can also allow for increased temporal resolution, such as only collecting XANES data rather than collecting EXAFS data, which require longer scans. Pasquini et al.54 used a novel approach of collecting single-energy XAS to track the oxidation state changes during OER with millisecond time resolution. They were successful in tracking the oxidation state of Co during cyclic voltammetry. This approach is a valid one for tracking metal oxidation state changes under potential control.
Beam damage of XAS samples
Using XAS requires a high-energy beam to come in contact directly with your sample, and therefore the possibility of beam damage should be considered as a best practice. It was shown in Diklic et al. 33 that even having too high a flux could cause the sample to heat up where the beam is located, which would change the electrocatalytic system. The X-ray beam may also cause radiolysis of the solvent or electrolyte. The generation of the radicals could impact the catalysis being studied55,56. Therefore, one must consider that having the beam on during the entire duration of an experiment might lead to changes. One method to test if the beam is causing damage to the sample is to move the beam to a different spot on the sample and check if the spectra match. Additionally, many high-flux beamlines have the capability of adding attenuators into the beam to decrease the photon flux on the sample. These can be used to mitigate beam damage to the sample. For long electrochemical experiments, one can run their experiment with the beam on during the experiment's entirety and run another with the beam only on at the end of your experiment. This should show if long-erm beam exposure caused changes in your sample. Finally, if the system is reversible41 one can pulse the potential and see if the sample goes back to the original state. If the sample looks the same, one can know that minimal beam damage has occurred. Taking this into consideration should help reduce the effects caused by the beam and provide a way of testing for beam damage.
Theoretical approaches to XAS in catalysis
Operando characterization usually reports an averaged signal over all species and sites available at the catalytic interface. However, it is sometimes a minority set of sites that govern the catalytic outcome, whereas a fraction of the surface might contain spectators or be permanently occupied by strong adsorbing reaction intermediates or reactants. For example, signals from CO adsorbates on a CO2R catalyst may be dominated by strongly CO-binding sites that do not participate in CO-CO coupling en route to C2 products. In some cases, the catalyst may be dynamic, where the structure and stoichiometry of the catalyst change on the timescale of the reaction. In addition, a diversity of material sizes, shapes, oxidation states and compositions can exist in parallel, and experimental XANES and EXAFS data will measure the average of the ensemble. This average effect on overall species present in the system is one of the inherent challenges in operando XAS data analysis, which may obscure the contributions of active sites. It becomes particularly pronounced in heterogeneous or evolving systems, such as catalysts under reaction conditions, where multiple structural states may coexist and dynamically interconvert. Furthermore, if the operando measurement is relatively slow, compared to interfacial rearrangement, the signal will be smeared over all accessible structures and not correspond to any one of them exactly.
However, theory needs to make its own strides to meet this challenge. Two approaches with contrasting philosophies can be mentioned for theoretical operando spectroscopy. One is to first identify the structure of the interface corresponding to the reaction conditions in the most realistic possible way, and then compute the spectra and compare them to the experiment. This approach requires simulating the phase diagram of the catalyst and taking into account the ensembles of the catalyst states and their dynamics. The second approach bypasses the complex simulations of the realistic interface and instead uses a “basis set” of states, such as typical oxidation states of a metal, or coordination environments of an atom. These are used to fit experimental spectra and their evolution in reaction conditions and qualitatively interpret the catalyst states.
Theoretical approaches to EXAFS
EXAFS arises when a core electron transitions into a continuum state, followed by the scattering of the photoelectron wave from a neighbouring atom and subsequent return to the absorbing atom57. The scattered photoelectrons give rise to constructive or destructive interference, providing information on bond distances and coordination numbers. A challenge in EXAFS is that it provides averaged information from all absorbing atoms in the system ensemble58. One should keep in mind that the average configuration as measured is not necessarily the most active one and should take into account potential diversity in, for example, catalyst particle sizes, shapes, and structural disorder, particularly at the catalyst surface of the bulk region59. To tackle these limitations, bottom-up approaches have been effective. These involve computing the EXAFS spectral contribution from each individual atom and conducting temporal averaging. An example investigated dendrimer-encapsulated nanoparticles and yielded insights into the nanoparticle structure by incrementally introducing structural disorder through interactions with their surface ligands60. The nanoparticle structures from AIMD simulations confirmed the intricate experimental setups by cross-validating theoretical EXAFS with corresponding experimental signals from analogous particle systems. Additionally, they revealed the limitations of conventional EXAFS fitting models in accurately portraying strongly and asymmetrically disordered systems, as discussed above. This was demonstrated by a direct comparison of AIMD-derived PDF with Gaussian functions describing the first-shell Au–Au interactions extracted from experimental EXAFS fits.
Operando EXAFS is a rigorous method for identifying the fine structure of reconstruction-derived components, utilizing the highly sensitive shifts in bond lengths resulting from the oxidation behaviour of metal species. It can be employed, for example, in the development and characterization of a novel catalyst for the oxygen evolution reaction (OER). A notable example is hydrous iridium oxide, a preeminent OER candidate material, which is synthesized through the electrochemical oxidation of Ir61. This material exhibits a highly active amorphous structure, but its structural complexity makes it exceptionally challenging to understand at the atomic scale and to elucidate the mechanisms occurring at local structural levels. Operando EXAFS, combining experimental precision with theoretical modelling, is highly effective in probing local structures within the first coordination shell and extending up to approximately 6 Å. This approach has successfully revealed unique structural features, such as elongated Ir–O bond lengths compared to crystalline counterparts and the formation of coordinatively unsaturated Ir and O species. Moreover, the sensitivity of EXAFS has enabled the identification of dynamic transformations of these local structures under operational conditions, including deprotonation processes, Ir vacancy formation, and bond contractions. Such insights provide an atomistic understanding of the activity-stability trade-off inherent to these amorphous materials.
However, the effectiveness of in-situ*/operando* EXAFS diminishes for longer-range structural contributions beyond the first coordination shell (extending up to 10 Å). This limitation arises due to the exponential increase in parameters required to comprehensively describe the more complex structural features at greater distances62. While the integration of theoretical models enhances the interpretation of local structures, accurately reconstructing atomic configurations from EXAFS data in the extended range remains a significant challenge. This underscores the need for advanced theoretical and computational approaches to address the complexities of long-range structural analysis. On the experimental side, limitations in the information provided can be addressed through complementary methods. For example, one can carry out pair distribution function (PDF) analysis which provides information on a greater length scale without requiring crystallinity63. This has, for example, been shown to elucidate structural changes within a Ni-based catalyst during water electrolysis64. This can be further combined with small-angle X-ray scattering (SAXS) which can provide information on average particle sizes65. X-ray diffraction can also be leveraged to pinpoint changes in long-range crystalline structure with a common example entailing voltage-dependent transformation of oxide to metal transitions66.
In addition to bottom-up computational approaches, multivariate curve resolution techniques, such as Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS), have been instrumental in decomposing complex EXAFS spectra into their constituent components67,68. MCR-ALS provides a robust framework for resolving overlapping spectral contributions from distinct atomic environments within a catalyst. By leveraging datasets acquired under varying conditions (e.g., temperature, pressure, or time), it identifies individual spectral components and their relative contributions, allowing researchers to correlate specific structural motifs with catalytic activity. However, the successful application of MCR-ALS relies on diverse and high-quality datasets to resolve individual contributions effectively. Its interpretation is subject to certain assumptions, such as spectral linearity and independence, and can be challenging in systems with significant spectral overlap or noise.
While MCR-ALS is a powerful tool for dissecting complex operando XAS datasets, its effectiveness is inherently tied to the quality and diversity of the input data, as well as the complexity of the system under study. Therefore, its application should be complemented by other analytical and computational approaches to fully unravel the intricacies of dynamic systems.
Theoretical approaches to XANES
Perhaps the most widely used approach to XANES calculations uses the finite difference method (FDM) and the Hedin−Lundqvist exchange-correlation potential implemented in the FDM near-edge structure (FDMNES) ab initio package69. FDMNES works in real space and builds a cluster of a specified radius (such as 5–6 Å) around the absorbing atoms of interest, after which the absorption spectra in the specified energy range are calculated for each nonequivalent absorbing atom and then combined. FEFF9 is also used, and it is based on the real-space Green’s function (RSGF) approach70. Another popular code is called OCEAN, which is based on ground-state DFT (done with QuantumESPRESSO or ABINIT) and the Bethe-Salpeter equation (BSE) solved with the programme NBSE.
Recently, XANES spectra were computed in electrochemical conditions of electrode charging and the presence of implicit electrolyte71. The approach was used to investigate the electrochemical stability of various IrO2 interfaces under realistic conditions of oxygen evolution (OER) reaction. The comparison of computed O K-edge XANES spectra to the experimental spectra suggested that electron-deficient surface oxygen species form at the interface, and surface hydroxyl groups undergo progressive oxidation at larger potentials. XANES spectra can also be computed using the FEFF code to, for example, model the effects of O and OH adsorption on Pt clusters72.
As XANES, like EXAFS, is an ensemble-averaged technique, interpreting the XANES signal from in-situ/operando experiments can be nontrivial for dynamic interfaces73. One approach to tackle the problem is to use ab initio molecular dynamics (AIMD) simulations, to get a distribution of thermally accessible cluster geometries, starting from the most stable structure. If the system visits sufficient space of local atomic environments during the dynamics, good agreement with experimental data can result. One can obtain insights regarding the oxidation state of the metal or the typical coordination of the catalysts. However, beyond geometries, catalysts can visit a wide variety of adsorbate coverages that may not be fully captured with traditional AIMD methods. Advanced approaches such as grand-canonical AIMD can provide access to these states under appropriate conditions. This information is essential for constructing a comprehensive catalytic mechanism and not readily accessible, thereby limiting the depth of attainable insights.
It has been shown that clusters do not stay in the vicinity of the global minimum structure, and often populate many distinct minima on the free energy surface, which can differ in geometry, reagent coverage, and interface with the support. One way to produce a purely computational ensemble-averaged spectrum is, to sum up the computed XANES spectra of all thermally-accessible isomers of the system, weighted by their population probability. However, including kinetic effects on interfacial isomerization is needed in low temperature systems like electrocatalysts where the system dynamics is governed by the kinetics74. Operando spectra, can be fitted using such a realistic ensemble of states, which may include a variety of the metal atom environments, bound adsorbates, other metals and atoms of the support, and sintering effects. The resultant spectra can well match the experimental data75,76. The main caveat is that fitting to a large set of computed XANES almost certainly results in overfitting.
This approach is less ambiguous (and overfitting minimized) for well-defined systems that feature fewer degrees of freedom, such as single-atom catalysts. For example, Cu–N–C single-atom catalysts with a well-defined initial Cu2+–N4 coordination were used for ORR (Fig. 4)77. Operando XANES combined with theoretical interpretation revealed that at the applied voltage, the structure of the active site changes to Cu–N3 and then to HO–Cu–N2, while the Cu is reduced from Cu2+ to Cu+. The Cu+–N3 site was then shown by DFT to feature a lower free energy profile for the ORR than the initial Cu2+–N4.
Fig. 4: Illustration of SAC and theoretical approach toward its investigation.
figure 4
a Schematic representation of the Cu-N-C SAC. (b) comparison between Cu K-edge XANES spectra recorded at 0.50 V (black line) and theoretical Cu-N3 XANES spectra (red line). (c) Free energy diagram of the ORR on Cu-N4 and Cu-N3 sites. Adapted with permission from Ref. 77. Copyright 2021 American Chemical Society.
Full size image
Machine learning approaches
The direct application of machine learning (ML) strategies can be used to simulate the catalytic interfaces to link to in-situ*/operando* experimental data78,79. An example of this entails a neural network (NN) method that was utilized to build the relationship between the XANES spectra and structural parameters for copper oxide cluster systems80. The average particle size and the oxidation state of metal can be determined from in-situ*/operando* XANES spectra, as the NN is trained for both metallic and oxide forms. The key descriptor in the approach is the coordination number of an atom and the method can be applied to gauge the changes in metal particle coordination during the reaction. The limitations of the method include the inability to distinguish clusters that feature the same coordination numbers, and challenges in treating mixed oxidation states in the same system, especially with partially oxidized Cu clusters. In particular, the isomers of mixed oxidation states of Cu with different numbers of coordinating oxygen atoms varies significantly and limit the applicability.
A recently developed ML method determined the chemical states and compositions of Cu species in supported Cu and CuPd clusters used for propane oxidative dehydrogenation (Fig. 5)68. A notable experimental observation is that the selectivity of the reaction changes from propylene oxide to propylene as the temperature increases. Because of the coexistence of mixed states of Cu in the catalyst during the reaction, MCR-ALS analysis, a method useful for isolating individual contributions to spectra without requiring a set of standards67,74, was applied to separate the spectra of different types of metal atoms from the operando XANES spectrum. Next, an NN was trained to obtain structural information, with nearest neighbour metal Cu-Pd/Cu coordination numbers (CNs) chosen as structural descriptors. The training set of XANES spectra was produced in silico, producing good fits and as a result, a correlation between descriptors of the atoms in the clusters and the catalyst activity and selectivity was established. Cu, rather than Pd, was found to be the source of the enhancement of catalytic activity and selectivity toward C3 products. The analysis also ascribed the enhancement of the catalytic performance with temperature to the presence of Cu+ species structures with high Cu−Cu and low Cu−Pd coordination numbers. Hence, some level of mechanistic understanding was achieved, though still without specific molecular detail.
Fig. 5: ML approaches to EXAFS modelling.
figure 5
a Examples of structures produced for theoretical training set for XANES, featuring diverse local coordinaitons and particle size. b True values of CN(Cu–Cu), CN(Cu–Pd) vs predicted values of CN(Cu–Cu), CN(Cu–Pd) for the first shell, CN(Cu–Cu), and CN(Cu–Pd), obtained using the CuxPd2–xO neural network model. Adapted with permission from refs. 68,132. Copyright 2017, 2021 American Chemical Society.
Full size image
In our view, ML-based strategies are mathematically rigorous and provide a qualitative idea of the relevant active sites, such as a metal in a certain oxidation state. However, if the intention is to connect this information to the reaction mechanism and computed reaction profiles, rates, and a kinetic model, this information is insufficient, and a more granular approach is necessary. While using the explicit computed standards can result in overfitting of the experimental in-situ*/operando* XANES due to similarities in the spectral signatures of various atoms, the identified present structures can serve as starting point for mechanistic calculations.
Vibrational spectroscopy
Vibrational spectroscopy is a frequently used technique to support the interpretation of electrochemical catalytic performance data. Here, UV/Visible/IR photons are used to measure molecular vibrations of the system. The presence and changes of the vibrational modes of the solvent, catalyst, and reactants/intermediates/products are detected and subsequently interpreted to yield mechanistic insights into the catalytic system. These experiments can be readily carried out under in-situ*/operando* conditions. In the case of Raman spectroscopy, the solvent signals (e.g. water bands) often do not disturb those of the catalyst or reaction intermediates.
One unique aspect is that the surface sensitivity of certain catalyst surfaces can be increased so that signals arising from the surface-bound intermediates are amplified. This is a powerful tool in catalysis research as it enables one to selectively view the reaction process on the catalyst surfaces via minimizing spectral contributions from the bulk system. Surface enhancement can be accomplished using metal electrodes such as copper, gold, and silver for surface-enhanced Raman spectroscopy (SERS)81 or surface-enhanced infrared absorption (SEIRA)82. From a pure spectroscopic point of view these systems are the easiest to handle as the metal catalysts itself does not give rise to Raman/IR signals. Investigations in this context include those of CO2R83,84 ORR85 and HER86 catalysts. However, the direct identity of the surface-adsorbed molecule is not always clear from a spectrum as a number of species and binding motifs may often give rise to vibrational modes in a similar frequency range.
Further, in-situ vibrational spectroscopy is not limited to the identification of reaction intermediates but can also give information about transient changes of the catalyst itself such as redox changes87 or oxide to metal transitions66. This is often much easier to detect but, particularly in the case of carbon-based catalysts, may be falsely assigned to a reaction intermediate. Thus, the biggest challenge within in-situ*/operando* vibrational spectroscopy arises from the analysis of the obtained spectra because the resultant data often allow several routes of interpretation. To this end, the next sub-sections will discuss best practices in arriving at the correct interpretation of in-situ*/operando* data.
Methodology
Electrochemical vibrational spectroscopy is conducted in tailored spectroelectrochemical cells that allow coupling of laser light or IR light to the working electrode for spectroscopic measurements under an applied voltage88. EC-Raman/SERS spectroscopy is carried out in confocal mode, using a microscope objective to focus the probing laser light onto the working electrode with the catalyst. In EC-SEIRA/IR spectroscopy, the measurement is carried out in attenuated total reflection mode. In this mode, an IR-active waveguide is used as support for the working electrode, and the IR light coupled in from the back side illuminates the electrode surface near-area.
The cell design beyond the spectral windows also dramatically impacts the processes at the catalyst/electrode surface, setting the frame for the type of mechanistic information extractable from the collected set of data89. Key reaction parameters, such as concentration (limits) of the substrate and mass transport, must be considered as these factors influence mechanistic pathways, and thus, data obtained from different cells cannot be directly compared. H-cells are more accessible for spectroelectrochemical measurements, but the reaction is restricted to the concentration limits of the substrate in the aqueous solution, e.g., 33 mM for CO2. Spectroelectrochemical cells accommodating gas diffusion electrodes (GDEs) are considered more practical systems as they can operate at high current densities and substrate concentrations close to industrially relevant conditions. While EC-Raman/SERS spectroscopy has been successfully employed using gas diffusion electrodes66,90, due to the higher technical requirements for conducting EC-SEIRA spectroscopy, the latter is not yet routine on such electrodes. However, EC-SEIRA spectroscopy has been carried out in electrochemical flow cells that can help alleviate mass transport limitations25.
Required controls
We present a workflow of best practice required controls in Fig. 6 that should accompany in-situ*/operando* vibrational spectroscopic experiments in order to attain a reliable interpretation of a catalytic mechanism. At first level, a vibrational peak can only be assigned to (attributed to) a reaction intermediate if it occurs only under reaction conditions and disappears after going back to the resting state (e.g. at non-catalytic potential). This would rule out the accumulation of impurities or surface contamination and poisoning species. However, spectra from spectator species may also rise and fall with an applied potential. It is preferred to check the differences between two defined external potentials (e.g. pre-catalytic and catalytic) rather than comparing them to the open circuit potential. The latter is sensitive to irreversible side processes and might change before and after catalysis. An example here lies in the investigation of MoSx catalysts used for hydrogen evolution. A recent work observed the rise of a band at 2530 cm-1 under cathodic potentials close to the HER onset potential91. The band was assigned to the S-H stretching vibration based on a comparison with literature, DFT calculations and isotopic H/D substitution. The peak disappeared when going back to anodic potential confirming it as a reaction intermediate rather than a vibrational mode arising from a restructured catalyst or surface poison.
Fig. 6: Measurement guide for vibrational spectroscopy.
figure 6
Checklist for measuring and interpreting in-situ*/operando* vibrational spectra.
Full size image
In addition, the peak corresponding to the reaction intermediate must not appear in the absence of substrate molecules, which can readily be checked for if the reactant is not the solvent itself (i.e. water in HER/OER). We recommend such spectra be added to the supplementary information of studies, particularly if there are numerous other species with vibrational modes that may show up in the in-situ*/operando* spectra (e.g. a carbon-based catalyst, electrolyte species…). In the context of CO2R, local changes in pH when protons are consumed under reaction conditions are accompanied by changes in the CO32−/HCO3− or PO43−/HPO42− ratios. This change will correspondingly show up in experimental spectra92. While this is useful for estimating pH93,94,95, these spectral features should not be mistaken for reaction intermediates. Thus, cross-referencing spectra with those of the electrolyte components should always be carried out.
With Raman spectroscopy, a particular challenge also lies in distinguishing between bands arising from reaction intermediates vs. those from the catalyst itself when carbon-based catalysts are used. These systems show no surface enhancement and the carbon matrix itself gives intense Raman signals between 1300 and 1700 cm−1 (graphite D and G band) as well as in the region between 2600 and 2800 cm−1 (G´ band)96. Regarding the catalytic sites, changes in the symmetry, coordination or oxidation state of the catalyst might lead to the rise or attenuation of vibrational bands close to the position of the expected substrate intermediate peak. For example, the reversible attenuation of a band at 593 cm−1 at cathodic potentials for ORR in a FePc/C catalyst was assigned to the loss of planarity in the system and not a reaction intermediate as it was visible in the absence of oxygen reactants97. It should be noted that also non reversible peaks and peaks that arise in the absence of the substrate give valuable information on the catalytic reaction regarding aspects of surface poisoning and catalyst transformation, respectively.
In the case where the signal-to-noise ratio (S/N) is below 5, the presence of the peak across several measurements would help to distinguish it from noise. Further, one must also be careful of over-interpretation in assigning peaks to a proposed reaction pathway. Spectator species and reaction intermediates in pathways responsible for only a small fraction of the total reactivity may also be built up on the catalytic surface and detected. Here, correlating the in-situ*/operando* spectra with theory as a best practice helps strengthen interpretation and the eventual elucidation of the major catalytic route/active sites that lead to the observed activity (see below).
A further requirement to interpret and assign peaks to proposed surface intermediates is to match their frequencies to what is found in the literature. The assignment should be based on a comprehensive and convincing work that deduced the identity of the band, rather than one that simply made a similar assignment. The accuracy of this comparison depends highly on the type of intermediate. For example, *CO is a readily seen species in CO2 reduction with a well-defined frequency range from 1800 to 2100 cm−1, but more complex, multi-atom species are less clear. A thorough literature comparison can also be used to exclude certain intermediates.
Isotope labelling
A clear assignment of a vibrational peak can be achieved by using isotope-labelled substrate molecules (e.g. 13CO2, D2O, H218O, etc.). Isotope shifts need to be in accordance with the expected peak shift from theory for this vibration, and, for stretching vibrations, typically fall between 20 and 70 cm-1. The best way to present the data is by illustrating the difference spectra from the standard and isotope-labelled experiments. Note that this must lead to a negative and a corresponding positive band and that one isolated positive or negative band does not represent an isotope shift. Proper difference spectroscopy requires that vibrational peaks not involved in the isotope shift, stay the same. This is sometimes very difficult to achieve and as a result, many publications show both spectra without subtraction or only limited parts of their spectra. As an example, isotope labelling of NiOx catalysts itself also has been used to provide into their OER reaction mechanism. For example, a shift in the region between 800 and 1000 cm−1 could be identified, which changed during OER when the 16O in the NiOxHx catalysts was replaced by 18O98. The broad peak was assigned to the O–O stretching vibration of a NiOO− intermediate based on the 50 cm−1 shift expected for this vibration. The peak was also present in the resting state and plausibly assigned to pre-OER lattice oxygen that participated in the catalytic cycle. This was also used in CO2R studies on Cu electrodes, in which Raman spectra using 12CO2/13CO2 and H2O/D2O were overlaid over top of one another and used to determine which species stemmed from the CO2 reactant as opposed to electrolyte and solvent species99. For reactions involving two separate reactants, such as electrochemical C–N bond formation from CO2 and N2/NO3−100, spectra can be acquired using separate 12C/14N, 13C/14N, 12C/15N, 13C/15N reactant feeds. This was recently carried out using IR spectroscopy to investigate the mechanism of urea formation from CO2 and N2 co-reduction in which a band at 1449 cm−1 was attributed to a C–N stretching vibration101.
Moreover, ab initio modeling can simulate isotope effects by predicting vibrational frequencies and isotopic shifts, providing a theoretical framework for comparison with experimental data102,103. These simulations help validate the assignment of vibrational peaks and offer deeper insights into the reaction mechanisms by correlating experimental shifts with specific intermediates or molecular structures.
Finally, the appearance of an intermediate only suggests that this intermediate is present in the catalytic cycle and does not exclude other reaction pathways. Steady-state spectra are more sensitive to long-lived rate-limiting reaction intermediates and very short-lived intermediates corresponding to an alternate, highly effective reaction pathway might not be visible. A positive correlation between catalytic activity, proven by higher currents and more product, and the intensity of the intermediate peak is, therefore, a strong hint that this intermediate is catalytically relevant. To this end, time-resolved measurements can provide complementary information to steady-state experiments. For example, time-resolved Raman has been used to detect *CO species on oxide-derived Cu shortly after the application of a reducing potential that was not present under steady-state conditions104,105, or detecting optimal *OH and *CO co-adsorption during dynamic pulsed conditions on oxide-derived Cu106.
Complementary analysis
To solidify the peak assignment, further independent information on the system has to be obtained (shall-measure). Typically, this is done by comparison with spectra of the predicted species derived from density functional theory (DFT) simulations. Peak positions must also clearly be visible. Asymmetric peaks likely represent a mixture of vibrations and should be treated as such. Complementary in-situ*/operando* techniques like UV-Vis absorption and ECMS can also be used to provide evidence for material transformation and surface intermediates, respectively. If no peak assignment can be derived from the literature, complementary analysis is mandatory to assign an observed vibrational band.
Theoretical approaches to IR and Raman spectroscopy in catalysis
From the computational point of view, it is imperative to have a reliable catalyst model for calculations of vibrational spectra. In order to produce the most realistic model, grand canonical sampling is used to optimize the structure and the coverage with and placement of the relevant adsorbate(s). The coverage should be governed by the chemical potential of the adsorbate which is influenced by factors like partial pressure and local pH. Ultimately, the starting point for the modelling is a computational phase diagram.
Concerning the best practices of IR modelling to link to in-situ*/operando* spectra, one approach is to use phonon calculations based on the relevant structures from the phase diagram. When multiple isomers of the system are thermally accessible, the IR spectrum of all these species should be computed and an ensemble-averaged spectrum can be obtained via a weighted average. The limitation here is the lack of anharmonic effects that would red-shift the soft modes in the system. In addition, in situations where the solid interfaces with the liquid phase, such as in electrocatalysis, one needs to average over the interfacial dynamics of the solvent. For that, the systems need to have adequate solvation and be treated with a molecular dynamics simulation to accumulate sufficient statistics of the solvent (and electrolyte) configurations. The convergence of such simulations can be judged by the stabilization of the system energy and of the radial distribution functions of various atom pairs, such as the adsorbate with the oxygen atom of the water. For the converged trajectories, the spectra can be obtained from the velocity autocorrelation functions, as follows107: All vibrational spectra calculated from AIMD are based on the Fourier transform of certain autocorrelation functions. The sum of all the correlation functions of a molecule is Fourier transformed to get the power spectrum of that molecule.
These established approaches were recently augmented with ML108,109. Using a dataset of precomputed IR spectra for an adsorbed CO on various sites on a Pt surface, an NN was trained to predict surface morphology from the IR spectrum. The approach is powerful because it does not require a lengthy search for the interfacial structure to begin with. Instead, using the fact that the adsorption site and adsorption energy, IR frequency, and surface structure are coupled, morphology can be derived from the spectrum110. Using synthetic IR spectra of CO on platinum, a multinomial regression via NN was implemented to learn probability distribution functions (pdfs) that describe adsorption sites. These pdfs were used to infer experimental behaviour, including (1) CO binding on different types of sites (atop, bridge, threefold, and fourfold) and (2) coordination environments (planes and edges). Figure 7a shows the spread of the CO stretching frequencies on these sites, computed with DFT. The algorithm included the removal of outliers from the training set, and the coverage was taken into account using a scaling factor for the frequencies. Figure 7b shows the outcome-predicted microsite structures from NN.
Fig. 7: ML approach towards vibrational spectroscopy data interpretation.
figure 7
a DFT frequencies and corresponding intensities for an adsorbed CO on different sites. Different symbols correspond to sites with different GCN. The inset is a zoomed-in image of the Pt–CO frequencies. b Inverse machine learning model predictions for types and environments of sites. Spectra and predictions are for CO on Pt(111) at 0.5 ML (green solid line and bars with forward slashes), Pt(111) at 0.17 ML (blue dotted line and bars with backslashes), Pt(110) at 1 ML (yellow dashed-dotted line and bars with horizontal lines), and Au/Pt core/shell nanoparticles at low coverage (purple dashed line and solid bars). The black error bars indicate the 95% prediction region. Adapted from ref. 108. Nature Communications (2020) under a CC BY 4.0 license.
Full size image
The approach is generally very powerful, but one main caveat is that more than the first and second coordination spheres of the binding site may be important for catalysis. The link to the specific molecular structure of the larger site and eventual mechanism can be limited for this reason. As with any ML method, the quality of the NN depends on the amount and quality of the data used for its training, and the computational efficiency of the approach depends on the time needed to obtain the computed dataset and to train the NN. While ML methods used for the analysis of vibrational spectroscopy are just emerging and they are not yet widely established methods, their use as a best practice reference is expected to grow as this direction correspondingly grows.
Differential electrochemical mass spectrometry (DEMS)
Electrochemical mass spectrometry is a powerful technique that can identify reactants, intermediates, and products en route to elucidating the mechanism of a reaction111. Here, the mass of various species is the primary characteristic used for their identification. The most common configuration used is differential electrochemical mass spectrometry (DEMS) and will be the focus of this perspective, though it is not the only one to be employed. In such setups, volatile species of interest (e.g. products and surface intermediates) can desorb from a catalyst and transfer directly to or diffuse to a mass spectrometer chamber where they are ionized and detected based on the mass/charge (m/z) ratio of the species or their fragments. This is a particularly convenient method to distinguish between potential reaction pathways and also to incorporate isotope labelled reactants for validation. This technique is particularly sensitive and consequently, minor species can be detected and time-resolved measurements (with ~1 s temporal resolution) can readily be carried out10. We note that DEMS is an online analytical method and because it is not directly observing intermediates on the surface but rather species that desorb, it is not strictly classified as in-situ or operando. However, it is still an important technique used in the field that is also prone to measurement errors and improper interpretation and, thus we dedicate a section of the text to discuss DEMS.
Cells of DEMS
The core component of the DEMS cell is the porous hydrophobic membrane assembled on a steel frit as the inlet to a mass spectrometer through a mild vacuum and high vacuum chamber (Fig. 8a)112. The porous membrane only allows the permeation of gas while rejecting the liquid electrolyte. When the porous membrane is set close to the surface of the working electrode, the generated volatile species are rapidly introduced to the MS because of the pressure difference between the two sides of the membrane. In a conventional cell of DEMS (Fig. 8a), a layer of Au film is sputtered on the porous membrane, and the catalyst is dispersed on top. Alternatively, through a thin layer cell, the working electrode and the porous membrane are separated by a thin layer of electrolyte with thickness of 50–100 μm and the produced volatile species diffuse across the thin electrolyte layer to reach the porous membrane.
Fig. 8: Overview of DEMS experiments.
figure 8
Illustration of typical DEMS measurements and their main considerations. a Structure of DEMS. b Conventional cell. c Dual thin layer cell. d Probe cell. e Typical data obtained from DEMS. Note, such cells represent only those capable of DEMS measurements and not parallel spectroscopic analysis.
Full size image
The small electrolyte volume also results in a rapid depletion of the reactants but also introduces mass transport limitations. Thus, the use of a flow cell geometry is important under high current densities (Fig. 8c). To decrease the large iR drop due to the long and narrow channels for the electrolyte, two counter electrodes are placed at both the inlet and the outlet of the electrolyte. The porous membrane is also used as a probe set near the working electrode, with the distance digitally controlled (Fig. 8d).
Applications of DEMS
Since carbon and nitrogen have a large family of volatile compounds, DEMS is very suitable for the investigation of carbon- and nitrogen-involved reactions like CO2R or nitrate reduction in addition to water electrolysis113,114,115,116,117. For example, during nitrate reduction to ammonia over Cu/Cu2O, the results from DEMS show a series of volatile products with the m/z value of 46, 30, 33, and 17, which are attributed to NO2, NO, NH2OH, and NH3, respectively118. Accordingly, the reaction pathway was proposed to proceed through the reduction from *NO3 to *NO2 then to *NO. In subsequent steps *NO is further reduced and hydrogenated with *H to successively form *HNO, *H2NO, *H2NOH, and finally *NH3119.
Through designing the isotope-labelled DEMS experiments, different reaction mechanisms can be clearly distinguished. To this end, a lattice-oxygen-mediated mechanism (LOM) is recently proposed, and experimentally verified through 18O-labelled DEMS measurements. Here, metal oxide electrocatalysts were first labelled by 18O via ex-situ electrolysis in H218O and then investigated via DEMS in an H216O electrolyte. Under OER conditions, the appearance of 36O2 indicated that 18O from the catalyst lattice became part of the O2 product via the LOM mechanism120,121. Owing to the high time resolution (about 10 ms per data point)122, a time-dependent behaviour can be resolved with DEMS. For example, differing OER mechanisms can be monitored in real-time during a cyclic voltammogram to determine the relative contributions of the LOM and conventional adsorbate evolution mechanisms (AEM) as a function of time and potential123.
Experimental parameters
The MS ion current is highly dependent on the vacuum degree of the MS chamber and any perturbation will influence the recorded MS signal. This phenomenon is frequently seen in gas evolution reactions in a thin layer cell, especially under large overpotentials124,125. The generated gas species will push the electrolyte in the thin layer away. On the surface of the porous membrane, the change from water to gas alters the pressure difference between both sides of the membrane, leading to the change in the vacuum degree in the MS. The gas molecules can also diffuse and distribute throughout the whole electrolyte, leading to a tailing and broadening of the MS signal122. Increasing the flow rate of the electrolyte can alleviate these issues by carrying the produced gas out of the thin layer region. However, a fast flow rate will decrease the transfer efficiency of the volatile species to the MS, thus lowering the intensity of the MS signal. If the effects of the generated gas cannot be completely eliminated by adjusting the flow rate, a conventional or probe cell may be more appropriate.
Another important feature of DEMS is the inevitable permeation of water from the aqueous electrolyte into the MS. Water molecules in the ionization chamber can react with the hot filament to produce H2 and O2. These species may contribute to the resulting MS spectra, complicating data analysis, especially for HER/OER studies. Other volatile species in the system or the residual species in the ionization chamber of the MS may also cause similar interference. To avoid interference, the filament of the MS should be preheated with the loading of the electrocatalytic system, but no bias before each experiment. The preheating period should be long enough (>30 min) until the baseline of the relevant species is stable126. At this moment, the interfered species permeation and reactions achieve balance, and the DEMS tests are ready to start. Another solution to eliminate the effect of water is to install a cold trap between the cell and the vacuum chamber so that the water vapour is trapped before it can enter the ionization chamber127.
The results of MS can be also influenced by factors such as the electrolyte and the counter electrode. The electrolyte used in a DEMS test should not produce the same species as the electrode reaction. For example, if CO2 is the targeted species of DEMS, KHCO3 is not suitable as it HCO3- exists in equilibrium with CO2, and therefore K2SO4 may be a more appropriate choice to avoid false positives. Further, the electrolyte should not react with the reactants128. Gases like CO2 and NO2 will react with commonly used KOH and, therefore may be undetected or underestimated. In this case, a conventional cell rather than one employing the thin layer geometry is the better choice. Finally, reactions on the counter electrodes may lead to false positives in the measurement. For example, OER or HER at the counter may give signals with the m/z of 32 and 2, respectively.
Data analysis
The results from DEMS only contain the information about m/z of the ionized species instead of the composition of the products, thereby requiring a layer of interpretation to detect them. The desorbed intermediates must be stable in the system for short durations and cannot be molecules that are extremely reactive. In addition, the ionization of the volatile species may produce many fragments with different m/z values. Accordingly, complementary characterization can validate the identification of proposed products and intermediates. Species with the same m/z may have different compositions like CO and N2 with the same m/z of 28. These can both be intermediates in the electrosynthesis of urea from CO2 and NO3−, for example. As a best practice, their identity can readily be determined by using isotope-labelled reactants like 15NO3- and 13CO2 as reactants. The resultant data would distinguish between 13CO (m/z = 29) and 15N2 (m/z = 30).
When elemental composition is confirmed, uncertainty still remains regarding the structure of the detected species due to the existence of the isomers. For example, species with m/z = 33 may be intermediates in the reduction of NO3− to NH3 and can stem from either *NH2OH129 or *ONH3130 depending on the reaction pathway. Elucidating the exact identity, in this case, would require complementary analysis like vibrational spectroscopy or DFT modelling. Even when the structure of the intermediate is resolved, DEMS cannot distinguish its adsorption motif on the catalytic site. For example, if CO is detected (m/z = 28) during the reduction of CO2, DEMS cannot distinguish whether *CO is adsorbed or in a bridge-type mode on the catalyst surface. Finally, it may be difficult to determine the origin of small fragments. N, with m/z = 14, can originate from N2 or from NH3, both of which may co-exist during the NO3- reduction to NH3.
We also stress that when it comes to directly linking DEMS data to a proposed mechanism, one should avoid over-interpretation. If an intermediate (ex. NO in the NO3− reduction reaction) is not detected, it may still be part of a catalytic mechanism, though adsorbed too strongly to the catalyst surface or present for only very short durations before it is converted to a subsequent species in the pathway. Similarly, if a certain species is indeed detected, then a reaction mechanism featuring this species is likely to occur. However, multiple reaction mechanisms, particularly on heterogeneous catalysts that can have different types of active sites, may be occurring in parallel. This renders it difficult to unambiguously attribute a certain mechanism to the one responsible for the macroscopically observed activity. Resolving these challenges can be aided through complementary analysis like vibrational spectroscopy or theoretical modelling and is encouraged.
It should finally be noted that DEMS is an quasi in-situ measurement rather than an operando one. While reaction products can be quantitatively detected in a proper setup131, some parameters, such as mass transfer, and electrolyte type, may be different from the optimal catalytic conditions in which performance benchmarking was carried out. Since volatile species are removed from the electrode surface during the measurement, the reaction equilibrium may also change. Further, this technique offers insights into only the identity of species that desorb from the catalyst surface and not information about the catalyst itself. For this, a setup that integrates either vibrational spectroscopy or XAS would have to be used.
Outlook
We highlight five topics crucial for heterogeneous catalysis. As previously mentioned, these topics include reactor design, X-ray absorption spectroscopy, vibrational spectroscopy, electrochemical mass spectrometry, and theoretical modelling. Our objective in each section is to describe information conferred by the technique, common pitfalls, and best practices that are enabled by the current or near-future state-of-the-art (Table 1).
Table 1 Summary of information, pitfalls and best practices associated with each technique/topic
Full size table
The increasing utilization of in-situ and operando methods drives progress in heterogeneous catalysis, but a lack of thorough analysis exists on extracting insights from these technologies. In this Perspective, we propose a technology-dependent framework to interpret in-situ and operando data, minimizing common pitfalls and providing effective explanations. Integration of various techniques, including vibrational spectroscopy, XAS, and ECMS, together with theoretical modelling enables a comprehensive understanding of active sites, intermediate structures, and reaction microenvironments. This approach compensates for the individual limitations of each technique and provides layered information, ultimately unveiling the intricate mechanisms underlying catalytic reactions.
In this concluding section, we highlight several key directions that would propel the growth of the field.
Continuing advances in cell design
In a catalytic process, maintaining conditions consistent with actual operation poses a significant challenge. Due to the complexity of environmental factors, instrument/cell design needs to be more comprehensive to simulate the interaction of various catalysts. Addressing these challenges involves continuous technological innovation, diverse combinations of equipment, and a thorough consideration of cost-effectiveness. This will drive the development of advanced materials processing technologies, such as laser cutting, 3D printing, and nanomanufacturing, enabling a more accurate simulation and direct measurement of parameters such as temperature, pressure, and flow rate in real operating environments.
A further point here would be to design setups that can be integrated with multiple measurements. As stressed in this review, each measurement only provides a limited set of information on the catalytic process. As technology advances and the instrumentation allows for a lower spatial footprint and higher flexibility, we envision that techniques like DEMS, XAS and vibrational spectroscopy can be simultaneously carried out for a comprehensive analysis of catalyst structure, surface-bound intermediates and reaction products This, of course, would be challenging to realize and would require cells that feature multiple widows and measurement ports. Such integrated measurements could also be useful in determining whether a spectroscopic technique alters the catalyst or reaction process (ex. XAS beam damage modifying performance).
Time-resolved measurements
Detecting short-lived intermediates poses a challenge that hinges on achieving ultrafast time resolution. It is crucial to ensure that measurement techniques can capture these intermediates, especially given their low coverage. Innovative methods and advanced instrumentation play a pivotal role in addressing this challenge, allowing researchers to probe and analyse fleeting species with greater precision. As instrumentation technology continues to advance, we can anticipate the emergence of measurement tools with higher precision and sensitivity. This will aid in more accurately capturing the transient and momentary existence of intermediates in catalytic reactions, providing robust support for in-depth investigations into catalytic mechanisms.
Model catalytic systems
The effective utilization of computational methods stands as a pivotal strategy; however, its successful application necessitates the establishment of precise models, with computational costs potentially serving as a limiting factor. The utilization of a model catalyst featuring a precisely defined structure will lead to a notable reduction in the computational workload and complexity. This streamlined representation facilitates more efficient computational analyses, easing the challenges associated with simulations and calculations in catalytic studies. Moreover, machine learning has gained recognition for its potential to guide experimental designs, reduce trial-and-error costs, and accelerate the experimental cycle. The integration of machine learning with both experimental and theoretical approaches holds promise in unravelling structure-performance relationships, facilitating continuous innovation in electrocatalysis technology.
Standardization and open information
The effective analysis and interpretation of data derived from in-situ and operando experiments are imperative given the extensive and intricate nature of the generated datasets. The absence of standardized protocols for these measurements may introduce methodological variations across studies, thereby posing challenges in making meaningful comparisons. To address this, a comprehensive strategy is essential. This involves the establishment of standardized protocols, the adoption of uniform data formats, the utilization of advanced data processing tools, and the promotion of interdisciplinary collaboration. We recommend that the community adhere to the FAIR data principles: Findability, Accessibility, Interoperability, and Reuse of digital assets. The implementation of such a multifaceted approach enhances our capacity to handle and interpret complex experimental data, ensuring both consistency and comparability.
To overcome these challenges, interdisciplinary collaboration and continuous innovation in experimental and computational techniques are essential for successful in-situ and operando applications. We anticipate that as the above-listed challenges continually are gradually overcome, progress in the field and consequent translation to real-world applications will further accelerate.
Data availability
There is data generated in this work.
Code availability
There is code generated in this work.
References
Descorme, C., Gallezot, P., Geantet, C. & George, C. Heterogeneous catalysis: a key tool toward sustainability. ChemCatChem 4, 1897–1906 (2012).
CASGoogle Scholar
De Luna, P. et al. What would it take for renewably powered electrosynthesis to displace petrochemical processes? Science 364, eaav3506 (2019).
ADSPubMedGoogle Scholar
Handoko, A. D., Wei, F., Jenndy, Yeo, B. S. & Seh, Z. W. Understanding heterogeneous electrocatalytic carbon dioxide reduction through operando techniques. Nat. Catal. 1, 922–934 (2018).
CASGoogle Scholar
Zhu, Y., Wang, J., Chu, H., Chu, Y.-C. & Chen, H. M. In-situ/operando studies for designing next-generation electrocatalysts. ACS Energy Lett. 5, 1281–1291 (2020).
CASMATHGoogle Scholar
Hasa, B., Zhao, Y. & Jiao, F. In-situ/operando characterization techniques of electrochemical CO2 reduction. Annu. Rev. Chem. Biomol. Eng. 14, 165–185 (2023).
CASPubMedMATHGoogle Scholar
Favaro, M., Artiglia, L. & Mun, B. S. In-situ/operando investigation of catalytic and electrocatalytic interfaces. J. Phys. D: Appl. Phys. 55, 060201 (2022).
ADSCASMATHGoogle Scholar
Watkins, N. B. et al. Hydrodynamics change Tafel slopes in electrochemical CO2 reduction on copper. ACS Energy Lett. 8, 2185–2192 (2023).
CASMATHGoogle Scholar
Verdaguer-Casadevall, A. et al. Probing the active surface sites for CO reduction on oxide-derived copper electrocatalysts. J. Am. Chem. Soc. 137, 9808–9811 (2015).
CASPubMedMATHGoogle Scholar
Lee, S. H. et al. Correlating oxidation state and surface area to activity from operando studies of copper CO electroreduction catalysts in a gas-fed device. ACS Catal. 10, 8000–8011 (2020).
CASMATHGoogle Scholar
Clark, E. L. & Bell, A. T. Direct observation of the local reaction environment during the electrochemical reduction of CO2. J. Am. Chem. Soc. 140, 7012–7020 (2018).
CASPubMedMATHGoogle Scholar
Farmand, M. et al. Electrochemical flow cell enabling operando probing of electrocatalyst surfaces by X-ray spectroscopy and diffraction. Phys. Chem. Chem. Phys. 21, 5402–5408 (2019).
CASPubMedMATHGoogle Scholar
Meyer, Q., Zeng, Y. & Zhao, C. In-situ and operando characterization of proton exchange membrane fuel cells. Adv. Mater. 31, 1901900 (2019).
CASGoogle Scholar
El-Nagar, G. A., Haun, F., Gupta, S., Stojkovikj, S. & Mayer, M. T. Unintended cation crossover influences CO2 reduction selectivity in Cu-based zero-gap electrolysers. Nat. Commun. 14, 2062 (2023).
ADSCASPubMedPubMed CentralGoogle Scholar
Lin, S.-C. et al. Operando time-resolved X-ray absorption spectroscopy reveals the chemical nature enabling highly selective CO2 reduction. Nat. Commun. 11, 3525 (2020).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Long, C. et al. Regulating reconstruction of oxide-derived Cu for electrochemical CO2 reduction toward n-propanol. Sci. Adv. 9, eadi6119 (2023).
Mendoza, D., Dong, S.-T. & Lassalle-Kaiser, B. In-situ/operando X-ray spectroscopy applied to electrocatalytic CO2 reduction: status and perspectives. Curr. Opin. Colloid Interface Sci. 61, 101635 (2022).
CASGoogle Scholar
Mom, R. V. et al. Assessment of the degradation mechanisms of Cu electrodes during the CO2 reduction reaction. ACS Appl. Mater. Interfaces 15, 30052–30059 (2023).
CASPubMedPubMed CentralMATHGoogle Scholar
Yang, F. et al. Operando analysis reveals potential-driven in-situ formation of single-Fe-atom electrocatalysts for green production of ammonia. Proc. Natl Acad. Sci. USA 120, e2301011120 (2023).
CASPubMedPubMed CentralGoogle Scholar
Timoshenko, J. & Roldan Cuenya, B. In-situ/operando electrocatalyst characterization by X-ray absorption spectroscopy. Chem. Rev. 121, 882–961 (2021).
CASPubMedMATHGoogle Scholar
Lee, C.-Y., Taylor, A. C., Nattestad, A., Beirne, S. & Wallace, G. G. 3D printing for electrocatalytic applications. Joule 3, 1835–1849 (2019).
CASMATHGoogle Scholar
Savage, T. et al. Machine learning-assisted discovery of flow reactor designs. Nat. Chem. Eng. 1, 522–531 (2024).
Achilli, E. et al. 3D-printed photo-spectroelectrochemical devices for in-situ and in operando X-ray absorption spectroscopy investigation. J. Synchrotron Radiat. 23, 622–628 (2016).
PubMedGoogle Scholar
Zhong, L., Chen, D. & Zafeiratos, S. A mini review of in-situ near-ambient pressure XPS studies on non-noble, late transition metal catalysts. Catal. Sci. Technol. 9, 3851–3867 (2019).
CASMATHGoogle Scholar
Brummel, O. et al. A versatile approach to electrochemical in-situ ambient-pressure X-ray photoelectron spectroscopy: application to a complex model catalyst. J. Phys. Chem. Lett. 13, 11015–11022 (2022).
CASPubMedMATHGoogle Scholar
Avilés Acosta, J. E., Lin, J. C., Un Lee, D., Jaramillo, T. F. & Hahn, C. Electrochemical flow reactor design allows tunable mass transport conditions for operando surface enhanced infrared absorption spectroscopy. ChemCatChem 15, e202300520 (2023).
Google Scholar
Lazar, J., Schnelting, C., Slavcheva, E. & Schnakenberg, U. Hampering of the stability of gold electrodes by ferri-/ferrocyanide redox couple electrolytes during electrochemical impedance spectroscopy. Anal. Chem. 88, 682–687 (2016).
CASPubMedGoogle Scholar
Noël, T., Cao, Y. & Laudadio, G. The fundamentals behind the use of flow reactors in electrochemistry. Acc. Chem. Res. 52, 2858–2869 (2019).
PubMedPubMed CentralMATHGoogle Scholar
Sherwin, C. et al. An optimised cell for in-situ XAS of gas diffusion electrocatalyst electrodes. ChemCatChem 16, e202400221 (2024).
CASMATHGoogle Scholar
Piccolo, L. et al. Operando X-ray absorption spectroscopy investigation of photocatalytic hydrogen evolution over ultradispersed Pt/TiO2 catalysts. ACS Catal. 10, 12696–12705 (2020).
CASGoogle Scholar
Sasaki, K., Marinkovic, N., Isaacs, H. S. & Adzic, R. R. Synchrotron-based in-situ characterization of carbon-supported platinum and platinum monolayer electrocatalysts. ACS Catal. 6, 69–76 (2016).
CASGoogle Scholar
Ostervold, L., Hoffman, A. S., Thompson, D., Bare, S. R. & Clark, E. L. A versatile electrochemical cell for operando XAS. ChemCatChem n/a, e202400072 (2024).
Google Scholar
Lopez-Astacio, H., Vargas-Perez, B. L., Valle-Perez, D., Pollock, C. J. & Cunci, L. Open-source electrochemical cell for in-situ X-ray absorption spectroscopy in transmission and fluorescence modes. J. Synchrotron Radiat. 31, 322–327 (2024).
Diklić, N. et al. Potential pitfalls in the operando XAS study of oxygen evolution electrocatalysts. ACS Energy Lett. 7, 1735–1740 (2022).
MATHGoogle Scholar
Li, L. et al. Operando observation of chemical transformations of iridium oxide during photoelectrochemical water oxidation. ACS Appl. Energy Mater. 2, 1371–1379 (2019).
CASMATHGoogle Scholar
Yang, Y. et al. In-situ X-ray absorption spectroscopy of a synergistic Co–Mn oxide catalyst for the oxygen reduction reaction. J. Am. Chem. Soc. 141, 1463–1466 (2019).
CASPubMedMATHGoogle Scholar
Imai, H. et al. In-situ and real-time monitoring of oxide growth in a few monolayers at surfaces of platinum nanoparticles in aqueous media. J. Am. Chem. Soc. 131, 6293–6300 (2009).
CASPubMedMATHGoogle Scholar
Czioska, S. et al. Increased Ir–Ir interaction in iridium oxide during the oxygen evolution reaction at high potentials probed by operando spectroscopy. ACS Catal. 11, 10043–10057 (2021).
CASMATHGoogle Scholar
Diercks, J. S. et al. Interplay between surface-adsorbed CO and bulk Pd hydride under CO2-electroreduction conditions. ACS Catal. 12, 10727–10741 (2022).
CASMATHGoogle Scholar
Binninger, T. et al. Electrochemical flow-cell setup for in-situ X-ray investigations. J. Electrochem. Soc. 163, H906 (2016).
CASMATHGoogle Scholar
Zhang, X. Y. et al. In operando identification of in-situ formed metalloid zincδ+ active sites for highly efficient electrocatalyzed carbon dioxide reduction. Angew. Chem. Int. Ed. 61, e202202298 (2022).
CASGoogle Scholar
Möller, T. et al. Electrocatalytic CO2 reduction on CuOx nanocubes: tracking the evolution of chemical state, geometric structure, and catalytic selectivity using operando spectroscopy. Angew. Chem. Int. Ed. 59, 17974–17983 (2020).
Google Scholar
Stevens, M. B. et al. Identifying and tuning the in-situ oxygen-rich surface of molybdenum nitride electrocatalysts for oxygen reduction. ACS Appl. Energy Mater. 3, 12433–12446 (2020).
CASMATHGoogle Scholar
Huang, H. & Russell, A. E. Approaches to achieve surface sensitivity in the in-situ XAS of electrocatalysts. Curr. Opin. Electrochem. 27, 100681 (2021).
CASMATHGoogle Scholar
Alesker, M. et al. Operando X-ray absorption spectroscopy of a Pd/γ-NiOOH 2 nm cubes hydrogen oxidation catalyst in an alkaline membrane fuel cell. Catal. Sci. Technol. 11, 1337–1344 (2021).
CASGoogle Scholar
Gao, Y. et al. Investigating degradation mechanisms in PtCo alloy catalysts: the role of Co content and a Pt-rich shell using operando high-energy resolution fluorescence detection X-ray absorption spectroscopy. ACS Appl. Mater. Interfaces 15, 52473–52484 (2023).
CASGoogle Scholar
Li, J. et al. Experimental proof of the bifunctional mechanism for the hydrogen oxidation in alkaline media. Angew. Chem. Int. Ed. 56, 15594–15598 (2017).
CASMATHGoogle Scholar
Cao, D. et al. Anomalous self-optimization of sulfate ions for boosted oxygen evolution reaction. Sci. Bull. 66, 553–561 (2021).
CASMATHGoogle Scholar
Wu, B. et al. In-situ X-ray absorption spectroscopy of metal/nitrogen-doped carbons in oxygen electrocatalysis. Angew. Chem. Int. Ed. 62, e202219188 (2023).
CASGoogle Scholar
Dzara, M. J. et al. Characterizing complex gas–solid interfaces with in-situ spectroscopy: oxygen adsorption behavior on Fe–N–C catalysts. J. Phys. Chem. C 124, 16529–16543 (2020).
CASGoogle Scholar
Velasco-Velez, J.-J. et al. The structure of interfacial water on gold electrodes studied by x-ray absorption spectroscopy. Science 346, 831–834 (2014).
ADSCASPubMedMATHGoogle Scholar
Wang, J. et al. In-situ X-ray spectroscopies beyond conventional X-ray absorption spectroscopy on deciphering dynamic configuration of electrocatalysts. Nat. Commun. 14, 6576 (2023).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Zhao, Y. et al. Oxygen evolution/reduction reaction catalysts: from in-situ monitoring and reaction mechanisms to rational design. Chem. Rev. 123, 6257–6358 (2023).
CASPubMedMATHGoogle Scholar
Acharya, P. et al. Temporal Ni K-edge X-ray absorption spectroscopy study reveals the kinetics of the Ni redox behavior of the iron–nickel oxide bimetallic OER catalyst. J. Phys. Chem. C 127, 11891–11901 (2023).
CASMATHGoogle Scholar
Pasquini, C. et al. Operando tracking of oxidation-state changes by coupling electrochemistry with time-resolved X-ray absorption spectroscopy demonstrated for water oxidation by a cobalt-based catalyst film. Anal. Bioanal. Chem. 413, 5395–5408 (2021).
CASPubMedPubMed CentralMATHGoogle Scholar
Albrahim, M. et al. Reduction and agglomeration of supported metal clusters induced by high-flux X-ray absorption spectroscopy measurements. J. Phys. Chem. C 125, 11048–11057 (2021).
CASGoogle Scholar
Gopakumar, G. et al. Radiation damage by extensive local water ionization from two-step electron-transfer-mediated decay of solvated ions. Nat. Chem. 15, 1408–1414 (2023).
CASPubMedPubMed CentralMATHGoogle Scholar
Zhu, Y. et al. Emerging dynamic structure of electrocatalysts unveiled by in-situ X-ray diffraction/absorption spectroscopy. Energy Environ. Sci. 14, 1928–1958 (2021).
CASMATHGoogle Scholar
Frenkel, A. I., Wang, Q., Sanchez, S. I., Small, M. W. & Nuzzo, R. G. Short range order in bimetallic nanoalloys: an extended X-ray absorption fine structure study. J. Chem. Phys. 138, 064202 (2013).
ADSPubMedGoogle Scholar
Yevick, A. & Frenkel, A. I. Effects of surface disorder on EXAFS modeling of metallic clusters. Phys. Rev. B 81, 115451 (2010).
ADSMATHGoogle Scholar
Yancey, D. F. et al. A theoretical and experimental examination of systematic ligand-induced disorder in Au dendrimer-encapsulated nanoparticles. Chem. Sci. 4, 2912–2921 (2013).
CASMATHGoogle Scholar
van der Merwe, M. et al. Unravelling the mechanistic complexity of the oxygen evolution reaction and Ir dissolution in highly dimensional amorphous hydrous iridium oxides. Energy Environ. Sci. 18, 1214–1231 (2025).
Kuzmin, A. & Chaboy, J. EXAFS and XANES analysis of oxides at the nanoscale. IUCrJ 1, 571–589 (2014).
CASPubMedPubMed CentralMATHGoogle Scholar
Diaz-Lopez, M. et al. Fast operando X-ray pair distribution function using the DRIX electrochemical cell. J. Synchrotron Radiat. 27, 1190–1199 (2020).
CASPubMedPubMed CentralMATHGoogle Scholar
Caliandro, R. et al. Structural dynamics of a nickel electrocatalyst during water splitting observed via the operando pair distribution function. Cell Rep. Phys. Sci. 5, 102341 (2024).
Pittkowski, R. K. et al. Monitoring the structural changes in iridium nanoparticles during oxygen evolution electrocatalysis with operando X-ray total scattering. J. Am. Chem. Soc. 146, 27517–27527 (2024).
CASPubMedPubMed CentralGoogle Scholar
Li, J., Al-Mahayni, H., Chartrand, D., Seifitokaldani, A. & Kornienko, N. Electrochemical formation of C–S bonds from CO2 and small-molecule sulfur species. Nat. Synth. 2, 757–765 (2023).
ADSCASGoogle Scholar
Rabeah, J. et al. Multivariate analysis of coupled operando EPR/XANES/EXAFS/UV–Vis/ATR-IR spectroscopy: a new dimension for mechanistic studies of catalytic gas–liquid phase reactions. Chem.—Eur. J. 26, 7395–7404 (2020).
CASPubMedMATHGoogle Scholar
Liu, Y. et al. Probing active sites in CuxPdy cluster catalysts by machine-learning-assisted X-ray absorption spectroscopy. ACS Appl. Mater. Interfaces 13, 53363–53374 (2021).
CASPubMedGoogle Scholar
Bunău, O. & Joly, Y. Self-consistent aspects of x-ray absorption calculations. J. Phys.: Condens. Matter 21, 345501 (2009).
PubMedMATHGoogle Scholar
Rehr, J. J., Kas, J. J., Vila, F. D., Prange, M. P. & Jorissen, K. Parameter-free calculations of X-ray spectra with FEFF9. Phys. Chem. Chem. Phys. 12, 5503–5513 (2010).
CASPubMedGoogle Scholar
Nattino, F. & Marzari, N. Operando XANES from first-principles and its application to iridium oxide. Phys. Chem. Chem. Phys. 22, 10807–10818 (2020).
CASPubMedMATHGoogle Scholar
Teliska, M., O’Grady, W. E. & Ramaker, D. E. Determination of O and OH adsorption sites and coverage in-situ on Pt electrodes from Pt L23 X-ray absorption spectroscopy. J. Phys. Chem. B 109, 8076–8084 (2005).
CASPubMedMATHGoogle Scholar
Zhai, H. & Alexandrova, A. N. Fluxionality of catalytic clusters: when it matters and how to address it. ACS Catal. 7, 1905–1911 (2017).
CASMATHGoogle Scholar
Saeki, M. et al. Application of an augmentation method to MCR-ALS analysis for XAFS and Raman data matrices in the structural change of isopolymolybdates. Anal. Sci. 36, 1371–1375 (2020).
CASPubMedMATHGoogle Scholar
Zandkarimi, B. et al. Interpreting the operando XANES of surface-supported subnanometer clusters: when fluxionality, oxidation state, and size effect fight. J. Phys. Chem. C 124, 10057–10066 (2020).
CASMATHGoogle Scholar
Poths, P., Sun, G., Sautet, P. & Alexandrova, A. N. Interpreting the operando X-ray absorption near-edge structure of supported Cu and CuPd clusters in conditions of oxidative dehydrogenation of propane: dynamic changes in composition and size. J. Phys. Chem. C 126, 1972–1981 (2022).
CASGoogle Scholar
Yang, J. et al. Dynamic behavior of single-atom catalysts in electrocatalysis: identification of Cu-N3 as an active site for the oxygen reduction reaction. J. Am. Chem. Soc. 143, 14530–14539 (2021).
CASPubMedGoogle Scholar
Timoshenko, J. & Frenkel, A. I. “Inverting” X-ray absorption spectra of catalysts by machine learning in search for activity descriptors. ACS Catal. 9, 10192–10211 (2019).
CASMATHGoogle Scholar
Guda, A. A. et al. Quantitative structural determination of active sites from in-situ and operando XANES spectra: from standard ab initio simulations to chemometric and machine learning approaches. Catal. Today 336, 3–21 (2019).
CASMATHGoogle Scholar
Liu, Y. et al. Mapping XANES spectra on structural descriptors of copper oxide clusters using supervised machine learning. J. Chem. Phys. 151, 164201 (2019).
Stiles, P. L., Dieringer, J. A., Shah, N. C. & Van Duyne, R. P. Surface-enhanced Raman spectroscopy. Annu. Rev. Anal. Chem. 1, 601–626 (2008).
CASGoogle Scholar
Osawa, M. in Near-field Optics and Surface Plasmon Polaritons (ed. Kawata, S.) 163–187 (Springer, 2006).
Dunwell, M. et al. The central role of bicarbonate in the electrochemical reduction of carbon dioxide on gold. J. Am. Chem. Soc. 139, 3774–3783 (2017).
CASPubMedMATHGoogle Scholar
Zhu, S., Jiang, B., Cai, W.-B. & Shao, M. Direct observation on reaction intermediates and the role of bicarbonate anions in CO2 electrochemical reduction reaction on Cu surfaces. J. Am. Chem. Soc. 139, 15664–15667 (2017).
CASPubMedGoogle Scholar
Galloway, T. A. & Hardwick, L. J. Utilizing in-situ electrochemical SHINERS for oxygen reduction reaction studies in aprotic electrolytes. J. Phys. Chem. Lett. 7, 2119–2124 (2016).
CASPubMedMATHGoogle Scholar
Yonezawa, Y. et al. In-situ electrochemical surface-enhanced Raman scattering observation of molecules accelerating the hydrogen evolution reaction. J. Electroanal. Chem. 800, 7–12 (2017).
CASMATHGoogle Scholar
Zhong, H. et al. Boosting the electrocatalytic conversion of nitrogen to ammonia on metal-phthalocyanine-based two-dimensional conjugated covalent organic frameworks. J. Am. Chem. Soc. 143, 19992–20000 (2021).
CASPubMedGoogle Scholar
Ly, K. H. & Weidinger, I. M. Understanding active sites in molecular (photo)electrocatalysis through complementary vibrational spectroelectrochemistry. Chem. Commun. 57, 2328–2342 (2021).
CASMATHGoogle Scholar
Liang, S., Altaf, N., Huang, L., Gao, Y. & Wang, Q. Electrolytic cell design for electrochemical CO2 reduction. J. CO2 Util. 35, 90–105 (2020).
CASGoogle Scholar
Ren, S. et al. Catalyst aggregation matters for immobilized molecular CO2RR electrocatalysts. J. Am. Chem. Soc. 145, 4414–4420 (2023).
CASPubMedMATHGoogle Scholar
Deng, Y. et al. Operando Raman spectroscopy of amorphous molybdenum sulfide (MoSx) during the electrochemical hydrogen evolution reaction: identification of sulfur atoms as catalytically active sites for H+ reduction. ACS Catal. 6, 7790–7798 (2016).
CASMATHGoogle Scholar
Li, J. & Kornienko, N. Electrochemically driven C–N bond formation from CO2 and ammonia at the triple-phase boundary. Chem. Sci. 13, 3957–3964 (2022).
CASPubMedPubMed CentralMATHGoogle Scholar
Dunwell, M. et al. Examination of near-electrode concentration gradients and kinetic impacts on the electrochemical reduction of CO2 using Surface-enhanced infrared spectroscopy. ACS Catal. 8, 3999–4008 (2018).
CASMATHGoogle Scholar
Lu, X. et al. In-situ observation of the pH gradient near the gas diffusion electrode of CO2 reduction in alkaline electrolyte. J. Am. Chem. Soc. 142, 15438–15444 (2020).
CASPubMedGoogle Scholar
Yang, K., Kas, R. & Smith, W. A. In-situ infrared spectroscopy reveals persistent alkalinity near electrode surfaces during CO2 electroreduction. J. Am. Chem. Soc. 141, 15891–15900 (2019).
CASPubMedPubMed CentralGoogle Scholar
Reich, S. & Thomsen, C. Raman spectroscopy of graphite. Philos. Trans. R. Soc. Lond. Ser. A: Math. Phys. Eng. Sci. 362, 2271–2288 (2004).
ADSCASMATHGoogle Scholar
Wei, J. et al. Probing the oxygen reduction reaction intermediates and dynamic active site structures of molecular and pyrolyzed Fe–N–C electrocatalysts by in-situ Raman spectroscopy. ACS Catal. 12, 7811–7820 (2022).
CASMATHGoogle Scholar
Lee, S., Banjac, K., Lingenfelder, M. & Hu, X. Oxygen isotope labeling experiments reveal different reaction sites for the oxygen evolution reaction on nickel and nickel iron oxides. Angew. Chem. Int. Ed. 58, 10295–10299 (2019).
CASGoogle Scholar
Chernyshova, I. V., Somasundaran, P. & Ponnurangam, S. On the origin of the elusive first intermediate of CO2 electroreduction. Proc. Natl Acad. Sci. USA 115, E9261–E9270 (2018).
ADSCASPubMedPubMed CentralGoogle Scholar
Li, J., Zhang, Y., Kuruvinashetti, K. & Kornienko, N. Construction of C–N bonds from small-molecule precursors through heterogeneous electrocatalysis. Nat. Rev. Chem. 6, 303–319 (2022).
CASPubMedGoogle Scholar
Chen, C. et al. Coupling N2 and CO2 in H2O to synthesize urea under ambient conditions. Nat. Chem. 12, 717–724 (2020).
ADSCASPubMedMATHGoogle Scholar
Xu, X. et al. Revealing *OOH key intermediates and regulating H2O2 photoactivation by surface relaxation of Fenton-like catalysts. Proc. Natl Acad. Sci. USA 119, e2205562119 (2022).
CASPubMedPubMed CentralGoogle Scholar
Shao, F. et al. In-situ spectroelectrochemical probing of CO redox landscape on copper single-crystal surfaces. Proc. Natl Acad. Sci. USA 119, e2118166119 (2022).
CASPubMedPubMed CentralGoogle Scholar
An, H. et al. Sub-second time-resolved surface-enhanced Raman spectroscopy reveals dynamic CO intermediates during electrochemical CO2 reduction on copper. Angew. Chem. Int. Ed. 60, 16576–16584 (2021).
CASGoogle Scholar
de Ruiter, J. et al. Probing the dynamics of low-overpotential CO2-to-CO activation on copper electrodes with time-resolved Raman spectroscopy. J. Am. Chem. Soc. 144, 15047–15058 (2022).
PubMedPubMed CentralMATHGoogle Scholar
Herzog, A. et al. Operando Raman spectroscopy uncovers hydroxide and CO species enhance ethanol selectivity during pulsed CO2 electroreduction. Nat. Commun. 15, 3986 (2024).
ADSCASPubMedPubMed CentralGoogle Scholar
Vimont, A., Thibault-Starzyk, F. & Daturi, M. Analysing and understanding the active site by IR spectroscopy. Chem. Soc. Rev. 39, 4928–4950 (2010).
CASPubMedMATHGoogle Scholar
Lansford, J. L. & Vlachos, D. G. Infrared spectroscopy data- and physics-driven machine learning for characterizing surface microstructure of complex materials. Nat. Commun. 11, 1513 (2020).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Lansford, J. L. & Vlachos, D. G. Spectroscopic probe molecule selection using quantum theory, first-principles calculations, and machine learning. ACS Nano 14, 17295–17307 (2020).
CASPubMedMATHGoogle Scholar
Hammer, B., Nielsen, O. H. & Nrskov, J. K. Structure sensitivity in adsorption: CO interaction with stepped and reconstructed Pt surfaces. Catal. Lett. 46, 31–35 (1997).
CASGoogle Scholar
Herl, T. & Matysik, F.-M. Recent developments in electrochemistry–mass spectrometry. ChemElectroChem 7, 2498–2512 (2020).
CASMATHGoogle Scholar
Baltruschat, H. Differential electrochemical mass spectrometry. J. Am. Soc. Mass Spectrom. 15, 1693–1706 (2004).
ADSCASPubMedMATHGoogle Scholar
Liu, H. et al. Ultra-low voltage bipolar hydrogen production from biomass-derived aldehydes and water in membrane-less electrolyzers. Energy Environ. Sci. 15, 4175–4189 (2022).
CASMATHGoogle Scholar
Kasian, O. et al. Degradation of iridium oxides via oxygen evolution from the lattice: correlating atomic scale structure with reaction mechanisms. Energy Environ. Sci. 12, 3548–3555 (2019).
CASMATHGoogle Scholar
Shao, J. et al. Scalable electrosynthesis of formamide through C–N coupling at the industrially relevant current density of 120 mA cm−2. Angew. Chem. Int. Ed. 61, e202213009 (2022).
ADSCASGoogle Scholar
Li, Y. et al. Electrocatalytic reduction of low-concentration nitric oxide into ammonia over Ru nanosheets. ACS Energy Lett. 7, 1187–1194 (2022).
ADSCASMATHGoogle Scholar
Li, T. et al. Ru-doped Pd nanoparticles for nitrogen electrooxidation to nitrate. ACS Catal. 11, 14032–14037 (2021).
CASGoogle Scholar
Wang, Y., Zhou, W., Jia, R., Yu, Y. & Zhang, B. Unveiling the activity origin of a copper-based electrocatalyst for selective nitrate reduction to ammonia. Angew. Chem. Int. Ed. 59, 5350–5354 (2020).
CASMATHGoogle Scholar
Wang, Y., Wang, C., Li, M., Yu, Y. & Zhang, B. Nitrate electroreduction: mechanism insight, in-situ characterization, performance evaluation, and challenges. Chem. Soc. Rev. 50, 6720–6733 (2021).
CASPubMedMATHGoogle Scholar
Shi, Z. et al. Confined Ir single sites with triggered lattice oxygen redox: Toward boosted and sustained water oxidation catalysis. Joule 5, 2164–2176 (2021).
CASMATHGoogle Scholar
Wen, Y. et al. Stabilizing highly active Ru sites by suppressing lattice oxygen participation in acidic water oxidation. J. Am. Chem. Soc. 143, 6482–6490 (2021).
CASPubMedMATHGoogle Scholar
Chen, W. et al. Effect of mass transfer and solution composition on the quantification of reaction kinetics by differential electrochemical mass spectrometry. J. Energy Chem. 56, 412–419 (2021).
ADSCASMATHGoogle Scholar
He, Z. et al. Activating lattice oxygen in NiFe-based (oxy)hydroxide for water electrolysis. Nat. Commun. 13, 2191 (2022).
Yao, Y., Zhu, S., Wang, H., Li, H. & Shao, M. A spectroscopic study of electrochemical nitrogen and nitrate reduction on rhodium surfaces. Angew. Chem. Int. Ed. 59, 10479–10483 (2020).
CASMATHGoogle Scholar
Wen, G. et al. Continuous CO2 electrolysis using a CO2 exsolution-induced flow cell. Nat. Energy 7, 978–988 (2022).
ADSCASMATHGoogle Scholar
He, F. et al. The effect of water on the quantification of volatile species by differential electrochemical mass spectrometry. Anal. Chem. 93, 5547–5555 (2021).
CASPubMedMATHGoogle Scholar
Lin, C. et al. In-situ reconstructed Ru atom array on α-MnO2 with enhanced performance for acidic water oxidation. Nat. Catal. 4, 1012–1023 (2021).
ADSCASMATHGoogle Scholar
Huang, Y. et al. Direct electrosynthesis of urea from carbon dioxide and nitric oxide. ACS Energy Lett. 7, 284–291 (2021).
ADSMATHGoogle Scholar
Han, S. et al. Ultralow overpotential nitrate reduction to ammonia via a three-step relay mechanism. Nat. Catal. 6, 402–414 (2023).
CASMATHGoogle Scholar
Sun, W.-J. et al. Built-in electric field triggered interfacial accumulation effect for efficient nitrate removal at ultra-low concentration and electroreduction to ammonia. Angew. Chem. Int. Ed. 60, 22933 (2021).
CASGoogle Scholar
Sedano Varo, E. et al. Gold nanoparticles for CO2 electroreduction: an optimum defined by size and shape. J. Am. Chem. Soc. 146, 2015–2023 (2024).
CASPubMedPubMed CentralMATHGoogle Scholar
Timoshenko, J., Lu, D., Lin, Y. & Frenkel, A. I. Supervised machine-learning-based determination of three-dimensional structure of metallic nanoparticles. J. Phys. Chem. Lett. 8, 5091–5098 (2017).
CASPubMedMATHGoogle Scholar
Download references
Acknowledgements
D.T. acknowledges support from the U.S. Department of Energy (DOE) Office of Science Graduate Student Research (SCGSR) Programme. S.R.B. acknowledges Co-ACCESS, part of the SUNCAT Center for Interface Science and Catalysis, which is supported by the U.S. DOE, Office of Science, BES, Chemical Sciences, Geosciences, and Biosciences Division. Work from A.P. and C.H. was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344 and was supported by the Center for Closing the Carbon Cycle, an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), under Award Number DE-SC0023427 with IM release number LLNL-JRNL-868780. A.N.A. and Y.L. acknowledge the support from the DE-SC0019152 grant. S.Y., S.C. and N.K. acknowledge the University of Bonn and the Deutsche Forschungsgemeinschaft grant KO 7060/4-1. IMW gratefully acknowledges financial support from the Deutsche Forschungsgemeinschaft (DFG) within CRC 1415 (Grant No. 417590517) and RTG 2861 (Grant No. 491865171).Y.S. acknowledges the support from the National Natural Science Foundation of China (No. 22275134).
Author information
Authors and Affiliations
Materials Science Division, Lawrence Livermore National Laboratory, Livermore, CA, 94550, USA
Aditya Prajapati & Christopher Hahn
Faculty of Chemistry and Food Chemistry, Technische Universität Dresden, 01062, Dresden, Germany
Inez M. Weidinger
Institute of Molecular Plus, Tianjin University, 300072, Tianjin, China
Yanmei Shi
Department of Chemistry and Biochemistry, University of California, Los Angeles, Los Angeles, CA, 90095, USA
Yonghyuk Lee & Anastassia N. Alexandrova
Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, Menlo Park, CA, 94025, USA
David Thompson & Simon R. Bare
Department of Chemistry and Biochemistry, University of Arkansas, Fayetteville, AR, 72701, USA
David Thompson
Institute of Inorganic Chemistry, University of Bonn, Gerhard-Domagk-Str. 1, 53121, Bonn, Germany
Shuai Chen, Shuai Yan & Nikolay Kornienko
Authors
Aditya Prajapati
View author publications
You can also search for this author in PubMedGoogle Scholar
2. Christopher Hahn
View author publications
You can also search for this author in PubMedGoogle Scholar
3. Inez M. Weidinger
View author publications
You can also search for this author in PubMedGoogle Scholar
4. Yanmei Shi
View author publications
You can also search for this author in PubMedGoogle Scholar
5. Yonghyuk Lee
View author publications
You can also search for this author in PubMedGoogle Scholar
6. Anastassia N. Alexandrova
View author publications
You can also search for this author in PubMedGoogle Scholar
7. David Thompson
View author publications
You can also search for this author in PubMedGoogle Scholar
8. Simon R. Bare
View author publications
You can also search for this author in PubMedGoogle Scholar
9. Shuai Chen
View author publications
You can also search for this author in PubMedGoogle Scholar
10. Shuai Yan
View author publications
You can also search for this author in PubMedGoogle Scholar
11. Nikolay Kornienko
View author publications
You can also search for this author in PubMedGoogle Scholar
Contributions
A.P., C.H., I.M.W., Y.S., Y.L., A.N.A., D.T., S.R.B., S.C., S.Y. and N.K., contributed towards literature analysis and manuscript writing.
Corresponding authors
Correspondence to Christopher Hahn, Inez M. Weidinger, Yanmei Shi, Anastassia N. Alexandrova, Simon R. Bare or Nikolay Kornienko.
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Reprints and permissions
About this article
Check for updates. Verify currency and authenticity via CrossMark
Cite this article
Prajapati, A., Hahn, C., Weidinger, I.M. et al. Best practices for in-situ and operando techniques within electrocatalytic systems. Nat Commun 16, 2593 (2025). https://doi.org/10.1038/s41467-025-57563-6
Download citation
Received:30 August 2024
Accepted:25 February 2025
Published:16 March 2025
DOI:https://doi.org/10.1038/s41467-025-57563-6
Share this article
Anyone you share the following link with will be able to read this content:
Get shareable link
Sorry, a shareable link is not currently available for this article.
Copy to clipboard
Provided by the Springer Nature SharedIt content-sharing initiative