Begin with the question the graph can actually answer
How do you read an electrochemical impedance spectroscopy distribution of relaxation times plot? Start by treating an EIS DRT plot as a way of displaying the time scales represented in measured impedance, not as an automatic list of reactions taking place inside an electrode. A distribution of relaxation times, often abbreviated DRT, is obtained by mathematically analysing frequency-dependent impedance. It may separate responses that overlap in a Nyquist arc, but a peak is evidence of a resolvable time-dependent contribution under a chosen model and data treatment. It is not a chemical identification label.
A reader looking for help interpreting DRT peaks should first locate the horizontal relaxation-time axis, the vertical distribution amplitude, and the frequency range from which the spectrum was reconstructed. Only after that should the reader examine which peak becomes taller, shifts, merges, or disappears between conditions. The graph is most informative when it remains connected to the original EIS experiment, its measurement conditions, and the uncertainty of the numerical inversion.
Think of the DRT as a translated view of an electrical measurement. Good translation preserves what the source measured and explains which interpretive steps have been added. It does not replace an unfamiliar waveform with a confident story about a battery, catalyst, coating, or fuel cell that the plot alone cannot substantiate.
Understand what impedance contributes before looking for peaks
Electrochemical impedance is the ratio of a small alternating potential perturbation to the resulting alternating current response, with a magnitude and phase relationship that vary with angular frequency. A conventional measurement represents impedance as a complex quantity with real and imaginary parts. Frequency can expose different responses because fast processes and slow processes do not follow changing excitation in the same way. The Nyquist plot and Bode plot are two familiar views of this information; DRT processing supplies another.
In an idealised resistance–capacitance process, a characteristic relaxation time is the product of resistance and capacitance. Real electrochemical interfaces rarely behave as a single perfect resistor coupled to a perfect capacitor. Surface heterogeneity, porous pathways, distributed reaction rates, diffusion, and measurement limitations broaden or distort ideal responses. The DRT therefore often looks like a range of contributions, not a row of exact isolated impulses.
A novice should distinguish the measured inputs from inferred outputs. Current and potential are recorded; impedance is calculated under assumptions about linear response; a DRT is estimated from the impedance. Each additional calculation is useful, but each also creates choices that must be described to anyone interpreting the final result.
Read the horizontal axis without reversing fast and slow
Most DRT plots put relaxation time, often written tau or τ, on the horizontal axis. It may be expressed in seconds and displayed logarithmically. When the axis increases to the right, a peak near a millisecond belongs to a faster characteristic response than a peak near ten seconds. Some software uses the inverse axis order or plots frequency rather than time. Always read the printed axis, not a remembered convention.
The elementary relation between characteristic time and characteristic frequency for a simple first-order relaxation is τ = 1/(2πf). A feature associated with a frequency near 1 kilohertz corresponds to a characteristic time near 0.16 milliseconds in that idealised mapping. This is not a license to convert every DRT peak into an exact process frequency, because distributed or non-Debye responses may complicate that relationship.
A log axis changes visual distance. Equal spacing may mean each step increases or decreases time tenfold, rather than by a fixed number of seconds. The span from 0.001 to 0.01 seconds covers a factor of ten, as does the span from 1 to 10 seconds. A peak that moves one division may therefore represent a substantial change in characteristic time.
Decide what the vertical axis measures
DRT software commonly labels its vertical axis γ(τ), gamma, distribution intensity, resistance contribution, or a related quantity. The units depend on the exact integral representation and whether the distribution is defined per natural logarithm of time, per decimal logarithm, or through a different kernel. Two programmes can plot curves with superficially similar shapes and different normalisations. A tall peak in one figure cannot be compared with a tall peak in another until their mathematical definitions match.
Area under a DRT curve may relate to a resistive contribution when the model uses an appropriate resistance-based representation. But the numerical area must be integrated in the correct coordinate: an integral with respect to d ln τ is not equivalent to an unweighted sum of ordinates sampled at arbitrary intervals. If values were rescaled or normalised, absolute area may no longer yield ohms without undoing the transformation.
Before translating a peak as more resistance, check the scale, definition, preprocessing, and colour legend. A visually greater amplitude does not necessarily mean a proportionally greater physical resistance, especially when peak width changes or regularisation settings differ.
Why two Nyquist arcs may become three DRT peaks
The principal appeal of DRT analysis is separation. Suppose an impedance spectrum contains a large depressed semicircle that appears to represent one interfacial process. Multiple relaxation contributions can overlap in the complex plane; the Nyquist outline may not reveal their boundaries clearly. Under favourable data quality and suitable inverse-model assumptions, a DRT can distinguish two or more concentrations of relaxation strength where only one broad arc appears.
This is a useful diagnostic clue rather than proof of distinct chemical mechanisms. The inverse calculation can split a broad response because of noise or an insufficient regularisation penalty. Conversely, a smoothing penalty can combine two genuine nearby processes into one wide hill. A claim that DRT has found three reactions must be corroborated with perturbation experiments, frequency-domain fit quality, and physically plausible assignments.
The disciplined wording is that the processed impedance contains three resolvable relaxation features under the stated analysis, and each feature can be investigated. That is much stronger educational writing than quietly replacing a modelling decision with certainty about microscopic chemistry.
What regularisation is doing behind the line
Reconstructing a relaxation distribution from finite, noisy impedance data is an inverse problem. Many different distributions may yield almost the same calculated spectrum, especially near the measured frequency limits. Regularisation imposes mathematical restraint. It may favour smoothness, non-negative amplitudes, sparsity, or another condition so that the inferred distribution does not chase noise point by point.
A regularisation parameter controls a trade-off. A weak penalty may preserve fine details but introduce oscillations, false narrow peaks, and unstable behaviour. A strong penalty often gives a calmer curve, at the price of broadening or erasing nearby contributions. A comparison of samples processed with different settings can become misleading even if the graph design is beautiful.
Readers should look for descriptions of the inversion algorithm, penalty or prior, grid of relaxation times, treatment of inductance, and method of selecting the tuning parameter. Authors may select a penalty through cross-validation, residual inspection, or other criteria. No selection technique can remove uncertainty completely; its purpose is to make the chosen reconstruction defensible and reproducible.
Ask how much frequency information the instrument really collected
A DRT peak does not arise from an unlimited observation of the cell. Its input EIS experiment covers a finite frequency window, perhaps many decades wide but not infinite. Very rapid responses may fall above the highest recorded frequency; very slow responses may fall below the lowest. The inversion algorithm can nevertheless produce a drawn curve outside the strongest data-supported interval.
At the edges, estimated features are vulnerable to extrapolation, baseline handling, and the representation chosen for series resistance or inductance. A peak near a boundary deserves more caution than an isolated stable peak well inside the measured window. More log-frequency points do not automatically rescue an experiment if the system was drifting as it was measured.
When interpreting a publication, locate the original frequency range. If it is missing, avoid repeating a precise assignment from a relaxation-time plot without qualification. When preparing a figure, show the frequency limits and provide the raw or reconstructable complex impedance if possible. Every claimed timescale needs a clear relationship to the actual acquired data.
Separate series resistance from dynamic relaxation
Ohmic resistance from electrolyte, wiring, contacts, and other nearly instantaneous contributions often appears as a high-frequency real-axis intercept in a Nyquist plot. A conventional DRT integral for relaxations may represent that resistance separately rather than as a finite-time peak. Therefore the absence of a very-fast DRT peak does not imply that the system has no bulk resistance.
If a report says a particular DRT peak is the electrolyte resistance, read its model carefully. The author may have embedded series terms into a chosen generalised distribution, or may have used informal naming that conceals the separation. For comparison, the safest practice is to report series resistance and relaxation contributions as distinct terms unless a validated formulation explicitly combines them.
This distinction matters during ageing studies. A cell may show increasing series resistance while its DRT peak shapes remain similar. Another cell may retain roughly the same high-frequency intercept but gain a growing interfacial time constant. Those are different physical observations and should produce different interpretive language, even if both raise total impedance.
Inductive loops require their own treatment
At high frequencies, stray inductance from cables, fixture geometry, and electrical connections can produce a positive imaginary impedance contribution under a common sign convention. Some electrochemical systems also display genuine low-frequency inductive-type behaviours associated with adsorbed intermediates or feedback processes. A straightforward non-negative RC-type DRT representation does not naturally describe every inductive response.
If inductive effects are ignored, an inversion may distort nearby peaks or leave systematic residuals. Some algorithms explicitly fit inductance as a separate term; others use more general kernels or signed distributions. Readers should identify which strategy was chosen before treating a negative contribution or unusual peak as evidence of chemical instability.
Do not assume that every loop below or above the axis means the same thing in two instruments. Verify whether the plotted Nyquist ordinate is Im(Z) or −Im(Z), and whether the source is a measured physical signal or an acquisition artefact. Consistent signs and an explicit method statement prevent a surprisingly large share of misread electrochemical figures.
What can be said about charge transfer?
Charge transfer describes electron exchange across an electrochemical interface. In a simplified system it can combine with double-layer charging into a characteristic response. A DRT feature at an intermediate time may therefore plausibly be associated with interfacial kinetics. But the location of a feature on the horizontal axis alone does not certify that assignment.
To test the charge-transfer hypothesis, examine how the feature responds to electrode potential, reactant concentration, temperature, surface area, and the availability of active sites. Compare the behaviour with complementary kinetics measurements and a model that predicts the observed trend. A feature that changes under a potential step may still include contributions from film evolution or transport.
Precise educational translation uses the phrase consistent with an interfacial kinetic contribution until other evidence justifies a stronger statement. It also distinguishes changes in a peak’s amplitude from shifts in its characteristic time. Both may accompany altered kinetics, but they are not interchangeable descriptors and may arise from different parameter changes in an equivalent model.
Surface films and coatings are not identified by colour
Passivation layers, corrosion products, polymer coatings, and battery solid-electrolyte interphases can contribute impedance. Some researchers assign higher-frequency features to films because films often introduce compact interfacial responses. That expectation can be a useful starting hypothesis, but its ordering is not a universal physical law.
A coating may be porous, heterogeneous, swelling, evolving with electrolyte uptake, or connected to multiple transport paths. Its signatures may overlap with charge transfer, contact resistance, or dielectric relaxation. The same figure may be interpreted differently if electrode geometry, operating state, or environmental conditions change.
Look for independent evidence such as spectroscopy, microscopy, controlled layer thickness, equivalent circuit consistency, or measurements before and after a deliberate surface modification. A DRT plot can motivate those tests. The plot itself cannot tell a reader the precise film composition, thickness, or protective quality merely because one region becomes larger. Translators should preserve the tentative status of assignments rather than presenting possible mechanisms as direct observations.
Diffusion does not always make a neat single peak
Mass transport commonly produces impedance that differs from an ideal narrow RC relaxation. Semi-infinite diffusion is often expressed through a Warburg-like frequency response, while finite-length diffusion and blocking boundaries produce different forms. A standard RC-kernel DRT may render transport as a broad feature, a tail, or a pattern whose shape depends on model assumptions and available frequencies.
If the slowest DRT region grows during a battery discharge or corrosive exposure, concentration transport may be involved. That does not establish a particular ionic diffusion coefficient. Geometry, thermodynamic factors, reaction coupling, and the selected diffusion boundary condition all matter when extracting transport parameters.
A good interpretation says what was measured, which transport model is being tested, and what alternatives remain. It avoids assigning every low-frequency peak to diffusion automatically. Physical systems can also show slow adsorption, porous-electrode redistribution, and cell-state drift, all of which may occupy overlapping temporal ranges. The appropriate next step is to vary conditions in a way that distinguishes these explanations.
Why temperature experiments improve peak interpretation
Temperature changes relaxation times through several possible mechanisms, including transport coefficients, charge-transfer kinetics, electrolyte conductivity, and thermally activated structural processes. Measuring EIS at several controlled temperatures can therefore help identify whether a DRT feature follows a plausible physical trend. A feature may shift to shorter relaxation time as a relevant process accelerates, but real systems can change composition or morphology at the same time.
Compare states only after the temperature is stable and the cell has had a suitable equilibration period. State of charge, electrolyte concentration, and mechanical stress should be controlled when possible. If the experimental sample evolves irreversibly while temperature is stepped, a clean Arrhenius interpretation is unjustified.
An Arrhenius-type plot may be useful for a specific extracted parameter if its theoretical meaning is established. However, a slope derived from a peak maximum is not automatically an activation energy for a named reaction. The quality of the underlying assignment matters as much as the apparent straightness of the fitted line.
State of charge and potential are not background details
Battery impedance is not a timeless fingerprint. The same cell can produce different DRT features at different states of charge, temperatures, and recent cycling histories. An electrode driven far from equilibrium may continue relaxing while the instrument sweeps through its frequency list. Such drift violates the stable-system assumption that ordinarily underpins straightforward EIS interpretation.
If comparing two peaks measured at different states, report the operating condition of each trace, the resting procedure, measurement amplitude, and direction of the preceding charge or discharge. A change in peak height might reflect reversible electrochemical state rather than permanent ageing.
This matters for phrases like degradation marker. A feature can be correlated with a worn cell under carefully matched test conditions; the same feature can also vary in a healthy cell as lithium content changes. The interpretive skill is to separate state dependence from durable material change. Without that separation, the DRT graph can support an attractive but false ageing narrative.
Check the small-signal assumption and stability
Impedance spectroscopy normally assumes that a sufficiently small perturbation probes a response near the operating point. If the excitation is too large, a nonlinear electrode can generate harmonics and amplitude-dependent impedance. If the system changes materially over the measurement period, frequencies acquired early and late may represent different physical states. Both issues threaten a simple DRT inversion.
A practical reading sequence is to check excitation amplitude, frequency ordering, duration, and stability indicators. Replicate spectra obtained under comparable conditions offer useful context. When authors report a linearity test or a consistency check, examine the outcome, not merely the assertion that one was performed.
No inverse method repairs invalid input physics by itself. A smooth, attractive distribution from drifting data may still be scientifically misleading. Conversely, a messy measured spectrum should not be silently smoothed until it matches an anticipated mechanism. State the measurement limitations and allow readers to judge the strength of the interpretation from the available evidence.
Kramers–Kronig checks are useful but not magical
Kramers–Kronig relations connect the real and imaginary parts of a suitable linear, causal, stable impedance response under the assumptions of the chosen formulation. Their numerical implementation can help detect inconsistency in measured data. However, limited frequency ranges and noise mean a real experiment is not judged by an abstract perfect test alone.
A consistency check is not a chemical fingerprint and cannot show that an equivalent circuit or DRT peak assignment is correct. A spectrum may pass a useful consistency criterion while several different physical models still fit its features. Likewise, failed consistency may motivate checking drift, wiring, perturbation size, or problematic frequencies before claiming a spectacular new electrochemical effect.
In translation, distinguish data quality from process identification. The first asks whether the measurement behaves like a plausible impedance response. The second asks what produced that response. The questions are related but not equivalent. Good experimental reporting keeps them separate and explains which evidence supports each conclusion.
What an equivalent circuit contributes to the interpretation
Equivalent circuits offer a compact physical or phenomenological description of impedance. A circuit containing a series resistor and two parallel resistance–capacitance branches may yield two relaxation times if the branches are sufficiently distinct. A DRT can help suggest whether one, two, or several timescales deserve further modelling. But the translation is not reversible in a unique way: different circuits may fit similar spectra.
Use DRT peaks as constraints or clues when proposing circuits, not as a command to add one element per plotted bump. A narrow peak may reflect a nearly idealised time constant; a broad peak can suggest distributed behaviour. Yet the correspondence depends on whether the chosen circuit and DRT kernel represent the same family of responses.
A useful report compares the circuit’s calculated impedance with the measured Nyquist and Bode plots, examines residuals, and states the physical rationale for each element. A circuit chosen only because its diagram seems familiar is weak evidence, even if fitting software returns an impressive number of decimal places.
Distinguish a peak moving from a peak growing
Suppose the DRT of a fresh electrode has a feature centred near 0.02 seconds. After a treatment the feature appears near 0.2 seconds. A horizontal shift to a longer time indicates a slower characteristic relaxation under that analysis; it does not by itself tell whether resistance rose or capacitance changed. In an ideal RC relationship, either parameter or both could account for a larger product.
Now suppose a peak stays at roughly the same time but doubles in height. Depending on the DRT normalisation and the change in its width, this may indicate more relaxation strength without a major shift of its centre. Again the mechanism must be tested. If the peak simultaneously broadens, looking only at maximum height can obscure area changes.
The most defensible comparison records central time, integrated contribution where meaningful, width, data interval, and uncertainty. A one-sentence summary can then say which attributes changed, rather than collapsing them all into the claim that impedance became worse.
Why shoulder features should be described carefully
A shoulder is a slope change on the side of a larger peak. It may indicate an overlapping contribution that cannot be fully resolved, but it can also arise from smoothing choices, data noise, or a broad non-Debye response. Treat shoulders as tentative features rather than hidden named reactions waiting to be revealed.
To investigate a shoulder, fit the original impedance with plausible alternative models, repeat the inversion using defensible regularisation strengths, and examine whether the shape persists across replicated measurements. If a separate local maximum appears only under a very weak penalty, it may not justify a separate mechanistic assignment. If it appears across conditions and follows a reproducible perturbation, the case becomes more persuasive.
Scientific language matters here. ‘An emerging slower-time shoulder’ reports a visible pattern. ‘A new diffusion mechanism’ asserts more than the graph may warrant. The first leaves space for additional analysis while still telling readers exactly what changed. That is a strength, not a weakness.
How to compare studies using different DRT algorithms
Cross-paper comparisons are especially hazardous because researchers may use different kernels, preprocessing rules, constrained optimisers, discretisations, and regularisation selection. Even when both plots use gamma against log tau, their amplitudes and widths may not be numerically equivalent. A sharp peak in one paper and a broad hill in another need not reflect a material difference.
Check whether authors report the full complex impedance, frequency window, fitting residual, uncertainty, and code or method settings. If possible, reprocess comparable raw spectra under one declared method before discussing absolute peak contributions. If not possible, keep the comparison qualitative and mark the methodological caveat.
Words such as identical, twice as large, or newly formed demand stronger comparability than words such as broadly similar timescale. A reader who understands this can make more reliable sense of the literature without treating every plotted colour as a common measurement scale. Technical communication improves when method differences are part of the interpretation rather than footnotes concealed after the conclusion.
A worked example with two processes and series resistance
Imagine a hypothetical cell measured at one temperature and state of charge. Its Nyquist plot shows one depressed arc followed by a modest slow-frequency extension. The corresponding DRT, processed with a consistent non-negative RC kernel, displays two peaks: one centred near 0.004 seconds and another near 0.4 seconds. The software separately reports a series resistance of 6 ohms. This is an illustrative dataset, not a result from a real experiment.
First, do not call the 6-ohm term a third DRT peak unless the model represents it that way. Second, do not name the 0.004-second peak the surface film and the 0.4-second peak charge transfer merely because of their relative order. Third, note that the slower feature may need particular scrutiny if the original spectrum reaches only moderately low frequencies.
A suitable report would describe two resolved relaxation regions plus an independent series term, identify the processing model, and propose experiments that test film, kinetic, or transport assignments. The analysis becomes valuable because it tells a future investigator what to measure, not because it pretends the final chemistry is already known.
A worked example of over-smoothing
Picture two simulated relaxation contributions whose characteristic times differ by less than one decade. With generous smoothing, the reconstructed DRT may show one large feature between them. With little smoothing, it may produce two maxima plus several small ripples near the boundary. Neither visual alone proves that the true system contains one or four processes.
The next step is not to choose the prettiest curve. Inspect predicted impedance and residuals for each reconstruction. Compare behaviour across justified parameter choices and repeated measurements. A synthetic test with known peak positions can reveal the method’s resolution under a similar noise level. If no choice consistently resolves two contributions, report the overlap.
A reader translating the result should use language that communicates limited resolution. Instead of saying ‘the two reactions merged’, one might say ‘the measured impedance does not reliably distinguish the nearby relaxation components under this analysis’. That sentence identifies a property of the available evidence rather than inventing a physical event.
A worked example of ageing at matched operating state
Suppose two EIS datasets come from the same battery at the same temperature, state of charge, rest time, perturbation amplitude, and instrumentation configuration. The later dataset has a similar series resistance but one medium-time DRT feature shows greater integrated strength. A reasonable first observation is that a time-dependent contribution became larger under closely matched measurement conditions.
It may be consistent with interfacial change during ageing, but the feature’s identity remains a hypothesis. Related evidence might include capacity fade, surface analysis, changes in electrochemical efficiency, or electrode-specific reference measurements. The peak’s alteration might also reflect a change in effective area or contact conditions. Importantly, the analyst should say whether the contribution was estimated with the same inverse settings in both datasets.
This example shows how to build a cautious causal story. First document the controlled comparison, next report the measured shift, then propose and test mechanisms. The strongest claims follow converging evidence; they do not lead the data by assuming every extra relaxation feature represents a newly formed degradation product.
A worked example of deceptive axis presentation
Imagine one figure whose x-axis reads log10(τ/s) with ticks −4, −3, −2 and another whose axis reads τ/s on a log scale with ticks 0.0001, 0.001 and 0.01. These express comparable characteristic-time positions using different notational conventions. A reader who mistakes −3 for negative time or compares −3 directly with 0.001 will misinterpret the graph.
Also imagine a panel arranged right-to-left for increasing relaxation time. If the reader assumes right always means slower, the direction of a reported shift may be reversed. This error is avoidable by taking ten seconds to transcribe the axis definition in ordinary language before reading the results.
In translation across languages or disciplinary styles, preserve symbols and units precisely. An equivalent written sentence might state ‘the peak moved from a time near one millisecond to a time near ten milliseconds’. That description works across plotting styles and reduces ambiguity for readers who do not use the same software.
A worked example near the low-frequency boundary
A coating study reports a broad peak at about 100 seconds, but the EIS scan’s lowest frequency is 0.1 hertz. For a simple first-order mapping, 0.1 hertz corresponds to about 1.6 seconds. The stated 100-second feature therefore lies far beyond the uncomplicated central mapping range and must be treated as strongly dependent on modelling or extrapolation choices.
This does not prove that the coating has no slow process. It means the dataset provides weak direct leverage on such a long relaxation. A much lower-frequency experiment could be attempted if the specimen remains stationary long enough. Alternatively, time-domain observations may illuminate very slow evolution, but they will have their own assumptions and limits.
A careful explanation distinguishes possible slow physics from a numerically reconstructed tail. Readers should not confuse a plotted estimated curve with uninterrupted experimental observation. The central question is not whether software can draw a peak; it is what the measured frequencies actually constrain.
A worked example of a temperature shift
Consider a catalytic electrode for which a DRT feature shifts from 0.3 seconds to 0.06 seconds as temperature rises, while its area also changes. It would be tempting to claim that charge transfer became five times faster. That claim is too simple. The characteristic relaxation time moved by a factor of five, but the underlying relation may contain changing resistance, capacitance, and multiple coupled processes.
The improved wording says that the dominant fitted relaxation in the specified region became faster with temperature. To assign the origin, one needs a controlled temperature series, a plausible physical model, stable electrode conditions, and corroborating kinetics. If the peak broadens or splits as temperature changes, fitting just the maximum’s coordinate may further oversimplify the response.
This example teaches the difference between a rate constant and a time marker in a transformed plot. They can be linked through a validated model, but are not synonyms. That distinction is especially important when translating specialised analytical prose for learners from other fields.
A worked example with an inductive artefact
Suppose a high-frequency Nyquist trace bends into an inductive region while a DRT reconstruction adds a narrow extreme-fast feature. The instrument lead arrangement has changed between runs. Before assigning that feature to electron transfer, check the fixture, cable loop, contact quality, and any software inductance compensation. A wiring change can alter the fast end of the measurement without any modification of electrode chemistry.
If the measured spectrum improves when the cables are shortened or arranged consistently, that supports an instrumental origin. If an inductive response persists under carefully controlled setups and changes systematically with electrode potential, a physical electrochemical interpretation becomes more credible. Both possibilities deserve consideration.
The practical language is ‘the high-frequency response may contain inductive contamination, and the associated reconstructed feature should not be assigned without control measurements’. This is more informative than deleting an awkward data region silently or proclaiming a new reaction step on the basis of one tiny peak.
A worked example of film growth with independent evidence
A researcher grows an oxide layer under a controlled potential, then collects EIS at the same test condition before and after treatment. The DRT shows a larger short-time contribution after treatment. Ellipsometry or another appropriate independent measurement confirms that the layer changed. This combination gives more support for relating the DRT variation to surface-film properties than the relaxation plot could supply alone.
Even then, the new layer could affect several electrical pathways at once. It may change ionic transport, dielectric response, interfacial roughness, and reaction kinetics. Calling a particular feature ‘the film resistance’ requires a model able to distinguish those contributions. If the feature’s breadth changes, consider whether the layer has become more heterogeneous rather than only thicker.
A well-written conclusion can connect the independent film observation to an impedance change while explicitly stating the remaining ambiguity. This makes the analysis both useful and scientifically honest. Readers see why complementary methods are not decorative additions but checks on the interpretation of inverse data.
A worked example comparing two instruments
Two laboratories measure nominally identical cells. Their DRT spectra disagree in high-frequency peak shape but align reasonably in the centre. One laboratory uses different cables, a higher maximum frequency, and another regularisation implementation. It would be premature to conclude that cell manufacturing or laboratory technique caused the difference.
A robust comparison starts with overlapping frequency ranges, shared sign conventions, matching temperature and state of charge, and comparable data quality. Raw impedance can be compared before any DRT inversion. If the raw spectra already disagree, investigate measurement conditions. If the raw spectra agree but DRT shapes differ, the inversion and normalisation deserve attention.
This logic prevents over-interpretation of derived graphics. It also offers a useful route for translation between teams: agree on the directly measured quantities first, specify every calculation afterward, and mark claims that depend on the chosen model. A derived plot is a lens, not a replacement for the original experiment.
A worked example where the peak area changes but height does not
A broad DRT feature narrows during a treatment while its maximum ordinate remains roughly unchanged. A superficial description calls the process ‘unchanged’ because the tallest point is at the same height. But if the chosen distribution’s integrated area corresponds to a resistance contribution, narrowing can reduce that contribution despite an unchanged maximum. The opposite can happen when a peak broadens.
Therefore a careful reader inspects width and integrated strength in addition to height and centre. Integration must respect the defined log-time coordinate and numerical normalisation; otherwise the answer may be wrong even when the arithmetic looks tidy. Uncertainty in the peak boundaries should also be reported, particularly when two features overlap.
The general lesson extends beyond electrochemistry. A graph can encode meaningful information in shape, not just height. Translation into prose should preserve all relevant geometric dimensions and state what operation converts them into a physical parameter.
How to describe uncertain peak assignments in a figure caption
A strong caption identifies the sample, state, measurement frequency range, inverse method, and axis conventions before offering mechanism labels. It can say that arrows mark reproducible features at selected characteristic times and that proposed assignments are based on complementary experiments. This gives the reader enough information to know where observation ends and interpretation begins.
Captions should not quietly introduce conclusions that the methods section cannot support. If confidence varies between two assignments, use different wording. For example, a well-established contact-related contribution might be described directly, while a speculative slow process can be described as potentially transport related. If the plots are normalised, state how and to what reference.
Clear captions are part of scientific literacy. They are frequently read before the body of an article, reused in teaching materials, or translated into another language. A caption that preserves units, assumptions, and qualifications can travel much farther without losing its meaning than one that simply announces a dramatic peak.
Questions to ask before interpreting another paper’s DRT
Ask whether the authors provide the raw Nyquist and Bode spectra; whether the electrode was stable during measurement; what frequency range and perturbation amplitude were used; how the time grid was selected; which kernel and regularisation were applied; and whether sensitivity analyses or replicate measurements are reported. Then ask whether labels such as charge transfer, film, and diffusion were established independently or merely inherited from common expectations.
Do not treat the absence of one methodological detail as proof that a study is worthless. It may mean that the claim must be read with limited confidence. Different articles serve different purposes, from teaching a method to reporting a fully validated mechanism. The level of certainty should match the evidence actually shown.
For readers learning scientific English, build a short vocabulary notebook distinguishing ‘measured’, ‘estimated’, ‘consistent with’, ‘attributed to’, and ‘demonstrated’. These terms represent different degrees of inference. Understanding them is as important as understanding ohms and seconds when translating technical claims accurately.
When DRT helps and when another analysis is better
DRT is particularly helpful when the reader suspects overlapping relaxation contributions and wants a timescale-oriented view to guide experiments or circuit selection. It can clarify which regions of a complex impedance spectrum change systematically under controlled conditions. It also gives teachers a way to explain how one measured signal can be represented in several mathematical domains.
DRT is a poor substitute for chemical identification. It cannot directly image reaction products, determine surface composition, or measure a microscopic diffusion coefficient without modelling assumptions. It may be uninformative when a response does not fit the chosen kernel or when the input impedance is unstable, strongly nonlinear, or too sparse. In those cases, improved EIS, another electrode configuration, spectroscopy, microscopy, or time-domain measurements may be more revealing.
Choosing not to interpret an uncertain peak is a legitimate analytical decision. Good technique involves matching the tool to the question. It is better to report a limited but reliable observation than a complete-looking mechanistic map assembled from weak assignments.
Make a reliable plain-language translation of the plot
A reader-facing summary can proceed in a disciplined order. First: state what was measured, including the cell condition and impedance frequency range. Second: identify how the DRT was calculated and whether it describes relaxation-time amplitude, area, or another normalisation. Third: describe the prominent features using characteristic times, shape, and reproducibility. Fourth: separate proposed mechanisms from direct observations. Fifth: explain what evidence would strengthen the assignments.
For example, ‘under matched operating conditions, the DRT analysis shows an increased medium-time relaxation contribution after cycling; its physical origin remains under investigation’ is useful and restrained. It says what the figure supports without pretending that a mathematically inferred feature carries a chemical label attached at birth.
The same approach works across disciplines. Accurate translation preserves entities, measurements, relations, caveats, and the direction of change. It also allows readers to learn new terms without collapsing every unfamiliar idea into a loose everyday analogy. Clear prose makes advanced science more accessible precisely because it keeps the technical boundaries visible.
Related reading and references
Context and meaning in technical translation
Cross-linguistic vocabulary transfer
eduKateSG science learning hub
2024 Joule review of DRT inversion