Analytical characterization data scope

Purpose

Preserve source-written records that can be used to train and validate machine-learning models and other AI systems, or help people recognize, interpret, or predict analytical characterization. This collection covers analytical records, not every experimental result in a paper.

Admission

Transcribe source-written analytical results from the paper’s own experiments on source-defined compounds, materials, mixtures, samples, states, or series. Each analytical record must identify its subject, method or measured property, result, and the conditions and comparators needed to understand it. It must be usable without consulting an omitted figure or reconstructing a value.

Decide eligibility from the particular quantity or observation, not the purpose of the experiment, its location in the paper, or the method’s name. An analytical result does not become ineligible because it was measured during synthesis, catalysis, degradation, calibration, or device testing. The exclusions below apply to individual results, not whole experiments.

Retain every source-defined analytical block that meets these requirements. In a mixed block, retain every eligible analytical result and sample member while applying the omissions below. Do not cherry-pick values or repeat a narrative restatement of the same data.

Retain

• Measured analytical properties, observed features, and composition explicitly written in prose, tables, captions, legends, or annotations; labels that merely identify a sample, axis, or trace are not results
• NMR peak lists
• Author-written local analytical assignments, including a specific identity, configuration, state, phase, or artefact established by the measurement, with the source’s evidence and uncertainty preserved
• Source-printed routine analytical reference values calculated from a stated composition, formula, or adduct
• Source-printed estimates, fits, ratios, and averages derived from a retained experimental measurement when the source states the basis and any model or assumption needed to understand the value
• Explicit non-detections from completed measurements
• Reaction-monitoring records that name the analytical method and report signals, species, or composition of defined sampled states, including time-resolved concentration, yield, or conversion series; the bare terminal outcomes and performance metrics listed below remain excluded

Omit

• Values, trends, peak lists, identities, or assignments inferred by the transcriber from spectra, plots, images, structures, or graphical position. Inspect rendered pages to verify written content; do not reconstruct results from graphics
• Simulations, predicted sample properties, calculated spectra or patterns, and assignments dependent on those predictions. This does not exclude the source-reported analytical estimates and fits allowed above. Retain an independently reported experimental assignment but omit its computational rationale
• Qualitative observations that do not state an observed feature, state, or result; relational observations that also omit the direction or comparator
• Proposed mechanisms, causes, and explanations beyond a local analytical feature assignment
• Biological activity and functional performance, including catalytic activity, sensing performance, energy storage or delivery, separation throughput, device efficiency, and durability metrics; also instrument, method, workflow, or platform calibration, accuracy, robustness, transferability, and optimization results. Time resolution or use of an analytical method does not make a performance metric eligible. Retention times of named standards, protein concentrations, and post-cycling phase compositions are analytical results; method-transfer accuracy, enzyme activity, and capacity retention are performance results
• Synthesis, work-up, purification, isolated yields, yield or conversion comparisons used to screen or optimize reaction conditions, bare terminal yield, conversion, or consumption outcomes, incidental isolation appearance, and preparation details beyond those needed to identify a retained measured state. An analytical result printed beside a preparation, such as an Rf value, remains eligible
• Literature-only comparator data and generic statements that identity, purity, structure, or literature agreement was confirmed without a textual analytical result
• Standalone database links, accession codes, and deposition numbers; if the source itself reports an eligible analytical result associated with one of these identifiers, retain the identifier—not externally hosted database content—as context within that result’s record
• Measurements that were unavailable, unacquired, or unsuccessful and yielded no analytical result; describe missing or untranscribed data in coverage notes as specified below
• Anything inferred, newly calculated, reconciled, or scientifically interpreted by the extractor

Context and coverage

Keep source-explicit conditions, calibration references, replicate counts, precision, limitations, and reporting qualifiers attached to an eligible result. They need not qualify as independent results. Preserve local conclusions tied to concrete analytical evidence; do not remove a clause merely because it says “confirmed,” “control,” “assay,” or “reproducibility.” Omit standalone method-performance claims and broader mechanistic explanations.

A single sampled-state record must report an analytical feature, species identification, or composition beyond a yield or fraction converted or consumed. Naming NMR and adding reaction conditions to “80 % yield after 2 h” does not make it eligible. An assigned resonance, product-to-byproduct ratio, or inventory of detected components can qualify, including at the reaction endpoint or within a screening experiment; the accompanying screening comparison remains excluded.

For ratios and percentages, retain the source’s reporting basis or denominator. A retained composition record may include component amounts and their total on a feed-normalized basis; do not relabel these as fractions of the analyzed mixture. A standalone process-yield or recovery result remains excluded.

Preserve useful notes that data were not found or that image data were not transcribed, including sample-specific placeholders with no transcribed result. These document coverage and are not analytical results. Distinguish a completed non-detection from a missing result, an untranscribed image, or an unavailable source. A page containing only coverage notes has no retained analytical data.

Representation

Collection-authored captions and row labels identify the source owner, method, and conditions without adding scientific meaning. Preserve result text, values, units, qualifiers, notation, and anomalies without scientific paraphrase. Correct notation only when the source unambiguously identifies the same chemical species and the correction changes no value, assignment, or uncertainty; otherwise preserve the anomaly. Never recompute a value or resolve a discrepancy.

Use Unicode superscript and subscript characters for typographic scripts. For example, write 76 m² g⁻¹, cm⁻¹, 10⁵, and CO₂, not baseline stand-ins such as m2 g-1, cm−1, or 10^5, and not HTML <sup> or <sub> elements. Apart from an unambiguous correction allowed above, do not rewrite the source notation.

Use one <article> and one <table> for each source-defined sample, state, or explicitly joint series. The <caption> contains its source name or description. Put a source identifier in an Identifier row when present. Other rows contain analytical results, their attached conditions and qualifiers, or explicit coverage notes. A coverage-only article must identify the sample and method, state what was not transcribed or found, and cite its source.

Represent an exact Main/SI duplicate once and cite both sources. Keep differing results separate with unambiguous provenance. Put page provenance in the article <footer>, using physical PDF pages for Main and printed SI page labels when present, otherwise physical SI PDF pages. Do not put page, figure, or table references in captions or method labels.

Cite the DOI in the document header and do not link to local files. If searched data were not found, say what was not found and where, using “was not found” or “were not found.” Describe the verified supporting-information situation precisely. State when it is appended to the main paper. Otherwise, if no separate supporting information file was found among the files available from the ChemRxiv record, state that and, when true, that the main paper refers to supporting information. Name a specific file format only when that format is itself a verified fact. Do not claim that supporting information does not exist. Do not create empty articles or append terminal punctuation to header notes, captions, cells, or footers.