From your first spectrum to the report.

Videos: learn, pause and repeat

Each workflow has a separately narrated Spanish and English movie, on the website and in WPF packages that include the player. No copy registration is required to read this public help.

1. Open Workflow videos and choose Prime or ResinAtOnce. Use the website language: Español plays Spanish audio; English plays English audio with its corresponding interface, poster and chapters.
2. Select Play; there is no autoplay. Use chapters to return to a step and CC to select Spanish or English captions synchronized with that movie.
3. Expand Steps, provenance and transcript to read or download the text. Changing captions alone does not change narration; change the page or WPF-player language for another voice.
4. In the application open Vídeos de uso / Workflow videos from Help and select Español or English. If an older copy does not include these media, use the web gallery; do not download executables from unrelated links.
5. Play the step, pause the video and practice in a copy of your project. Check the displayed version: buttons can be positioned differently in an earlier executable.

• From project to DOSY map →
• Peak attenuation and diffusion →
• Interpret context with Beatrix →

2D field and lateral-projection components

The spatial Gaussian dictionary reconstructs the 2D view; its atoms are not the lateral-projection component count. Several ppm signals can contribute to the same lateral lobe.

1. Select DOSY or MwOSY and Original or Model to decompose the projection. Use Residual only as a signed diagnostic, not as a positive mixture. Check protocol and source; the total blue curve is not an individual Gaussian.
2. New maps automatically use lateral-modes-v1. For a saved map use Recalculate lateral-projection components and preserve the new decomposition as an explicit review. Opening a file does not replace its historical model.
3. The fit detects lobes by prominence and local support and fits the unsmoothed active projection in log10(D) or log10(M). Three supported lobes can yield three components; three are not always imposed, nor is every 2D atom drawn as another lateral component.
4. Review total, fitted sum, components, signed residual and positive unassigned signal on the same scale. If a broad curve covers both peaks and valleys, compare local support and Original view before accepting it.
5. Advanced settings allow review of prominence, component limit and exploratory manual fitting. The standard maximum is ten, not ten mandatory components. The historical dictionary-marginal-v1 diagnostic remains separate from lobe counting.
6. Select a component to inspect its ppm contributions. Recalculating the projection alone neither reinverts the signal nor changes D or reapplies standards, temperature or molecular calibration. Cancelling retains the previous decomposition.

Export DOSY, MwOSY and attenuations to PNG

Save a high-resolution figure without capturing the whole screen or recalculating the analysis. The PNG preserves the visible framing, representation and theme.

1. Open the completed plot. Select DOSY or MwOSY, Original/Model/Residual, standards, component and zoom. Reset zoom first to show everything; export does not change it automatically.
2. In Prime select PNG · DOSY / MwOSY… on the map. In the peak window use PNG beside attenuation, ln(I/I₀) or the calculated distribution.
3. In Local signal · TRAIn / compare, after calculation, use PNG · attenuation… or PNG · distribution…. A pending plot is not exported as a result.
4. The proposed size is 4800 pixels wide at 300 dpi. Choose 2400/150, 7200/600 or a custom size. Height retains the visible plot’s proportions; resize the window first for a taller figure.
5. Review pixel dimensions and physical size before saving. Limits are 512–12000 pixels wide, 72–1200 dpi, 16000 high and 64 million total pixels. Increasing dpi alone does not create more pixels.
6. Choose filename and folder; confirm only if you want to replace an existing PNG. Cancelling creates no figure. For an empty area or excessive-size warning, open and resize the plot, then retry with an admissible size.

Distributions, moments and uncertainty: what to report

Review the distribution and its contract, not just a central D. The shared workbench supports method comparison and saved reviews without overwriting accepted results.

1. Open Scientific workbench from the local signal to retain the same region, observations and factors. Compare methods with identical inputs and protocols, not a local signal against a global band summing other ppm positions.
2. Check whether values are integrated bin amplitudes or densities per linear or logarithmic unit. Record units, boundaries, orientation, support and threshold. Transforming all D→M preserves integrated weights; only density transformations require a Jacobian.
3. Declare number, mass or unknown signal-response weighting. Mn, Mw and dispersity depend on that declaration; M(central D) is a different quantity. Mean Rh also generally differs from Rh at mean D.
4. Compare full workbench support (0%) with historical summaries that may use 1%. The figure’s visual threshold is different. Do not change historical protocols to force agreement or interpret visual Gaussian area as concentration.
5. In Uncertainty document fit, gradients, standard, temperature, viscosity and calibration, with covariances and shared errors where applicable. Empty means unknown: residuals do not automatically replace independent sigma. Stability is sensitivity, not a confidence interval.
6. Explicitly accept and apply a compatible review, keeping the previous one. For a whole-map family profile review equation, units and domain; applying it creates a derived review. Save the project and open the report and exported CSV/contracts to check their review and support.

Resolve a warning and report a reproducible problem

Keep the complete message and context before changing parameters. An access, drawing or export problem is not necessarily a solver failure.

1. If Continue is disabled, read registration status. A pending request requires approval of that copy; a support ticket or researcher account does not authorize it. Check status again after approval.
2. If Copy identifier fails with OpenClipboard, select and manually copy the identifier where possible. Do not change installation identity or delete its files to attempt registration again.
3. If no attenuations are accepted, review phase, noise, signed signal, useful points, b and exclusions. Keep the previous result and attempt diagnostics; do not relax criteria just to produce a figure.
4. If editing and result differ, restore saved settings or calculate a new review. Do not mix another ramp’s result with the editable selection. Retain non-convergence or unknown-noise status in the report.
5. In Support and suggestions specify Prime or ResinAtOnce, version, steps, ramp/ppm region, method, parameters, factors, literal message and screenshot. Attach only a minimal authorized copy, without keys, tokens or personal or confidential experimental data.

Open Support and suggestions → · Read frequently asked questions →

A selected peak: attenuation, distribution and linear fit

First choose what you want to estimate. A monoexponential fit, a local distribution and a global-projection band are different results; they need not give the same value.

1. Double-click the upper spectrum. Prime marks the measured local maximum and opens its attenuation. Check the marked ppm and extraction half-width: the same region is extracted throughout the ramp, without following each spectrum’s maximum. To compare an exact map bin, double-click inside the map or use Local signal / compare.
2. Check standards, temperature and medium before calculating. This panel retains a fixed snapshot of the conditions. To change them, close it, open Prepare DOSY / MwOSY and select the signal again; Rh requires a reviewed viscosity.
3. Calculate D / Mw and distribution runs the local inversion using the selected method. D only · no distribution computes a monoexponential fit and shows its D as a line: it does not create a distribution or dispersity. Review method, grid, convergence and residual before retaining it.
4. Open ln(I/I₀) · linear fit and points. The all-valid-points fit and selected-points fit show slope, intercept, R² and D. NaN/Inf, invalid b and nonpositive intensities are documented as inadmissible for the logarithm; they are not clipped or removed from the original inversion.
5. To explore exclusions, select Automatic · improve R², or edit Use under Points · manual selection and select Apply manual selection. All valid restores the complete diagnostic. Initial automatic-selection limits retain at least six points and 80% of the b range, removing no more than 20%; review limits and reasons, not only final R².
6. Add to results and Save linear fit are separate actions; save the project afterwards. To include a workbench supplement, accept its review, open Scientific reviews, choose the target and justification, and select Apply review explicitly. A local supplement applied to a map does not replace its global inversion.

Review drift without altering originals

Standard values provide a starting point; Advanced only exposes the settings. Alignment requires an independent review and explicit application.

1. Check whether the header identifies standard or custom/imported values. Opening or closing Advanced changes no numbers. Saved settings remain active even when the section is collapsed; Restore standard values is an explicit action for the next calculation, not an update to earlier results.
2. For a conventional DOSY ramp, select Review / align or follow Beatrix’s contextual prompt. Compare interval, anchor experiment, ppm shift, spectral shape and area-versus-b curve. Unresolved means insufficient evidence, not proven absence of drift.
3. Use Advanced · alignment only when you need to review the anchor, intervals or maximum drift. After editing, select Review with these settings. Check signal loss/fill and rejection reasons; do not relax safeguards to obtain a more attractive fit.
4. Select only reviewed intervals, confirm the review and record your reason. Create aligned copy retains the previous project and creates another project. Review standards and conditions and recalculate D and maps on that copy; earlier fits are not transferred and alignments are not accumulated.

SPEN and ultrafast: choose the correct workflow

There are two distinct routes: exploratory ultrafast D–T₂ notebook processing, and guided reconstruction with a documented kernel and calibration. Opening an example or figure runs neither route.

1. Under Examples and references, review DOI, licence, manifest and hash before extracting a package into a new folder. Published SPEN-DOSY data retain their provenance. Marek 1502 is a local example: no public licence is declared and its experimental data are not distributed here.
2. To reproduce exploratory processing of 1502, open Scientific workbench → Data / R&D → Ultrafast CPMG · open SER, then Process notebook / D–T₂. Review digital filtering, cropping, orientation, Gmax and chirp bandwidth; explicitly confirm the assumptions used. Retain acquisition temperature and do not copy TBO calibration from another experiment.
3. Choose TRAIn2D · C# · in-house adaptation or FISTA 2D: they are different methods. Under Advanced, review the selected method’s criterion: TRAIn2D uses term_fac relative to the NNLS residual; FISTA uses its regularization and KKT criterion. Parameter changes require recalculation; a failed method is never automatically replaced.
4. For the physically documented route, open SPEN / ultrafast · guided workflow. Import CPMG SER or a uniform SPEN/EPSI contract and follow Data → Echoes → Phase / axis → Profile / kernel → Inversion → Controls. Supply order and timing, spatial axis, phase, measured profile and compatible kernel before reconstructing. Do not confirm a review you cannot support.
5. Inspect observed signal, prediction, signed residuals, projections and boundary accumulation. In the guided workflow, compare against a conventional DOSY control from the same preparation and conditions, with tolerance declared before comparison. Save the workflow and export result and provenance; adding it to the project/report requires explicit review.

01 · Choose a product and start

Prime is the general-purpose diffusion workspace. ResinAtOnce retains the resin-specific workflow.

1. Open your existing Windows Prime package; check the executable version. This website does not enable a new public download.
2. Open Bruker data or a .prime.json project. Use Synthetic demo only to learn the interface.
3. Follow acquisition → preparation → calculation → review → save → report.

↗ Enlarge

02 · Import a Bruker ramp

The conventional DOSY import route reads processed 1r spectra, retaining amplitudes and metadata.

1. Select Open Bruker and the acquisition with numbered experiment folders. Set ramp, pdata number and points per bin.
2. Review title, nucleus, solvent, sequence, TE, RG, NS, probe and gradients. Do not combine incompatible acquisitions.
3. Use binning 1 to retain full resolution; binning speeds up calculations but changes the averaged region.

↗ Enlarge

03 · Navigate, zoom and change theme

Spectrum, map and projection share context; the side panel holds the controls.

1. Use Theme ◐ for light/dark and Focus F11 to give the map more space. Effects can be disabled.
2. Fit signals frames visible bands; Full range restores the entire axis. Use the wheel to zoom.
3. The visual threshold is at the upper right. Change Original/Model through representation settings.

↗ Enlarge

04 · Probe and b encoding

Probe detection and gradient calibration are separate decisions.

1. Open Probe / calibration. Automatic reads PROBHD; a documented manual label is also available.
2. Retain reviewed b or import a b manifest in s/m². You may also review Gmax or calibrate with a known solute in the same medium and temperature.
3. Preview factors and b by experiment, document the source and apply to a derived copy. Recalculate afterwards.

↗ Enlarge

05 · Solvent, temperature and viscosity

Open Prepare DOSY / MwOSY → Medium / temperature.

1. Check detected SOLVENT and TE. Select an available correlation or enter manual viscosity with source and Pa·s units.
2. Use fixed η at a reference temperature (293 K proposed) or temperature-dependent η. Confirm applicability to your composition.
3. Thermal normalisation of D is a separate option: choose output temperature and law. Check factors before applying.

↗ Enlarge

06 · TTMS and multiple internal standards

Detecting a signal does not mean approving a standard.

1. Open Internal standards and select Detect / recheck. Review position, attenuation, local D and documented Rh.
2. Confirm which substance you added and its region. The laboratory TTMS protocol uses Rh = 3.65 Å, conditional on identity and medium.
3. If expected D is missing, review η(Tref) beside the standard. For multiple standards, use Manage multiple and check compatibility before enabling them.

↗ Enlarge

07 · Calculate without an internal standard

An internal standard is optional, not a prerequisite for D.

1. Disable Apply confirmed standards or choose the no-internal-standard mode.
2. Review b, temperature and remaining assumptions; recalculate.
3. Check WITHOUT INTERNAL STANDARD in the result and report. Standard definitions can remain saved.

08 · Apply corrections without duplication

Prime retains measured D, fReference, fTemperature and their product.

1. Use Check factors to preview. Save parameters does not replace the previous result.
2. Apply and recalculate archives the previous state and recalculates maps and regions with the same settings.
3. Review final D = measured D × fReference × fTemperature and the associated conditions.

09 · Calculate DOSY and select a method

Calculate map uses edited method and conditions; the previous map retains its own.

1. Choose coherent TRAIn, MF-NNLS256, TRAIn-MF, RAI-S, DOME-S, SILT or the local Diffusion-Inversion route for your problem.
2. Review ppm interval, D range, bins, iterations and regularisation. TRAIn uses termFac; it is not the same parameter as λ.
3. Calculate and inspect convergence, residual and attenuations. Use Cancel to interrupt; the last complete result is retained.

↗ Enlarge

10 · Omitted or rejected attenuations

The detector evaluates noise, useful signal and attenuation coverage.

1. Open Attenuations. Check whether a column is continuous, discrete, a standard, excluded or omitted, and read its reason.
2. If no attenuations are accepted, select a diagnostic row to inspect its exact points and noise estimate.
3. Review phase, baseline, region and b; save the diagnostic. Adjust policy only with experimental justification.

11 · Double-click: calculate a specific peak

The marker identifies the local maximum in the lowest-b spectrum; extraction stays fixed throughout the ramp.

1. Zoom into the region and double-click the peak. Check the marker and extraction half-width.
2. Calculate D / Mw and distribution runs inversion. D only gives a monoexponential fit, not a distribution.
3. Review the values and use Add to results; then save the project.

↗ Enlarge

12 · ln(I/I₀) and linear regression

The linear tab is available before inversion.

1. Open ln(I/I₀) · linear fit and points. Read slope, intercept, R², measured/final D and OLS error.
2. Review I₀ and b₀: they come from the positive point at the lowest available b. The horizontal axis is (b−b₀)/10⁹ and the intercept is free.
3. NaN/Inf, invalid b or I≤0 are excluded only from the logarithm, with a reason in the table.

↗ Enlarge

13 · Automatic and manual point selection

Review compares all valid points with the selected subset and keeps exclusions visible.

1. Automatic · improve R² searches within limits: by default ≥6 points, ≥80% b coverage and ≤20% valid-point exclusions.
2. For manual decisions, select rows and press Apply manual selection. All valid restores the baseline fit.
3. Save the linear fit to include it in the project and report, with mask, reasons and comparison.

↗ Enlarge

14 · Distributions, regularisation and components

A 1D distribution describes a specific region; it is not the projection of the entire map.

1. Select local method, grid and parameters. Reset parameters proposes values; it does not validate their optimality for your sample.
2. Compare signal/prediction, residuals, boundaries and sensitivity to λ or termFac. A non-converged iterate is diagnostic.
3. In Separate components, review valleys, descriptive Gaussians, residual and component moments. Original bins are retained.

↗ Enlarge

15 · Compare local TRAIn and DOSY

Use Local signal · TRAIn / compare to inspect the original bin and its settings.

1. Select the same interval, ppm binning, acquisitions, b, method, grid and parameters.
2. First compare measured D and original distribution; then fReference and fTemperature.
3. Distinguish local components, MF rank and global-projection bands. They are different objects.

16 · MwOSY: choose the molecular calibration

Changing the axis does not turn apparent mass into validated absolute mass.

1. In Prepare DOSY / MwOSY → 3 choose D, high-dilution PS, high-dilution PPG, shared concentration or a custom curve.
2. PS and PPG are two high-dilution curves. There is one concentration model, developed with PS; it does not depend on choosing PS/PPG.
3. Review viscosity, temperature, conditions and domain. The shared implementation requires ¹H/CDCl₃ and output near 293 K; other media require a custom calibration.

↗ Enlarge

17 · Weighings and concentration

Only the concentration model needs C = solute mass / solvent mass, in mg/mg.

1. Prime looks for masses with units in the title of the first experiment in the ramp. They may occur anywhere in the text.
2. The first is solute; the second is solute + solvent. Subtract them to obtain solvent. mg/g and decimal comma/point are supported.
3. If information is missing or ambiguous, fields remain empty. Enter both weighings or explicitly review a historical C; zero is not invented.

18 · Gaussian field and active projection

Modelling is descriptive and retains positive unfitted signal.

1. Gaussian modelling + retained signal is enabled by default. Change it in advanced settings.
2. The projection is recalculated from the active representation. Review error, visually added area and model residual.
3. Select a projection band or map cross-peak. Inspect peak, mean and Rh or mass moments for the complete band.

19 · Review and apply alignment

Beatrix flags possible shifts; the application never aligns without review.

1. Open Review / align and compare intervals, experiments, shifts and before/after attenuation.
2. Adjust anchor and limits if needed. Unresolved means the method lacks sufficient evidence to shift that case.
3. Confirm your reason and create an aligned copy. The previous state is archived; review standards and recalculate results.

↗ Enlarge

20 · Convection diagnosis and review

Diagnosis distinguishes detected, not detectable and inconclusive.

1. Select the signal and open R&D → Analyse ramps from this preparation. Compatible ramps and known calibration are required.
2. Read Guide and final correction: noise, sequence moments, stability and sensitivity must meet the requirements.
3. Where eligible, calculate physical inversion or document another control acquisition. Explicitly activate the final revision for that signal.

21 · Beatrix and assignment suggestions

Cards start from the same peak, band and result you are viewing.

1. Hover over a signal to see hypotheses. Click to pin context and open cards and sources.
2. Review nucleus, solvent, accompanying windows, multiplicity and other band cross-peaks. Open the source before assigning.
3. History is labelled historical. Pause prompts or configure local AI/API providers; keys are not part of the project.
4. Local inference requires matching model, corpus and manifest. A checksum mismatch blocks that inference; rules and documentation remain available. Use a matching package; do not disable the check or attribute rule-based guidance to an LLM.

↗ Enlarge

22 · Local library, HTTP and external algorithms

Local Diffusion-Inversion and an HTTP connection are different routes.

1. The local selector runs the bundled C# library without an HTTP server or MATLAB. Review parameters for each method.
2. For HTTP, configure the endpoint and JSON/SI contract. Review data leaving the computer and authorise the request.
3. An external executable requires path, arguments, hash and timeout. Inspect the response before using it.

23 · SPEN, ultrafast and provenance-tracked examples

Each modality has its own data, sequence, contract and limits; no universal reconstruction is claimed.

1. Open Examples and references: SPEN-DOSY and UF-DEXSY include publications, repositories, licences and SHA-256. Local ultrafast example 1502 includes user-supplied raw data and code without claiming a publication of its own.
2. SPEN-DOSY is inspected as a published package; its spatial reconstruction is not a native Prime method. UF-DEXSY studies diffusion/exchange and must not be treated as ppm–D DOSY.
3. For ultrafast D–T₂ use the specific workflow in the following chapter. The user’s new ultrafast DOSY sequence remains untested: do not silently substitute conventional b for its contract.

↗ Enlarge

24 · Batches, projects and reports

The saved result, not a pending edit, is the basis of traceability.

1. In Ramps / batch choose ramps and review each set of conditions.
2. Save the .prime.json project. Reusable settings retain algorithm parameters, not sample identity or transferable calibration.
3. Export report creates a new revision with editable DOCX, HTML, CSV and JSON; PNG from the view, a manifest and completion marker. Check completion before using a batch.

25 · Ultrafast D–T₂: TRAIn2D and FISTA

The native C# module separates diffusion and effective T₂, not chemical shift and diffusion.

1. Open example 1502 or compatible raw data through the ultrafast module. Review nucleus, temperature, echo count, complex points, acquisition order, spatial window and reference profile.
2. Check spatial-transform and b assumptions before inversion. Example 1502 lacks a measured coil profile and uses an exploratory window; its scale is not presented as final quantitative calibration.
3. Choose TRAIn2D or FISTA, review D/T₂ grids, stopping, residual and parameter sensitivity. Save contract, parameters, data provenance and output separately.

26 · Audit and safe recovery

Prime 0.12.5 and ResinAtOnce 0.48.1 share tested scientific-core and calculation-state fixes.

1. After a failed or cancelled attempt, check status and selection before adding results. Do not attribute an earlier graph to the new attempt.
2. Local separation requires a positive increasing D grid and finite nonnegative weights. Invalid data produce a diagnostic, not a negative-area component.
3. Review JSON files use atomic replacement, with a .bak copy when replacing an existing file. Keep the manifest and check the final file; cancellation is not a new result.