GanJiangUser manual

Start by choosing the right function. GanJiang brings observed-pattern analysis, structure calculations and crystal viewing into one workspace.

01

Phase identification

Search for candidate phases from an observed pattern.

02

Identification + refinement

Identify phases, then refine candidate structures.

03

Known-phase refinement

Refine supplied CIF structures against your data.

04

PDF card calculation

Compare theoretical reflections with observed peaks.

05

XRD simulation

Calculate diffraction patterns from CIF / MCIF files.

06

Crystal viewer

Inspect the atomic arrangement in a CIF.

Prepare the inputs for your task and record the actual measurement conditions before configuring the analysis.

Pattern formats

Submit one observed pattern per task. If an instrument-specific file cannot be parsed, export a two-column CSV or XY file from the instrument software. Renaming the extension does not convert the file.

Two-column text typically contains 2θ in degrees and intensity, with one sample per row. Keep angle and intensity aligned. State whether the file contains a header and whether background subtraction, smoothing or normalization has been applied.

Measurement conditions

Record the sample ID, radiation source or wavelength, horizontal-axis units, scan range and instrument parameter file, if available. Software defaults are not evidence of the measurement conditions.

1. Select New task, then choose the required function on the left.
2. Use + in the task input box to attach the specific files. Check their names; do not upload a whole test-data archive as one pattern.
3. Enter the analysis goal, sample ID and known conditions in the same box. You can extend the prefilled request. The model selector does not set the phase-search model or radiation.
4. Send the request. When the confirmation form appears, check the task type, single/multiphase mode, search channels, elements and formula fields.
5. The final values in the confirmation form govern the calculation. Submit the form after editing it; closing or cancelling it does not start computation.
6. Check execution messages. If the task is still awaiting plan confirmation, submit the form instead of uploading the same files again.

Continue the original task when adding information for the same sample. Create a new task for another sample or a different set of settings, and include the sample ID and conditions in its name to keep outputs distinct.

Search for candidate crystalline phases from one observed pattern without structural refinement. Unknown-phase identification does not require a CIF. Leave the formula field empty for tests without formula hints.

Identify a single phase in the attached pattern without refinement.
Known elements: Ca Fe Si O
No formula hints. Use XMatcher and XQueryer-LW in the single-phase configuration.
Radiation: [actual conditions]. Preprocessing: [actual treatment or none].

Check the result

Check channel success before reviewing candidate rankings and validation. Top 1 means the highest-ranked candidate, not an automatically confirmed phase. Without a candidate list, there is no Top 1 to select.

Multiphase means multiple crystalline phases in one sample pattern, not multiple uploaded patterns. Select Phase identification, attach one mixed pattern and choose multiphase mode in the form. Use unknown phase count when the count is not established.

The multiphase form offers these two channels. Keep both default channels or select one for a controlled test. Candidates are pooled for validation. Single-phase XQueryer-LW and XMatcher selections do not replace the multiphase settings.

Enter elements separately from formulas

Identify multiple phases in the attached pattern without refinement or formula hints.
Known elements: Ca Al Si O Ni S
Use XMatcher-AutoMix and XDecomposer in the multiphase configuration.
Radiation: [actual conditions]. Phase count: [unknown or established count].

Identify phases first, then test how candidate structures fit the observed pattern. Supply one pattern and measurement conditions; known elements support the identification stage. Check identification and refinement results separately.

1. Select Identification + refinement, attach the pattern and specify single phase, multiphase or unknown phase count.
2. Submit the initial identification form after checking elements, formula hints and search channels.
3. Review the candidates and their evidence. When the refinement form appears, check the candidate phases or structure files to be used.
4. Select an available refinement engine and specify the instrument file, if any, fitting range, parameter constraints and cycles. Do not release all atomic parameters merely to improve the fit.
5. Confirm refinement and keep the task session. If communication is interrupted, check the original task before starting the same calculation again.
6. Review calculated and observed curves, differences, fit metrics, structure files and validation status, then download the task files.

Identify phases in the attachment, then refine evidence-supported candidates.
Sample: [ID]. Elements: [known elements or unknown]. Radiation: [actual conditions].
Prioritize scale, background and unit-cell parameters; keep atomic occupancies fixed.
Provide the whole-pattern fit, difference curve, fit metrics and final validation status.

For samples with existing structure models, skip phase search and refine the supplied CIFs directly. Attach one CIF for a single phase or one per phase for a mixture, together with one complete observed pattern.

1. Select Known-phase refinement and use + to attach the pattern and all CIFs.
2. Check the formula, cell and atomic information in every CIF. Avoid duplicate structures and omitted phases.
3. Confirm the phase mode, structure list, engine and instrument configuration. Use the corresponding IRF file when supplying a FullProf instrument file.

Supply one observed pattern and one or more known CIF structures to calculate theoretical reflections and compare them with observed peaks. This function does not perform unknown-phase identification or Rietveld refinement.

1. Select PDF card calculation and attach the pattern and all phase CIFs.
2. Check filenames, wavelength and peak detection/matching parameters in the form.
3. Submit and review the theoretical reflection table and observed peak table separately, including assignments and unmatched peaks.
4. Retain phase names and sources when changing visibility, colours or annotations. After editing assignments, export the current version using the report controls.

Initial values are software defaults and must be checked for the dataset. Download theory.csv, peaks.csv, observed.csv, parameters.json or the result bundle. Initial files contain automatic assignments; later manual edits require a new export.

Calculate theoretical powder diffraction from CIF or MCIF files without an observed pattern. Upload one or more structures to compare component patterns and their total.

1. Select XRD simulation and attach the structures; separate files represent separate phases.
2. Check wavelength, 2θ range, sampling step, peak shape, width and mixing parameter in the form.
3. Submit and inspect parsing status for every file. If some files fail, only the successful phases are completed.
4. Adjust visibility, colour, relative display scale and width in the panel. Choose total, overlay or stacked views and inspect the HKL peak table.
5. Use the panel CSV or SVG export for the current view. Changing wavelength, range or sampling step requires a newly confirmed calculation.

Initial downloads include peaks.csv, pattern.csv, parameters.json and a ZIP, using the confirmed settings.

1. Select Crystal viewer on the left to open the structure viewer.
2. Choose one .cif file using the file picker, or drag it into the drop zone if one is provided.
3. Wait for parsing and verify the current filename and structure information. Load the correct CIF when switching samples so the previous model is not mistaken for the current one.
4. Use the view controls to inspect atomic arrangements. Return to XRD simulation when you need a diffraction pattern.

Read elements from a CIF

Open the CIF in a text editor and search for _chemical_formula_structural, _chemical_formula_sum or _atom_site_type_symbol. Element symbols identify species; numbers give stoichiometry. An expanded cell formula may have larger numbers than the simplest formula but contains the same element types.

If parsing fails

Confirm the file is extracted and is actually a CIF, and check that cell parameters and atomic sites are complete. Renaming a CSV pattern or an image does not make it a structure file.

Check execution status first

Save and review

Return through Recent and check the sample ID, attachments, parameters and time. Save original inputs, confirmed settings, output files and error records. Use the download links in the result attachments; screenshots do not replace complete CSV files or result bundles.

View PDF page 13

The software organizes evidence. Do not accept a phase from a single score, coverage value or low residual alone. Data conditions, peak positions, intensity trends, chemistry and review status must agree.

1. Verify inputs. Confirm wavelength, 2θ range, background, elemental constraints and sample metadata.
2. Review candidates. Read the phase name, formula, space group and database ID before the ranking score.
3. Inspect the pattern. Check principal peak positions and investigate unexplained peaks or systematic intensity errors.
4. Read validation. Inspect Inspector, the peak-position check and coverage. “Passed” still requires manual verification.
5. Audit refinement. Compare observed and calculated curves, the difference curve, fit indices and parameter plausibility.
6. Report the result. State only conclusions supported by consistent evidence. Retain inputs, CIFs, the task ID and exports.

View PDF page 14

Select a candidate combination to preview its theoretical peaks and structures. The four combinations below reproduce the manual example. Percentages are relative diffraction contributions.

• C2Li4O6 is a doubled-formula representation of Li2CO3; ClNa represents NaCl. Candidate 3 matches the known two-phase composition.
• Its score is 55.71 and explained intensity is 51.4%, but validation failed. Report it as a candidate only.
• The third phases in combinations 1, 2 and 4 may reflect over-decomposition. Require diagnostic peaks and a chemically credible sample origin before acceptance.

View PDF page 15

• Candidate 1 leads the ranking and passed software review, so it may proceed to manual confirmation. Peak positions are broadly close, but theoretical and observed intensities differ. Passing is not final identification.
• Candidate 2 failed the peak-position check. Candidate 3 has coverage of 0.938 but failed review; coverage must not override validation.
• Confirm allowed elements, radiation or wavelength, diagnostic peaks, material ID, space group and downloaded CIF before reporting the phase.

Selecting a candidate changes the preview of theoretical peaks and structure; it does not change the ranking or scientific conclusion.

View PDF page 16

• Intensity factors change display contribution; global or independent FWHM controls broadening. Neither value is a phase fraction.
• Simulation checks peak positions, HKL indices and overlap; it does not replace identification from experimental data.
• Blue P1 markers represent Li2CO3 and dark-red P2 markers represent NaCl. Position matches support phase assignment, but a normalized display cannot be converted directly into phase content.
• Intensity mismatch may arise from preferred orientation, crystallite size or strain, background treatment and instrument conditions. Use refinement for quantitative evaluation.

View PDF page 17

• The starting model is the user-supplied Mn2O3.cif with 80 atoms. Lattice and volume changes are below 0.03%, indicating a small unit-cell adjustment.
• Principal peaks mostly overlap and the difference curve remains near zero over most of the range. Local residuals at strong peaks still require profile, background and preferred-orientation checks.
• Small cell changes do not prove acceptance. Record Rwp, Rp and χ², and confirm physically plausible parameters without abnormal correlations.

View PDF page 18

Sharp residuals remain at several intense peaks. Review the model, profile, background and preferred orientation.

Acceptance checklist

• Confirm matching space group, cell setting and CIF identity. A dash does not mean zero change.
• Principal peak positions and intensity trends agree; the difference curve has no structured or intense residuals.
• Compare Rwp, Rp and χ² under comparable data quality and the same strategy. Low Rwp alone is insufficient.
• Cell, profile, background, preferred-orientation and atomic parameters remain physically plausible.
• For quantitative analysis, phase fractions sum to approximately 100%. Report uncertainty, constraints and unmodelled phases.

View PDF page 19

GanJiang organizes XRD processing, phase search, candidate review, PDF-card matching, theoretical-pattern simulation and structure refinement. Outputs support an evidence chain; final conclusions still require sample context and manual verification.

Standard workflow

1. Prepare XRD / CIF.
2. Create a task.
3. Review candidates.
4. Run validation.
5. Refine if needed.
6. Export and archive.

View PDF page 20

Confirm the signed-in account and determine the amount, payment method and invoice requirement. Balance, plans, pricing and payment channels follow the current in-app display.

1. Open the account page. Use the account or avatar entry to open the account, quota or top-up page.
2. Select an item. Choose an available plan or amount. Verify currency, validity, scope and refund terms.
3. Confirm the order. Verify account, item, amount and payment channel. Do not expose payment data in chat or screenshots.
4. Complete payment. Follow the payment page and wait for a successful order status. Do not resubmit an unconfirmed order.
5. Refresh quota. Return to the account page and refresh the balance or quota. If missing, check order and payment status first.
6. Retain records. Keep the order number, receipt and invoice information as required by your organization.

View PDF page 21

Open the configuration page

1. Click the gear-shaped Settings entry at the lower left.
2. Select Large Model in the left navigation.
3. Click Add Model in the Custom Models area.

View PDF page 22

Required-field example

View PDF page 23

1. Save. Click Save and confirm the display name under Saved Models. Use Refresh if the list does not update.
2. Return to chat. Close Settings and click the current model name near the send button.
3. Select. Choose the new entry under Custom. Distinguish it from identically named System entries.
4. Confirm. Verify that the target configuration is displayed below the input box.
5. Test. Start a normal chat and send “Reply only: configuration successful.” Confirm a normal response.

Demonstration configuration, not callable

If the workflow selects models automatically, perform the first test in a normal chat that allows manual selection.

View PDF page 24

Capability-validation boundaries

• After the basic text test, separately validate tool use, image understanding and other required capabilities.
• A successful text response does not prove every advanced capability is available.
• Retain original advanced values. On parameter errors, remove nonessential fields first.

View PDF page 25

View PDF page 26

Method 1: use a system model provided with the account

1. If Go to Sign In is shown, sign in first. If already signed in, confirm network access.
2. Select Rediscover when available. If no model appears, sign out and in again. Restart the application if the backend cannot be reached.
3. In a normal chat, open the model selector below the input box and choose an entry under System.
4. Send a short message and confirm a normal response.

Method 2: configure a personal model service (optional)

1. Open Settings → Large Model and select Add Model.
2. Enter the configuration name, Model ID, API Key and Base URL. Set other fields according to the provider, then save.
3. Return to chat, choose the entry under Custom and send a test message.

View PDF page 27

This is the normal environment-preparation process. The application checks required dependencies and starts installation when any are missing.

1. Review which dependency is being processed and wait for completion. Follow the prompts if a system terminal opens.
2. Enter the computer administrator password only when the system requests authorization, not the GanJiang password. Terminal input may remain invisible; press Enter after typing.
3. Return to Settings → Environment Variables and select Detect Again.
4. Confirm that required dependencies are detected. Install optional dependencies when their functions are needed.

If installation fails

• Expand View Detailed Output on the preparation page and record the failure message.
• Open Settings → Environment Variables and retry Install for supported missing dependencies.
• If no terminal is found, run the command shown by the application in the system terminal, then select Detect Again.

View PDF page 28

1. Use the project’s download entry and select the correct operating system and processor architecture.
2. Open the browser download panel and read the complete message for the file.

View PDF page 29

1. Check the email address or phone number, then select Get Code.
2. Distinguish sending failure from non-delivery. If sending fails, check the network and record the error. If sent, check the inbox, SMS messages, spam folder and blocking rules.
3. After a successful request, the button shows a 60-second countdown. Do not submit repeated requests during this period.
4. Reopen GanJiang, enter the account again and request a new code.

View PDF page 30

Provide the following information in one report so that the issue can be reproduced and diagnosed.