1. Dataset overview

This dataset contains 28 CSV files: six figure-specific summary tables and 22 loading-history files. The files describe 11 grain-based granite model configurations, each subjected to uniaxial compression and direct tension, to investigate the effects of explicit mica cleavage, cleavage orientation, and nominal mica content on mechanical response and microcracking.

The simulations were performed using PFC2D version 7.00.161. Numerical results were extracted from saved model states without rerunning the simulations. The loading-history files contain 679,975 recorded loading-stage samples in total. The six summary tables reuse these configurations; they do not represent additional simulations or independent replicates.

The files provide stress–strain histories, selected crack-count histories, peak strengths, derived elastic moduli, and end-of-loading crack statistics. This dataset does not include particle coordinates, individual fracture geometries, native PFC model states, model source code, or third-party experimental comparison data.

2. Model configurations

• baseline_long_axis: 7% nominal mica content, with cleavage aligned with each mica grain’s long axis.
• baseline_no_cleavage: 7% nominal mica content, with explicit smooth-joint cleavage disabled. This configuration is not mica-free.
• content_original_03, content_original_05, content_original_09, and content_original_11: original mica-content series with nominal mica contents of 3%, 5%, 9%, and 11%, respectively, and grain-long-axis cleavage. The 7% member is baseline_long_axis, not an additional simulation.
• orientation_00, orientation_30, orientation_45, orientation_60, and orientation_90: 7% nominal mica content, with prescribed cleavage angles of 0°, 30°, 45°, 60°, and 90°, respectively, measured relative to the horizontal specimen axis. The loading axis is vertical.

Mica percentages are prescribed model input fractions, not independently measured three-dimensional volume fractions. One saved realization is provided for each configuration and loading type. The alternative later mica-content series is not included.

3. File inventory

The following list describes all 28 CSV files individually. Filenames are case-sensitive as written below. All files can be placed in the same directory.

3.1 Figure-specific summary tables: 6 files

Figure numbers refer to the manuscript supplied with the study. Mechanical-property tables contain one row per configuration. Crack-count tables contain one row per configuration and loading type, including both compression and direct tension.

3.2 Loading-history files: 22 files

Each file contains the recorded loading-stage stress–strain and selected crack-count histories for one configuration and loading type. compression denotes uniaxial compression; tension denotes direct tension.

4. File format and variable definitions

All 28 data files are UTF-8, comma-delimited CSV files with a header row. Decimal points use a period. Scientific notation may occur. A blank field indicates that a value is not applicable; it does not represent zero. Files can be read using standard spreadsheet software or CSV-compatible analysis tools without PFC.

4.1 Loading-history variables

All 22 loading-history files contain the following six columns:

Each row is a stored simulation sample, not an independent replicate. The loading-history files do not provide separately recorded parallel-bond or smooth-joint count histories. Final counts for these classes are provided in the figure-specific crack-count tables.

4.2 Mechanical-property variables

The following definitions apply to the three files ending in _mechanical_properties.csv:

4.3 End-of-loading crack-count variables

The following definitions apply to the three files ending in _crack_counts.csv:

In these models, parallel bonds represent grain-boundary contacts, flat joints represent intragranular contacts, and smooth joints represent explicit mica cleavage. Flat-joint counts can represent failed interface elements and should not be interpreted as the number of visually independent macroscopic cracks. In the configuration without explicit cleavage, mica intragranular contacts remain included among flat-joint contacts. All counts are unscaled integers, regardless of any powers-of-ten multipliers used in manuscript figures.

5. Processing and interpretation

Loading-stage selection

Only samples after the saved loading-start cycle are included in the 22 loading-history files. Compression seating samples are excluded. No interpolation, smoothing, baseline stress subtraction, or image digitization was used. Every recorded sample within the selected loading stage is retained.

Tracked and sampled strengths

Tracked strengths are obtained from saved running-maximum variables. Sampled strengths are calculated from recorded loading histories. A sampled maximum may be slightly lower than the tracked maximum because histories are stored at discrete intervals. These are distinct quantities, not contradictory strength measurements.

Derived elastic moduli

The E20_50 moduli are calculated by ordinary least-squares regression with an intercept, fitting stress in MPa against dimensionless strain over stored pre-peak samples between 20% and 50% of tracked peak strength. The fitted slope is divided by 1000 to express the modulus in GPa. These values are explicitly derived quantities; they are not asserted to reproduce the manuscript’s original modulus calculation exactly. Fitted sample counts are not counts of independent simulations.

End-state crack statistics

The crack-count tables describe the final post-peak state, with a configured termination factor of 0.8 of peak stress magnitude. Discrete termination steps can produce a small deviation from exactly 0.8. These tables must not be labelled as peak-state crack distributions.

Reuse across figures

The grain-long-axis baseline is reused in the Figure 6, Figure 9, and Figure 11 summaries. Corresponding repeated rows refer to the same simulations and must not be treated as independent observations. The five mica-content levels consist of four original content-series configurations plus this shared 7% baseline.

6. Scope and limitations

These files support numerical comparisons of mechanical response and crack counts. They do not constitute a complete simulation-regeneration package and do not contain experimental measurements or spatial crack maps. The dataset contains one realization per configuration and loading type and does not establish variability across independent replicates. Trends or thresholds inferred from these discrete configurations should be interpreted within that scope.