Data and code from: Nanoscale interfaces suppress fatigue damage in mineralized tissues
Data files
Sep 09, 2026 version files 14.13 GB
-
analyzeRotation.m
1.88 KB
-
calculate_fibril_crack_orientation.m
12.82 KB
-
monotonic_prefracture_tomography.zip
2.15 GB
-
Monotonic_raw_stiffness.csv
387.39 KB
-
Raman_bulk_data.zip
159.66 KB
-
README.md
12.12 KB
-
SEM_corrected_smooth_stiffness.csv
196.71 KB
-
SEM_raman.zip
32.27 KB
-
SEM_raw_stiffness.csv
120.92 KB
-
tildeK0p15_corrected_smooth_stiffness.csv
177.79 KB
-
tildeK0p15_postfatigue_tomography.zip
2.86 GB
-
tildeK0p15_prefatigue_tomography.zip
2.12 GB
-
tildeK0p15_raman.zip
32.22 KB
-
tildeK0p15_raw_stiffness.csv
109.50 KB
-
tildeK0p25_corrected_smooth_stiffness.csv
82.47 KB
-
tildeK0p25_postfatigue_tomography.zip
1.74 GB
-
tildeK0p25_prefatigue_tomography.zip
1.76 GB
-
tildeK0p25_raman.zip
32.10 KB
-
tildeK0p25_raw_stiffness.csv
82.65 KB
-
tildeK0p30_corrected_smooth_stiffness.csv
995 B
-
tildeK0p30_postfatigue_tomography.zip
1.76 GB
-
tildeK0p30_prefatigue_tomography.zip
1.75 GB
-
tildeK0p30_raman.zip
32.38 KB
-
tildeK0p30_raw_stiffness.csv
1.01 KB
Abstract
Nanoscale structural heterogeneity is a ubiquitous design motif in biogenic materials, but how it contributes to the structural longevity of biological systems remains unknown. Using bone as a model hierarchical material, we combine in-situ fatigue loading with synchrotron X-ray tomography and radiography to directly observe crack propagation in real time at 21nm and 100ms resolution. We find that mineralized collagen fibrils decelerate crack growth via branching along fibril axes, while cracks oriented orthogonal to fibrils undergo periodic deceleration at interfibrillar interfaces. Collectively, these mechanisms suppress the rate of damage accumulation under physiological loads by up to an order of magnitude, bridging the timescale gap between mechanical damage and cell-mediated repair. Beyond bone, these findings highlight the separation of length scales between nanoscale interfaces and bulk tissue as a design motif for delaying fatigue and enhancing the durability of biomaterials, implants, and engineered materials.
Dataset DOI: 10.5061/dryad.bzkh189rm
Description of the data and file structure
Dataset supporting the manuscript: "Nanoscale Interfaces Suppress Fatigue Damage in Mineralized Tissues."
This dataset contains experimental data and analysis files used in the study of fatigue deformation and crack propagation mechanisms in mineralized collagen fibrils using in-situ synchrotron X-ray radiography, tomography, and mechanical testing.
The dataset includes:
- Mechanical stiffness data
- Pre- and post-fatigue 3D tomography datasets
- MATLAB scripts used to quantify fibril and crack orientations (γ and θ angles)
- Raman Spectroscopy data
Files and variables
File: SEM_raw_stiffness.csv
Description:
Raw stiffness measurements extracted during fatigue loading prior to indent compliance correction or smoothing. These measurements represent the stiffness evolution during cyclic loading in the SEM loaded beam.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Measured stiffness of the sample at the corresponding cycle.
File: SEM_corrected_smooth_stiffness.csv
Description:
Stiffness values after correction for experimental indent compliance and application of Gaussian smoothing to reduce measurement noise as described in the manuscript.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Sample stiffness after compliance correction and smoothing.
File: tildeK0p15_raw_stiffness.csv
Description: Raw stiffness measurements obtained during in-situ X-ray fatigue experiments conducted at initial applied maximum stress intensity of Kmax ≈ 0.15 MPa√m.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Measured stiffness of the sample at the corresponding cycle.
File: tildeK0p15_corrected_smooth_stiffness.csv
Description: Corrected and smoothed stiffness measurements for fatigue experiments conducted at Kmax ≈ 0.15 MPa√m. The stiffness values have been corrected for experimental indent compliance and smoothed to reduce measurement noise, as described in the manuscript.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Sample stiffness after compliance correction and smoothing.
File: tildeK0p25_raw_stiffness.csv
Description: Raw stiffness measurements obtained during in-situ X-ray fatigue experiments conducted at initial applied maximum stress intensity of Kmax ≈ 0.25 MPa√m.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Measured stiffness of the sample at the corresponding cycle.
File: tildeK0p25_corrected_smooth_stiffness.csv
Description: Corrected and smoothed stiffness measurements for fatigue experiments conducted at Kmax ≈ 0.25 MPa√m. The stiffness values have been corrected for experimental indent compliance and smoothed to reduce measurement noise, as described in the manuscript.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Sample stiffness after compliance correction and smoothing.
File: tildeK0p30_raw_stiffness.csv
Description: Raw stiffness measurements obtained during in-situ X-ray fatigue experiments conducted at an initial applied maximum stress intensity of Kmax ≈ 0.3 MPa√m.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Measured stiffness of the sample at the corresponding cycle.
File: tildeK0p30_corrected_smooth_stiffness.csv
Description: Corrected and smoothed stiffness measurements for fatigue experiments conducted at Kmax ≈ 0.3 MPa√m. The stiffness values have been corrected for experimental indent compliance and smoothed to reduce measurement noise, as described in the manuscript.
Variables
- Cycle: Loading cycle number during fatigue testing.
- Stiffness (N/m): Sample stiffness after compliance correction and smoothing.
File: Monotonic_raw_stiffness.csv
Description: Raw stiffness measurements obtained during an in-situ X-ray monotonic loading experiment used as a comparison to cyclic fatigue experiments.
Variables
- Time (s): Elapsed time during the monotonic loading experiment after contact of the tip to the beam.
- Stiffness (N/m): Measured stiffness of the sample at the corresponding time point.
File: monotonic_prefracture_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset acquired from the monotonic loading specimen prior to fracture. The dataset contains reconstructed volumetric images used to visualize the mineralized collagen fibril microstructure. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p15_prefatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.15 MPa√m prior to cyclic loading. The dataset represents the initial microstructure before fatigue damage accumulation. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p15_postfatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.15 MPa√m after cyclic loading. The dataset captures final crack propagation and microstructural changes induced by fatigue. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p25_prefatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.25 MPa√m prior to cyclic loading. This dataset represents the initial microstructure before fatigue damage. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p25_postfatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.25 MPa√m after cyclic loading. The dataset captures final crack propagation and fibrillar deformation associated with fatigue damage. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p30_prefatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.30 MPa√m prior to cyclic loading. The dataset represents the initial microstructure before fatigue loading. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: tildeK0p30_postfatigue_tomography.zip
Description: Three-dimensional synchrotron X-ray tomography dataset of the fatigue sample tested at Kmax ≈ 0.30 MPa√m after cyclic loading. The dataset shows crack propagation following fatigue loading. The data are provided as a stack of TIFF images within a compressed ZIP archive. The reconstructed voxel size is 21 nm, corresponding to a slice spacing of 21 nm between consecutive images.
File: Raman_bulk_data.zip
Description: Raman spectroscopy data acquired from the bulk trabecular bone sample used in this study. The dataset contains raw spectral data provided in CSV format and was used to assess the compositional variability of the parent tissue, as described in the manuscript.
Variables
- Raman Shift (cm⁻¹): Raman shift is reported in reciprocal centimeters (cm⁻¹).
- Intensity (a.u.): Measured Raman signal intensity at each Raman shift value, normalized to maximum intensity, reported in arbitrary units (a.u.).
File: tildeK0p15_raman.zip
Description: Raman spectroscopy data acquired from the fatigue-tested microbeam specimen subjected to cyclic loading at Kmax ≈ 0.15 MPa√m. The dataset contains post-fatigue Raman measurements provided in CSV format and was used to evaluate compositional changes due to x-ray dosing.
Variables
- Raman Shift (cm⁻¹): Raman shift is reported in reciprocal centimeters (cm⁻¹).
- Intensity (a.u.): Measured Raman signal intensity at each Raman shift value, normalized to maximum intensity, reported in arbitrary units (a.u.).
File: tildeK0p25_raman.zip
Description: Raman spectroscopy data acquired from the fatigue-tested microbeam specimen subjected to cyclic loading at Kmax ≈ 0.25 MPa√m. The dataset contains post-fatigue Raman measurements provided in CSV format and was used to evaluate compositional changes due to x-ray dosing.
Variables
- Raman Shift (cm⁻¹): Raman shift is reported in reciprocal centimeters (cm⁻¹).
- Intensity (a.u.): Measured Raman signal intensity at each Raman shift value, normalized to maximum intensity, reported in arbitrary units (a.u.).
File: tildeK0p30_raman.zip
Description: Raman spectroscopy data acquired from the fatigue-tested microbeam specimen subjected to cyclic loading at Kmax ≈ 0.30 MPa√m. The dataset contains post-fatigue Raman measurements provided in CSV format and was used to evaluate compositional changes due to x-ray dosing.
Variables
- Raman Shift (cm⁻¹): Raman shift is reported in reciprocal centimeters (cm⁻¹).
- Intensity (a.u.): Measured Raman signal intensity at each Raman shift value, normalized to maximum intensity, reported in arbitrary units (a.u.).
File: SEM_raman.zip
Description: Raman spectroscopy data acquired from the fatigue-tested microbeam specimen subjected to cyclic loading in the SEM. The dataset contains post-fatigue Raman measurements provided in CSV format and was used as a control to evaluate compositional changes due to x-ray dosing.
Variables
- Raman Shift (cm⁻¹): Raman shift is reported in reciprocal centimeters (cm⁻¹).
- Intensity (a.u.): Measured Raman signal intensity at each Raman shift value, normalized to maximum intensity, reported in arbitrary units (a.u.).
Code/software
File: calculate_fibril_crack_orientation.m
Description: MATLAB script used to determine the orientations between collagen fibrils and crack propagation paths observed in tomography datasets. The script calculates the angular parameters γ (gamma) and θ (theta) as defined in the manuscript to quantify crack deflection relative to fibrils and interfaces.
File: analyzeRotation.m
Description: MATLAB script used as the function that determines fibril orientations using a rotating profile line.
Software:
MATLAB: MATLAB was used to analyze tomography-derived geometric measurements and calculate fibril and crack orientation parameters.
Dragonfly: Dragonfly was used to visualize and analyze the reconstructed 3D tomography datasets.
Human subjects data
The datasets associated with this repository were generated from human bone tissue obtained through a commercial tissue bank (Articular Engineering, Inc.). All data have been fully de-identified prior to receipt by the investigators and contain no personally identifiable information (PII) or protected health information (PHI). The repository includes only experimental data (e.g., imaging, mechanical testing, and Raman spectroscopy measurements) and does not contain information that could be used to identify the donor. The donor's age and sex are reported solely to provide relevant biological context for the study.
Sample fabrication
Bulk human trabecular bone specimens were acquired from a tissue bank (Articular Engineering, Inc.) from a 69-year-old female donor with no history of bone disease. Bone was harvested from the proximal epiphyseal region spanning the femoral head to the underlying distal subchondral trabecular bone. This region consists predominantly of trabecular bone immediately beneath the articular surface. Samples were sectioned into blocks of approximately 20 ×20 ×10 mm3 and shipped fresh in sterile media (50:50 DMEM: F12 with 10% FBS) to maintain tissue viability and mechanical properties prior to testing.
Samples were cleaned by successive washes in isopropanol (IPA) and deionized (DI) water, polished to a precision of 1 µm, and coated with a thin (∼75 𝑛𝑚) conductive Pt–Pd layer to mitigate charging effects. Microscale (50x10x10 µm3) beams were fabricated using the Focused Ion Beam – Scanning Electron Microscope (FIB-SEM) at Drexel University, operating at 30 keV accelerating voltage. The beams were oriented such that lamellae were predominantly parallel to the beam span, replicating physiologically relevant crack-arrest orientations observed in bone. Each beam included a small notch as an initial flaw. Through-thickness starter notches were patterned with an initial length 𝑎𝑜, with 𝑎𝑜/𝑊=0.35−0.4 for all specimens, where W is the specimen width. The use of FIB allowed control of notch geometry and high reproducibility of notch dimensions. These beams were subsequently extracted via lift-out and mounted onto silicon supports (2x2 mm2 wafers) using ion beam-induced platinum deposition glue, forming a microscale three-point bending geometry (Fig.1B).
In-situ loading setup and imaging
Samples were transported to the Brookhaven National Laboratory (BNL) synchrotron facility for in-situ X-ray imaging at beamline 18-ID, employing monochromatic X-rays at 9.4 keV and obtaining a resolution of 21 nm. Nanomechanical fatigue experiments were conducted using an in-situ testing apparatus (KLA NanoFlip) equipped with a spherical diamond tip. The microscale beams were initially loaded at a normalized stress intensity, K̃ = Kmax / KIC = 0.15, followed by a constant amplitude cyclic loading (1 Hz, R=0.5) to drive fatigue crack propagation. The loading frequency (1 Hz) used here is within the physiological range of human gait and daily activity, and is therefore representative of in vivo loading rates. The load amplitude was selected to achieve controlled, subcritical crack-driving forces (small fractions of 𝐾𝐼𝐶), consistent with fracture mechanics-based fatigue testing approaches in which the driving force is the primary control variable (as is common in ASTM fatigue frameworks). Because crack growth per cycle depends on the applied driving force, the effective number of cycles is inherently coupled to the chosen amplitude. Mechanical parameters including load, displacement, stiffness, and cycle number were continuously recorded throughout testing. Concurrently, radiographic images were captured every 100 ms to visualize real-time spatial and temporal changes in mineralized collagen fibril structure near the crack tip. High-resolution 3D tomographic scans (21 nm resolution) were acquired before and after fatigue loading to quantify structural alterations. These same fatigue load experiments were repeated in the SEM to account for the effects of X-ray irradiation.
Crack length calculation.
The effective crack length, Δ𝑎, was calculated using the contact stiffness measured during the duration of the experiment. Using a compliance calibration procedure, the stiffness at the start of the experiment (cycle 1) was identified as the maximum measured stiffness, and inverted to determine the bone specimen’s initial compliance. The main assumption is that the minimum of the measured dynamic compliance represents the initial effective compliance, 𝐶0, of the fabricated specimen with an initial crack length 𝑎0.
The relationship between the load-line compliance, 𝐶, and the normalized crack length,𝑎𝑖/𝑊, for a three-point bend fracture specimen was calculated using:
𝑎𝑖/𝑊 = 0.997 − 3.58𝑈 − 1.51𝑈2 − 110𝑈3 +1232𝑈4 − 4400𝑈5
where
𝑈= 1/ ((𝐹𝐶)1/2 +1)
Here, 𝐹 is a calibration constant specific to each specimen, determined using initial geometry and compliance conditions. This is done by setting𝑎/W = 𝑎0/𝑊 and 𝐶= 𝐶0 for each specimen. The instantaneous crack extension is then obtained by solving for 𝑎𝑖/𝑊 using the measured compliance 𝐶𝑖.
This process allows for continuous tracking of crack propagation as the specimen is subjected to cyclic loading, providing crack extension data over the entire experimental duration. This smoothed crack extension data was used to calculate crack growth rates, 𝑑𝑎/𝑑𝑁, and stress intensity factor values, 𝐾𝐼.
Radiography and Tomography processing
Radiography images were acquired during fatigue loading at a temporal resolution of 100 ms per frame. To ensure uniformity across images, flat-field and background correction were applied using the following normalization formula: 𝑁𝑜𝑟𝑚 = (𝑖𝑚𝑎𝑔𝑒−𝑑𝑎𝑟𝑘) / (𝑓𝑙𝑎𝑡−𝑑𝑎𝑟𝑘). Corrected radiographs were used to calculate the crack mouth opening displacement (CMOD)
Pre- and post-fatigue X-ray tomography was performed on microscale beams to capture the 3D arrangement of mineralized collagen fibrils and crack propagation relative to the introduced notch. Tomographic datasets were reconstructed and visualized using open-source software packages including TXM Sandbox, Dragonfly, and 3D Slicer.
Crack Segmentation
Crack paths were segmented from reconstructed tomography data using intensity thresholding in 3D Slicer, leveraging contrast differences between cracks and surrounding bone matrix. Threshold values were iteratively optimized to isolate cracks accurately while minimizing artifacts. Resulting binary masks were converted to fitted 2D surface meshes along the crack’s main vertical axis (aligned with the maximum tensile stress direction) and any oblique crack branches. These fitted surfaces were used to quantify crack path orientation. Cracks identified in the pre-fatigue tomography were labeled as pre-existing. By overlaying pre- and post-fatigue 3D reconstructions, these pre-existing cracks were traced in the post-fatigue dataset. Newly formed cracks were categorized either as the main crack or as crack branches. The main crack was defined as the dominant fracture extending along the notch axis, while branches were defined as cracks propagating obliquely from the notch in the xz-plane.
Fibril Orientation Measurements
Reconstructed tomography volumes were subdivided into smaller volumes of size 2.5 µm × 2.5 µm × 0.2 µm. For each subvolume, an intensity profile was extracted by sweeping a horizontal line through the center of the medial xy-plane, rotating it incrementally through a range of angles. For each rotation angle, the number of local maxima in the intensity profile was counted using a minimum peak width of 50 nm and a minimum peak prominence of 0.1. The angle that produced the greatest number of peaks was taken to be perpendicular to the in-plane collagen fibril orientation. In cases where multiple angles produced the same number of peaks, the angle with the smallest mean peak width was selected. After determining the fibril orientation, half-prominence valley widths in the intensity profile were measured to characterize the size of low-intensity interfibrillar material where crack growth rate accelerates. To evaluate crack deflection relative to fibril organization, the angle between the crack face normal and fibril orientation was quantified. Specifically, the crack-fibril alignment metric, 𝛾, was defined as 𝛾=1 −|n·f|, where n is the unit normal vector of a crack face (extracted from fitted 2D mesh facets), and f is the unit vector representing fibril orientation in the nearest subvolume. Similarly, the crack orientation relative to the applied loading direction was defined as 𝜃=1 −|n·e𝑦|, where e𝑦 is the unit vector along the loading axis.
Raman spectroscopy
Raman spectra were acquired using a Horiba Raman system (Singh Center for Nanotechnology) equipped with a 532 nm laser and a 600 grooves/mm grating. A 50x objective lens was used to focus the laser on the sample surface, with each spectrum collected using an acquisition time of 10 seconds and averaged over 30 accumulations to improve the signal-to-noise ratio. A third-order polynomial was fit to the background fluorescence and subtracted to perform baseline correction. The resulting spectra were then smoothed using a five-point moving mean average. Each spectrum was normalized to the intensity of the 𝐶𝐻2 peak, and intensity ratios between the 𝐶𝐻2 and 𝐴𝑚𝑖𝑑𝑒 𝐼 peaks were calculated. Spectra were collected from multiple regions across the sample surface.