OPPOSING CONTACT LOADS AND A PROPOSED CENTRAL SPLIT
================================================

This is an independent 2D linear-elastic finite-element benchmark motivated by
the proposed two-sided excitation sketch. It answers a narrower physical
question: can opposite compressive contacts generate tensile stress normal to
a central plane? Yes, in this reference geometry. The Brazilian-disk mechanism
is established; this calculation verifies our implementation of that mechanism.
It does not predict fracture or demonstrate acoustic energy focusing.

Reference problem
-----------------
Circular disk: 100 mm diameter, 20 mm thickness.
Arbitrary isotropic elastic constants: E=70 GPa, Poisson ratio=0.2.
Static plane stress; small deformation; homogeneous material; no cracks.
Uniform inward radial pressure over top and bottom arcs, each spanning +/-7.5
degrees around the loading diameter. Each end has a 100 N VERTICAL RESULTANT.
P in the formulas is 100 N, not the 200 N sum of both opposed loads.

Mesh: constant-strain triangles, one quarter reflected into four quadrants to
preserve both symmetries. Contact-arc endpoints align on every mesh. Pressure
loads are consistent nodal edge tractions on the polygonal disk boundary.
Rigid motion is removed by center ux=uy=0 plus rightmost-node uy=0. Those
constraints are compatible with the exact symmetry; negligible reaction forces
are checked so they do not secretly supply additional loading.

Central analytic comparison
---------------------------
For narrow diametric loading, with tension positive:
    sigma_x = 2P/(pi*D*t), sigma_y = -3*2P/(pi*D*t).

The uniform-pressure finite-arc center solution used here is independently
obtained by integrating rotated diametric point-load stress tensors:
    sigma_x = [2P/(pi*D*t)] * [sin(2alpha)-alpha]/sin(alpha)
    sigma_y = -[2P/(pi*D*t)] * [sin(2alpha)+alpha]/sin(alpha)
where alpha is the contact half-angle IN RADIANS. The pressure/resultant
normalization is P=2*p*R*t*sin(alpha). Other presentations sometimes normalize
P to the scalar arc-integrated load 2*p*R*t*alpha; mixing those definitions
changes the finite-arc correction. Tests include direct numerical integration
of the rotated point-load expression and its concentrated-load limit.

The reported FEM center estimate is an area-weighted average of stresses in
triangles sharing the center node. That region's radius is R/nr and shrinks
with mesh refinement. It is not falsely described as a pointwise stress at a
node (CST elements have constant stress inside each element).

Results and verification
------------------------
Mesh nodes: 193 -> 769 -> 1729.
Central sigma_x error versus finite-arc solution: 10.1% -> 1.84% -> 0.594%.
Finest sigma_x: +31.010 kPa, normal to the proposed vertical central plane.
Finest sigma_y: -94.703 kPa, compressive along that plane/loading direction.
Finite-arc analytic sigma_x: +31.195 kPa; sigma_y: -95.039 kPa.
Narrow-load analytic sigma_x: +31.831 kPa.

Eleven tests cover constant-strain patch reproduction, rigid-motion zero strain,
geometry convergence, independent analytic comparisons, monotone central-stress
convergence, stress signs, force/moment balance, negligible support reactions,
mirror symmetry, strain-energy/work identity, and load/modulus scaling.
The reference stores about 11.32 microjoules of elastic energy. The equality
U=0.5*F^T*u assumes a quasi-static linear ramp from zero load. It is NOT measured
electrical input, consumed fracture energy, or a fracture-efficiency result.

Figure interpretation
---------------------
NUMERICAL_disk_stress.svg/.png show a schematic and element stress maps.
Red positive sigma_x means tension across the proposed vertical plane.
Negative sigma_y means compression in the loading direction. The color scales
are labeled separately; the sigma_x scale is asymmetric about zero.
The dashed line is an annotation, not an actual crack or a simulation of a
split. No curved propagation path is assumed from the original sketch.

Reproduce with NumPy only (no plotting dependency):

python efficiency/generate_disk_benchmark.py --skip-plots
python -m unittest discover -s efficiency -p test_disk_stress.py -v

The supplied SVG/PNG are static files already generated and require no Python
plotting package to view. Regenerating figures optionally requires Matplotlib:

python -m pip install matplotlib
python efficiency/generate_disk_benchmark.py

The solver and browser app do not depend on Matplotlib. During development,
plotting packages were installed in a temporary directory outside the delivered
project. --plot-deps can optionally identify such a directory. It is not bundled.

Output files
------------
results/NUMERICAL_disk_benchmark.json: assumptions, convergence, balances, limits.
results/NUMERICAL_disk_stress.svg/.png: scientific stress figure.
results/NUMERICAL_disk_element_stresses.csv: per-element centroid and stress.
results/NUMERICAL_disk_mesh_fields.npz: nodes, triangles, displacements, stresses,
and applied nodal forces for independent inspection with numpy.load.
results/disk-stress-test-report.json and disk-stress-test-log.txt: actual checks.

What remains untested
--------------------
No real specimen was measured, loaded or broken. E and nu are reference values,
not measured quartz/diamond properties. The model has no strength, flaws,
fracture energy, material anisotropy, 3D stress, actuator force limits, timing,
waves, resonance, damping, fatigue, adhesive, jaw compliance or heat transfer.
Central-stress convergence does not validate contact-edge peak stresses.

To turn this mechanism into a specimen prediction requires measured geometry,
anisotropic material properties where appropriate, real contact loading, a
calibrated crack-initiation/propagation model, and withheld physical tests. The
known compressive-contact mechanism does not establish that two sticker-sized
sources can generate the required forces or an arbitrary selected fracture path.

Primary research sources
------------------------
Wessling et al., A statistical DEM approach for modelling heterogeneous brittle
materials (2021; published volume 2022), section 2.1. The paper documents the
Brazilian test mechanism and explains why actual crack initiation can depart
from the ideal center. Its fracture model is not reproduced here:
https://link.springer.com/article/10.1007/s40571-021-00434-w

Numerical modelling of the contact condition of a Brazilian disk test and its
influence on the tensile strength of rock (2017), finite contact-arc discussion:
https://www.sciencedirect.com/science/article/abs/pii/S136516091730045X

These references substantiate the physical benchmark. Agreement with an
analytic elastic solution verifies numerical implementation; it does not
validate predictions of actual fracture.
