R07: completed numerical contact-law sensitivity study
Project: DESTROYING BOULDERS, Ashley Kalkowski

The frozen 144-integration matrix completed in 1183.997664 seconds. All 144
integrations met the declared numerical constraints. All 18 required traces
were retained at full resolution. There were no failed, invalid, omitted or
unrun matrix rows. This is an uncalibrated beam/contact calculation, not a
physical test of a device or a fracture demonstration.

The separate computational audit within this project passed. It checked all
144 row identities and frozen inputs; all 48 linear-parent replays; all 18
traces and 1,224,000 saved RK4 transitions; all 36 timestep comparisons; all
18 five-to-seven-mode screens; all 24 nonlinear-versus-linear energy-change
records; and the complete artifact inventory, byte counts and hashes.
Audit runtime was 6.233913 seconds. Independent means a separate numerical
implementation, not an outside laboratory or external replication.

At the fine timestep with seven modes, the change in peak first-mode energy
for adaptive versus unprobed fixed control was:

Case                       Linear      Force matched    Storage matched
I4-00 reference             -0.8923%       -0.8907%          -0.8948%
I4-01 lower frequency       +6.4599%       +6.3139%          +5.9269%
I4-02 higher frequency     +10.7743%      +10.9417%         +10.9847%
I4-F-error combined F       -6.8420%       -6.9381%          -7.1945%
I4-H-error combined H       -8.6365%       -8.5758%          -8.5564%
I4-H-zero flight removed   +10.2612%      +10.3864%         +10.3895%

All 18 seven-mode comparisons passed the declared timestep and mode screens;
9 favored adaptive control and 9 favored fixed control. No comparison was
unresolved under the declared sensitivity rule. Signs were unchanged across
the three contact assumptions and across five/seven retained modes in these
selected cases. These screens are not confidence intervals or a proof of
continuum convergence.

Both protocols have a 10 mJ ideal mechanical budget. Adaptive main impact
receives 9.9 mJ, with 0.1 mJ charged for the inherited ideal probe; fixed control
receives 10 mJ. The largest positive relative comparison, +10.9847%, corresponds
to 3.520540 mJ adaptive versus 3.172094 mJ fixed first-mode peak energy: an
absolute difference of 0.348446 mJ, or 3.48446% of the common 10 mJ budget.
It is not a 10.9847% electrical efficiency or fracture improvement.

Changing only the contact assumption changed individual seven-mode, fine-step
target energies by -0.44095% to +4.86673% relative to the linear reference.
Force matching and storage matching are different conventions at one frozen
0.25 mm reference indentation; neither nonlinear law is calibrated to a
measured contact. These values do not rank physical actuator designs.

The H counterfactual changes the adaptive/fixed comparison from a decline to
an improvement when the inherited flight-time error is removed. This result
appears under all three contact assumptions. Removing that error changes both
protocols' actual contact trajectories. It supports investigating delivery
timing in this model; it does not establish the effect in a boulder.

Numerical diagnostics

Maximum whole-energy residual fraction across 144 runs: 2.176090401448029e-6
(0.0002176090401448029 percent; declared threshold 0.5 percent).
Maximum coarse/fine target-energy change: 3.2421142905914024e-6
(0.00032421142905914024 percent; declared threshold 2 percent).
Maximum five/seven-mode target-energy change: 0.0013354976421324771
(0.13354976421324771 percent; declared threshold 2 percent).
Maximum independently reconstructed RK4 transition discrepancy:
1.1102230246251565e-16 with the declared componentwise scaling.

What this establishes

The frozen program produced internally consistent, repeatable numerical
results under the tested passive contact laws, mode counts and time steps.
The unfavorable cases remain in the record. The model does not support a
claim that adaptive timing always improves energy transfer.

What remains to establish

Physical force/indentation calibration, actuator and attachment losses,
measured modal properties, repeated impact/reset costs, three-dimensional
stress fields, crack initiation and propagation, debris containment and actual
specimen fracture are outside this result. No diamond, boulder, steel pole or
mountain was tested. The study does not establish novelty, patentability,
whole-system efficiency, a universal optimum or a manufacturable fracturing
product. A physical force/velocity and contact-calibration experiment is a
necessary next evidence step before using this model to choose real hardware.

Evidence references

Frozen final summary SHA256:
e965d2b5d433a68479c72adbad3435d13b9e7e0ebcd81ae9e1605c8a3fe889a8
Independent final audit SHA256:
7a1a5064ca25a8286f27ff0c80fe18e5bde9eee7288329994660a5a170057064
Audit: INDEPENDENT_RUN_AUDIT.json
Method and reproduction scope: REVIEW-METHOD.txt
Scientific figures: figures-v2/FIGURE-MANIFEST.json
Plot data: figures-v2/contact_law_figure_data.csv

All four figures were inspected visually. The second rendering spaces bar
labels and uses scientific notation for small residuals so nonzero residuals
do not round to displayed zero. Numerical records and raw traces were not
changed. All 68,001 states are passed to each raw-trace plot; contact force
uses a labeled contact-window view and the energy ledger shows the full run.
