Test a conjecture¶
Consider the family
For which values of \(a\) is \(f_a\) starlike in the unit disk? This example starts with a failed case, finds the exact reason, and then solves the whole family.
Run it¶
python examples/research_workflows/conjecture_counterexample.py \
--a 1 \
--output /tmp/gfa-conjecture
What you will see¶
The result marks \(a=1\) as non-starlike and gives the witness \(z=-3/4\). At that point,
so its real part is negative. The SVG shows the same logarithmic derivative, and the JSON records the exact witness.
Solve the whole family¶
Away from the removable value at zero,
If \(0<a\leq1/2\), then
for every \(|z|<1\). The endpoint \(a=1/2\) is included because the inequality is strict inside the disk.
If \(a>1/2\), the derivative vanishes at \(z=-1/(2a)\) inside the disk. Such a function cannot be univalent, and therefore cannot be starlike.
The corrected statement is
Try the endpoint:
python examples/research_workflows/conjecture_counterexample.py \
--a 1/2 \
--output /tmp/gfa-conjecture-endpoint
A failed sufficient test is not a counterexample¶
Some coefficient criteria are sufficient but not necessary. If such a test fails, it has learned nothing about functions outside the criterion. The Koebe dilation \(z/(1-z/4)^2\) is starlike even though the coefficient sum used by one of the package's simple tests exceeds its threshold.
By contrast, the negative value of \(\operatorname{Re}(zf'/f)\) at an interior point is a direct counterexample. The workflow keeps those two outcomes separate.