Relative formalization in Isabelle/HOL of a result in inverse problems

In the world of physics and engineering, there is a class of puzzles known as inverse problems. Imagine you are standing outside a sealed, opaque box. You cannot see inside, but you can touch the surface and measure how electricity flows across it. By applying a voltage at one point and measuring the resulting current at another, you gather a set of data. The inverse problem asks: based on these external measurements, can you reconstruct the exact shape and material properties of the hidden object inside? This is not just a theoretical game; it is the mathematical engine behind medical imaging, oil exploration, and non-destructive testing of materials. The challenge lies in the fact that many different internal structures can produce nearly identical external signals, making the solution unstable or impossible to find without very specific conditions.

One specific type of material that complicates these puzzles is a composite made of different regions, where the ability to conduct electricity changes abruptly from one region to the next. If these regions are shaped like simple geometric blocks and the material properties within each block follow a predictable, smooth pattern, mathematicians have long suspected that the external measurements should be enough to uniquely identify the internal layout. However, proving this suspicion requires navigating a labyrinth of complex equations and assumptions about how waves and fields behave at the boundaries between these regions. For decades, these proofs have relied on human intuition and the acceptance of certain deep mathematical truths that were too tedious to verify by hand.

A recent experiment by Cătălin I. Cârstea takes a different approach to this old problem. Instead of relying solely on human verification, the researcher used a computer system designed to check mathematical logic with absolute precision. The goal was to take a specific, complex theorem regarding these piecewise materials and translate it into a language the computer could understand and verify. This process, known as formalization, involves breaking down a mathematical argument into tiny, atomic steps, ensuring that every single logical jump is justified by a rule the computer knows to be true. The researcher did not try to prove every underlying fact from scratch, which would have taken years; instead, he built a structure where the main result rests on a foundation of other established mathematical truths, which the computer treats as trusted building blocks.

The work focused on a theorem stating that if two different internal conductivity patterns produce the exact same electrical measurements on the surface of a bounded, three-dimensional object, then those two patterns must actually be identical. The object in question is divided into a finite number of regions, and within each region, the material properties are described by polynomials—mathematical expressions that are smooth and predictable. The theorem holds true provided the regions are arranged in a specific, well-behaved way and the measurements are taken on a sufficiently large portion of the surface.

To achieve this, the researcher used artificial intelligence tools to help translate the dense, specialized language of the original paper into the strict syntax required by the computer. This was not a simple copy-paste job; it required a careful, iterative process where the computer and the human worked together to ensure the meaning was preserved. The computer, acting as a rigorous referee, checked every step of the proof, confirming that the conclusion followed logically from the starting assumptions and the trusted mathematical facts. The entire process, from translation to final verification, was completed in less than a week, a task that would have been prohibitively expensive and time-consuming using traditional methods.

The result is a machine-checked proof that the main theorem is correct, assuming the trusted mathematical facts it relies upon are also correct. The computer confirmed that the logic holds together without any gaps or hidden errors. This does not mean the theorem is a new discovery in the sense of finding a new physical law; rather, it is a new level of certainty about an existing result. The experiment demonstrates that with the help of modern tools, it is now feasible to verify complex mathematical claims in a matter of days rather than years, turning the verification process into a routine step that can be applied to other difficult problems in science and engineering.

The paper also highlights the challenges of this translation process. Because the original mathematical text is written for human readers, it often omits details that a computer needs to make a decision. The researcher had to ensure that the computer's version of the problem matched the human version exactly, a task that involved checking the translation multiple times and using different tools to spot any discrepancies. The final output is a collection of code and logical statements that anyone with access to the computer system can run to see the proof for themselves. This transparency offers a new kind of confidence in mathematical results, one that comes not from the reputation of the author or the peer review process, but from the unyielding logic of the machine.

Ultimately, this work serves as a proof of concept. It shows that the barrier to formalizing complex mathematical research is lowering, making it possible to bring a higher standard of verification to fields that rely on intricate theoretical foundations. While the specific theorem about electrical conductivity is just one example, the method used here suggests a future where the most difficult mathematical arguments can be subjected to the same level of scrutiny as a scientific experiment, ensuring that the foundations of our understanding are as solid as they appear.