Schanuel’s conjecture

Start with 1 and the square root of 2, written √2. No choice of rational weights—positive or negative fractions—can cancel them unless both weights are zero. Yet √2 satisfies a polynomial equation: its square minus 2 is zero. These are different kinds of dependence. Schanuel asks: take any list of n complex numbers with no rational weighted cancellation, then add their exponential values—powers of e, about 2.718, extended to complex inputs. Must the combined list contain at least n numbers with no nonzero polynomial relation? Such a relation is an equation made from sums and products of whole-number powers, with rational coefficients. Complex numbers allow an imaginary part as well as an ordinary real part.[1] [2]

Can taking exponentials supply enough numbers without hidden polynomial relations? Complex inputs may have real and imaginary parts. They are rationally linearly independent if no weighted sum with rational coefficients is zero unless all weights are zero. For any such inputs $z_1,\ldots,z_n$, form the field $\mathbb Q(z_1,\ldots,z_n,e^{z_1},\ldots,e^{z_n})$: numbers obtained from these generators and rational numbers using addition, subtraction, multiplication and division by nonzero numbers. Its transcendence degree is the maximum number of generators that can be chosen without a nonzero rational-coefficient polynomial relation. Schanuel predicts that this count is at least $n$. It does not require all $2n$ displayed quantities to be algebraically independent; inputs can themselves be algebraic.[1] [2]

For the example 1 and √2, the exponential values are e and e raised to √2. Both starting numbers solve rational polynomial equations, so this example belongs to a known theorem. The unresolved statement must work for every allowed complex list, including inputs outside that special case. A theorem covering almost every list can still miss exceptional ones.[1]

Waldschmidt distinguishes this universal assertion from a stronger independence property that holds for almost every tuple in the measure-theoretic sense. “Almost every” permits an exceptional set of measure zero, and does not settle the specified tuples of greatest arithmetic interest. The paper’s elliptic and quasi-elliptic extensions are separate claims. A proof must handle every rationally linearly independent complex tuple, not obtain a count only for algebraic inputs or generically chosen values.[1]

Knowing that two numbers separately satisfy no rational polynomial equation does not tell us whether they have a polynomial relation together. For example, e and e² are individually transcendental but obey the relation “the second is the square of the first.” Schanuel would provide a common rule for how many genuinely independent quantities remain when exponentials are added, including a prediction that e and π have no such joint relation.[1] [2]

The conditional consequence for $e$ and $\pi$ can be seen directly. Choose inputs $1$ and $i\pi$, where $i^2=-1$. Their real and imaginary parts show rational linear independence; their exponentials are $e$ and $-1$. The conjecture requires transcendence degree at least two in the field generated by $1,i\pi,e,-1$. The algebraic quantities $1,-1,i$ contribute no transcendence degree, so $e$ and $\pi$ must be algebraically independent. This remains a consequence conditional on Schanuel, not a proof of that independence. It illustrates how one rule can convert an exponential identity into a prediction about otherwise elusive numerical relations.[2]