Geometric mean on unit circle

Warm up

The geometric mean of two numbers is the square root of their product. For example, the geometric mean of 9 and 25 is 15.

More generally, the geometric mean of a set of n numbers is the nth root of their product.

Alternatively, the geometric mean of a set of n numbers the exponential of their average logarithm.

\left(\prod_{i=1}^n x_i\right)^{1/n} = \exp\left(\frac{1}{n} \sum_{i=1}^n \log x_i\right)

The advantage of the alternative definition is that it extends to integrals. The geometric mean of a function over a set is the exponential of the average value of its logarithm. And the average of a function over a set is its integral over that set divided by the measure of the set.

Mahler measure

The Mahler measure of a polynomial is the geometric mean over the unit circle of the absolute value of the polynomial.

M(p) = \exp\left( \int_0^1 \log \left|p(e^{2\pi i \theta})\right| \, d\theta\right)

The Mahler measure equals the product of the absolute values of the leading coefficient and roots outside the unit circle. That is, if

p(z) = a \prod_{i=1}^n(z - a_i)


M(p) = |a| \prod_{i=1}^n\max(1, |a_i|)


Let p(z) = 7(z − 2)(z − 3)(z + 1/2). Based on the leading coefficient and the roots, we would expect M(p) to be 42. The following Mathematica code shows this is indeed true by returning 42.

    z = Exp[2 Pi I theta]
    Exp[Integrate[Log[7 (z - 2) (z - 3) (z + 1/2)], {theta, 0, 1}]]

Multiplication and heights

Mahler measure is multiplicative: for any two polynomials p and q, the measure of their product is the product of their measures.

M(pq) = M(p)\,M(q)

A few days ago I wrote about height functions for rational numbers. Mahler measure is a height function for polynomials, and there are theorems bounding Mahler measure by other height functions, such as the sum or maximum of the absolute values of the coefficients.

Related posts