Pretentious number theory refresher

“Pretentious” number theory replaces some zero-driven arguments by measuring how closely a multiplicative function pretends to be a simple model (typically \(n^{it}\) or a Dirichlet character). It is especially effective for comparative experiments: which model explains the data best?

Core definitions

Let \(f,g\) be multiplicative functions bounded by 1 in magnitude on primes. The pretentious distance up to \(x\) is

\[ \mathbb{D}(f,g;x)^2 =\sum_{p\le x}\frac{1-\Re\big(f(p)\overline{g(p)}\big)}{p}. \]

Interpretation:

  • \(\mathbb{D}(f,g;x)\) small means \(f(p)\) and \(g(p)\) “agree” on most primes (in a \(1/p\)-weighted sense).

  • \(\mathbb{D}(f,g;x)\) large means \(f\) cannot be explained well by \(g\) on primes.

A common model family is \(g(n)=\chi(n)n^{it}\) with a Dirichlet character \(\chi\) and real \(t\).

What experiments usually visualize or measure

  • Fit \(t\) to minimize \(\mathbb{D}(f,n^{it};x)\) and plot the best-fit \(t(x)\).

  • Compare distances \(\mathbb{D}(f,\chi;x)\) across characters \(\chi\) to see which congruence structure matches \(f\).

  • Use the distance to explain why partial sums of \(f(n)\) may be large or small.

Practical numerical caveats

  • Always define the prime cutoff (use the same \(x\) for fair comparisons).

  • Distance is dominated by small primes; if you want to study “large-prime behavior,” consider plotting contributions by prime ranges.

  • When fitting \(t\), the objective can have shallow minima; report robustness (e.g., multiple initial guesses).

References

See References.

[Granville, 2009, Granville and Soundararajan, 2007]

Experiments in this repository

  • E120 — Pretentious distance atlas for core multiplicative functions (μ, λ, φ/n, …).