A function g(n) is an average order of f(n) if the partial sums satisfy Σ_{k≤x} f(k) ∼ Σ_{k≤x} g(k) as x→∞; g captures the cumulative typical size of f when summed up to x.

Principle

Replace pointwise comparison by asymptotic equality of cumulative sums so that global summatory behavior is modeled; average order is a device for predicting large-scale sum behavior rather than individual values.

Demonstration

The divisor function d(n) (number of positive divisors) has average order log n because Σ_{k≤x} d(k) = x log x + (2γ −1) x + o(x), so g(n)=log n models the summatory growth.

Misapplication

Interpreting average order as a claim about f(n) for specific n (i.e., expecting f(n) ∼ g(n) pointwise) or using an average order without verifying the asymptotic equivalence of sums.

Consequence

When a simple g is known, it yields asymptotic estimates for large sums and guides probabilistic and analytic arguments about the global distribution of f.

Reversal

Normal order is a stricter, almost-everywhere pointwise statement (f(n)/g(n)→1 for almost all n) and need not follow from knowledge of the average order; averages can be dominated by exceptional values.

Boundary

Pertains to summatory asymptotics over initial intervals [1,x]; it does not control local fluctuations, rare large values, or convergence of f(n) itself, and it requires sufficiently precise asymptotic analysis of Σ_{k≤x} f(k).

Semantic Tension

Tension with normal order and pointwise asymptotics: average order is a global summatory notion, while normal order concerns typical individual behaviour; they agree for some functions but diverge for others.

Synthesis

Average order summarizes the aggregate growth of an arithmetic function via an asymptotic comparison of cumulative sums, providing a tractable model for global summatory behavior even when pointwise control is unavailable.