An analytic device that estimates bilinear or quadratic sums by measuring variance (dispersion) of partial sums of arithmetic sequences; it converts control of a global mean and second moment into bounds on correlations or deviations from expected distribution.

Principle

Formulate the quantity of interest as a bilinear form and bound its variance by expanding a second moment, using orthogonality, completion, and averaging to force cancellation except for structured diagonal contributions.

Demonstration

Use a bilinear decomposition of a multiplicative sequence a_n and a test sequence b_m to bound sums over moduli q of |Sum_{n} a_n e(n/q)|^2; Linnik-style calculations convert such second moments into manageable sums that detect irregular distribution of primes or character sums in progressions.

Misapplication

Applying the method to sequences without sufficient averaging or to single, very sparse sums; treating a small sample of data as representative so that the dispersion estimate is dominated by few terms and yields misleadingly weak or spurious bounds.

Consequence

When applicable, produces quantitative bounds on correlations and deviations (for example, limiting exceptional moduli or bounding mean-square errors), and isolates the structured (diagonal) part responsible for any persistent bias.

Reversal

Instead of studying dispersion (variance) one could study linear correlations or additive structure directly; a small dispersion indicates pseudorandomness while a large dispersion points to detectable structure or bias.

Boundary

Requires bilinear decomposition and enough averaging (ranges in variables or moduli) to exploit cancellation; it does not replace pointwise estimates, and it is ineffective for single-term control or for sequences lacking orthogonality or multiplicative-type structure.

Semantic Tension

Competes with the circle method and large-sieve techniques: all convert global averages into local information but differ in transform tools and the way off-diagonal contributions are treated; choosing among them depends on available structure and ranges.

Synthesis

The dispersion method measures variance of bilinear forms to convert averaged second-moment information into bounds on distributional irregularities, separating diagonal structure from off-diagonal noise whenever sufficient averaging and orthogonality are available.