Power Sieve over Number Fields

• Power sieve over number fields is a family of methods that blend arithmetic reciprocity, local congruence conditions, and advanced analytic tools like Poisson summation and geometry of numbers.
• These techniques yield explicit quantitative results such as quadratic large sieve bounds, power-moduli count estimates, and power-saving error terms in orbit counting problems.
• The approach systematically converts local algebraic constraints into global asymptotic statements, enabling precise control over error margins and existence results.

Power sieve over number fields denotes a family of sieve-theoretic methods in which arithmetic over a number field is combined with reciprocity, local congruence conditions, Poisson summation, geometry of numbers, or smoothness estimates to detect power structure and extract global counting information. In the literature, the phrase does not refer to a single fixed formalism. It includes Heath-Brown-style large-sieve arguments for ideal-indexed character families, Weyl-differencing methods for $k$-th power moduli, Selberg $\Lambda^2$-sieves inserted into orbit-counting problems, and beta-sieve arguments for norm-form equations. In a broader algorithmic sense, related “power-sieving” language also appears in number-field-sieve subroutines and local-global root extraction over number fields (Goldmakher et al., 2011, Gao et al., 2021, Shankar et al., 2013, Shute, 2022, Bernard et al., 2023).

1. Conceptual structure

Across its variants, the subject is organized by a recurrent local-to-global pattern. Arithmetic objects are encoded either by ideals, by values of binary or univariate forms, or by integral orbits of algebraic groups. “Bad” or “forbidden” behavior is then expressed through local residue conditions, local splitting constraints, or divisibility by prime ideals. The sieve converts these local conditions into upper or lower bounds for global counting functions. This suggests a common template: reciprocity or congruence data supply a local model, Poisson summation or geometry of numbers provides uniform counting in families, and the sieve mechanism turns that input into quantitative asymptotics or nonvanishing statements (Goldmakher et al., 2011, Gao et al., 2021, Shankar et al., 2013, Shute, 2022).

Setting	Main mechanism	Representative outcome
Ideal-indexed character families	Reciprocity + Poisson summation + large sieve	Quadratic large sieve bound of scale $(MN)^\varepsilon(M+N)$
$k$-th power moduli in imaginary quadratic fields	Weyl differencing + Poisson summation on $\mathcal O_K$	Large sieve with terms $Q^{k+1}$, $NQ^{1-1/\kappa}$, $N^{1-1/\kappa}Q^{1+k/\kappa}$
Bhargava-style orbit counting	Selberg $\Lambda^2$-sieve on local congruence classes	$S_5$-quintic field count with error $\Lambda^2$0
Norm-form equations	Rosser–Iwaniec beta sieve	Hasse principle under explicit thresholds $\Lambda^2$1 and $\Lambda^2$2

A central terminological point is that “power sieve” is sometimes used narrowly for Heath-Brown-style detection of power structure, and sometimes more broadly for any sieve in which powers, powerful values, or power moduli govern the arithmetic constraints. The broader usage appears explicitly in discussions of norm forms, exponent semigroups, and root extraction, but some papers state that the term is only heuristic in those contexts.

2. Hecke families and the quadratic large sieve over number fields

A foundational analytic form of the subject is the quadratic large sieve over number fields. In this framework, the basic obstacle is that the classical Jacobi symbol does not have a direct ideal-indexed analogue over a general number field. The solution in "A quadratic large sieve inequality over number fields" is the notion of an $\Lambda^2$3-th order Hecke family: a collection $\Lambda^2$4 of primitive Hecke characters attached to squarefree ideals, together with a reciprocity law

$\Lambda^2$5

a bounded-order condition, and a compatibility condition ensuring primitivity of products when ideal classes match appropriately (Goldmakher et al., 2011).

The central estimate is the quadratic large sieve inequality

$\Lambda^2$6

This is presented as the exact analogue of Heath-Brown’s classical quadratic large sieve over $\Lambda^2$7, and it also generalizes Onodera’s result for $\Lambda^2$8. The proof develops Poisson summation for Hecke characters over number fields, introduces smoothed quantities $\Lambda^2$9 and $(MN)^\varepsilon(M+N)$0, and uses a self-improvement mechanism: from a bound of the form $(MN)^\varepsilon(M+N)$1 with $(MN)^\varepsilon(M+N)$2, one deduces $(MN)^\varepsilon(M+N)$3, and iteration drives the exponent to $(MN)^\varepsilon(M+N)$4.

Two features are structurally decisive. First, reciprocity allows one to exchange the roles of the sieving variable and the dual variable, which is the hallmark of power-sieve arguments. Second, the squarefree restriction is essential. The paper notes that without it the inequality fails badly; a sequence concentrated on square ideals can make the left-hand side of size $(MN)^\varepsilon(M+N)$5. In that sense, the theorem is not merely a large-sieve estimate over number fields, but an explicitly power-sieve-compatible analytic tool.

3. Power moduli in imaginary quadratic fields

A second major branch concerns large sieve inequalities for power moduli in imaginary quadratic fields. Here the ambient field is

$(MN)^\varepsilon(M+N)$6

with additive character naturally attached to the trace,

$(MN)^\varepsilon(M+N)$7

The main theorem of "The Large Sieve with Power Moduli in Imaginary Quadratic Number Fields" states that for $(MN)^\varepsilon(M+N)$8, $(MN)^\varepsilon(M+N)$9, and any complex sequence $k$0,

$k$1

(Gao et al., 2021).

The proof is explicitly described as a power sieve argument over $k$2. After rewriting the exponential sum in the $k$3-basis $k$4, the problem becomes a two-dimensional large sieve problem in $k$5. A Poisson summation formula on $k$6, together with a Dirichlet approximation theorem in $k$7 by fractions from $k$8, reduces the analysis to counting congruence classes. The power-sieve step enters through repeated Weyl differencing: after specializing the moduli to $k$9, the phase is differenced $\mathcal O_K$0 times until it becomes linear. The exponent $\mathcal O_K$1 is the characteristic output of this repeated Cauchy–Schwarz procedure.

The same paper also establishes a sparse-moduli theorem in terms of a local density parameter $\mathcal O_K$2, together with specialized results for square moduli and prime moduli in class-number-one fields. At the same time, it records a boundary of the method: real quadratic fields are said to be more complicated because of their infinite unit groups, so the argument does not transfer directly.

4. Selberg $\mathcal O_K$3-sieve in orbit counting

A distinct, but closely related, direction uses the sieve not to detect power moduli directly, but to upgrade orbit-counting asymptotics over number fields to power-saving asymptotics. In "Counting $\mathcal O_K$4-fields with a power saving error term", the arithmetic objects are parametrized by Bhargava’s representation

$\mathcal O_K$5

for

$\mathcal O_K$6

whose integral orbits correspond to isomorphism classes of pairs $\mathcal O_K$7, with $\mathcal O_K$8 a quintic ring and $\mathcal O_K$9 a sextic resolvent. The final counting theorem is

$Q^{k+1}$0

with

$Q^{k+1}$1

for $Q^{k+1}$2-quintic fields of discriminant bounded by $Q^{k+1}$3 and fixed signature (Shankar et al., 2013).

The key innovation is to insert a Selberg $Q^{k+1}$4-sieve precisely where degenerate or nonmaximal objects are removed by local congruence conditions. For each prime $Q^{k+1}$5, Bhargava’s construction supplies local subsets $Q^{k+1}$6 and $Q^{k+1}$7, with complements $Q^{k+1}$8 and $Q^{k+1}$9. Degenerate objects are characterized by forbidden simultaneous local conditions modulo distinct primes. A naive inclusion–exclusion argument yields only an $NQ^{1-1/\kappa}$0-type remainder, whereas the Selberg sieve combines the local densities multiplicatively and produces

$NQ^{1-1/\kappa}$1

The sieve rests on a uniform geometry-of-numbers input for congruence classes: $NQ^{1-1/\kappa}$2 for a translate $NQ^{1-1/\kappa}$3 of $NQ^{1-1/\kappa}$4. It also uses local density limits

$NQ^{1-1/\kappa}$5

and a nonmaximality estimate

$NQ^{1-1/\kappa}$6

for squarefree $NQ^{1-1/\kappa}$7. The resulting exponent $NQ^{1-1/\kappa}$8 is the precise power saving in the error term. This paper is explicit that the method is not special to quintic fields: it applies whenever arithmetic objects are parametrized as integral orbits, the counting is done in a fundamental domain, bad objects are cut out by finitely many local congruence conditions, and uniform asymptotics are available in congruence classes.

5. Beta sieve, norm forms, and powerful values

The Rosser–Iwaniec beta sieve yields another major family of number-field power-sieve arguments. In "Polynomials represented by norm forms via the beta sieve", the object of study is the affine equation

$NQ^{1-1/\kappa}$9

After homogenizing $N^{1-1/\kappa}Q^{1+k/\kappa}$0 to a binary form and counting lattice points $N^{1-1/\kappa}Q^{1+k/\kappa}$1 in regions $N^{1-1/\kappa}Q^{1+k/\kappa}$2, the paper defines a sifting function

$N^{1-1/\kappa}Q^{1+k/\kappa}$3

The analytic input is a level-of-distribution estimate for binary forms and a Buchstab decomposition, after which the beta sieve supplies lower and upper bound constants $N^{1-1/\kappa}Q^{1+k/\kappa}$4 and $N^{1-1/\kappa}Q^{1+k/\kappa}$5. The main Hasse-principle result in the quadratic-factor range assumes that $N^{1-1/\kappa}Q^{1+k/\kappa}$6 satisfies the Hasse norm principle and that all irreducible factors of $N^{1-1/\kappa}Q^{1+k/\kappa}$7 have degree at most $N^{1-1/\kappa}Q^{1+k/\kappa}$8; if

$N^{1-1/\kappa}Q^{1+k/\kappa}$9

then the Hasse principle holds for $\Lambda^2$0. For cubic factors, a second theorem requires a structured congruence condition on the bad primes and

$\Lambda^2$1

(Shute, 2022).

The same sieve input is applied there to new cases of the Harpaz–Wittenberg conjecture and to rational points on fibrations. A notable limitation is equally explicit: degree $\Lambda^2$2 factors are out of reach because the sieve decomposition forces incompatible inequalities in the control of the $\Lambda^2$3 terms.

A Diophantine variant appears in "On the powerful values of polynomials over number fields". For a number field $\Lambda^2$4, an element of $\Lambda^2$5 is $\Lambda^2$6-powerful if every prime ideal dividing its principal ideal occurs with exponent at least $\Lambda^2$7, and a polynomial is $\Lambda^2$8-powerful if every irreducible factor occurs with multiplicity at least $\Lambda^2$9. The paper studies monic degree-$S_5$0 polynomials whose irreducible factors all have multiplicity strictly less than $S_5$1, but whose values at prescribed points $S_5$2 are nonzero $S_5$3-powerful elements. Assuming a quantitative bounded-degree form of Vojta’s conjecture, it proves finiteness of the exceptional set $S_5$4 for

$S_5$5

together with two-sided bounds

$S_5$6

Under periodicity of one of the associated sequences $S_5$7, it further proves that $S_5$8 for all sufficiently large $S_5$9 (Salami, 2017). This is not a classical large-sieve theorem, but it belongs to the same power-sieve philosophy: repeated prime factors in many values are forced to arise from algebraic structure unless one is in a small exceptional set.

6. Scope, adjacent usages, and limitations

The subject has a broad perimeter, and several nearby literatures use “power” and “sieving” language in ways that are related but not identical. One recurrent misconception is to identify the power sieve over number fields with the Number Field Sieve. The latter is a factorization or discrete-logarithm algorithm whose polynomial selection, lattice sieving, and square-root or root-extraction stages take place over number fields, but its main purpose is computational rather than the derivation of analytic counting theorems. "Montgomery's method of polynomial selection for the number field sieve" analyzes the polynomial-selection stage through modular geometric progressions; "An Implementation of the Extended Tower Number Field Sieve using 4d Sieving in a Box and a Record Computation in $\Lambda^2$00" describes ExTNFS as a tower-based NFS variant using lattice sieving over number fields, with empirical evidence that 4-dimensional box sieving was faster in practice than 4-dimensional hypersphere sieving in the reported $\Lambda^2$01 record computation (Coxon, 2014, Robinson, 2022).

A second boundary issue concerns broader “power-sieving” metaphors. "Computing $\Lambda^2$02-th roots in number fields" describes local-global root extraction algorithms as fundamentally “power-sieving” tools because they reconstruct exact $\Lambda^2$03-th roots from factored input by combining local root extraction, Chinese remaindering over prime ideals, and relative norm compatibility. "Power-integral matrices over number fields" states explicitly that it does not use the term “power sieve” as a formal method, but has a clear sieving flavor in its study of the exponent semigroup

$\Lambda^2$04

where the Drazin decomposition and pseudo-determinant act as a structural filter on allowable exponents (Bernard et al., 2023, Chinn et al., 29 Jun 2026).

The principal technical limitations are method-specific rather than terminological. In the quadratic large sieve, squarefreeness is indispensable. In the power-moduli large sieve, the imaginary quadratic setting is essential to the current argument. In the beta-sieve norm-form problem, quartic factors remain beyond the method. In the powerful-values problem, the main finiteness theorem is conditional on a quantitative version of Vojta’s conjecture. These limitations indicate that “power sieve over number fields” is best understood as a family resemblance among techniques rather than a single theorem schema.

The unifying significance of the subject is that it supplies a systematic way to convert local algebraic constraints in number fields into global statements with explicit savings, thresholds, or existence results. Whether the ambient objects are Hecke characters, power moduli, arithmetic orbits, norm-form values, or root-extraction data, the method succeeds when one has enough uniformity in congruence classes and enough reciprocity or distribution theory to make the local information multiplicative.