KEVOS
ArticlesServicesCase studiesAboutContact
ArticlesServicesCase studiesAboutContact
← ArticlesSmoothness and Sub-exponential ComplexityEngineering · Engineering MathematicsLesson 867/884← PrevNext →
ArticlePublished 7 Aug 20262 min readBy Kevin Joginsmoothnesssub-exponentialDickman functionL-notation
On this page

Ask about this page

KEVOS AISmoothness and Sub-exponential Complexity

KEVOS knowledge first · trusted web sources when needed

Modern Factoring Methods

Smoothness and Sub-exponential Complexity

Smooth numbers, the Dickman function, and how balancing smoothness probability against factor base size produces sub-exponential running times.

Engineering / MathematicsModern Factoring Methods2 min readKV-MATH-0665

Almost every modern factoring and class group algorithm rests on the same calculation: how likely a random number is to factor entirely over a fixed set of small primes, and how to choose that set optimally.

Smooth numbers

B-smooth
An integer all of whose prime factors are at most B.
Semi-smooth
Smooth except for one prime slightly above the bound. Exploited by the large prime variation.
Smoothness probability
The proportion of integers up to a size that are smooth to a given bound.

The Dickman estimate

The proportion of integers up to a bound that are smooth to a considerably smaller bound is estimated by the Dickman function, evaluated at the ratio of the logarithms.

Proportion of x-sized integers that are y-smooth ~ u^(-u), u = log x / log yThe Dickman estimate; accurate for the ranges of interest.

Key point

The estimate says that smoothness probability decays like u to the minus u. This decay is fast but not catastrophic, and it is gentle enough that a balance point exists — which is exactly why sub-exponential algorithms are possible.

The optimisation

A larger factor base makes each candidate more likely to be smooth but requires more relations and a larger matrix. Balancing the two costs gives the optimal bound.

Optimising the smoothness bound

  1. Write down the two costsSearch cost is the reciprocal of the smoothness probability, times the number of relations needed.
  2. DifferentiateWith respect to the smoothness bound.
  3. SolveThe balance point gives the optimal bound and the resulting complexity.

L-notation

L_n(a, c) = exp( c (log n)^a (log log n)^(1-a) )Interpolates between polynomial at a = 0 and exponential at a = 1.
Sub-exponential complexities in L-notation
AlgorithmComplexity
CFRACL(1/2)
Quadratic sieveL(1/2)
ECML(1/2) in the size of the factor, not the number
Number field sieveL(1/3)
Class group methodsL(1/2) in the discriminant

Key point

ECM is the odd one out and the difference matters: its cost depends on the size of the factor found, not the size of the number. That is why it is the right tool for finding medium factors of very large numbers, where the sieves are hopeless.

Why the number field sieve is better

The sieves examine values of a polynomial. The quadratic sieve's values are around the square root of the number; the number field sieve's are far smaller, so they are much more likely to be smooth. That improvement is what moves the exponent from one half to one third.

Note

The heuristic content of these analyses is the assumption that the values examined behave like random integers of their size for smoothness purposes. This is unproven and universally believed, and matches observation closely.

Source. Henri Cohen, A Course in Computational Algebraic Number Theory, Springer GTM 138 — 10.1. Structural reference unverified: the source file was not available during authoring; chapter and section numbers are taken from the published edition and have not been checked against a physical copy.

Related pages

  • Factor Base Selection and Smoothness
  • Quadratic Sieve Factor Base Selection
  • The Continued Fraction Factorisation Method

Continue learning

Primality Certificates and Independent VerificationArticle · Engineering MathematicsNEXT LESSON →The Continued Fraction Factorisation MethodArticle · Engineering MathematicsAtkin-Morain Elliptic Curve Primality ProvingArticle · Engineering MathematicsThe Schnorr-Lenstra Class Group Factoring MethodArticle · Engineering Mathematics
KEVOS · Engineering, manufacturing and project improvement
ArticlesServicesCase studiesAboutContact
© 2026 KEVOS®