Division and Slope Factorization of p-Adic Polynomials · Division and Slope Factorization of...

Preview:

Citation preview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic Polynomials

Division and Slope Factorization of p-AdicPolynomials

Xavier Caruso, David Roe Tristan Vaccon

Univ.Rennes 1, Univ. Pittsburgh, 立教大学

July 22nd, 2016

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;

e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation, e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation, e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation,

e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation, e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;

Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation, e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Why should one work with p-adic numbers ?

p-adic methodsWorking in Qp instead of Q, one can handle more efficiently thecoefficients growth ;e.g. Dixon’s method (used in F4), Polynomial factorization viaHensel’s lemma.

p-adic algorithmsGoing from Z/pZ to Zp and then back to Z/pZ enables morecomputation, e.g. the algorithms of Bostan et al. and Lercier et al.using p-adic differential equations ;Kedlaya’s and Lauder’s counting-point algorithms via p-adiccohomology ;

My personal (long-term) motivationComputing (some) moduli spaces of p-adic Galois representations.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra.

This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra. This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra. This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.

Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra. This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra. This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.

Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Studying polynomial computations over p-adics

A building blockAt ISSAC 2015, we have studied the p-adic stability of somecomputations in linear algebra. This year, we study basic operationsrelated to polynomial computations.

More motivationsUnderstanding basic operations related to field extensions, inparticular division and quotients.Understanding the behaviour of precision during factorisation: overQp or kJT K, or as an intermediate to factorisation over Q.

Today’s highlightsOptimal tracking of precision for modular multiplication anddiffused digits.Precision for slope factorization algorithms.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

What are p-adic numbers?

p refers to a prime number

p-adic numbers are numbers written in p-basis of the shape:a = . . . ai . . . a2 a1 a0 , a−1 a−2 . . . a−n

with 0 ≤ ai < p for all i .

Addition and multiplication on these numbers are defined by applyingSchoolBook algorithms.

The valuation vp(a) of a is the smallest v such that av ̸= 0.

The p-adic numbers form the field Qp.

A p-adic number with no digit after the comma is ap-adic integer.

The p-adic integers form a subring Zp of Qp.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Summary on p-adics

Proposition

Zp/pZp = Z/pZ.

∀k ∈ N,Zp/pkZp = Z/pkZ.

A first ideaQp is an extension of Q where one can perform calculus, as simplyas over R.We are closer to arithmetic : we can reduce modulo p.

Remark

Qp R Zp Zp

Q Z/pZ Z/pZ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Summary on p-adics

Proposition

Zp/pZp = Z/pZ.

∀k ∈ N,Zp/pkZp = Z/pkZ.

A first ideaQp is an extension of Q where one can perform calculus, as simplyas over R.We are closer to arithmetic : we can reduce modulo p.

Remark

Qp R Zp Zp

Q Z/pZ Z/pZ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Summary on p-adics

Proposition

Zp/pZp = Z/pZ.

∀k ∈ N,Zp/pkZp = Z/pkZ.

A first ideaQp is an extension of Q where one can perform calculus, as simplyas over R.

We are closer to arithmetic : we can reduce modulo p.

Remark

Qp R Zp Zp

Q Z/pZ Z/pZ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Summary on p-adics

Proposition

Zp/pZp = Z/pZ.

∀k ∈ N,Zp/pkZp = Z/pkZ.

A first ideaQp is an extension of Q where one can perform calculus, as simplyas over R.We are closer to arithmetic : we can reduce modulo p.

Remark

Qp R Zp Zp

Q Z/pZ Z/pZ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Summary on p-adics

Proposition

Zp/pZp = Z/pZ.

∀k ∈ N,Zp/pkZp = Z/pkZ.

A first ideaQp is an extension of Q where one can perform calculus, as simplyas over R.We are closer to arithmetic : we can reduce modulo p.

Remark

Qp R Zp Zp

Q Z/pZ Z/pZ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Definition of the precision

Finite-precision p-adics

Elements of Qp can be written∑+∞

i=k aipi , with ai ∈ J0, p − 1K, k ∈ Zand p a prime number.While working with a computer, we usually only can consider thebeginning of this power serie expansion: we only consider elements of thefollowing form

∑d−1i=l aipi + O(pd) , with l ∈ Z.

Definition

The order, or the absolute precision of∑d−1

i=k aipi + O(pd) is d .

ExampleThe order of 3 ∗ 7−1 + 4 ∗ 70 + 5 ∗ 71 + 6 ∗ 72 + O(73) is 3.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Definition of the precision

Finite-precision p-adics

Elements of Qp can be written∑+∞

i=k aipi , with ai ∈ J0, p − 1K, k ∈ Zand p a prime number.While working with a computer, we usually only can consider thebeginning of this power serie expansion: we only consider elements of thefollowing form

∑d−1i=l aipi + O(pd) , with l ∈ Z.

Definition

The order, or the absolute precision of∑d−1

i=k aipi + O(pd) is d .

ExampleThe order of 3 ∗ 7−1 + 4 ∗ 70 + 5 ∗ 71 + 6 ∗ 72 + O(73) is 3.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Definition of the precision

Finite-precision p-adics

Elements of Qp can be written∑+∞

i=k aipi , with ai ∈ J0, p − 1K, k ∈ Zand p a prime number.While working with a computer, we usually only can consider thebeginning of this power serie expansion: we only consider elements of thefollowing form

∑d−1i=l aipi + O(pd) , with l ∈ Z.

Definition

The order, or the absolute precision of∑d−1

i=k aipi + O(pd) is d .

ExampleThe order of 3 ∗ 7−1 + 4 ∗ 70 + 5 ∗ 71 + 6 ∗ 72 + O(73) is 3.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Definition of the precision

Finite-precision p-adics

Elements of Qp can be written∑+∞

i=k aipi , with ai ∈ J0, p − 1K, k ∈ Zand p a prime number.While working with a computer, we usually only can consider thebeginning of this power serie expansion: we only consider elements of thefollowing form

∑d−1i=l aipi + O(pd) , with l ∈ Z.

Definition

The order, or the absolute precision of∑d−1

i=k aipi + O(pd) is d .

ExampleThe order of 3 ∗ 7−1 + 4 ∗ 70 + 5 ∗ 71 + 6 ∗ 72 + O(73) is 3.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

p-adic precion vs real precisionThe quintessential idea of the step-by-step analysis is the following :

Proposition (p-adic errors don’t add)Indeed,

(a + O(p k )) + (b + O(p k )) = a + b + O(p k ).

That is to say, if a and b are known up to precision O(pk), then so isa + b.

RemarkIt is quite the opposite to when dealing with real numbers, because ofRound-off error :

(1 + 5 ∗ 10−2) + (2 + 6 ∗ 10−2) = 3 + 1 ∗ 10−1 + 1 ∗ 10−2.

That is to say, if a and b are known up to precision 10−n, then a + b is

known up to 10(−n +1 ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

p-adic precion vs real precisionThe quintessential idea of the step-by-step analysis is the following :

Proposition (p-adic errors don’t add)Indeed,

(a + O(p k )) + (b + O(p k )) = a + b + O(p k ).

That is to say, if a and b are known up to precision O(pk), then so isa + b.

RemarkIt is quite the opposite to when dealing with real numbers, because ofRound-off error :

(1 + 5 ∗ 10−2) + (2 + 6 ∗ 10−2) = 3 + 1 ∗ 10−1 + 1 ∗ 10−2.

That is to say, if a and b are known up to precision 10−n, then a + b is

known up to 10(−n +1 ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

p-adic precion vs real precisionThe quintessential idea of the step-by-step analysis is the following :

Proposition (p-adic errors don’t add)Indeed,

(a + O(p k )) + (b + O(p k )) = a + b + O(p k ).

That is to say, if a and b are known up to precision O(pk), then so isa + b.

RemarkIt is quite the opposite to when dealing with real numbers, because ofRound-off error :

(1 + 5 ∗ 10−2) + (2 + 6 ∗ 10−2) = 3 + 1 ∗ 10−1 + 1 ∗ 10−2.

That is to say, if a and b are known up to precision 10−n, then a + b is

known up to 10(−n +1 ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

p-adic precion vs real precisionThe quintessential idea of the step-by-step analysis is the following :

Proposition (p-adic errors don’t add)Indeed,

(a + O(p k )) + (b + O(p k )) = a + b + O(p k ).

That is to say, if a and b are known up to precision O(pk), then so isa + b.

RemarkIt is quite the opposite to when dealing with real numbers, because ofRound-off error :

(1 + 5 ∗ 10−2) + (2 + 6 ∗ 10−2) = 3 + 1 ∗ 10−1 + 1 ∗ 10−2.

That is to say, if a and b are known up to precision 10−n, then a + b is

known up to 10(−n +1 ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Precision formulae

Proposition (addition)

(x0 + O(pk0)) + (x1 + O(pk1)) = x0 + x1 + O(pmin(k0,k1))

Proposition (multiplication)

(x0 + O(pk0)) ∗ (x1 + O(pk1)) = x0 ∗ x1 + O(pmin(k0+vp(x1),k1+vp(x0)))

Proposition (division)

xpa + O(pb)ypc + O(pd) = x ∗ y−1pa−c + O(pmin(d+a−2c,b−c))

In particular, 1pcy + O(pd) = y−1p−c + O(pd−2c)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Precision formulae

Proposition (addition)

(x0 + O(pk0)) + (x1 + O(pk1)) = x0 + x1 + O(pmin(k0,k1))

Proposition (multiplication)

(x0 + O(pk0)) ∗ (x1 + O(pk1)) = x0 ∗ x1 + O(pmin(k0+vp(x1),k1+vp(x0)))

Proposition (division)

xpa + O(pb)ypc + O(pd) = x ∗ y−1pa−c + O(pmin(d+a−2c,b−c))

In particular, 1pcy + O(pd) = y−1p−c + O(pd−2c)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Precision formulae

Proposition (addition)

(x0 + O(pk0)) + (x1 + O(pk1)) = x0 + x1 + O(pmin(k0,k1))

Proposition (multiplication)

(x0 + O(pk0)) ∗ (x1 + O(pk1)) = x0 ∗ x1 + O(pmin(k0+vp(x1),k1+vp(x0)))

Proposition (division)

xpa + O(pb)ypc + O(pd) = x ∗ y−1pa−c + O(pmin(d+a−2c,b−c))

In particular, 1pcy + O(pd) = y−1p−c + O(pd−2c)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsIntroduction

Interesting behaviour

Large modular multiplicationSetting: 10 significant digits in Q2.

Modulus P = X 2 + 2.

A = 5 − 2X + 4X 2.

Compute iteratively (naively) An = An mod P.

Behaviour of significant digits

n 100 101 102 103 104 105 106

relative precision of lc(An) 10 9 8 7 6 5 4

Question:How can we obtain satisfying / optimal behaviour regarding toprecision?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

The Main lemma of p-adic differential precision

Lemma (CRV14)Let f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + B) = f (x) + f ′(x) · B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

The Main lemma of p-adic differential precision

Lemma (CRV14)Let f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + B) = f (x) + f ′(x) · B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

The Main lemma of p-adic differential precision

Lemma (CRV14)Let f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + B) = f (x) + f ′(x) · B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

The Main lemma of p-adic differential precision

Lemma (CRV14)Let f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + B) = f (x) + f ′(x) · B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

x

B

f (x)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

x

B

f (x)

f ′(x)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

x

B

f (x)

f ′(x)

f ′(x) · B

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

xx + B

B

f (x)

f ′(x)

f ′(x) · B

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

xx + B

f

B

f (x)

f ′(x)

f ′(x) · B

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Geometrical meaning

Interpretation

xx + B

f

B

f (x)

f ′(x)

f (x) + f ′(x) · B

f ′(x) · B

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Lattices

LemmaLet f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + ) = f (x) + f ′(x) · .

RemarkThis allows more models of precision, like

(x , y) = (1 + O(p10), 1 + O(p)).

RemarkOur framework can be extended to (complete) ultrametric K -vectorspaces (e.g. being Fp((X ))n, Q((X ))m, R((ε))s).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Lattices

LemmaLet f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough,

f (x + B) = f (x) + f ′(x) · B.

RemarkThis allows more models of precision, like

(x , y) = (1 + O(p10), 1 + O(p)).

RemarkOur framework can be extended to (complete) ultrametric K -vectorspaces (e.g. being Fp((X ))n, Q((X ))m, R((ε))s).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Lattices

LemmaLet f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough, for any open Zp-latticeH ⊂ B

f (x + H) = f (x) + f ′(x) · H.

RemarkThis allows more models of precision, like

(x , y) = (1 + O(p10), 1 + O(p)).

RemarkOur framework can be extended to (complete) ultrametric K -vectorspaces (e.g. being Fp((X ))n, Q((X ))m, R((ε))s).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Lattices

LemmaLet f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough, for any open Zp-latticeH ⊂ B

f (x + H) = f (x) + f ′(x) · H.

RemarkThis allows more models of precision, like

(x , y) = (1 + O(p10), 1 + O(p)).

RemarkOur framework can be extended to (complete) ultrametric K -vectorspaces (e.g. being Fp((X ))n, Q((X ))m, R((ε))s).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Lattices

LemmaLet f : Qn

p → Qmp be a (strictly) differentiable mapping.

Let x ∈ Qnp. We assume that f ′(x) is surjective.

Then for any ball B = B(0, r) small enough, for any open Zp-latticeH ⊂ B

f (x + H) = f (x) + f ′(x) · H.

RemarkThis allows more models of precision, like

(x , y) = (1 + O(p10), 1 + O(p)).

RemarkOur framework can be extended to (complete) ultrametric K -vectorspaces (e.g. being Fp((X ))n, Q((X ))m, R((ε))s).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Higher differentials

What is small enough ?How can we determine when the lemma applies ?When f is locally analytic, it essentially corresponds to

+∞∑k=2

1k! f (k)(x) · Hk ⊂ f ′(x) · H.

This can be determined with Newton-polygon techniques.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Higher differentials

What is small enough ?How can we determine when the lemma applies ?

When f is locally analytic, it essentially corresponds to

+∞∑k=2

1k! f (k)(x) · Hk ⊂ f ′(x) · H.

This can be determined with Newton-polygon techniques.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Higher differentials

What is small enough ?How can we determine when the lemma applies ?When f is locally analytic, it essentially corresponds to

+∞∑k=2

1k! f (k)(x) · Hk ⊂ f ′(x) · H.

This can be determined with Newton-polygon techniques.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

p-Adic Precision

Higher differentials

What is small enough ?How can we determine when the lemma applies ?When f is locally analytic, it essentially corresponds to

+∞∑k=2

1k! f (k)(x) · Hk ⊂ f ′(x) · H.

This can be determined with Newton-polygon techniques.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Handling the precision

Lattice precisionTo each polynomial, we attach a Zp-lattice (given by a basis of thislattice).

Lattice and divisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Proof.

A = BQ + R,

A + δA = (B + δB)(Q + δQ) + (R + δR),δA = QδB + BδQ + δR (at first order),

δA − QδB = BδQ + δR.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Handling the precision

Lattice precisionTo each polynomial, we attach a Zp-lattice (given by a basis of thislattice).

Lattice and divisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Proof.

A = BQ + R,

A + δA = (B + δB)(Q + δQ) + (R + δR),δA = QδB + BδQ + δR (at first order),

δA − QδB = BδQ + δR.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Handling the precision

Lattice precisionTo each polynomial, we attach a Zp-lattice (given by a basis of thislattice).

Lattice and divisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Proof.

A = BQ + R,

A + δA = (B + δB)(Q + δQ) + (R + δR),

δA = QδB + BδQ + δR (at first order),δA − QδB = BδQ + δR.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Handling the precision

Lattice precisionTo each polynomial, we attach a Zp-lattice (given by a basis of thislattice).

Lattice and divisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Proof.

A = BQ + R,

A + δA = (B + δB)(Q + δQ) + (R + δR),δA = QδB + BδQ + δR (at first order),

δA − QδB = BδQ + δR.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Study of the division

Handling the precision

Lattice precisionTo each polynomial, we attach a Zp-lattice (given by a basis of thislattice).

Lattice and divisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Proof.

A = BQ + R,

A + δA = (B + δB)(Q + δQ) + (R + δR),δA = QδB + BδQ + δR (at first order),

δA − QδB = BδQ + δR.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Lattice precision and modular multiplication

Composed derivativesIt is easy to handle in an optimal way the modular multiplication byapplying:

For A × B,δ(A × B) = AδB + BδA.

For remainder R in A = BQ + R,

δ(R) = δA − QδB mod B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Lattice precision and modular multiplication

Composed derivativesIt is easy to handle in an optimal way the modular multiplication byapplying:

For A × B,

δ(A × B) = AδB + BδA.

For remainder R in A = BQ + R,

δ(R) = δA − QδB mod B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Lattice precision and modular multiplication

Composed derivativesIt is easy to handle in an optimal way the modular multiplication byapplying:

For A × B,δ(A × B) = AδB + BδA.

For remainder R in A = BQ + R,

δ(R) = δA − QδB mod B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Lattice precision and modular multiplication

Composed derivativesIt is easy to handle in an optimal way the modular multiplication byapplying:

For A × B,δ(A × B) = AδB + BδA.

For remainder R in A = BQ + R,

δ(R) = δA − QδB mod B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Lattice precision and modular multiplication

Composed derivativesIt is easy to handle in an optimal way the modular multiplication byapplying:

For A × B,δ(A × B) = AδB + BδA.

For remainder R in A = BQ + R,

δ(R) = δA − QδB mod B.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Toward numerical understanding

Displaying precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Toward numerical understanding

Displaying precisionLet H ⊂ Qn

p be a lattice.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Toward numerical understanding

Displaying precisionLet H ⊂ Qn

p be a lattice.

H0 = π1(H) ⊕ · · · ⊕ πn(H)

provides the best precision oncoordinates.It is the smallest diagonal latticecontaining H.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Toward numerical understanding

Displaying precisionLet H ⊂ Qn

p be a lattice.

H0 = π1(H) ⊕ · · · ⊕ πn(H)

provides the best precision oncoordinates.It is the smallest diagonal latticecontaining H.

Diffused digitsThe number of diffused digits of precisionof H is the length of H0/H.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

More on diffused digits

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

More on diffused digits

Diffused digitsThe number of diffused digits of precisionof H is the length of H0/H.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

More on diffused digits

Diffused digitsThe number of diffused digits of precisionof H is the length of H0/H.

Another definitionHere, the number of diffused digits is:

− logp (|H0/H|) .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Diffused digits: example

Diffused digits: the latticeLet

H =⟨ 1

pp2

,

pp2

2p4

,

2pp2

2p3

⟩.

Then

H0 =

1p

p2

.

Number of diffused digits

The SNF of H0/H is:

1p

p

. Hence 2 diffused digits.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Diffused digits: example

Diffused digits: the latticeLet

H =⟨ 1

pp2

,

pp2

2p4

,

2pp2

2p3

⟩.

Then

H0 =

1p

p2

.

Number of diffused digits

The SNF of H0/H is:

1p

p

. Hence 2 diffused digits.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Diffused digits: example

Diffused digits: the latticeLet

H =⟨ 1

pp2

,

pp2

2p4

,

2pp2

2p3

⟩.

Then

H0 =

1p

p2

.

Number of diffused digits

The SNF of H0/H is:

1p

p

.

Hence 2 diffused digits.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Diffused digits: example

Diffused digits: the latticeLet

H =⟨ 1

pp2

,

pp2

2p4

,

2pp2

2p3

⟩.

Then

H0 =

1p

p2

.

Number of diffused digits

The SNF of H0/H is:

1p

p

. Hence 2 diffused digits.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Comparison: ∏ni=1 Ai mod M

Modulus M nGain of precision

Jagged Lattice(not dif. + dif.)

X 5 + X 2 + 1(Irred. mod 2)

10 0.2 0.2 + 0.050 4.2 4.2 + 0.0100 11.2 11.2 + 0.0

X 5 + 1(Sep. mod 2)

10 0.4 0.9 + 6.050 5.6 11.1 + 42.0100 13.6 27.0 + 87.0

X 5 + 2(Eisenstein)

10 6.2 6.2 + 0.050 44.0 44.0 + 0.0100 92.5 92.5 + 0.0

(X + 1)5 + 2(Shift Eisenstein)

10 0.6 4.7 + 1.450 7.1 42.6 + 1.4100 15.1 91.8 + 1.4

X 5 + X + 2(Two slopes)

10 1.7 7.9 + 9.850 8.1 70.7 + 59.8100 16.1 152.6 + 125.9

Figure: Precision for modular multiplication

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilities

Input: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

no diffused digits

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: two possibilitiesInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

×Ai

Zp[X ] Zp[X ]

diffused digits

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with lattice

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 0

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 1

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 2

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 3

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 3

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 4

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 5

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 5

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 10

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsDivision and Differential Precision

Modular Multiplication

Qualitative understanding: long-termInput: P, A1, . . . , An ∈ Zp[X ]d known at precision O(pN)Output: the product A1A2 · · · An mod P

1. A = 1 + O(pN)2. for i in 1, 2, . . . n:3. A = A · Ai

4. return A

Zp[X ] Zp[X ]

without lattice with latticei = 100

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Newton polygon of a polynomial

DefinitionLet f (X ) = a0 + · · · + anX n ∈ Qp[x ] (remark: QpJxK would also be fine).

Let U = {(0, v(a0)), . . . , (n, v(an))}.We define the Newton polygon of f , NP(f ), as the lower convex hullof U.By lower convex hull, we mean the points of the convex hull of U belowthe straight line from (0, (a0)) to (n, v(an)).

PropositionLet Newtf be the (point-wise) biggest convex mapping below U. Thenthe graph of Newtf is (the lower frontier of) NP(f ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Newton polygon of a polynomial

DefinitionLet f (X ) = a0 + · · · + anX n ∈ Qp[x ] (remark: QpJxK would also be fine).Let U = {(0, v(a0)), . . . , (n, v(an))}.

We define the Newton polygon of f , NP(f ), as the lower convex hullof U.By lower convex hull, we mean the points of the convex hull of U belowthe straight line from (0, (a0)) to (n, v(an)).

PropositionLet Newtf be the (point-wise) biggest convex mapping below U. Thenthe graph of Newtf is (the lower frontier of) NP(f ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Newton polygon of a polynomial

DefinitionLet f (X ) = a0 + · · · + anX n ∈ Qp[x ] (remark: QpJxK would also be fine).Let U = {(0, v(a0)), . . . , (n, v(an))}.We define the Newton polygon of f , NP(f ), as the lower convex hullof U.

By lower convex hull, we mean the points of the convex hull of U belowthe straight line from (0, (a0)) to (n, v(an)).

PropositionLet Newtf be the (point-wise) biggest convex mapping below U. Thenthe graph of Newtf is (the lower frontier of) NP(f ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Newton polygon of a polynomial

DefinitionLet f (X ) = a0 + · · · + anX n ∈ Qp[x ] (remark: QpJxK would also be fine).Let U = {(0, v(a0)), . . . , (n, v(an))}.We define the Newton polygon of f , NP(f ), as the lower convex hullof U.By lower convex hull, we mean the points of the convex hull of U belowthe straight line from (0, (a0)) to (n, v(an)).

PropositionLet Newtf be the (point-wise) biggest convex mapping below U. Thenthe graph of Newtf is (the lower frontier of) NP(f ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Newton polygon of a polynomial

DefinitionLet f (X ) = a0 + · · · + anX n ∈ Qp[x ] (remark: QpJxK would also be fine).Let U = {(0, v(a0)), . . . , (n, v(an))}.We define the Newton polygon of f , NP(f ), as the lower convex hullof U.By lower convex hull, we mean the points of the convex hull of U belowthe straight line from (0, (a0)) to (n, v(an)).

PropositionLet Newtf be the (point-wise) biggest convex mapping below U. Thenthe graph of Newtf is (the lower frontier of) NP(f ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3

NP(P)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Vocabulary

DefinitionA slope of the Newton polygon of f is an element of Newt ′

f ([0, n]).

DefinitionIf λ is a slope of Newtf , we call segment of slope λ of Newtf the set{(x , Newtf (x))⧸Newt ′

f (x) = λ}.

DefinitionThe length of this slope is the length of its projection on the x -axis.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Fundamental theorem of Newton polygons

Theoremf has a root of valuation λ iff −λ is a slope of Newtf .

Moreover, the number of roots of f (with multiplicity) of valuation λ, isthe length of the segment of slope −λ of Newtf .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Fundamental theorem of Newton polygons

Theoremf has a root of valuation λ iff −λ is a slope of Newtf .Moreover, the number of roots of f (with multiplicity) of valuation λ, isthe length of the segment of slope −λ of Newtf .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3

NP(P)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3

NP(P)

One slope for 0 of length 2, one slope 1/2 of length 2.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3

NP(P)

One slope for 0 of length 2, one slope 1/2 of length 2.

Two roots of valuation 0, two of valuation −1/2.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Basic operations

Proposition (Addition)If f and g are two polynomials, then the Newton polygon of f + g can belower-bounded by taking the lower convex hull for the vertices of Newtfand Newtg .

Proposition (Multiplicativity)If f and g are two polynomials, then the Newton polygon of fg has forslopes that of f and g, with length the sum of that of f and g.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Basic operations

Proposition (Addition)If f and g are two polynomials, then the Newton polygon of f + g can belower-bounded by taking the lower convex hull for the vertices of Newtfand Newtg .

Proposition (Multiplicativity)If f and g are two polynomials, then the Newton polygon of fg has forslopes that of f and g, with length the sum of that of f and g.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Some remarks

PropositionIf P ∈ Qp[X ] is irreducible, then all its roots have the same valuation.Hence, NewtP has only one slope.

RemarkThe converse is false. For instance, (X − 1)(X − 2) over Q5.

CorollaryIf NewtP has more than one slope, P is not irreducible.

RemarkThere are good irreducibility criterion based on testing whether one slopecan be obtained by multiplication (namely, Dumas and Eisenstein).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Some remarks

PropositionIf P ∈ Qp[X ] is irreducible, then all its roots have the same valuation.Hence, NewtP has only one slope.

RemarkThe converse is false. For instance, (X − 1)(X − 2) over Q5.

CorollaryIf NewtP has more than one slope, P is not irreducible.

RemarkThere are good irreducibility criterion based on testing whether one slopecan be obtained by multiplication (namely, Dumas and Eisenstein).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Basics

Some remarks

PropositionIf P ∈ Qp[X ] is irreducible, then all its roots have the same valuation.Hence, NewtP has only one slope.

RemarkThe converse is false. For instance, (X − 1)(X − 2) over Q5.

CorollaryIf NewtP has more than one slope, P is not irreducible.

RemarkThere are good irreducibility criterion based on testing whether one slopecan be obtained by multiplication (namely, Dumas and Eisenstein).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

What is the Newton polygon of the remainder in the division of Aby B (in Qp[X ])?What is the Newton polygon of the quotient in the division of A byB (in Qp[X ])?

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

NP(R)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

NP(R)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of puX n by B : puX n = BQ + R

valuation

orderdd − 1

NP(B)

n

u puX n

NP(R)

n − d

NP(Q)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of A by B : A = BQ + R

valuation

orderdd − 1

NP(B)NP(A)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of A by B : A = BQ + R

valuation

orderdd − 1

NP(B)NP(A)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Euclidean division

Euclidean division and Newton polygon

Lemma (Division lemma)

Division of A by B : A = BQ + R

valuation

orderdd − 1

NP(B)NP(A)NP(R) NP(Q)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Precision: Return on Modular Multiplication

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Precision: Return on Modular Multiplication

Handling the precision

Lattice precisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Newton precisionFor A = BQ + R with A known with precision-polygon φ, we can applythe previous construction to φ divided by B to obtain the precision on Qand R.

RemarkThis proved to be useful to handle precision for the computation of thecharacteristic polynomial.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Precision: Return on Modular Multiplication

Handling the precision

Lattice precisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Newton precisionFor A = BQ + R with A known with precision-polygon φ, we can applythe previous construction to φ divided by B to obtain the precision on Qand R.

RemarkThis proved to be useful to handle precision for the computation of thecharacteristic polynomial.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Precision: Return on Modular Multiplication

Handling the precision

Lattice precisionA and B are known with precision lattice HA and HB . Then (HQ , HR) aregiven by the Euclidean division of HA − QHB by B.

Newton precisionFor A = BQ + R with A known with precision-polygon φ, we can applythe previous construction to φ divided by B to obtain the precision on Qand R.

RemarkThis proved to be useful to handle precision for the computation of thecharacteristic polynomial.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsNewton Polygons

Precision: Return on Modular Multiplication

Comparison: ∏ni=1 Ai mod M

Modulus M nGain of precision

Jagged Newton Lattice(not dif. + dif.)

X 5 + X 2 + 1(Irred. mod 2)

10 0.2 0.2 0.2 + 0.050 4.2 4.2 4.2 + 0.0100 11.2 11.2 11.2 + 0.0

X 5 + 1(Sep. mod 2)

10 0.4 0.4 0.9 + 6.050 5.6 5.6 11.1 + 42.0100 13.6 13.6 27.0 + 87.0

X 5 + 2(Eisenstein)

10 6.2 6.2 6.2 + 0.050 44.0 44.0 44.0 + 0.0100 92.5 92.5 92.5 + 0.0

(X + 1)5 + 2(Shift Eisenstein)

10 0.6 0.6 4.7 + 1.450 7.1 7.1 42.6 + 1.4100 15.1 15.1 91.8 + 1.4

X 5 + X + 2(Two slopes)

10 1.7 1.7 7.9 + 9.850 8.1 8.1 70.7 + 59.8100 16.1 16.1 152.6 + 125.9

Figure: Precision for modular multiplication

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Factoring respecting slopes

TheoremLet f ∈ Qp[X ]. Then,

We can write f =∏

i fi .The fi ’s are all of one slope.They have respectively different slopes.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Factoring respecting slopes

TheoremLet f ∈ Qp[X ]. Then,

We can write f =∏

i fi .

The fi ’s are all of one slope.They have respectively different slopes.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Factoring respecting slopes

TheoremLet f ∈ Qp[X ]. Then,

We can write f =∏

i fi .The fi ’s are all of one slope.

They have respectively different slopes.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Factoring respecting slopes

TheoremLet f ∈ Qp[X ]. Then,

We can write f =∏

i fi .The fi ’s are all of one slope.They have respectively different slopes.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3.

NP(P)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3.

NP(P)

NP(f1)

NP(f2)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3.

NP(P)

NP(f1)

NP(f2)

P = (2 + 3X + X 2) × (1 + 3X 2)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

A Newton polygon

1

2

0 1 2 3 4

P = 2 + 3X + 7X 2 + 9X 3 + 3X 4, over Q3.

NP(P)

NP(f1)

NP(f2)

P = (2 + 3X + X 2) × (1 + 3X 2)

Remark: 2 + 3X + X 2 = (1 + X )(1 + 2X ).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

A Newton iteration

The iterationAlready found in Polynomial root finding over local rings and applicationto error correcting codes by Berthomieu, Lecerf, Quintin:

Ai+1 := Ai + (ViP mod Ai)Bi+1 := P \quo Ai+1

Vi+1 := (2Vi − V 2i Bi+1) mod Ai+1

The resultAi , Bi , Vi converge quadratically to A, B, V such that AB = P, V is theinverse of B modulo A and A, B have the desired Newton polygons.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Ideas on the proof

About the proofWe monitor Ri = Ai+1 − Ai and Si = P mod Ai .

We prove that both NP(Ri) and NP(Si) goes to infinity.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step. Euclidean division.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step. Euclidean division.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step. Euclidean division.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step. Euclidean division.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step. Euclidean division.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step.

κ

0 dd−1 d+1

A0

NF (P)

NP(R0)NP(S0)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step.

κ

0 dd−1 d+1

A0

NF (P)

NP(R0)NP(S0)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

First step.

κ

0 dd−1 d+1

A0

NF (P)

NP(R0)NP(S0)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

Second Step.

κ

0 dd−1 d+1

A0

NF (P)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

Second Step.

2κκ

0 dd−1 d+1

A0

NF (P)

NP(R1)NP(S1)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

Second Step.

2κκ

0 dd−1 d+1

A0

NF (P)

NP(R1)NP(S1)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

Third Step.

κ

0 dd−1 d+1

A0

NF (P)

NP(R2)NP(S2)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

A Newton scheme

Illustration

Third Step.

κ

0 dd−1 d+1

A0

NF (P)

NP(R2)NP(S2)

slope = λ1

slope = λ0φ

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

Table of contents

1 Division and Differential Precisionp-Adic PrecisionStudy of the divisionModular Multiplication

2 Newton PolygonsBasicsEuclidean divisionPrecision: Return on Modular Multiplication

3 Slope factorizationA Newton schemeApplying differential precision

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

What about precision?

SettingLet FA : P 7→ A(1) be the application such that P = A(1)B(1), withA(1), B(1) corresponding to the slopes before/after the breakpoint d .

DifferentialThe application FA : P 7→ A(1) is of class C1. Its differential at somepoint P is the linear mapping

dP 7→ dA(1) = (V (1) dP) mod A(1)

where A(1)B(1) = P and V (1) is the inverse of B(1) modulo A(1).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

What about precision?

SettingLet FA : P 7→ A(1) be the application such that P = A(1)B(1), withA(1), B(1) corresponding to the slopes before/after the breakpoint d .

DifferentialThe application FA : P 7→ A(1) is of class C1. Its differential at somepoint P is the linear mapping

dP 7→ dA(1) = (V (1) dP) mod A(1)

where A(1)B(1) = P and V (1) is the inverse of B(1) modulo A(1).

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

Some numerical results: A 7→ AB 7→ A.

Polynomials PrecisionMean gain of precision

Jagged Newton Lattice

dA B absolute −14.5 −14.7 0.0

relative −1.5 −3.6 0.0

dA B absolute 0.0 0.0 0.0

relative −0.3 −1.2 0.0

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precision

Step-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.

Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.

Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computations

Diffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.

Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

To sum up

On p-adic precisionStep-by-step analysis: as a first step. Can show differentiability andnaïve loss in precision during the computation.Differential calculus: intrinsic and can handle both gain and loss.Lattice precision: achieving and understandig the best precision.

On polynomial computationsDiffused digits for modular multiplication.Newton iteration for slope factorisation.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

References

Initial article

Xavier Caruso, David Roe and Tristan Vaccon Tracking p-adicprecision, ANTS XI, 2014.

Linear Algebra

Xavier Caruso, David Roe and Tristan Vaccon p-adic stability inlinear algebra, ISSAC 2015.

Polynomial Computations

Xavier Caruso, David Roe and Tristan Vaccon Division andSlope Factorization of p-Adic Polynomials, ISSAC 2016.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Division and Slope Factorization of p-Adic PolynomialsSlope factorization

Applying differential precision

Thank you for your attentionThanks

x + B

f

B

f ′(x)

f (x) + f ′(x) · B

f ′(x) · B

x

x + O(pN′) y + O(pM′) ⊂ f (x) + O(pN)

Recommended