Alphabeta Math
TheoremStatement: AI-adaptedProof: AI-adaptedSession-authored (Fable 5 assisted)precheck passjudge pass (z-ai/glm-5.2)verified 2026-07-26 (claude-opus-5)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Minkowski's inequality for finite sums (rational exponent)

Statement

Let nNn \in \mathbb{N}, let a0,,an1a_0, \dots, a_{n-1} and b0,,bn1b_0, \dots, b_{n-1} be reals, and let pQp \in \mathbb{Q} with p1p \ge 1 (Order on the rationals). Then

(k<nak+bkp)1/p    (k<nakp)1/p+(k<nbkp)1/p.\Big(\sum_{k<n}|a_k + b_k|^{p}\Big)^{1/p} \;\le\; \Big(\sum_{k<n}|a_k|^{p}\Big)^{1/p} + \Big(\sum_{k<n}|b_k|^{p}\Big)^{1/p}.

All exponents appearing are positive rationals, so every power is defined for a nonnegative base by Rational powers ara^r of a positive base and its supplementary clause 0r=00^{r} = 0.

The conjugate exponent is rational exactly because pp is. For p>1p > 1 the proof runs through Hölder with q=p/(p1)q = p/(p-1), and a quotient of rationals with nonzero denominator is a rational (Arithmetic on the rationals). Had pp been an arbitrary real, qq would still be a real, but akp|a_k|^{p} would already be undefined: the whole statement lives inside the rational exponents built on this page, as the closing remark of the page explains.

On the case p=1p = 1. It reads k<nak+bkk<nak+k<nbk\sum_{k<n}|a_k+b_k| \le \sum_{k<n}|a_k| + \sum_{k<n}|b_k| and follows by summing the two-term triangle inequality (The triangle inequality) termwise. It is not literally the finite-sum triangle inequality Triangle inequality for finite sums, which compares k<nck\big|\sum_{k<n} c_k\big| with k<nck\sum_{k<n}|c_k| for one list; combining the two at p=1p = 1 gives the familiar k<n(ak+bk)k<nak+k<nbk\big|\sum_{k<n}(a_k+b_k)\big| \le \sum_{k<n}|a_k| + \sum_{k<n}|b_k|.

Facts & Assumptions

Given: A natural nn, reals a0,,an1a_0, \dots, a_{n-1} and b0,,bn1b_0, \dots, b_{n-1}, and a rational p1p \ge 1. Write S:=k<nak+bkpS := \sum_{k<n}|a_k+b_k|^{p}, U:=(k<nakp)1/pU := \big(\sum_{k<n}|a_k|^{p}\big)^{1/p}, V:=(k<nbkp)1/pV := \big(\sum_{k<n}|b_k|^{p}\big)^{1/p}, and ck:=ak+bkp1c_k := |a_k+b_k|^{\,p-1} when p>1p > 1.

[L1]

Hölder's inequality (Hölder's inequality for finite sums (rational exponents)): for conjugate rationals p,q>1p, q > 1, k<nxkyk(k<nxkp)1/p(k<nykq)1/q\sum_{k<n}|x_k y_k| \le \big(\sum_{k<n}|x_k|^{p}\big)^{1/p}\big(\sum_{k<n}|y_k|^{q}\big)^{1/q}.

[L2]

Laws of finite sums (Laws of finite sums and finite products, Finite sums and finite products, by recursion): additivity, scaling, monotonicity, and the fact that a sum of nonnegative terms is nonnegative.

[L3]

Rational power laws (Laws of rational exponents, Rational powers ara^r of a positive base, Existence and uniqueness of nn-th roots: a unique a1/n0a^{1/n} \ge 0 with (a1/n)n=a(a^{1/n})^n = a): for u>0u > 0 and rationals r,sr, s: ur>0u^{r} > 0, ur+s=urusu^{r+s} = u^{r}u^{s}, (ur)s=urs\big(u^{r}\big)^{s} = u^{rs}, u1=uu^{1} = u and ur=(ur)1u^{-r} = \big(u^{r}\big)^{-1}; and 0r=00^{r} = 0 for every rational r>0r > 0.

[L4]

Absolute values (Basic properties of the absolute value, Absolute value in an ordered field, The triangle inequality): x0|x| \ge 0, xy=xy|xy| = |x||y|, and x+yx+y|x+y| \le |x| + |y|.

[L5]

Rational arithmetic (Arithmetic on the rationals, The rationals as equivalence classes of pairs of integers, Order on the rationals), carried out in the totally ordered field Q\mathbb{Q} (The rationals form a totally ordered field, which is what makes the order comparisons below legitimate, and which supplies totality, compatibility with addition and closure of the positives under multiplication but NOT 1>01 > 0; that is The multiplicative identity is positive, valid in Q\mathbb{Q} because Q\mathbb{Q} is an ordered field): for rational p>1p > 1 one has p1>0p - 1 > 0 and, since 1>01 > 0, also p>p1>0p > p - 1 > 0, so the number q:=p/(p1)q := p/(p-1) is a rational with q>1q > 1, 1p+1q=1\frac1p + \frac1q = 1, (p1)q=p(p-1)q = p and 11q=1p1 - \frac1q = \frac1p.

[L6]

Order arithmetic: Order is preserved by adding a constant and by adding inequalities and Sign rules for products and monotonicity of multiplication state adding inequalities and scaling by a positive element for the STRICT order only, so the nonstrict forms used below (adding two \le, and scaling a \le by a nonnegative element) are those statements together with the case of equality, which is settled by trichotomy (Ordered field); and the inverse of a positive element is positive (Inverses of positives are positive, and reciprocation reverses order, claim 1).

Proof

technique · direct
1.1

Every quantity is defined and nonnegative: ak+bkp|a_k + b_k|^{p}, akp|a_k|^{p} and bkp|b_k|^{p} are nonnegative because p>0p > 0, hence so are SS, UU and VV.

L2L3L4
1.2

The case p=1p = 1: summing the two-term triangle inequality termwise and using monotonicity and additivity gives k<nak+bkk<n(ak+bk)=k<nak+k<nbk\sum_{k<n}|a_k+b_k| \le \sum_{k<n}\big(|a_k| + |b_k|\big) = \sum_{k<n}|a_k| + \sum_{k<n}|b_k|, and since x1/1=xx^{1/1} = x this is exactly the assertion at p=1p = 1.

L2L3L4
1.3

The case S=0S = 0: the left-hand side is 01/p=00^{1/p} = 0, which is at most the nonnegative right-hand side.

L2L3
2.1

Assume from now on p>1p > 1 and S>0S > 0, and put q:=p/(p1)q := p/(p-1), a rational with q>1q > 1 conjugate to pp, so that (p1)q=p(p-1)q = p and 11q=1p1 - \frac1q = \frac1p.

step 1.1step 1.2step 1.3L5
3.1

Splitting each term: for u0u \ge 0 one has up=u1+(p1)=uup1u^{p} = u^{1 + (p-1)} = u \cdot u^{\,p-1}, valid for u>0u > 0 by the addition law and for u=0u = 0 because both sides are 00; applying this with u=ak+bku = |a_k+b_k| and then the triangle inequality, multiplied by the nonnegative factor ck=ak+bkp1c_k = |a_k+b_k|^{\,p-1}, gives ak+bkp(ak+bk)ck=akck+bkck|a_k+b_k|^{p} \le \big(|a_k| + |b_k|\big) c_k = |a_k| c_k + |b_k| c_k for every k<nk < n.

step 2.1L3L4L6
3.2

The auxiliary list has qq-th power sum SS: ckq=(ak+bkp1)q=ak+bk(p1)q=ak+bkpc_k^{\,q} = \big(|a_k+b_k|^{\,p-1}\big)^{q} = |a_k+b_k|^{(p-1)q} = |a_k+b_k|^{p} for ak+bk>0|a_k+b_k| > 0 by the iterated-power law, and both sides are 00 when ak+bk=0a_k + b_k = 0; hence k<nckq=S\sum_{k<n}|c_k|^{q} = S.

step 2.1L3L4
4.1

Summing the termwise bound: Sk<nakck+k<nbkckS \le \sum_{k<n}|a_k| c_k + \sum_{k<n}|b_k| c_k.

step 3.1L2
4.2

Applying Hölder to the pairs (ak,ck)(a_k, c_k) and to (bk,ck)(b_k, c_k), and using akck=akck|a_k c_k| = |a_k| c_k since ck0c_k \ge 0: k<nakckUS1/q\sum_{k<n}|a_k| c_k \le U\,S^{1/q} and k<nbkckVS1/q\sum_{k<n}|b_k| c_k \le V\,S^{1/q}.

step 3.2step 2.1L1L4
5.1

Combining, S(U+V)S1/qS \le (U + V)\,S^{1/q}.

step 4.1step 4.2L6
6.1

Dividing by S1/q>0S^{1/q} > 0, which is legitimate because S>0S > 0, and computing S(S1/q)1=S1S1/q=S11/q=S1/pS \big(S^{1/q}\big)^{-1} = S^{1}S^{-1/q} = S^{\,1 - 1/q} = S^{1/p}, we obtain S1/pU+VS^{1/p} \le U + V; together with the case p=1p = 1 and the case S=0S = 0 this proves the inequality for every rational p1p \ge 1.

step 5.1step 2.1step 1.2step 1.3L3L5L6

Depends on

Used by

Dependency tree · next 3 levels

Direct dependencies and their dependencies through the next three levels: 78 results over 30 levels. An arrow runs from a result to what uses it, and this result sits at the bottom with a heavier outline. Click the chart to enlarge it.

Sources