Alphabeta Math
ExampleConstruction: AI-adaptedVerification: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6.1-sol)
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.

Five finite proofs of Bernoulli, with rational exponents and the quadratic refinement

Example

Five proof routes verify the same integer statement: for x≥−1 and a natural n, (1+x)n≥1+nx, with equality exactly when n≤1 or x=0. The routes are ordinary induction, geometric-sum comparison, an exact nonnegative supporting-line remainder, AM-GM, and a finite binomial/positive-weight expansion. They cover −1≤x<0 as well as x≥0; 00=1 is the integer-power convention. In addition, with rational r and x≥−1 (requiring x>−1 when r<0), (1+x)r≥1+rx for r≥1 or r≤0, and (1+x)r≤1+rx for 0≤r≤1. Equality is exactly x=0 or r∈{0,1}. For rational r≥2, the promised quadratic refinement follows in the stronger form (1+x)r≥1+rx+(r/2)x2; hence it holds with any natural coefficient m satisfying 2m≤r, including the coefficient customarily written ⌊r/2⌋. For 1≤r<2 that floor coefficient is zero and the ordinary bound applies. The stronger refinement has equality exactly when x=0 or r=2. At x=−1, the supplementary clause of Rational powers ar of a positive base supplies 0r=0 for rational r>0, and Integer powers am supplies 00=1; negative powers of zero remain undefined. No irrational exponent or exponential/general convexity theorem is asserted.

Facts & Assumptions

Given: A natural n and a real x≥−1, with y:=1+x≥0; rationally, a rational r and the base y>0.

[L1]

Finite sums and products (Finite sums and finite products, by recursion, Laws of finite sums and finite products): recursion clauses, splitting, additivity, scaling, constant sums and products, monotonicity; the constant product ∏k<mc=cm follows by induction from the matching base values 1 and successor multiplication by c in [L3] and [L8]; and the strict comparison obtained by splitting off one strictly larger term and adding the strict inequality of [L2].

[L2]

Order arithmetic (Ordered field, Order is preserved by adding a constant and by adding inequalities, Sign rules for products and monotonicity of multiplication): the order is total and transitive, adding constants and adding strict inequalities preserve it, multiplication by a positive element preserves it and by a negative element reverses it.

[L3]

Powers (Integer powers am, Laws of integer exponents, Monotonicity of x↦xn and of n↦an): u0=1, uj+1=uju, um+n=umun, (um)n=umn, (uv)n=unvn; u≥0 gives uj≥0; and u↦uj is strictly increasing on the nonnegatives for j≥1.

[L4]

Squares (Squares of nonzero elements are positive, Multiplication by zero: 0⋅a=0): u2>0 for u≠0 and 02=0, so u2≥0 with equality only at u=0.

[L5]

Bernoulli's inequality (Bernoulli's inequality (1+x)n≥1+nx): for x≥−1 and natural n, (1+x)n≥1+nx, with equality exactly when n≤1 or x=0; its published proof is the ordinary induction with that equality analysis.

[L6]

Difference factorisation (Factorisation of bn−an, and the resulting Lipschitz estimate): bn−an=(b−a)∑k=0n−1akb n−1−k for n≥1.

[L7]

AM-GM (The arithmetic mean, geometric mean inequality): for nonnegative entries of length m≥1, ∏≤(mean)m, with equality exactly for constant lists; and weighted AM-GM (Weighted AM-GM inequality with rational weights): for xk>0 and rationals wk≥0 summing to 1 when read as real numbers, ∏xkwk≤∑wkxk.

[L9]

Rationals and their embedding (Every rational has a positive-denominator representative, The naturals embed in the integers, Order on the rationals, Order on the integers, The unique embedding of ℚ into an ordered field, Canonical naturals are positive and strictly increasing, Inverses of positives are positive, and reciprocation reverses order): every rational has a representative m/N with m a natural and N≥1 when it is nonnegative; the canonical embedding ι:Q→R is injective and order-preserving, and the real number m is positive for m≥1; inverses of positives are positive.

[L10]

Rational powers and their laws (Rational powers ar of a positive base, Laws of rational exponents, Monotonicity of r↦ar and of a↦ar): ar+s=aras, (ar)s=ars, a1=a for a>0; ar>0; 0r=0 for rational r>0 and 00=1; and for fixed positive exponent the base map is strictly increasing, so yr=1 with y>0 forces y=1.

Verification

technique · direct; five routes for the integer statement, then the rational forms and the quadratic refinement
1.1L3L5algebra

The degenerate cases, common to all routes: if n=0 both sides are 1; if n=1 both sides are 1+x; if x=0 both sides are 1, since 1n=1 for every natural n; so equality holds whenever n≤1 or x=0, and it remains to prove the strict inequality (1+x)n>1+nx for n≥2 and x≠0 with x≥−1.

1.2L5

Route 1, ordinary induction: the published theorem [L5] states the inequality with exactly the claimed equality clause and its published proof is the ordinary induction, so this route is verified by quoting [L5] as a whole; no duplicate induction is minted.

1.3L1L2L6algebra

Route 2, geometric-sum comparison: for n≥2 and y≥0, y≠1, [L6] gives yn−1=(y−1)∑k<nyk; if y>1 then every term yk is at least 1 and the term k=1 is strictly larger, so the sum is strictly greater than n by the strict sum comparison of [L1] and y−1>0 gives yn−1>n(y−1), while if 0≤y<1 every term is at most 1 and the term k=1 is strictly smaller, so the sum is strictly less than n and multiplication by the negative y−1 reverses the comparison in [L2] to give again yn−1>n(y−1); substituting y=1+x gives (1+x)n>1+nx for n≥2, x≠0.

1.4L1L4L6algebra

Route 3, exact nonnegative remainder: for n≥2 and y≥0 one has the identity yn−ny+(n−1)=(y−1)2∑j=0n−2(n−1−j)yj: indeed [L6] gives yn−1=(y−1)∑k<nyk, so yn−ny+(n−1)=(y−1)(∑k<nyk−n), and ∑k<nyk−n=∑k=1n−1(yk−1)=∑k=1n−1(y−1)∑j<kyj=(y−1)∑j=0n−2(n−1−j)yj, since each yj occurs once for each k>j with k≤n−1, of which there are n−1−j; every coefficient n−1−j is a positive natural and every yj≥0, so the last sum is at least its constant term n−1>0, and (y−1)2≥0 with equality only at y=1 by [L4]; hence yn−ny+(n−1)>0 for y≠1, that is (1+x)n>1+nx for n≥2 and x≠0, with the identity giving 0=0 and equality at x=0.

1.5L3L7algebra

Route 4, AM-GM applied to (yn,1,…,1): for n≥2 and y≥0 the n entries yn,1,…,1 are nonnegative, their product is yn, and their geometric mean is (yn)1/n=y because y≥0 and yn=yn; [L7] gives y≤(yn+(n−1))/n, that is yn≥ny−(n−1)=1+n(y−1)=1+nx; if y≠1 the list is not constant, so [L7]'s equality clause makes the inequality strict, and y=1 means x=0; this proves the integer statement for n≥2 by a fourth route, with strictness for x≠0.

1.6L1L3L8L9algebrainduction

Route 5, finite Pascal rows: set b0,0=1, b0,k=0 for k≥1, and define bm+1,0=1, bm+1,k=bm,k+bm,k−1 for k≥1. This is recursion on m in the set of functions N→N, with a successor map sending a row to the displayed new row. Coefficients in field expressions denote their canonical images. Induction gives bm,k=0 for k>m, since both predecessor entries then vanish, and all coefficients are nonnegative. It also gives bm,0=bm,m=1 and bm,1=m: the diagonal successor is 0+1, and the first column increases by 1. The expansion (u+v)m=∑k=0mbm,kukvm−k holds at m=0; multiplying the row-m expansion by u+v and shifting the u-part by one index gives the coefficient bm,k+bm,k−1 at each interior index, and coefficient 1 at the endpoints, proving the successor expansion. Finally kbm,k=mbm−1,k−1 for 1≤k≤m, by induction on m. At m=1 it is 1=1, and the cases k=1,m follow from the endpoint and first-column values. For 2≤k≤m−1, the preceding-row identities and Pascal recursion give kbm,k=kbm−1,k+kbm−1,k−1=(m−1)bm−2,k−1+(m−1)bm−2,k−2+bm−1,k−1=mbm−1,k−1, closing the induction.

1.7L1L3L7L9L10algebra

Strictness transfer for weighted AM-GM: let x0,…,xm−1>0 and let wk≥0 be rationals with ∑k<mwk=1 as real numbers. The common-denominator and repeated-block construction in the proof of Weighted AM-GM inequality with rational weights, steps 1.1 through 5.1, gives wk=nk/N with nk natural, N≥1, and a list of N positive entries, repeating xk exactly nk times. Its sum is ∑k<mnkxk and its product is ∏k<mxknk. Put H:=∏k<mxkwk>0 and M:=∑k<mwkxk>0. The iterated-power law in [L10], applied to each factor, and the natural product law in [L3], applied inductively, give HN=∏k<mxknk; the repeated list has mean M. AM-GM in [L7] therefore gives HN≤MN, with equality exactly when the repeated list is constant. By [L3], H≤M with equality under the same condition; in particular the weighted inequality is strict whenever two bases carrying positive weights differ.

1.8L2L7L10algebra

Rational Bernoulli forms, inequality directions, for y>0 and rational r: (a) if r>1, the weights 1/r and 1−1/r are rational, nonnegative and sum to 1, so [L7] applied to the positive bases yr and 1 gives y=(yr)1/r1 1−1/r≤(1/r)yr+(1−1/r), and multiplying by the positive r gives yr≥1+r(y−1); (b) if r<0, the weights −r/(1−r) and 1/(1−r) are rational, nonnegative and sum to 1, and with bases y and yr the geometric side is y−r/(1−r)(yr)1/(1−r)=y0=1 by the addition law of [L10], so 1≤(−r y+yr)/(1−r) and multiplying by the positive 1−r gives yr≥1+r(y−1); (c) if 0<r<1, the weights r and 1−r give yr=yr1 1−r≤r y+(1−r)=1+r(y−1); in every case the powers of 1 used are 1 by [L10].

2.1step 1.6L1L3algebra

Route 5 for x>0: the expansion of step 1.6 at (u,v)=(x,1) gives (1+x)n=∑k=0nbn,kxk=1+nx+∑k=2nbn,kxk with bn,0=1, bn,1=n, all coefficients ≥0 and bn,n=1, so (1+x)n≥1+nx+xn>1+nx for n≥2 and x>0; equality at x≠0, n≥2 is impossible in this range.

2.2step 1.6L1L3algebra

Route 5 for −1≤x<0: put t:=−x∈(0,1] and y:=1−t=1+x≥0; the expansion of step 1.6 gives 1=(t+y)n=∑k=0nbn,ktky n−k and, using the coefficient identity k bn,k=n bn−1,k−1 and the expansion at n−1, also nt=∑k=1nk bn,ktky n−k; subtracting the first identity from the second, nt−(1−yn)=∑k=1n(k−1)bn,ktky n−k≥0 with k=n term (n−1)tn>0 for n≥2, so nt>1−yn and yn>1−nt=1+nx; this proves the integer statement for the negative range x∈[−1,0) by positive weights, without discarding alternating terms.

2.3step 1.7step 1.8L10algebra

Rational equality and strictness: for y>0 and rational r≠0,1, if x≠0, that is y≠1, then in each case of step 1.8 the two bases carrying positive weight differ: for r>1, yr=1 would force y=1 by [L10]; for 0<r<1 the bases are y and 1; and for r<0, yr=y would imply yr−1=1, hence y1−r=1 by taking the inverse through [L10], again forcing y=1 since 1−r>0. Thus the strictness transfer of step 1.7 makes the weighted inequality strict and, after the positive multiplications of step 1.8, the rational Bernoulli inequality is strict as well; if x=0, that is y=1, every base equals 1 and both sides equal 1, and the cases r=0 and r=1 give y0=1=1+0 and y1=y=1+x identically, so equality holds exactly for x=0 or r∈{0,1}.

2.4step 1.8L10algebra

Endpoint x=−1: for rational r>0 the supplementary clause gives 0r=0 by [L10], so the left-hand side is 0 and the right-hand side is 1−r; for r≥1 the assertion 0≥1−r holds with equality exactly at r=1, and for 0<r≤1 the reversed assertion 0≤1−r holds with equality exactly at r=1; at r=0 the integer convention gives 00=1=1+0, an equality, and negative rational exponents are not defined at 0 and are excluded by the domain of the statement.

3.1step 1.1step 1.2step 1.3step 1.4step 1.5step 2.1step 2.2L5

The integer statement is proved with five routes and the sharp equality case: Route 1 is [L5] quoted in step 1.2; Route 2 is the geometric-sum comparison of step 1.3; Route 3 is the supporting-line remainder of step 1.4, whose identity has positive coefficients and double zero exactly at y=1; Route 4 is the AM-GM application of step 1.5; and Route 5 is the binomial expansion, which gives strictness for x>0 in step 2.1 and for −1≤x<0 in step 2.2; all routes cover x≥−1, the degenerate equalities n≤1 or x=0 are those of step 1.1, and every route gives strict inequality for n≥2, x≠0.

3.2step 1.8step 2.3step 2.4L10

The rational forms hold on the stated domain: for x>−1 the inequalities are the three cases of step 1.8, their equality and strictness pattern is step 2.3, and the endpoint x=−1 is step 2.4; hence (1+x)r≥1+rx for r≥1 or r≤0 and (1+x)r≤1+rx for 0≤r≤1, in each case exactly on the stated domain, with equality exactly when x=0 or r∈{0,1}.

4.1step 3.2L3L4L10algebra

The quadratic refinement: let r≥2 be rational and x≥−1. If x>−1, put y:=1+x>0, z:=y2>0 and s:=r/2, a rational with s≥1; the rational Bernoulli lower bound of step 3.2 applied to z and s gives zs≥1+s(z−1), and zs=(y2)r/2=y r while s(z−1)=(r/2)((1+x)2−1)=(r/2)(2x+x2), so yr≥1+rx+(r/2)x2; equality in this chain holds exactly when z=1, that is x=0, or s=1, that is r=2; if x=−1 the left side is 0r=0 and the right side is 1−r+r/2=1−r/2≤0 with equality exactly at r=2; and if m is a natural with 2m≤r, then r/2−m≥0 and x2≥0 give (r/2)x2≥m x2, so the stronger bound implies (1+x)r≥1+rx+m x2; for 1≤r<2 there is no positive natural m with 2m≤r, the coefficient r/2 lies below 1, and the ordinary lower bound is the case m=0.

5.1step 3.1step 3.2step 4.1L5L7∎

Summary: the integer Bernoulli statement holds with the equality clause n≤1 or x=0 by five independent finite routes, two of which (Routes 4 and 5) are new here while Route 1 is the published induction and Routes 2 and 3 are the difference-of-powers and supporting-line arguments (steps 1.2 through 1.5, 2.1, 2.2, 3.1); the rational-exponent forms hold with the equality clause x=0 or r∈{0,1} including the directly verified endpoint (steps 1.8, 2.3, 2.4, 3.2); and the quadratic refinement (1+x)r≥1+rx+(r/2)x2 for r≥2, holds with equality exactly when x=0 or r=2 (step 4.1), which also proves the coefficient weakening and the ordinary bound for 1≤r<2 without assigning them that equality clause; no route invokes a derivative, logarithm, exponential, convergence, or an unproved optimization theorem.

Depends on

Used by

Nothing in the library uses this result yet.

Dependency tree · two levels

62 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources