Alphabeta Math
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.

✓ 9 results · all verified · 7 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 2 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

limsup, liminf, and Subsequential Limits: Examples and Counterexamples

1 · Prerequisites

2 · Summary

3 · Logical flowchart

4 · Definitions, theorems and proofs

None yet.

5 · Examples, counterexamples and false statements

ExampleConstruction: AI-adaptedVerification: AI-generatedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

(−1)k has lim inf⁡=−1 and lim sup⁡=1, so it does not converge

Example

Let (sk) be the alternating sequence of The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1, the sequence usually written sk=(−1)k, characterised by s0=1 and sk+1=−sk. Then

lim inf⁡ksk=−1,lim sup⁡ksk=1,

so the two differ and (sk) neither converges nor diverges to ±∞ (A real sequence converges to L∈R iff lim inf⁡xk=lim sup⁡xk=L, and diverges to ±∞ iff both equal ±∞).

This is the smallest example in which the inequality lim inf⁡≤lim sup⁡ is strict, and it shows exactly what the gap measures: the sequence keeps returning to two different values, and neither of them can be the limit because the other keeps interrupting.

Facts & Assumptions

[L2]

A strictly increasing index map satisfies nj≥j (A strictly increasing index map satisfies nk≥k).

[L5]

Absolute value: ∣t∣=1 forces t=1 or t=−1 (Basic properties of the absolute value, Absolute value in an ordered field).

[L7]

A real sequence converges to L∈R exactly when lim inf⁡=lim sup⁡=L, and diverges to ±∞ exactly when both equal ±∞ (A real sequence converges to L∈R iff lim inf⁡xk=lim sup⁡xk=L, and diverges to ±∞ iff both equal ±∞).

Verification

technique · direct
1.1

Every value of the sequence is 1 or −1, since ∣sk∣=1.

givenL1L5
1.2

For every n∈N both values occur at some index ≥n: sen=1 with en≥n, and son=−1 with on≥n.

givenL1L2
2.1

Hence Tn={1,−1} for every n. Its least upper bound in R‾ is 1, since 1 bounds both elements from above, using −1<1, and any upper bound is ≥1 because 1∈Tn; dually its greatest lower bound is −1.

step 1.1step 1.2L3L4L6
3.1

Therefore the family of tail suprema is the one-element family {1}, whose greatest lower bound is 1, so lim sup⁡ksk=1; and the family of tail infima is {−1}, whose least upper bound is −1, so lim inf⁡ksk=−1.

step 2.1L3L4
4.1

Since −1≠1, there is no L with lim inf⁡ksk=lim sup⁡ksk=L, so by [L7] the sequence converges to no real number and diverges to neither +∞ nor −∞.

step 3.1L6L7∎

Remarks

ExampleConstruction: AI-adaptedVerification: AI-generatedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

The block sequence 1/1; 1/2,2/2; 1/3,2/3,3/3; … has subsequential limit set exactly [0,1]

Example

Write ι(n)=n⋅1R for the canonical natural. Let c:N→N×N be the function supplied by the recursion theorem (The recursion theorem) from the starting element (1,1) and the function

F(m,j):=(m, j+1)  if j<m,F(m,j):=(m+1, 1)  if j=m,

write ck=(mk,jk), and define the sequence of reals

xk  :=  ι(jk)ι(mk)(k∈N).

Its first terms are

11;12,22;13,23,33;14,…

the blocks being the successive values of m. Then the subsequential limit set (Subsequential limit of a real sequence, and the subsequential limit set) is the whole unit interval (Intervals of R: the nine order-convex forms, nondegeneracy, and length),

SL⁡(x)=[0,1],

and consequently lim sup⁡kxk=1 and lim inf⁡kxk=0 (Limit superior and limit inferior of a real sequence as inf⁡nsup⁡k≥nxk and sup⁡ninf⁡k≥nxk in R‾).

The recursion replaces the block bookkeeping. Presenting the sequence by the partial sums 1+2+⋯+(m−1) that mark where each block begins would require inverting that count at every index. Carrying the pair (m,j) along instead makes each term's block and position immediately available, and the three facts the argument needs, that 1≤jk≤mk, that mk≤k+1, and that every admissible pair occurs, are then three short inductions.

Facts & Assumptions

Given: The recursion ck=(mk,jk) described above, the sequence xk=ι(jk)/ι(mk), and a real number y with 0≤y≤1.

[L1]

Recursion theorem: for a set A, an element a∈A and f:A→A there is a unique g:N→A with g0=a and gk+1=f(gk) (The recursion theorem, The natural numbers N (von Neumann)).

[L2]
[L3]

Well-ordering principle: every nonempty subset of N has a least element (The well-ordering principle).

[L4]

Index maps: if ni<ni+1 for every i then n is strictly increasing, and then ni≥i; the composite (xni) is a subsequence, and L∈SL⁡(x) means some subsequence converges to L (A strictly increasing index map satisfies nk≥k, Sequences of reals: bounded, eventually, frequently, tails, subsequences, Subsequential limit of a real sequence, and the subsequential limit set, Limits and Cauchy sequences of reals).

[L5]

Canonical naturals: ι(n)>0 and invertible for n≥1, ι is strictly increasing, ι(1)=1, and ι(n+n′)=ι(n)+ι(n′) (Canonical naturals are positive and strictly increasing).

[L6]

Order arithmetic: claim 4 of Sign rules for products and monotonicity of multiplication, Order is preserved by adding a constant and by adding inequalities and Inverses of positives are positive, and reciprocation reverses order state the strict forms, that multiplication by a positive element preserves <, that inequalities may be translated and added, and that 0<u<v gives 0<1/v<1/u; adjoining the case of equality, where the two sides coincide, gives the nonstrict forms used below. Products of nonnegative inequalities multiply in the nonstrict form stated by Multiplying inequalities of positives, and the order is total (Ordered field, Complete ordered field (least-upper-bound property)).

[L7]

Reciprocal Archimedean property: for every real η>0 there is a natural p≥1 with 1/p<η (For every ε>0 in a complete ordered field there is a natural n≥1 with 1/n<ε, Every complete ordered field is Archimedean).

[L8]

Limits preserve non-strict inequalities (Limits preserve non-strict inequalities); a convergent sequence is bounded and a sequence diverging to ±∞ is unbounded (Every convergent sequence is bounded, Divergence to +∞ and to −∞).

[L9]

The interval [0,1]={t∈R:0≤t≤1}, with least element 0 and greatest element 1 (Intervals of R: the nine order-convex forms, nondegeneracy, and length).

[L11]

Absolute value: ∣t∣≤η if and only if −η≤t≤η; and the order on N is total with k<k+1 (Basic properties of the absolute value, Absolute value in an ordered field, Order on the natural numbers, ≤ is a linear order on N, Discreteness: σ(n) is the immediate successor).

Verification

technique · constructive
1.1

The recursion theorem applied to A=N×N, the element (1,1) and the function F gives a unique c with c0=(1,1) and ck+1=F(ck), so (mk,jk) and hence xk are well defined once mk≥1 is known.

givenL1
1.2

For every k one has 1≤jk≤mk: this holds at k=0, where (m0,j0)=(1,1); and if 1≤jk≤mk then either jk<mk, so that ck+1=(mk,jk+1) with 1≤jk+1≤mk, or jk=mk, so that ck+1=(mk+1,1) with 1≤1≤mk+1. The claim follows by induction on k; in particular mk≥1 always, so ι(mk)>0 and xk is defined.

givenL2L5L11
1.3

For every k one has mk≤k+1: at k=0, m0=1; and mk+1 is mk or mk+1, so mk+1≤mk+1≤k+2. This is again an induction on k.

givenL2L11
1.4

For every natural m≥1 and every j with 1≤j≤m there is k with (mk,jk)=(m,j). Induct on m. For m=1 the only admissible j is 1, realised at k=0. Assume the claim for m and apply it at j=m to get k with (mk,jk)=(m,m); then ck+1=(m+1,1), and a second induction on i shows ck+1+i=(m+1,1+i) for every i≤m: it holds at i=0, and if it holds at i<m then 1+i<m+1, so ck+2+i=(m+1,2+i). Hence every j with 1≤j≤m+1 is realised in block m+1.

givenL1L2L11
1.5

Let y be a real with 0≤y≤1.

givenL9
2.1

For every k one has 0<xk≤1: from 1≤jk≤mk we get 0<ι(jk)≤ι(mk), and dividing by ι(mk)>0 gives 0<xk≤1.

step 1.2L5L6
2.2

For every natural m≥1 the set { j∈N:1≤j≤m and y ι(m)≤ι(j) } is nonempty, since j=m belongs to it because y≤1 gives y ι(m)≤ι(m); let J(m) be its least element. Then y≤ι(J(m))/ι(m), and moreover ι(J(m))/ι(m)≤y+1/ι(m): if J(m)=1 this reads 1/ι(m)≤y+1/ι(m), true because y≥0; and if J(m)>1 then j:=J(m)−1 satisfies 1≤j≤m and j<J(m), so minimality gives ι(j)<y ι(m), that is ι(J(m))−1<y ι(m), and dividing by ι(m)>0 gives the bound. Hence 0≤ι(J(m))/ι(m)−y≤1/ι(m).

step 1.5L3L5L6L11
3.1

For every L∈SL⁡(x) one has 0≤L≤1, since a subsequence of (xk) converging to L satisfies 0≤xni≤1 at every index by step 2.1, and limits preserve non-strict inequalities. So SL⁡(x)⊆[0,1].

step 2.1L4L8L9
3.2

For every natural m≥1 the set { k∈N:(mk,jk)=(m,J(m)) } is nonempty by step 1.4, since 1≤J(m)≤m; let K(m) be its least element. Then xK(m)=ι(J(m))/ι(m), so ∣xK(m)−y∣≤1/ι(m) by step 2.2, and m=mK(m)≤K(m)+1 by step 1.3.

step 1.4step 1.3step 2.2L3L11
4.1

Define f:N→N by f(n):=K(n+2); then f(n)≥(n+2)−1=n+1>n and ∣xf(n)−y∣≤1/ι(n+2). The recursion theorem applied to N, the element K(1) and the function f gives n:N→N with n0=K(1) and ni+1=f(ni); it is strictly increasing, so ni≥i, and ∣xni+1−y∣≤1/ι(ni+2)≤1/ι(i+2).

step 3.2L1L4L5L6construct
5.1

The subsequence (xni) converges to y: given a real ε>0, take a natural p≥1 with 1/p<ε; every i≥p satisfies i≥1, so step 4.1 applied at i−1 gives ∣xni−y∣≤1/ι(i+1), and i+1>p gives 1/ι(i+1)<1/p<ε. Hence y∈SL⁡(x), and since y was an arbitrary element of [0,1], [0,1]⊆SL⁡(x).

step 4.1step 1.5L4L5L6L7L11
6.1

Therefore SL⁡(x)=[0,1]. The sequence is bounded by step 2.1, so every subsequence of it is bounded and none diverges to ±∞; hence SL⁡‾(x)=SL⁡(x)=[0,1], whose greatest element is 1 and least element 0, and [L10] gives lim sup⁡kxk=1 and lim inf⁡kxk=0.

step 5.1step 3.1step 2.1L8L9L10discharge-construct∎

Remarks

  • Every point of [0,1] is approached, and the rate is the block width. In block m the terms are 1/m,2/m,…,m/m, spaced 1/m apart and covering (0,1], so any target in [0,1] has a term of block m within 1/m of it. Since blocks of every width occur, and occur arbitrarily late, this produces a subsequence converging to the target. That is the whole idea; steps 2.2 and 3.2 only make the choice of term canonical, by taking a least element rather than an arbitrary one, so that no choice principle is used.

  • The set is closed, as it must be. [0,1] contains the limit of every convergent sequence of its own points, which is what If each yj is a subsequential limit of (xk) and yj→y∈R, then y is a subsequential limit of (xk) predicts for any subsequential limit set. This example shows the prediction is not vacuous: the set here is an entire interval, in contrast with the two-point set of (−1)k has lim inf⁡=−1 and lim sup⁡=1, so it does not converge.

  • Neither endpoint is a value of the sequence in the case of 0. Every term is >0 by step 2.1, so 0∈SL⁡(x) is a genuine limit and not an attained value, while 1 is attained, once in every block. Subsequential limits need not be values, and values need not be subsequential limits.

  • Why the sequence is not written by a closed formula. The classical presentation defines xk by first solving m(m−1)/2≤k<m(m+1)/2 for m, which needs a least-element argument at every index and a quadratic estimate to get m≤k+1. The recursion carries the block and position forward instead, and the estimate mk≤k+1 becomes the one-line induction of step 1.3.

ExampleConstruction: AI-adaptedVerification: AI-generatedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

ak=2−k+(−1)k has lim inf⁡ak+1/ak=1/8, lim sup⁡ak+1/ak=2 and lim⁡ak1/k=1/2

Example

Let (sk) be the alternating sequence of The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1, let tk:=2 when sk=1 and tk:=1/2 when sk=−1, and put

ak  :=  2−ktk(k∈N),

the sequence usually written ak=2−k+(−1)k. Writing qk:=ak+1/ak for the ratios and rk:=ak+11/(k+1) for the roots, which is an1/n reindexed by n=k+1 as For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak requires,

lim inf⁡kqk=18,lim sup⁡kqk=2,lim⁡krk=12,

so also lim inf⁡krk=lim sup⁡krk=1/2. In addition ak→0.

The point. The ratios oscillate across 1, taking the values 1/8 and 2 alternately, so no statement of the form "the ratios are eventually below some λ<1" is available; the roots, by contrast, converge to 1/2<1. Any criterion reading the ratios alone is silent here, and one reading the roots is not. That is the concrete form of the dominance recorded in For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak.

Facts & Assumptions

Given: The alternating sequence (sk), the auxiliary tk∈{2,1/2}, the sequence ak=2−ktk, the ratios qk=ak+1/ak and the roots rk=ak+11/(k+1), all as in FALSE: lim sup⁡ak1/k=lim sup⁡ak+1/ak for every positive sequence.

[L1]

For this sequence: every ak is positive, qk∈{1/8,2} with both values occurring at arbitrarily large indices, lim inf⁡kqk=1/8, lim sup⁡kqk=2, and rk→1/2, so lim inf⁡krk=lim sup⁡krk=1/2 (FALSE: lim sup⁡ak1/k=lim sup⁡ak+1/ak for every positive sequence).

[L2]

The chain lim inf⁡kqk≤lim inf⁡krk≤lim sup⁡krk≤lim sup⁡kqk (For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak).

[L4]

Powers: 2−k=(1/2)k and tk≤2, so ak≤2⋅(1/2)k; and ak>0 (Integer powers am, Laws of integer exponents, Rational powers ar of a positive base, Sign rules for products and monotonicity of multiplication).

Verification

technique · direct
1.1

The three values in the display are exactly what [L1] records for this sequence, together with lim inf⁡krk=lim sup⁡krk=1/2, which follows from the convergence of (rk) to 1/2.

givenL1L3
1.2

The sequence is null: 0<ak=2−ktk≤2⋅(1/2)k for every k, and (1/2)k→0 because ∣1/2∣<1, so 2⋅(1/2)k→0 and the squeeze gives ak→0.

givenL4L5L6
2.1

The ratio quantities differ from one another and from the root quantities: 1/8<1/2<2, so lim inf⁡kqk<lim inf⁡krk=lim sup⁡krk<lim sup⁡kqk. In particular the chain [L2] holds here with both outer inequalities strict and the middle one an equality, and the ratios do not determine the roots.

step 1.1L1L2L6
3.1

So (ak) is a positive null sequence whose root sequence converges to 1/2<1 while its ratio sequence has lim sup⁡kqk=2>1 and lim inf⁡kqk=1/8<1, that is, the ratios oscillate across 1 while the roots settle strictly below it.

step 1.2step 2.1L1L6∎

Remarks

  • Where the numbers come from. The exponent −k+(−1)k changes by −1+(−1)k+1−(−1)k=−1∓2 from one index to the next, giving ratios 2−3=1/8 and 21=2; the root divides the exponent by the index, so the bounded oscillation contributes 2±1/(k+1)→1 and only the linear part −k survives, giving 2−1=1/2. The full computation is in FALSE: lim sup⁡ak1/k=lim sup⁡ak+1/ak for every positive sequence.

  • The same sequence reappears for series. With these ak the series ∑kak converges, and the root criterion sees it while the ratio criterion does not. That use belongs to the series page and is not made here.

  • Strictness of the middle inequality needs a different witness. Here lim inf⁡krk=lim sup⁡krk, since the roots converge. A sequence making all three inequalities of the chain strict is A positive sequence making all three inequalities of the ratio-to-root chain strict.

ExampleConstruction: AI-adaptedVerification: AI-adaptedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

A positive sequence making all three inequalities of the ratio-to-root chain strict

Example

Let (sk) be the alternating sequence of The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1 and define

ak:=2−k  when sk=1,ak:=3−k  when sk=−1.

This interleaves the two geometric sequences 2−k and 3−k, taking the first at even indices and the second at odd ones. With qk:=ak+1/ak and rk:=ak+11/(k+1) as in For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak,

lim inf⁡kqk=0,lim inf⁡krk=13,lim sup⁡krk=12,lim sup⁡kqk=+∞,

so the chain of that theorem reads

0  <  13  <  12  <  +∞

with all three inequalities strict.

Where each comparison lives. The first two, 0<1/3 and 1/3<1/2, are comparisons of real numbers and hold in R; they hold in R‾ as well only because the extended order restricts on R to the order of R (The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined). The third, 1/2<+∞, is not a comparison in R at all: +∞ is not a real number, and the inequality is the instance of "every real is below the greatest element" in R‾. So the outer two values of the chain are of different kinds here, and only the extended line can hold all four at once.

Facts & Assumptions

Given: The alternating sequence (sk) with index maps e,o (The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1); the sequence ak defined above; the ratios qk=ak+1/ak; and the roots rk=ak+11/(k+1).

[L1]

The alternating sequence: ∣sk∣=1, sk+1=−sk, sej=1, soj=−1, with e, o strictly increasing, so ej≥j and oj≥j; also o0=σ(0)≥1 (The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1, A strictly increasing index map satisfies nk≥k).

[L3]

The order on R‾ is total, +∞ is greatest, every real is <+∞, and the order restricts on R to the order of R (The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined).

[L4]

Powers: 2−k=(1/2)k and 3−k=(1/3)k; xmxm′=xm+m′ and (xy)m=xmym for integer exponents and nonzero bases; xm>0 for x>0; (x−n)1/n=x−1 for x>0 and n≥1 (Integer powers am, Laws of integer exponents, Rational powers ar of a positive base, Laws of rational exponents, Existence and uniqueness of n-th roots: a unique a1/n≥0 with (a1/n)n=a).

[L5]

Geometric sequences: ∣ρ∣<1 implies ρk→0, and ∣ρ∣>1 implies ∣ρ∣k→+∞ (For ∣r∣<1 the sequence rk is null, and for ∣r∣>1 the sequence ∣r∣k diverges to +∞, Limits and Cauchy sequences of reals, Divergence to +∞ and to −∞).

[L7]

The order on N is total, so any two indices have a common upper bound (Order on the natural numbers, ≤ is a linear order on N).

[L8]

The chain lim inf⁡kqk≤lim inf⁡krk≤lim sup⁡krk≤lim sup⁡kqk (For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak).

Verification

technique · direct
1.1

Each sk is 1 or −1, so ak is well defined, and ak>0 for every k since positive powers of positive bases are positive.

givenL1L4L6
1.2

For every n∈N there are indices k,k′≥n with sk=1 and sk′=−1, namely k=en and k′=on; and there are indices l,l′≥n with sl+1=1 and sl′+1=−1, namely l=ej−1 and l′=oj−1 for any j≥n+1, these being natural numbers because ej≥j≥1 and oj≥j≥1, and satisfying l≥j−1≥n and l′≥j−1≥n.

givenL1L7
1.3

Since sk+1=−sk, the ratios are qk=3−(k+1)/2−k=3−1(2/3)k when sk=1, and qk=2−(k+1)/3−k=2−1(3/2)k when sk=−1; in both cases qk>0.

givenL1L4L6
1.4

Likewise the roots are rk=(2−(k+1))1/(k+1)=2−1 when sk+1=1, and rk=(3−(k+1))1/(k+1)=3−1 when sk+1=−1.

givenL1L4
2.1

By steps 1.2 and 1.4 the tail range of (rk) at every index n is exactly {1/2,1/3}, whose least upper bound is 1/2 and greatest lower bound 1/3, since 1/3<1/2 and both belong to the set. Hence lim sup⁡krk=1/2 and lim inf⁡krk=1/3.

step 1.2step 1.4L2L3L6
2.2

lim sup⁡kqk=+∞. Fix n and a real M. Since ∣3/2∣>1, the sequence (3/2)k diverges to +∞, so there is K with 2−1(3/2)k>M for all k≥K; taking j at least as large as both n and K and putting k:=oj, we get k≥j≥n and sk=−1, hence qk=2−1(3/2)k>M. So no real bounds the tail range of (qk) above, its least upper bound in R‾ is +∞ for every n, and lim sup⁡kqk is the greatest lower bound of {+∞}, namely +∞.

step 1.2step 1.3L2L3L5L6L7
2.3

lim inf⁡kqk=0. Fix n. All qk are positive, so 0 is a lower bound of the tail range. If ℓ>0 were a lower bound, then, since ∣2/3∣<1 gives (2/3)k→0 and hence 3−1(2/3)k<ℓ for all k≥K for some K, taking j at least as large as both n and K and putting k:=ej would give k≥j≥n, sk=1 and qk=3−1(2/3)k<ℓ, contradicting that ℓ is a lower bound. So every lower bound is ≤0 and the greatest lower bound of each tail range is 0; hence lim inf⁡kqk is the least upper bound of {0}, namely 0.

step 1.2step 1.3L2L3L5L6L7
3.1

Collecting the four values, the chain [L8] reads 0≤1/3≤1/2≤+∞, and each inequality is strict: 0<1/3 and 1/3<1/2 hold in R and therefore in R‾, while 1/2<+∞ holds because +∞ is the greatest element of R‾ and is distinct from every real. So no two of the four quantities coincide.

step 2.1step 2.2step 2.3L3L6L8∎

Remarks

ExampleConstruction: AI-generatedVerification: AI-generatedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

The four standard limits n1/n→1, a1/n→1, nα/(1+p)n→0 and xk/k!→0, computed

Example

The four standard limits of this page, written as sequences on N and instantiated. Throughout ι(n)=n⋅1R is the canonical natural, with ι(0)=0.

classical formas a sequence on Nvaluesource
n1/n→1(k+1)1/(k+1)1n1/n→1
a1/n→1, a>0a1/(k+1)1For every a>0, a1/n→1
nα/(1+p)n→0ι(k)α/(1+p)k0For every p>0 and every positive rational α, nα/(1+p)n→0
xk/k!→0xk/k!0For every real x, xk/k!→0

Two of the four need an index shift and two do not, and the reason is visible in the classical forms: in the first two the index sits in the exponent as 1/n, which is not a rational number at n=0, so those families begin at n=1 and are written here with n=k+1. In the last two the index sits in the base or in a factorial, both of which are defined at 0, so no shift is needed and the sequences begin at k=0 with the values 0 and 1 respectively.

The instances computed below are:

21/(k+1)→1,ι(k)22k→0,2kk!→0,(−3)kk!→0,ι(k)2k!→0.

The last of these is not one of the four; it is the composite that orders the three scales, and it is obtained from two of them by the product rule.

Facts & Assumptions

Given: The canonical naturals ι(n)=n⋅1R with ι(0)=0; the factorial k!=∏j<kι(j+1) of For every real x, xk/k!→0; rational powers (Rational powers ar of a positive base) and integer powers (Integer powers am).

[L1]

(k+1)1/(k+1)→1, with 1≤n1/n≤1+2/n1/2 for every natural n≥1 (n1/n→1).

[L2]

For every real a>0, a1/(k+1)→1; and for real b≥1 and natural n≥1, 1≤b1/n≤1+(b−1)/ι(n) (For every a>0, a1/n→1).

[L3]

For every real p>0 and rational α>0, ι(k)α/(1+p)k→0 (For every p>0 and every positive rational α, nα/(1+p)n→0).

[L4]

For every real x, xk/k!→0 (For every real x, xk/k!→0).

Verification

technique · direct
1.1

The first standard limit gives (k+1)1/(k+1)→1, together with the explicit two-sided bound 1≤(k+1)1/(k+1)≤1+2/(k+1)1/2 valid at every k∈N, since k+1≥1.

givenL1
1.2

The second gives 21/(k+1)→1 and (1/2)1/(k+1)→1, both bases being positive; for the first of these the explicit bound of [L2] with b=2 reads 1≤21/n≤1+1/ι(n) for every natural n≥1.

givenL2L6
1.3

The third, with p=1 and α=2, gives ι(k)2/2k→0; with p=1/2 and α=1/2 it gives ι(k)1/2/(3/2)k→0. Both p are positive reals and both α are positive rationals, so [L3] applies in each case.

givenL3L6
1.4

The fourth, with x=2 and with x=−3, gives 2k/k!→0 and (−3)k/k!→0; no hypothesis on x is needed, in particular no positivity.

givenL4
2.1

Multiplying the first limit of step 1.3 by the first of step 1.4 and using 2k≠0 gives ι(k)2/k!=(ι(k)2/2k)(2k/k!)→0⋅0=0.

step 1.3step 1.4L5L6
2.2

Multiplying the limit of step 1.1 by the first of step 1.2 gives (k+1)1/(k+1) 21/(k+1)→1⋅1=1.

step 1.1step 1.2L5
3.1

The four limits and the two composites are therefore established as displayed, and together they order the three growth scales: a fixed power of n is beaten by every geometric sequence of ratio >1 by step 1.3, every geometric sequence is beaten by the factorial by step 1.4, and consequently a fixed power of n is beaten by the factorial by step 2.1.

step 1.1step 1.2step 1.3step 1.4step 2.1step 2.2∎

Remarks

  • The two bounds quoted in steps 1.1 and 1.2 are the useful part in practice. They convert the qualitative statement into a rate: n1/n is within 2/n1/2 of 1, and a1/n within (a−1)/n of 1 when a≥1. The second rate is faster, and the difference is real: in n1/n the base itself grows with the index.

  • Why α is rational and p is real. The exponent α must be rational because rational powers are all this library has; the base 1+p may be any real >1 because it is raised only to integer powers. The asymmetry is a fact about what has been constructed, not about the mathematics, and it disappears once real exponents are available.

  • The composite in step 2.1 is the one usually quoted as "factorials beat polynomials". It is not proved directly anywhere on this page: it is the product of two of the four standard limits, and the product rule (Algebra of limits: sums, scalar multiples, products and quotients) is what assembles it.

  • Nothing here uses lim sup⁡. All four are ordinary limits, and the page's machinery is needed only to prove them, not to state them; the connection to the rest of the page is that For ak>0: lim inf⁡ak+1/ak≤lim inf⁡ak1/k≤lim sup⁡ak1/k≤lim sup⁡ak+1/ak is the tool that makes several of them routine once one of them is known.

CounterexampleConstruction: AI-adaptedVerification: AI-adaptedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

xk=(−1)k, yk=(−1)k+1 give lim sup⁡(xk+yk)=0<2=lim sup⁡xk+lim sup⁡yk

Statement refuted

That the subadditivity of lim sup⁡(xk+yk)≤lim sup⁡xk+lim sup⁡yk whenever the right-hand side is defined in R‾, and dually for lim inf⁡ can be improved to an equality: that for all sequences (xk), (yk) of reals whose limit superiors have a defined sum, lim sup⁡k(xk+yk)=lim sup⁡kxk+lim sup⁡kyk. The claim is recorded and refuted as FALSE: lim sup⁡(xk+yk)=lim sup⁡xk+lim sup⁡yk; this item is the named witness.

Facts & Assumptions

Given: The alternating sequence (sk) of The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1; the sequences xk:=sk and yk:=−sk, which are the families usually written (−1)k and (−1)k+1.

[L1]

For this pair of sequences: xk+yk=0 for every k, lim sup⁡kxk=1, lim sup⁡kyk=1 and lim sup⁡k(xk+yk)=0 (FALSE: lim sup⁡(xk+yk)=lim sup⁡xk+lim sup⁡yk).

[L2]

Subadditivity: lim sup⁡k(xk+yk)≤lim sup⁡kxk+lim sup⁡kyk whenever the right-hand side is defined (lim sup⁡(xk+yk)≤lim sup⁡xk+lim sup⁡yk whenever the right-hand side is defined in R‾, and dually for lim inf⁡).

Counterexample

technique · direct
1.1

The two sequences xk=sk and yk=−sk are sequences of reals whose termwise sum is constantly 0, and their limit superiors are both the real number 1.

givenL1L3
1.2

Both limit superiors being real, their sum is defined in R‾ and equals the field sum 1+1=2.

givenL1L3L4
2.1

The limit superior of the sum sequence is 0, while the sum of the limit superiors is 2, and 0<2; so the inequality of [L2] holds here and is strict, and the equality asserted by the refuted claim fails.

step 1.1step 1.2L1L2L4∎

Remarks

CounterexampleConstruction: AI-generatedVerification: AI-generatedprecheck passjudge pass (z-ai/glm-5.2)audited 2026-07-26Open item page →

xk=1+(−1)k, yk=1+(−1)k+1 give lim sup⁡(xkyk)=0<4

Statement refuted

That the submultiplicativity of For bounded nonnegative sequences, lim sup⁡(xkyk)≤(lim sup⁡xk)(lim sup⁡yk) can be improved to an equality: that for all bounded nonnegative sequences (xk), (yk) of reals, lim sup⁡k(xkyk)=(lim sup⁡kxk)(lim sup⁡kyk).

Facts & Assumptions

[L3]

Absolute value: ∣t∣=1 forces t=1 or t=−1 (Basic properties of the absolute value, Absolute value in an ordered field).

[L4]

Order and field arithmetic: 0<1, so 0<2=1+1 and 0<4=2⋅2; 1+1=2, 1−1=0, 1+(−1)=0 and 1−(−1)=2; a product with a zero factor is 0 (The multiplicative identity is positive, Order is preserved by adding a constant and by adding inequalities, Sign rules for products and monotonicity of multiplication, Ordered field, Complete ordered field (least-upper-bound property)).

[L5]

Submultiplicativity: for bounded nonnegative sequences, lim sup⁡k(xkyk)≤(lim sup⁡kxk)(lim sup⁡kyk), all three quantities being real (For bounded nonnegative sequences, lim sup⁡(xkyk)≤(lim sup⁡xk)(lim sup⁡yk)).

Counterexample

technique · direct
1.1

Each sk is 1 or −1. When sk=1 the pair (xk,yk) is (2,0), and when sk=−1 it is (0,2). In either case 0≤xk≤2 and 0≤yk≤2, so both sequences are bounded and nonnegative, and xkyk=0 because one of the two factors is 0.

givenL1L3L4
1.2

For every n both cases occur at an index ≥n: sen=1 with en≥n and son=−1 with on≥n.

givenL1
2.1

Hence Tn(x)={0,2} and Tn(y)={0,2} for every n, each with least upper bound 2 in R‾, since 2 bounds both elements and belongs to the set; so lim sup⁡kxk=lim sup⁡kyk=2. The product sequence is constantly 0, so Tn(xy)={0} and lim sup⁡k(xkyk)=0.

step 1.1step 1.2L2L4
3.1

The hypotheses of [L5] are met by step 1.1, and the inequality it gives reads 0≤2⋅2=4. Since 0<4, it is strict, so the equality asserted above fails for this pair and the refuted claim is false.

step 2.1step 1.1L4L5∎

Remarks

CounterexampleConstruction: AI-generatedVerification: AI-generatedprecheck passverified 2026-08-08 (gpt-5.6-terra-codex-subscription)Open item page →

A sequence with lim sup⁡=+∞: the greatest subsequential limit exists only in R‾

Statement refuted

That The limit superior is itself a subsequential limit in R‾ and is the greatest one can be stated inside R: that for every sequence (xk) of reals the set SL⁡(x) of real subsequential limits (Subsequential limit of a real sequence, and the subsequential limit set) has a greatest element and that element is lim sup⁡kxk.

The witness below has a nonempty SL⁡(x) with a greatest element, so the failure is not that the real set is empty: it is that the greatest element of SL⁡(x) is 0 while lim sup⁡kxk=+∞. The dominant behaviour of the sequence is invisible to SL⁡(x) and is recorded only by SL⁡‾(x) (Convergence in R‾ and the extended subsequential limit set: L∈R‾ is an extended subsequential limit when some subsequence converges to L, or diverges to L=±∞).

Facts & Assumptions

Given: The alternating sequence (sk) and the index maps e,o of The even and odd index maps and the alternating sequence: strictly increasing e,o with N their disjoint union, and the unique (sk) with s0=1, sσ(k)=−sk, which satisfies ∣sk∣=1, s∘e≡1 and s∘o≡−1; the canonical naturals ι(k)=k⋅1R with ι(0)=0; and the sequence xk:=ι(k) when sk=1 and xk:=0 when sk=−1.

[L3]

The order on R‾ is total, +∞ is greatest, every real is <+∞ and >−∞, and the order restricts on R to the order of R (The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined).

[L5]

Canonical naturals: ι is strictly increasing with ι(k)≥0, and for every real M there is a natural p≥1 with M<ι(p) (Canonical naturals are positive and strictly increasing, Every complete ordered field is Archimedean).

[L6]

A convergent sequence of reals is bounded, a limit is unique, and a sequence agreeing with a constant from some index on converges to that constant (Every convergent sequence is bounded, A sequence has at most one limit, Convergence depends only on the tail, Sequences of reals: bounded, eventually, frequently, tails, subsequences).

Counterexample

technique · direct
1.1

Each sk is 1 or −1, so (xk) is a well-defined sequence of reals with xk≥0 for every k; moreover xej=ι(ej) and xoj=0 for every j.

givenL1L5L7
1.2

The subsequence along o is constantly 0, and o is strictly increasing, so 0∈SL⁡(x).

givenL1L4L6
2.1

For every n∈N the tail supremum sup⁡Tn(x) is +∞. Given a real M, take a natural p≥1 with M<ι(p) and an index j at least as large as both n and p; then ej≥j≥n, so xej=ι(ej)∈Tn(x), and ej≥j≥p gives ι(ej)≥ι(p)>M. So no real number bounds Tn(x) above, and the least upper bound in R‾ must be +∞.

step 1.1L1L2L3L5L7
3.1

Every real subsequential limit of (xk) equals 0. Let n be strictly increasing with xni→L∈R; the subsequence is then bounded, say ∣xni∣≤B for every i. Suppose sni=1 for arbitrarily large i: taking a natural p≥1 with B<ι(p) and such an index i≥p, we get xni=ι(ni)≥ι(i)≥ι(p)>B, contradicting the bound. So there is I with sni=−1, hence xni=0, for every i≥I; a sequence equal to 0 from an index on converges to 0, so L=0 by uniqueness of limits.

step 1.1step 2.1L1L4L5L6L7
4.1

Consequently lim sup⁡kxk is the greatest lower bound of the family {+∞}, namely +∞, while SL⁡(x)={0} by steps 1.2 and 3.1, whose greatest element is the real number 0. Since 0≠+∞, the refuted claim fails for this sequence.

step 2.1step 1.2step 3.1L2L3∎

Remarks

CounterexampleConstruction: AI-generatedVerification: AI-generatedprecheck passverified 2026-08-08 (gpt-5.6-terra-codex-subscription)Open item page →

Null times divergent has no rule: xk=1/k with yk=ck gives product limit c, and with yk=k2 gives divergence

Statement refuted

That the product 0⋅(+∞), left undefined by The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined, could be given a value compatible with limits: that there is v∈R‾ such that for all sequences of reals with xk→0 (Limits and Cauchy sequences of reals) and yk→+∞ (Divergence to +∞ and to −∞) the products xkyk have the single limiting behaviour named by v.

Equivalently: that knowing a factor is null and the other diverges to +∞ determines anything at all about the product. It does not, and the two undefined entries in the arithmetic of R‾ are undefined for exactly this reason.

Facts & Assumptions

Given: The canonical naturals ι(n)=n⋅1R; the sequence xk:=1/ι(k+1); for a real c>0 the sequence yk(c):=c ι(k+1); and the sequence zk:=ι(k+1) ι(k+1).

[L1]

Canonical naturals: ι(n)>0 and invertible for n≥1, ι is strictly increasing, and ι(n)≥1 for n≥1 (Canonical naturals are positive and strictly increasing, Order on the natural numbers, ≤ is a linear order on N).

[L2]

Archimedean facts: for every real η>0 there is a natural p≥1 with 1/p<η, and for every real M there is a natural p≥1 with M<ι(p); and 0<u<v gives 0<1/v<1/u (For every ε>0 in a complete ordered field there is a natural n≥1 with 1/n<ε, Every complete ordered field is Archimedean, Inverses of positives are positive, and reciprocation reverses order).

[L3]

Convergence to a real and divergence to +∞; to establish convergence it suffices to produce a threshold for every real ε>0; a constant sequence converges to its value (Limits and Cauchy sequences of reals, Sequences of reals: bounded, eventually, frequently, tails, subsequences, Divergence to +∞ and to −∞).

[L4]

A sequence diverging to +∞ is unbounded and therefore does not converge to any real (Divergence to +∞ and to −∞, Every convergent sequence is bounded); a limit, when it exists, is unique (A sequence has at most one limit).

Counterexample

technique · direct
1.1

The sequence xk=1/ι(k+1) is well defined, positive, and converges to 0: given a real ε>0, take a natural p≥1 with 1/p<ε; for k≥p we have ι(k+1)>ι(p)>0, hence 0<xk<1/p<ε.

givenL1L2L3
1.2

For every real c>0 the sequence yk(c)=c ι(k+1) diverges to +∞: given a real M, the quotient M/c is real, so there is a natural p≥1 with M/c<ι(p), and for k≥p we get ι(k+1)>ι(p)>M/c, hence yk(c)=c ι(k+1)>M after multiplying by c>0.

givenL1L2L3L5
1.3

The sequence zk=ι(k+1)ι(k+1) diverges to +∞: given a real M, take a natural p≥1 with M<ι(p); for k≥p we have ι(k+1)≥1 and ι(k+1)>ι(p)>M, so zk≥ι(k+1)>M.

givenL1L2L3L5
2.1

For every real c>0 the product sequence is constant: xkyk(c)=(1/ι(k+1)) c ι(k+1)=c for every k, so it converges to c.

step 1.1step 1.2L3L5
2.2

The product with (zk) is xkzk=(1/ι(k+1))ι(k+1)ι(k+1)=ι(k+1), which diverges to +∞ by the argument of step 1.3 with the single factor, and therefore converges to no real number.

step 1.1step 1.3L1L2L3L4L5
3.1

Now take the three pairs (x,y(1)), (x,y(2)) and (x,z). In each, the first sequence is null and the second diverges to +∞, so each pair satisfies the hypotheses of the refuted claim; but the three products converge to 1, converge to 2, and converge to +∞ in the extended sense. Since 1≠2 and limits are unique, no single v∈R‾ describes all three, and the claim is false.

step 2.1step 2.2L4L5L6∎

Remarks

Sources