Escape from a smaller interval
Write 𝜃𝑚 ={𝑚! 𝑆}. The
next criterion compares a fractional part with the actual tail: 𝜃𝑚−1 must leave [0,𝐸𝑚/𝑚) at arbitrarily large indices.
The interval has width less than 2/𝑚2. Since 𝐸𝑚 contains 𝑆, this equivalence is not yet a finite
test; the second assertion supplies a sufficient test using the finite
quantities Δ𝑚 and 𝜀𝑚.
Proposition 4.3 (lower-interval criterion).
𝑆∉ℚ⟺∀𝐵 ∃𝑚>𝐵: 𝐸𝑚≤𝑚𝜃𝑚−1.(20)
For 𝑚 ≥3
the finite condition
𝑚Δ𝑚≤1+𝜀𝑚or1+𝜀𝑚+2𝑚≤𝑚Δ𝑚(21)
implies the escape inequality in (20). Cofinally
many instances of (21) therefore
imply 𝑆 ∉ℚ.
Proof. If 𝑆 is
rational, the remainders 𝜃𝑚−1 vanish eventually while
𝐸𝑚 >0, so the escape inequality
fails eventually. Conversely, suppose 𝑚𝜃𝑚−1 <𝐸𝑚 for every
sufficiently large 𝑚. By (13) this gives
𝑚𝜃𝑚−1 <1, so 𝑎𝑚(𝑆) =0 and 𝜃𝑚 =𝑚𝜃𝑚−1. Fix 𝑁 after this recurrence begins. Then
𝜃𝑁+𝑘 =((𝑁 +𝑘)!/𝑁!)𝜃𝑁 <1 for
every 𝑘 ≥0, forcing 𝜃𝑁 =0. Hence 𝑁!𝑆 is an integer and 𝑆 is rational.
For the finite implication, suppose 𝑚𝜃𝑚−1 <𝐸𝑚. The tail recurrence
is 𝑚𝐸𝑚−1 =1 +𝜀𝑚 +𝐸𝑚, so
𝐸𝑚−1 >𝐸𝑚/𝑚 >𝜃𝑚−1 ≥0, and
with 0 <𝐸𝑚−1 <1 the
strict-successor identity gives Δ𝑚 =𝐸𝑚−1 −𝜃𝑚−1. Hence
1+𝜀𝑚<𝑚Δ𝑚=1+𝜀𝑚+𝐸𝑚−𝑚𝜃𝑚−1≤1+𝜀𝑚+𝐸𝑚<1+𝜀𝑚+2𝑚,
which
contradicts (21). ◻
There are two comparisons to keep separate: a finite sufficient test
for escape at one index, and an equivalence when indices are unbounded.
The finite test excludes only the open interval 1 +𝜀𝑚 <𝑚Δ𝑚 <1 +𝜀𝑚 +2/𝑚;
it accepts either endpoint and all values outside. The width 2/𝑚 refers to 𝑚Δ𝑚; the interval for Δ𝑚 has width 2/𝑚2. At one index this is a sufficient
test for escape, obtained by replacing 𝐸𝑚 with the upper bound 2/𝑚.
At arbitrarily large integer indices, however, the finite test itself
characterises irrationality:
𝑆∉ℚ⟺(21) holds for arbitrarily large integers 𝑚.
Only the forward implication remains to be shown.
If 𝑆 is irrational, Theorem 4.1 gives
𝑏𝑚 ≠1 at arbitrarily large
indices. By (14), each such
index satisfies 𝑚Δ𝑚 ≤1 +𝜀𝑚 or 𝑚Δ𝑚 >2 +𝜀𝑚. Since 2/𝑚 <1 for 𝑚 ≥3, either alternative implies (21). The
converse is Proposition 4.3. This
combines the two criteria; it does not prove either unbounded occurrence
for 𝑆. The equivalence is over all
integer indices. A restriction to primes gives only a sufficient
condition here.
For comparison at a single index, the exact classification is
𝐸𝑚≤𝑚𝜃𝑚−1⟺𝑏𝑚≠1 or 𝑎𝑚(𝑆)=𝑚−1(𝑚≥3).(22)
To verify the classification, use 0 <𝐸𝑗 <1 to write
𝑍𝑗=⌊𝑗!𝑆⌋+𝟏{𝜃𝑗≥𝐸𝑗}.
Substitution in the carry recurrence gives
𝑏𝑚=1−𝑎𝑚(𝑆)+𝑚𝟏{𝜃𝑚−1≥𝐸𝑚−1}−𝟏{𝜃𝑚≥𝐸𝑚}.
Since 0 ≤𝑎𝑚(𝑆) ≤𝑚 −1, a unit carry has
exactly two possibilities: both indicators are zero and 𝑎𝑚(𝑆) =0, or both are one and 𝑎𝑚(𝑆) =𝑚 −1. In the first case 𝑚𝜃𝑚−1 =𝜃𝑚 <𝐸𝑚; in the
second, 𝑚𝜃𝑚−1 ≥𝑚 −1 >𝐸𝑚. Conversely, 𝑚𝜃𝑚−1 <𝐸𝑚 <1 forces 𝑎𝑚(𝑆) =0 and both indicators to be zero,
using 𝑚𝐸𝑚−1 =1 +𝜀𝑚 +𝐸𝑚.
It therefore forces a unit carry with zero digit. This proves (22),
including the equality case in its left-hand inequality.
Thus escape can occur even when 𝑏𝑚 =1, provided the digit is 𝑚 −1. This happens at 𝑚 =52: exact rational arithmetic gives
𝑏52 =1 together with 52 Δ52 >1 +𝜀52 +2/52,
the margin being about 0.5689674908. This example separates the
two events at a single index: the finite test can hold at a unit carry.
Their occurrence at arbitrarily large indices nevertheless gives
equivalent rationality criteria. Since 52 is composite, the example supplies no
prime-index instance.