The Collatz Conjecture

{ }
For all 𝑎𝑎0 ∈ 1,3,5,7, … of the Collatz conjecture sequence 𝑎𝑎0, , 𝑎𝑎1 , 𝑎𝑎2 , … 𝑎𝑎𝑥𝑥 , there exist 𝑎𝑎𝑥𝑥 : 𝑎𝑎0 > 𝑎𝑎𝑥𝑥
By Mohlomi Cliff Makhetha
cliffmohlomilouw@gmail.com

Abstract

An investigation of the Collatz conjecture sequence 𝑎𝑎0, , 𝑎𝑎1 , 𝑎𝑎2 , … 𝑎𝑎𝑥𝑥 where 𝑎𝑎0 is an element of the set 𝑆𝑆, and S is
a family of subsets that belong to the partition of the set {1,3,5,7, … }. The investigation finds that there exist a family
set T of subsets that belong to 𝑆𝑆 where the sequence 𝑎𝑎0, , 𝑎𝑎1 , 𝑎𝑎2 , … 𝑎𝑎𝑥𝑥 satisfies the inequality 𝑎𝑎0 > 𝑎𝑎𝑥𝑥 . With the use the
monotonic lattice paths the investigation concludes that:
∀𝑎𝑎0 ∈ {1,3,5,7, … } of the Collatz conjecture sequence 𝑎𝑎0, , 𝑎𝑎1 , 𝑎𝑎2 , … 𝑎𝑎𝑥𝑥 , ∃𝑎𝑎𝑥𝑥 : 𝑎𝑎0 > 𝑎𝑎𝑥𝑥 … 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 1

Introduction
1937 Mr. L. Collatz posed a conjecture also known as the mapping problem, Hasse’s algorithm, or Katutani’s
problem which states. Let 𝑎𝑎0 be a positive integer if iterating:
𝑔𝑔 1
𝑎𝑎𝑥𝑥−1 → 𝑎𝑎𝑥𝑥−1 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑥𝑥−1 𝑖𝑖𝑖𝑖 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
𝑎𝑎𝑥𝑥 = � 2
𝑓𝑓
𝑎𝑎𝑥𝑥−1 → 3𝑎𝑎𝑥𝑥−1 + 1 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑥𝑥−1 𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜𝑜𝑜
Then 𝑎𝑎0 will always iterate 𝑎𝑎𝑥𝑥 : 𝑎𝑎𝑥𝑥 = 1. The members of the Collatz conjecture sequence are also known as the
hailstone numbers. [1,6] This paper attempt to demonstrate a logical and mathematical approach to prove
𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 1. The mathematical approach will lend help from the branches of set theory, number theory, graph
theory, and combinatory mathematics.
1) S is a family of subsets that belong to the partition of the set {1,3,5,7, … }
Let ℕ = {1,2,3,4, … }, ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = {1,3,5,7, … }, 𝑎𝑎𝑎𝑎𝑎𝑎 ℕ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 = {2,4,6,8, … } We start by representing the set ℕ𝑜𝑜𝑜𝑜𝑜𝑜 as a
union of two disjoint subsets:
ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = {4𝑛𝑛 − 1: 𝑛𝑛 ∈ ℕ } ∪ {4𝑛𝑛 − 3: 𝑛𝑛 ∈ ℕ } … 1
Next, we represent each of the two disjoint subsets from … 1 by two new disjoint subsets:
ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = {8𝑛𝑛 − 1: 𝑛𝑛 ∈ ℕ } ∪ {8𝑛𝑛 − 3: 𝑛𝑛 ∈ ℕ } ∪ {8𝑛𝑛 − 5: 𝑛𝑛 ∈ ℕ } ∪ {8𝑛𝑛 − 7: 𝑛𝑛 ∈ ℕ } … 2
Again, we can represent each of the four disjoint subsets from … 2 by two new disjoint subsets:
ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = {16𝑛𝑛 − 1: 𝑛𝑛 ∈ ℕ } ∪ {16𝑛𝑛 − 3: 𝑛𝑛 ∈ ℕ } ∪ {16𝑛𝑛 − 5: 𝑛𝑛 ∈ ℕ } ∪ … ∪ {16𝑛𝑛 − 15: 𝑛𝑛 ∈ ℕ } … 3
If we repeat the process of representing each of the disjoint subsets from … 3 by two new disjoint subsets indefinitely,
we can formulate the first identity by induction where 𝑢𝑢 ∈ ℕ.
𝑣𝑣=2𝑢𝑢−1

ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = � {2𝑢𝑢 𝑛𝑛 − 2𝑣𝑣 + 1: 𝑛𝑛 ∈ ℕ } … 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 1

𝑣𝑣=1

Since each of the subsets in … 𝑖𝑖𝑖𝑖𝑖𝑖. 1 is a disjoint subset, we can formulate the second identity:
𝑣𝑣=2𝑢𝑢−1

∅ = � {2𝑢𝑢 𝑛𝑛 − 2𝑣𝑣 + 1: 𝑛𝑛 ∈ ℕ } … 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 2

𝑣𝑣=1
Note 1: 𝑆𝑆(𝑢𝑢) = {2𝑢𝑢 𝑛𝑛 − 2𝑣𝑣 + 1: 𝑛𝑛 ∈ ℕ } where 𝑆𝑆(𝑢𝑢) is a family of subsets that belong to the partition of the set ℕ𝑜𝑜𝑜𝑜𝑜𝑜
and by implication the cardinality of 𝑆𝑆(𝑢𝑢) = 2𝑢𝑢−1 .

2) The mapping problem

For this paper let the elements of a 2 × 1 matrix be the initial element 𝐼𝐼 (𝑛𝑛) ∈ 𝑆𝑆(𝑢𝑢) and the resultant element 𝑅𝑅 (𝑛𝑛),
and we use three operations to model a mapping ∀𝑆𝑆(𝑢𝑢) in a form of a tree diagram in figure 1 [8]:
1 0
𝐼𝐼(𝑛𝑛) 𝑓𝑓 1 𝐼𝐼(𝑛𝑛)
� � →� ( ) ( ) ( )
𝑅𝑅(𝑛𝑛) 3� ∙ �𝑅𝑅(𝑛𝑛)�⇔ 𝑅𝑅 𝑛𝑛 ⊂ ℕ𝑜𝑜𝑜𝑜𝑜𝑜 𝑎𝑎𝑎𝑎𝑎𝑎 𝐼𝐼 𝑛𝑛 ≤ 𝑅𝑅 𝑛𝑛 … 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 1
𝐼𝐼(𝑛𝑛)
𝐼𝐼(𝑛𝑛) 𝑔𝑔 1 0 𝐼𝐼(𝑛𝑛)
� � → �0 1� ∙� � ⇔ 𝑅𝑅 (𝑛𝑛) ⊂ ℕ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑎𝑎𝑎𝑎𝑎𝑎 𝐼𝐼 (𝑛𝑛) < 𝑅𝑅(𝑛𝑛) … 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 2
𝑅𝑅(𝑛𝑛) 2
𝑅𝑅(𝑛𝑛)

⇔ 𝐼𝐼 (𝑛𝑛) > 𝑅𝑅(𝑛𝑛), 𝑜𝑜𝑜𝑜 𝑅𝑅(𝑛𝑛) ⊂ (ℕ𝑜𝑜𝑜𝑜𝑜𝑜 ∪ ℕ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ) … 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 3

𝐼𝐼(𝑛𝑛)
Note 2: Given any matrix � � from figure 1 such that 𝐼𝐼(𝑛𝑛) ≠ 𝑅𝑅(𝑛𝑛), if we let 𝑎𝑎0 = 𝐼𝐼(𝑛𝑛) and iterate:
𝑅𝑅(𝑛𝑛)
𝑔𝑔 1
𝑎𝑎𝑥𝑥−1 → 𝑎𝑎𝑥𝑥−1 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑥𝑥−1 ⊂ ℕ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
𝑎𝑎𝑥𝑥 = � 2
𝑓𝑓
𝑎𝑎𝑥𝑥−1 → 3𝑎𝑎𝑥𝑥−1 + 1 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑥𝑥−1 ⊂ ℕ𝑜𝑜𝑜𝑜𝑜𝑜
Then 𝐼𝐼(𝑛𝑛) will always return to 𝑅𝑅(𝑛𝑛) implying a Collatz conjecture sequence 𝑎𝑎0, , 𝑎𝑎1 , 𝑎𝑎2 , … 𝑎𝑎𝑥𝑥 where 𝑎𝑎𝑥𝑥 = 𝑅𝑅(𝑛𝑛) will
be produced.
{16𝑛𝑛 − 5}
Example 1: the matrix set � �
{54𝑛𝑛 − 14}
𝑓𝑓 𝑔𝑔 𝑓𝑓 𝑔𝑔 𝑔𝑔 𝑓𝑓
{16𝑛𝑛 − 5} → {48𝑛𝑛 − 14} → {24𝑛𝑛 − 7} → {72𝑛𝑛 − 20} → {36𝑛𝑛 − 10} → {18𝑛𝑛 − 5} → {54𝑛𝑛 − 14}
{64𝑛𝑛 − 61}
Example 2: the matrix set � �
{36𝑛𝑛 − 34}
𝑓𝑓 𝑔𝑔 𝑓𝑓 𝑔𝑔 𝑔𝑔 𝑔𝑔
{64𝑛𝑛 − 61} → {192𝑛𝑛 − 182} → {96𝑛𝑛 − 91} → {288𝑛𝑛 − 272} → {144𝑛𝑛 − 136} → {72𝑛𝑛 − 68} → {36𝑛𝑛 − 34}
 𝐼𝐼(𝑛𝑛) {2𝑛𝑛 − 1}
Let the starting matrix � �=� � and use the three operations to model the following tree diagram:
𝑅𝑅(𝑛𝑛) {2𝑛𝑛 − 1}

Figure 1
 3) Analysis of the tree diagram
According to the Collatz conjecture if 𝑎𝑎0 ∈ ℕ𝑜𝑜𝑜𝑜𝑜𝑜 , we start by applying the 𝑓𝑓 function and thereafter we apply the 𝑔𝑔
𝑓𝑓𝑔𝑔𝑥𝑥
function 𝑥𝑥 number of times thus 𝑎𝑎0 �⎯� 𝑎𝑎𝑥𝑥 ∶ 𝑎𝑎𝑥𝑥 ∈ ℕ𝑜𝑜𝑜𝑜𝑜𝑜 this brings us to a theorem where 𝑘𝑘, 𝑎𝑎𝑎𝑎𝑎𝑎 𝑥𝑥 ∈ ℕ:
𝐼𝐼(𝑛𝑛)
Theorem 1: For any matrix � � in figure 1 : 𝐼𝐼 (𝑛𝑛) ∈ 𝑆𝑆(𝑢𝑢) where 𝑢𝑢 = 2𝑘𝑘 from figure 1, 𝐸𝐸𝐸𝐸. 1 is true ⇔ 𝑥𝑥1 + 𝑥𝑥2 +
𝑅𝑅(𝑛𝑛)
𝑥𝑥3 + ⋯ + 𝑥𝑥𝑘𝑘 = 2𝑘𝑘 and the solutions to 𝐸𝐸𝐸𝐸. 1 are the monotonic lattice paths along the edges of a grid with 𝑘𝑘 × 𝑘𝑘
square cells. A monotonic path is one which starts in the lower left corner, finishes in the upper right corner, which
does not touch and pass above the diagonal line drawn from the start point to the end point. it consists entirely of
edges pointing rightwards or upwards. Counting such paths is equivalent to counting Dyck words [2,5,7] where:
3𝐼𝐼(𝑛𝑛)+1
3� �+1
3�…� 2𝑥𝑥1 �… �+1
2𝑥𝑥2
𝑓𝑓𝑔𝑔𝑥𝑥1 𝑓𝑓𝑔𝑔𝑥𝑥2 𝑓𝑓𝑔𝑔𝑥𝑥3 … 𝑓𝑓𝑔𝑔𝑥𝑥𝑘𝑘
�𝐼𝐼 (𝑛𝑛) �⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯� 𝑅𝑅 (𝑛𝑛) ∶ 𝐼𝐼 (𝑛𝑛) > 𝑅𝑅(𝑛𝑛)� = 𝐼𝐼 (𝑛𝑛) > 𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝐼𝐼 (𝑛𝑛) ≠ 1 … 𝐸𝐸𝐸𝐸. 1
2𝑥𝑥𝑘𝑘

fg
And a monotonic lattice paths along the edges of a grid with 𝑘𝑘 × 𝑘𝑘 square cells where X = → stands for "moving one
g 𝑓𝑓𝑔𝑔2
block to the right”, Y = → stands for "moving one block up", and XY = �⎯� stands for "moving one block to the right
and then one block up"

The image above including image (a) and (b) Ref: (stackoverflow.com/questions/34244047)
Proof of theorem 1 by boundary of the monotonic lattice paths with minimum and maximum turning:

3𝐼𝐼(𝑛𝑛) + 1
3� �+1
2𝑥𝑥1
3 �… � �…� + 1
2𝑥𝑥2
𝐼𝐼(𝑛𝑛) >
2𝑥𝑥𝑘𝑘
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) + 3𝑘𝑘−1 + 3𝑘𝑘−2 ∙ 2𝑥𝑥1 + 3𝑘𝑘−3 ∙ 2𝑥𝑥1 +𝑥𝑥2 +. . . +3𝑘𝑘−𝑘𝑘 ∙ 2𝑥𝑥1 +𝑥𝑥2 +𝑥𝑥3 +⋯+𝑥𝑥𝑘𝑘−1
𝐼𝐼 (𝑛𝑛) >
2𝑥𝑥1 × 2𝑥𝑥2 × … × 2𝑥𝑥𝑘𝑘
Since 𝑥𝑥1 + 𝑥𝑥2 + 𝑥𝑥3 + ⋯ + 𝑥𝑥𝑘𝑘 = 2𝑘𝑘 → 𝑥𝑥1 + 𝑥𝑥2 + 𝑥𝑥3 + ⋯ + 𝑥𝑥𝑘𝑘−1 = 2𝑘𝑘 − 𝑥𝑥𝑘𝑘 :
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−1 3𝑘𝑘−2 3𝑘𝑘−3 3𝑘𝑘−𝑘𝑘
𝐼𝐼 (𝑛𝑛) > + + + + ⋯ +
22𝑘𝑘 22𝑘𝑘 22𝑘𝑘−𝑥𝑥1 22𝑘𝑘−𝑥𝑥1 −𝑥𝑥2 22𝑘𝑘−𝑥𝑥1 −𝑥𝑥2 −𝑥𝑥3 −⋯−𝑥𝑥𝑘𝑘−1
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−1 3𝑘𝑘−2 3𝑘𝑘−3 3𝑘𝑘−𝑘𝑘
𝐼𝐼 (𝑛𝑛) − > + + + ⋯ + … 𝐸𝐸𝐸𝐸. 2
22𝑘𝑘 22𝑘𝑘 22𝑘𝑘−𝑥𝑥1 22𝑘𝑘−𝑥𝑥1 −𝑥𝑥2 2𝑥𝑥𝑘𝑘

The RHS of 𝐸𝐸𝐸𝐸. 2 is minimum ⇔ the term 𝑥𝑥𝑘𝑘 = 𝑘𝑘 + 1 thus we find the limit for.
 3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−1 3𝑘𝑘−2 3𝑘𝑘−3 3𝑘𝑘−𝑘𝑘
𝐼𝐼 (𝑛𝑛) − > + + + ⋯ +
22𝑘𝑘 22𝑘𝑘 22𝑘𝑘−1 22𝑘𝑘−2 2𝑘𝑘+1
𝑠𝑠=𝑘𝑘
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−𝑠𝑠
lim �𝐼𝐼 (𝑛𝑛) − � > lim �� �
𝑘𝑘→∞ 22𝑘𝑘 𝑘𝑘→∞ 22𝑘𝑘−𝑠𝑠+1
𝑠𝑠=1
∴ 𝐼𝐼 (𝑛𝑛) > 0 (a) min monotonic lattice turning path [10]

The RHS of 𝐸𝐸𝐸𝐸. 2 is maximum ⇔ 𝑥𝑥1 = 𝑥𝑥2 = 𝑥𝑥3 = ⋯ = 𝑥𝑥𝑘𝑘 = 2 thus we find the limit for.
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−1 3𝑘𝑘−2 3𝑘𝑘−3 3𝑘𝑘−𝑘𝑘
𝐼𝐼 (𝑛𝑛) − > + + + ⋯ +
22𝑘𝑘 22𝑘𝑘 22𝑘𝑘−2 22𝑘𝑘−4 22𝑘𝑘−2𝑘𝑘
𝑠𝑠=𝑘𝑘
3𝑘𝑘 ∙ 𝐼𝐼 (𝑛𝑛) 3𝑘𝑘−𝑠𝑠
lim �𝐼𝐼 (𝑛𝑛) − � > lim �� �
𝑘𝑘→∞ 22𝑘𝑘 𝑘𝑘→∞ 22𝑘𝑘−2𝑠𝑠+2
𝑠𝑠=1
∴ 𝐼𝐼 (𝑛𝑛) > 1
(b) max monotonic lattice turning path [10]

Therefore Theorem 1 is true and the solutions to 𝐸𝐸𝐸𝐸. 1 are the monotonic lattice paths along the edges of a grid with
𝑘𝑘 × 𝑘𝑘 square cells. A monotonic path is one which starts in the lower left corner, finishes in the upper right corner,
which does not touch and pass above the diagonal drawn from the start point to the end point.

Note 4: ∃𝐶𝐶𝑘𝑘−1 number of monotonic lattice path solutions that satisfy theorem 1, where:

1 2𝑘𝑘 − 2
𝐶𝐶𝑘𝑘−1 = � � … 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 [9]
𝑘𝑘 𝑘𝑘 − 1

Example 3: let 𝑘𝑘 = 3 thus 𝐶𝐶𝑘𝑘−1 = 2 monotonic lattice path solutions that satisfy the conditions of theorem 1,
where the solutions: (𝑥𝑥1 , 𝑥𝑥2 , … , 𝑥𝑥𝑘𝑘 ) = (1, 1, 4), 𝑎𝑎𝑎𝑎𝑎𝑎 (1, 2, 3).

{64𝑛𝑛 − 41} 𝑓𝑓𝑔𝑔1 𝑓𝑓𝑔𝑔1 𝑓𝑓𝑔𝑔4

E.g. The matrix � � ∶ 𝐼𝐼 (𝑛𝑛) = {2𝑢𝑢 ∙ 𝑛𝑛 − 2𝑣𝑣 + 1: 𝑛𝑛 ∈ ℕ} where, CCS = {64𝑛𝑛 − 41} �⎯⎯⎯⎯⎯⎯⎯� {27𝑛𝑛 − 17}
{27𝑛𝑛 − 17}

Example 4: let 𝑘𝑘 = 5 thus 𝐶𝐶𝑘𝑘−1 = 14 monotonic lattice path solutions that satisfy the conditions of theorem 1,
where the solutions: (𝑥𝑥1 , 𝑥𝑥2 , 𝑥𝑥3 , … , 𝑥𝑥𝑘𝑘 ) = (1, 2, 2, 2, 3), (1, 1, 2, 2, 4), (1, 1, 1, 3, 4), ( 1, 2, 1, 1, 5), (1, 1, 2, 1, 5),
(1, 1, 1, 2, 5), (1, 1, 1, 1,6), (1, 2, 1, 3, 3), (1, 1, 2, 3, 3), , (1, 1, 1, 4, 3), (1, 2, 2, 1, 4), (1, 1, 3, 1, 4), (1, 1, 3, 2, 3),
𝑎𝑎𝑎𝑎𝑎𝑎 (1, 2, 1, 2, 4).
 {1024𝑛𝑛 − 853}
E.g. The matrix set � � ∶ 𝐼𝐼 (𝑛𝑛) = {2𝑢𝑢 ∙ 𝑛𝑛 − 2𝑣𝑣 + 1: 𝑛𝑛 ∈ ℕ} where,
{243𝑛𝑛 − 202}

𝑓𝑓𝑔𝑔1 𝑓𝑓𝑔𝑔2 𝑓𝑓𝑔𝑔2 𝑓𝑓𝑔𝑔2 𝑓𝑓𝑔𝑔3

CCS = {1024𝑛𝑛 − 853} �⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯� {243𝑛𝑛 − 202}

4) Analysis of the monotonic lattice path solutions of Theorem 1

Recall that ∀𝐼𝐼 (𝑛𝑛) ∈ 𝑆𝑆(𝑢𝑢) in figure 1, the cardinality of 𝑆𝑆(𝑢𝑢) = 2𝑢𝑢−1 , and
𝑣𝑣=2𝑢𝑢−1 𝑣𝑣=2𝑢𝑢−1

ℕ𝑜𝑜𝑜𝑜𝑜𝑜 = � 𝑆𝑆(𝑢𝑢) , 𝑎𝑎𝑎𝑎𝑎𝑎 ∅ = � 𝑆𝑆(𝑢𝑢)

𝑣𝑣=1 𝑣𝑣=1

According to Theorem 1 and 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 3 ∃𝛾𝛾𝑘𝑘−1 matrix from figure 1 such that 𝐼𝐼 (𝑛𝑛) ≥ 𝑅𝑅(𝑛𝑛) 𝑤𝑤ℎ𝑒𝑒𝑒𝑒 𝑢𝑢 = 2𝑘𝑘 where:
𝑠𝑠=𝑘𝑘
𝑘𝑘−1 𝑘𝑘−2 𝑘𝑘−3 𝑘𝑘−𝑘𝑘
4𝑘𝑘−𝑠𝑠 2𝑠𝑠 − 2
𝛾𝛾𝑘𝑘−1 = 𝐶𝐶0 ∙ 4 + 𝐶𝐶1 ∙ 4 + 𝐶𝐶2 ∙ 4 + ⋯ + 𝐶𝐶𝑘𝑘−1 ∙ 4 =� � �
𝑠𝑠 𝑠𝑠 − 1
𝑠𝑠=1

Example 5: let 𝑘𝑘 = 2 since the cardinality of 𝑆𝑆(𝑢𝑢) = 8 thus there are eight matrices in figure 1 where 𝐼𝐼 (𝑛𝑛) ∈ 𝑆𝑆(𝑢𝑢)
and 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 1 & 2 are satisfied.
{16𝑛𝑛 − 1} {16𝑛𝑛 − 3} {16𝑛𝑛 − 5} {16𝑛𝑛 − 7} {16𝑛𝑛 − 9} {16𝑛𝑛 − 11} {16𝑛𝑛 − 13} {16𝑛𝑛 − 15}
� �;� � ;� �;� �;� �;� � ;� � ; 𝑎𝑎𝑎𝑎𝑎𝑎 � �
{81𝑛𝑛 − 1} {12𝑛𝑛 − 2} {27𝑛𝑛 − 7} {12𝑛𝑛 − 5} {27𝑛𝑛 − 14} {12𝑛𝑛 − 8} {9𝑛𝑛 − 7} {12𝑛𝑛 − 11}
and 𝛾𝛾𝑘𝑘−1 = 5 such that five out of the eight matrices in figure 1 satisfy the inequality 𝐼𝐼 (𝑛𝑛) ≥ 𝑅𝑅 (𝑛𝑛);
{16𝑛𝑛 − 3} {16𝑛𝑛 − 7} {16𝑛𝑛 − 11} {16𝑛𝑛 − 13} {16𝑛𝑛 − 15}
� �;� �; � � ;� � ; 𝑎𝑎𝑎𝑎𝑎𝑎 � �
{12𝑛𝑛 − 2} {12𝑛𝑛 − 5} {12𝑛𝑛 − 8} {9𝑛𝑛 − 7} {12𝑛𝑛 − 11}
Note 5: Let us assume the limit as 𝑘𝑘 → 𝜔𝜔 where 𝜔𝜔 is a really big number is:
𝑠𝑠=𝑘𝑘
4𝑘𝑘−𝑠𝑠 2𝑠𝑠 − 2
lim � � � = 22𝜔𝜔−1
𝑘𝑘→𝜔𝜔 𝑠𝑠 𝑠𝑠 − 1
𝑠𝑠=1

𝑠𝑠=𝑘𝑘
4𝑘𝑘−𝑠𝑠 2𝑠𝑠 − 2
ln � lim � � �� = ln(22𝜔𝜔−1 )
𝑘𝑘→𝜔𝜔 𝑠𝑠 𝑠𝑠 − 1
𝑠𝑠=1

𝑠𝑠=𝑘𝑘
4𝑘𝑘−𝑠𝑠 2𝑠𝑠 − 2
lim ln �� � �� = ln(22𝜔𝜔−1 )
𝑘𝑘→𝜔𝜔 𝑠𝑠 𝑠𝑠 − 1
𝑠𝑠=1

𝑠𝑠=𝜔𝜔
1 2𝑠𝑠 − 2
ln �� 𝑠𝑠
� �� = ln(22𝜔𝜔−1 ) − ln(4𝜔𝜔 )
𝑠𝑠 × 4 𝑠𝑠 − 1
𝑠𝑠=1

1 2𝑠𝑠 − 2
ln �∑𝑠𝑠=𝜔𝜔 � ��
𝑠𝑠=1 𝑠𝑠 × 4𝑠𝑠 𝑠𝑠 − 1
= −1 … 𝐸𝐸𝐸𝐸. 3
ln(2)
𝐸𝐸𝐸𝐸. 3 is true if and only if 𝜔𝜔 → ∞
Note 6: let us take the ratio:
4𝑘𝑘−𝑠𝑠 2𝑠𝑠 − 2
𝛾𝛾𝑘𝑘−1 ∑𝑠𝑠=𝑘𝑘
𝑠𝑠=1 𝑠𝑠 � �
= 𝑠𝑠 − 1
22𝑘𝑘−1 22𝑘𝑘−1
 𝛾𝛾𝑘𝑘−1
and one will find the lim = 1 … 𝐸𝐸𝐸𝐸. 4
𝑘𝑘→∞ 22𝑘𝑘−1
𝛾𝛾𝑘𝑘−1
Note 6: lets plot 𝐸𝐸𝐸𝐸. 4 ∶ 𝑦𝑦 = 𝑎𝑎𝑎𝑎𝑎𝑎 𝑥𝑥 = 𝑘𝑘
22𝑘𝑘−1

∴ 𝐸𝐸𝐸𝐸. 3 𝑎𝑎𝑎𝑎𝑎𝑎 𝐸𝐸𝐸𝐸. 4 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 1 𝑄𝑄. 𝐸𝐸. 𝐷𝐷.

References
[1] J.J. O’Connor and Robertson E.F. (2006) Scotland. School of Mathematics and statistic University of St Andrews.

[2] Instructor Sushmita Ruj (2014) Discrete mathematics lecture notes. Scribe: Nishhant Kumar and Subhadip
Singha

[3] Peter J. Cameron, (Combinatorics Cambridge University Press, 1994. MR1311922)

[4] Jian Yi She lattice path, and bounded partition, (J. Combin. Theory Ser, DOI 10.1016/0097-3165(93)90026-5
MR1213132)

[5] Toufik Mansour and Mark Shattuck. Counting Dyck paths according to the maximum distance between peaks
and valleys. Department of Mathematics University of Haifa, 31905 Haifa, Israel and the Department of
Mathematics University of Tennessee, Knoxville, TN 37996.
[6] Marcus Elia and Amanda Tucker. 2015 Conservative integers and the Collatz conjecture. Department of
Mathematics, SUNY Geneseo, Geneseo, NY and the Department of Mathematics, University of Rochester, Rochester,
NY.
[7] Michael Wallner diploma thesis. Lattice Path Combinatorics. Diskrete Mathematik und Geometrie Vienna
University of Technology. Under the Supervisor: Univ.-Prof. Dipl.-Ing. Dr. Michael Drmota
[8] Jeremy Gunawardena. 2006 Matrix algebra for beginners. Department of Systems Biology Harvard Medical
School 200 Longwood Avenue, Cambridge, MA 02115, USA.
[9] Philippe Flajolet and Robert Sedgewick. 2009 Analytic Combinatorics. Cambridge University Press.
[10] Sri Gopal Mohanty. 1979 Lattice Path Counting and Applications, volume 37 of Probability and Mathematical
Statistics. Academic Press, New York, USA.