Resolution of the 3n + 1 problem using

inequality relation between indices of 2 and 3

Gaurav Goyal

Department of Mechanical Engineering

Indian Institute of Technology, Delhi, India
email: Gaurav.Goyal@mech.iitd.ac.in

April 8, 2023

Abstract
Collatz conjecture states that an integer n reduces to 1 when cer-
tain simple operations are applied to it. Mathematically, it is written
as 2z = 3k n + C, where z, k, C ≥ 1. Suppose the integer n vio-
lates Collatz conjecture by reappearing, then the equation modifies
to 2z n = 3k n + C. The article takes an elementary approach to this
problem by stating that the inequality 2z > 3k must hold for n to vi-
olate the Collatz conjecture. It leads to the inequality z > 3k/2 that
helps obtain bounds on the value of 3k /2z and 2z − 3k . It is found
that the 3n + 1 series loops for 1 and negative integers. Finally, it is
proved that the 3n + 1 series shows pseudo-divergence but eventually
arrives at an integer less than the starting integer.

Keywords— Collatz conjecture, 3n+1, inequality relations

1 Introduction
Collatz conjecture, or the 3n + 1 problem, is a simple arithmetic function applied
to positive integers. If the integer is odd, triple it and add one. It is called the odd

1
step. If the integer is even, it is divided by two and is denoted as the even step. It
is conjectured that every integer will eventually reach the number 1. Much work
has been done to prove or disprove this conjecture [1–4].
The problem is easy to understand, and since it has attracted much attention
from the general public and experts alike, the literature is endless. Still, the
efforts made to tackle the 3n + 1 problem can generally be categorized under the
following headings:

• Experimental or computational method: This method uses computational

optimizations to verify Collatz conjecture by checking numbers for conver-
gence [5–7]. Numbers as large as 1020 have shown no divergence from the
conjecture.

• Arguments based on probability: The heuristic argument suggests that, on

average, the sequence of numbers tends to shrink in size so that divergence
does not occur. On average, each odd number is 3/4 of the previous odd
integer [8].

• Evaluation of stopping times: Many researchers seem to work on the 3n + 1

problem from this approach [9–13]. In essence, it is sought to prove that the
Collatz conjecture yields a number smaller than the starting number.

• Mathematical induction: It is perhaps the most common method to “prove”

the Collatz conjecture. The literature involving this particular method seems
endless [14, 15].

The issue is that the Collatz conjecture is a straightforward arithmetic op-

eration, while the methods used are not. The mismatch is created because the
problem has attracted the attention of brilliant people in mathematics who are
used to dealing with complex issues with equally complex tools. Therefore, an
elementary analysis of the problem might be lacking.
This article takes a rudimentary approach to the Collatz conjecture and treats it
as a problem of inequality between indices of 2 and 3. The inequality relation will
be turned into equality using variables. The values of these variables will be inves-
tigated, and it will be shown that the Collatz conjecture does not need complex
analysis.

2
2 Prerequisite
Consider that n is an odd integer, and the following function f is applied.

3n + 1, if n is odd
f (n) =
n if n is even
2

A sequence is formed by performing this operation repeatedly, taking the re-

sult at each step as the input for the next. Collatz conjecture states that, for all
n, f k (n) = 1 for some non-negative integer k, where the function is applied to n
exactly k times. Let the sequence of integers obtained be:

1st odd, z1 evens, 2nd odd, z2 evens, · · · , k th odd, za evens.

The function f k (n) is computed to be

z1 +1}
 { 3n+1
  



 3 2 +1


2

 
z 

2
    



 3 2 z3 +1 


 
 
 
 

3 .. +1
.

 


 


 


 

f k (n) =
2 za
3k n + 3k−1 + 3k−2 2z1 + · · · + 30 2z1 +z2 +···+za−1
f k (n) =
2z1 +z2 +···+za−1 +za
3k n + C
f k (n) = (1)
2z

where

z = z1 + z2 + · · · + za−1 + za
C = 3k−1 + 3k−2 2z1 + · · · + 30 2z1 +z2 +···+za−1 (2)

It is noted that (z, C) > 0 are integers, and C is always odd.

3
3 Relation between z and k
Suppose the integer n > 1 violates Collatz conjecture by reappearing in the se-
quence, i.e., f k (n) = n. Therefore,

2z n = 3k n + C (3)

Modify Eq. (3) to the following forms:

C
n= (4)
2z
− 3k
3k C
n = zn+ z (5)
2 2
C = (2z − 3k )n (6)

The following inequality is established from Eq. (3):

2z > 3k (7)
z k
(23 ) > (32 )
3 2

z k
8 3 > (8 + 1) 2
z k
83 > 82
3k
z> (8)
2

The minimum integer solution of Eq. (8) is (k, z) = (1, 2).

4 Calculation of 3k /2z and 2z − 3k

For Eq. (4) to be valid and noting that the difference of integers is also an integer,

4
 2z − 3k ≥ 1
3k 1
1− z
≥ z
2 2
3k 3
z
≤ (9)
2 4
3k 3
z
= −p (10)
2 4

Where the minimum value of z is used to find the upper bound on 3k /2z and
p ∈ [0, 3/4).
Bounds on 2z − 3k is found using Eq. (9)

3k 3
z
≤
2 4
2z
2z − 3k ≥
4
z k 2z
2 −3 = +q (11)
4

Where q ≥ 0.

5 Convergence to 1 in the 3n + 1 series

The value of C calculated from Eq. (11) and (6) is C = (2z /4 + q)n. The value of
C and 3k /2z is substituted in Eq. (5).

   z 
3 2 n
n= −p n+ +q
4 4 2z
q
p= (12)
2z

Equation (12) is valid for two cases discussed in the following sections.

5
5.1 Case 1: p = q = 0
For p = 0, Equation (10) gives (k, z) = (1, 2).
Similarly, for q = 0, Eq. (12) gives p = 0, which in turn gives (k, z) = (1, 2) from
Eq. (10).

q
5.2 Case 2: p = 2z
Substitute the value of q in Equation (11) to obtain

2z
C= n + 2z np
4  
z 1
C=2 n +p (13)
4

Compare Eq. (2) and (13) by rearranging as follow

 
z 1
2 n +p = 3k−1 + 3k−2 2z1 + · · · + 30 2z1 +z2 +···+za−1
4
 k−1
3k−2 2z1 30 2z1 +z2 +···+za−1
  
1 3
2z n +p =2 z
+ + ··· +
4 2z 2z 2z
 z1 +z2 +···+za−1
3k−1 3k−2 2z1 31 2z1 +···za−2
  
1 2
2z n +p =2 z
+ z + + ··· +
4 2z 2 2z 2z
   
1 1
2z n + p = 2z + some fraction (14)
4 2za

Compare coefficients to obtain n = 1 and za = 2. Value of p and z can not

be determined from Eq. (14), but since n = 1, the solution is (n, k, z, p, q) =
(1, 1, 2, 0, 0).

6 Non-convergence to 1 in 3n + 1 series
The previous section demonstrated that no positive integer n violates the Collatz
conjecture by reappearing in the series other than 1. The convergence to 1 is
possible when Eq. (7) is true. Therefore, for non-convergence or divergence to
occur, the following is true.

6
 3k > 2z (15)

Let an integer m be obtained after applying 3n + 1 rule repeatedly. The Eq.

(3) modifies to

2z m = 3k n + C
n m o
z k
C=n 2 −3 (16)
n

Since C > 0

m
2z − 3k > 0
n
2z Q − 3k > 0
3k < 2z Q

Where m/n = Q and Q > 0.

When Q = 1, the value of C becomes negative. Therefore, it is required that n < 0
for C be positive according to Eq. (16). Using a negative value of n is equivalent
to modifying the Collatz conjecture to the 3n − 1 series. It is already known that
3n − 1 series does not converge to 1 for integers like 5, 7, 17. Therefore, the 3n + 1
series does not converge to 1 only for a negative integer.

7 Pseudo-divergence in the 3n + 1 series

Consider that any odd integer n > 0 can be expressed as n = 2a + 2b + · · · + 2z − 1,
where a > b > · · · > z and the indices need not be consecutive integers. For
simplicity, take n = 2m − 1 and apply odd step

O1 = 3n + 1
O1 = (2 + 1)(2m − 1) + 1
O1 = 2m+1 + 2m − 2

7
 Since the RHS is even, apply the even step

E1 = 2m + 2m−1 − 1

Which is odd, and hence an odd step will follow. A few iterations of odd and
even steps are listed.

O2 = 2m+2 + 2m−1 − 2
E2 + 2m+1 + 2m−2 − 1
O3 = 2m+2 + 2m+1 + 2m−1 + 2m−2 − 2
E3 = 2m+1 + 2m + 2m−2 + 2m−3 − 1
O4 = 2m+3 + 2m+1 + 2m−3 − 2
E4 = 2m+2 + 2m + 2m−4 − 1

At first glance, the sequence is increasing and might diverge to infinity as the
power of 2 increases. Note that the value of m is decreasing as well as increasing.
The decrease is consistent with the step but the increase lags. After m steps,
the second last term in the RHS becomes 20 , which cancels −1, and the lowest
remaining term is 2m .
The expression of the integer after m steps is

Em = 2m+α + · · · + 2m
= 2m {2α + · · · + 1}

Where α < m. The above integer is a multiple of 2m ; hence, the even step
follows for m times in succession. The final integer is 2α + · · · + 1, less than 2m .
The same general conclusion is derived for n = 2a + 2b + · · · + 2z − 1. Hence, it is
concluded that the 3n + 1 sequences show pseud-divergence but eventually arrive
at an integer that is less than the starting integer.

8
8 Conclusion
This article shows that the Collatz conjecture can be treated as an inequality
relation between the indices of 2 and 3. Values of 3k /2z and 2z −3k are determined
and substituted in 2z n = 3k n + C. It is found that the 3n + 1 series loops for 1
and negative integers. Finally, it is shown that the 3n + 1 series shows pseudo-
divergence but eventually arrives at an integer less than the starting integer.

References
[1] J. C. Lagarias, The ultimate challenge: The 3x+ 1 problem. American Math-
ematical Soc., 2010.

[2] J. C. Lagarias, “The 3x+ 1 problem: An annotated bibliography (1963–

1999),” The ultimate challenge: the 3x, vol. 1, pp. 267–341, 2003.

[3] J. C. Lagarias, “The 3x+ 1 problem: An annotated bibliography, ii (2000-

2009),” arXiv preprint math/0608208, 2006.

[4] J. C. Lagarias, “The 3x+ 1 problem: An annotated bibliography,” preprint,

2004.

[5] D. Barina, “Convergence verification of the collatz problem,” The Journal of

Supercomputing, vol. 77, no. 3, pp. 2681–2688, 2021.

[6] A. Rahn, E. Sultanow, M. Henkel, S. Ghosh, and I. J. Aberkane, “An algo-

rithm for linearizing the collatz convergence,” Mathematics, vol. 9, no. 16,
p. 1898, 2021.

[7] E. Yolcu, S. Aaronson, and M. J. Heule, “An automated approach to the

collatz conjecture.,” in CADE, pp. 468–484, 2021.

[8] K. Barghout, “On the probabilistic proof of the convergence of the collatz
conjecture,” Journal of Probability and Statistics, vol. 2019, 2019.

[9] R. Terras, “A stopping time problem on the positive integers,” Acta Arith-
metica, vol. 3, no. 30, pp. 241–252, 1976.

9
[10] T. Tao, “Almost all orbits of the collatz map attain almost bounded values,”
in Forum of Mathematics, Pi, vol. 10, p. e12, Cambridge University Press,
2022.

[11] S. G. Jiang, “Collatz total stopping times with neural networks,” World Sci-
entific Research Journal, vol. 7, no. 1, pp. 296–301, 2021.

[12] D. Applegate and J. Lagarias, “Lower bounds for the total stopping time of
3n+1 iterates,” Mathematics of computation, vol. 72, no. 242, pp. 1035–1049,
2003.

[13] M. Chamberland, “Averaging structure in the 3x+1 problem,” Journal of

Number Theory, vol. 148, pp. 384–397, 2015.

[14] W. Ren, “A new approach on proving collatz conjecture,” Journal of Mathe-

matics, vol. 2019, 2019.

[15] M. Orús-Lacort and C. Jouis, “Analyzing the collatz conjecture using the
mathematical complete induction method,” Mathematics, vol. 10, no. 12,
p. 1972, 2022.

10