Date: May 10 2023
Summary: A computational treatment and exploration of abundant and deficient numbers
Keywords: #blog #abundant #deficient #number #theory #julia #programming #perfect #aliquot #sequence #archive
Not Available
I was at the gym working out when I started thinking about locker numbers in the men's locker room. I was reminded of perfect numbers and was thinking about perfect number examples. I started testing random numbers in my mind and noticed numbers which had divisors that summed up to greater than their number and also less than their number. I had no idea about the existence of abundant and deficient numbers and got curious about these numbers and to see what characteristics I could find about them.
In Number Theory, there exist three species of numbers that depend on the divisors of a given number (excluding the number itself as a divisor). Here are the three species and their simple characteristics:
Deficient Numbers - these numbers have divisors whose sum is never greater than the number being examined. An example is the number which has as divisors and – those divisors only sum up to .
Perfect Numbers - these numbers have divisors which sum to exactly to the number being examined. An example is the number which has as divisors which sum together to .
Abundant Numbers - these numbers have divisors whose sum is greater than the number being examined. An example is the number whose divisors are and sum to .
As it turns out, there are infinite deficient, perfect, and abundant numbers. However, only around 50 perfect numbers have ever been discovered to this day! is the smallest perfect number but then perfect numbers grow to be hundreds of digits long! For that reason, this fun exploration will really only explore abundant and deficient numbers.
Out of curiosity, I wanted to know if there were any trends to be noticed in the divisors of the deficient and abundant number species. So, I whipped together some code to explore this within Julia (if you are not interested in the code, you can skip it and just go to the results for each section). To get started, I first defined a function to calculate divisors of a number:
import Primes: factor
function divisors(n)
d = Int64[1]
for (p, e) in factor(n)
t = Int64[]
r = 1
for i in 1:e
r *= p
for u in d
push!(t, u * r)
end
end
append!(d, t)
end
return sort!(d)
end
With this function defined, now, I am going to calculate some deficient and deficient numbers (since perfect numbers are hard to calculate, I am going to look up a few to explore). To do that, we will use the following snippet to find abundant and deficient numbers:
i = 1
deficient_numbers = []
abundant_numbers = []
while true
divisor_sum = divisors(i)[1:end-1] |> sum
if divisor_sum < i && length(deficient_numbers) != 1000
push!(deficient_numbers, i)
elseif divisor_sum > i && length(abundant_numbers) != 1000
push!(abundant_numbers, i)
end
i += 1
length(abundant_numbers) == 1000 && length(deficient_numbers) == 1000 ? break : continue
end
We are set to explore further these numbers!
As a first pass, let's calculate the divisors of the abundant numbers and plot their frequency:
import DataStructures: counter
import UnicodePlots: barplot
abundant_divisors = vcat(divisors.(abundant_numbers)...) |> counter |> Dict |> sort;
vs = collect(values(abundant_divisors));
ks = collect(keys(abundant_divisors));
barplot(ks[1:20], vs[1:20], xlabel = "Count", ylabel = "Divisors", title = "Divisor Count for First 1000 Abundant Numbers")
Which gives the following plot:
Divisor Count for First 1000 Abundant Numbers
┌ ┐
1 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 1 000
2 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 995
3 ┤■■■■■■■■■■■■■■■■■■■■■■ 677
4 ┤■■■■■■■■■■■■■■■■■■■■■ 623
5 ┤■■■■■■■■■■ 308
6 ┤■■■■■■■■■■■■■■■■■■■■■■ 672
7 ┤■■■■■■■ 216
8 ┤■■■■■■■■■■■ 347
9 ┤■■■■■■■■ 229
Divisors 10 ┤■■■■■■■■■■ 303
11 ┤■■■■ 116
12 ┤■■■■■■■■■■■ 336
13 ┤■■■ 96
14 ┤■■■■■■■ 211
15 ┤■■■■■ 139
16 ┤■■■■■■ 188
17 ┤■■ 68
18 ┤■■■■■■■ 224
19 ┤■■ 61
20 ┤■■■■■■■ 201
└ ┘
Count
Without any real methodology, what I notice is that there seems to be an interesting pattern where certain divisors are being repeated more than others as more and more divisors are found. It almost feels like a kind of decaying sequence where counts seems to spike on any multiple of or more consistently than any other number. Even though, it seems like multiples of are not as consistent.
Now, let's calculate the divisors of the deficient numbers and plot their frequency:
import DataStructures: counter
import UnicodePlots: barplot
deficient_divisors = vcat(divisors.(deficient_numbers)...) |> counter |> Dict |> sort;
vs = collect(values(deficient_divisors));
ks = collect(keys(deficient_divisors));
barplot(ks[1:20], vs[1:20], xlabel = "Count", ylabel = "Divisors", title = "Divisor Count for First 1000 Deficient Numbers")
Which gives the following plot:
Divisor Count for First 1000 Deficient Numbers
┌ ┐
1 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 1 000
2 ┤■■■■■■■■■■■ 337
3 ┤■■■■■■■ 220
4 ┤■■■■ 124
5 ┤■■■■■■ 168
7 ┤■■■■ 119
8 ┤■■ 51
9 ┤■■ 73
10 ┤■ 36
Divisors 11 ┤■■■ 82
13 ┤■■ 71
14 ┤■ 25
15 ┤■ 43
16 ┤■ 18
17 ┤■■ 56
19 ┤■■ 50
21 ┤■ 31
22 ┤■ 22
23 ┤■ 42
25 ┤■ 33
└ ┘
Count
What's interesting here is that I did not see any immediate pattern or phenomena with these divisors at first glance. However, when I examined the plot using a log10 scale, I then saw that consistently, the counts for odd divisors for outnumber those for even divisors:
Divisor Count for First 1000 Deficient Numbers
┌ ┐
1 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 1 000
2 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 337
3 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■ 220
4 ┤■■■■■■■■■■■■■■■■■■■■■■■ 124
5 ┤■■■■■■■■■■■■■■■■■■■■■■■■ 168
7 ┤■■■■■■■■■■■■■■■■■■■■■■■ 119
8 ┤■■■■■■■■■■■■■■■■■■■ 51
9 ┤■■■■■■■■■■■■■■■■■■■■ 73
10 ┤■■■■■■■■■■■■■■■■■ 36
Divisors 11 ┤■■■■■■■■■■■■■■■■■■■■■ 82
13 ┤■■■■■■■■■■■■■■■■■■■■ 71
14 ┤■■■■■■■■■■■■■■■ 25
15 ┤■■■■■■■■■■■■■■■■■■ 43
16 ┤■■■■■■■■■■■■■■ 18
17 ┤■■■■■■■■■■■■■■■■■■■ 56
19 ┤■■■■■■■■■■■■■■■■■■■ 50
21 ┤■■■■■■■■■■■■■■■■ 31
22 ┤■■■■■■■■■■■■■■■ 22
23 ┤■■■■■■■■■■■■■■■■■■ 42
25 ┤■■■■■■■■■■■■■■■■■ 33
└ ┘
Count (Log10 Scale)
Out of curiosity, I wondered if there could be any overlap of abundant and deficient numbers' divisors with their respective aliquot sequences. Now, an aliquot sequence is a rather fun thing. It has the following form:
I decided to implement a small algorithm to compute the aliquot sequence for a given number as follows:
function aliquot_sequence(num; max_itrs = missing)
sequence = [num]
s = num
while true
s = sum(divisors(s)) - s
if !ismissing(aliquot_sequence) && length(sequence) == max_itrs
return nothing
elseif s == 0
push!(sequence, s)
break
elseif in(s, sequence)
break
else
push!(sequence, s)
end
end
return sequence
end
In my implementation, I decided to limit the sequence to no repeating sequence values for a number. Let's plot these sequence values and see what could be seen as before.
NOTE: As a limitation, some of these sequences have an immensely high number of iterations which cause my computer to explode (looking at you, abundant number )! For that reason, I am only calculating sequences for an abundant number that has only 10 maximum iterations within their aliquot sequence.
Let's calculate the aliquot sequences for abundant numbers that have at most terms within their sequence:
import DataStructures: counter
import UnicodePlots: barplot
abundant_aliquot_sequences = []
for i in 1:1000000
divisor_sum = divisors(i)[1:end-1] |> sum
if divisor_sum > i
seq = aliquot_sequence(i, 10)
!isnothing(seq) ? push!(abundant_aliquot_sequences, seq) : continue
end
i += 1
length(abundant_aliquot_sequences) == 500 ? break : continue
end
abundant_aliquot_terms = vcat(abundant_aliquot_sequences...) |> counter |> Dict |> sort;
vs = collect(values(abundant_aliquot_terms));
ks = collect(keys(abundant_aliquot_terms));
barplot(ks[1:20], vs[1:20], xlabel = "Count", ylabel = "Terms", title = "Aliquot Term Count for 500 Abundant Numbers")
Which yields the plot:
Aliquot Term Count for 500 Abundant Numbers
┌ ┐
0 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 448
1 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 448
3 ┤ 3
4 ┤ 3
6 ┤ 3
7 ┤■ 9
8 ┤■ 9
9 ┤ 3
10 ┤ 1
Terms 11 ┤ 4
12 ┤ 1
13 ┤■ 11
14 ┤ 1
15 ┤ 3
16 ┤ 1
17 ┤ 3
18 ┤ 2
19 ┤■ 8
20 ┤ 1
21 ┤ 4
└ ┘
Count
Here, I really cannot discern any relatable pattern as well as significance that can be tied back to abundant numbers. I am not sure if there is a way to tie significance back to abundant numbers at all in this scenario.
Let's calculate the aliquot sequences for abundant numbers that have at most terms within their sequence:
import DataStructures: counter
import UnicodePlots: barplot
deficient_aliquot_sequences = []
for i in 1:1000000
divisor_sum = divisors(i)[1:end-1] |> sum
if divisor_sum < i
seq = aliquot_sequence(i, 10)
!isnothing(seq) ? push!(deficient_aliquot_sequences, seq) : continue
end
i += 1
length(deficient_aliquot_sequences) == 500 ? break : continue
end
deficient_aliquot_terms = vcat(deficient_aliquot_sequences...) |> counter |> Dict |> sort;
vs = collect(values(deficient_aliquot_terms));
ks = collect(keys(deficient_aliquot_terms));
barplot(ks[1:20], vs[1:20], xlabel = "Count", ylabel = "Terms", title = "Aliquot Term Count for 500 Deficient Numbers")
Which yields the plot:
Aliquot Term Count for 500 Deficient Numbers
┌ ┐
0 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 487
0 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 487
1 ┤■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ 487
2 ┤ 1
3 ┤■■ 28
4 ┤■■ 27
5 ┤ 1
6 ┤■ 10
7 ┤■■ 24
8 ┤■■ 23
Terms 9 ┤■■ 26
10 ┤■ 10
11 ┤■ 17
12 ┤ 2
13 ┤■ 15
14 ┤■ 9
15 ┤■■ 25
16 ┤■ 7
17 ┤■ 13
18 ┤ 1
19 ┤■■ 25
└ ┘
Count
Again, I really cannot discern any relatable pattern as well as significance that can be tied back to deficient numbers.
This was a small exploration that I wanted to do of these numbers to see if I could find any patterns or significance within aspects of these numbers. It seems like there may be some present within the factors of abundant and deficient numbers, but when looking at their corresponding aliquot sequences, I am unable to determine anything from a computational sense. To that end, I was also curious about how effective computation can be in helping to derive or provide hints about what may underlie these numbers. In short, it would appear that computation is quite helpful to give rise to initial questions. For example, I'd be curious to what extent the patterns I noticed within abundant and deficient numbers prolong for and if they are actually legitimate observations. At that point, one could then start applying basic data science skills to group, explore, and summarize potential trends within these numbers.
For now, my curiosity is sated and it might be worth a return to in the future. One thing this blog post did make me think about is analogies. The idea of deficient, perfect, and abundant numbers are really fascinating as it lends itself to analogs within set theory relationships (like many-to-one -> deficient number, one-to-one -> perfect number, one-to-many -> abundant number). I wonder if it could be used as analogy outside of mathematics strictly and in terms like healthcare (sub-type of a disease -> deficient number, canonical disease diagnosis -> perfect number, disease family -> abundant number). Might be worth further exploration in the future.
There was a nice discussion from within the Julia Discourse about this post talking about implementation details of some of the functions I was using and how to handle large numbers in computation. In particular, there was suggestion on using types like BigInt
or BigFloat
to handle these large numbers (such as the 8th perfect number). Interestingly, to calculate divisors, one user (gtgt) suggested the following approach which was quite beyond my thinking to calculate divisors for a given value. Here was their approach:
You can use the fact that if then to avoid allocating a vector to store the divisors.
This was then followed by a programming implementation:
using Primes
function sum_divisors(n)
s = one(n)
for (p, e) in Primes.factor(n)
s *= (p^(e + 1) - 1) ÷ (p - 1)
end
s
end
function get_abundant_and_deficient_numbers(n::T) where T <: Integer
# get the first n abundant and deficient numbers
n_abundants = 0
n_deficients = 0
abundants = sizehint!(T[], n)
deficients = sizehint!(T[], n)
k = 1
while n_abundants < n || n_deficients < n
σ = sum_divisors(k)
if σ > 2k && n_abundants < n
n_abundants += 1
push!(abundants, k)
elseif σ < 2k && n_deficients < n
n_deficients += 1
push!(deficients, k)
end
k += 1
end
abundants, deficients
end
I haven't had a chance to test that new implementation but I would imagine, being that it is far more type stable, that it would be more efficient. However, I still feel like we need to have safeguards for large number computation.
Within the Category Theory Zulip community, there was another great discussion about viewing these number species through the lens of categories.
David Egolf and John Carlos Baez had some fantastic ideas within that discussion that I'll excerpt here:
David: [I] wonder if the concepts of "deficient", "perfect" and "abundant" generalize to certain kinds of categories. I suppose what we would need is:
a way to say if one object is a divisor of another object
a way to add objects
a way to compare the size of objects
If we have the three things above available to us, then we can
describe an object as "abundant" if the sum of its divisor objects is larger than .
[...] Here's an initial idea for "categorifying" the above list of three requirements, in a category with coproducts:
Say that divides if the coproduct of with itself some finite number of times is isomorphic to
Let the sum of two objects be their coproduct (so the sum is defined up to isomorphism)
Say that if there is a monomorphism from to
Applying this to the category of finite sets I think gives us something similar to the usual notions of divisiblity, addition, and ordering for the natural numbers. Another approach for divisibility might be, in a category with products:
Say that divides if there exists some so that the product of and is isomorphic to
John: Both these approaches work, in the sense that starting from the category of finite sets and functions we get the usual concept of divisibility for natural numbers.
I encourage you to go see the rest of that discussion there if you found this perspective interesting! There was a lot more spoken of there, but I found this really fascinating. For me, my category theoretic skills still are not up to par just yet to track everything David and John are saying, so I am leaving these excerpts here for future reference to perhaps come back to this.
Zelko, Jacob. A Fun Exploration of Perfect, Abundant, and Deficient Numbers. https://jacobzelko.com/05102023043333-perfect-abundant-deficit. May 10 2023.