Patterns in Pythagoras
3C. Even More Patterns in Right Triangles
and Pythagorean Triples
David W. Hansen
© 2008
5. Complete Primitive Pythagorean Triples
In any primitive Pythagorean triple (a. b. c), b + c = (2mn) + (m2 + n2)
= m2 + 2mn + n2
= (m + n)2,
_____
so b + c is a perfect square, and Ö b + c = m + n, (1)
which is a factor of a, since a = m2 – n2 = (m + n)(m – n). (2)
_____ _____
Now, if m – n = 1 in (2) above, then a = m + n = Ö b + c , from (1) above. Thus, Ö b + c = a,
which is so nice that we shall say:
______
Any triple (a, b, c) in which m = n+1, and thus Ö b + c = a, is called a complete triple.
________________________________________________________________________
Note that m – n = 1 implies that m = n + 1.
Let’s have some fun and construct a table of complete primitive Pythagorean triples sorted in ascending order by their values of m and n. Perhaps we shall find some interesting patterns! As stated above, we must have m – n
= 1 (or m = n + 1) for each of these triples to be complete.
Table 5 (Complete Triples)
m n a b c
---------------------------------------------
2 1 3 4 5
3 2 5 12 13
4 3 7 24 25
5 4 9 40 41
6 5 11 60 61
So, can you see any interesting patterns? Well, we see that in each complete triple, a is an odd integer, and c is one more than b. This is always true for complete triples since m = n + 1, so
a = m2 – n2 = (n+1)2 – n2 = n2 + 2n +1 – n2 = 2n + 1, (3)
which is an odd integer. The values of a thus form a set of increasing consecutive odd integers starting at 3.
b = 2mn = 2(n + 1)(n) = 2n2 + 2n, (4)
and c = m2 + n2 = (n + 1)2 + n2 = 2n2 + 2n + 1. (5)
Comparing (4) and (5) above, we see that indeed, c = b + 1. Thus, in a complete triple (a, b, c),
m = n+1, a = 2n+1, and c = b+1. (6)
Are there other patterns? Yes, since a < b for each triple in Table 5, this suggests that all complete triples are regular. Now, n is always ³ 1, so
n > 1 / Ö2 » 0.707. Squaring both sides of this inequality, we have n2 > 1 / 2,
or 2n2 > 1. Adding 2n to both sides of this inequality gives us 2n2 + 2n > 2n + 1,
which is just b > a from (4) and (3) above, respectively.
Thus, a < b, and a complete triple is regular. Combining this result with (3), (4), (5), and (6), we have
Theorem 37
_____
A primitive Pythagorean triple (a, b, c) is said to be complete if and only if Ö b + c = a.
For any complete triple, m = n+1, a = 2n+1, b = 2n2 + 2n, c = 2n2 + 2n + 1, and c = b+1.
In addition, a < b, so all complete triples are regular.
____________________________________________________________________________
Any more patterns? Look at Table 5 again, taking note of the sums and differences of the a’s and b’s.
In the first complete triple, (3,4,5), the sum of a and b is 7, and in the second triple, (5,12,13), the sum of a and b is 17, which is not much help. However, the difference of b and a in this second triple is 7, the same value as the sum of a and b in the first triple! Is this just a coincidence? Let’s see.
In the second triple (5,12,13), the sum of a and b is 5+12 = 17, and in the third triple (7,24,25), the difference between b and a is 24 – 7, which is also 17! Now, this looks good!
Checking the other triples in Table 5, we find that for every one of the complete triples in the table, the sum of
a and b for any given triple equals the difference of b and a in the next succeeding triple. Thus, for example, the sum of a and b in the fourth triple (9,40,41) is 49, which is equal to the difference of b and a (60 – 11 = 49) in the fifth triple (11,60,61).
It appears that if (a1, b1, c1) is any given complete triple in a list of complete triples sorted in ascending order by their a-values, and (a2, b2, c2) is the next succeeding triple in this list, then
a1 + b1 = b2 – a2. (7)
A very interesting way to look at this is to add a2 to both sides of (7) above, giving us
a1 + b1 + a2 = b2. (8)
Then, in our list of complete triples, (a1, b1, c1) and its successor triple (a2, b2, c2) look like this:
a b c
--------------------------------------------
. . . . . . . . .
a1 b1 c1
¬ a1 + b1 + a2 = b2
a2 b2 c2
. . . . . . . . .
Let’s look at Table 5 again.
Table 5 (Complete Triples)
m n a b c
----------------------------------------------
2 1 3 4 5
¬ 3 + 4 + 5 = 12
3 2 5 12 13
¬ 5 + 12 + 7 = 24
4 3 7 24 25
¬ 7 + 24 + 9 = 40
5 4 9 40 41
¬ 9 + 40 + 11 = 60
6 5 11 60 61
Here we can clearly see that for any two successive complete triples, a1 + b1 + a2 = b2. How very nice.
But this is not all! Take a look at the values of the a’s and c’s for any two successive complete triples in Table 5 as shown below.
Table 5 (Complete Triples)
m n a b c
-----------------------------------------------------
2 1 3 4 5
¬ 3 + 5 + 5 = 13
3 2 5 12 13
¬ 5 + 13 + 7 = 25
4 3 7 24 25
¬ 7 + 25 + 9 = 41
5 4 9 40 41
¬ 9 + 41 + 11 = 61
6 5 11 60 61
Thus, we see that for any two successive triples (a1, b1, c1) and (a2, b2, c2) in Table 5,
a1 + c1 + a2 = c2. (9)
We can easily prove (9) if (8) is true, since
a1 + c1 + a2 = a1 + (b1 + 1) + a2
= (a1 + b1 + a2) + 1
= b2 + 1 substituting from (8)
= c2.
Thus, a1 + c1 + a2 = c2. (10)
But, is (8) true? To prove (8), we shall try to prove that if R = (a1, b1, c1) is any given complete triple in a list of complete triples sorted in ascending order by their a’s, and S = (a2, b2, c2) is the next succeeding triple in this list, then a1 + b1 + a2 = b2.
Let m1, n1 and m2, n2 be the m and n values used to generate the complete triples R = (a1, b1, c1), and
S = (a2, b2, c2), respectively. Then, since R and S are both complete, we know that
m1 = n1 + 1 and m2 = n2 + 1,
or n1 = m1 - 1 and n2 = m2 - 1. (11)
First, let’s carry out some preliminary steps.
From (3) above, we have a = 2n+1,
so a1 = 2n1 + 1 and a2 = 2n2 + 1. (12a)
Using (11), we have a1 = 2 (m1 – 1) + 1 = 2m1 – 1,
and a2 = 2(m2 – 1) + 1 = 2m2 – 1. (12b)
Also from (3) above, we know that the a’s form a set of increasing consecutive odd integers,
so a2 = a1 + 2, (13)
and substituting from (12a), we get 2n2 + 1 = (2n1 + 1) + 2,
or 2n2 = 2n1 + 2,
so n2 = n1 + 1 or n1 = n2 - 1 (14)
Again, from (13), we have a2 = a1 + 2,
and substituting from (12b), we get 2m2 – 1 = (2m1 - 1) + 2,
2m2 = 2m1 + 2,
and m2 = m1 + 1 or m1 = m2 - 1 (15)
Using these preliminary results, we shall now show that a1 + b1 + a2 = b2. Starting with the left-hand side of this equation, we have a1 + b1 + a2
= m12 – n12 + 2m1n1 + m22 – n22 (Theorem 5)
= (m1 + n1)(m1 – n1) + 2m1n1 + (m2 + n2)(m2 – n2) (factoring)
= (m1 + n1)(1) + 2m1n1 + (m2 + n2)(1) [ By Theorem 28, m = n + 1, so m – n = 1. Thus,
m1 – n1 = 1 and m2 – n2 = 1. ]
= m1 + n1 + 2m1n1 + m2 + n2
= (m2 – 1) + (n2 – 1) + 2(m2 – 1)(n2 – 1) + m2 + n2 [ from (15) and (14) ]
= m2 – 1 + n2 – 1 + 2m2n2 – 2m2 – 2n2 + 2 + m2 + n2
= 2m2n2 = b2
So, a1 + b1 + a2 = b2, and we have done it! This proof, together with (10) above, gives us
Theorem 38
If (a1, b1, c1) is any given complete triple in a list of complete triples sorted in ascending
order by their a-values, and (a2, b2, c2) is the next succeeding triple in this list, then
a1 + b1 + a2 = b2 and a1 + c1 + a2 = c2.
______________________________________________________________________
Theorem 38 makes it very easy to construct a table of complete triples. To illustrate this, let’s suppose we have the complete triple (5,12,13) and wish to write out quickly and easily the next two successive triples in a list of complete triples sorted in ascending order by their values of a. We simply write down in a row the three values of a, b, and c from the starting triple: namely, 5, 12, and 13, and write three blanks underneath them for the values of the next triple. (See table A below.) Then, we place a 7 for the new a-value under the a-value of 5 since the values of a are increasing consecutively by 2. (See table B below.) Next, we write the sum of the two a-values (5 and 7) and the b-value (12), giving us 5 + 7 + 12 = 24 for the new b-value, as shown in Table C. Lastly, we add 1 to this new b-value and place this number as the new c-value as shown in table D, since c = b + 1. Thus, (7, 24, 25) is the next successive complete triple in our list.
5 12 13 5 12 13 5 12 13 5 12 13
_ _ _ → 7 _ _ → 7 24 _ → 7 24 25
A B C D
Continuing in this manner to get the next triple, we get
5 12 13 5 12 13 5 12 13 5 12 13
7 24 25 → 7 24 25 → 7 24 25 → 7 24 25
_ _ _ 9 _ _ 9 40 _ 9 40 41
E F G H
and so (9, 40, 41) is the second successive complete triple after (5, 12, 13). The third successive complete triple after (5, 12, 13) is (11, 60, 61), obtained as shown below.
5 12 13 5 12 13 5 12 13 5 12 13
7 24 25 7 24 25 7 24 25 7 24 25
9 40 41 → 9 40 41 → 9 40 41 → 9 40 41
_ _ _ 11 _ _ 11 60 _ 11 60 61
More Patterns?
Let’s use the technique of Theorem 38 to construct a larger table of complete Pythagorean triples sorted in ascending order by their a-values to see if there are any patterns we have missed.
Table 6 (Complete Triples)
m n a b c
-----------------------------------------
2 1 3 4 5
3 2 5 12 13
4 3 7 24 25
5 4 9 40 41
6 5 11 60 61
7 6 13 84 85
8 7 15 112 113
9 8 17 144 145
10 9 19 180 181
11 10 21 220 221
12 11 23 264 265
13 12 25 312 313
Notice in Table 6 that the value of n tells us the number of the triple. For example, (11,60,61) is the fifth triple
in Table 6, and 5 is its value of n. The twelfth triple in the table, (25,312,313), has12 as its value of n. In fact, we shall
call the value of n associated with the nth triple in our list, the number of the triple. It’s easy to find n for the nth
triple (a, b, c) if we know the value of a. Since a = 2n +1, we can find n by solving this equation for n, getting
n = (a- 1)/2. So, we have
The number of the nth complete triple (a, b, c) in a list of complete triples sorted in ascending order
by their a-values is its value of n. It can be obtained by using the equation n = (a –1) / 2.
____________________________________________________________________________________
Note also that for any given triple in Table 6, the m and n values for the next triple are each one greater than those of the given triple, and the m and n values of the preceding triple are each one less than those of the given triple. Thus, for the triple (an, bn, cn) with m and n values of m and n, the next triple in order, (an+1, bn+1, cn+1),
will have m and n values of m+1 and n+1, respectively, and the preceding triple, (an-1, bn-1, cn-1), will have m and n values of m - 1 and n - 1, respectively. This gives us
(an - 1, bn - 1, cn - 1), where an - 1 = 2(n-1) + 1,
bn - 1 = 2(m-1)(n-1), and
cn - 1 = (m-1)2 + (n-1)2,
(an, bn, cn), where an = 2n+1,
bn = 2mn, and
cn = m2 + n2,
(an + 1, bn + 1, cn + 1), where an + 1 = 2(n+1) + 1,
bn + 1 = 2(m+1)(n+1), and
cn + 1 = (m+1)2 + (n+1)2.
Now, in Table 6 there are several very interesting patterns hidden in this maze of seemingly unrelated data. Can you find any of them? Before reading on, why not see if you can find some. It’s well worth the effort and lots of fun if you do! (I’ll wait.)
- - - - - - - - - - - - - - - - - - - - - - - - - -
Did you find any patterns? One I was fortunate enough to discover is this:
Add any two consecutive b-values in Table 6, such as the first two: 4 + 12 = 16, or the second and third ones: 12 + 24 = 36, or the fifth and sixth ones: 60 + 84 = 144. In each case, the sum is a perfect square! Check it out for yourself - the sum of any two consecutive values of b from Table 6 is a perfect square. We have
b1 + b2 = 4 + 12 = 16 = 42 = (2x2)2
b2 + b3 = 12 + 24 = 36 = 62 = (2x3)2
b3 + b4 = 24 + 40 = 64 = 82 = (2x4)2
b4 + b5 = 40 + 60 = 100 = 102 = (2x5)2, and in general,
bn - 1 + bn = (2n)2 for n = 2, 3, 4, . . .
But that is not all! Let’s add two consecutive a-values and compare their sum to the corresponding sum of the b-values above. We get
a1 + a2 = 3 + 5 = 8 and 2(a1 + a2) = 2(8) = 16 = b1 + b2,
a2 + a3 = 5 + 7 = 12, and 3(a2 + a3) = 3(12) = 36 = b2 + b3,
a3 + a4 = 7 + 9 = 16, and 4(a3 + a4) = 4(16) = 64 = b3 + b4,
a4 + a5 = 9 + 11 = 20 and 5(a4 + a5) = 5(20) = 100 = b4 + b5, and in general n(an - 1 + an) = bn - 1 + bn = (2n)2 for n = 2, 3, 4, . . .
Here is a proof.
Let (an, bn, cn) be the nth triple and (an-1, bn-1, cn-1) be the (n-1)st triple in a list of complete triples
sorted in ascending order by their a’s. Then, since in general, a = 2n + 1 and b = 2mn, we have
an - 1 = 2(n-1) + 1 = 2n – 1, an = 2n + 1, bn - 1 = 2(m – 1)(n – 1), and bn = 2mn. Thus,
1. n (an - 1 + an ) = n [ (2n - 1) + (2n + 1) ] = n (4n) = 4n2 = (2n)2. (16)
2. bn - 1 + bn = 2(m – 1)(n – 1) + 2mn = 2(n)(n– 1) + 2(n +1)n
= 2(n2 - n + n2 + n) = 2(2n2) = 4n2 = (2n)2. (17)
Since (16) and (17) are identical, we have thus proved
Theorem 39
If (an, bn, cn) is the nth complete triple and (an - 1, bn - 1, cn - 1) is the preceding triple in
a list of complete triples sorted in ascending order by their a-values, then
bn - 1 + bn = n(an - 1 + an ) = (2n)2, for n = 2, 3, 4, . . .
___________________________________________________________________
A very nice theorem!
Let’s list below some other patterns that I have found and which you also might have discovered.
Each pattern applies to a set of two successive complete triples, either (an, bn, cn) and its successor
(an +1, bn + 1, cn + 1), or its predecessor (an - 1, bn - 1, cn - 1), and is listed first and then illustrated using
(9, 40, 41) and (11,60,61), the 4th and 5th triples from Table 6 above.
First, we get the sums, the first two of which we have already proved (Theorem 38).
The Sums:
1. an + bn + an+1 = bn+1 a4 b4 c4
9 40 41
a4 + b4 + a5 = b5
a5 b5 c5
9 + 40 + 11 = 60 11 60 61
2. an + cn + an+1 = cn+1 a4 b4 c4
9 40 41
a4 + c4 + a5 = c5
a5 b5 c5
9 + 41 + 11 = 61 11 60 61
3. bn + cn+1 = bn+1 + cn a4 b4 c4
9 40 41
b4 + c5 = b5 + c4
a5 b5 c5
40 + 61 = 60 + 41 11 60 61
4. cn + cn+1 = bn + bn+1 + 2 41 + 61 = 40 + 60 + 2
5. bn + bn+1 = (an+1 – 1)2 40 + 60 = (11 – 1)2
= (an + 1)2 40 + 60 = (9 + 1)2
= [(an + an+1)/2]2 40 + 60 = [ (9 + 11)/2 ]2
The Differences:
1. For any two consecutive complete triples, the difference between their c-values equals the difference between their b-values, which equals the sum of their a-values.
cn+1 – cn = bn+1 – bn = an+1 + an
= 2(an+1 – 1) = 2(an + 1)
a4 b4 c4
9 40 41
a5 b5 c5
11 60 61
c5 – c4 = b5 – b4 = a5 + a4
61 – 41 = 60 – 40 = 9 + 11
= 2(a5 – 1) = 2(a4 + 1)
= 2(11 – 1) = 2(9 + 1) = 20
2. The Sum – Product (Theorem 39): For any two consecutive complete triples, the sum of their b-values equals the sum of their a-values multiplied by the number of the higher triple.
bn - 1 + bn = n ( an - 1 + an ) = (2n)2
m n a b c
--------------------------------------
2 1 3 4 5
3 2 5 12 13
4 3 7 24 25
5 4 9 40 41
6 5 11 60 61
b1 + b2 = 2(a1 + a2) 4 + 12 = 2(3 + 5) = (2x2)2 = 16
b2 + b3 = 3(a2 + a3) 12 + 24 = 3(5 + 7) = (2x3)2 = 36
b3 + b4 = 4(a3 + a4) 24 + 40 = 4(7 + 9) = (2x4)2 = 64
b4 + b5 = 5(a4 + a5) 40 + 60 = 5(9 +11) = (2x5)2 = 100
The Cross-Products:
Consider now the cross product of the a and b values for any two consecutive complete triples by looking at the first two triples below. Their cross-product is 3(12) – 4(5) = 16 = 42, a perfect square! But that is not all. Add their two b values (4 and 12) together. This sum is equal to the cross-product! For the second and third triples, we have a cross-product of 5(24) – 7(12) = 36 = 62, another perfect square! And the sum of their two b values (12 + 24) also equals their cross-product. The same is true for the third and fourth triples. There’s certainly something going on here!
m n a b c anbn + 1 – an + 1bn bn + bn+1
-------------------------------------------------------------------------------------------------------------
2 1 3 4 5
3(12) – 5(4) = 36 – 20 = 16 = 4+12 = 42
3 2 5 12 13
5(24) – 7(12) = 120 – 84 = 36 = 12+ 24 = 62
4 3 7 24 25
7(40) – 9(24) = 280 – 216 = 64 = 24 + 40 = 82
5 4 9 40 41
In general, it appears that
an bn + 1 – an + 1 bn = bn + bn+1 , for n = 0, 1, 2, 3, . . .
Let’s try to prove this. Accordingly,
an bn + 1 – an + 1 bn = (2n+1)[ 2(m+1)( n+1)] – [ 2(n+1) + 1] (2mn)
= 2 [ (2n+1)(m+1)( n+1) – 2mn( n+1) – mn ]
= 2 [(2n+1)( mn + m + n + 1) – 2mn2 – 2mn – mn]
= 2[2mn2 + 2mn + 2n2 + 2n + mn + m + n + 1 – 2mn2 – 3mn]
= 2 [2n2 + 3n + m + 1] = 2 [2n2 + 3n + (n+1)+1]
= 2 [2n2 + 4n + 2] = 4(n2 +2n + 1) = 4(n+1)2. (18)
bn + bn + 1 = 2mn +2( m+1)( n+1) = 2(mn + mn + m + n+ 1)
= 2[2mn + m + (n+1)] = 2[2mn + m + m] = 2[2mn + 2m]
= 4 [mn + m] = 4 [(n+1)n + (n+1)] = 4 [n2 +2n + 1] = 4(n+1)2 (19)
Since (18) and (19) are identical, we have proved
1. an bn + 1 – an + 1 bn = bn + bn+1 = 4(n + 1)2.
We can also show that an bn + 1 – an + 1 bn = (n+1)(cn – cn+1) = (an + 1)2
= (an+1 – 1)2 = [ (an + an+1) / 2 ] 2
a4 b4 c4
9 40 41
a5 b5 c5
11 60 61
a4b5 – a5b4 = 9(60) – 11(40) = 100
b4 + b5 = 40 + 60 = 4(4+1)2
( n+1)( c5 – c4 ) = 5(61 – 41)
(a4 + 1)2 = (9 + 1)2
(a5 – 1)2 = (11 – 1)2
[ (a4 + a5) / 2 ] 2 = [(9 + 11)/2] 2 = 100
Here are two more cross-product formulas, one for the a and c values; the other for the b and c values.
2. ancn+1 – an+1cn = (an + 1)2 – 2
________
3. cnbn+1 – bncn+1 = bn+1 – bn = cn+1 – cn = 2(1 + Ö (1 + 2bn ) = 2(an + 1)
Those proofs which are not shown for some of the patterns above can easily be proved by simply rewriting the formulas in terms of m’s and n’s, using a = 2n+1, b = 2mn, c= m2 + n2, and m = n+1.
Well, we’re done here for now, and as we can clearly see, there are many patterns to be found. As Ecclesiastes the Preacher has truly said:
. . . of (the) making (of) many books there is no end, and much study is a weariness of the flesh.
Ecclesiastes 12:12
Perhaps we should paraphrase this as:
. . . of (the) making (of) many formulas there is no end, and so many patterns are a weariness to the flesh.
Be that as it may, we can certainly say that the field of primitive right triangles and Pythagorean triples is rich in patterns.Perhaps with further exploration and study, we may find other fine patterns whose beauty it will be a joy to experience. Until then, I hope that you have enjoyed reading about the many beautiful patterns we have found here as much as I have enjoyed writing about them. Let me close with the following observation by the brilliant number-theorist Godfrey Harold Hardy.
I believe that mathematical reality lies outside us, that our function is to discover or observe it,
and that the theorems which we prove, and which we describe grandiloquently as our "creations,"
are simply the notes of our observations.
A Mathematician's Apology ( London 1941 )