Patterns in Pythagoras
 
3B. Even More Patterns in Right Triangles
andPythagorean Triples
 
 David W. Hansen
© 2008
 
4. Unit and Constant Difference Triples
 
     A unit primitive Pythagorean triple? Just what is that? Well, as the name suggests, it is a primitive Pythagorean triple (a, b, c), where the difference between a and b is the unit 1. (3,4,5) is a unit Pythagorean triple because b - a = 4 - 3 = 1.   (21,20,29) is a unit triple because a - b = 21 - 20   = 1. Unit triples can be regular (a < b) as in (3,4,5) or inverted (a > b) as in (21,20,29). If |a – b| = 1, then we call the primitive Pythagorean triple (a, b, c) a unit triple. Are there unit triples other than the two just mentioned? Let’s see.
 
     A.  If b = a+1, then, the triple (a, a+1, c) is a regular unit triple, and a2 + (a+1)2   =   c2. Solving for a, we get                                                                                                                                     _______    
              a2 + a2 + 2a + 1 = c2,         2a2 + 2a + (1 - c2) = 0,       and          a = ( - 1 ±  Ö 2c2 -1 ) / 2.
 
     B.  If b = a -1, then, the triple (a, a -1, c) is an inverted unit triple, and a2 + (a-1)2 = c2. Solving for a, we
get                                                                                                                                    _______
              a2 + a2 - 2a + 1 = c2,         2a2 - 2a + (1 - c2) = 0,        and          a = ( 1 ±    Ö  2c2 -1 ) / 2.
 
   Now a must be an integer, so in both cases A and B above,  2c2 –1  must be a perfect square in order for its square root to be an integer. Furthermore, since c is odd, then c2 is odd,  2c2 is even,  2c2 – 1 is odd,
  ______                        ______                                                                                ______
Ö 2c2 - 1  is odd, 1 ±   Ö 2c2 -1  is even and thus divisible by 2. So, a  =  ( 1  ±    Ö 2c2 -1 ) / 2 is an integer. 
 
Thus, to find all unit triples, we simply need to find all those values of c for which 2c2 – 1 is a perfect square. 
 
     Let’s set up a table of the permissible values of c written in ascending order, together with their corresponding
values of 2c2 - 1, and if these are perfect squares, their square roots, the values of a, b, and c, and the values of m and n.
 
                                                                                             Table 3
                                                                                      ______
                                                    c       2c2 -1            Ö  2c2 - 1             a      b      c             m      n  
                                              -------------------------------------------------------------------------------------------                       
                                                    5           49                    7                   3      4     5              2      1
                                                  13         337                    -
                                                  17         577                    -
                                                  25       1249                    -
                                                  29       1681                  41                 21    20   29             5      2
                                                  37       2737                    -
                                                  41       3361                    -
                                                  53       5617                    -
 
                                                        .   .   .                     .   .   .                    .   .   .                    .   .   .
 
                                                 157     49297                  -
                                                 169     57121                239             119  120   169          12      5
                                                 173     59857                  -
                                                 181     65521                  -
                                             ---------------------------------------------------------------------------------------------
     
     Now, as seen in Table 3, we have calculated the values of 2c2 - 1 for all permissible values of c up to 181 and found just one new unit triple. Are there others? If so, can we find a simpler and more pleasing way to find them with less computation involved? Let’s look for a pattern in Table 3 that might help us.
 
     Writing down the three unit triples we have found and their corresponding values of m and n in the table below, we get
                                                                           (a, b, c)               m      n
                                                                  -------------------------------------------
                                                                           (3, 4, 5)               2       1
                                                                        (21, 20, 29)           5       2
                                                                     (119, 120, 169)     12       5
 
What can we see? Well, the m value of the first unit triple (2) is the same as the n value of the second triple, and the m value of the second unit triple (5) is the same as the n value of the third unit triple! If this pattern continues, then the m value of the third unit triple (12) will be the same as the n value of the fourth unit triple, as shown below.
                                                                          (a, b, c)              m        n
                                                              ---------------------------------------------------
                                                                          (3, 4, 5)                2         1
                                                                      (21, 20, 29)             5         2
                                                                  (119, 120, 169)        12         5
                                                                         (?, ?, ?)               m?      12
 
But what will the m value of the fourth unit triple be?
     If we look closely at the values of m and n for the first two unit triples as shown below,
 
                                                                     Unit triple 1               2          1                                       
                                                                                      
                                                                     Unit triple 2              5           2
 
we see that the sum of the three bold-faced values , 2 + 1 + 2, equals the underlined value, 5. Thus, the m value of the second unit triple equals the combined sum of the m and n values of the first unit triple and the n value of the second unit triple.
 
     Similarly, looking at the values of m and n for the second and third unit triples,
 
                                                                     Unit triple 2                 5          2
                                           
                                             
                                                                     Unit triple 3               12          5
 

we see that the sum of the three bold-faced values, 5 + 2 + 5, equals the underlined value, 12. Thus, the m value of the third unit triple equals the sum of the m and n values of the second unit triple and the n value of the third unit triple.
 
     Now if the pattern continues, and looking at the values of m and n for the third and fourth unit triples as
shown below, we surmise that the m value for the fourth unit triple will equal the combined sum of the m and n values of the third unit triple and the n value of the fourth unit triple; namely, 12 + 5 + 12 = 29.
 
                                                                      Unit triple 3                12          5
                                         
                                                                      Unit triple 4                m?       12
 
Thus, the m value for the fourth unit triple would be 29. Let’s check this.
 
     Using m = 29, n = 12, will we obtain a unit Pythagorean triple?   Let’s see!    a  =  292  -  122  =  697,  
b  =  2(29)(12)  =  696, and c  =  292 + 122  =  985, and we see that indeed  (697, 696, 985)  is the fourth unit triple, since a  -  b  =  1. How VERY nice!  
 
     Adding this unit triple to our list of unit triples, we get
 
                                                                           (a, b, c)                m      n
                                                               -----------------------------------------------                  
                                                                           (3, 4, 5)                 2       1
                                                                        (21, 20, 29)              5       2
                                                                     (119, 120, 169)         12       5
                                                                     (697, 696, 985)         29     12
 
Notice that the first unit triple is regular, the second triple is inverted, the third one is regular, and the fourth one is inverted. Does this pattern continue for all unit triples? Are there more unit triples? Are there an infinite number of unit triples? We surely hope so!
 
     Now what have we found so far? It seems that when we have a list of unit triples sorted in ascending order by their c values, then the m value of a given unit triple U equals the n value of the next unit triple, the combined sum of the m and n values of U and the n value of the next unit triple equals the m value of the next unit triple, and the unit triples alternate between being regular and inverted triples.
 
     In general, in a list of unit triples sorted in ascending order by their c values, if U1 is a unit triple with
m = m1, n = n1, and U2 is the next unit triple with m = m2, n = n2, then the list looks like this:
 
                                                                          (a, b, c)              m       n
                                                               ----------------------------------------------
                                                                             .  .  .               . . .      . . .
 
                                                                               U1                m1       n1
 
                                                                               U2                m2       n2
 
                                                                              .  .  .               . . .     . . .
                                                                -----------------------------------------------       
    
and we surmise that      n2   = m1    and    m2   = m1 + n1 + n2 ,  or, since  n2  =  m1, then   m2  =  2m1 + n1,        making the list now look like this: 
                                                                           (a, b, c)              m              n
                                                                  ------------------------------------------------------
                                                                               . . .                 . . .            . . .
 
                                                                                U1                 m1              n1
 
                                                                                U2              2m1+n1         m1
 
                                                                               . . .                   . . .             . . .
                                                                   ----------------------------------------------
 
Let’s see if we can prove this conjecture. Accordingly, we let U1 be any regular unit triple (a1, b1, c1) in a list of unit triples sorted in ascending order by their c values, with m = m1,    n = n1. Since U1 is a regular unit triple, we know that           
                                                                           b1     =         a1          +      1,
or                                                                  2m1n1    =    m12 – n12   +     1.                                                     (1)
and multiplying (1) by -1, we have            - 2m1n1    =     n12 – m12   –     1                                                      (2)
 
Next, we let U2 be the next triple (a2, b2, c2) in the list with its m and n values given by m = 2m1 + n1 and n = m1. Will U2 be a unit triple for these values of m and n? To show that it is, we must show that either b2 = a2 + 1 or
b2 = a2  - 1. 
 
Now,                                               b2  =  2mn  =  2(2m1 + n1)m1  =  4m12 + 2m1n1,                                       (3)
 
and                                             a2 - 1  =  m2 - n2 - 1  =  (2m1 + n1)2  - m12 - 1
                                                             =  4m12 + 4m1n1 + n12 - m12 - 1
                                                             =  4m12 + 4m1n1 + (n12 - m12 - 1)
                                                             =  4m12 + 4m1n1 + ( - 2m1n1),  (substituting from (2) above)
                                                             =  4m12 + 2m1n1, which is (3) above.
 
Thus, b2  =  a2  -  1, and U2 is a unit triple. Furthermore, since b2 < a2, it is inverted !
 
     If U1 had been an inverted unit triple  (b1 =  a1 - 1), our proof would have shown that U2, the next triple in the list, would be a regular unit triple  (b2 =  a2 + 1). Thus, in a list of unit triples sorted in  ascending order by their c values, the triples will alternate between regular and inverted triples, and since (3,4,5), the first unit triple, is regular, the list will begin with a regular unit triple.
     Note that for any given unit triple, we can always find the next one, so there are an infinite number of unit triples!   All of this gives us the following theorem.
 
           Theorem 29
 
           If U is any unit triple in a list of unit triples sorted in ascending order by their c values,
           with m = m1, and n = n1, then the next unit triple in this list has m and n values given by
           m = 2m1 + n1 and n = m1. If U is regular, the next triple will be inverted. If U is inverted,
           the next triple will be regular. There are an infinite number of unit triples, half of which are
           regular and half of which are inverted.                                                                                   
           ________________________________________________________________________
 
 
     To use Theorem 29, we must be able to find the values of m and n for any given unit triple (a, b, c). Here’s how.  Since   c = m2 + n2   and   a = m2 - n2,   then
 
                                         c + a   = (m2 + n2) + (m2 - n2)   =   2m2,   or    m2 = (c + a) /2.
and,                                  c - a  = (m2 + n2) - (m2 - n2)    =    2n2,   or     n2 = (c - a) /2.
                                                                                                 _______                  ______
So, for a primitive Pythagorean triple (a, b, c),   m = Ö (c+a)/2 and n = Ö (c-a)/2 .                             (4)
 
Example 1. Find the next unit triple after (697, 696, 985). To use Theorem 29, we must first find the values of m and n for (697, 696, 985). Using (4) above, we get
                                                     ___________                                                   __________
                              m  =  m1  =  Ö (985+697)/2    =  29      and       n  =   n1  =  Ö (985-697)/2   =   12,
 
which confirm the values shown in the table on page 19. Now, using Theorem 29, we have for the next unit triple,
m  =  2m1 + n1  =  2(29)  + 12  =  70, and n  =  m1  =  29. 
    To be pictorial, we could have used the following tables (which are more fun):
 
 
                     29    12                         29     12                          29            12                        29       12
                                         →                                 →                                           →
                     ?     ?                           ?      29                     (29+12+29)     29                       70       29
 
These values for m and n give us a  =  702 – 292  =  4059, b = 2mn  =  2(70)(29)  =  4060, and c = 702 +292
=  5741, thus generating (4059, 4060, 5741), the next unit triple after (697, 696, 985). Note that this new unit triple is regular, since (697, 696, 985) was inverted.       
 
Example 2U = (137903, 137904, 195025) is a unit triple. For the unit triple just preceding U, a) determine if it is regular or inverted, and b) find this triple.
 
     a) Since U is a regular unit triple, (a < b), then by Theorem 29, the unit triple preceding U must be inverted
         (b > a).
 
     b) We must first find m and n for U. Thus, using (4) above, we get for U,
                                       ________________                                            ________________
                           m  =  Ö(195025+137903)/2   =  408,       and       n  =  Ö(195025+137903)/2   =  169.
 
Now. to find the preceding unit triple, we must “reverse” the procedure in Theorem 29.
 
     Pictorially, we can use the following tables to work backwards to find the m1 and n1 values for the preceding unit triple. As shown in the tables below, we put m = 408 and n = 169 in row 2 of Table A. Then, since m1 = n, we put 169 as the first entry in row 1, giving us Table B. Since the sum of both the entries in row 1 and the second entry in row 2 must equal the first entry in row 2, we subtract the sum of the first entry in row 1 and the second entry in row 2 to find the second entry in row 1 (Table C below). This gives us Table D below, which tells us  m1 = 169,  n1 = 70.
 
                         Table A                        Table B                               Table C                               Table D
                         ?     ?                       169      ?                     169    (408 -169 -169)                  169     70 
                                         →                                  →                                               →  
                         408 169                       408    169                    408             169                          408   169  
 
     Algebraically, using Theorem 29, we can find the m1 and n1 of the preceding unit triple by solving the equations   m  = 2m1 + n1   and    n = m1   for m1 and n1. Substituting the values of m  =  408 and n  =  169 into these two equations, we get
                                                                               408 = 2m1 + n1                                                                  (5)
 
and                                                                          169 = m1.                                                                          
 
Thus, m1 = 169, and substituting this value of m1 into (5), we get      408 = 2(169) + n1,   or    n1 = 70.        
 
      Either pictorially or algebraically, m1  =  169  and  n1  =  70, giving us  a  =  1692  -  702   =   23661,  b  =  2mn
=  2(169)(70)  =  23660, and c  =  1692 + 702  = 3 3461, thus generating (23661 23660, 33461), the inverted unit triple (23661 > 23660) which precedes (137903, 137904, 195025).
 
     Adding the information and results found in examples 1 and 2 to the table of unit triples above, we get Table 4 below, which lists the first seven unit triples together with their values of m and n.
 
                                                                        Table 4  (Unit Triples)
                                                               (a, b, c)                              m      n
                                                        -----------------------------------------------------------
                                                               (3, 4, 5)                              2       1
 
                                                            (21, 20, 29)                           5       2
 
                                                         (119, 120, 169)                      12       5
 
                                                         (697, 696, 985)                      29      12
 
                                                      (4059, 4060, 5741)                   70      29
 
                                                   (23661, 23660, 33461)              169      70
 
                                               (137903, 137904, 195025)            408    169
 
          Now, does there exist any triple in which the difference between a and b is 1 and the difference between b and c is also 1? If so, we shall call that triple a pure unit triple. Let’s try to find some.
 
    For a Pythagorean triple (a, b, c), if a and b differ by 1, then b = a + 1, and if b and c differ by 1, then c = b + 1
= (a + 1) + 1 = a + 2. Then, substituting for b and c in  a2   +   b2    =    c2,
we have
                                                                 a2    +   (a + 1)2     =      (a + 2)2,
                                                                 a2 + a2 + 2a + 1    =    a2 + 4a + 4,
and simplifying, we get                          a2 – 2a – 3  =  (a – 3)(a + 1)  =  0.
 
Solving this last equation, we have a  =  3 or - 1.  Since a > 0, we must pick a = 3, giving us a = 3,  b = 3 + 1 = 4, and c  =  3 + 2  =  5. Thus, (3,4,5) is a pure unit triple, and, in fact, the only pure unit triple!                                    
 
Constant Differences between a, b, and c
 
     Now, we have just finished looking for and finding patterns in those primitive Pythagorean triples where the difference between the a and b values is always 1. Let’s now look at those triples where the difference between their c and b values, their c and a values, and their and b values is always a constant other than 1.
 
The difference between c and b   Since  c – b   =  (m2 + n2) – 2mn = (m – n)2, then c – b is always a perfect square. Furthermore, since c is always odd and b is always even, then
 
                                                           c – b is always an odd perfect square.
 
Let c – b = d2, where d is some given odd positive integer. Then, we have
 
                                             c  –  b   =   (m – n)2  =   d2,    or    m  –  n   =   d.
 
 Thus,   m = n + d,    and substituting this into the generating formulas for a, b, and c, we get
 
                                               a   =   m2 – n2   =   (n + d)2 –   n2     =    2nd + d2
                                               b   =      2mn      =        2(n + d)n       =    2n(n + d)
                                               c   =   m2 + n2                                   =    (n + d)2 + n2
 
Now, n must not be a multiple of d, for if n = fd for some integer f, then substituting this into the above equations, we get
 
                            a   =   2d(fd) + d2   =   d2(2f + 1),           b   =   2fd(fd + d)   =   d2(2f2 + 2f),
 
and                      c   =   (fd + d)2 + (fd)2   =   f2d2 + 2fd2 + d2 + f2d2    =   d2(2f2 + 2f + 1).
 
Thus, a, b, and c will all have a common factor of d2, and (a, b, c) will not be a primitive Pythagorean triple
This gives us
 
         Theorem 30 (The difference between c and b)
 
         If (a, b, c) is a primitive Pythagorean triple, then c – b is always an odd perfect square, say d2.
         All primitive Pythagorean triples (a, b, c) for which c – b = d2, where d is an odd positive integer,
         are given by   a = 2dn + d2,   b = 2n(n + d),   and   c = (n + d)2 + n2, where n is any positive
         integer not a multiple of d
         ______________________________________________________________________________
 
Example 3Find formulas for all primitive Pythagorean triples (a, b, c) for which the difference between c and b is 9. We have c – b = 9 = 32 = d2. Thus, d = 3, and from Theorem 30, we get
 
                               a   =   2(3)n + 32   =   6n + 9,        b   =   2n(n + 3),          c   =   (n + 3)2 + n2
 
for n a positive integer not a multiple of 3. Below is a table of these triples for n = 1, 2, 4, 5, 7, and 8. Note that n ≠ 3, 6, 9, . . . , which are multiples of 3.
 
                       n         a       b       c            c – b                      n         a       b       c            c – b
                 -------------------------------------------------------------------------------------------------------------------
                       1       15       8     17       17 –   8 = 9                5        39     80     89       89  –  80 = 9
                       2       21     20     29       29 – 20 = 9                7        51   140   149     149 – 100 = 9
                       4       33     56     65       65 – 56 = 9                8        57   176   185     185 – 176 = 9
 
Example 4Are there any primitive Pythagorean triples (a, b, c) for which the difference between c and b is 13, 36, 49, or 123?   Since 13, 36, and 123 are not odd squares, there are no primitive triples for these values. However, 49 = 72 is an odd square with d = 7. So, a = 2(7)n + 72 = 14n + 49, b = 2n(n + 7), and  c = (n + 7)2 + n2, and the triples (14n + 49, 2n(n + 7), (n + 7)2 + n2)  for n a positive integer not a multiple of 7, all have a difference of 49 between their b and c values. This is so because
 
                c  –  b    =   (n + 7)2 + n2 – 2n(n + 7)   =   n2 + 14n + 49 + n2   –  2n2  – 14n   =   49 √    
 
The difference between c and a    Since c – a  =  (m2 + n2) – (m2 – n2)  =  2n2, then
 
                                                     c – a is always twice a perfect square.
 
 
                      Here is a table showing the possible values of c – a for n  =  1, 2, 3, . . . , 15.
 
            n              1      2      3      4       5       6       7        8       9       10     11      12      13      14      15
       ------------------------------------------------------------------------------------------------------------------------------------
         c – a           2      8     18    32     50     72     98     128   162    200    242    288    338    392    450   
 
 
     Now, m and n are of opposite parity (one even, one odd), so m – n is always odd; that is, m – n = 2k - 1
for some positive integer k. Solving this for m, we get m = n + 2k – 1 and substituting this for m in the generating formulas for a, b, and c, we have
 
                                          a    =   m2  –  n2                                   =   (n + 2k – 1)2  –  n2                                   (6)
                                          b    =    2mn           =   2(n + 2k – 1)     =        2n(n + 2k – 1)                                    (7a)
                                          c    =   m2 + n2                                     =     (n + 2k – 1)2  +  n                               (7b)
 
Now, a, b, and c must have no common factors. Let p be a prime factor of n. Then, n = pg for some positive integer g, and substituting this value of n into (7a), we see that
 
                                         b  =  2n(n + 2k – 1)  =  2pg(n + 2k – 1)   =   p[2g(n + 2k – 1)],  
      
which shows that b has a factor of p.                                 Furthermore, substituting this value of n into (6), we get
 
                                          a   =   (n + 2k – 1)2 – n2     =     [pg  +  (2k – 1)]2 – (pg)2,
 
and looking at this last expression, we see that if 2k – 1 is a multiple of p (that is, 2k – 1  =  ph for some positive integer h), then
                                       a   =    [pg + (2k – 1)]2 – (pg)2   =   [pg + ph]2 – (pg)2
 
                                            =    [p(g + h) ]2 – (pg)         =     p2(g + h)2   –   p2g2
                                    
                                            =      p [ p(g + h2 – pg2 ],          which shows that a has a factor of p.
 
Thus, 2k – 1 cannot be a multiple of a prime factor p of n, or else a and b will have a common factor of p, and
(a, b, c) will not be primitive.
 
     Now, if 2k – 1 = ph for some integer h, then both p and h must be odd, since 2k – 1 is odd. Rewriting
 
2k – 1 =   ph as   2k  =   ph + 1   =   ph + p + 1 – p    =    p(h + 1) – (p – 1),
 
we get                                                     k   =   p[(h+1)/2] – (p – 1)/2.
 
Now, p and h are both odd, so (h+1)/2 is an integer, say w, as is (p – 1)/2 also. This gives us
 
                                                      k   =   pw – (p – 1)/2,    for   w = 1, 2, 3, . . .
 
Thus, if (a, b, c) is to be primitive, then k must not be equal to pw – (p – 1)/2, in (6), (7a), and (7b), giving us
 
         Theorem 31  (The difference between c and a)
 
          If (a, b, c) is a primitive Pythagorean triple, then c – a is always twice a perfect square, namely, 2n2,
          and
                    a = (n + 2k – 1)2  –  n2,           b = 2n(n + 2k – 1),            c = (n + 2k – 1)2  + n2,
 
          where n and k are any positive integers with k ≠ pw – (p – 1)/2 for p an odd prime factor of n,
          and w = 1, 2, 3, . . .
           _______________________________________________________________________________
 
 
Example 5Find formulas for all primitive Pythagorean triples (a, b, c) for which the difference between c and a is 98. Using Theorem 31, we know that c – a   = 2n2 = 98. Thus, n2 = 98, n = 7, and p = 7. Then,
 
                          a   =   (2k + 6)2 – 49,        b   =   14(2k + 6),          c   =   (2k + 6)2 + 49
 
                             for k ≠   7w – 3, w = 1, 2, 3,  . . . , that is,    k ≠ 4, 11, 18, 25, . . .     
 
 
Below is a table of these triples for k = 1, 2, 3, 5, 6, 7, 8, and 9.
 
                          k          a       b       c             c – a               k         a        b        c             c – a 
                     --------------------------------------------------------------------------------------------------------------
                          1        15    112    113              98                6       275    252    373             98
                          2        51    140    149              98                7       351    280    449             98
                          3        95    168    193              98                8       435    308    533             98
                          5      207    224    305              98                9       527    336    625             98
 
 
Example 6Find formulas for all primitive Pythagorean triples (a, b, c) for which the difference between
c and a is 450. From Theorem 31, we know that c – a   = 2n2 = 450, so, n2 = 225, and n = 15 = 3(5). Thus, n has two odd prime factors; namely, p = 3 and p = 5.        
 
        For  p = 3, we get   k ≠ pw – (p – 1)/2 = 3w – 1 , and for  p = 5, we get   k ≠   5w – 2.
 
Thus, for n = 15, we have                       a   =   (2k + 14)2 – 225
                                                                b   =   30(2k + 14)
                                                                c   =   (2k + 14)2 + 225
 
                       for k ≠ 3w – 1, w = 1, 2, 3, 4, . . .     and       k ≠   5w – 2, w = 1, 2, 3, 4, . . .
 
that is,                  k ≠ 2, 5, 8, 11, 14, 17, . . .           and              k ≠   3, 8, 13, 18, 23, . . .
 
Below is a table of these triples for k = 1, 4, 6, 7, 9, and 10. Note that k ≠ 2, 3, 5, 8, 11, 13, . . .
 
                          k          a       b       c             c – a               k          a        b        c           c – a 
                     --------------------------------------------------------------------------------------------------------------
                          1        31    480    481             450               7       559    840   1009          450
                          4      259    660    709             450               9       799    960   1249          450
                          6      451    780    901             450             10       931  1020   1381          450
                         
Example 7Find formulas for all primitive Pythagorean triples (a, b, c) for which the difference between
c and a is 2. From Theorem 31, we know that c – a  =  2n2  =  2, so n2 = 1, and n = 1. Note that since n has no prime factors, there are no values for p and thus no restrictions on k.   Thus, for n = 1, we have
 
                                                                a   =   (2k)2 – 1   =   4k2 – 1                                                           (8a)
                                                                b   =    2(2k)        =    4k                                                                  (8b)
                                                                c   =   (2k)2 + 1   =   4k2 + 1                                                           (8c)
 
Below is a table of these triples for k = 1 through 6.
 
                                                              k            a        b        c          c – a
                                                          ---------------------------------------------------
                                                              1            3        4        5             2 
                                                              2         15        8    17             2
                                                              3          35     12      37             2
                                                              4         63      16    65             2
                                                              5          99      20    101             2
                                                              6        143      24    145             2
 
Now, looking at rows 2, 3, and 4 in this table, we can see a very interesting pattern. The sum of the bold-face entries in the left diagonal (from upper left to lower right) is 15 + 12 + 65 = 92, and the sum of the underlined bold-face entries in the right diagonal (from lower left to upper right) is 63 + 12 + 17, which is also 92! Thus, the sums of the entries in the diagonals of these three consecutive rows are equal!
     Let’s look at another three consecutive rows; say, rows 4, 5, and 6. The sum of the entries in the left diagonal is 63 + 20 + 145 = 228, and the sum of  the entries in the right diagonal is 143 + 20 + 65 = 228. Again, these two sums are equal!
     As another example, look at rows 6, 7, and 8 in the table of Example 5 above. Here we see that the sum of the left diagonal entries (275 + 280 + 533 = 1088) is the same as the sum of the right diagonal entries
(435 + 280 + 373 = 1088).
     Thus, it seems that for any table of primitive Pythagorean triples where c – a is a constant value, the sums of the entries in the diagonals of any three consecutive rows in these tables are equal.
     To show that this pattern is always true, let’s consider any table of primitive Pythagorean triples where c – a is a constant and which is arranged in increasing order of their a-values. Let n be the number of the nth row of this table, and an, bn, and cn be the a, b, and c values of the primitive Pythagorean triples in this row. Then, the sum of the entries in the left diagonal of rows n, n+1, and n+2 is  an + bn+1 + cn+2  and the sum of the entries in the right diagonal is  an+2 +  bn+1 +  cn. We must show that these two sums are equal.
     Now, c – a is a constant, so  cn+2 – an+2  =  cn – an. Adding an, an+2, and bn+1 to both sides of this equation, we get     
                                                               cn+2 – an+2    =    cn – an
                                cn+2 – an+2 + an + an+2 + bn+1   =    cn – an + an + an+2 + bn+1
                                                        cn+2 + an + bn+1   =    cn + an+2 + bn+1
                                                       an + bn+1 + cn+2    =    an+2 + bn+1 + cn,
 
proving that these two diagonal sums are indeed equal. This gives us
 
           Theorem 32  (Diagonal sums for constant c – a)
 
            Let an, bn, and cn be the a, b, and c values of the primitive Pythagorean triple in the nth row of any
            table of primitive Pythagorean triples (a, b, c) which is arranged in increasing order of their a-values,
            where c – a is a constant. Then, the sums of the entries of the left and right diagonals of any three
            consecutive rows in this table are equal; that is,   an +  bn+1 + cn+2   =   an+2 +  bn+1 + cn.
            ________________________________________________________________________________
 
     Now Theorem 32 applies to any table of triples for which c – a is a constant, but two more interesting and even extraordinary patterns can be seen in the table of Example 7 shown again below for c – a = 2.
 
                                                              k            a        b        c          c – a
                                                          ---------------------------------------------------
                                                              1            3        4        5             2 
                                                              2         15        8      17             2
 
                                                              3          35      12      37             2
 
                                                              4          63      16     65             2
                                                              5          99      20  101            2
 
                                                              6        143      24    145             2
 
Notice that the sum of the a and b values of the first triple (3 + 4) plus the b value of the second triple
(3 + 4 + 8) equals the a-value of the second triple,15, and notice that the sum of the b and c values of the fourth triple (16 + 65) plus the b value of the fifth triple (16 + 65 + 20) equals the c-value of the fifth triple,101
How nice!
 
     In general, it seems that if  ak, bk, and ck are the a, b, and c values of a triple in the kth row of this table, then
 
                                                             1)   ak + bk + bk+1   =   ak+1
and                                                       2)   bk + ck + bk+1   =   ck+1                   Let’s see.
 
From (8a), (8b), and (8c) above, we have    ak   =   4k2 – 1,   bk   =   4k, and  ck   =    4k2 + 1, so
 
                                       ak+1   =   4(k+1)2 – 1       and       ck+1   =    4(k+1)2 + 1.
 
Thus,   ak + bk + bk+1   =   (4k2 – 1) + 4k + 4(k + 1)   =   4k2 + 8k + 4 – 1    =   4(k2 + 2k + 1) – 1 
                                      =    4(k + 1)2 – 1   =   ak+1,   proving 1),
and      bk + ck + bk+1   =   4k + (4k2 + 1) + 4(k + 1)   =   4k2 + 8k + 4 + 1   =   4(k2 + 2k + 1) + 1 
                                      =    4(k + 1)2 + 1   =   ck+1,   proving 2),  and giving us
 
      Theorem 33
 
       Let an, bn, and cn be the a, b, and c values of the primitive Pythagorean triple (a, b, c) in the nth row
       of any table of primitive Pythagorean triples arranged in increasing order of their a-values, where
       c – a = 2. Then, 1)   an + bn +  bn+1   =   an+1,    and    2)   bn + cn +  bn+1   =   cn+1.    
       ________________________________________________________________________________
 
Example 8Use Theorem 33 to construct the first 6 triples of the table in Example 7, where c – a = 2, starting with the triple (3, 4, 5). From (8b) of Example 7, we know that b = 4k for k = 1, 2, 3, . . . , so the b-values will always be consecutive multiples of 4. Let’s put them in our table to get
 
                                                                    k            a        b        c
                                                              --------------------------------------
                                                                    1            3        4        5
                                                                    2                      8
                                                                    3                    12
                                                                    4                    16
                                                                    5                    20
                                                                    6                    24
 
 
Now, to find the a-value for row 2, we simply add the sum of the a and b-values from row 1 to the b-value from row 2, getting  (3 + 4) + 8 = 15, and similarly, to find the c-value for row 2, we simply add the sum of the b and c-values from row 1 to the b-value from row 2, getting (4 + 5) + 8 = 17. Our table is now
 
 
                                                                    k            a        b        c
                                                              --------------------------------------
                                                                    1            3        4        5
                                                                    2          15        8      17
                                                                    3                     12
                                                                    4                     16
                                                                    5                     20
                                                                    6                     24
 
 
Next, to find the a-value for row 3, we simply add the sum of the a and b-values from row 2 to the b-value from row 3, getting (15 + 8) + 12 = 35, and to find the c-value for row 3, we simply add the sum of the b and c-values from row 2 to the b-value from row 3, getting (8 + 17) + 12 = 37. The table now becomes
 
                                                                    k            a        b        c
                                                              --------------------------------------
                                                                    1            3        4        5
                                                                    2          15        8       17
                                                                    3          35       12     37
                                                                    4                     16
                                                                    5                     20
                                                                    6                     24
 
 
Continuing in this same fashion for row 4, we get  (35 + 12) + 16 = 63 for the a-value, and (12 + 37) + 16 = 65 for its c-value. Following this procedure, we end up with 
 
 
                                                                    k            a        b        c
                                                              --------------------------------------
                                                                    1            3        4         5
                                                                    2          15        8       17
                                                                    3          35      12       37
                                                                    4          63      16      65
                                                                    5          99      20     101
                                                                    6        143      24     145
 
 
     c) The difference between a and b. We can write a – b as   a – b = (m2 – n2) – 2mn, but this does not lead to any simple algebraic expression. The best we can do is to interchange the left and right-hand sides of this equation, getting 
                                                               (m2 – n2) – 2mn   =  a – b,
or                                                                m2 – 2mn – n2  =  a – b.                         Adding 2n2 to both sides of this equation, we get                                        m2 – 2mn + n2  =  a – b + 2n2,
or                                                                          (m – n2   =   2n2 + a – b.                Taking the positive square root of both sides of this equation, we have                                         __________       
                                                                                m – n    =  √ 2n2 + a – b  
                                                                                                             __________
or                                                                                m       =     n + √ 2n2 + a – b                                             (9)
                                                         __________
Since m and n are integers, then  √ 2n2 + a – b   must be an integer also, requiring that 2n2 + a – b be a perfect square. But how do we find those values of n, a, and b which make this a perfect square? 
 
     To find some patterns, let’s look at Table 2 in section 3 above and calculate the various values for a – b for all the triples found there. We will omit any differences of 1 or -1 (unit differences), since we have already discussed these earlier in this section. Proceeding in order from top to bottom in Table 2 and omitting repetitions and those values greater than 300, we get from the second triple in the table, (15, 8, 17),  a – b  =  15 – 8  =  7, from the third triple,
(35, 12, 37), a – b  =  35 – 12  =  23, and so forth. Thus, we have a – b   =
 
               7,  23, 47, 79, 79, 119,  -17, 167, 223, 287, 41,  -31, 73, 113, 161, 217, 281, 103, 71, 17, 151, 49, 97, 271, 89.   
 
Since we really don’t care whether the differences are positive or negative, let’s write down just the absolute values of these differences in numerical order, giving us   |a – b|   =
 
       7, 17, 23, 31, 41, 47, 49, 71, 73, 79, 89, 97, 103, 113, 119, 151, 161, 167, 217, 223, 271, 223, 271, 281, 287          (9a)
 
Now, are there any patterns here? Let’s look at the differences between these numbers from 7 through 113 to try and find some.
 
                   |a – b|             7,   17,   23,   31,   41,   47,   49,   71,   73,   79,   89,   97,   103,   113
        -------------------------------------------------------------------------------------------------------------------------------
               differences            10    (6     8)    (10    6      2)    22     (2      6     10     8)    6       10
                                                         14               18                                 26
 
Alas, there doesn’t seem to be any pattern in these differences! The first difference is 10, which then decreases by 4 to the second difference 6. We would have liked the second difference to be 14, an increase of 4, leading to the arithmetic sequence 10, 14, 18, 22, 26,   . . . However, the sum of the second and third differences is 14, which is what we want! And the sum of the fourth, fifth, and sixth differences is 18, as desired! And the seventh difference is 22! And the sum of the eighth, ninth, tenth, eleventh differences is 26. Wow! There is a pattern! (If we try hard enough.) 
 
     Let’s rewrite the table above by eliminating the numbers 23, 41, 47, 73, 79, 89, 103, and 113, getting
 
                                                       |a – b|                7     17     31     49     71     97
                                               -----------------------------------------------------------------------
                                                   differences               10     14     18     22     26
 
and the following table:
 
                                                                k                1       2      3       4       5       6
                                                         ---------------------------------------------------------------
                                                             |a – b|           7     17     31     49     71     97
 
     Using the values in this table, let’s write |a – b| as a second-degree polynomial in k; that is, for some integers d, e, and f,
                                                                      |a – b|  =  dk2 + ek + f.                                                              (10)
 
From the table, if k = 1, then |a – b| = 7, and (10) becomes       7   =    d  +   e + f.                                     (10a)
If k = 2, then |a – b| = 17, and (10) becomes                             17  =   4d  + 2e + f.                                     (10b)
If k = 3, then |a – b| = 31, and (10) becomes                             31  =   9d  + 3e + f.                                     (10c)
 
Subtracting (10a) from (10b), we get                                         10  =  3d + e.                                              (11a)
Subtracting (10b) from (10c), we get                                         14  =  5d + e.                                              (11b)
 
Finally, subtracting (11a) from (11b), we get                       4 = 2d,   or   d = 2. 
 
Substituting d = 2 into (11a) gives us e = 4, and substituting d = 2 and e = 4 into (10a) gives us f = 1.
 
Thus, (10) becomes                                   |a – b| = 2k2 + 4k + 1   for   k = 1, 2, 3, . . . ,                              (12)
 
and we have found a formula for some of the a – b differences. But what about the rest?
 
     Writing down those remaining differences from (9a) to which (12) does not apply, we have
 
                23,  41,  47,  71,  73,  79,  89,  103,  113,  119,  151,  161,  167,  217,  223,  271,  281,  287                    (13) 
 
Are there any patterns here? It certainly doesn’t seem so! However, to get formula (12), we used those differences which formed the arithmetic sequence 10, 14, 18, 22, . . .   Now, in (13) the first difference is 23, then 41, which is a difference of 18. Perhaps we can find numbers from (13) whose differences form the arithmetic sequence 18, 22, 26, 30, . . . Let’s see. Choosing specific numbers from (13) so as to form our desired sequence, we get
 
 
                                                         |a – b|              23     41     (63)     89     119                                           (13a)  
                                               -----------------------------------------------------------------------
                                                   differences               18      22       26      30   
 
where we have had to put in 63 as a value for |a – b| to get our desired sequence even though it is not in the list (9a). 
 
Forming the following table,
                                                                k                1       2      3       4       5      
                                                         ---------------------------------------------------------------
                                                             |a – b|          23     41    63     89    119
 
let’s use the values in this table, as we did earlier in a similar manner, to write |a – b| as a second-degree polynomial in k; that is, for some integers d, e, and f,
 
                                                                      |a – b| = dk2 + ek + f.                                                                 (14)
 
From the table, if k = 1, then |a – b| = 23, and (14) becomes    23   =    d  +   e + f.                                     (14a)
If k = 2, then |a – b| = 41, and (14) becomes                             41   =   4d  + 2e + f.                                     (14b)
If k = 3, then |a – b| = 63, and (14) becomes                             63   =   9d  + 3e + f.                                     (14c)
 
Subtracting (14a) from (14b), we get                                         18   =   3d + e.                                             (15a)
Subtracting (14b) from (14c), we get                                         22   =   5d + e.                                             (15b)
 
Finally, subtracting (15a) from (15b), we get                       4 = 2d,   or   d = 2. 
 
Substituting d = 2 into (15a) gives us e = 12, and substituting d = 2 and e = 12 into (14a) gives us f = 9.
 
Thus, (14) becomes                                   |a – b| = 2k2 + 12k + 9   for   k = 1, 2, 3, . . . ,                             (16)
 
     Now, comparing (12) with (16), we can see another pattern emerging.
   
                                          |a – b|   =   2k2  +   4k  +  1   =   2k2  +  4(1)k  +  12
 
                                           |a – b|   =   2k2  + 12k  +  9   =   2k2 +  4(3)k  +  32
 
If this pattern continues, we would expect that   |a – b|   =   2k2 + 4(5)k + 52,   and in general,
 
                                       |a – b|  =   2k2 + 4sk + s2   for   k = 1, 2, 3, . . . and s = 1, 3, 5, . . .                        (17)
 
Now, if we use formula (17) for |a – b| in (9) above, will this help us to find primitive Pythagorean triples with constant a – b differences? Let’s see!
 
     In (9), the relationship necessary between m and n for a given difference a – b was found to be
                                                                                      __________
                                                                  m   =   n + √ 2n2 + a – b .                                                              (9)
 
1) If a > b, then a – b is positive, and we substitute 2k2 + 4sk + s2 for a – b in (9) above to get
                                                                                ___________________
                                                            m   =   n + √ 2n2 +  2k2 + 4sk + s2                                                    (18)
 
Now the last three terms of the expression under the radical sign, 2k2 + 4sk + s2, remind us of (2k + s)2, but of course, (2k + s)2 = 4k2 + 4sk + s2, not 2k2 + 4sk + s2. We need another 2k2, and we can get it if we pick n = k.
Then, the expression under the radical sign becomes
  
                                  2n2 + 2k2 + 4sk + s2   =   4k2 + 4sk + s2   =   (2k + s)2,   a perfect square!,
                                                                                        __________________
and substituting n = k into (18), we get       m   =   n + √ 2n2 + 2k2 + 4sk + s2
                                                                                         _____________               ________
                                                                            =   k + √ 4k2 + 4sk + s2     = k + √ (2k + s)2           
                                                                          
                                                                            =    k + 2k + s  =  3k + s.
 
So, when m  = 3k + s  and  n = k, the difference between the a and b values of (a, b, c) will be the constant value  2k2 + 4sk + s2, and a will be greater than b. (a  >  b).
 
2) If a < b, then a – b is negative, and we substitute  - (2k2 + 4sk + s2) for  a – b  in (9) above to get
                                                                                ___________________
                                                            m   =   n + √ 2n2 – 2k2 – 4sk – s2                                                    (19)
 
If, as in 1) above, we let n = k, then the expression underneath the radical sign becomes - 4sk – s2, which is
negative and thus not possible. But, what if we pick n = k + s ? Then, the expression under the radical sign
becomes
                                                2n2 – 2k2 – 4sk – s2   =   2(k + s)2  – 2k2  – 4sk – s2   
 
                                                      =  2k2 +   4sk + 2s2 – 2k2 – 4sk – s2   =   s2, again a perfect square!   
                                                                                                   ______________________
and substituting n = k + s into (19), we get    m   =   k + s  + √ 2(k + s)2 – 2k2 – 4sk – s2
                                                                                                   ___                
                                                                               =   k + s + √  s2     =  k  +  2s.           
                                                                          
So, when  m = k + 2s  and  n = k + s, the difference between the a and b values of (a, b, c) will be the constant value  - 2k2 – 4sk – s2, and a will be less than b. (a  <  b).
 
     As can be seen from the formulas for m and n, if k is a multiple of s, then m and n will have a common factor of s, and (a, b, c) will not be a primitive Pythagorean triple. All of this gives us the following theorem.
 
           Theorem 34
 
           Let  (a, b, c)  be a primitive Pythagorean triple.   Then the possible values for | a – b | are
           given by   2k2 + 4sk + s2; that is,  | a – b | = 2k2  +  4sk  +  s2, where s is any odd positive
           integer, and k is any positive integer not a multiple of s. The generating numbers m and n
           for (a, b, c) are
                                       1)   m = 3k + s     and   n = k,            if a > b,
           and                      2)   m =   k + 2s   and   n = k + s,      if a < b.
           _________________________________________________________________________
 
Example 9Find three primitive Pythagorean triples (a, b, c) for which a – b = 73. Since a – b is positive, we know that a > b, so we will use part 1) of Theorem 34 to find the values of m and n for these triples. To do this, we must first solve | a – b|  =  2k2 + 4sk + s2  =  73 for s and k. Choosing odd positive integers for s, we get
 
   for s = 1,   2k2  +  4k +   1 = 73,    2k2 +  4k – 72 = 0, and   k2 +   2k – 36 = 0, which is not factorable,
   for s = 3,   2k2 + 12k +   9 = 73,   2k2 + 12k – 64 = 0, and   k2  +  6k – 32 = 0, which is also not factorable,
   for s = 5,   2k2 + 20k + 25 = 73,   2k2 + 20k – 48 = 0, and   k2 + 10k – 24 = 0, so (k + 12)(k – 2) = 0, and k = 2.
  
Thus, s = 5 and k = 2, and substituting these values into 1) of Theorem 34, we get m = 3(2) + 5 = 11, and n = 2,
giving us  a = 112 – 22 = 117, b = 2(11)(2) = 44, and c = 112 + 22 = 125, and so (117, 44, 125) is one primitive Pythagorean triple for which a – b = 73. But, how do we find two more?
 
     In our discussion of unit triples earlier in this section, we discovered a way to find the next unit triple after any given unit triple in a list of triples arranged in increasing order of their a values. (See Theorem 29.) Since, for unit triples, the difference between their a and b values is always constant, | a – b | = 1, perhaps Theorem 29 will apply to primitive Pythagorean triples with constant value differences between their a and b values other than 1. Let’s see.
 
     Starting with m1 = m = 11 and n1 = n = 2, and using Theorem 29, we get
 
                                          m2  =  2m1 + n1  =  2(11) + 2  =  24      and     n2  =  m1  =  11.
 
Continuing in this fashion, we get the table shown below in which the triples all have a constant difference of 73 (in absolute value) between their a and b values.
 
                                              m       n                  a             b              c                a – b
                                        --------------------------------------------------------------------------------
                                              11      2                117          44           125                73
                                              24    11                455        528           697              - 73
                                              59    24              2905      2832         4057                73
                                            142    59            16683    16756       23645              - 73
                                            343  142            97485    97412     137813                73
 
     We choose (117, 44, 125), (2905, 2832, 4057), and (97485, 97412, 137813) as our three triples since
a – b is to be positive.
 
     Apparently, Theorem 29 does apply to primitive Pythagorean triples for which the difference between their a and b values is a constant other than 1! Let’s see if we can prove this.
 
     Accordingly, we let A1 be any triple (a1, b1, c1) with m and n values of m1 and n1, and a1 – b1 = d,  a given positive integer. Next, we let A2 be another triple (a2, b2, c2) whose m and n values are m2 = 2m1 + n1   and n2 = m1. Then,
                                          a2  –  b2   =          m22       –  n22    –      2m2n2
 
                                                          =   (2m1 + n1)2  –  m12  –  2(2m1 + n1)m1
 
                                                          =   4m12 + 4m1n1 + n12 – m12 – 4m12 – 2m1n1
 
                                                          =   2m1n1 + n12 – m12
 
                                                          =  -  (m12 – n12 – 2m1n1)
 
                                                           =  -  (a1 – b1)   =   - d.
 
Thus, we get
 
            Theorem 35
 
             If A = (a, b, c) is a primitive Pythagorean triple with   m = m1,   n = n1,   and   a – b = d,
             a given positive integer, then in the triple whose m and n values are given by m = 2m1 + n1
             and n = m1,  a – b   = - d.
             __________________________________________________________________________
 
Example 10Find three primitive Pythagorean triples (a, b, c) for which b – a = 41. Since b – a is positive, we know that a < b, so we will use part 2) of Theorem 34 to find the values of m and n for these triples. To do this, we must first solve | a – b| = 2k2 + 4sk + s2 = 41 for s and k. To help us find s and k easily, let’s find the values of
2k2 + 4sk + s2 for various values of s and k, putting them in Table 4 below.
 
Table 4
Values of |a – b| = 2k2 + 4sk + s2
for s = 1, 3, 5, . . .  , 15 and k = 1, 2, 3, . . .  , 7.
 
                                                                                               k
 
                                                                1        2        3         4         5        6        7
                                              --------------------------------------------------------------------------
                                                  1            7      17       31       49       71       97      127
                                                  3          23      41       63       89     119     153      191
                                                  5          47      73     103     137     175     217      263
                                        s        7          79    113     151     193     239     289      343
                                                  9        119    161     207     257     311     369      431
                                                11        167    217     271     329     391     457      527
                                                13        223    281     343     409     479     553      631    
                                                15        287    353     423     497     575     657      743          
                                              ---------------------------------------------------------------------------
 
     As we can see from Table 4, 2k2 + 4sk + s2 equals 41 when s = 3 and k = 2. Then, from part 2) of Theorem 34, we have m = k + 2s = 2 + 2(3) = 8, and n = k + s = 2 + 3 = 5, which gives us (39, 80, 89) as one triple for which
b – a = 41. To find more triples, we use Theorem 35 to get
 
                                              m       n                     a                b                c              b – a 
                                         ----------------------------------------------------------------------------------
                                                8       5                   39              80              89                41
                                              21       8                 377            336            505              - 41
                                              50     21               2059          2100          2941                41
                                            121     50             12141        12100        17141              - 41
                                            292   121             70623        70664        99905                41
       
and thus (39, 80, 89), (2059, 2100, 2941), and (70263, 70664, 99905) are three triples for which b – a = 41.
(If we wish |a – b| = 41, then all five triples in the table satisfy this requirement.)
 
     Note: In (13a) above, we put in 63 as a value for |a – b| to get a desired sequence even though it was not in the list (9a). As we can see above, 63 is indeed in Table 4 and is thus a possible value for |a – b|.
 
Example 11Find three primitive Pythagorean triples (a, b, c) for which a – b = 119. Since a – b is positive, we know that a > b, so we will use part 1) of Theorem 34 to find the values of m and n for these triples. To do so, we must first solve | a – b|   =   2k2 + 4sk + s2   = 119  for  s  and  k.   Looking in Table 4, we find that
2k2 + 4sk + s2 equals 119 in two different places;
 
                                  1) when  s = 9 and k = 1,      and       2) when s = 3 and k = 5.
 
Then, from part 1) of Theorem 34, we have
 
        1) for s = 9, k = 1,  m = 3k + s = 3(1) + 9, m = 12, and n = k, n = 1, which gives us (143, 24, 145)
            as one triple for which a – b = 119, and
 
        2) for s = 3, k = 5, m = 3k + s = 3(5) + 3, m = 18, and n = k, n = 5, which gives us (299, 180, 349)
            as another triple for which a – b = 119.
 
 
    To find more triples, we use Theorem 35 to get
 
 
                 1)                          m       n                    a                b                c               a – b 
                                          ----------------------------------------------------------------------------------
                                              12       1                 143              24            145              119
                                              25     12                 481            600            769            - 119
                                              62     25               3219          3100          4469              119
                                            149     62             18357        18476        26045            - 119
                                            360   149           107399      107280      151801              119
      
                and thus (143, 24, 145), (3219, 3100, 4469), and (107399, 107280, 151801) are three triples for       
                which a – b = 119.
 
 
 
                 2)                          m       n                    a                b                c               a – b 
                                          ----------------------------------------------------------------------------------
                                              18       5                 299            180            349              119
                                              41     18               1357          1476          2005            - 119
                                            100     41               8319          8200        11681              119
                                            241   100             48081        48200        68081            - 119
                                            582   241           280643      280524      396805              119
      
                and thus (299, 180, 349), (8319, 8200, 11681), and (280643, 280534, 396805) are three triples for       
                which a – b = 119.
 
 
     There is an interesting pattern in Table 4.  Here are the first three rows of Table 4, written to show the differences between each two adjacent row entries (in bold-face type).
 
 
                                                                                          
                                                           1        2        3         4         5         6         7
                                         ---------------------------------------------------------------------------
                                             1           7      17       31       49       71       97      127
                                                              10       14       18       22       26      30
 
                                             3          23      41       63       89     119     153      191
                                                              18       22       26       30      34        38        
 
                                             5          47      73     103     137     175     217      263
                                                              26     30       34       38      42       46
 
 
Notice that the differences between adjacent row entries in row 1, namely, 10, 14, 18, 22, . . .  , form an arithmetic sequence starting at 10 and then increasing by 4 for each succeeding difference.  The same is true for rows 2 and 3 except their arithmetic sequences start at 18 and 26, each 8 units higher than the previous row’s starting value, respectively.   Below is Table 4 again, showing that this pattern continues throughout the table. 
 
 
 
Table 4 (Rows)
Values of |a – b| = 2k2 + 4sk + s2
for s = 1, 3, 5, . . .  , 15 and k = 1, 2, 3, . . .  , 7.
 
                                                                                          k
 
                                                           1        2        3         4         5         6         7
                                         ---------------------------------------------------------------------------
                                             1            7      17       31       49       71       97       127
                                                              10      14       18        22       26      30
 
                                             3          23      41       63       89     119     153       191
                                                              18      22       26       30       34        38        
 
                                             5          47      73     103     137     175     217       263
                                                              26      30       34       38       42        46
 
                                  s         7          79    113     151     193     239     289       343
                                                              34     38       42       46       50        54
 
                                             9        119    161     207     257     311     369      431
                                                              42     46       50       54       58        62
 
                                           11        167    217     271     329     391     457       527
                                                             50      54       58        62       66        70
 
                                           13        223    281     343     409     479     553       631    
                                                             58      62       66        70       74        78
                                                             
                                           15        287    353     423     497     575     657       743         
                                                             66      70       74        78       82        86
 
                                       ----------------------------------------------------------------------------
 
 
     We can easily find more values in Table 4, if needed, by simply extending the arithmetic sequences as far as we like. For example, to extend the values in the row where s = 9, we simply add 4 to 62, the last difference in its sequence, getting 66 for the next difference in the sequence. We then add this difference to 431, the last entry in the row for s = 9, getting 431 + 66 = 497 as the next entry in this row.
 
 
                                             9        119    161     207     257     311     369       431         (497)
                                                              42     46       50       54       58        62          (66)
 
 
    Below are the first four columns of Table 4, written to show the differences between each two adjacent column entries (in bold-face type).   Notice that the differences between adjacent column entries in column 1, namely, 16, 24, 32, 40, . . . , form an arithmetic sequence starting at 16 and then increasing by 8 for each succeeding difference. The same is true for columns 2, 3, and 4 except their arithmetic sequences start at 24, 32, and 40, each 8 units higher than the previous column’s starting value, respectively.
 
 
 
 
Table 4 (Columns)
Values of |a – b| = 2k2 + 4sk + s2
for s = 1, 3, 5, . . .  , 15 and k = 1, 2, 3, . . .  , 7.
 
                                                                                          k
                                                           1                      2                      3                      4
                                         ------------------------------------------------------------------------------------
                                             1            7                    17                    31                    49
                                                               16                   24                     32                    40
                                             3          23                    41                    63                    89
                                                               24                   32                     40                    48
                                             5          47                    73                   103                 137
                                                               32                   40                     48                    56
                                  s         7          79                  113                   151                  193  
                                                               40                   48                     56                    64
                                             9        119                  161                   207                  257
                                                               48                   56                     64                    72
                                           11        167                  217                   271                  329
                                                               56                   64                     72                    80
                                           13        223                  281                   343                  409    
                                                               64                   72                     80                    88
                                           15        287                  353                   423                  497         
                                         --------------------------------------------------------------------------------------
 
     Another interesting pattern in Table 4 can be found by looking at any square block in the table, such as the 3 x 3 square shown below in red.
Table 4
Values of |a – b| = 2k2 + 4sk + s2
for s = 1, 3, 5,  . . .  , 15 and k = 1, 2, 3, . . .  , 7.
 
                                                                                          k
                                                                1        2        3         4         5         6         7
                                              --------------------------------------------------------------------------
                                                  1            7      17       31       49       71       97      127
                                                  3          23      41       63       89     119     153      191
                                                                                                                                                  d              f
                                                  5          47      73     103     137     175     217      263             
                                       s         7          79    113     151     193     239     289      343               .        .        .
                                                  9        119    161     207     257     311     369      431             
                                                                                                                                                  e        .        g
                                                11        167    217     271     329     391     457      527
                                                13        223    281     343     409     479     553      631    
                                                15        287    353     423     497     575     657      743         
                                              ---------------------------------------------------------------------------
 
If we add the two numbers at the ends of the left diagonal, call them d and g, then d + g = 73 + 257 = 330, and if we add the two numbers at the ends of the right diagonal, call them e and f, then e + f = 161 + 137 = 298. Then, (d + g) – (e + f) = 330 – 298 = 32. Not so very interesting until we realize that this is always true for any 3 x 3 square in the table.  For example, in the 3 x 3 square that is at the bottom right of Table 4 above, we have d = 391, e = 575, f = 527, and g = 743,  and (d + g) – (e + f) = (391 + 743) – (575 + 527) = 1134 – 1102 = 32.  
 
     This pattern is true for other sizes of square blocks, also. For example:
 
     In the 2 x 2 square shown below in orange, we have d = 223, e = 287, f = 281, and g = 353, and (d + g) –
(e + f) = (223 + 353) – (287 + 281) = 576 – 568 = 8, which is true for all 2 x 2 squares.  
 
     In the 4 x 4 square shown below in red, we have d = 137, e = 329, f = 263, and g = 527, and (d + g) – (e + f) = (137 + 527) – (329 + 263) =   664 – 592 = 72, which is true for all 4 x 4 squares.  
 
Table 4
Values of |a – b| = 2k2 + 4sk + s2
for s = 1, 3, 5, . . .  , 15 and k = 1, 2, 3, . . .  , 7.
 
                                                                                          k
                                                                1        2        3         4         5         6         7
                                              --------------------------------------------------------------------------
                                                  1            7      17       31       49       71       97      127
                                                  3          23      41       63       89     119     153      191
                                                                      
                                                  5          47      73     103     137     175     217      263
                                       s         7          79    113     151     193     239     289      343
                                                  9        119    161     207     257     311     369      431                                                                       
                                                11        167    217     271     329     391     457      527
 

                                                13        223    281     343     409     479     553      631    
                                                15        287    353     423     497     575     657      743         
                                          
                                               ---------------------------------------------------------------------------
 
     Let’s put all this information into a table and try to find a general expression for (d + g) – (e + f). Accordingly, we have      
                                           Square size                           (d + g) – (e + f)
                                       ----------------------------------------------------------------------------------
                                                2 x 2                    8   =    8(1)    =    8(12)    =    8(2 – 1)2
                                                3 x 3                  32   =    8(4)    =    8(22)    =    8(3 – 1)2
                                                4 x 4                  72   =    8(9)    =    8(32)    =    8(4 – 1)2
                                                n x n                                      8(n – 1)2
 
The data in the table suggests that if we have an n x n square, then   (d + g)  –  (e + f)   =   8(n – 1)2.     (20) Let’s try to prove this.  Look at the 6 x 6 square below.
 
                                                             k                                                                 k + (n-1)
                                                             3       4      5      6       7        8 → [3 + 5 = 3 + (6-1) ]
                                                             
                                                           
                                          s       7         d        .      .       .       .         f
                                                              
                                                   9          .                                          .
 
                                                 11          .                                          .
                              
                                                 13          .                                          .
 
                                                 15          .                                          .
 
           17 → [7 + 10 = 7 + 2(6-1)]        e        .        .        .       .      g
                                     s + 2(n-1)
 
Notice that for this 6 x 6 square,
 
1)      d and f have the same value of s (namely, 7), but the k-value of f, 8, is 5 units (6 – 1) greater than the k-value of  d; namely, 3 + (6 – 1).
2)      e and g have the same value of s (namely, 17), but this value of s is 10 units [ 2(5) = 2(6 – 1 ] greater than the s-value of d and f; namely, 7 + 2(6 – 1), and the k-value of g, 8, is 5 units (6 – 1) greater than the k-value of e; namely, 3 + (6 – 1).
 
In general, for an n x n square,
 
1)      d and f have the same values of s, but the k-value of f is n – 1 units greater than the k-value of d; namely, k + (n – 1). Thus, if
                                               d   =          2k2           +           4sk          +    s2,
           then                                         f   =   2(k + n – 1)2    +   4s(k + n – 1)   +    s2.
      
2)       e and g have the same value of s, but it is 2(n – 1) units greater than the s-values of d and f; namely,
       s + 2(n – 1) = s + 2n – 2, and the k-value of g is n – 1 units greater than the k-value of e;  namely.
       k + (n – 1). Thus, if   
                                d    =         2k2           +             4sk                         +           s2,
           then                         e    =         2k2           +     4(s + 2n – 2)k                +   (s + 2n – 2)2,
            and                         g    =   2(k + n – 1)2  +    4(s + 2n – 2)(k + n – 1)  +   (s + 2n – 2)2
 
Now,   d + g   =   [ 2k2 + 4sk + s2 ]   +   [2(k + n – 1)2 + 4(s + 2n – 2)(k + n – 1) + (s + 2n –     22 ],        (21)                                            
 
and      e + f   =   [ 2k2 + 4(s + 2n – 2)k + (s + 2n – 2)2 ] + [ 2(k + n – 1)2 + 4s(k + n – 1) + s2 ].              (22)
 
Subtracting (22) from (21), the bold-faced italicized terms cancel, and we get
 
           (d + g) – (e + f)    =    [ 4sk   +   4(s + 2n – 2)(k + n – 1) ]     –    [ 4(s + 2n – 2)k   +   4s(k + n – 1) ]
 
 =           [ 4sk + 4(sk + sn – s + 2nk + 2n2 – 2n – 2k – 2n + 2) ]     –    [ 4sk + 8nk – 8k + 4sk + 4sn – 4s ]
 
 =           4sk + 4sk + 4sn – 4s + 8nk + 8n2 – 8n – 8k – 8n + 8 – 4sk – 8nk + 8k – 4sk – 4sn + 4s
 
 =            8n2 – 16n + 8   =    8(n2 – 2n + 1)   =    8(n – 1)2,   which proves (20).   Thus, we have
 
      Theorem 36
   
      For any primitive Pythagorean triple (a, b, c),   |a – b| = 2k2  +  4sk  +  s2, where s is any odd positive
      integer, and k is any positive integer not a multiple of s. Consider any table of values of |a – b|, where
      the row headings are values of s = 1, 3, 5, , and the column headings are values of k = 1, 2, 3, … .
      Then, for any n x n square block in this table, (d + g) – ( e + f)   =   8(n – 1)2, where d and g are the
      end values of the left diagonal, and e and f are the end values of the right diagonal of the square.
      __________________________________________________________________________________
 
                                                   Part A                                                         Part C 
                                                                                        Home