SOME SEQUENCES GENERATED BY SPIRAL SIEVING METHODS

H.W.GOULD
West Virginia University, M organ town, West Virginia 26506
The object of this note is to point out some curious sequences which may be generated by natural number spirals
and rotating grids. The method is a combination of the spiral introduced by Ulam [2] in his studies of prime number
distribution and a well known technique employed in cryptographic work. We illustrate with Fibonacci numbers.
Ulam considers a spiral numbering of the lattice points in the plane, i.e., by starting at (0,0) and proceeding
counterclockwise in a spiral so that
(0,0)^1, (7,0)-+2, (1,1}* $, (0,1)-* 4, (-1,1)-* 5, (-1,0)-* 6, (-1,-1)-+7, (0,-1)^8,
(1-1) ~>9, (2-1) -10, (2,0) - /1, (2,1) - 12, (0,2) - 13, (-1,2) -* 14, etc.
This mapping gives us the spiral below.
(D
f"""" - -
17
18
19
20
21
16
5
6
7
22
15
4
1
8
23
14
3
2
9
24
13
12
11
10
25
27
26
A nice illustration of the basic Ulam spiral makes up the front cover of the March 1964 Scientific American. In the
same issue Martin Gardner [1] gives an account of Ulam's work. Briefly, Ulam marks the primes (1,2,3,5,7,11,) in
the spiral and studies the visual display for patterns or almost-patterns in the prime number sequence. By use of a
computer at Los Alamos he is able to generate displays having around 65,000 points in them. It would be of interest
to try something of the same sort for the Fibonacci, Lucas, and other recurrent sequences, however the writer does
not have available such versatile equipment as that used by Ulam and his colleagues at Los Alamos, so we have little
to suggest about possible patterns in a spiral display of Fibonacci numbers. Of course, the fact that we now know [ 3] ,
[4] that 1 and 144 are the only square Fibonacci numbers does tell us that the diagonals 1,9
ff
25,49, and 4, 16,
36, 64, . will be conspicuously blank in such a display.
Now, there is a technique in cryptographic work which makes use of a rotating grid. We can best illustrate by means
of an example. Consider the message, ' INTUITION LIKE A FLASH OF LIGHTNING LASTS ONLY FOR A
SECOND." We write this in a square array
I N T U I T !
ON L I K EA
F L A S HOF
L I G H T N I
NG L A S T S
ON LY F 0 R
A S E C D N D
393
394 SOME SEQUENCES GENERATED BY SPIRAL SIEVING METHODS [DEC.
and then impose a prepunched grid, e.g., of the form (where an X indicates a hole)
[x
X

I n ,
X
1 X
X
- J
X
X
X
X
X
X
X
X
and copy out the visible letters, which are (serially, row by row) ITTIAOHTSOLOC. We then rotate the grid counter-
clockwise through 90 and again copy out the visible letters, which are IOLESIHIMLTAIM. Two more rotations gives
us UMKAFGHGSYSODand NILHFLHNAIMFRE. Running these four groups together and breaking the whole up in-
to convenient blocks then gives us the enciphered message. To decipher, one merely places the grid on a sheet of pa-
per, writes in the letters serially, row by row, thirteen at a time here, rotating the grid until all four positions are used,
removes the grid and reads off the message. Here we have used a 7 by 7 grid which leaves the middle point fixed (H).
This is unsatisfactory for cryptographic work in some cases and most ordinary uses involve an even-order grid.
The effect of an odd-order grid in the case of superposition on the natural number spiral is to partition the natural
numbers into four sets, any two of which have onlv the number 1 in common.
It is clear that the very special cryptographic grid cannot be made from the Fibonacci sequence (or the prime num-
ber sequence) without adding and/or deleting elements, since any given square annulus of the grid must be so design-
ed that one-fourth of its lattice points are punched, and in such a way that the same hole does not appear under
successive rotations of 90 until the original position is assumed. We shall not discuss how this can be effected.
We modify the rotating grid as follows. On the original natural number spiral (1) superimpose a square sheet of
paper which will just cover the first (2n - 1)
2
natural numbers, unity being kept at the center. IVSake a grid by punch-
ing the sheet wherever an element a
k
(k = 1,2, 3, -) of a given sequence appears in the natural number spiral. We
shall call this the (counterclockwise) spiral grid of the sequence \a
k
\. We next rotate the spiral grid through 90
and read off from the natural number spiral a new sequence generated by the spiral grid of our original sequence.
With any given sequence there will be associated three new sequences, and by turning the grid over (making it a clock-
wise spiral grid) we can generate four other sequences. Clearly all these eight sequences will be somehow related,
For a grid measuring 2n - / by 2n - 1 (n>2) there will be the natural numbers from 7 through (2n - 1)
2
with the outer square annulus containing the successive natural numbers from (2n - 3)
2
+ 1 to (2n - I)
2
. If an
element a
k
of our given sequence lies in the outer square annulus, then so will the corresponding element b'
k
of any
of the associated sequences obtained by use of the grid. It is possible to work out complicated formulas relating b
k
to a
k
depending upon the position of an element in the annulus. For example, any two diagonally opposite ele-
ments in the outer annulus have numerical difference 4n.
We give below, in Table 1, a few values for the sequences generated by the counterclockwise spiral grid of the Fib-
onacci sequence (I, I I , I I I , IV) and also for the clockwise grid (T, IT, I I I ' , I V) .
Here, d = d
k
is the minimum positive difference between terms in the sequences, or
d = d
k
= min (b'
k
- b{) > 0,
U
(with d
0
M. = Q)
for Counterclockwise (I - SV), or for Clockwise (V ~ I V) .
In our table, ak = Fk+u with
Fk+1 = Fk + Fk-1. Fo = 0, Fj= I
St is convenient to begin our sequence with F
2
instead of making some rules abouthowto interpret 0, 1, 1,2,3, - .
(The indistinguishability of f y and F
2
prevents us from calling the ordinary Fibonacci sequence a subset of the
set of all natural numbers.)
There is no reason to confine our attention to spirals based on a square. Ulam's work with the sequence of primes
quite naturally fits in well with such a spiral because quadratic polynomials Ax + Bx + C are often so rich in
1974] SOIVSE SEQUENCES GENERATED BY SPIRAL SIEVIWG METHODS
395
k
1
2
3
4
5
6
7
8
9
10
11
12
13
a*
I

2
3
5
8
13
21
34
55
89
144
233
377
i
II
r~
2
4
5
7
17
25
40
63
99
156
249
397
b
k '
III
. _
4
6
7
9
13
21
46
71
109
168
265
417
Table 1
etc.
IV
1
3
6
8
9
17
25
28
79
119
132
281
437
r

3
5
6
8
17
25
34
67
103
134
265
405
II'
1~~
2
5
7
8
13
21
40
75
113
146
281
425
IIS'
~
2
4
7
9
17
25
46
51
83
158
233
365
IV
1
3
4
6
9
13
21
28
59
93
122
249
385
d
(T
1
1
1
1
4
4
6
8
10
| 12
16
I 20
primes for integral values of x (Euler's polynomial x +x + 41 being the most well-known example). However, to
exhibit other properties of a sequence, as well as to generate variations of a given sequence, it is natural to pass on to
figurate numbers as the basis of our spirals. That is, we may consider a polygon of m sides.
Consider, for example, a pentagonal spiral as shown below.
21
20
19
23
22
11
10
18
17
24
25
12
5 6
13 14
31
16
30
15
26 27 28 29
It would be of interest to examine the distribution of primes, Fibonacci numbers, etc., in an extended pentagon with
thousands of points, and of course this would require quite an elaborate computer set-up.
It is fairly easy to type out a pentagonal spiral on ordinary typing paper with 456 points and this is sufficient to
give an idea of how the pentagonal spiral grid of the Fibonacci sequence can be used to generate curious sequences.
Here, of course, we shall have in all ten sequences. The sequences are tabulated below in Table 2.
The number d tabulated in the last column is defined as before by
(3) d = min (b
i
k
-h{) > Q
(for I - V or I' - V) , and it is not difficult to see that for any given value of k the numbers II - V determined
by our grid will differ from the Fibonacci number a
k
by a multiple of the number d. The reader may find it of in-
terest to try and develop a general formula for d in terms of k and m (generalizing to an /7?-gon).
396 SOlVfE SEQUENCES GENERATED BY SPIRAL SIEVING METHODS [DEC.
Table 2
_Jk
" 1
2
3
4
5
6
7
8
9
10
11
12
13
3k
I
1
2
3
5
0
13
21
34
55
89
144
233
377
b
k
n_
1
3
4
6
10
15
24
38
60
95
152
243
389
b
k
in
1
2
4
5
7
12
27
42
65
101
160
253
341
...
IV
1
3
5
6
9
14
30
46
70
77
168
263
353
etc.
V
1
2
4
6
11
16
18
50
75
83
176
273
365
!'
1
3
4
6
10
15
19
38
59
77
156
241
371
II'
1
2
4
5
7
12
22
42
64
83
164
251
383
III'
1
3
5
6
.9
14
25
46
69
89
172
261
335
IV
1
2
4
6
11
16
28
50
74
95
180
271
347
V
1
2
3
5
8
13
31
34
54
101
148
231
359
d
0 '
1
1
1
1
1
3
4
5
6
8
10
12
The visual display of perfect squares in a pentagonal spiral turns out to be a simple trefoil spiral appearing some-
what as diagrammed below.
This is easily verified to be in accord with the fact that the three arms of the spiral are formed by squares of form
(3n)
2
, (3n + 1)
2
, and (3n+2)
2
, respectively.
1974]
SOME SEQUENCES GENERATED BY SPIRAL SIEVING METHODS 397
Finally, we turn to the case of a triangular spiral grid. Because of the hexagonal rotational character in this case,
one may generate 12 sequences for a given spiral grid, 6 counterclockwise and 6 clockwise. A portion of the triangu-
lar spiral appears below.
39 7 35
15 13 50
26 3 22
8 6 34
16 1 12
4 2 21
9 10 5
18 19 11
The 12 sequences generated by a triangular spiral grid based on the Fibonacci numbers are tabulated in Table 3.
Table 3
d ..d* I
1
2
3
5
8
13
21
34
55
89
il
1
4
6
8
15
22
23
35
57
70
III
1
3
4
7
10
16
25
39
61
97
IV
1
2
8
10
18
26
27
40
63
77
V
1~
2
4
6
9
19
29
44
49
105
VI
~~1
3
6
10
12
30
31
45
51
84
r
1
3
4
6
9
15
27
40
i
5 1
99
II'
1
3
8
10
16
29
30
44
55
81
III'
~~7~
2
4
5
8
18
31
45
57
107
IV
1
4
6
10
19
21
22
34
61
67
V
1
2
3
7
10
12
23
35
63
91
VI'
~1
2
6
8
13
25
26
39
49
74
~~i 0
1
1
1
1
3
4
5
6
8
0
1
2
2
3
4
4
5
6
7
Here, d is based on either l - l i l - V or I* 111' V* while d* is based on II - IV - VI or I I ' - I V - V I ' .
This is because I I , IV, and VI arise from the hexagonal effect. Thus it seems of interest to list d as based on both
triangular pattern and hexagonal.
With this much as an introduction to the notion of a spiral grid for generating variations of a given sequence, we
shall close this account. Our purpose has been mainly to exhibit the results of some calculations and suggest possible
avenues of research. Various questions could be posed. For example: What can be said about divisibility properties
of the new sequences? What can be said about when such sequences will satisfy simple recurrence relations? Does
any of this shed light on when a Fibonacci number may be a figurate number? Can simple formulas be written for
the various derived sequences? What is a simple formula for the number we have called dl
REFERENCES
1. Martin Gardner, "Mathematical Games, The Remarkable Lore of the Prime Numbers," Scientific American, 210
(1964), March, No. 3, 120 et seq.
2. M.L. Stein, S.M. Ulam, and M.B. Wells, "A Visual Display of Some Properties of the Distribution of Primes,"
Amer. Math. Monthly, 71 (1964), pp. 516-520.
3. Oswald Wyler, Solution of Problem bmQ.Amer. Math. Monthly, 71 (1964), pp. 220-222.
4. John H.E. Cohn, "Square Fibonacci Numbers, etc.," The Fibonacci Quarterly, Vol. 2, No. 2 (April, 1964),pp.
109-113.