rec.puzzles Archive

Which numbers n have more divisors than any smaller number, and how large is the number of divisors d(n)...

arithmetic/highly.composite New

New entry, edited from material collected for the archive and not yet in a released version.

Problem

Which numbers n have more divisors than any smaller number, and how
large is the number of divisors d(n) for n up to 100,000,000?

[ Editor: the opening sentence above is the editor's summary; the
original posting follows. ]

The divisor function d(n) is the sum of number of divisors that divide
integrally a number n.  The first few values of d(n), such that both n
and d(n) are larger than the previous, are given in table 1.

                      Divisor Function d(n)
     n     d(n)            n      d(n)            n       d(n)
     1       1             24       8            180       18
     2       2             36       9            240       20
     4       3             48      10            360       24
     6       4             60      12            720       30
    12       6            120      16            840       32

                           Table 1.

Can anyone extend this table to cover n<=100,000,000 or provide a
source if this information is available?  d(n) is needed in a C-program
to allocate space for some arrays, and some idea of its size is needed.

-- Randall L. Rathbun <Randall.Rathbun@SanDiego.NCR.COM>
Show solution
The numbers you are interested in are frequently referred to as
`highly composite numbers', and have been studied by Ramanujan and
Erdos, among others.  The two sequences you want appear as numbers 196
and 385 in Sloane's "A Handbook of Integer Sequences".  Ramanujan was
rather fond of calculating himself and probably calculated many values;
see his "Collected Papers", p. 78-128, or his article "Highly Composite
Numbers", Proc. London Math. Soc. 14 (1915) 347-400.  (I would be
willing to bet that a table exists in these articles, but since both
books have mysteriously disappeared from our math library, I am unable
to verify this.)

About ten years ago, I wrote a little APL program to compute these
numbers.  I don't absolutely guarantee that the numbers are correct,
but at least they match the numbers in Sloane's table as far as
d(n) = 336.

 HIGHLY COMPOSITE NUMBERS

 d(n)       n
  1         1
  2         2
  3         4
  4         6
  6        12
  8        24
  9        36
 10        48
 12        60
 16       120
 18       180
 20       240
 24       360
 30       720
 32       840
 36      1260
 40      1680
 48      2520
 60      5040
 64      7560
 72     10080
 80     15120
 84     20160
 90     25200
 96     27720
100     45360
108     50400
120     55440
128     83160
144    110880
160    166320
168    221760
180    277200
192    332640
200    498960
216    554400
224    665280
240    720720
256   1081080
288   1441440
320   2162160
336   2882880
360   3603600
384   4324320
400   6486480
432   7207200
448   8648640
480  10810800
504  14414400
512  17297280
576  21621600
600  32432400
640  36756720
672  43243200
720  61261200
768  73513440
800 110270160

-- Jeff Shallit, Department of Computer Science, University of Chicago

I suggest you read the Thesis of Guy Robin of the University of Limoges
(France), who studied these numbers and worked out an algorithm that
generates all highly composite numbers (as one calls them) between N0
and N1.  The address of the University is :

U. F. R. des Sciences
Departement de Mathematique
123 avenue Albert Thomas
F-87060 Limoges Cedex
France