Multiply Perfect Numbers of Three Different Primes

Author(s): R. D. Carmichael
Source: Annals of Mathematics , Oct., 1906, Second Series, Vol. 8, No. 1 (Oct., 1906), pp.
49-56
Published by: Mathematics Department, Princeton University

Stable URL: https://www.jstor.org/stable/1967428

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

is collaborating with JSTOR to digitize, preserve and extend access to Annals of Mathematics

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
 MULTIPLY PERFECT NUMBERS OF THREE DIFFERENT PRIMES*

BY R. D. CARMICHAEL

THE object of the present paper is to complete the investigation of the

existence of multiply perfect numbers of three different primes.t
J. Westlund has shownt that the only such numbers of multiplicity 3 of
the form Plal P2W2 p3, where Pi < p2 < P3p, Pi, P2' s being different primes, are
23 3*.5 and 25. 3.7. I have shown ? that there is none of multiplicity 3 of
the form plaipasap23,2 where P1 P2 9P3 are as before; and also)1 that there
exist no multiply perfect odd numbers containing (only) three different
primes.
It is easily shown that the multiplicity of the numbers under considera-
tion is either 2 or 3. The following proposition completes the investigation
of the latter case:
There exist no multiply perfect numbers of multiplicity 3 of the form
Pl'P2 pas3+l, where Vi' P2' p3 are different primes and P1 <P2 <p3 and
as> 0.
For this case, we should have, corresponding to equation (3) of a pre-
vious note (see ANNALS, 1. C. P. 153), the following equation:
2%a+2-1
(1) p38++pf 3+ * * *+p3 + 1 = 2r3

where the values of r and s are to be considered, and r > 0 and 8 0. Since
ps > 3, it is of one of the forms 6n ? 1.

* Presented to the American Mathematical Society, September 8, 1906.

t It can be shown without difficulty that there exists no multiply perfect number which
is the power of a prime; also, that every multiply perfect number, containing only two different
primes, is of the form 2n-1 (2n - 1), where 2n - I is prime; and, moreover, that every such
number is a perfect number of multiplicity 2. This paper therefore completes the investigation
for numbers of fewer than 4 different primes.
$ ANNALS OF MATHEMATICS, vol. 2, p. 172; July, 1901.
? ANNALS OF MATHEMATICS, vol. 7, p. 158; April, 1906.
AMECAN MATHEMATICAL MONTHLY, Vol. 18, P. 85; 1906.
(49)

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
 50 CARMICHAEL [October

CASE 1. Letp3 = 6n + 1. Equation (1) now becomes

(2) (6ni + 1)2as+2 1 - 2r 1 3

(2) ~~~~~6n
Therefore,

(3) (6ut)2a3+I + + (as + 1) (2a3 + 1)6it + 2(a3 + 1) = 2r * 38.

Now, let a3 + 1 = 2A * 3A t, where t is not divisible by 2 or 3. Every term

of equation (3), except the last two, is divisible by 2A+2 . 3*+-1. Dividing
through by 2A+1 3's, we may write (q an integer quotient)

(4) 6q + t [3n (2A+1 * 3A t - 1) + 1] = 27-A-1 . 38-A

where 2a3 + 1 is replaced by its value 2A+1 3* t - 1. The left member of
(4) is not divisible by 3. Hence, 8 - p = 0. q is not less than 1; for it is
evidently positive and not zero. Also, 2,+1 * 3* . t - 1 - 3 and t - 1. Hence,

(5) 2r-A12 ?9n+7, n-1.

Since t is odd, the conditions of (4) and (5) can be satisfied only when n is
odd. In this case 6n + 1 may be written in the form 4m - 1. Substituting
this value in equation (2) and reducing, we have

(6) (4m)2as+2 + . + (a3 + 1) (2a3 + 1) (4m)2 - 2(a3 + 1) 4m

= 2r+1 . 38(2m - 1.).

Now, let m = 2. 1, I an odd factor. Then 2A++ +3 divides every term but the
last of the first member of (6) with an even quotient, and the last with an odd
quotient; hence it is the highest power of 2 in that member. Then we must
haveX+ x+ 3 =r+ 1; or, r-X-1=x+ 1. Again,since6n? 1=4m -1,
9n + 7 = 6m + 4. These values substituted in (5) give
2x+1 _ 6m + 4 = 2x+1 * 31 + 4,
a clearly impossible inequality. Hence, case I yields no numbers of the type
here considered.
CASE II. Let p3 = (in - 1. Substituting in equation (1), using the
fraction for first member, we have by reduction

(7) (6n)2aa+2 + * + (a3 + 1) (2a3 + 1) (6n)2 - 2(a3 + 1)6n

= 2r+1 . 38 (3n - 1).

Again put a3 + 1 = 2A * 31A t; also let n = 2x . 3Y * 1, where I is

bv 2 or 3. Here 2A+3x+3 * 3,+3y+2 divides every term of the first member of

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
1906] MULTIPLY PERFECT NUMBERS 51

(7) except the last two. Therefor

we may write (q being an integer, which is readily shown to be positive and
> 0)

(8) 22,+l . 32Y-+1 q + lt[2x . 3+1 1(2A+ 1. 3"t- 1) - 1]

= 2r-Akx-1 . 38-*-Y-1(2x 313+1 1 - 1).

The left member is not divisible by 3; neither is the quantity (2x * 3Y+1 1 - 1)
in the second member. Hence,

8- - y-1 0.

If x ?0, that is if n is even, we have 6n-1 = 4m-1, where m

- 2x-1 . 3y+1. Putting 4m - 1 for 6n - 1 as the value of ps in eq
we reproduce equation (6). With the present value of m, 2A+x+2 divides
every term except the last of the first member of this equation with an even
quotient, and the last with an odd quotient; hence it is the highest power of
2 in that member. Then we must have x + x + 2 = r + 1; or r-X-x- 1
= 0. This condition and 8 - - y - I = 0 above found reduces equation
(8) to the form

(8') 2sx+1 * 32y+1 * q + It[2x 3Y+1 1(2+l 3* 3 t-1)-1] = 2x 3Y+ 1-1,

which is clearly impossible, since q is positive and 2A+ 1 * 3 t - 1 = 2a3

+ 1 - 3. This absurdity is the result of the supposition that n is even.
If n is odd, 6n - 1 may be written in the form 4mn + 1. Substituting
this value in equation (1) and reducing, we have

(9) (4mn)2aa+l + * + (as + 1)(2a3 + 1)4nm + 2(a3 + 1) = 2r * 38,

Here 2A+1 divides every term but the last of the first member of (9) with an
even quotient, and the last with an odd quotient; hence, it is the highest
power of 2 in that member. Therefore X + 1 = r. Substituting this value
in equation (8), remembering that s - p - y - 1 = 0, and that now x
we have

(8") 2 . 32+1 q + It[3Y+1 1 (2A+1 . 3 ,t- 1)- 1]=3X+1 1 1,

which is clearly impossible. This case therefore yields no numbers of the

type under consideration. Hence the proposition as announced above.
We propose now to prove the following theorem:

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
 52 CARMICHAEL [Octobel

There exist no m
ferent primes and
Let the form be 2a, pa2pa3. From the definition of multiply perfect
numbers we readily have (by means of the formula for the sum of the divisors
of a number),
2aj+ _a I - pa2 + pa 2-1 +;**+ 2+1.p+P
(10) 2= a+~ p~2+p 2 +. *+ P2 , + 1 p"? *p3? 1

Either a2 or a3 is odd, as otherwise all the numerators would be odd, which

is inadmissible, the left member being even. Take a3 odd. (It is to be
noticed that in this case no assumption has been made with regard to any
relations between P2 and p3; and hence to take specifically a8 odd causes no
loss of generality.) Equation (10) may now be written

2a,+1
Sal
_ 1 p2
pan
p2 + (p3?
p
1)(1 +?p +p + *+p

Suppose, first
in the numerator above has an odd number (> 1) of terms, it can always be
separated into two different odd factors, of the same degree in pR and e
prime to p3; thus, in the case supposed, a3-1 = ; an integer; the
factors are therefore

1+p +p3 + p s + *p * +&(-3+3p + *) _ *V3 +_l)

31 p+P3+* +3.

These two odd factors, being each prime to p3, introduce at least two odd
primes different from p3, unless they are both powers of the same prime.
First suppose the latter case. Let r be a prime different from PA, and let
+ 3+ p2 + ***+ 2xt = .a ; 1 3+ _p2 _ p3 + + . . 2x
where ,3 < a. Since each quantity is a power of the prime r, their dif-
ference is divisible by the less. Hence,

PA')3 + P3 p3 p . . ~+' p2Z)=an integer.

Now 203 is prime to the denominator; hence the quantity in parenthesis in

the numerator must be a multiple of the denominator. Call the quantity
in parenthesis N and the denominator D. Then
2xIVxl3
p3 _ - < @

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
1906] MULTIPLY PERFECT NUMBERS 53

Also,

ff p2w- =(.p2 1)_y + l =pi N-

3 P~~13 123'~ 3
for p3 _ 3. But
j2jx-1 <p~2X-1 ( -1) =2x -2x-1
P3 3 (P3 =13 P3
Hence, N < D; therefore TN is not a multiple of D. Hence 1 + p3 + p3
? * ** +p = and 1-p3 + p2 -p3 + + pax are not powers of the same
prime. These two factors therefore introduce (at least) two odd primes
different from p3, which is inadmissible. Hence, only two conclusions are
now possible, namely: the last factor in the numerator of (11) contains
only one term, in which case a3 = 1; or, it contains an even number of
terms.
Take first the case when there is only one term in the expression.
That term is unity. For this case equation (10) may be written

(12) 2 =2ai+1
.2 _1
2'21A
+ 1)2?+.
PAP++1

Now 2a1+1 _ 1 necessarily contains the f

factor 2,L- 1 (which is a prime, pa), which it does in case 12 2 + + P2 +
is a power of 2. Hence,

p2a2= 2a+ 1 - 1, or = (21A- 1)-1(2a1+1-1)l

In the former case, we should also have

]42 + + P2 + 1 = 2a1+1-M (2- 1) = 2a1 - 2a+1-.

The two equations are mutually exclusive. Hence we have left to consider
the case
2ai+1-I - 1
p2 = (2- 1)-1(2a +_ 1) = 2at+- + - 1

This case also requires

2 + P2 +I = 2-
The last two equations are mutually exclusive.
We must therefore consider the other case, that in which 1 + p2 + p4
+ * + pas-1 contains an even number of terms. The expression is then
divisible by I + p2, which introduces a third prime; that is, 12. The result

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
 54 CARMICHAEL [October

of dividing out this factor is 1 +14 +p8 + * .. If this were divisible by

1 +p4 a fourth prime would be introduced; for, 1 + p4 is prime to p3 and
contains only the factor 2 in common with p2 + 1, and is moreover not a
power of 2. This fourth prime is inadmissible. Hence, 1 + p4 + p8 +
must contain an odd number of terms; for otherwise it is divisible by
1 + 4. If this number of terms is greater than 1, the expression may be
separated into two odd factors, each prime to p3, introducing at least one
other prime. (Compare similar case above.) This is inadmissible; hence, the
expression can contain only one term, and this is therefore equal to 1. Hence,
a3 ?I 3; and, therefore, since it is odd and the value 1 has already been ex-
cluded, a3 = 3, when p8 + 1 is a power of 2.
Substituting in equation (10) the values a3 = 3, pi + 1 = 2*, we have

2 2a+1 - 1 1 +P2 + * * +p 2,-+1(22*-1 - 21 + 1)

Sal pat ~ ~ (2* - 1 )3
Now, if a2 is even, 1 + P2 + * + p2a is odd and a, = g

1 +P2 + * * * +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~pa,, = (2, - 1)3

and therefore pv"= (21L+ -- 1)(22I-1 - 2,L + 1). Expanding these two
equations, we have respectively

1+P2++ ***+P2= 23*- 3 * 22*+3 * 2M 1;

a. = 2 31L- 221%+1 - 22*-1 + 2,*+1 + 2,% - 1.

By the former of the last two equations IA> 1; for such values of , the
equations become mutually exclusive. Hence, when p3 + 1 is a power of 2, a2
is odd.
Now, if P2 + 1 is a power of 2, - 2, then a2 = 3, by the same reasoning
as above for a8. For this case we have from equation (10),

(13) 2= 2ai+1 _ 1 2A+1(22A-I - 2A + 1) 2*+1(22I-- - 21 + 1)

(1) = 2%l (2Ak 1)3 (2* 1)3
By clearing of fractions and reducing it is easily shown that this equation can
not be satisfied in any case. Hence, p3 + 1 and P2 + 1 are not both powers
of 2.
If we supposed either P2 + 1 orp3 + 1 to be not a power of 2, a course
of reasoning very similar to that by which, in case p3 + 1 is a power of 2, it

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC6 UTC
All use subject to https://about.jstor.org/terms
1906] MULTIPLY PERFECT NUMBERS 55

was shown above that as = 3, would bring us to the conclusion that the cor-
responding a equals 1, provided that that a is odd.
Suppose now that P2 + 1 is not a power of 2. Also letp3 + l = 2* as
before. Then, a2 is odd, as seen above; and is hence = 1. Substituting now
in equation (10), we have

(14) 2 2a1+1 - 1 P2 + 1 21L+1(22*-1 - 2* + 1)

- 2a_ - R2 (2's - 1)3

whence, necessarily, P2 = 22*-1 - 2, + 1. Writing this value for P2 in equa-

tion (14) and simplifying, we have

(15) (21 - 1)3 = (21*+2 - 1) (221-2 - 21-1 + 1)

since, by (14), a, = 1A + 1. No value of , satisfies

We have now examined every possible case in which p3 + 1 is a power
of 2, and a3 is odd, and find no numbers of the type in consideration; simi-
larly we may completely dispose of the case in which P2 + 1 is a power of 2
and a2 odd; and that without assuming in this case that a3 is odd. And since
necessarily a2 or as is odd, this effectually disposes of every case in which
either P2 + 1 or ps + 1 is a power of 2, unless it can happen that o
even while the corresponding p + 1 is a power of 2, a case which is shown
below to be impossible. Hence, we have left only the case in which neither
P2 + 1 nor p3 + 1 is a power of 2. As already seen, a2 or as must be odd.
We may, without loss of generality, take as to be odd. Then, as we have
seen above, as = 1. Now, suppose a2 to be even; say a2 = 2v. Substituting
these values in equation (10) we have

2 2a-+1_1 p ++p v2I + . . +p2

Sa2v- a
+ 1 p3 +

Since pa + 1
tain no odd prime butp2. Hence, p3 + 1 = 2?Tp'/, where x > 0, y > 0. Also,
y < 2v; for, to satisfy the equation above we also require 2a2+1 _ 1 = some
power Of P2. Also p, = p2v + * * + p2 + 1, for this last member does
contain 2. These values for p3 + 1 and p3 are mutually exclusive. Hence,
in this case also, a2 is odd. Hence, as for the case a2 odd, p2 + 1 is not
a power of 2; and, therefore, by previous considerations, a2 = 1. (This also
shows to be impossible the case above supposed; namely, one a even and the

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms
 56 CARMICHAEL

correspondingp + 1 a power of 2; for it is now seen that since one a is ne-

cessarily odd and its correspondingp + 1 not a power of 2, the other a is odd
also. Of course a here refers to either a2 or a3, but not to a1.)
Substituting in equation (10) the values a2 = 1, a3 = 1, above developed,
we have

(16) 2 = 2ai P2 PA +

Now, in this case, neither P2 + 1 nor p3 + 1 is a power of 2. Hence, the

numerator has three odd factors, which may or may not be different. In either
case these three factors can not be divided out by two odd primes. This case
therefore yields no numbers of the type in consideration.
This completes the investigation of the several cases of our last theorem.
We are now able to state in conclusion the following general proposition, all
cases having been examined either in this paper or in those already referred
to :-
There are but two multiply perfect numbers of three different primes;
and these two are 23 * 3 * 5 and 25 * 3 * 7. The multiplicity of each is 3.

HARTSELE, ALABAMA.

This content downloaded from

49.145.226.222 on Sun, 05 Sep 2021 00:07:49 UTC
All use subject to https://about.jstor.org/terms