Summer's over

Let d ( n ) d(n) be the number of positive factors, including 1 1 and n n , of a positive integer n n . Find the sum of all n n such that d ( n ) = n 3 d(n) = \dfrac{n}{3} .

Any elegant proofs of why these are the only such values of n n are welcome but not required.


The answer is 51.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

4 solutions

The only such values of n n are 9 , 18 9, 18 and 24 24 , giving a sum of 51 51 . Now for a quick proof of why these are the only such values of n n .

Let n = i = 1 m ( p i ) a i n = \prod_{i=1}^m{(p_{i})^{a_{i}}} be the prime factorization of n n .

Now by induction we can show that 2 k ( k + 1 ) 2^{k} \ge (k + 1) for all natural numbers k k . Since 2 2 is the least prime we can then say that p k ( k + 1 ) p^{k} \ge (k + 1) for all primes p p and natural k k .

Now suppose n n has 7 7 as a prime factor. We can show by induction that 7 k > 3 ( k + 1 ) 7^{k} \gt 3(k + 1) for all natural k k . Then

n = 7 a j i j m ( p i ) a i > 3 ( a j + 1 ) i j m ( a i + 1 ) = 3 d ( n ) n = 7^{a_{j}} * \prod_{i \ne j}^m{(p_{i})^{a_{i}}} \gt 3(a_{j} + 1) * \prod_{i \ne j}^m{(a_{i} + 1)} = 3*d(n) .

This implies that if n n has 7 7 as a prime factor then n 3 d ( n ) n \ne 3*d(n) . This is in fact the case for all primes 7 \ge 7 , and so the prime factorization of n n is of the form 2 a 3 b 5 c 2^{a} * 3^{b} * 5^{c} for some non-negative integers a , b , c a,b,c .

Now we can prove by induction that 5 k > 3 ( k + 1 ) 5^{k} \gt 3(k + 1) for k 2 k \ge 2 , so c = 0 c = 0 or c = 1 c = 1 .

We can also prove by induction that 3 k > 3 ( k + 1 ) 3^{k} \gt 3(k + 1) for k 3 k \ge 3 , and so b = 1 b = 1 or b = 2 b = 2 . (Note that n n must be divisible by 3 3 so b 0 b \ne 0 .)

Yet another induction proof gives us that 2 k > 3 ( k + 1 ) 2^{k} \gt 3(k + 1) for k 4 k \ge 4 , and so a a has possible values of 0 , 1 , 2 0, 1, 2 and 3 3 .

(Sorry I'm leaving out the proofs by induction, but this solution is long enough as it is.)

Now if c = 1 c = 1 , then we would have 2 a 3 b 5 = 3 ( a + 1 ) ( b + 1 ) 2 2^{a} * 3^{b} * 5 = 3*(a + 1)*(b + 1)*2 . This would imply that 5 5 divides either ( a + 1 ) (a + 1) or ( b + 1 ) (b + 1) . But given the bounds we've established for a a and b b this is impossible, so we must have c = 0 c = 0 . Thus n = 2 a 3 b n = 2^{a} * 3^{b} .

Now if b = 1 b = 1 then

2 a 3 b = 3 ( a + 1 ) ( b + 1 ) 3 2 a = 6 ( a + 1 ) 2 a 1 = a + 1 a = 3 2^{a} * 3^{b} = 3(a + 1)(b + 1) \Longrightarrow 3*2^{a} = 6(a + 1) \Longrightarrow 2^{a-1} = a + 1 \Longrightarrow a = 3 .

This gives us n = 24 n = 24 as a solution.

If b = 2 b = 2 then we would have

2 a 9 = 9 ( a + 1 ) 2 a = a + 1 a = 0 , 1 2^{a} * 9 = 9(a + 1) \Longrightarrow 2^{a} = a + 1 \Longrightarrow a = 0,1 .

This gives us solutions n = 9 n = 9 and n = 18 n = 18 .

Thus the only possible values for n n are 9 , 18 9, 18 and 24 24 , giving us a sum of 51 \boxed{51} .

Shouldn't it be 2 a 3 b 5 = 3 ( a + 1 ) ( b + 1 ) 2 ? 2^{a}*3^{b}*5 = 3*(a + 1)*(b + 1)*2?

Spore Rome - 6 years, 8 months ago

Log in to reply

Yes, you're right. I'll just make the edit now. Thanks for catching that mistake. :)

Brian Charlesworth - 6 years, 8 months ago

Nice :D.Maybe we can show by this inequality argument that the equation d ( n ) . k = n d(n).k=n has only a finite number of solutions for a fixed positive integer k?

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

Yes, I was thinking about how to prove the general case as well. At first glance it looks as though this inequality argument could be used as a template, but there would be some details to work out, (as always). I was also wondering if there might be some formula for the number of solutions S ( k ) S(k) to the equation d ( n ) k = n d(n)*k = n for positive integers k k .

We have S ( 1 ) = 2 S(1) = 2 , since d ( 1 ) = 1 , d ( 2 ) = 2 d(1) = 1, d(2) = 2 . S ( 2 ) = 2 S(2) = 2 , since d ( 8 ) = 4 , d ( 12 ) = 6 d(8) = 4, d(12) = 6 are the only possibilities. From this question we have that S ( 3 ) = 3 S(3) = 3 . I'm getting S ( 4 ) = 1 S(4) = 1 , with d ( 36 ) = 9 d(36) = 9 , S ( 5 ) = 2 S(5) = 2 , with d ( 40 ) = 8 d(40) = 8 and d ( 60 ) = 12 d(60) = 12 . and S ( 6 ) = 1 S(6) = 1 , with d ( 72 ) = 12 d(72) = 12 . Continuing, S ( 7 ) = 2 S(7) = 2 , with d ( 56 ) = 8 d(56) = 8 and d ( 84 ) = 12 d(84) = 12 .

Hmmm.... This is interesting; a lot of further questions are coming to mind here. This seems like the type of problem that has already been dealt with, but if not, it's a potential thesis topic for someone. :)

Brian Charlesworth - 6 years, 8 months ago

Log in to reply

What if k is a rational number?

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

@Bogdan Simeonov Hadn't thought of that option. Well, my first thought is to flip things around and create (yet another) function k ( n ) k(n) which gives the value for k k in the equation d ( n ) k = n d(n)*k = n for a given positive integer n n .

So k ( 1 ) = 1 , k ( 2 ) = 1 , k ( 3 ) = 3 2 , k ( 4 ) = 4 3 , k ( 5 ) = 5 2 , k ( 6 ) = 3 2 k(1) = 1, k(2) = 1, k(3) = \frac{3}{2}, k(4) = \frac{4}{3}, k(5) = \frac{5}{2}, k(6) = \frac{3}{2} , etc..

Clearly k ( n ) 1 k(n) \ge 1 for all n n . For any prime p p we would have k ( p ) = p 2 k(p) = \frac{p}{2} , so k ( n ) k(n) would have no maximum as there is no greatest prime. So the set of all k ( n ) k(n) 's would be countably infinite but there would probably be a lot of gaps compared to the set of all rationals 1 \ge 1 . Again, more thesis material here. :)

Brian Charlesworth - 6 years, 8 months ago

Log in to reply

@Brian Charlesworth Oooh, by the way when k is a prime>3, I found that the only solutions are n=8p,12p

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

@Bogdan Simeonov Good observation. Indeed, for primes p > 3 p \gt 3 we have

d ( 8 p ) = d ( 2 3 p ) = 4 2 = 8 d(8p) = d(2^{3} * p) = 4*2 = 8 and

d ( 12 p ) = d ( 2 2 3 p ) = 3 2 2 = 12 d(12p) = d(2^{2} * 3 * p) = 3*2*2 = 12 .

The only solution pairs ( a , b ) (a,b) to

d ( 2 a 3 b p ) = ( a + 1 ) ( b + 1 ) 2 = 2 a 3 b d(2^{a}*3^{b}*p) = (a + 1)*(b + 1)*2 = 2^{a}*3^{b}

are ( 3 , 0 ) (3,0) and ( 2 , 1 ) (2,1) , as you have noted.

Brian Charlesworth - 6 years, 8 months ago

Log in to reply

@Brian Charlesworth Actually, the argument can be used for any power of a prime.Let n = d ( n ) . p a n=d(n).p^a .Now let n = p a + x . N n=p^{a+x}.N where x is a nonnegative integer and N is not divisible by p.Then p x . N = d ( p a + x . N ) p^x.N=d(p^{a+x}.N) , which leads to N d ( N ) = a + x + 1 p x \frac{N}{d(N)}=\frac{a+x+1}{p^x}

But now the left hand side must be greater than or equal to 1, which is true for only finitely many x and thus it is solvable.

When a=2, we are solving N d ( N ) = x + 3 p x \frac{N}{d(N)}=\frac{x+3}{p^x} .But if p>3 and x>0, then RHS is less than one, a contradiction.Thus if p>3, we need to solve N d ( N ) = 3 \frac{N}{d(N)}=3 , which we know has solutions N=9,18,24.That means that n = d ( n ) . p 2 n=d(n).p^2 as the only three solutions n = 9. p 2 , 18. p 2 n=9.p^2,18.p^2 and 24. p 2 24.p^2 for p>3.The cases p=2,3 can be solved easily.

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

@Bogdan Simeonov Another great observation. (I was a bit confused by " n = p ( a + x ) N n = p(a + x)*N " until I figured out you meant " n = p a + x N n = p^{a+x}*N ".)

So when a = 3 a = 3 , for p > 3 p \gt 3 we end up needing to solve N d ( N ) = 4 \frac{N}{d(N)} = 4 , which has solution N = 36 N = 36 . This means that n = d ( n ) p 3 n = d(n)*p^{3} has only the solution n = 36 p 3 n = 36*p^{3} for p > 3 p \gt 3 .

Things get more interesting as a a increases, but you've definitely found a method for dealing with powers of primes, which is a big step towards dealing with all integers via prime factorization. :)

Brian Charlesworth - 6 years, 8 months ago

Log in to reply

@Brian Charlesworth Yeah, sorry about that typo !I wonder if there is some multiplicative property we can use (Since we know how to solve it for the prime powers).

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

@Bogdan Simeonov Found something about the general case:Assume every prime p in the prime factorization of k is bigger than v p ( k ) + 2 v_p(k)+2 .Now let k = p 1 a 1 . . . p k a k , n = p 1 a 1 + x 1 . . . p k a k + x k . N k=p_1^{a_1}...p_k^{a_k},n=p_1^{a_1+x_1}...p_k^{a_k+x_k}.N .Then N d ( N ) \frac{N}{d(N)} = a 1 + x 1 + 1 p 1 x \frac{a_1+x_1+1}{p^x_1} ... a k + x k + 1 p k x \frac{a_k+x_k+1}{p^x_k} .But since p i > a i + 2 p_i>a_i+2 we need to have every x_i=0.So now we are dealing with N / d ( N ) = ( a 1 + 1 ) . . . ( a k + 1 ) N/d(N)=(a_1+1)...(a_k+1) which is exactly d(k).Well isn't this weird?!

Bogdan Simeonov - 6 years, 8 months ago

Log in to reply

@Bogdan Simeonov Weird and cool, although I'm still trying to work out the consequences of this observation. In the meantime, I've found the sequence for the function S ( k ) S(k) on OEIS here . (Note: it has the n n and k k the reverse of our usage.)

The associated subsequences listed in the "Comments" section are interesting as well. Apparently Paul Erdos has spent some time working on this, which means that it is a problem worth spending time on. :)

Brian Charlesworth - 6 years, 8 months ago

@Brian Charlesworth CD made another observation in his comment which was simple but clever.I wonder what else we can get from this problem :D

Bogdan Simeonov - 6 years, 8 months ago

Solution in 2 rows. The number of divisor of n n between n 4 \frac{n}{4} and n n is less or equal than 4 4 . So n 4 + 4 > n 3 \frac {n}{4}+4>\frac {n}{3} . It implies that n < 48 n<48 . We know that 3 n 3\mid n , so we only need check the cases.

How can you say that the number of divisors will be less than or equal to 4?

Pranjal Jain - 6 years, 8 months ago

Very nice bounding.

Calvin Lin Staff - 6 years, 8 months ago
C D
Sep 20, 2014

Another solution is to notice that d ( n ) d(n) can never be greater than n × 2 \lfloor\sqrt{n}\rfloor \times 2 because for every pair of factors of n n one must be above and one must be below n \sqrt{n} (both can be equal to s q r t n sqrt{n} but that complicates the solution).

d ( n ) d(n) = n × 2 \sqrt{n} \times 2 at numbers such as n = 2 , 6 , 12 n = 2,6,12 .

For d ( n ) = n 3 d(n) = \frac{n}{3} to be true n × 2 n 3 \lfloor\sqrt{n}\rfloor \times 2 \geq \frac{n}{3} must be true.

If n × 2 n 3 \lfloor\sqrt{n}\rfloor \times 2 \geq \frac{n}{3} is false for a perfect square (or equality exists) no value of n n higher than the perfect square exists. Therefore, we need to find the highest perfect square for which this is true. Because it is a perfect square, we have: n = n \lfloor\sqrt{n}\rfloor = \sqrt{n} . So we have:

n × 2 n 3 \sqrt{n} \times 2 \geq \frac{n}{3}

n × 6 n \sqrt{n} \times 6 \geq n

The highest value for which this is true is at n = 36 n = 36 . As equality exists, it is not true for n > 36 n > 36 . Now we only must check the values of n n below 37 which are divisible by 3.

We find n = 9 , 18 n=9,18 and 24 24 are the only solutions the answer is 51 \boxed{51} .

We can extend this to the general case: d ( n ) × k = n d(n) \times k = n , where k k is a fixed integer.

The maximum number of solutions for n n can be found by finding the highest square for which:

n × 2 k n \sqrt{n} \times 2k \geq n is true, which is at n = ( 2 k ) 2 n = (2k)^{2} .

n n must not be greater than ( 2 k ) 2 (2k)^{2} and must also be divisible by k k so the maximum value of solutions to d ( n ) d(n) is 4 k 4k . I believe it is possible to prove that the number of solutions cannot be greater than a value lower than 4 k 4k but cannot prove it myself.

I welcome any corrections/additions.

Well we can search for a function of n that is always less than or equal to d(n), because that would give us a lower bound on the solutions.

Bogdan Simeonov - 6 years, 8 months ago
Atomsky Jahid
Jun 16, 2016

Let, n = 3 k p i q i n=3^{k} \prod p_{i}^{q_{i}} where p i p_{i } is any particular prime number excluding 3. Now, d ( n ) = ( k + 1 ) ( q i + 1 ) d(n)=(k+1) \prod (q_{i}+1) The numbers we are trying to find must satisfy d ( n ) = n 3 d(n)=\frac{n}{3} or, ( k + 1 ) ( q i + 1 ) = 3 k 1 p i q i (k+1) \prod (q_{i }+1)=3^{k-1} \prod p_{i }^{q_{i }} or, k + 1 3 k 1 = p i q i ( q i + 1 ) \frac{k+1}{3^{k-1}}=\frac{\prod p_{i }^{q_{i }}}{\prod (q_{i }+1)} ...(1)

Now, p i q i ( q i + 1 ) 1 \frac{\prod p_{i }^{q_{i }}}{\prod (q_{i }+1)} \geq 1 where equality occurs when all q i = 0 q_{i }=0 or when p i q i = 2 1 \prod p_{i }^{q_{i }}=2^1 . Hence, the following must be true. k + 1 3 k 1 1 \frac{k+1}{3^{k-1}} \geq 1 This happens for k 2 k \leq 2 . As we can see k = 2 k=2 makes LHS of equation (1) equal to 1. How many ways can we make RHS equal to 1? In two ways, to be precise. One way is to make all q i = 0 q_{i}=0 and another way is to set p i q i = 2 1 \prod p_{i }^{q_{i }}=2^1 . These two cases respond to two numbers 3 2 = 9 3^2=9 and 3 2 2 1 = 18 3^2 2^1=18 . For k = 1 k=1 the LHS becomes equal to 2. To get 2 from RHS we need to have 2 as a prime in p i q i \prod p_{i }^{q_{i }} . A little exploration shows that p i q i = 2 3 \prod p_{i }^{q_{i }}=2^3 can make the RHS exactly equal to 2 and this corresponds to the number 3 1 2 3 = 24 3^1 2^3=24 . Therefore, the answer is 9 + 18 + 24 = 51 9+18+24=\boxed{51} .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...