For a finite non-empty set of primes , let be the size of the set , and be the largest possible number of consecutive positive integers, each of which is divisible by at least one member of .
(i) Show that , with equality if and only if ;
(ii) Show that .
Solutions — 2
Solution 1
In the sequel we will consider being made of the primes , with .
a. By the Chinese Remainder Theorem there will exist some such that , hence . Then the set of consecutive integers has the desired property, hence . When , within any set of consecutive integers at most one is divisible by any , hence by the Pigeonhole Principle there will be one not divisible by any of the primes in . On the other hand, when , we will make again use of the Chinese Remainder Theorem, so there will exist some such that , hence , where and the extra requirement that . It follows that the set of consecutive integers has the desired property, hence .
b. Now, let a set made of consecutive integers have the desired property. For any , the number of its elements which are divisible by will satisfy the inequality
Then, by the Principle of Inclusion/Exclusion, one has
The first term is clearly equal to , while the second is equal to
therefore , and so will be .
Solution 2
Alternative Solution (ii). (F. Chindea)
Not only an elegant alternative approach, but vastly improving on the required bound. Start by proving the following
LEMMA. If the integer arithmetic sequences , , are such that they cover consecutive integers, then they cover .
(Particular case of a Crittenden & Vanden Eynden Theorem)
Proof. Let be a primitive root of unity. Then if and only if , hence is divisible by at least one if and only if . Denote , , .
Consider the polynomial , of degree , and with , . Then for any integer , hence
Assume the consecutive integers are covered, so, according with the above, . Since and , by simple induction one gets for any integer (a linear recurrence relation of order containing consecutive null terms generates an all-null sequence).
Now, in our case, if we assume , then the arithmetic sequences given by , , will cover consecutive integers, hence they will cover . But clearly a large enough prime is not divisible by any , so this is absurd, therefore .