#A67 INTEGERS 19 (2019) CARDINALITY OF A FLOOR FUNCTION SET Randell Heyman School of Mathematics and Statistics, University of New South Wales, Sydney, New South Wales, Australia [email protected] Received: 5/13/19, Revised: 9/16/19, Accepted: 12/22/19, Published: 12/24/19 Abstract Fix a positive integer X. We quantify the cardinality of the set X/n : 1 n X . We discuss restricting the set to those elements that are prime,{bsemiprimec or similar. } 1. Introduction Throughout we will restrict the variables m and n to positive integer values. For any real number X we denote by X its integer part, that is, the greatest integer b c that does not exceed X. The most straightforward sum of the floor function is related to the divisor summatory function since X = 1 = ⌧(n), n nX6X ⌫ nX6X k6XX/n nX6X where ⌧(n) is the number of divisors of n. From [2, Theorem 2] we infer X = X log X + X(2γ 1) + O X517/1648+o(1) , n − nX6X ⌫ ⇣ ⌘ where γ is the Euler–Mascheroni constant, in particular γ 0.57722. ⇡ Recent results have generalized this sum to X f , n nX6X ✓ ⌫◆ where f is an arithmetic function (see [1], [3] and [4]). In this paper we take a di↵erent approach by examining the cardinality of the set X S(X) := m : m = for some n X . n ⇢ ⌫ Our main results are as follows. INTEGERS: 19 (2019) 2 Theorem 1. Let X be a positive integer. We have S(X) = p4X + 1 1. | | − j k Theorem 2. We have S(X) = 2pX + O(1). | | 2. Proofs of Main Theorems Throughout let 1 + p4X + 1 b = − , (1) 2 and note that X = b + 1. b We define 2 sets: X S (X) = m : m = , m b (2) 1 n ⇢ ⌫ and X S (X) = m : m = , n(n 1) X . (3) 2 n − ⇢ ⌫ We first show that S(X) = S (x) S (x), which will allow the use of the inclusion- 1 [ 2 exclusion principle to formulate the cardinality of S(X). It is clear that S (X) 1 ✓ S(X) and S (X) S(X). So S (X) S (X) S(X). Next, if m S(X) then 2 ✓ 1 [ 2 ✓ 2 m = X/n with either n(n 1) X or n(n 1) > X. If n(n 1) X then − − − m S (X). If n(n 1) > X then, solving the quadratic equation, we have 2 2 − 1 + p4X + 1 n > = b + 1. 2 Then X X m = = b. n n ⌫ So m S (X) and therefore S(X) S (X) S (X), which proves that S(X) and 2 1 ✓ 1 [ 2 S (X) S (X) are equal. 1 [ 2 Now, using the inclusion-exclusion principle, S(X) = S (X) + S (X) S (X) S (X) . (4) | | | 1 | | 2 | − | 1 \ 2 | INTEGERS: 19 (2019) 3 We now calculate the number of elements of S1(X). Let m be an arbitrary positive integer with 1 + p4X + 1 m b = − . 2 This means that m2 + m X 0, − which implies that m(m + 1) X. Thus X 1 m(m + 1) ≥ and therefore X X 1. m − m + 1 ≥ X X Since the interval from m to m+1 has length at least 1, there must be an integer n such that X X < n , (5) m + 1 m from which X n < mn X. − In turn this implies that X X 1 < m . n − n This means that m = X/n and so m S (X). From (2) there are b possible b c 2 1 b c values of m. From (5) we see that we can always find an n to give us any of these values of m. Therefore the numbers 1, 2, . , b are the only elements of S (X) and b c 1 so S (X) = b . (6) | 1 | b c We now consider the cardinality of S (X). We show that n(n 1) < X implies 2 − X/n and X/(n 1) are distinct. We have b c b − c X X X X X X = + , n 1 − n n 1 − n − n 1 n − ⌫ ⌫ − ⇢ − ⇢ where represents, as usual, the fractional part of the real number. So {·} X X X = + t, (7) n 1 − n n(n 1) − ⌫ ⌫ − where t ( 1, 1). Recalling that n(n 1) X, we have 2 − − X 1. n(n 1) ≥ − INTEGERS: 19 (2019) 4 X X Substituting into (7) we see that n 1 n > 0, which implies that X/n − − b c and X/(n 1) are distinct. Sincej n(nk 1) X we have, solving the quadratic b − c − ⌅ ⇧ equation, X n (8) b and so X S (X) = = b + 1. (9) | 2 | b b c ⌫ We now consider S (X) S (X) . We have seen that | 1 \ 2 | S (X) = 1, 2, , b . 1 { · · · b c} From (8) we see that X n = b + 1. b So the values of n in S (X) are 1, 2, b + 1 and therefore 2 · · · b c X X S (X) = , , , X . 2 b + 1 b · · · ( b c⌫ b c⌫ ) The set S (X) S (X) will be non-empty if 1 \ 2 X = b d , b + 1 c b − c b − c⌫ for some c, d 0. From this we deduce that X < b + 1 d, so that X < ≥ b+1 c − ((b + 1) d)(b + 1 c). Recalling that X = b(b + 1)−we have − − b + 1 c(b + d + 1) d(b + 1) > 0, − − which is only possible if c = d = 0. Thus there will be at most one element of S (X) S (X) and this one element will occur if, and only if, 1 \ 2 X = b . b + 1 b c b c⌫ In fact, X S (X) S (X) = 1 b . | 1 \ 2 | − b + 1 − b c ✓b c⌫ ◆ Combining this equation with (4), (6) and (9) and simplifying we obtain X S(X) = b + . | | b c b + 1 b c⌫ INTEGERS: 19 (2019) 5 Removing b, by using 1, we have 1 + p4X + 1 X S(X) = − + . | | 2 6 1+p4X+1 7 ⌫ 6 2 7 6 7 4j k5 So to prove Theorem (1) it will suffice to show that 1 + p4X + 1 X − + p4X + 1 + 1 = 0. (10) 2 6 1+p4X+1 7 − ⌫ 6 2 7 6 7 j k 4j k5 Let a be the largest square number less than or equal to X. So X = a2 + s, 0 s < a. We show that (10) is true in 2 all-encompassing cases. For the first case let a2 + s + 1/4 1 a. We have − 2 ≥ 1 +pp4X + 1 1 + 4(a2 + s) + 1 − = − 2 2 ⌫ $ p % 1 = a2 + s + 1/4 = a, (11) − 2 ⌫ p and 1 + p4X + 1 1 + 4(a2 + s) + 1 1 = = a2 + s + 1/4 + = a + 1. 2 2 2 ⌫ $ p % ⌫ p So X a2 + s a s = = a − = a 1. (12) 6 1+p4X+1 7 a + 1 − a + 1 − 6 2 7 ⌫ ⌫ 6 7 4j k5 Also, p4X + 1 = 4a2 + 4s + 1 = 4a2(1 + s/a + 1/a2) = 2a. (13) j k jp k j k Substituting (11), (12) and (13) into (10) shopws that Theorem (1) is true in this case. For the second case let a2 + s + 1/4 1 < a. We have − 2 1 + p4Xp+ 1 1 + 4(a2 + s) + 1 − = − 2 2 ⌫ $ p % 1 = a2 + s + 1/4 = a 1, (14) − 2 − ⌫ p INTEGERS: 19 (2019) 6 and 1 + p4X + 1 1 + 4(a2 + s) + 1 1 = = a2 + s + 1/4 + = a. 2 2 2 ⌫ $ p % ⌫ p So X a2 + s s = = a + = a. (15) 6 1+p4X+1 7 a a 6 2 7 ⌫ 6 7 j k 4j k5 Also, p4X + 1 = 4a2 + 4s + 1 = 4a2(1 + s/a + 1/a2 = 2a. (16) j k jp k jp k Substituting (14), (15) and (16) into (10) shows that Theorem (1) is true in this case. This completes the proof of Theorem 1. Theorem 2, that is S(X) = 2pX + O(1), | | follows immediately from Theorem 1. 3. Discussion 3.1. Primes in S(X) We can generalize S(X) by considering elements of S(X) that are divisible by some positive integer d X. This is interesting in its own right but could also form the basis for calculating something much more interesting: the number of primes, semiprimes or similar in S(X). Let X X S (X) = m : m = for some n X, d . d n | n ⇢ ⌫ ⌫ A standard approach to express Sd(X) would be to follow a path involving an indicator function, di↵erences of floor functions, the function and exponential sums, hoping that we can bound the exponential sums (here (y) = y y 1/2). − b c − Unfortunately this is not the case here. The process yields the following result. Lemma 3. We have X X 4X1/2 b b c rd X X Sd(X) = + + O(1), | | 3d X r dj + 1 − dj r=1 j= − ✓ ✓ ◆ ✓ ◆◆ X Xrd where a = X/b. INTEGERS: 19 (2019) 7 A proof is given in the Arxiv version of this paper (see [5, Section 5]). Calculating various sums using Maple suggests that the double sum cannot successfully be bound. In fact Maple suggests that the double sum is asymptotically equivalent to 2X1/2/3d. If this argument is correct then 2X1/2 S (X) , d ⇠ d as one would expect heuristically. 3.2. Trivial Bounds In the absence of a better approach we outline some trivial bounds on S (X) .