Citation: Wanxi Yang, Mao Li, Yulu Feng, Xiao Jiang. On the integrality of the first and second elementary symmetricfunctions of 1,1/2s2,...,1/nsn[J]. AIMS Mathematics, 2017, 2(4): 682-691. doi: 10.3934/Math.2017.4.682
[1] | Yulu Feng, Min Qiu . Some results on p-adic valuations of Stirling numbers of the second kind. AIMS Mathematics, 2020, 5(5): 4168-4196. doi: 10.3934/math.2020267 |
[2] | Aleksa Srdanov . Fractal form of the partition functions p (n). AIMS Mathematics, 2020, 5(3): 2539-2568. doi: 10.3934/math.2020167 |
[3] | Phakhinkon Phunphayap, Prapanpong Pongsriiam . Explicit formulas for the $p$-adic valuations of Fibonomial coefficients II. AIMS Mathematics, 2020, 5(6): 5685-5699. doi: 10.3934/math.2020364 |
[4] | Gunduz Caginalp . A renormalization approach to the Riemann zeta function at — 1, 1 + 2 + 3 + …~ — 1/12.. AIMS Mathematics, 2018, 3(2): 316-321. doi: 10.3934/Math.2018.2.316 |
[5] | Ziyu Dong, Zhengjun Zhao . An application of $ p $-adic Baker method to a special case of Jeśmanowicz' conjecture. AIMS Mathematics, 2023, 8(5): 11617-11628. doi: 10.3934/math.2023588 |
[6] | Ling Zhu . Completely monotonic integer degrees for a class of special functions. AIMS Mathematics, 2020, 5(4): 3456-3471. doi: 10.3934/math.2020224 |
[7] | Xi-Fan Huang, Miao-Kun Wang, Hao Shao, Yi-Fan Zhao, Yu-Ming Chu . Monotonicity properties and bounds for the complete p-elliptic integrals. AIMS Mathematics, 2020, 5(6): 7071-7086. doi: 10.3934/math.2020453 |
[8] | Fei Wang, Bai-Ni Guo, Feng Qi . Monotonicity and inequalities related to complete elliptic integrals of the second kind. AIMS Mathematics, 2020, 5(3): 2732-2742. doi: 10.3934/math.2020176 |
[9] | Naila Mehreen, Matloob Anwar . Some inequalities via Ψ-Riemann-Liouville fractional integrals. AIMS Mathematics, 2019, 4(5): 1403-1415. doi: 10.3934/math.2019.5.1403 |
[10] | Yan Ma, Di Han . On the high-th mean of one special character sums modulo a prime. AIMS Mathematics, 2023, 8(11): 25804-25814. doi: 10.3934/math.20231316 |
A well-known result in number theory states that the harmonic sum
Hn(1):=n∑k=11k |
is never an integer for n>1. The first published proof went back to 1915 by Leopold Theisinger. In 1946, Erdös and Niven proved that the multiple harmonic sum
Hn({1}r)=∑1≤k1<⋯<kr≤n1k1⋯kr |
is not an integer with finite exceptions. In 2012, Chen and Tang showed a stronger result stating that Hn({1}r) is an integer only for (n,r)=(1,1) and (n,r)=(3,2).
For an n-tuple vector s=(s1,s2,⋯,sn) with si∈Z+ and n≥2, we define the first ordinary multiple harmonic sum H(1)n(s1,…,sn) as
H(1)n(s1,…,sn):=n∑i=11isi, |
and the second ordinary multiple harmonic sum H(2)n(s1,…,sn) as
H(2)n(s1,…,sn):=∑1≤i<j≤n1isijsj, |
and the second star multiple harmonic sum H∗(2)n(s1,…,sn) as
H∗(2)n(s1,…,sn)=∑1≤i≤j≤n1isijsj. |
We shall prove that H(1)n(s1,…,sn), H(2)n(s1,…,sn) and H∗(2)n(s1,…,sn) are not integers except some special cases.
Let p be a prime and vp(q) be the p-adic valuation of rational number q, that is, if q=apnbpm with gcd(a,p)=gcd(b,p)=1 and m,n∈Z+, then vp(apnbpm)=n−m. It is well known that the following two statements are true:
(1). For any x,y∈Q, one has
vp(x+y)≥min(vp(x),vp(y)), |
and the equality holds if vp(x)≠vp(y).
(2). For any x1,x2,…,xn∈Q, one has
vp(x1+x2+⋯+xn)≥min(vp(x1),vp(x2),…,vp(xn)), |
and the equality holds if there exists an i such that vp(xi)<vp(xj) for all integers j with j≠i.
To express concisely, we take the following abbreviations: H(2)n:=H(2)n(s1,…,sn), H∗(2)n:=H∗(2)n(s1,…,sn). We denote the sum H(2)n when si is fixed and sj→∞ by H(2)n(si=k) and H(2)n(sj→∞) respectively.
Now we state our main results.
Theorem 1.1. Let n be an integer with n≥2 and si∈Z+ for 1≤i≤n. Then H(1)n is never an integer.
Theorem 1.2. Let n be an integer with n≥2 and si∈Z+ for 1≤i≤n. Then each of the following is true:
(ⅰ). H∗(2)n is never an integer.
(ⅱ). H(2)n is never an integer except that n=3,s2=s3=1, in which case, H(2)n is an integer.
Remark 1.1. We can get an approximating value by the following expansion
H(2)n=12((n∑i=11isi)2−n∑i=11i2si). |
Actually, if we fix the valuation of n and s1,…,sn, we can get an approximate value by hand from the expansion above.
In the next two sections, we first show that when n=9 and n=21, H(2)n is not an integer, i.e. H(2)9 and H(2)21 are not integers. And the remaining cases will be treated in the final section. The proofs of Theorems 1.1 and 1.2 are also given in the final section.
In this section, we show the fact that H(2)9 is not an integer. At first, we have following lemmas.
Lemma 2.1. For a fixed prime p and a positive integer n with p<n, let k:=[np],
Tp:=∑1≤i<j≤n,vp(i)vp(j)≥11isijsj |
and
m:=min(−vp(psp),−vp((2p)s2p),…,−vp((kp)skp)). |
If vp(Tp)<m, then H(2)n∉Z+.
Proof. Since vp(1isijsj)≥m for any 1≤i≠j≤n with vp(i)vp(j)=0, it then follows that
vp(H(2)n−Tp)≥m. |
On the other hand, since vp(Tp)<m, we have vp(Tp)<m≤vp(H(2)n−Tp). Thus
vp(H(2)n)=vp(Tp+(H(2)n−Tp))=vp(Tp)<0, |
the last inequality is true because vp(Tp)<m<0. So H(2)n is not an integer.
Remark 2.1. Let tp=Tppsp(2p)s2p⋯(kp)skp. Then
vp(tp)=vp(Tp)+k∑i=1vp((ip)sip). |
So vp(Tp)<m if and only if
vp(tp)<m+k∑i=1vp((ip)sip). | (2.1) |
Remark 2.2. For a fixed k with 2≤k≤n, if sk<s′k, then
H(2)n(s1,…,sk,…,sn)>H(2)n(s1,…,s′k,…,sn), |
Moreover, for any tk satisfies sk≥tk, we denote all H(2)n(tk) as H(2)n(sk≥tk), then
H(2)n(sk→∞)<H(2)n(sk≥tk)≤H(2)n(sk=tk). |
Lemma 2.2. We have that H(2)9 is not an integer.
Proof. By Lemma 2.1 and Remark 2.1, we consider the cases satisfying (2.1) for p=3. Namely,
v3(t3)<m+s3+s6+2s9=min(s3+s6,s3+2s9,s6+2s9), | (2.2) |
where m=min(−s3,−s6,−2s9), t3=3s3+6s6+9s9. Next, we consider the following cases.
Case 1. s3=s6=2s9. Since s3=2s9≥2, so 2+2s3<3s3.Then we have t3=3s3+6s3+32s9=3s3(2+2s3)<32s3, which implies that
v3(t3)<2s3=min(s3+s6,s3+2s9,s6+2s9) |
Case 2. There is only one of s3,s6,2s9 equals to min(s3,s6,2s9). Hence it is obvious that v3(t3)=min(s3,s6,2s9)<min(s3+s6,s3+2s9,s6+2s9).
Case 3. s6=2s9<s3. Then one has t3=3s3+32s9(1+22s9). But 1+22s9=1+4s9≡2(mod3), so we deduce that
v3(t3)=v3(3s3+32s9(1+22s9))=v3(32s9(1+22s9))=2s9<4s9=min(s3+s6,s3+2s9,s6+2s9) |
Case 4. s3=2s9<s6. Thus t3=2×32s9+6s6=2×32s9+2s63s6, since 3 does not divide 2 or 2s6, then it follows that
v3(t3)=v3(2×32s9)=2s9<4s9=min(s3+s6,s3+2s9,s6+2s9). |
Case 5. s3=s6<2s9. We divide the proof into the following subcases.
Case 5.1. 2≤s3<2s9<2s3. Then we have t3=3s3+6s3+32s9≤3s3(1+2s3)+32s3−2=3s3(1+2s3+3s3−2). Since s3≥2, so 1+2s3+3s3−2<3s3, which implies that
v3(t3)<2s3=min(s3+s6,s3+2s9,s6+2s9) |
as excepted.
Case 5.2. 2≤s3<2s3≤2s9. One has
v3(t3)=v3(3s3(1+2s3)+32s9)=v3(3s3(1+2s3))<2s3=min(s3+s6,s3+2s9,s6+2s9). |
Case 5.3. 1=s3=s6<2s9. In this case, we consider H(2)9(s2=4,s3=s4=s5=s6=s7=s8=s9=1). Then it is easy to see that
v3(t3)=2=min(s3+s6,s3+2s9,s6+2s9), |
i.e. (2.2) does not hold any more. Thus we have to investigate (2.1) for another prime. Actually, we choose p=2, that is, we will prove the following inequality
v2(t2)<min(s2+2s4+1,s2+1+3s8,2s4+1+3s8,s2+2s4+3s8). |
Notice that 3s8,s2,2s4≥1, then we only need to show
v2(t2)<min(s2+2s4,s2+3s8,2s4+3s8)+1. | (2.3) |
Subcase 5.3.1. s2=1. Then we deduce that t2=22×3+22s4+3+23+3s8+22s4+3s8 which implies that
v2(t2)=2<2+min(2s4,3s8)=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
Subcase 5.3.2. s2>1, and there's only one of s2,2s4,3s8 equals to min(s2,2s4,3s8). Just like Case 2 stated, v2(t2)=min(s2,2s4,3s8)+1, so
v2(t2)<min(s2+2s4,s2+3s8,2s4+3s8)+1. |
Subcase 5.3.3. 3s8=2s4<s2. Then we have t2=2s2+2s4+1+24s4+3×2s2+1+3×22s4+2. Obviously, we only need to compare 2s4+2 with s2+2s4.
(1) 2s4+2≠s2+1. Then
v2(t2)=min(2s4+2,s2+1)≤2+2s4<4s4+1=min(s2+2s4,s2+3s8,2s4+3s8)+1, |
(2) 2s4+2=s2+1. We deduce that
v2(t2)=2s4+3<4s4+1=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
Subcase 5.3.4. 1<s2=2s4<3s8. It infers that
t2=22s4+3s8+1+24s4+3×23s8+1+3×22s4+2. |
Moreover, it's easy to see 2s4+3s8+1>max(4s4,3s8+1,2s4+2).
(1) s4≥3, so s2≥6. Then we have 2s4+2≤3s8+1 and 2s4+2<4s4, which implies that if 2s4+2=3s8+1, then
v2(t2)=v2(22s4+3s8+1+24s4+3×22s4+3)=2s4+3<1+4s4=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
and if 2s4+2<3s8+1 we have
v2(t2)=2+2s4<1+4s4=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
(2) s4=2, so s2=4. Then we have 2s4+2<min(4s4,3s8+1), which infers that
v2(t2)=2s4+2<1+4s4=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
(3) s4=1 and s8=1, so s2=2. Then t2=27+3×24, and we derive that
v2(t2)=4<7=min(s2+2s4+1,s2+1+3s8,2s4+1+3s8). |
(4) s4=1, s8≥2, so s2=2.
In this subcase, the inequality (2.3) does not hold. In fact, we have t2=23s8+3+26+3×23s8+1 and min(s2+2s4,s2+3s8,2s4+3s8)+1=1+min(4,2+2s8). Then it follows that v2(t2)=min(6,3s8)>1+min(4,2+2s8) (s8≥2).
But fortunately, we can calculate the approximate value. Recall that s3=s4=s6=1 and s8≥2.
If s5≥3, then by the expansion in Remark 1.1 and Remark 2.2 we have
1<H(2)9(s3=s4=s6=1,s2=2,s5,s7,s8,s9→∞)<H(2)9(s3=s4=s6=1,s2=2,s5≥3,s8≥2)≤H(2)9(s3=s4=s6=s7=s9=1,s2=s8=2,s5=3)<2. |
Hence H(2)9(s3=s4=s6=1,s2=2,s5≥3,s8≥2) is not an integer.
If s5=2, we have to classify the case by s7,s9. Indeed, one has
1<H(2)9(s3=s4=s6=1,s2=s5=2,s8≥2,s7=1,s9≥2)<2 |
and
1<H(2)9(s3=s4=s6=1,s2=s5=2,s8≥2,s7≥2,s9=1)<2. |
Moreover H(2)9(s3=s4=s6=1,s2=s5=2,s8≥2,s7=s9=1) is a decreasing function with the single variable s8. Thus we derive that
H(2)9(s3=s4=s6=s7=s9=1,s2=s5=s8=2)≈2.02, |
and by Remark 2.2,
1<H(2)9(s3=s4=s6=s7=s9=1,s2=s5=2,s8≥3)<1.99, |
which yields that H(2)9(s3=s4=s6=1,s2=s5=2,s8≥2) is not an integer.
if s5=1, as the discussion above, we get that
2<H(2)9(s3=s4=s6=s5=1,s2=2,s8≥2,s7=s9=1)<3, |
2<H(2)9(s3=s4=s6=s5=1,s2=2,s8≥2,s7=1,s9>1)<3, |
2<H(2)9(s3=s4=s6=s5=1,s2=2,s8≥2,s9=1,s7>1)<3, |
1<H(2)9(s3=s4=s6=s5=1,s2=2,s8≥2,s7>1,s9>1)<2. |
so H(2)9(s3=s4=s6=s5=1,s2=2,s8≥2) is not an integer.
Subcase 5.3.5. 1<s2=3s8<2s4. Then
t2=22s4+3s8+1+26s8+3×22s4+1+3×23s8+2. |
(1) s4=2. We can get an approximate value in the same way.
i.e. 1<H(2)9(s3=s6=s8=1,s2=3,s4=2)<2. It implies that H(2)9(s3=s6=s8=1,s2=3,s4=2) is not an integer.
(2) s4≥3 and 2s4>3s8+1. Then
v2(t2)=3s8+2<1+6s8=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
(3) s4≥3 and 2s4=3s8+1. Then we have t=26s8+2+26s8+3×23s8+3, which implies that
v2(t2)=3s8+3<1+6s8=min(s2+2s4,s2+3s8,2s4+3s8)+1. |
Thus we finish the proof of Lemma 2.2.
In this section, we take the same results of Lemma 2.1 and Remark 2.1 in proof. And we have the following lemma. This is also needed in the proof of Theorem 1.2.
Lemma 3.1. We have that v7(H(2)21)<0. So H(2)21 is not an integer.
Proof. By Lemma 2.1 and Remark 2.1, we consider the cases satisfying (2.1) for p=7. That is,
v7(t7)<m+s7+s14+s21=min(s7+s14,s7+s14,s14+s21) | (3.1) |
where m=min(−s7,−s14,−s21), t3=7s7+14s14+21s21.
Case 1. s7=s14=s21. Then we have t7=7s7(1+2s7+3s7)<7s77s7=72s7 and it implies v7(t7)<2s7=min(s7+s14,s7+s14,s14+s21).
Case 2. There's only one of s7,s14,s21 equals to min(s7,s14,s21). Just like the process we took before, we have
v7(t7)=min(7s7,14s14,21s21)<min(s7+s14,s7+s14,s14+s21). |
Case 3. s7=s14<s21. We can simplify t7=7s7(1+2s7)+7s213s21. Since 7 does not divide 1+2s7, then
v7(t7)=min(s7,s21)=s7<2s7=min(s7+s14,s7+s14,s14+s21). |
Case 4. s14=s21<s7. Consider the following cases:
(1) s14<2s14≤s7. Thus 2s14+3s14<7s14. So we have 7s14(2s14+3s14)<72s14<7s7. It means
v7(t7)=v7(7s14(2s14+3s14)+7s7)=v7(7s14(2s14+3s14))<2s14=min(s7+s14,s7+s14,s14+s21). |
(2) s14<s7<2s14. Then t7=7s14(7s7−s14+2s14+3s14)<7s14(7s14−1+5s14)<72s14. Hence v7(t7)<2s14=min(s7+s14,s7+s21,s14+s21).
Case 5. s7=s21<s14. Then t7=7s7(1+3s7)+7s142s14. Since 7 divides 1+3s7 sometimes, let v:=v7(1+3s7).
Case 5.1. v≠s14−s7. Then v7(t7)=min(s7+v,s14)≤s7+v. Since 1+3s7<7s7, so v<s7, which implies that
v7(t7)≤s7+v<2s7=min(s7+s14,s7+s21,s14+s21). |
Case 5.2. v=s14−s7. Since 3s7≡−1(mod7v), then 32s7≡1(mod7v).
On the other hand, it is easy to see 3 is a primitive root module 7, which yields that the order of 3 modulo 7v is (7−1)7v−1=φ(7v) (see [6], Theorem 3.6), that is, 3 is a primitive root module 7v. Then 2s7≥φ(7v)=6×7v−1. Hence s7≥3×7v−1≥3v, so we have s7+v2≤s7−v. Then it follows that
a=7s7+14s14+21s21=7s7+14s7+v+21s7=7s7+v(1+3s77v+4s7+v2)<7s7+v(3s7−v+4s7−v)<7s7+v7s7−v=72s7. |
Thus v7(t7)<2s7=min(s7+s14,s7+s14,s14+s21).
Therefore, Lemma 3.1 is proved.
In this section, we present the proofs of Theorems 1.1 and 1.2. We begin with the proof of Theorem 1.1.
Proof of Theorem 1.1. If n≥2, then by Bertrand's postulate, there exists one prime p such that n2<p≤n. Then p≤n<2p. Thus one has
H(1)n=∑1≤i≤n,i=p1isi+∑1≤i≤n,i≠p1isi=1psp+∑1≤i≤n,i≠p1isi. |
Since vp(1isi)=0>vp(1psp)=−sp for any 1≤i≤n with i≠p, then
vp(H(1)n)=vp(1psp)=−sp<0. |
i.e. for n≥2, H(1)n is never an integer.
In order to prove Theorem 1.2, we need the following lemmas.
Lemma 4.1. Let n≥2. If there is a prime p and p∈(n3,n2], then H(2)n is not an integer.
Proof. Since p∈(n3,n2], then 2p≤n<3p. Hence
H(2)n=∑1≤i<j≤n,vp(i)vp(j)>01isijsj+∑1≤i<j≤n,vp(i)vp(j)=01isijsj=12s2ppsp+s2p+∑1≤i<j≤n,vp(i)vp(j)=01isijsj, |
Since for any 1≤i≠j≤n with vp(i)vp(j)=0, we have
vp(1isijsj)≥min(−sp,−s2p)>−(sp+s2p)≥vp(12s2ppsp+s2p). |
The last inequality holds for the case of p=2. Then
vp(H(2)n)=vp(12s2ppsp+s2p)≤−(sp+s2p)<0. |
So if n≥2 and there is a prime p such that p∈(n3,n2], then H(2)n is not an integer.
Let
H=⋃p∈P[2p,3p) |
where P is the set consisting of all primes. Then for any integer n, it is easy to see that n∈H if and only if there exists one prime p, such that p∈(n3,n2]. Moreover, we have the following lemma.
Lemma 4.2. ([3]) The set Z+∖H is finite and max(Z+∖H)=21.
We can now give the proof of Theorem 1.2 as the conclusion of this paper.
Proof of Theorem 1.2. We divide the proof into two parts. First of all, we prove that H(2)n is never an integer except that n=3,s2=s3=1.
Case 1. n∈H. Then by Lemma 4.1, we have H(2)n is not an integer.
Case 2. n∉H. By Lemma 4.2, we have Z+∖H is finite and max(Z+∖H)=21. So if n>21, then n∈H. Furthermore, we have that
H=[4,6)⋃[6,9)⋃[10,15)⋃[14,21)⋃p>7,p∈P[2p,3p) |
Thus it follows that if n∉H, n equals to one of 2,3,9,21. Next, we will prove that none of H(2)2,H(2)3,H(2)9 and H(2)21 is an integer.
Case 2.1. n=2. Then it is obvious that
H(2)2=11s12s2∉Z, |
so H(2)2 is not an integer.
Case 2.2. n=3. Then we have
H(2)3=11s12s2+11s13s3+12s23s3=1s1+2s2+3s31s12s23s3. |
Since (2s2−1)(3s3−1)≥2, we have 2s23s3≥2s2+3s3+1. Thus we get that H(2)3≤1, and H(2)3=1 if and only if s2=s3=1.
Case 2.3. n=9 and n=21. Then Lemma 2.2 and Lemma 3.1 give us the desired result.
Secondly, we give a brief proof to the fact that when n≥2, H∗(2)n is never an integer. This will finish the proof of Theorem 1.2.
If n≥2, then by Bertrand's postulate, there is at least one prime p such that n2<p≤n. Then p≤n<2p, and vp(i)=0 or 1 for any integer i with 1≤i≤n. Thus one has
H∗(2)n=∑1≤i≤j≤n,vp(i)=vp(j)=11isijsj+∑1≤i≤j≤n,vp(i)vp(j)=01isijsj=1p2sp+∑1≤i≤j≤n,vp(i)vp(j)=01isijsj. |
Since 0≥vp(1isijsj)>vp(1p2sp)=−2sp for any 1≤i,j≤n with vp(i)vp(j)=0, then
vp(H∗(2)n)=vp(1p2sp)=−2sp<0. |
So when n≥2, H∗(2)n is never an integer.
Therefore Theorem 1.2 is proved.
The authors express their sincere gratitude to S.F.Hong and Qiuyu Yin for their excellent support. They thank the referees for their valuable comments and suggestions which help improve this article.
All authors declare no conflicts of interest in this paper.
[1] | Y.G. Chen and M. Tang, On the elementary symmetric functions of 1, 1/2, ..., 1/n, Am. Math. Mon., 119 (2012), 862-867. |
[2] | P. Erdös and I. Niven, Some properties of partial sums of the harmonic series, B. Am. Math. Soc., 52 (1946), 248-251. |
[3] | S.F. Hong and C.L. Wang, The elementary symmetric functions of reciprocals of the elements of arithmetic progressions, Acta Math. Hung., 144 (2014), 196-211. |
[4] | N. Koblitz, p-Adic numbers, p-adic analysis and zeta-functions, GTM 58, Springer-Verlag, New York, 1984. |
[5] | Y.Y. Luo, S.F. Hong, G.Y. Qian and C.L.Wang, The elementary symmetric functions of a reciprocal polynomial sequence, C.R. Math., 352 (2014), 269-272. |
[6] | M. B. Nathanson, Elementary methods in number theroy, GTM 195, Springer-Verlag, New York, 2000. |
[7] | C.L. Wang and S.F. Hong, On the integrality of the elementary symmetric functions of 1, 1/3, ..., 1=(2n -1), Math. Slovaca, 65 (2015), 957-962. |
1. | Qiuyu Yin, Shaofang Hong, Liping Yang, Min Qiu, Multiple reciprocal sums and multiple reciprocal star sums of polynomials are almost never integers, 2019, 195, 0022314X, 269, 10.1016/j.jnt.2018.06.005 | |
2. | Y. L. Feng, S. F. Hong, X. Jiang, Q. Y. Yin, A generalization of a theorem of Nagell, 2019, 157, 0236-5294, 522, 10.1007/s10474-018-00903-4 | |
3. | S. F. Hong, M. Qiu, On the p-adic properties of Stirling numbers of the first kind, 2020, 161, 0236-5294, 366, 10.1007/s10474-020-01037-2 |