Loading [MathJax]/jax/output/SVG/jax.js

The Mahler measure of (x+1/x)(y+1/y)(z+1/z)+k

  • Received: 01 October 2019 Revised: 01 December 2019
  • Primary: 11F67, 11R06; Secondary: 33C20

  • In this paper we study the Mahler measures of reciprocal polynomials (x+1/x)(y+1/y)(z+1/z)+k for k=16, k=104±603, 4096 and k=2024±7657. We prove six conjectural identities proposed by Samart in [16].

    Citation: Huimin Zheng, Xuejun Guo, Hourong Qin. The Mahler measure of (x+1/x)(y+1/y)(z+1/z)+k[J]. Electronic Research Archive, 2020, 28(1): 103-125. doi: 10.3934/era.2020007

    Related Papers:

    [1] Huimin Zheng, Xuejun Guo, Hourong Qin . The Mahler measure of $ (x+1/x)(y+1/y)(z+1/z)+\sqrt{k} $. Electronic Research Archive, 2020, 28(1): 103-125. doi: 10.3934/era.2020007
    [2] Dmitry Krachun, Zhi-Wei Sun . On sums of four pentagonal numbers with coefficients. Electronic Research Archive, 2020, 28(1): 559-566. doi: 10.3934/era.2020029
    [3] Hai-Liang Wu, Zhi-Wei Sun . Some universal quadratic sums over the integers. Electronic Research Archive, 2019, 27(0): 69-87. doi: 10.3934/era.2019010
    [4] Rong Zhang, Liangchen Wang . Global dynamics in a competitive two-species and two-stimuli chemotaxis system with chemical signalling loop. Electronic Research Archive, 2021, 29(6): 4297-4314. doi: 10.3934/era.2021086
    [5] Heesung Shin, Jiang Zeng . More bijections for Entringer and Arnold families. Electronic Research Archive, 2021, 29(2): 2167-2185. doi: 10.3934/era.2020111
    [6] Fedor Petrov, Zhi-Wei Sun . Proof of some conjectures involving quadratic residues. Electronic Research Archive, 2020, 28(2): 589-597. doi: 10.3934/era.2020031
    [7] Haiyu Liu, Rongmin Zhu, Yuxian Geng . Gorenstein global dimensions relative to balanced pairs. Electronic Research Archive, 2020, 28(4): 1563-1571. doi: 10.3934/era.2020082
    [8] Maoji Ri, Shuibo Huang, Canyun Huang . Non-existence of solutions to some degenerate coercivity elliptic equations involving measures data. Electronic Research Archive, 2020, 28(1): 165-182. doi: 10.3934/era.2020011
    [9] Lirong Huang, Jianqing Chen . Existence and asymptotic behavior of bound states for a class of nonautonomous Schrödinger-Poisson system. Electronic Research Archive, 2020, 28(1): 383-404. doi: 10.3934/era.2020022
    [10] Zhi-Wei Sun . New series for powers of $ \pi $ and related congruences. Electronic Research Archive, 2020, 28(3): 1273-1342. doi: 10.3934/era.2020070
  • In this paper we study the Mahler measures of reciprocal polynomials (x+1/x)(y+1/y)(z+1/z)+k for k=16, k=104±603, 4096 and k=2024±7657. We prove six conjectural identities proposed by Samart in [16].



    Let f be a nonzero Laurent polynomial in C[T1,, Tn,T11,,T1n]. The logarithmic Mahler measure of f is defined as

    m(f)=1010log|f(e2πiθ1, ..., e2πiθn)|dθ1dθn.

    There have been a lot of numerical evidence and theoretical results implying that there exist non-trivial relations between Mahler measure of certain kind of Laurent polynomials and special values of L-functions. The first example is provided by Smyth[17]:

    m(x+y+1)=334πL(χ3,2)=L(χ3,1),

    where χD denotes the Jacobi symbol (D).

    Suppose P is a two-variable Laurent polynomial. If P(x,y)=0 happens to define an elliptic curve E over Q, then m(P) can sometimes be related to L(E,2) by Beilinson's regulator maps. It is Deninger who first noticed this and conjectured in [5] that

    m(1+X+1/X+Y+1/Y)=L(E15,0),

    where E15 is an elliptic curve with conductor 15. This conjecture is proved by Rogers and Zudilin in [11] and [12]. After Deninger's discovery, Boyd proposed hundreds of conjectures in this direction [4].

    In 2008, Bertin [1,2] investigated the Mahler measure of three-variable Laurent polynomials:

    m(x+x1+y+y1+z+z1+k).

    Later, Rogers [10] considered the following functions:

    f2(k)=2m((x+1x)(y+1y)(z+1z)+k),f4(k)=4m(x4+y4+z4+1+k1/4xyz),

    where kC is a constant. Samart's work shows that there are some appealing connections between the special values of the above functions and the special value of L-functions of certain modular forms [15]. Actually Samart made many more conjectures of this type in [15] and [16]. The following five identities conjectured in [15] have been proved by Guo, Peng and Qin in [7]:

    f2(64)=2L(g32,0)+2L(χ4,1),f2(512)=L(g64,0)+L(χ8,1),f4(1024)=85(5L(g20,0)+2L(χ4,1)),f4(12288)=409(L(g36,0)+2L(χ3,1)),f4(82944)=4013(L(g52,0)+2L(χ4,1)),

    where gN denotes some newform with rational coefficients in the space of cusp forms of weight 3 and level N.

    In this article, we prove six identities conjectured in 2015 in [16]:

    f2(16)=8M12,f2(104603)=16(4M12(4)+36M12+15d3+8d4),f2(104+603)=12(4M12(4)36M12+15d38d4),f2(4096)=47(M7(4)+8d4),f2(20247657)=114(4M7(4)+384M7+32d4+11d7),f2(2024+7657)=12(4M7(4)384M732d4+11d7).

    Here, dk=L(χk,1), MN=L(gN,0), MND=L(gNχD,0) and gNχD is the quadratic twist of gN by χD.

    In addition to the complicated computation of the lattice sums, which is mainly due to the distinct forms of the integral basis of the number fields Q(3) and Q(7), some techniques that work well in [7] fail in our cases. For example, the method of finding the identities of the corresponding modular forms fails, as the L-functions are no longer equal to each other in one of our cases. In order to prove the identities of the special values of L-functions, we have to develop another method, see Theorem 4.2 in this paper. On the other hand, it seems that the identities like f2(1)=8M7 in [16] cannot be proved by any of our previous techniques, although they appear very similar to the ones we have verified. In fact we are still not able to prove them.

    As usual we denote the Dedekind eta function by η(τ). Then we have Δ(τ)=η(τ)24, where Δ is the modular discriminant. Define

    s2(q(τ))=Δ(τ+12)Δ(2τ+1).

    Write f(τ) for the classical Weber modular function eπi24η((τ+1)/2)η(τ). The following relation holds (cf. [15], Lemma 2.2)

    s2(q(τ))=f(2τ)24.

    Let f1(τ)=η(τ/2)η(τ), f2(τ)=2η(2τ)η(τ). We have f(τ)f1(τ)f2(τ)=2.

    Moreover, f(τ)24, f1(τ)24, f2(τ)24 are exactly the three roots of the cubic equation

    (x16)3j(τ)x=0,

    where j(τ) is the classical j-invariant. One can see more details on page 288 of [3].

    Lemma 2.1. We have the following special values of f(τ)24:

    (1)f((1+3)/2)24=s2(q((1+3)/4))=16.

    (2) f(1+3)24=s2(q((1+3)/2))=104603.

    (3) f((3+3)/3)24=s2(q((3+3)/6))=104+603.

    (4) f(7)24=s2(q(7/2))=4096.

    (5) f(1+7)24=s2(q((1+7)/2))=20247657.

    (6) f((7+7)/7)24=s2(q((7+7)/14))=2024+7657.

    Proof. (1)We have

    (f(τ+1)f1(τ+1)f2(τ+1))=(0ζ10ζ10000ζ2)(f(τ)f1(τ)f2(τ)) (2.1)

    where ζ=eπi24(cf. page 1647 of [19]). Setting x=16 in equation (x16)3j(τ)x=0, we get j(τ)=0, which implies τ can be taken as 1+32.

    It follows from (x16)3=0 that

    f(τ)24=16,f1(τ)24=16,f2(τ)24=16,

    and we are done.

    Setting x=104±603 in equation (x16)3j(τ)x=0, one gets j(τ)=54000. The three roots of the equation (x16)354000x=0 are

    {x1=256,x2=104603,x3=104+603}.

    By Weber's table VI [18], f(3)=32, i.e., f(3)24=28=256, hence

    {f1(3)24,f2(3)24}={104+603,104603}.

    By numerical computations, one can deduce that f1(3)24=104+603 and f2(3)24=104603.

    Using (2.1) we obtain

    f(1+3)24=f1(3)24=104603.

    (3) Note that the transformation τ1τ fixes f, and exchanges f1, f2. Hence,

    f1(1/3)24=f2(3)24=104603, which implies

    f(1+13)24=f1(13)24=104+603.

    (4) By Weber's table VI [18], we have f(7)=2, which implies

    f(7)24=4096.

    (5) & (6): Setting x=2024±7657 in equation (x16)3j(τ)x=0, we get j(τ)=16581375. Solving the equation (x16)316581375x=0 one obtains three roots:

    {x1=4096,x2=2024+7657,x3=20247657}.

    From above, we have known that f(7)24=4096. Thus

    {f1(7)24,f2(7)24}={2024+7657,20247657}.

    By numerical computation, one can decide that

    f1(7)24=20247657

    and

    f2(7)24=2024+7657.

    Therefore

    f(1+7)24=f1(7)24=20247657.

    Moreover, we have

    f1(7/7)24=2024+7657,

    which implies

    f(1+77)24=f1(7/7)24=2024+7657.

    Lemma 2.2.([15], Proposition 2.1(i)). Assume that q(0,1). If Im(τ)12, then

    f2(s2(q(τ)))=2Im(τ)π3(A+16B)

    where

    A=m,nZ4(mRe(τ)+n)2|mτ+n|6m,nZ1|mτ+n|4,B=m,nZ4(4mRe(τ)+n)2|4mτ+n|6m,nZ1|4mτ+n|4.

    The restriction on q can be removed and the lower bound of Im(τ) is unnecessary. Applying a similar argument as in [8], one can show that the identity in this lemma holds everywhere. Indeed, both f2(k) and 2Im(τ)π3(A+16B) are the real parts of holomorphic functions at least when kC[64,64] and they coincide on a non-discrete set of points, see [10] and [15].

    The L-function with respect to a quadratic Dirichlet character χ is denoted by L(s,χ). Let χ4 be the character (1), χ8()=(2), χ8()=(2) and χ3()=(3), etc.

    We will use the notations as in [6]. For any integer n, write Ln(s)=L(χn,s), in particular, L1(s)=ζ(s) the Riemann zeta function. The following results prove to be crucial to our computation.

    Lemma 2.3. (Glasser & Zucker [6], Table Ⅵ). For any complex s with Re(s)>1,

    (1) x,yZ1(x2+3y2)s=2(1+212s)L1(s)L3(s)

    (2)x,yZ1(x2+7y2)s=2(121s+212s)L1(s)L7(s)

    (3) x,yZ1(x2+12y2)s=(1+22s+224s)L1(s)L3(s)+L4(s)L12(s)

    (4) x,yZ1(x2+28y2)2=(121s+322s223s+224s)L1(s)L7(s)+L4(s)L28(s)

    Remark 1. The following results are useful for our computation: L1(2)=π26, L12(2)=318π2, L28(2)=2π277.

    Remark 2. For the sake of brevity, let us do the following convention of notations in this paper. Let R(l,m) be any function of l and m. We denote

    (a,b)R(l,m)=(l,m)(a,b)(mod2)R(l,m),

    where a,bZ. As usual, the symbol (a,b)R(l,m) means the sum does not include (a,b)=(0,0). Similarly, the notation Σxy(d) means Σxy(modd), where the primed summation sign means to sum over all integer pairs except (x,y)=(0,0).

    Following Samart's method, we will calculate A and B in the statement of Lemma 2.2 for our case, where τ=1+34 by Lemma 2.1(1).

    A=m,nZ(16((m+n)+3n)2((m+n)2+3n2)316((m+n)2+3n2)2)=l,nZ(16(l+3n)2(l2+3n2)316(l2+3n2)2)(let l=m+n)=l,nZ(16(l2+9n2)(l2+3n2)316(l2+3n2)2)(l,nZln/(l2+3n2)3=0)=l,nZ(16(l2+9n2)(l2+3n2)332(l2+3n2)(l2+3n2)3+16(l2+3n2)2)=l,nZ16(l23n2)(l2+3n2)3+l,nZ16(l2+3n2)2=AM+AD,

    and

    B=m,nZ(4(m+n)2(n2+2mn+4m2)31(n2+2mn+4m2)2)=l,mZ(4l2(l2+3m2)31(l2+3m2)2)=l,mZ(2(l23m2)(l2+3m2)3+1(l2+3m2)2)=BM+BD.

    Thus,

    AD+16BD=l,nZ16(l2+3n2)2+l,mZ16(l2+3m2)2=0.

    Let

    f(τ)=x,yZ(x23y22)qx2+3y2=η(2τ)3η(6τ)3.

    By lemma 2.7 in [15], f(τ)S3(Γ0(12),χ3). Thus,

    AM+16BM=l,mZ(16(l23m2)(l2+3m2)3+32(l23m2)(l2+3m2)3)=l,mZ48(l23m2)(l2+3m2)3=96L(f,3)=16π333L(f,0).

    Therefore, we obtain:

    2Im(τ)π3(A+16B)=2Im(τ)π3(AM+16BM)=8L(f,0),

    as desired.

    In this subsection, K=Q(3). By Lemma 2.1, we know that

    s2(q(1+32))=f(1+3)24=104603.

    So in this case

    τ=1+32,Im(τ)=32,Re(τ)=12.

    To ease the notation, we set

    AM=ml(2)32(l23m2)(l2+3m2)3,BM=l,mZ2(l212m2)(l2+12m2)3,AD=ml(2)16(l2+3m2)2,BD=l,mZ1(l2+12m2)2,C=l,mZ1(l2+3m2)2,X=(1,1)1(l2+3m2)2.

    By Samart's formulas in Lemma 2.2, we have

    A=m,nZ(64(2n+m)2((2n+m)2+3m2)316((2n+m)2+3m2)2)=ml(2)(64l2(l2+3m2)316(l2+3m2)2)(l=2n+m)=ml(2)32(l23m2)(l2+3m2)3+ml(2)16(l2+3m2)2=AM+AD,

    and

    B=m,nZ(4(n+2m)2(16m2+4mn+n2)31(16m2+4mn+n2)2)=m,nZ(4(n+2m)2((n+2m)2+12m2)31((n+2m)2+12m2)2)=l,mZ(4l2(l2+12m2)31(l2+12m2)2)(l=n+2m)=l,mZ2(l212m2)(l2+12m2)3+l,mZ1(l2+12m2)2=BM+BD.

    We continue to set

    f(τ)=x,yZ(x23y22)qx2+3y2=η(2τ)3η(6τ)3S3(Γ0(12),χ3).

    From the proof of Lemma 2.4 in [15], we know that h=xy(mod2)(x23y22)qx2+3y2=0. As a multiple of the L-function associate to h, AM=0. Thus

    f(τ)=xy(2)(x23y22)qx2+3y2.

    On the other hand, we let

    fχ4(τ)=xy(2)(x23y22)χ4(x2+3y2)qx2+3y2.

    It follows from [15] (cf. page 248, last line) that fχ4(τ)S3(Γ0(48),χ3). We see that

    fχ4(τ)=(x,y)(1,0)(mod2)(x23y22)qx2+3y2(x,y)(0,1)(mod2)(x23y22)qx2+3y2,

    and

    (x,y)(1,0)(mod2)(x23y22)qx2+3y2=12(f(τ)+fχ4(τ)).

    Therefore,

    l,mZ2(l212m2)(l2+12m2)3=l,mZm even2(l23m2)(l2+3m2)3=(0,0)2(l23m2)(l2+3m2)3+(1,0)2(l23m2)(l2+3m2)3=94L(f,3)+2L(fχ4,3).

    Applying the functional equation (the sign can be settled by finite precision numerical computation, it is '+' for our case)

    (N2π)sΓ(s)L(f,s)=±(N2π)3sΓ(3s)L(f,3s), (3.1)

    we obtain

    2Im(τ)π3(AM+16BM)=163π3(94π363L(f,0)+2π3483L(fχ4,0))=6L(f,0)+23L(fχ4,0).

    Now let us deal with the Dirichlet L-function part. It is clear that

    ml(2)1(l2+3m2)2=(0,0)1(l2+3m2)2+(1,1)1(l2+3m2)2.

    One can easily check that

    {IOK}={αOK|α=x+y32,x1(mod2)y1(mod2)}{αOK|α=x+y3,xZyZ},

    where means disjoint union.

    Thus we have

    6IOK1N(I)2=(1,1)16(x2+3y2)2+x,yZ1(x2+y2)2,

    which implies

    X=116(6ζK(2)C).

    Here we used the fact that |U(K)|=6.

    By Lemma 2.3, we know that the following results hold

    C=2(1+2122)ζK(2)=94ζK(2). (3.2)

    Hence,

    X=116154ζK(2)=1564L1(2)L3(2). (3.3)

    Therefore,

    2Im(τ)π3(AD+16BD)=232π3(ml(2)16(l2+3m2)2+16l,mZ1(l2+12m2)2)=163π3(((0,0)1(l2+3m2)2+(1,1)1(l2+3m2)2)+l,mZ1(l2+12m2)2)=163π3(964π26L3(2)1564π26L3(2)+6964π26L3(2)+318π2L4(2))=1636π(4564L3(2)+33L4(2))=833π(45644π33d3+33π2d4)=52d3+43d4.

    Here, we used the functional equation

    (πk)2s2Γ(2s2)L(χk,1s)=(πk)s+12Γ(s+12)L(χk,s). (3.4)

    In conclusion,

    s2(q(τ))=2Im(τ)π3(A+16B)=2Im(τ)π3(AM+16BMAD+16BD)=6L(f,0)+23L(fχ4,0)+52d3+43d4=16(4M12(4)+36M12+15d3+8d4).

    In this subsection, we still denote Q(3) by K. By Lemma 2.1, we have

    s2(q(3+36))=f(3+36)24=104+603.

    So in this case,

    τ=3+36, Im(τ)=36, Re(τ)=12.

    To ease the notation, we set

    AM=ml(2)288(3l2m2)(3l2+m2)3,BM=l,mZ18(3l24m2)(3l2+4m2)3,AD=ml(2)122(3l2+m2)2,BD=l,mZ32(3l2+4m2)2,C=l,mZ1(3l2+m2)2,X=(1,1)1(3l2+m2)2,D=l,mZl even1(3l2+m2)2,Y=(1,0)1(3l2+m2)2.

    We need to calculate the values of A and B in Samart's formula, where

    A=m,nZ(123(2n+m)2(3(2n+m)2+m2)3122(3(2n+m)2+m2)2)=ml(2)(123l2(3l2+m2)3122(3l2+m2)2)(l=2n+m)=ml(2)288(3l2m2)(3l2+m2)3+ml(2)122(3l2+m2)2=AM+AD,

    and

    B=m,nZ(4(4m12+n)2(n2+4mn+16m23)31(n2+4mn+16m23)2)=m,nZ(4(n+2m)2((n+2m)2+4m23)31((n+2m)2+4m23)2)=l,mZ(334l2(3l2+4m2)332(3l2+4m2)2)(l=n+2m)=l,mZ18(3l24m2)(3l2+4m2)3+l,mZ32(3l2+4m2)2=BM+BD.

    From the computation in the last section, we know that AM=0. Define f(τ) just in the same way as in the last section. We obtain

    l,mZ(4m23l2)(3l2+4m2)3=l,mZm even(m23l2)(3l2+m2)3=(0,0)(m23l2)(m2+3l2)3+(1,0)(m23l2)(m2+3l2)3=2l,mZ4(m23l2)432(m2+3l2)3+2(1,0)(m23l2)2(m2+3l2)3=18L(f,3)+212(L(f,3)L(fχ4,3))=98L(f,3)L(fχ4,3)=98π363L(f,0)π3483L(fχ4,0).

    Thus,

    2Im(τ)π3(AM+16BM)=1633π3l,mZ18(3l24m2)(3l2+4m2)3=963π3(98π363L(f,0)π3483L(fχ4,0))=18L(f,0)+2L(fχ4,0).

    We now treat the part involving the Dirichlet L-Series. By Lemma 2.3, we have

    D=l,mZ1(l2+12m2)2=(1+24+26)L1(2)L3(2)+L4(2)L12(2)
    =6964π26L3(2)+318π2L4(2), (3.5)

    where we used Remark 2.4 that L12(2)=318π2. It is clear that Y=CXD. By identities (2)-(5), we do the following calculation

    2Im(τ)π3(AD+16BD)=12233π3(ml(2)1(m2+3l2)2+l,mZ1(4m2+3l2)2)=12233π3((1,1)1(m2+3l2)2+(1,0)1(m2+3l2)2)=12233π3(X+Y)=12233π3(C2XD)=12233π3(4564π26L3(2)318π2L4(2))=12233π3(4564π264π33d3318π2π2d4)=152d34d4.

    Thus

    s2(q(τ))=2Im(τ)π3(A+16B)=2Im(τ)π3(AM+16BMAD+16BD)=18L(f,0)+2L(fχ4,0)+152d34d4=12(4L(fχ4,0)36L(f,0)+15d38d4).

    In this section, let K=Q(7). Then Cl(OK)=1OK is PID, U(K)={±1} and 2 split completely in K. Let

    w=1+72

    and

    =(w).

    Hence,

    2OK=ˉ.

    On the other hand, let us do the following settings

    C=IOK1N(I)2=ζK(2),F=l,mZ1(l2+7m2)2,

    where I denotes the ideals in OK. In addition we set

    =(l,mZ+2(0,0)ml(2))1(l2+7m2)2.

    By Lemma 2.1, we know that

    s2(q(72))=f(7)24=4096.

    So in this case,

    τ=72,Im(τ)=72,Re(τ)=0.

    To ease the notation, we set

    AM=l,mZl even32(l27m2)(l2+7m2)3,BM=m,nZ2(n228m2)(n2+28m2)3,AD=l,mZl even16(l2+7m2)2,BD=m,nZ1(n2+28m2)2,

    We have

    A=m,nZ(256n2(7m2+4n2)316(7m2+4n2)2)=l,mZl even32(l27m2)(l2+7m2)3+l,mZl even16(l2+7m2)2=AM+AD,

    and

    B=m,nZ(4n2(28m2+n2)31(28m2+n2)2)=m,nZ2(n228m2)(n2+28m2)3+m,nZ1(n2+28m2)2=BM+BD.

    Let g7=η(τ)3η(7τ)3=m,nNχ4(mn)mnqm2+7n28. We have following result.

    Theorem 4.1. Let

    g=xy1(2)(x27y22)qx2+7y2.

    Then,

    gχ4=g7χ4

    Proof. Let Λ=(2) and define ϕ(a)=α2 for a generator of an integral ideal a which satisfies α1(Λ).

    In fact, if (α) is coprime to Λ, then α1(Λ). To see this, consider the norm of α, it is odd (since (α)+Λ=OK) and moreover,

    N(α)=N(x+y1+72)=(x+y2+72y)(x+y272y)=(x+y2)274y2=(x+y2)2+74y2=x2+xy+2y2x2+xy(Λ).

    Hence,

    (x,y)(1,0)(mod2).

    Thus if (α) is coprime to Λ, then α1(Λ). It is clear that ϕ is multiplicative and satisfies ϕ(αOK)=α2. Therefore, ϕ is a Hecke charactor which satisfies the condition of Theorem 1.31 in [9]. So

    g(τ)=aϕ(a)qN(a)=n(aN(a)=nϕ(a)qn)

    is a newform in S3(Γ0(28),χ7).

    As y is even, we can rewrite the norm as

    N(α)=N(x+y(1+7))(y=y/2)=N(x+y+y7)=N(x+y7)(x=x+y,y=y).

    Here, x,y satisfy that (x,y)Z and xy1(mod2) since x=xy should be odd from above.

    Note that α=x+y7 and β=xy7 share the same norm value, so we have:

    g(τ)=12(α,Λ)=1(ϕ(α)+ϕ(β))qN(α)=1212xy1(2)(x2+y2(7)+2xy7+x2+y2(7)2xy7)qx2+7y2=xy1(2)(x27y22)qx2+7y2.

    So one can deduce that gχ4=g7χ4 whose level divides 112 by considering the Sturm Bound (with the help of SageMath[13]). Indeed, the Sturm Bound for our case is at most 48. In addition,

    g7=q3q2+5q47q73q8+9q96q11+21q1411q1627q18+18q22+18q23+25q2535q2854q29+45q32+45q3638q37+58q4330q4454q46+49q49+O(q50),
    g=q7q7+9q96q11+18q23+25q2554q2938q37+58q43+49q49+O(q50),

    and

    gχ4=g7χ4=q+7q7+9q9+6q1118q23+25q2554q2938q3758q43+49q49+O(q50).

    Since

    AM+16BM=l,mZl even32(l27m2)(l2+7m2)3+l,mZm even32(l27m2)(l2+7m2)3=32((0,0)(0,1)+(0,0)+(1,0))=32((1,0)(0,1))=64L(g7χ4,3),

    one can get

    2Im(7/2)π3(AM+16BM)=7π3(AM+16BM)=7π364L(g7χ4,3)=7π364π31127L(g7χ4,0)=47L(g7χ4,0).

    Here we applied the functional equation (1) to L(g7χ4,3).

    Now let us deal with the Dirichlet L-Series part, or more exactly, the computation of the term

    l,mZm even1(l2+7m2)2l,mZl even1(l2+7m2)2.

    Firstly, by

    ml(2)=m,lZl,mZm evenl,mZl even+2(0,0),

    we obtain

    =l,mZm even+l,mZl even.

    Thus,

    l,mZm even1(l2+7m2)2l,mZl even1(l2+7m2)2=2l,mZm even1(l2+7m2)2.

    By Lemma 2.3,

    l,mZm even1(l2+7m2)2=x,yZ1(x2+28y2)2=4164L1(2)L7(2)+L4(2)L28(2).

    On the other hand, we have

    =l,mZ+2(0,0)(0,0)(1,1)=l,mZ+(0,0)(1,1).

    In addition, one can easily check that

    {IOK}={αOK|α=x+y72,x1(mod2)y1(mod2)}{αOK|α=x+y7,xZyZ}

    which implies

    16(1,1)1(l2+7m2)2=2ζK(2)F.

    By Lemma 2.3,

    =4132ζK(2)=4132L1(2)L7(2),

    Therefore,

    16(l,mZm even1(l2+7m2)2l,mZl even1(l2+7m2)2)=16(2l,mZm even)=32(4164L1(2)L7(2)+L4(2)L28(2))164132L1(2)L7(2)=32L4(2)L28(2).

    Since L28(2)=2π2/(77) and by the functional equation (5) we have

    L4(2)=π2L4(1)=π2d4.

    Hence,

    2Im(τ)π3(AD+16BD)=7π332L4(2)L28(2)=7π332π2d42π277=327d4.

    Finally, we obtain

    2Im(τ)π3(A+16B)=2Im(τ)π3(AMAD+16BM+16BD)=47(L(g7χ4,0)+8d4),

    as desired.

    To ease the notation, in this subsection we set

    AM=l,mZl even32(l27m2)(l2+7m2)3,BM=m,nZ2(n228m2)(n2+28m2)3,AD=l,mZl even16(l2+7m2)2,BD=m,nZ1(n2+28m2)2.

    Following Samart's method, we will calculate A and B in the statement of Lemma 2.2 for our case, where τ=1+72 by Lemma 2.1(2).

    A=m,nZ(64(2n+m)2((2n+m)2+7m2)316((2n+m)2+7m2)2)=ml(2)(64l2(l2+7m2)316(l2+7m2)2)(Letl=2n+m)=ml(2)32(l27m2)(l2+7m2)3+ml(2)16(l2+7m2)2=AD+AM,andB=m,nZ(4(n+2m)2((n+2m)2+28m2)31((n+2m)2+28m2)2)=l,mZ(4l2(l2+28m2)31(l2+28m2)2)(Letl=n+2m)=l,mZ2(l228m2)(l2+28m2)3+l,mZ1(l2+28m2)2=BM+BD.

    To prove this identity, we need to find another expression for g7 defined in last subsection.

    Theorem 4.2. We have

    g7=xy(2)(x27y28)qx2+7y24S3(Γ0(7),χ7).

    Proof. Take modulus Λ=(1) and define ϕ(a)=α2 for a generator α of a which satisfies α1(Λ). It is clear that ϕ is multiplicative and satisfies ϕ(αOL)=α2. Therefore, ϕ is a Hecke character which satisfies the conditions of Theorem 1.31 in [9]. So

    G7(τ)=aϕ(a)qN(a)=n(aN(a)=nϕ(a)qn)

    is a newform in S3(Γ0(7),χ7). Here, the sum in the definition of G7(τ) is over all ideals in OK.

    Rewrite α=x+y1+72 as α=2x+y+y72=x+y72, x,yZ and x=2x+y, y=y. Then we have xy(mod 2).

    Note that α=x+y7 and β=xy7 share the same norm value, so we have

    G7(τ)=12I=(α)(ϕ(α)+ϕ(β))qN(α)=1212xy(2)((x+y72)2+(xy72)2)qx2+7y24=1212xy(2)(x2+y2(7)+2xy74+x2+y2(7)2xy74)qx2+7y24=xy(2)(x27y28)qx2+7y24.

    Finally one can show that g7=G7 with the help of SageMath [13] and Sturm bound.

    Theorem 4.3. With the notation above, the following identity holds:

    3L(g7,3)=2(1,1)(l27m2)(l2+7m2)3+(1,0)(l27m2)(l2+7m2)3+(0,1)(l27m2)(l2+7m2)3.

    Proof. Firstly, by the fact that,

    (0,0)=4|l,2||m+2||l,4|m+2||l,2||m+4|l,4|m,

    we have

    L(g7,3)=8((1,1)l27m2(l2+7m2)3+(0,0)l27m2(l2+7m2)3)=8((1,1)l27m2(l2+7m2)3+(124+128+)((1,1)+(1,0)+(0,1)))=12815(1,1)+815(1,0)+815(0,1). (4.1)

    Secondly, we have

    {IidealOK}=S1S2S3S4,

    where

    S1={(α)|αOK,((2),(α))=1},S2={(α)|αOK,(α),ˉ(α)},S3={(α)|αOK,ˉ(α),(α)},S4={(α)|αOK,2|(α)}.

    And it is clear that:

    S2={(α)|αOK,α=wjβ,j1,(β)S1},S3={(α)|αOK,α=ˉwjβ,j1,(β)S1},S4=S4,1S4,2S4,3.

    where

    S4,1={(α)|αOK,α=2jβ,j1,(β)S1},S4,2={(α)|αOK,α=2jwkβ,where j,k1,(β)S1},S4,3={(α)|αOK,α=2jˉwkβ,where j,k1,(β)S1}.

    Write g7 as following form

    g7=I=(α)α2qN(α),

    i.e.,

    g7=I=(α)IS1α2qN(α)+I=(α)IS2α2qN(α)+I=(α)IS3α2qN(α)+I=(α)IS4α2qN(α).

    We can find that

    I=(α)IS1α2qN(α)=g,I=(α)IS2α2qN(α)=I=(α)αS1j=1(wjα)2qN(wjα)=j=1w2jI=(α)αS1α2q2jN(α)
    =j=1w2jg(2jτ),I=(α)IS3α2qN(α)=I=(α)αS1j=1(ˉwjα)2qN(ˉwjα)=j=1ˉw2jI=(α)αS1α2q2jN(α)=j=1ˉw2jg(2jτ).

    In addition,

    I=(α)IS4α2qN(α)=I=(α)IS4,1α2qN(α)+I=(α)IS4,2α2qN(α)+I=(α)IS4,3α2qN(α),

    where

    I=(α)IS4,1α2qN(α)=I=(α)αS1j=1(2jα)2qN(2jα)=j=122jI=(α)αS1α2q22jN(α)=j=122jg(22jτ),I=(α)IS4,2α2qN(α)=I=(α)αS1k=1j=1(2kwjα)2qN(2kwjα)=k=122kj=1w2jI=(α)αS1α2q22k2jN(α),I=(α)IS4,3α2qN(α)=I=(α)αS1k=1j=1(2kˉwjα)2qN(2kˉwjα)=k=122kj=1ˉw2jI=(α)αS1α2q22k2jN(α).

    Let h(τ)=j=1(w2j+ˉw2j)g(2jτ). Then

    g7=g(τ)+h(τ)+j=122jg(22jτ)+k=122kh(22kτ).

    Note that

    L(h,3)=j=1(w2j+ˉw2j)23jL(g,3).

    Therefore,

    L(g7,3)=L(g,3)(1+j=1w2j+ˉw2j23j)(1+124+128+)=1615L(g,3)(1+j=1((w28)j+(ˉw28)j))=1615L(g,3)(1+w2/81w2/8+ˉw2/81ˉw2/8)=1623L(g,3)=823((1,0)+(0,1)).

    Combining the relation (6), we have

    (1,1)=146((1,0)+(0,1)).

    Thus

    23(1,1)+13((1,0)+(0,1))=823((1,0)+(1,1))=L(g7,3),

    as desired.

    By functional equations, one can show that

    L(g7,0)=774π3L(g7,3),
    L(g7χ4,0)=1127π3L(g7χ4,3).

    In addition

    114(4M7(4)+384M7)=4141127π3L(g7χ4,3)+38414774π3L(g7,3)=7π3(32L(g7χ4,3)+48L(g7,3))=167π3(2L(g7χ4,3)+3L(g7,3)). (4.2)

    On the other hand,

    7π3(AM+16BM)=167π3(ml(2)2(l27m2)(l2+7m2)3+l,mZm even2(l27m2)(l2+7m2)3)=327π3(ml(2)+l,mZm even)=327π3((1,1)+(1,0))=327π3((1,1)+12(1,0)+12(0,1)+12((1,0)(0,1)))=327π3((1,1)+12(1,0)+12(0,1)+L(gχ4,3)). (4.3)

    Comparing (7) and (8) and by Theorem 4.2, 4.3, we are done.

    AD+16BD=(0,0)16(l2+7m2)2(1,1)16(l2+7m2)2+l,mZm even16(l2+7m2)2                  =16((0,0)1(l2+7m2)2(1,1)1(l2+7m2)2+l,mZm even1(l2+7m2)2).

    From last subsection, one can get

    AD+16BD=16(564C364C+4164C+L4(2)L28(2))=334C+L4(2)L28(2)=334L1(2)L7(2)+16L4(2)L28(2)=1114L(χ7,1)+167L(χ4,1)=1114d7+167d4,

    as desired. Here we used the following facts:

    L1(2)=π26;L28(2)=2π277;(Remark 2.4)L4(2)=π2d4;L7(2)=4π77d7.(by functional equation)

    To ease the notation, we set:

    AM=ml(2)7225(7l2m2)(7l2+m2)3,BM=m,nZ272(7l24m2)(7l2+4m2)3,AD=ml(2)7216(7l2+m2)2,BD=m,nZ72(7l2+4m2)2.

    in this subsection.

    Following Samart's method, we will calculate A and B in the statement of Lemma 2.2 for our case, where τ=7+714 by Lemma 2.1(3).

    A=m,nZ(64(2n+m)2((2n+m)2+m27)316((2n+m)2+m27)2)=ml(2)(64l2(l2+m27)316(l2+m27)2)(Letl=2n+m)=ml(2)(7225(7l2m2)(7l2+m2)3+7225(7l2+m2)(7l2+m2)37216(7l2+m2)2)=ml(2)7225(7l2m2)(7l2+m2)3+ml(2)7216(7l2+m2)2=AM+AD,

    and

    B=m,nZ(734(n+2m)2(7(n+2m)2+4m2)372(7(n+2m)2+4m2)2)=m,nZ(734l2(7l2+4m2)372(7l2+4m2)2)(Letl=n+2m,)=m,nZ(272(7l24m2)(7l2+4m2)3+272(7l2+4m2)(7l2+4m2)372(7l2+4m2)2)=m,nZ272(7l24m2)(7l2+4m2)3+m,nZ72(7l2+4m2)2=BM+BD.

    By functional equations, we have

    L(g7,0)=774π3L(g7,3),
    L(g7χ4,0)=1127π3L(g7χ4,3).

    And,

    12(4M7(4)384M7)=421127π3L(g7χ4,3)3842774π3L(g7,3)=7π3(224L(g7χ4,3)336L(g7,3))=1127π3(2L(g7χ4,3)3L(g7,3)). (4.4)

    On the other hand,

    77π3(AM+16BM)=2247π3(ml(2)(l27m2)(l2+7m2)3l,mZl even(l27m2)(l2+7m2)3)=2247π3((1,1)(0,1))=2247π3((1,1)12(1,0)12(0,1)+12((1,0)(0,1)))=2247π3((1,1)12(1,0)12(0,1)+L(gχ4,3)). (4.5)

    By Theorem 4.2 and Theorem 4.3, comparing (9) and (10), we are done.

    AD+16BD=(0,0)7216(l2+7m2)2(1,1)7216(l2+7m2)2+l,mZl even7216(l2+7m2)2                  =7216((0,0)1(l2+7m2)2(1,1)1(l2+7m2)2+l,mZl even1(l2+7m2)2).

    From previous computation, we have

    l,mZl even1(l2+7m2)2=l,mZm even1(l2+7m2)2+.

    Hence,

    AD+16BD=7216(564C364C4164CL4(2)L28(2)+4132C)=72(334L1(2)L7(2)16L4(2)L28(2))=72(1114L(χ7,1)167L(χ4,1))=72(1114d7167d4),

    as desired. Here we used the following facts:

    L1(2)=π26;L28(2)=2π277;(Remark 2.4.)L4(2)=π2d4;L7(2)=4π77d7(by functional equation)

    The first author would like to thank Dr.Wei Lu for deducing the very useful form of g7 in theorem 4.2 in an elegant way, which inspires him to come up with the proof given in this paper. The authors would like to thank the referee for many valuable comments and suggestions which have been incorporated herein.



    [1] M. J. Bertin, Mahler's measure and L-series of K3 hypersurfaces, Mirror Symmetry, AMS/IP Stud. Adv. Math., Amer. Math. Soc. Providence, RI, 38 (2006), 3–18.
    [2] Bertin M. J. (2008) Mesure de Mahler d'hypersurfaces K3. J. Number Theory 128: 2890-2913.
    [3] Birch B. J. (1969) Weber's class invariants. Mathematika 16: 283-294.
    [4] Boyd D. W. (1998) Mahler's measure and special values of L-functions. Experimental Math. 7: 37-82.
    [5] Deninger C. (1997) Deligne periods of mixed motives, K-theory and the entropy of certain Zn-actions. J. Amer. Math. Soc. 10: 259-281.
    [6] Glasser M. L., Zucker I. J. (1980) Lattice sums. New York: Theoretical Chemistry - Advances and Perspectives V , Academic Press.
    [7] X. Guo, Y. Peng and H. Qin, Three-variable Mahler measures and special values of L-functions of modular forms, Ramanujan J..
    [8] Lalin M. N., Rogers M. D. (2007) Functional equations for Mahler measures of genus-one curves. Algebraic Number Theory 1: 87-117.
    [9] K. Ono, The Web of Modularity: Arithmetic of the Coefficients of Modular Forms and Q-series, , Amer.Math. Soc., Providence, RI, 2004.

    MR2020489

    [10] Rogers M. (2009) New 5F4 hypergeometric transformations, three-variable Mahler measures, and formulas for 1/π. Ramanujan J. 18: 327-340.
    [11] M. Rogers and W. Zudilin, From L-series of elliptic curves to Mahler measures, Compositio Math., 148 (2012), 385–414. (MR 2904192)

    10.1112/S0010437X11007342

    [12] Rogers M., Zudilin W. (2014) On the Mahler measure of 1+X+1/X+Y+1/Y. Intern. Math. Res. Not. 2014: 2305-2326.
    [13] The Sage Developers, SageMath, the Sage Mathematics Software System (Version 8.6), 2019, http://www.sagemath.org.
    [14] F. Rodriguez-Villegas, Modular Mahler Measures I, Topics in number theory (University Park, PA, 1997), 1748, Math. Appl., 467, Kluwer Acad. Publ., Dordrecht, 1999.
    [15] Samart D. (2013) Three-variable Mahler measures and special values of modular and Dirichlet L-series. Ramanujan J. 32: 245-268.
    [16] Samart D. (2015) Mahler measures as linear combinations of L-values of multiple modular forms. Canad. J. Math. 67: 424-449.
    [17] Smyth C. J. (1981) On measures of polynomials in several variables. Bull. Austral. Math. Soc. 23: 49-63.
    [18] H. Weber, Lehrbuch der Algebra, Bd. III, F. Vieweg & Sohn, Braunschweig, 1908.
    [19] Yui N., Zagier D. (1997) On the singular values of Weber modular functions. Math. Comp. 66: 1645-1662.
  • This article has been cited by:

    1. Qiao He, Dongxi Ye, On conjectures of Samart, 2022, 167, 0025-2611, 545, 10.1007/s00229-021-01279-6
  • Reader Comments
  • © 2020 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3270) PDF downloads(373) Cited by(1)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog