ON LARA RODRÍGUEZ’ FULL CONJECTURE FOR DOUBLE ZETA VALUES IN FUNCTION FIELDS 7 1 RYOTAROHARADA 0 2 n a J Abstract. ThispaperdiscussesfourformulaeconjecturedbyJ.A.LaraRo- 7 dríguez on certain power series in function fields, which yield a ’harmonic 1 product’ formulaforThakur’s doublezetavalues. Weproveaffirmatively the firsttwoformulae. Whilewedetectandcorrecterrorsinthelasttwoformulae, ] andprovethecorrected ones. T N Contents . h 0. Introduction 1 t a 1. Notations and Definitions 2 m 1.1. Notations 2 [ 1.2. Definition of multiple zeta values in F t 2 2. Lara Rodríguez’ full conjecture and couqn[te]r-examples 3 1 v 2.1. Statements 3 2 2.2. Remarks and Counter-examples 4 5 3. Main results 6 5 Acknowledgments 18 4 References 18 0 . 1 0 7 0. Introduction 1 : v Allegedly, in 1776, the double zeta values (multiple zeta values with depth 2) i were firstly introduced by L. Euler in [4] where he also described three types of X relations for double zeta values with non-mathematical proofs and unconventional r a notations(theywerereformulatedwithmathematicalproofsandconventionalmod- ernnotationsin[5]). Itis saidthatthe multiple zetavalueswererediscoveredafter the silence of more than two centuries. In the last quarter century, it got known thatthey haveconnectionto number theory([3], [14]), knottheory ([9])andquan- tumfieldtheory([1])andsoon. Findinglinear algebraicrelationsformultiplezeta / valuesisoneofourfundamentalissues. Especially,the shuffleproductformulaand the harmonic product formula were discussed in detail in [6]. In2004,thefunctionfieldanaloguesofthemultiplezetavalueswereinventedby D. S. Thakur in [11]. He showed the existence of the ’harmonic product’ formula for them in [13]. While in [7], J. A. Lara Rodríguez conjectured its precise formu- lation in the case of depth 2 with bounded weights. This conjecture contained five formulae. The first formula was proved by himself in [8]. By using H. J. Chen’s Date:January17,2017. 1 2 RYOTAROHARADA result in [2], we will prove affirmatively the second and third formulae in Theorem 9 and 10. Whereas we detect and correct errors in the fourth and fifth formulae, and prove corrected ones in Theorem 11 and 12. 1. Notations and Definitions 1.1. Notations. We recall the following notation used in [8]. q a power of a prime number p, q=pl. F a finite field with q elements. q A the polynomial ring F t . q [ ] A the set of monic polynomials in A. + Ad+ the set of elements of A+ of degree d. F t the rational function field in the variable t. q F 1 (t ) the completion of F t at . q q (( / )) ( ) ∞ Int x =⎧⎪⎪0 if x is not an integer, ( ) ⎨1 if x is an integer. ⎪⎪ ⎩ 1.2. DefinitionofmultiplezetavaluesinF t . Firstwerecallthepowersums. q For s∈Z and d∈Z , we write [ ] ≥0 1 S s = ∈F t . d( )∶ ∑ as q( ) a∈Ad+ For positive integers s ,s ,...,s ∈Z and d∈Z , we put 1 2 n >0 ≥0 S s ,s ,...,s =S s S s ⋯S s ∈F t . d( 1 2 n)∶ d( 1) ∑ d2( 2) dn( n) q( ) d>d2>⋯>dn≥0 For s∈Z , the Carlitz zeta values are defined by ≥0 1 ζ s ∶= ∈F 1 t . ( ) ∑ as q(( / )) a∈A+ Thakur generalized this definition to that of multiple zeta values for F t in [11]. q For s ,s ,...,s ∈Z , [ ] 1 2 n >0 ζ s ,s ,...,s ∶= S s ⋯S s ( 1 2 n) ∑ d1( 1) dn( n) d1>d2>⋯>dn≥0 1 = ∈F 1 t . dega1>dega∑2>⋯>degan≥0as11⋯asnn q(( / )) a1,a2,...,an∈A+ For a,b∈Z , we define >0 ∆ a,b ∶=S a S b −S a+b . d d d d ( ) ( ) ( ) ( ) H.J.Chenprovedthe followingformulaforthe powersumsin[2]Theorem3.1and Remark 3.2. Proposition 1 (Chen’s formula). For r,s,d∈Z , the following relation holds. ≥1 j−1 j−1 ∆ r,s = −1 s−1 + −1 r−1 S i,j . d( ) ∑ {( ) (s−1) ( ) (r−1)} d( ) i+j=r+s q−1∣j i,j≥1 Here we put a =0 for a,b∈Z with a<b. (b) ≥0 3 WecandeterminethevalueofthebinomialcoefficientsmodulopbyusingLucas’s theorem ([10] Section 3). Proposition2 (Lucas’sTheorem). Let p be aprime numberand m,n∈Z . Then ≥0 we have m m m ≡ 0 ⋯ k mod p (n) (n0) (nk) where m=m +m p+⋯+m pk and n=n +n p+⋯+n pk m ,n ∈ 0,1,...,p−1 0 1 k 0 1 k i i for i=0,1,...,k are p-adic expansions of m and n. ( { } ) 2. Lara Rodríguez’ full conjecture and counter-examples Lara Rodríguez conjectured several relations for Thakur’s double zeta values in [7]. We recall it in Section 2.1. We detect some typos and errorsin his formulae in Section 2.2. 2.1. Statements. The following is one of those conjectures which he called the full conjecture ([7] Conjecture 2.8). It yields ”full” descriptions of the ’harmonic product’ formula for specific double zeta values (cf. [7] Section 1). Conjecture 3 (Lara Rodríguez’ full conjecture). For n,d∈Z and general q, we ≥1 have 2 (1) ∆ qn+1,qn =Int S 2,2qn−1 d d ( ) (q) ( ) qn−1 q−1 − S 3+ j−1 q−1 ,2qn−2− j−1 q−1 . d ∑ ( ( )( ) ( )( )) j=1 (2) qn+q−2 q−1 ∆ qn−1,qn+1 =− S 2+ j−1 q−1 ,2qn−2− j−1 q−1 . d d ( ) ∑ ( ( )( ) ( )( )) j=1 (3) 2 ∆ qn−1,qn+1 =Int S 2,qn+qn−1−1 d d ( ) (q) ( ) qn−1−1 q−1 − S 3+ j−1 q−1 ,qn+qn−1−2+ j−1 q−1 . d ∑ ( ( )( ) ( )( )) j=1 For 0≤i≤n, we have (4) 2 ∆ qn+1,qn+1−qi =Int S 2,2qn−qi d d ( ) (q) ( ) qn−qi q−1 − S 3+ j−1 q−1 ,2qn−qi−1− j−1 q−1 d ∑ ( ( )( ) ( )( )) j=1 qn−qi q−1 + S 3+ j−1 q−1 ,2qn−qi−1− j−1 q−1 . d ∑ ( ( )( ) ( )( )) j=qnq−−1qi+1 4 RYOTAROHARADA 2.2. Remarks and Counter-examples. Remark 4. Actually,in[7]Conjecture2.8(2.8.1),LaraRodríguezconjecturedone more relation ∆ qn,qn−1 =−S qn,qn−1 . d d ( ) ( ) However he proved it in his later paper [8] Theorem 6.3. Remark 5. The equation (1) was stated as [7] (2.8.2). In the case when q = 2, this coincide with second formula in [11] Section 4.1.3. The equation (1) will be affirmatively proven in Theorem 9. Remark 6. The equation(2)wasstatedas[7](2.8.3). Inthecasewhenq=2,this coincide with third formula in [11] Section 4.1.3. Again, the equation (2) will be affirmatively proven in Theorem 10. Remark 7. The equation (3) was stated as (2.8.4) in [7] (in the case when q =2, this coincide with fourthformula in [11] Section4.1.3). It looks that (3) contains a typo, and furthermore it requires an additional term to correct it. Indeed it is quite curious to expect such an equality among the values with differentweights(thesumofthefirstandthesecondcomponentsofdoubleindices): Inthe righthandside ofthe equation(3),the firsttermis withweightqn+qn−1+1 while the summandof the secondterm is with weight qn+qn−1+1+2 j−1 q−1 . In the case when q=2, d=2 and n=3, the equation (3) claims ( )( ) (5) ∆ 4,9 =S 2,11 −S 3,10 −S 4,11 −S 5,12 , 2 2 2 2 2 ( ) ( ) ( ) ( ) ( ) while Chen’s formula says j−1 j−1 (6) ∆ 4,9 = − S i,j 2( ) ∑ {( 8 ) ( 3 )} 2( ) i+j=13 i,j≥1 ≡S 2,11 +S 3,10 +S 4,9 +S 5,8 +S 9,4 mod 2. 2 2 2 2 2 ( ) ( ) ( ) ( ) ( ) Therefore we must have (7) S 2,11 +S 3,10 +S 4,11 +S 5,12 2 2 2 2 ( ) ( ) ( ) ( ) −S 2,11 −S 3,10 −S 4,9 −S 5,8 −S 9,4 ≡0 mod 2. 2 2 2 2 2 ( ) ( ) ( ) ( ) ( ) However, S 4,11 +S 5,12 −S 4,9 −S 5,8 −S 9,4 2 2 2 2 2 ( ) ( ) ( ) ( ) ( ) ≡S 9,4 +S 5,12 +S 5,8 +S 4,11 +S 4,9 mod 2 2 2 2 2 2 ( ) ( ) ( ) ( ) ( ) =S 9,4 +S 5 1+S 12 +S 5 1+S 8 2 2 1 2 1 ( ) ( )( ( )) ( )( ( )) +S 4 1+S 11 +S 4 1+S 9 2 1 2 1 ( )( ( )) ( )( ( )) =S 9,4 +S 5 S 12 +S 8 +S 4 S 11 +S 9 2 2 1 1 2 1 1 ( ) ( )( ( ) ( )) ( )( ( ) ( )) 5 Each term is calculated to be S 9,4 2 ( ) 33 ti+t32+t31+t30+t26+t25+t22+t17+t16+t15+t8+t6+t5+t2 i=0 ≡ {∑ } t22 t+1 19 t2+t+1 9 t2+1 5 ( ) ( ) ( ) ⋅ t+1 6 t2+1 5 mod 2, ( ) ( ) S 5 S 12 +S 8 2 1 1 ( )( ( ) ( )) 29 ti+t26+t25+t23+t21+t20+t19+t16+t13+t12+t9+t8+t5 ≡ ∑i=0 t22 t+1 19 t2+t+1 9 t2+1 5 ( ) ( ) ( ) ⋅ t2+t+1 4 t+1 2 mod 2, ( ) ( ) t12+t5+t4+t3+t2+t+1 t3 t2+t+1 8 t2+1 5 S 4 S 11 +S 9 ≡ { } ( ) ( ) mod 2. 2( )( 1( ) 1( )) t22 t+1 19 t2+t+1 9 t2+1 5 ( ) ( ) ( ) The degrees of numerators of S 9,4 , S 5 S 12 +S 8 and S 4 S 11 + 2 2 1 1 2 1 ( ) ( )( ( ) ( )) ( )( ( ) S 9 are49,39and41respectively. Thuswefindthedegreeofeachnumeratoris 1 ( )) different while they have the same denominators. Then it follows that S 4,11 + 2 S 5,12 −S 4,9 −S 5,8 −S 9,4 ≢0 mod 2 and this contradicts to ((7). T)his 2 2 2 2 ( ) ( ) ( ) ( ) gives the counter-example of (3). Therefore, we may correct (3) as follows. 2 (8) ∆ qn−1,qn+1 =Int S 2,qn+qn−1−1 d d ( ) (q) ( ) qn−1−1 q−1 − S 3+ j−1 q−1 ,qn+qn−1−2− j−1 q−1 . d ∑ ( ( )( ) ( )( )) j=1 However, the above equation is not correct, due to a lack of an additional terms which is explained below: When q=3,d=1 and n=3, (8) claims (9) ∆ 9,28 =−S 3 −S 5 −S 7 −S 9 . 1 1 1 1 1 ( ) ( ) ( ) ( ) ( ) But according to Chen’s formula, we have j−1 j−1 ∆ 9,28 = − + S i . 1( ) ∑ { ( 27 ) ( 8 )} 1( ) i+j=37 2∣j i,j∈Z≥1 By Lucas’s theorem, we find that the coefficient of S i ’s vanish modulo 3 except 1 −S 3 ,−S 5 ,−S 7 ,−S 9 and S 19 . That is, ( ) 1 1 1 1 1 ( ) ( ) ( ) ( ) ( ) (10) ∆ 9,28 =−S 3 −S 5 −S 7 −S 9 +S 19 . 1 1 1 1 1 1 ( ) ( ) ( ) ( ) ( ) ( ) By the definition of power sum, 1 1 1 t19 t+2 19+ t+1 19 t+2 19+t19 t+1 19 S1(19)= t19 + t+1 19 + t+2 19 = ( ) t1(9 t+)1 1(9 t+)2 19 ( ) . ( ) ( ) ( ) ( ) The numerator of the right hand side has 219 ≡ −1 mod 3 as a constant term. Therefore S 19 does not vanish modulo 3. Thus (9) contradicts to (10). So this 1 ( ) 6 RYOTAROHARADA suggeststhatweneedadditionaltermstocorrectit. InTheorem11,wecorrectthe equation (3) as the equation (17) and prove it. Remark 8. The equation (4) was stated as (2.8.5) in [7]. Again, it looks that the equation(4)containsatypobecausethesummationofthethirdterminrighthand side runs over the empty sum. We correct the equation (4) as the equation (26) and prove it in Theorem 12. 3. Main results In this section, we provethe first half of Lara Rodríguez’full conjecture in The- orem 9 and 10, we correct and prove the second half of the conjecture in Theorem 11 and 12. Precisely in Theorem 9 and 10, we show that the equations (1) and (2) hold. In Theorem11 and 12, we correctthe equations (3) and (4) as the equations (17) and (26) respectively and give their proofs. Theorem 9. For n and d∈Z , the equation (1) holds. ≥1 Proof. Case 1 (the case when q=2). By Chen’s formula for q=2, j−1 j−1 ∆ 2n+1,2n = + S i,j . d( ) i+j=∑2n+1+1{(2n−1) ( 2n )} d( ) i,j≥1 When j =2n+1, we have 2n+1−1 = 2n+1−1 . So we obtain ( 2n−1 ) ( 2n ) j−1 j−1 + =0 (2n−1) ( 2n ) for j=2n+1. When 1≤j<2n, it is clear that j−1 j−1 + =0+0=0. (2n−1) ( 2n ) When 2n≤j ≤2n+1−1, let j−1=j +j ⋅2+⋯+j ⋅2n 0 1 n be the 2-adic expansion of j−1. The 2-adic expansion of 2n−1 is given as follows 2n−1=1+2+22+⋯+2n−1. By using Lucas’s theorem, j−1 n−1 j j ≡ k n mod 2, (2n−1) ∏(1)(0) k=0 j−1 n−1 j j ≡ k n mod 2. ( 2n ) ∏(0)(1) k=0 Thus we obtain j−1 ≡1 mod 2 j−1=2n−1 or 2n+1−1, (2n−1) ⇔ j−1 ≡1 mod 2 j =1 ( 2n ) ⇔ n 7 We always have j =1 for all j with 2n≤j ≤2n+1−1. Therefore, n j−1 + j−1 ≡⎧⎪⎪ 1 mod 2 if 2n≤j ≤2n+1−1, (2n−1) ( 2n ) ⎨ 0 mod 2 if j≤2n−1 or j=2n+1. ⎪⎪ ⎩ Thus Chen’s formula for q=2 becomes ∆ 2n+1,2n = S i,j . d d ( ) i+j=∑2n+1+1 ( ) 2n≤j≤2n+1−1 Replacingj with2n+1−j,wehave2n−1≤2n+1−j≤2n+1−1andthus2≤j ≤2n+1. Therefore 2n+1 ∆ 2n+1,2n =S 2,2n+1−1 + S j,2n+1+1−j d d d ( ) ( ) ∑ ( ) j=3 2n−1 =S 2,2n+1−1 − S j+2,2n+1−1−j . d d ( ) ∑ ( ) j=1 So we obtain (1). Case 2 (the case when q=pl ≠2). By Chen’s formula, we have j−1 j−1 ∆ qn+1,qn = −1 qn−1 + −1 qn S i,j . d( ) i+j=∑2qn+1{( ) (qn−1) ( ) ( qn )} d( ) q−1∣j We obtain j−1 j−1 j−1 j−1 −1 qn−1 + −1 qn = − ( ) (qn−1) ( ) ( qn ) (qn−1) ( qn ) (we note that the above equation holds for p=2 because the characteristic is 2 in this case). When 0<j≤qn−1 with q−1j, it is easily seen that ∣ j−1 j−1 (11) − =0−0=0. (qn−1) ( qn ) When qn+q−2≤j ≤2qn−2 with q−1j, We put the p-adic expansions of j−1 and qn−1 as follows ∣ j−1=j +j p+⋯+j pln, 0 1 ln qn−1=p−1+ p−1 p+⋯+ p−1 pln−1. ( ) ( ) Applying Lucas’s theorem, we have j−1 ln−1 j j ≡ k ln mod p, (qn−1) ∏ (p−1)( 0 ) k=0 j−1 ln−1 j j ≡ k ln mod p. ( qn ) ∏ (0)( 1 ) k=0 Thus it follows that j−1 ≢0 mod p j =p−1 k∈ 0,1,⋯,ln−1 , (qn−1) ⇔ k ( { }) j−1 ≢0 mod p j ≠0. ( qn ) ⇔ ln 8 RYOTAROHARADA By the condition qn+q−1≤j ≤2qn−2, we have qn+q−1=pln+pl−1≤j≤p−2+ p−1 p+⋯+ p−1 pln−1+pln=2qn−2. ( ) ( ) So we always have j = 1. Then jln = 1 for j with qn+q−1 ≤ j ≤ 2qn−2 and q−1j. It follows thaltn ( 1 ) ∣ j−1 ≡1 mod p. ( qn ) If j =p−1 for all k∈ 0,1,...,ln−1 we have j−1=qn−1+qn=2qn−1 because k we always have j = 1{. This contrad}icts to the condition qn+q−2 ≤j ≤2qn−2. ln Thus we have j−1 ≡0 mod p. (qn−1) for j with qn+q−2≤j≤2qn−2 and q−1j. Therefore ∣ j−1 j−1 (12) − ≡1 mod p (qn−1) ( qn ) for j with qn+q−2≤j≤2qn−2 and q−1j. ∣ Therefore, by (11) and (12), we obtain j−1 − j−1 ≡⎧⎪⎪ 0 mod p if 0<j ≤qn−1 with q−1∣j, (qn−1) ( qn ) ⎨ 1 mod p if qn+q−2≤j≤2qn−2 with q−1j. ⎪⎪ ∣ ⎩ Then Chen’s formula becomes ∆ qn+1,qn =− S i,j . d d ( ) i+j=∑2qn+1 ( ) qn+q−2≤j≤2qn−2 q−1∣j Putting i as 3+ j−1 q−1 and j as 2qn−2− j−1 q−1 , we have qn+q−2≤ 2qn−2− q−1 (j−1)(≤2qn)−2 and thus 1≤j≤(qn−1.)(There)fore ( )( ) q−1 qn−1 q−1 ∆ qn+1,qn =− S 3+ j−1 q−1 ,2qn−2− j−1 q−1 . d d ( ) ∑ ( ( )( ) ( )( )) j=1 Combining Case 1 and Case 2, we obtain the equation (1). (cid:3) Theorem 10. For n and d∈Z , the equation (2) holds. ≥1 Proof. By Chen’s formula, j−1 j−1 ∆ qn−1,qn+1 =− −1 qn−1 + S i,j . d( ) i+j∑=2qn( ) {( qn ) (qn−2)} d( ) q−1∣j We have j−1 j−1 j−1 j−1 −1 qn−1 + = + ( ) {( qn ) (qn−2)} ( qn ) (qn−2) (we note that the above equation holds for q=pl with p=2 because the character- istic is 2). When j <qn−1 with q−1j, it is clear that ∣ j−1 j−1 (13) + =0+0=0. ( qn ) (qn−2) 9 When qn−1≤j ≤qn with q−1j , we have j =qn−1 because j satisfies q−1j. ∣ ∣ Thus in this case we have j−1 j−1 qn−2 qn−2 (14) + = + =0+1=1. ( qn ) (qn−2) ( qn ) (qn−2) When qn<j≤2qn−2 with q−1j,we set the p-adicexpansionofj−1 as follows ∣ j−1=j +j p+⋯j pln j ∈ 0,1,...,p−1 . 0 1 ln k ( { }) By Lucas’s theorem, we have j−1 ln−1 j j ≡ k ln mod p. ( qn ) ∏ (0)( 1 ) k=0 So we have j−1 ≡1 mod p j =1. ( qn ) ⇔ ln Wealwayshavej =1because2qn−2= p−2 + p−1 p+ p−1 p2+⋯+ p−1 pln−1+pln ln and qn=pln<j≤2qn−2. Therefore ( ) ( ) ( ) ( ) j−1 ≡1 mod p. ( qn ) Next we will prove j−1 ≡0 mod p. Again using Lucas’s theorem, we obtain (qn−2) j−1 j ln−1 j 1 ≡ 0 k (qn−2) (p−2) ∏ (p−1)(0) k=1 by qn−2=p−2+ p−1 p+⋯+ p−1 pln−1+0⋅pln. Thus ( ) ( ) j−1 ≢0 mod p p−2≤j ≤p−1 and j =p−1 for all k∈ 1,2,...,ln−1 . (qn−2) ⇔ 0 k { } If j = p−2 and j = p−1 for all k ∈ 1,2,...ln−1 , we have j −1 = p−2+ 0 k p−1 p+⋯+ p−1 pln−1+pln = 2qn−{2. However, j}= 2qn−1 is not divisible (by q−)1. If j (= p−)1 and j = p−1 for all k ∈ 1,2,...ln−1 , we have j−1 = 0 k p−1+ p−1 p+⋯+ p−1 pln−1+pln=2qn−1. Bu{tj =2qn is not}divisible by q−1. ( ) ( ) Thus we always have j−1 ≡0 mod p. (qn−2) Therefore we have j−1 j−1 (15) + ≡1 mod p ( qn ) (qn−2) for j with qn−1≤j≤2qn−2 and q−1j. ∣ By (13), (14) and (15), we obtain (16) j−1 + j−1 ≡⎧⎪⎪ 0 mod p if j <qn−1 with q−1∣j ( qn ) (qn−2) ⎨ 1 mod p if qn−1≤j≤2qn−2 with q−1j. ⎪⎪ ∣ ⎩ Therefore Chen’s formula becomes ∆ qn−1,qn+1 =− S i,j . d d ( ) i+j∑=2qn ( ) qn−1≤j≤2qn−2 q−1∣j 10 RYOTAROHARADA Replacingj with2qn−2− j−1 q−1 ,wehaveqn−1≤2qn−2− j−1 q−1 ≤2qn−2 and thus 1≤j ≤ qn+q−2. T(here)f(ore ) ( )( ) q−1 qn+q−2 q−1 ∆ qn−1,qn+1 =− S 2+ j−1 q−1 ,2qn−2− j−1 q−1 . d d ( ) ∑ ( ( )( ) ( )( )) j=1 Combining Case 1 and Case 2, we obtain the equation (2). (cid:3) As we saw in Remark 7, the equation (3) contains errors. We correct them as follows. Theorem 11. For d,n∈Z , we have ≥1 (17) 2 ∆ qn−1,qn+1 =Int S 2,qn+qn−1−1 d d ( ) (q) ( ) qn−1−1 q−1 − S 3+ j−1 q−1 ,qn+qn−1−2− j−1 q−1 d ∑ ( ( )( ) ( )( )) j=1 +S 2qn−1+1,qn−qn−1 . d ( ) Proof. By 1≤j≤ qn−1−1, we have q−1 qn+q−2≤qn+qn−1−2− j−1 q−1 ≤qn+qn−1−2. ( )( ) By replacing qn+qn−1−2− j−1 q−1 by j, we see that it is enough to prove ( )( ) 2 (18) ∆ qn−1,qn+1 =Int S 2,qn+qn−1−1 − S i,j d d d ( ) (q) ( ) qn+q−2≤j∑≤qn+qn−1−2 ( ) q−1∣j i+j=qn+qn−1+1 +S 2qn−1+1,qn−qn−1 , d ( ) which is a reformulation of (17). We note that Chen’s formula says (19) ∆ qn−1,qn+1 = −1 qn j−1 + −1 qn−1−1 j−1 S i,j . d( ) i+j=qn∑+qn−1+1{( ) ( qn ) ( ) (qn−1−1)} d( ) q−1∣j i,j∈Z≥1 Case 1 (the case when q=2). The equation (19) becomes j−1 j−1 ∆ 2n−1,2n+1 = + S i,j . d( ) i+j=2n∑−1+2n+1{( 2n ) (2n−1−1)} d( ) When 0≤j−1<2n−1−1, it is easily seen that j−1 j−1 (20) + =0+0=0. ( 2n ) (2n−1−1) When 2n−1−1≤j−1<2n, it is clear that j−1 =0. ( 2n )