Preview only show first 10 pages with watermark. For full document please download

Evaluation Of Two-center Overlap Integrals Using Slater Type Orbitals In Terms Of Bessel Type Orbitals

Evaluation of Two-Center Overlap Integrals Using Slater Type Orbitals In Terms of Bessel Type Orbitals

   EMBED


Share

Transcript

  Journal of Mathematical Chemistry Vol. 37, No. 1, January 2005 (© 2005) Evaluation of two-center overlap integrals using slatertype orbitals in terms of bessel type orbitals Telhat ¨Ozdo˘gan*Metin OrbaySalih De˘girmenci Department of Physics, Faculty of Education, Ondokuz Mayis University, 05189, Amasya-Turkey E-mail: [email protected] Received 3 April 2004; revised 29 June 2004 Using the definition of STOs in terms of BTOs, we have presented analyticalformula for two-center overlap integrals. The obtained formula contains generalizedbinomial coefficients and Mulliken integrals  A k  and  B k . Taking into account the recentadvances on the efficient calculation of Mulliken integrals (Harris,  Int. J. QuantumChem.,  100 (2004) 142), we have obtained many more satisfactory results for two-centeroverlap integrals, for arbitrary quantum numbers, scaling parameters, and location of atomic orbitals. KEY WORDS:  Slater type orbital, Bessel type orbital, two-center overlap integral PACS No:  31.15.+q AMS Subject Classification:  81V55, 81–08 1. Introduction The  ab initio  calculations using the LCAO–MO approach, where molec-ular orbitals are built from a linear combination of atomic orbitals, stronglydepend on the choice of the basis functions for the reliability of the electronicdistributions they provide. A good atomic orbital basis should satisfy two prag-matic conditions for analytical solutions of the appropriate Schr¨odinger equa-tion, namely the cusp at the origin [1] and exponential decay at infinity [2].Among the basis functions used in the literature, the Slater type orbitals (STOs)satisfy the aforementioned requirements [3,4]. Unfortunately, the notorious prob-lems arise in the evaluation of multicenter integrals when STOs used. Due tothe advance in applied mathematics and computer science, there are a progres-sive interest in the use of STOs in multicenter integrals, (see [3,4] and quotedtherein). Since the calculations of multicenter integrals over STOs is extremelydifficult, STOs are expressed as linear combinations of Bessel type orbitals ∗ Corresponding author.27 0259-9791 / 05 / 0100-0027 / 0 © 2005 Springer Science+Business Media, Inc.  28  T. ¨ Ozdo˘  gan et al./Evaluation of two center (BTOs) [5] then and all the multicenter integrals can be calculated more easily(see [6] detailed discussion of BTOs).Among the multicenter integrals, two-center overlap integral constitutes thebasic building block of many more complicated multicenter integrals since theexpectation value of one-electron operators can be expressed as linear combi-nation of two-center overlap integrals, and also since two-electron integrals canbe calculated by taking quadrature of two-center overlap integrals. At the sametime, these type integrals also contribute to the total energy of the moleculewhich is required to a precision sufficient for small fraction changes to be evalu-ated. In practice, an error in the range of 10 − 8  –10 − 10 in integrals will deterioratethe energy by an amount of 10 − 3 atomic units (a.u.) [4].Recently, we have presented analytical, series and recurrence relations forthe evaluation of multicenter integrals over STOs [7]. In this work, we aredealing with the evaluation of two-center overlap integrals using STOs as linearcombinations of BTOs, which seems to be quite promising for use in  ab initio calculations. 2. General definitions and basic formulas In the most real case, a STO is defined as follow: χ nlm (α,  r)  =  N  n (α)r n − 1 e − αr S  lm (θ,ϕ),  (1)where  α  is scaling parameter and  N  n (α)  is the normalization coefficient given by N  n (α)  =  ( 2 α) n + 12 √  ( 2 n) !(2)and  S  lm (θ,ϕ)  denotes the real spherical harmonic [8]: S  lm (θ,ϕ)  =  P  l | m | ( cos θ ) m (ϕ),  (3)in which  P  l | m |  is the normalized associated Legendre function and the   m (ϕ)  isdefined by  m (ϕ)  =  1 √  π ( 1  +  δ mo )   cos mϕ  for  m    0 , sin | m | ϕ  for  m <  0 .  (4)Using the expansion formula for unnormalized STOs in terms of BTOs [5],we re-express normalized STOs as finite linear combination of BTOs as χ nlm (α,  r)  =   n (α) n − l  p = ˆ p K pnl B mp,l  (α,  r),  (5)  T. ¨ Ozdo˘  gan et al./Evaluation of two-center  29 where  n (α)  =  α 1 − n N  n (α),  (6a) K pnl  =  ( − 1 ) n − l − p (n  −  l) ! (l  +  p) !2 l + p ( 2 p  −  n  −  l) ! ( 2 n  −  2 l  −  2 p) !!  ,  (6b)and ˆ p  =   (n  −  l)/ 2 if   n  −  l  even ,(n  −  l  +  1 )/ 2 if   n  −  l  odd .  (7)In equation (5), a BTO is defined as follow [5]:B mnl  (α,  r)  =  (αr) l 2 n + l (n  +  l) ! ˆ k n − 12 (αr)S  lm (θ,ϕ),  (8)where  ˆ k n − 12 is reduced Bessel function defined by ˆ k n − 12 (z)  =  z − 1 e − zn  j  = 1 ( 2 n  −  j   −  1 ) ! (j   −  1 ) ! (n  −  j ) !2 j  − n z j  .  (9) 3. Two-center overlap integrals over STOs Two-center overlap integrals examined in the present work have the follow-ing form: S  nlm,n ′ l ′ m ′  α,β ;  R   =    χ ∗ nlm (α,  r a )χ n ′ l ′ m ′  (β,  r b )dV,  (10)where  χ nlm (α,  r a )  and  χ n ′ l ′ m ′  (β,  r b )  are STOs located on centers  a  and  b , and   R is the radius vector given by   R  ≡  R ab  =  r b  −  r a .Substituting equation (5) into equation (10), we obtain the following rela-tion for two-center overlap integrals over STOs S  nlm,n ′ l ′ m ′  α,β ;  R   =   nn ′  (α,β)  p,q K nl,n ′ l ′ pq  I  plm,ql ′ m ′  α,β ;  R  ,  (11)where  nn ′  (α,β)  =   n (α) n ′  (β),  (12a) K nl,n ′ l ′ pq  =  K nlp  K n ′ l ′ q  .  (12b)  30  T. ¨ Ozdo˘  gan et al./Evaluation of two center In equation (11),  I  plm,ql ′ m ′  is the integral of the form: I  plm,ql ′ m ′  α,β ;  R   =    B mpl  (α,  r a )B m ′ ql ′  (β,  r b )dV,  (13)and called two-center overlap integrals between two BTOs. In the literature,this type of integrals are calculated by one center expansion method (reduc-tion theorem) [9] and Fourier transform convolution theorem [10]. It is knownthat Fourier transform convolution theorem injure from some possible insta-bility problems [11] and we think that one center expansion method may alsoinjure from possible instability problems due to the fact that this method con-tains approximations.To overcome the possible instability problems occurring in the calculationof two-center overlap integrals, using rotation coefficients for two-center overlapintegrals [12], we express the integral in equation (11) as I  plm,ql ′ m ′  α,β ;  R   = min ( l,l ′ )  λ = 0 T  λlm,l ′ m ′  (θ,ϕ)I  plλ,ql ′ λ (α,β ; R),  (14)where  T  λlm,l ′ m ′  (θ,ϕ)  is rotation coefficients for two-center overlap integrals and I  plλ,ql ′ λ  α,β ;  R   =    B λpl  (α,  r a )B λql ′  (β,  r b )dV.  (15)Using the definition of the reduced Bessel functions given by equation (9) intoequation (15), we express the overlap integral of two BTOs by the followingformula: I  plλ,ql ′ λ (α,β ; R)  =  γ  ll ′ pqp,q  i,j  η pi η qj  α l + i − 1 β l ′ + j  − 1 ×    r l + i − 1 a  r l ′ + j  − 1 b  e − αr a − βr b P  lλ ( cos θ  a )P  l ′ λ ( cos θ  b )dV,  (16)where γ  ll ′ pq  =  (p  +  l) !!  q  +  l ′  !!  − 1 ,  (17)and η pκ  =  F  κ − 1 ( 2 p  −  κ  −  1 )F  p − κ  ( 2 p  −  2 κ)(p  −  κ) !2 κ − p .  (18)Here  F  m (n)  is usual binomial coefficient defined by F  m (n)  =  n ! m ! (n  −  m) ! .  (19)  T. ¨ Ozdo˘  gan et al./Evaluation of two-center  31 As can be seen equation (16), it is necessary to have an equation for theexpansion of the product of two normalized associated Legendre functions cen-tered on  a  and  b : T  lλ,l ′ λ (θ  a ,θ  b )  =  P lλ ( cos θ  a ) P l ′ λ ( cos θ  b ).  (20)Recently [7a], we have presented the expansion formula for the product of two normalized associated Legendre functions in ellipsoidal coordinate system (µ,ν,ϕ)  as T  lλ,l ′ λ (µ,ν)  =  k,k ′  u,s a kk ′ us  lλ,l ′ λ   (µν) s (µ  +  ν) l − 2 (k + k ′ + λ) + 2 u (µ  −  ν) l ′  ,  (21)where the expansion coefficients are a kk ′ us  lλ,l ′ λ   =  C klλ C k ′ l ′ λ ( − 1 ) u F  u  k  +  k ′  +  λ  × F  s  l  −  2 k  −  λ  +  2 u,l ′  −  2 k ′  −  λ  ,  (22)and C klm  =  ( − 1 ) k 2 2 k + m  2 l  +  12  F  l − k  (l  +  m)F  k + m (l  −  k)F  2 k  (l  −  m)F  k  ( 2 k)  1 / 2 (23)and the ranges of the summation indices  k,k ′ ,u,  and  s  are as follows:0    k    E  l − λ 2  ,  0    k ′    E  l ′  −  λ 2  , 0    u    k  +  k ′  +  λ  ,  0    s    l  +  l ′   −  2  k  +  k ′  +  λ   +  2 u.  (24)In equation (22),  F  m  N,N  ′   is called generalized binomial coefficients [12,13] and recently we have re-expressed this quantity in terms of usual binomialcoefficients as [14] F  m  N,N  ′   = m  i,j  = 0 ( − 1 ) N  ′ − j  F  i  (N)F  j   N  ′  δ m,i + j  .  (25)Consequently, using equation (21) and ellipsoidal form of the radial partincluded in equation (16), two-center overlap integrals between two BTOs takethe form: I  plλ,ql ′ λ (α,β ; R)  =  γ  ll ′ pqp,q  i,j  η pi η qj  α l + i − 1 β l ′ + j  − 1  R 2  l + l ′ + i + j  + 1 ×  kk ′ ,us a kk ′ us  lλ,l ′ λ  Q si + 2 (k + k ′ + λ − u),j   (p,t ),  (26)