All extrapolations in W2 theory are normally done on components of
the *total atomization energy*, not of the *total energy*. (Some 
reliance on cancellation between atom and molecule is relied upon.)
For the benefit of people who would like to report "W2 total energies",
we here put W2 values for the total energies of the atoms. 

         SCF limit   CCSD-SCF   (T)         frozen core  core corr. E(W2)       scalar rel. spin-orbit  E(W2)   
                     val.limit  val.limit   limit                   nonrel.                                      
H(2S)     -0.500022   0.000000   0.000000    -0.500022   0.000000    -0.500022  -0.000007   0.000000    -0.500028
B(2P)    -24.529141  -0.071323  -0.001687   -24.602151  -0.044682   -24.646833  -0.006795  -0.000046   -24.653674
C(3P)    -37.688735  -0.098181  -0.002685   -37.789601  -0.048308   -37.837909  -0.015050  -0.000135   -37.853094
N(4S)    -54.401056  -0.126121  -0.003194   -54.530371  -0.051852   -54.582223  -0.029379   0.000000   -54.611602
O(3P)    -74.812681  -0.188512  -0.004282   -75.005475  -0.054737   -75.060212  -0.052387  -0.000355   -75.112954
F(2P)    -99.411966  -0.252012  -0.005386   -99.669364  -0.057573   -99.726938  -0.087173  -0.000614   -99.814725
Al(2P)  -241.876883  -0.056632  -0.001562  -241.935077  -0.272005  -242.207082  -0.435782  -0.000340  -242.643205
Si(3P)  -288.854531  -0.082216  -0.003258  -288.940004  -0.288872  -289.228876  -0.602539  -0.000682  -289.832097
P(4S)   -340.718740  -0.108115  -0.004998  -340.831853  -0.300512  -341.132365  -0.814307   0.000000  -341.946673
S(3P)   -397.507228  -0.161199  -0.006677  -397.675104  -0.307406  -397.982510  -1.078965  -0.000892  -399.062367
Cl(2P)  -459.483910  -0.213836  -0.008326  -459.706072  -0.316593  -460.022664  -1.405037  -0.001340  -461.429041


For comparison, here are limits with best basis from electron affinities paper:
G. de Oliveira, J. M. L. Martin, F. De Proft, and P. Geerlings, Phys. Rev. A, in press.

                  ACV5Z 1st row      
         best [a]   best [a]    AVQZ      best [b]    best      best [c]    best [c]    best
         SCF limit CCSD(T)-SCF  CCSDT-    FCI-CCSDT   valence   (2s,2p)core (1s)core    nonrelativ.
                     limit      CCSD(T)              corr.only                          energy
H(2S)    -0.500000   0.000000   0.000000   0.000000   -0.500000              0.000000    -0.500000
B(2P)   -24.529137  -0.072918  -0.000413   0.000007  -24.602462             -0.051632   -24.654094
C(3P)   -37.688692  -0.100690  -0.000491  -0.000018  -37.789891             -0.055328   -37.845219
N(4S)   -54.400932  -0.129001  -0.000421   0.000010  -54.530344             -0.059151   -54.589495
O(3P)   -74.812392  -0.192368  -0.000354   0.000018  -75.005096             -0.062518   -75.067614
F(2P)   -99.411405  -0.256940  -0.000189  -0.000059  -99.668593             -0.065784   -99.734378
Al(2P) -241.876880  -0.058220  -0.000461   0.000003 -241.935558  -0.322868  -0.074003  -242.332428
Si(3P) -288.854522  -0.085568  -0.000708  -0.000078 -288.940876  -0.315580  -0.072670  -289.329127
P(4S)  -340.718740  -0.112615  -0.000758  -0.000124 -340.832237  -0.331245  -0.076841  -341.240323
S(3P)  -397.507217  -0.168245  -0.000799  -0.000162 -397.676422  -0.338780  -0.077982  -398.093184
Cl(2P) -459.483896  -0.222719  -0.000682  -0.000201 -459.707497  -0.344386  -0.078633  -460.130516

[a] AVnZ (n=4,5,6) for H, B, C, N, O, F; n=3,4,5 for Al and Si; dAVnZ (n=3,4,5) for P, S, Cl
[b] AVQZ for B and Al; AVTZ for C and Si; AVDZ for rest
[c] extrapolated from MTavqz and MTav5z basis sets for 2nd row; from ACVQZ and ACV5Z basis
sets for 1st row

Note that limits for core correlation are much less reliable than those for valence correlation,
since the basis set increments involved are much larger. (Most of this of course cancels between
atom and molecule, which is why you can get away with using a basis set like MTsmall for inner-shell
correlation.)

Last updated July 21, 1999 by comartin@wicc.weizmann.ac.il