Half Sandwich Compounds

C₅H₅In

This page contains information regarding the calculations I am doing with Gaussian94 on the compound cyclopentadienyl indium. This compound has already been studied in our lab and we have determined several key parameters in the structure.

For the main isotopomer ¹¹⁵In

Parameter	Value
C₅H₅-In (A)	2.322(9)
C-C (A)	1.424(24)
eQq (MHz)	-119.99(3)

    My goal is to duplicate these values using restricted Hartree Fock and Density Functional Theory.
    I have just completed my first successful Gaussian94 restricted Hartree Fock calculations, here is the input file which contains the Z-matrix and pseudopotential.
    The Restricted Hartree-Fock (RHF) method was chosen as a starting point to obtain a structure near the (computed) minimum energy surface, which will then be used as input into the DFT programs.
    I have been modifying the number of independent basis functions on each of the centers, some bring me closer to the answer, while others do not, a few output files are incomplete because the program crashed for various reasons.
    Restricted Hartee Fock Caclulation with In psuedopotential In(5s,5p,4d)[3/21/4] C(9s,5p)[6111/41] cpin3.out. The main parameter Cp-In distance is much to large, by .16 A.
    Restricted Hartee Fock Caclulation with In psuedopoetntial In(5s,5p,4d)[3/21/31] C(9s,5p)[6111/41] cpin4.out. The inclusion of a free 'd' orbital on the Indium atom improves the Cp- In bond length to within .11 A of the measured value.
    Restricted Hartee Fock Caclulation with In psuedopotential In(5s,5p,4d)[3/21/211] C(9s,5p)[6111/41] cpin5.out. Increased 'd' character on the metal atom made the parameters slightly worse.
    Restricted Hartee Fock Caclulation with In psuedopotential In(5s,5p,4d)[21/21/211] C(9s,5p)[6111/41] cpin6.out . More metal 's' character increases the accuracy to within .097 A.
    Restricted Hartee Fock Caclulation with In psuedopotential In(5s,5p,4d)[21/111/211] C(9s,5p)[6111/41] cpin7.out. Increased metal 'p' character only improves the values slightly.
    Restricted Hartee Fock Caclulation with In psuedopotential In(5s,5p,4d)[21/111/211] C(9s,5p)[6111/311] cpin8.out . An attempt is made to improve the error by introducing more pi character into the ring, however this change only decreases the C-C bond length which was already lower than the experimental value and moves the In slightly closer to the ring.

Density Functional Theory

This method was chosen in order to deal with the large number of electrons in this system and get a calculation that runs a finite amount of time. DFT incorporates a relatively high level of theory into a computation with only a minimal increase in computation time. This is accomplished by minimizing the number of integrals that need to be calculated during each iteration.

My first DFT calculation was done with the same parameters as in cpin8.out above and changing the method to B3LYP. The results were worse than the last RHF calculation for the In parameters, however the C-C bond length did slightly improve. Also, at this point, a switch was found that will calculate electric field gradients for the electrons in the molecule, this switch was turned on so that once the field gradient at the In nucleus due to the other nuclei was calulated, the quadrupole coupling might be calculated. For a peek at these results see dft.out.

Since the structural results overall got worse, I decided to use a method recommended by Dr. Barfield for transition metals (yes I know In is in the p-block, but it's close enough). This method is known as B3PW91, and the initial results were promising, again using the same psuedopotential parameters as in cpin8.out. These results are shown in the file dft2.out.

Next the psuedopotential was modified further to In(5s,5p,4d)[111/111/1111] C(9s,5p)[51111/211] H[211] These results are in the file, dft3.out. They show a marked increase in accuracy over the previous calculation.

Sticking with the B3PW91 method, I further modified the psuedopotential parameters to In(5s,5p,4d)[111/111/1111] C(9s,5p)[51111/11111] H[1111] this calculation is in progress, this did not significantly alter the results see dft4.out, frustrated I took the final step in basis set madness and stepped up to the parameters In(5s,5p,4d)[111/111/1111] C(9s,5p)[111111111/11111] H[1111]. These results are in the file dft6.out, as you can see the main parameter is still about 5% away from our measured value.

At this point I decided to take some more advice from Dr. Barfield, who had also suggested that I try one of Gaussian's ready made basis sets instead of a psuedopotential, there were only two to choose from and I chose LanL2DZ, which contains parameters for all elements up to bismuth in double zeta functions. This method gave remarkable answers with 1% accuracy on the structural parameters for output on this method see the files DZ.out and DZ2.out.

These are results that I am confident with, it makes me wish I had tried LanL2DZ first, but I'm sure I would have tried the psuedopotential afterwards anyway. As for calculation of the quadrupole coupling, I still have some work to do in determining whether or not this is feasible. The coupling has been calculated for small nuclei, however, for large nuclei there is not just field gradients due to the other nuclei in the molecule and the valence (bonding) electrons, there are also asymmetric field gradients due to polarization of the core electrons that are usually assumed to be spherically symmetric! So until I find out more about this polarizability of the core electrons I am going to move on and try to find more structural data on other compounds of interest to our lab.

CpTl

This molecule, CpTl is isoelectronic with CpIn, since In and Tl are both in the first column of the p-block of the periodic table. We measured rotational spectra of eight different isotopomers of this species and were able to determine the complete structure of the molecule including the amount of H to carbon plane bending,

Parameter	Value
Tl-C₅ (A)	2.41(1)
C-C (A)	1.43
C-H (A)	1.081
C₅-H (degrees)	1.0(2)

Considering the similarities betwen this compound and CpIn, I immediately tried to use LanL2DZ and B3WP91 (see Tl.out), to much dismay the results were about as inaccurate as my initial calculations on CpIn with psuedopotential. Again the metal to carbon plane distance was much to big by about 5%. Since this molecule has considerably more core electrons, I deemed the descrepancies between LanL2DZ's results to relativistic effects.

I am now running calculations using a psuedopotential and DFT in order to try and increase the accuracy of these calculations. The first of these calculations changed the conformation of the molecule to more experimentally reasonable parameters, however the calculation failed to converge, here is the end of this output file.

I believe that thes particular compounds will require relativistic core potentials in order to resolve the differences in their agreements between theory and experiment.