Geometries

Gas Phase Calculations

In order to test the reliability of the semiempirical calculations, additional work was done on the radical anion of nitrobenzene. Restricted open-shell Hartree-Fock (ROHF) geometry optimizations with the 6-31G* basis set and restricted open-shell second order pertubation theory (ROMP2) single-point calculations with the cc-pVDZ basis set at the ROHF geometries were performed. The radical anion of nitrobenzene was shown in previous work to have C2v symmetry (12). As seen above, AM1-SM2 calculations predict the nitro groups in several of the compounds examined here to be rotated out of the plane of the ring and moreover to be pyramidal.

These two graphs illustrate the energetic cost of pyramidalization of the nitrobenzene radical anion as calculated at the ROMP2/cc-pVDZ//ROHF/6-31G* level. The minimum energy structure in the top graph starts as a planar C2v conformation. Forcing the oxygen atoms out of plane increases the energy. The lower graph, on the other hand, illustrates the C2v conformer with the nitro group perpendicular to the plane of the benzene ring to be a saddle point. Pyramidalizing the nitro group decreases the energy to a minimum at about 20 deg. out of plane. However, at this level of theory the barrier to overall rotation of the nitro group is approximately 25 kcal/mol.

To examine the effects of an ortho substituent, the rotation of the nitro group in 2-nitrotoluene radical anion was compared with nitrobenzene radical anion. Again ROMP2/cc-pVDZ single point calculations with an ROHF/6-31G* optimized geometry were used, and the results are shown here. With a methyl substituent ortho to the nitro group, the barrier to rotation drops to 13.0 kcal/mol.

Additional work was done to confirm the planar geometry of nitrobenzene in solution. Density functional theory using the B3LYP method with a 6-311G** basis set and accounting for solvation effects using a Kirkwood-Onsager reaction field model with a solvent dielectric of 30.0 (HMPA) predict carbon hyperfine coupling constants of -6.0 G at C1 and 5.4 G at C4 in the planar C2v geometry. Similar calculations were performed on a geometry in which the oxygen atoms were constrained to lie 20 deg. out of the plane of the nitrogen and benzene ring (i.e., the nitro group is pyramidalized). The coupling constants were found to be 5.8 G at C1 and 3.0 G at C4. Experimentally determined coupling constants in HMPA are C1=-7.0 G and C4=6.1 G. Overall, the computational results are consistent with a planar geometry for nitrobenzene radical anion (13). We note in support that density functional methods also give electron spin density hyperfine coupling constants in good agreement with experiment for the phenoxyl radical (14).

An electron spin resonance study of 1,3-dialkyl-2-nitroaromatic radical anions has shown that the nitro group is rotated out of the aromatic ring based on an observed increase in spin localization on the nitro group. The nitrogen was further proposed to be pyramidal for these radical anions (15).

So, although the gas-phase calculations discussed above indicate there to be a substatial barrier to rotating the nitro group out of the plane of the aromatic ring in the radical anions, it is apparent that substituent effects can significantly affect the rotational potential. As discussed in the next section, solvation effects further influence this rotation.

Effects of Solvation

As noted in the Results section, AM1-SM2 calculations predict the nitro group to be rotated as much as 90 deg out of the plane of the aromatic ring. In aqueous solution, the preferred geometry of 2,4-dinitroaniline rotates the ortho nitro group perpendicular to the aromatic ring and breaks the hydrogen bonds present in the preferred structure in the gas phase. This figure illustrates the differences in the 2,4-dinitroaniline geometries in the gas phase and solution. The solvated geometry shown here is favored by 9.2 kcal/mol over a geometry similar to the one found in the gas phase. Also shown are the gas-phase and aqueous geometries of 2,4-dinitrophenol. Once again, the effect of aqueous solvation is to rotate the nitro group out of plane in the radical anion. However, in this instance the nitro group is rotates less completely in order to maintain some hydrogen bonding between the hydroxyl hydrogen and a nitro group oxygen.

Two other compounds, 2,4-dinitroanisole and 2,4-dinitrobenzaldehyde are also predicted to have quite different gas-phase and solvated geometries. In order for the ortho nitro group to rotate out of the plane of the benzene ring in 2,4-dinitroanisole, the methoxy group rotates into the ring plane. 2,4-Dinitrobenzaldehyde, on the other hand, undergoes both a 180 deg rotation of the aldehyde group and a 90 rotation of the nitro group out of plane in passing from the gas phase to aqueous solution. Further discussion on the importance of rotation of the aldehyde group may be found further below.

As noted in the Results section, 2,4-dinitrobromo- and 2,4-dinitrochlorobenzene both have two low-energy geometries in solution, one with the ortho nitro group in the plane of the aromatic ring, and one with it rotated out of that plane. The predicted regioselectivities of reduction are different for the two structures. In the chloro case, this may well rationalize the observation of limited regioselectivity in the reduction step. However, the bromo case is quite selective experimentally. This probably illustrates the necessary limitations in this qualitative approach to assessing regioselectivity. Actual regioselectivity is dependent on the rate of competing protonations. Moreover, if there are multiple low-energy minima, the number of rates that must be considered increases. Obviously, evaluation of charge localization does not really provide any quantitative insight into these rate phenomena (although it might be interesting to attempt a QSAR-like analysis using calculated properties in an effort to correlate the observed selectivities).

So, the above examples, taken together with the gas-phase calculations described in the preceding subsection, suggest that the potential for rotation of the ortho nitro group can undergo a qualitative change when the effects of solvation (and sometimes substitution) are included. Specifically, solvation often preferentially stabilizes the out-of-plane geometry because that geometry permits greater localization of partial negative charge on the ortho group. This effect works in concert with the steric influence of the substituent to which the nitro group is ortho, as judged by the gas-phase nitrotoluene calculations. Thus, the aqueous geometries predicted by the AM1-SM2 model are presumed to be reliable (and, of course, the high predictive ability of the model provides further support for this contention).

Solvation Polarization of the Electronic Structures

This figure illustrates the importance of solvation on the electronic structure of the radical anions. Pictured is a comparison of the gas phase (center) and solvated (right) electrostatic potential surface of 2-bromo-4,6-dinitroaniline. The gas phase potential surface shows almost no charge localization in comparison to the solvated potential surface. The larger concentration of negative charge on the ortho nitro group rationalizes the experimentally observed reduction selectivity. This enhanced differentiation of the two nitro groups as a result of aqueous solvation occurs for all of the compounds presented above.

Comparison of Point Charges and Electrostatic Potential

The table below compares for each nitro group in those substrates having only one low-energy geometry the sum of the Mulliken and CM1 point charges on the nitrogen and the two oxygen atoms. In all compounds but one, 2,4-dinitrobenzaldehyde, the selectivity of the reduction reaction can in fact be predicted on this less informative basis. That is, the nitro group bearing the larger negative charge is preferentially reduced. For the molecules presented here, this analysis may be accomplished using either the Mulliken or the CM1 point charges. The latter, however, provide a more accurate description of the electronic structure, as they are designed to reproduce experimental charge distributions in organic molecules.

Table 1 Calculated Mulliken and CM1 point charges

Compound                                 Mulliken      CM1


2,4-dinitroaniline            ortho     -0.9870      -0.8830

                              para      -0.2457      -0.2358


2-bromo-4,6-dinitroaniline    ortho     -0.5408      -0.4280

                              para      -0.2196      -0.2536


2,4-dinitroanisole            ortho     -0.9561      -0.8830

                              para      -0.2022      -0.1928


2,4-dinitrobenzaldehyde       ortho     -0.2054      -0.2005

                              para      -0.3931      -0.3525


2,4-dinitrophenol             ortho     -0.8837      -0.7906

                              para      -0.2052      -0.2021

  
2,4-dinitrotoluene            ortho     -0.2283      -0.2332

                              para      -0.6092      -0.4914


2-amino-4,6-dinitrotoluene    ortho     -0.6168      -0.4829

                              para      -0.2205      -0.2245


2,4,6-trinitrotoluene         ortho     -0.1939      -0.1866

                              ortho     -0.2039      -0.2020

                              parai     -0.4294      -0.3483

In one instance, however, analysis of the electrostatic potential surface analysis proves critical to predicting the selectivity of nitro group reduction: the simplified analysis discussed above fails. 2,4-Dinitrobenzaldehyde is the case in point. Rather than any site associated with the two nitro groups, it is the carbonyl oxygen that is the most negative site in this compound. (See Results section for electrostatic potential surface) In the solvated structure, this oxygen rotates so as to lie in close proximity to the ortho nitro group--presumably this is another example of the tendency to localize negative charge as a function of solvation. It appears that protonation of the nitro group is assisted by the concentration of negative charge in the region of the ortho position. Protonation of the carbonyl is unlikely, since the resulting product is 8.7 kcal/mole higher than the product from protonation of the adjacent ortho nitro oxygen. This argument is consistent with the experimentally observed selectivity.

2,4-dinitrophenolate (reduction performed at pH of 7.0, see Experimental section for more details) is experimentally observed to be selectively reducted at the ortho nitro group. As seen here, one would predict the para nitro group to be selectively reduced based on charge localization in the radical dianion. Given the accuracy of the charge localization model in all of the other systems, and given the thermodynamic barrier to reducing an anion, we would argue that reduction probably occurs with the small fraction of 2,4-dinitrophenol (i.e., the neutral conjugate acid) present and not through the dianion.

Return to Table of Contents

Summary

In summary, it appears that analyses of the AM1-SM2 electrostatic potential surfaces for a series of substituted aromatic compounds having two or three nitro groups provides a reliable method of predicting the regioselectivity of their reduction to aromatic amines. The electrostatic potential surface provides a slightly more reliable method than just a comparison of sums of point charges on each nitro group. Inclusion of solvation effects is critical in order to observe any electrostatic preference for the reduction of one nitro group over the others.

Return to Table of Contents