You are on page 1of 21

F. Bernardi, et al.

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications
Fernando Bernardi, Andrea Bottoni * and Marco Garavelli*
Dipartimento di Chimica G.Ciamician, Universita di Bologna, via Selmi 2, 40126 Bologna, Italy , e-mail: mgara@ciam.unibo.it

Abstract In this review we report the results of DFT investigations which have been carried out in different fields of organic and organometallic chemistry, including radical reactivity, structure and reactivity of organometallic compounds, and biochemical/biophysical properties of long chain unsaturated systems. Many of the most popular non-local corrected functionals (e.g. B3LYP, BHLYP, BLYP, BP86) have been benchmarked both versus experimental and high level ab initio (e.g. MP2, MP4, CAS-SCF/CAS-PT2) data, resulting in an impressive agreement. The DFT approach appears to be a powerful tool, which can be used as a valid alternative to more traditional correlated methods, to achieve mechanistic information of chemical/ physical interest in the modelling of organic and biochemical systems. In particular, in the examples selected in this review, we discuss the results obtained for the addition reaction of alkyl radicals to double bonds and for the hydrogen/ chlorine abstraction reaction by alkyl and silyl radicals from various organic substrates. Moreover, binding interactions (i.e. geometries and energies) in organometallic compounds are shown to be satisfactorily reproduced via DFT and examples of nickel-catalyzed [2 2] cycloaddition reaction and homogeneous Ziegler-Natta catalysis are investigated. Finally, a DFT modelling for the singletoxygen quenching ability and radical trapping activity of carotenes is presented. The simulated data provide a rationale for the protective action of carotenes observed in biological tissues and elucidates the physical and chemical mechanisms involved in the reactivity of carotenes versus oxygen and radicals.

1 Introduction
In the last two decades, density functional theory (DFT) [1 11] has emerged as a practical and versatile tool to obtain accurate information on molecular systems of chemical interest. The popularity of DFT-based methods stems in large measure from its computational expedience that allows to acquire, even for large molecules, thermochemical data, force fields and frequencies, transition state structures, NMR, PES, ESCA, IR and Raman spectra. Thus this approach, which includes the dynamic correlation effects, represents a valid alternative to the HF theory, or to more traditional post-HF methods such as Moller-Plesset theory, coupled-cluster or configuration interaction, which are highly demanding in terms of CPU time. Nowadays DFT is put into practice almost exclusively using the Kohn-Sham equations [4], which are formally similar to the Hartree-Fock (HF) equations. The funda-

* To receive all correspondence Key words: DFT, radical reactions, organometallic compounds, carotenes, singlet-oxygen

mental approximation adopted in the practical applications of DFT is the Local Spin Density (LSD) [5] where a homogeneous electron gas model for the electron density is used. In spite of this crude approximation, the LSD approach is able to provide useful results for a variety of inorganic and organic molecular systems. In particular, results that are in better agreement with the experiment than those obtained from HF computations, have been obtained for transition metal complexes, transition metal clusters, polymers, metal surfaces and interfaces [6]. Other classes of problems, however, are not satisfactorily described at the LSD level [7]. For instance, bond energies are typically overestimated and the geometry of molecules containing hydrogen bonds is not correctly reproduced. One way to correct the errors of the local density approximation has been to introduce density-gradient terms (nonlocal corrections). Since in this way the exchange and correlation contributions are much more accurately described, the nonlocal methods afford much better estimates of bond energies than the simpler local approaches [8 9]. Becke, for instance, has demonstrated that the nonlocal methods provide bond energies of the same quality as those computed with the Poples G2 method [10]. Also, it has been demonstrated that the nonlocal approach can be

128

Quant. Struct.-Act. Relat., 21 (2002)

WILEY-VCH Verlag GmbH, 69469 Weinheim, Germany, 0931-8771/02/0207-0128 $ 17.50+.50/0

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

satisfactorily applied to the study of transition metal complexes where it can improve significantly the description of metal-ligand and metal-metal bonds [11]. The research carried out in this field during the last two decades has lead to the development of a variety of local and nonlocal functionals, which can be used within the Kohn-Sham formalism. These functionals have been implemented in various quantum-mechanical packages [12 16] and are now available for molecular routine computations. The objective of this review is to discuss a number of representative examples where the DFT results, obtained using various functionals and localized basis sets, are compared to experimental data or to the results obtained with traditional correlated methods. This type of calibration, to establish strengths and weaknesses of the various functionals, is nowadays particularly important since DFT methods are more and more replacing the HF approach in the standard description of medium and large-size molecules. In the next section, to illustrate the potentiality of the DFT approach, we shall present the results obtained for some important classes of molecular systems and chemical reactions. So doing, we hope to clearly show that DFT may be routinely used as an efficient tool to explore and analyze organic reactivity, as well as to gain useful information of mechanistic interest in biochemical and biophysical problems. As it will be demonstrated below, when properly combined and complemented with technical skills and computational knowledge, DFT can frequently provide results of impressive accuracy and reliability. In particular, we shall discuss the results of DFT applications (analyzing also the performances of various functionals) in the case of (a) radical reactions, (b) structure of organometallic compounds, (c) transition metal catalyzed reactions, and (d) reactivity of bio-organic molecules polyenic systems.

[19] and E(NON-LOCAL)c to the correlation functional of Lee, Yang and Parr (LYP) [20] which includes both local and non-local terms. The coefficients in expression 1 are those determined by Becke: a1 0.80, a2 0.20, a3 0.72, a4 0.19 and a5 0.81. The other hybrid method (BHLYP) is characterized by the following parameters: a1 0.50, a2 0.50, a3 0.50, a4 0.00 and a5 1.00. The two pure DFT functionals differ only in the correlation term. The functional denoted as BLYP has the following expression: E(S)x E(B88)x E(LYP)c while that denoted as BP86 corresponds to: E(S)x E(B88)x E(P86)c (3) (2)

2 Computational Details
The DFT computations reported in this paper have been carried out using various hybrid and pure functionals as implemented in the Gaussian 92/DFT [14], Gaussian 94 [15] and Gaussian 98 [16] series of programs. Following the Gaussian formalism these functionals are denoted as BHLYP, B3LYP (hybrid functionals) and BLYP, BP86 (pure functionals) and can be written in the general form: a1E(S)x a2E(HF)x a3E(B88)x a4E(LOCAL)c a5E(NON-LOCAL)c

(1)

In expression 1 E(S)x is the Slater exchange [3, 4, 17], E(HF)x the Hartree-Fock exchange, E(B88)x the Beckes 1988 non-local exchange functional corrections [18], E(LOCAL)c is a local correlation functional and E(NONLOCAL)c is the gradient-corrected correlation contribution. In the hybrid Beckes three-parameter exchange functional B3LYP [18] E(LOCAL)c corresponds to the Vosko, Wilk and Nusair (VWN) local correlation functional
Quant. Struct.-Act. Relat., 21 (2002)

where E(P86)c includes the local functional of Perdew along with his gradient corrections [21]. The CAS-SCF computations have been carried out either with the package available in the Gaussian programs [14 16] or using the MOLCAS [22] software. For the CAS-SCF wave-functions the dynamical correlation energy effects have been included using the multi-reference perturbation approach suggested by Andersson et al. (CAS-PT2) [23]. The active space used will be specified in the examples discussed below. The molecular structures and the minimum energy paths (MEP) reported in the following section have been fully optimized using the gradient methods available in the Gaussian packages [14 16]. Frequency computations have been carried out at the HF, MP2, DFT and CAS-SCF levels of theory to determine the nature of the various critical points and to evaluate the energy thermal corrections. In the case of an unrestricted Moller-Plesset approach, projected MP2 and MP4 energies have been used. The use of projected values, when dealing with an unrestricted approach, is discussed below. Localized basis sets have been used in all the computations employing, when possible, the basis sets available in the Gaussian pakages [14 16], as documented below for each reported example. In many examples reported here, HF, MP2 and DFT computations have been carried out using unrestricted wavefunctions. This is the elective procedure for studying open shell species (such as radicals, diradicals, 1O2, triplets, etc.) as the ones investigated in this paper. Unfortunately, the resulting wavefunction is not, in general, an hS2i eigenstate. This means that its energy is spin-contaminated (generally, contamination arises by the first higher multiplicity state, i.e. the T1 state for spin-contaminated singlet energies). Although unrestricted DFT methods in general (and the B3LYP functional in particular) have the capability of reducing the effects of spin contamination (see, for instance, the results discussed in section 3), if such a contamination remains significant, results can be corrected by spin129

F. Bernardi, et al.

projection. For this purpose, we have used the approximate spin-correction procedure proposed by Yamaguchi et al. [24a b], where the singlet spin-corrected (1E(SC)) energy is evaluated by computing the unrestricted DFT singlet (1E(UB)) and triplet (3E(UB)) energies, and applying the formula: Y(UB) cs 1f cT 3f
1

(4) (5) (6)

E(SC) 1E(UB) fSC[1E(UB) 3E(UB)]


1 c2 hS i T 3 2 2 1 cS hS i 1hS2 i 2

fSC

As will be pointed out in the following sections, this procedure has been recently applied by Houk et al. [24c] in the study of the two-step diradical mechanism of the Diels-Alder reaction between butadiene and ethylene. We have chosen this spin-correction procedure because it seems to provide reasonable energies for singlet diradicals and species, which are similar to those presented here. Another very important technical issue in DFT-based investigations concerns wavefunction stability. For the energies to be meaningful, we must check that they represent the correct variational solution of the SCF procedure, i.e. they must be local minima in the wavefunction-coefficients space (with the specified degrees of freedom taken into consideration). Therefore, it is always crucial to check wavefunction stability in order to grant the correctness of the corresponding energy.

3 Radical Reactions
3.1 The Addition of Alkyl and Halogenoalkyl Radicals to the Ethylene Double Bond Several experimental [25a d] and theoretical [25e g] mechanistic studies have been carried out over the last three decades on these radical reactions that are a powerful tool to form new CC bonds and represent the key step in many polymer processes. We report here the results that we have obtained in a computational study on reaction (7)
.

functionals have been used. To verify the validity of a single reference approach in describing the transition state region, the potential energy surfaces for the reactions involving . CH3, .CH2CH3, and .CH2F have been re-investigated at the CASSCF level of theory. The active space is that required to describe correctly the forming of the new CC s bond and the simultaneous breaking of the CC p bond. It consists of three electrons in three orbitals i.e. the p (doubly occupied) and the p* (empty) orbitals associated with the CC olefin bond and the singly occupied ps orbital associated with the non-bonding electron of the alkyl and halogenoalkyl radicals. The 6-31G* basis set [14 16] has been used in all computations. For the various transition states the values of the most relevant geometrical parameters and of the corresponding energies relative to reactants are collected in Table 1. These energies include the zero-point vibrational energy corrections (ZPE) and can be compared to the experimental activation energies collected in the table. The UMP4 energy values are reported in parenthesis and have been obtained from single-point computations on the UMP2 optimized structures. A schematic representation of the transition state structure is given in Figure 1. At all computational levels the structure of the transition state is not very sensitive to the nature of the attacking radical. For the simplest system (R .CH3), at the UHF level, the angle of attack of the approaching radical to the olefin ( < ra angle) is 109.18. As a consequence of the formation of the new CC bond a considerable rehybridization of the carbon atom takes place and a lengthening of the olefin bond, which is losing its double bond character, is observed. The forming new CC bond and the olefin bond are 2.232 and 1.383 respectively. These parameters do no change very much with the increasing size of the attacking alkyl radical. A point of interest in these computations concerns the adequacy of the Hartree-Fock theory in describing the transition states. For .CH3, .CH2CH3 and . CH2F the geometrical parameters obtained with the CASSCF method are almost identical to the corresponding UHF values. This result is in agreement with the form of the CAS-

R H2C CH2 ! products

(7)

where R .CH3, .CH2CH3, .CH(CH3)2, .C(CH3)3 .CH2F, .CF3, .CCl3 These reactions are particularly suitable for a theoretical investigation since they have been experimentally investigated in the gas phase [25a d, g] and, consequently, the obtained activation energies are not affected by the solvent effect and can be directly compared with the theoretical values. To carry out this computational study the unrestricted Hartree-Fock, MP2 and MP4 (UHF, UMP2 and UMP4) methods and the unrestricted DFT approach with two pure (BLYP, BP86) and two hybrid (BHLYP, B3LYP) 130

Figure 1. A schematic representation of the transition structure for the addition of an alkyl radical to the ethylene double bond. The values of the geometrical parameters r, a and < ra are reported in Table 1.

Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

Table 1. Transition state optimized structuresa and energies relative to reactants ( Ea )b computed for the reaction .R C2H4 at various levels of theory. The experimental activation energies ( Ea,exp )are also reported. cThe symbols used for geometrical parameters are those reported in Figure 1. UHF
.

CAS-SCF

UMP2 ( UMP4) 2.261 1.344 109.6 8.9 (9.8) 2.256 1.344 111.0 7.6 (8.7) 2.249 1.343 110.3 5.9 (7.2) 2.249 1.343 111.4 4.6 (6.0) 2.250 1.343 111.5 6.1 (7.0) 2.291 1.338 105.9 1.2 (1.9) 2.238 1.342 107.4 4.1 (5.1)
b

BHLYP 2.310 1.351 109.4 8.2 2.288 1.353 110.9 8.5 2.265 1.355 110.8 8.5 2.250 1.357 111.9 8.7 2.299 1.350 111.5 6.3 2.374 1.343 105.5 1.9 2.237 1.355 108.3 7.4

B3LYP 2.363 1.356 109.8 6.6 2.334 1.358 111.6 7.2 2.305 1.361 111.4 7.5 2.280 1.365 109.2 8.0 2.355 1.355 111.7 4.3 2.442 1.349 105.7 1.0 2.266 1.362 109.2 6.4

BLYP 2.423 1.364 110.3 5.2 2.384 1.367 112.2 5.9 2.344 1.371 111.9 6.6 2.305 1.375 112.9 7.4 2.419 1.363 111.9 3.7 2.535 1.356 105.2 1.0 2.296 1.371 108.8 5.3

BP86 2.473 1.358 110.3 4.1 2.433 1.361 112.5 4.7 2.399 1.364 111.4 4.9 2.361 1.367 112.4 5.3 2.484 1.358 111.6 2.7 2.689 1.349 105.2 1.1 2.362 1.363 109.5 3.6

R CH3 ( Ea, exp. 7.3 kcal mol1) r 2.245 2.245 a 1.382 1.378 < ra 109.1 109.8 11.6 16.3 Ea R .CH2CH3 ( Ea, exp. 6.9 kcal mol1) r 2.232 2.234 a 1.383 1.379 < ra 109.9 110.8 11.5 16.6 Ea R .CH( CH3)2 ( Ea, exp. 6.9 kcal mol1) r 2.210 a 1.385 < ra 110.7 11.8 Ea R .C( CH3)3 ( Ea, exp. 7.1 kcal mol1) r 2.200 a 1.387 < ra 111.6 12.5 Ea R .CH2F ( Ea, exp. 4.3 kcal mol1) r 2.242 2.241 a 1.379 1.376 < ra 110.3 110.9 8.6 13.5 Ea R .CF3 ( Ea, exp. 2.4 kcal mol1) r 2.299 a 1.372 < ra 106.6 4.5 Ea R .CCl3 ( Ea, exp. 6.3 kcal mol1) r 2.199 a 1.383 < ra 108.4 10.1 Ea
a

Bond lengths are in ngstroms and angles in degrees.

Values in kcal mol1. c See Ref. 25g.

SCF wavefunction that is dominated by the SCF configuration. The inclusion of the dynamic correlation at the UMP2 level has the effect of making the r distance longer and the C C double bond shorter (for .CH3, for instance, r and a become 2.261 and 1.344 respectively). Thus the UMP2 predicts earlier transition structures than the UHF method. The UMP2 geometrical parameters compare rather well with those obtained with the various DFT functionals. The most important changes found are a further lengthening of the r distance with a consequent increase of the reactant-like character of the transition state. These geometrical modifications are small when the two hybrid functionals are used, but become more important with the pure BLYP and BP86 functionals. Even if the fundamental structural features of the transition states are satisfactorily described by the UHF method, at this computational level the activation energies Ea are strongly overestimated and the inclusion of the
Quant. Struct.-Act. Relat., 21 (2002)

dynamic correlation effects is essential to obtain reasonable Ea values. This is not surprising, since the accurate prediction of the energy barriers for radical reactions is a difficult problem and it is well known that high levels of theory are needed to reproduce the experimental results [26]. A significant decrease of the activation barriers is observed when projected UMP2 energies are considered. However, even if a general better agreement with the experiment is found, in two cases ( .CH3 and .CH2F) the activation barriers are still overestimated and for .C(CH3)3, .CF and .CCl they are quite underestimated. Also, the 3 3 trend of the computed activation energy that decrease along the series .CH3 ! .CH2CH3 ! .CH(CH3)2 ! .C(CH3)3, is in disagreement with that of the experimental values which remain almost constant. In addition, the trend observed when we compare .C(CH3) (4.6 kcal mol1) and .CH2F (6.1 kcal mol1) is in contrast with the experiment (7.1 and 4.4 kcal mol1 respectively). Furthermore, it is worth to 131

F. Bernardi, et al.

point out that the UMP2 barriers are in better agreement with the experiment than the corresponding UMP4 values. The DFT barriers strongly depend on the functional used in the computations. The BHLYP functional still overestimates the barriers even if the trend along the series .CH3 ! .CH2CH3 ! .CH(CH3)2 ! .C(CH3)3 ! .CH2F, ! .CF3, ! .CCl3 is in much better agreement with the experiment. This trend does not vary when the two pure BLYP and BP86 functionals are used, but in both cases the activation energies become underestimated. The best agreement with the experiment is found at the B3LYP level. With this functional the difference between the experimental and theoretical value is in all cases, except .CF3, less than 1 kcal mol1. 3.2 Hydrogen Abstraction from Halo-substituted Methanes by Methyl Radical We focus now our attention on the hydrogen abstraction reaction from fluoro-, chloro and bromomethanes by methyl radicals (Eq. 8):

CHnXm .CH3 ! .CHn-1Xm CH4

(8)

where m 1, 2, 3, n 3, 2, 1 for X F, Cl and m 1, 2, n 2, 1 for X Br. For these reactions an irregular ordering of the activation energies has been experimentally observed [25d]. The activation energy of fluoromethanes decreases from CH3F to CH2F2, then increases again for CHF3 [25a d, 27]. For chloromethanes and bromomethanes the activation energy decreases regularly along the series CH3Cl ! CH2Cl2 ! CHCl3 and CH3Br ! CH2Br2. In the former case the Ea trend parallels that of the C H bond energies, while in the latter cases this trend becomes opposite [27]. For these reactions a computational investigation [27] on the structure and energy of reactants and transition states has been carried out at the UHF, UMP2 and unrestricted DFT levels with the 6-31G* basis set. The three BHLYP, B3LYP and BLYP functionals have been used. In all cases it has been found that the hydrogen abstraction from halomethanes proceeds in one step. The transition state, which is schematically represented in Figure 2, is characterized by a collinear or nearly collinear

Table 2. Transition state optimized structures a and energies relative to reactants ( Ea ) b computed for the hydrogen abstraction from halomethanes by methyl radicals at various levels of theory. The experimental activation energies ( Ea, exp ) are also reported.c The symbols used for geometrical parameters are those reported in Figure 2. UHF CH3F .CH3 ( Ea, exp 11.8 kcal mol1) a 1.353 b 1.360 28.85 Ea CH2F2 .CH3 ( Ea, exp 10.4 kcal mol1) a 1.356 b 1.349 28.54 Ea CHF3 .CH3 ( Ea, exp 13.6 kcal mol1) a 1.376 b 1.322 30.13 Ea CH3Cl .CH3 ( Ea, exp 9.4 kcal mol1) a 1.354 b 1.349 28.27 Ea CH2Cl2 .CH3 ( Ea, exp 7.2 kcal mol1) a 1.348 b 1.345 25.94 Ea CHCl3 .CH3 ( Ea, exp 5.8 kcal mol1) a 1.343 b 1.343 23.60 Ea CH3Br .CH3 ( Ea, exp 10.1 kcal mol1) a 1.357 b 1.343 28.62 Ea CH2Br2 .CH3 ( Ea, exp 8.7 kcal mol1) a 1.352 b 1.338 26.26 Ea
a

UMP2 1.319 1.348 21.48 1.320 1.344 16.15 1.349 1.312 17.33 1.320 1.339 16.07 1.308 1.345 12.49 1.298 1.349 9.14 1.331 1.325 16.80 1.320 1.328 12.58

BHLYP 1.325 1.358 15.77 1.324 1.354 14.64 1.353 1.320 16.03 1.326 1.348 15.42 1.314 1.355 12.48 1.302 1.361 9.78 1.331 1.339 15.86 1.317 1.348 12.74

B3LYP 1.320 1.388 10.71 1.316 1.390 9.12 1.345 1.351 10.31 1.323 1.374 10.75 1.303 1.393 7.64 1.287 1.407 4.94 1.328 1.365 11.35 1.303 1.387 7.95

BLYP 1.320 1.418 7.53 1.308 1.430 5.65 1.337 1.387 6.67 1.324 1.399 7.92 1.295 1.432 4.72 1.272 1.460 2.03 1.326 1.391 8.64 1.292 1.427 5.02

Values in ngstroms.

Values in kcal mol1. c See Ref. 27.

132

Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

difference between the computed and experimental value is in all cases within 1.5 kcal mol1. As mentioned in the previous section, a further point of interest is represented by the capability of the DFT (B3LYP) method of reducing the effects of spin contamination. This is evident from the comparison of the hS2i values obtained at the UHF, UMP2 and unrestricted B3LYP levels and collected in Table 3.
Figure 2. A schematic representation of the transition structure for the hydrogen abstraction from halomethanes by methyl radical. The values of the geometrical parameters a and b are reported in Table 2.

3.3 Hydrogen and Chlorine Abstraction from Chloromethanes by Silyl Radicals. While alkyl radicals react with haloakanes mainly via hydrogen abstraction, many heteroatom-centered radicals preferentially abstract a halogen atom from organic halides (see Eq. 9). R 'n M . X R ! RnM-X R . M B, Si, Ge, P, transition metal; X F, Cl, Br, I

arrangement of the three atoms involved in the process. The values of the breaking and forming bonds (parameters a and b) are reported in Table 2. For fluoromethanes, at the UHF level, the transition state slightly advances toward the product when the number of fluorine atoms in the substrate increases. At the UMP2 level, with the inclusion of the dynamic correlation, the transition state becomes more reactant-like. A similar effect has been observed at the DFT level with all three functionals. Similar results have been obtained for chloro and bromomethanes: in both cases the inclusion of correlation determines an increase of the reactant-like character of the transition structure. The experimental (Ea, exp) and computed (Ea) activation energies at various levels of theory are reported in Table 2 (the activation energies have been evaluated as the difference between the energies of the transition states and those of reactants and include the ZPE corrections). The UHF activation barriers are in all cases overestimated. This result is similar to that discussed in the previous section for the addition to the olefin bond. The projected UMP2 values are significantly lower than the corresponding UHF data, but they are still quite overestimated (the error with respect to the experiment is in all cases larger than 55%). Once again at the DFT level the energy barriers depend significantly on the type of functional. Even if with the hybrid BHLYP functional the barriers are still overestimated, their trend from CHF3 (16.04 kcal mol1) to CH3F (15.77 kcal mol1) is now the same as that experimentally found. On the other hand the pure BLYP functional provides Ea values which are quite underestimated. The best agreement with the experiment is obtained at the B3LYP level of theory, where the
Table 3. Values of hS2i computed at the UHF, UMP2 and unrestricted DFT ( B3LYP ) levels with the 6-31G* basis set. UHF CH3F CH3 CH2F2 .CH3 CHF3 .CH3 CH3Cl .CH3 CH2Cl2 .CH3 CHCl3 .CH3 CH3Br .CH3 CH2Br2 .CH3
.

(9)

In particular, the halogen abstraction carried out by silyl radicals has been widely investigated and a large amount of experimental data on the reactivity of silicon-centered radicals toward various organic halides is now available [28, 29]. For instance Cadman et al. studied the relative rates of chlorine abstraction from alkyl chlorides by the trimethylsilyl radical. These authors determined a barrier of 4.06 kcal mol1 for the chlorine abstraction from H3CCl [29]. Aloni et al. found that the activation barrier for the chlorine abstraction by trichlorosilyl radicals from chloromethane decreases when the number of halogen atoms increases [29]. In this section we present the results of a theoretical study [29] of the hydrogen and chlorine abstraction by the silyl H3Si . and trichlorosilyl Cl3Si . radicals from ClCH3, Cl2CH2 and Cl3 CH. The computations have been carried out using the UHF method, the Moller-Plesset perturbation theory MP2 up to

UMP2 0.7629 0.7627 0.7624 0.7629 0.7627 0.7621 0.7634 0.7636

B3LYP 0.7571 0.7570 0.7569 0.7571 0.7669 0.7566 0.7572 0.7570

0.7890 0.7885 0.7882 0.7888 0.7879 0.7861 0.7896 0.7891

Figure 3. A schematic representation of the transition structure for the hydrogen and chlorine abstraction from chloromethanes by silyl and trichlorosilyl radicals. The values of the geometrical parameters a and b are reported in Tables 4 and 5.

Quant. Struct.-Act. Relat., 21 (2002)

133

F. Bernardi, et al.

second order (UMP2) and the unrestricted DFT method with the two functionals B3LYP and BLYP. A schematic representation of the transition structures corresponding to the hydrogen (TS1) and chlorine (TS2) abstraction from the chloromethane molecule is given in Figure 3. The values of the forming (a) and breaking (b) bonds computed at different levels of theory are collected in Tables 4 for H3Si . and Table 5 for Cl3Si .. In these tables the values of the activation energies (Ea) are also reported. These activation energies have been computed using the following equation: Ea DH nRT (10)

Table 4. Optimum values of the most relevant geometrical parametersa of the transition states TS1 and TS2 with the corresponding activation energies ( Ea )b for the reaction between silyl radical and chloromethanes at various levels of theory . The symbols used for the geometrical parameters are those reported in Figure 3. UHF
.

UMP2

B3LYP

BLYP

where R is the gas constant, T the absolute temperature, n represents the molecularity of the reaction (2 for the case investigated here) and DH is the activation enthalpy. The molecular enthalpy is computed as H E Hth, where E is the quantomechanical energy and the thermal corrections to enthalpy (Hth) is given by: Hth ZPE Evib Erot Etr RT (11) In Eq. 11 Evib, Erot and Etr are the vibrational, rotational and translational contributions to the energy, respectively. For the three substrate molecules considered here, both hydrogen and chlorine abstractions proceed in one step and the three atoms involved in the process are characterized by a collinear arrangement. Furthermore, while in the hydrogen abstraction the two fragments H3Si . and Cl3Si . approach the hydrogen atom of the substrate in a staggered conformation, in the chlorine abstraction they are arranged in an eclipsed conformation. To discuss the structure of the transition states the quantity q is introduced. This parameter, reported in the tables, conveniently describes the nature of the various transition structures at different computational levels. For the H abstraction q dR(Si-H)/ dR(C-H) where dR(Si-H) a/R(Si-H)eq and dR(C-H) b/R(C-H)eq. Here a and b are the lengths of the forming Si H and breaking C H bonds, respectively and R(Si-H)eq and R(C-H)eq are the corresponding equilibrium distances in the product (silane) and reactant (choromethane). In a similar way, for the Cl abstraction, d is defined as q dR(Si- Cl)/dR(C-Cl) where dR(Si-Cl) a/R(Si-Cl)eq and dR(C-Cl) b/R(C-Cl)eq. In this case a and b are the lengths of the forming Si-Cl and breaking CCl bonds, respectively, and R(Si-Cl)eq and R(C-Cl)eq are the corresponding equilibrium distances in the product (chlorosilane) and reactant (choromethane). A value of 1 for q indicates a transition state where the two bonds are broken and formed to the same extent; a value lower or greater than 1 corresponds to a more product-like or to a more reactant-like transition state respectively. Inspection of Tables 4 and 5 points out the product-like character of TS1 (q< 1) and the reactant-like character of TS2 (q> 1). The trend of the computed q values indicates that TS1 becomes less product-like while TS2 becomes more 134

(a) H3Si ClCH3 Hydrogen Abstraction ( TS1) a 1.711 1.644 b 1.445 1.512 q 0.865 0.798 28.33 23.78 Ea Chlorine Abstraction ( TS2) a 2.568 2.435 b 2.056 2.001 q 1.078 1.051 22.10 15.04 Ea (b) H3Si . Cl2CH2 Hydrogen Abstraction ( TS1) a 1.715 1.648 b 1.431 1.493 q 0.872 0.809 26.37 19.90 Ea Chlorine Abstraction ( TS2) a 2.609 2.472 b 2.020 1.975 q 1.105 1.075 20.28 15.01 Ea (c) H3Si . Cl3CH Hydrogen Abstraction ( TS1) a 1.717 1.646 b 1.416 1.480 q 0.881 0.814 24.08 16.30 Ea Chlorine Abstraction ( TS2) a 2.654 2.514 b 1.991 1.950 q 1.137 1.105 17.80 9.88 Ea
a

1.642 1.552 0.776 17.43 2.541 2.010 1.097 8.32

1.639 1.600 0.752 15.69 2.606 2.082 1.091 5.58

1.661 1.506 0.807 14.01 2.633 1.975 1.149 6.38

1.663 1.539 0.791 11.96 2.740 1.971 1.203 3.80

1.671 1.476 0.827 11.04 2.741 1.935 1.218 3.73

1.676 1.500 0.816 8.84 2.909 1.925 1.305 1.55

Values in ngstroms.

Values in kcal mol1.

reactant-like when more chlorine atoms are introduced in the substrate. The inclusion of the dynamic correlation at the UMP2 level has the effect of making TS1 more product-like (q is smaller than 1 and further decreases) and TS2 less reactant-like (q is larger than 1 and decreases). A similar effect is observed for TS1 at the B3LYP and BLYP levels, where again q decreases with respect to the UHF and UMP2 values. An opposite trend is observed for TS2: in this case q significantly increases when the DFT is used. Since the activation barriers for the chlorine abstraction by the silyl radical H3Si . from chloromethanes are not experimentally available, to roughly estimate the accuracy of the various theoretical methods, the computed activation energies reported in Table 4 can be compared with the experimental values determined for the chlorine abstraction by the triethylsilyl radical Et3Si . from chloromethanes. These barriers are 4.06, 2.06 and 1.14 kcal mol1 for ClCH3, Cl2CH2 and Cl3CH, respectively [29]. From Table 4 it is evident that the corresponding UHF values (22.10, 20.28
Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

Table 5. Optimum values of the most relevant geometrical parameters a of the transition states TS1 and TS2 with the corresponding activation energies ( Ea ) b for the reaction between trichlorosilyl radical and chloromethanes at various levels of theory. The symbols used for the geometrical parameters are those reported in Figure 3. UHF
.

UMP2

B3LYP

BLYP

(a) Cl3Si ClCH3 Hydrogen Abstraction ( TS1) a 1.699 1.633 b 1.420 1.491 q 0.887 0.811 26.48 20.20 Ea Chlorine Abstraction ( TS2) a 2.475 2.376 b 2.063 1.993 q 1.055 1.046 20.78 10.78 Ea (b) Cl3Si . Cl2CH2 Hydrogen Abstraction ( TS1) a 1.709 1.642 b 1.402 1.464 q 0.901 0.830 25.79 16.74 Ea Chlorine Abstraction ( TS2) a 2.522 2.419 b 2.034 1.969 q 1.080 1.072 20.16 9.36 Ea (c) Cl3Si . Cl3CH Hydrogen Abstraction ( TS1) a 1.709 1.642 b 1.402 1.464 q 0.901 0.830 24.97 13.83 Ea Chlorine Abstraction ( TS2) a 2.522 2.419 b 2.034 1.969 q 1.080 1.072 18.55 7.38 Ea
a

1.615 1.569 0.763 16.99 2.438 2.027 1.060 7.33

1.598 1.664 0.712 16.44 2.476 2.047 1.068 4.98

1.642 1.510 0.804 14.36 2.536 1.989 1.116 6.18

1.629 1.581 0.762 13.22 2.616 1.995 1.150 3.80

1.642 1.510 0.804 12.10 2.536 1.989 1.116 4.81

1.629 1.581 0.762 10.76 2.616 1.995 1.15 2.72

MP2 methods, the chlorine abstraction is highly favored with respect to the hydrogen abstraction. However, it is difficult in the present case to establish what functional performs better, since in these computations the effects of the ethyl groups bonded to silicon are neglected. The trend of the energy barriers computed for the Cl3Si . radical (see Table 5) is the same as that found for H3Si . : the chlorine abstraction is favored and the activation energies decrease with the increasing number of chlorine atoms in the substrate. An experimental estimate of the barrier for the chlorine abstraction, which can be used as a reference, provides 6.18, 5.60 and 4.40 kcal mol1 for ClCH3, Cl2CH2 and Cl3CH, respectively [29]. Once again the barriers are largely overestimated not only at the UHF level (the error is higher than 200%), but also at the UMP2 level. These values greatly improve using the DFT approach. The B3LYP computed activation energies are 7.33, 6.18 and 4.81 kcal mol1 for ClCH3, Cl2CH2 and Cl3CH respectively and become 4.98, 3.80 and 2.72 kcal mol1 at the BLYP level. Thus the best agreement with the experiment is found at the B3LYP level where the error ranges between 9% and 18%. As stressed in the previous section, in the case of the hydrogen abstractions from halomethanes by methyl radical, it is worth discussing again the ability of DFT-based methods to reduce the effect of spin contamination on the wave function. This decrease of spin contamination, which should provide more reliable structures and energies, is evident from the hS2i values reported in Table 6 and computed at the UHF, UMP2, B3LYP and BLYP levels for both TS1 and TS2.

4 Structure of Organometallic Compounds and Modelling Homogeneous Catalysis


4.1 Nickel-Ethylene Complexes We describe in this section the results obtained in the investigation of the singlet potential energy surface for the bis(ethylene)-Ni complex. These clusters provide very simple models for understanding the nature of the metalolefin bond and for investigating the mechanism of catalyzed processes [30, 31]. Complexes of this type, in fact, seem to be involved in homogeneously catalyzed [2 2] cycloaddition reactions. The potential surface has been described at the CAS-SCF/CAS-PT2 and DFT levels with the B3LYP, BLYP and BP86 functionals. The CAS-SCF computations have been carried out using the atomic natural orbital (ANO) basis suggested by Bauschlicher et al. for the nickel atom and the Dunning cc-pVDZ basis for the carbon and hydrogen atoms (more details on these basis sets can be found in Ref. 30). The active space used to build the CAS wave-function includes the p and p* orbitals of the two ethylene moieties and the 4s and 3d orbitals of the nickel atom. The size of the active space has been increased in the single-point CAS-PT2 computations on the CAS optimized structures. In this case for each doubly occupied 3d orbital 135

ngstroms. Values in a

Values in kcal mol1.

and 17.80 kcal mol1) are much larger. However, it is worth to point out that these barriers, even if overestimated, are in all cases smaller than the corresponding barriers required by the H abstraction (28.33, 26.37 and 24.08 kcal mol1). This trend agrees with the experimental observation that silyl and chloro silyl radicals react via halogen abstraction and not hydrogen abstraction. As previously observed for the other radical reactions, when the projected UMP2 approach is used, these barriers significantly decrease, even if they remain quite overestimated. Once again the DFT approach provides much better results. The two sets of energy barriers obtained at this level are both quite close to the experimental values used as a reference (Et3Si . radical). The values obtained with the B3LYP functional are 8.32, 6.38 and 3.73 kcal mol1, while at the BLYP level these values become 5.58, 3.80 and 1.55 kcal mol1 respectively. Also, as found with the HF and
Quant. Struct.-Act. Relat., 21 (2002)

F. Bernardi, et al.

Table 6. Values of hS2i computed at the UHF, UMP2 and unrestricted DFT ( BLYP and B3LYP ) levels with the 6-31G* basis set. UHF H3Si . ClCH3 H3Si . Cl2CH2 H3Si . Cl3CH Cl3Si . ClCH3 Cl3Si . Cl2CH2 Cl3Si . Cl3CH H3Si . ClCH3 H3Si . Cl2CH2 H3Si . Cl3CH Cl3Si . ClCH3 Cl3Si . Cl2CH2 Cl3Si . Cl3CH UMP2 B3LYP BLYP 0.7535 0.7532 0.7529 0.7533 0.7530 0.7527 0.7549 0.7538 0.7529 0.7547 0.7530 0.7529

Hydrogen Abstraction ( TS1) 0.7885 0.7624 0.7562 0.7872 0.7622 0.7559 0.7852 0.7821 0.7554 0.7883 0.7624 0.7561 0.7880 0.7626 0.7558 0.7869 0.7623 0.7555 Chlorine Abstraction ( TS2) 0.8543 0.7900 0.7606 0.8504 0.7895 0.7592 0.8442 0.7885 0.7572 0.8507 0.7861 0.7605 0.8525 0.7880 0.7558 0.8523 0.7883 0.7576

Table 7. Optimized geometrical parameters a and relative energies ( E) b for the D2d, C2v(planar), C2v(bent) and D2h bis(ethylene)nickel complexes computed at various levels of theory. The energy values reported in parenthesis have been obtained at the CASPT2 level. The symbols used for the geometrical parameters are those reported in Figure 4. CAS( CAS-PT2) a b c e e' E a b c e e' E a b c e e' E a b c e e' E
a

B3LYP D2d 1.396 1.990 1.990 15.5 15.5 0.00 C2v(planar) 1.396 2.009 1.976 15.5 17.8 5.04 C2v(bent) 1.381 2.047 2.047 11.7 11.7 12.80 D2h 1.379 2.050 2.050 11.6 11.6 12.84

BLYP 1.408 2.013 2.013 15.2 15.2 0.00 1.409 2.029 1.998 15.4 17.9 5.64 1.398 2.053 2.053 13.1 13.1 13.24 1.392 2.068 2.068 11.3 11.3 14.53
b

BP86 1.406 1.986 1.986 15.2 15.2 0.00 1.408 1.996 1.972 16.2 18.3 5.03 1.397 2.022 2.022 13.4 13.4 14.09 1.390 2.038 2.038 11.5 11.5 15.94
Values in kcal

1.399 2.021 2.021 18.8 18.8 0.00 (0.00) 1.397 2.022 2.015 18.5 20.2 6.89 (5.07) 1.397 2.055 2.055 18.7 18.7 18.05 (17.31) 1.373 2.121 2.121 11.6 11.6 36.56 (18.96)

not taking part in the bond, a correlating 4d orbital has been added. For the DFT computations a local spin density (LSD)-optimized basis set of double-x quality in the valence shell plus polarization functions (DZVP) has been used [30]. Four critical points, two with C2v symmetry (bent and planar) and two with D2h and D2d symmetry, respectively, have been located. The structures of these complexes are schematically represented in Figure 4, while the values of the most important geometrical parameters and the corresponding energies are reported in Table 7. The D2d structure, where the two planes containing the CC bonds of the two ethylenes and the metal atom form a dihedral angle of 908, has the lowest CASSCF energy. The binding of the ethylene to the nickel atom causes a

Bond lengths are in ngstroms and angles in degrees. mol1.

Figure 4. A schematic representation of the various structures located for the Ni(C2H4)2 complex. The values of the geometrical parameters a, b, c, e and e' are reported in Tables 7.

significant lengthening of the CC bond and a nonnegligible rehybridization of the carbon atoms. The CC bond becomes 1.399 (1.332 in the free ethylene) and the two methylene hydrogen atoms are bent 18.88 out of the ethylene molecular plane (see the out of plane angle e formed by the bisector of the HCH angle and the CC direction schematically represented in Figure 4b). The C2v (planar) structure is only 6.89 kcal mol1 above the D2d species, while the C2v(bent) form is 11.16 kcal mol1 higher than the C2v(planar) complex. The highest in energy species is the D2h structure, which lies 36.56 kcal mol1 above D2d. The inclusion of the dynamic correlation energy at the CASPT2 level does not change the energetic order of the four structures, but affects the various energy gaps. While the energy differences D2d C2v(planar) and D2d C2v(bent) only slightly decrease (they are now 5.07 and 17.31 kcal mol1 respectively), the difference D2d-D2h strongly decreases (it becomes 18.96 kcal mol1). The geometrical structures obtained at the DFT level with the three functionals are almost identical to those found at the CAS-SCF level, suggesting that, for these systems, the
Quant. Struct.-Act. Relat., 21 (2002)

136

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

inclusion of the dynamic correlation does not affect critically the geometry. A significant variation is observed only in the C2v(bent) structure where the f angle (see Figure 4c) significantly increases (159.28, 158.08 and 170.18 at the BLYP, BP86 and B3LYP levels respectively). The most interesting result is that at the DFT level the energetic order of the four critical points is the same as that found at the CAS-PT2 level and the energy differences do not change significantly. The DFT results also indicate that the BP86 functional provides the best agreement with the CAS-PT2 data. Full Hessian matrix computations were performed at the DFT (BP86) level of theory to characterize the nature of the various critical points. These computations pointed out that the D2d structure is the only real minimum of the surface (all real frequencies) and that the other critical points are saddle points of index 1 (C2v(planar), one imaginary frequency), index 2 (C2v(bent), two imaginary frequencies) and index 3 (D2h, three imaginary frequencies). 4.2 Ethylene Dimerization Catalyzed by Ni(0) Complexes The [2 2] cycloadditions represent an effective tool for the synthesis of four-membered rings. Since these reactions are characterized by large activation energies, they must be carried out at high temperature (400 700 8C). However, in the presence of transition metal complexes used as catalysts, these reactions proceed much faster and under milder conditions [31, 32]. A vast amount of experimental work is now available in the literature [32], where it is demonstrated that Rh(I), Pd(II), Pt(II), Ni(0) complexes are effective catalysts for [2 2] cycloaddition reactions. These reactions have also much theoretical interest since transition metals apparently remove the symmetry constraints that make them thermally forbidden in a concerted approach, according to the Woodward-Hoffman symmetry rules. To explain this evidence two different mechanistic schemes are usually proposed. The first hypothesis is a concerted mechanism where the two new CC bonds are formed simultaneously. The role played by the metal is that of providing suitable d orbitals, that, after combination with the olefin p orbitals, make the reaction symmetry-allowed. The second hypothesis corresponds to a non-concerted mechanism, which involves the formation of 1 : 1, and 1 : 2 metal-olefin complexes followed by the formation of a metal-carbon s bonded intermediate (metallacyclopentane). The metallacyclopentane intermediate can lead to the cyclopropane product by reductive elimination. We report the results of a DFT computational study [31] on a model-system which emulates a [2 2] cycloaddition catalyzed by Ni(0) complexes. The model discussed here is formed by a Ni(PH3)2 fragment that can bond either one or two ethylene molecules leading to the Ni(PH3)2C2H4 or Ni(PH3)2(C2H4)2 complexes (1 : 1 and 1 : 2 metal-olefin complexes) that are assumed to represent two possible active forms of the catalysts. For both complexes the reaction with an additional ethylene molecule has been
Quant. Struct.-Act. Relat., 21 (2002)

investigated using the B3LYP functional in the unrestricted form. It has been proven, in fact, that this approach can provide a reliable description of both the concerted and diradical pathways in the case of [4 2] and [2 2] cycloadditions [31, 24c]. All the computations have been carried out with the pseudopotential LANL2DZ [14 16] basis which has been demonstrated to be capable of satisfactorily describing the ethylene and bis(ethylene)-nickel complexes [30]. The following steps of the catalyzed reaction are discussed: (i) The formation of the ethylene and bis(ethylene)-nickel complexes Ni(PH3)2C2H4 or Ni(PH3)2(C2H4)2 (active forms of the catalyst); (ii) the attack of a free ethylene on the active catalysts (formation of biradical intermediates); (iii) the intramolecular coupling of the diradical leading to the nickelacyclopentane. The two complexes arising from the interaction of the Ni(PH3)2 fragment with one (M1) or two (M2) ethylene molecules are shown in Figure 5, with the values of the most important geometrical parameters and those of the relative energies. M1 is a planar tricoordinated complex, while M2 is a tetracoordinated complex, 2.79 kcal mol1 lower than M1.

Figure 5. A schematic representation of the ethylene-nickel (M1) and bis(ethylene)-nickel (M2) complexes and of the anti biradical transition state TS1. The energies (kcal mol1) are relative to M2 a non-interacting ethylene molecule (asymptotic limit). Bond lenghts are in ngstroms and angles in degrees.

137

F. Bernardi, et al.

Figure 6. A schematic representation of the anti biradical intermediate M3, of the syn biradical intermediate M4, and of the nickelacyclopentane M5. The energies (kcal mol1) are relative to M2 a non-interacting ethylene molecule (asymptotic limit). Bond lenghts are in ngstroms and angles in degrees.

Figure 7. A schematic representation of the anti biradical transition state TS2 and of the anti biradical intermediate M6, associated with the attack of one ethylene molecule on the M1 complex. The energies (kcal mol1) are relative to M1 a noninteracting ethylene molecule. Bond lenghts are in ngstroms and angles in degrees.

The most remarkable feature found in the investigation of the reaction surface, is that an additional ethylene molecule attacks the complex M2 not at the metal center, but at one carbon of the two ethylene ligands. This attack, which involves the transition state TS1 (energy barrier of 35.80 kcal mol1), leads to the formation of the intermediate M3, 24.26 kcal mol1 higher in energy than the asymptotic limit (M2 a non-interacting ethylene molecule). M3 is represented in Figure 6. These structures (TS1 and M3) are both characterized by an anti orientation of the attacking ethylene with respect to the ethylene ligand and are similar to those determined for the non-catalyzed reaction [31]. The new forming CC bond is 1.900 in TS1 and becomes almost completed (1.584 ) in the intermediate M3. The most interesting aspect, which characterizes the electronic structure of TS1 and M3, is that one of the two unpaired electrons is mainly localized on the terminal methylene and the other on the nickel atom. A further investigation of the reaction surface has shown that a rotation around the new CC bond leads from the anti 138

intermediate to a syn intermediate M4, which is 6.91 kcal mol1 higher in energy than M3. In the new structural arrangement the two unpaired electrons easily couple to form, with a negligible barrier, the nickelacyclopentane M5. This complex is 34.8 kcal mol1 lower in energy than M4 and 3.01 kcal mol1 under the asymptotic limit. The computations have demonstrated that the ring closure to form the new Ni-C bond in the metallacyclopentane leads to the elimination of the ethylene ligand not involved in the reaction. M4 and M5 are represented in Figure 6. An anti attack of one ethylene molecule has also been considered on the M1 complex, the other possible active form of the catalyst and it has been found that an anti transition state and an anti intermediate (TS2 and M6 represented in Figure 7) exist. However in this case the activation energy required to form M6 from M1 is higher (39.79 kcal mol1) than that found for M2 and, more interesting, almost identical to the value of 40.34 kcal mol1 found, at the same computational level, for the noncatalyzed reaction [31]. These results suggest that a catalytic
Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

effect, even if not very large for the simple model-system discussed here, exists only for the 1 : 2 and not for the 1 : 1 nickel-ethylene complexes. The 1 : 1 complexes more easily coordinate an additional ethylene at the metal center, as previously seen. These DFT studies enforce the hypothesis, based on experimental observation, that olefin dimerization proceeds through a non-concerted mechanism involving an equilibrium between bis(olefin)-metal complexes and metallacyclopentanes. 4.3 Homogeneous Ziegler-Natta catalysis The Ziegler-Natta polymerization of olefins is an important process used in the industry to obtain long polymeric chains. The process is extremely fast and proceeds with high stereoselectivity. According to the most commonly accepted mechanism (see Scheme 1) [33], the active form of the catalyst is characterized by a vacant site in the metal coordination sphere which binds an olefin to form a metalalkyl-olefin complex. In a subsequent step the olefin

Scheme 1.

Scheme 2.

molecule inserts into the metal-alkyl bond leading to a new alkyl complex characterized by a longer chain (growing chain). The resulting complex has a new vacant site on the metal that can bind another olefin molecule. Even if a great deal of experimental [34] and theoretical [33] work has been carried out on this reaction, many mechanistic details are still obscure. In particular the nature of the active form of the catalyst and of the metal-alkyl-olefin complexes needs to be elucidated. If we consider, for instance, the commonly used two-component Ziegler-Natta catalyst TiCl4-AlR3, the real active form originated by the catalyst-cocatalyst interaction could correspond either to a bimetallic complex or to a solvent-separated ion-pair as shown in Scheme 2. In a recent paper [33] a DFT(B3LYP) investigation has been carried out on both mechanistic hypothesis i.e. the bimetallic complex and the solvent separated ion-pair. We discuss here in details only the results obtained for the latter hypothesis, since in this case it is possible to compare the DFT data with those obtained at the MP2 and CAS-PT2 levels of theory. A model-system formed by the Cl2TiCH 3 cationic species interacting with one ethylene molecule has been used to emulate the positive fragment of the solventseparated ion-pair. The computations have been carried out with two different basis sets. The simpler one is the MIDI4 basis of Huzinaga augmented by two sets of p functions on the titanium atom (exponents 0.083 and 0.028). The more accurate basis is formed by the 6-31G* basis for carbon, aluminium, chlorine and hydrogen and by the WatchersHay basis for titanium (a (14s, 11p, 6d) primitive set contracted to [8s, 6p, 4d]). More details on these basis sets are reported in Ref. 34. The structures of the critical points located on the potential energy surface are represented in Figure 8. The values of the most relevant geometrical parameters and the energy values relative to reactants (Cl2TiCH 3 a non interacting ethylene molecule) are reported in the figure. A p complex m1 between the Cl2TiCH 3 moiety and the ethylene molecule forms without any barrier. This complex is much more stable than reactants: 34.80 and 37.88 kcal mol1 at the MIDI4 and 6-31G* levels respectively. TS1 is a four-centered structure corresponding to the transition state for the ethylene insertion: this requires the overcoming of a barrier of 7.67 kcal mol1 at the MIDI4 level (5.65 kcal mol1 with the 6-31G* basis). The transition state leads to the insertion product m2, which is a propyl complex characterized by an approximately planar four-centered structure. The insertion product m2 is significantly more stable than reactants: 40.35 and 45.55 kcal mol1 are the exothermicity values obtained with the MIDI4 and the 6-31G* basis, respectively A comparison of the DFT results obtained at the two levels of accuracy (MIDI4 and 6-31G*) indicates that both the geometrical parameters and the energy values are not very sensitive to the basis set. For this reason, to validate the DFT approach, the reaction has been re-investigated at the MP2 and CAS-SCF/CAS-PT2 levels with the MIDI4 basis set. The structures have been fully re-optimized at the MP2 139

Quant. Struct.-Act. Relat., 21 (2002)

F. Bernardi, et al.

the CAS-SCFand DFT values are larger than those found in the comparison between MP2 and DFT. This is probably due to the fact that at the CAS-SCF level the largest part of the dynamic correlation is neglected. For this reason the CASSCF energy values are not expected to be reliable: the exothermicity of the reaction decreases notably ( 27.30 kcal mol1), while the insertion barrier becomes much larger (14.54 kcal mol1). It is interesting to note that, when the dynamic correlation is included by means of single-point CAS-PT2 computations on the CAS-SCF structures, the energy values become very similar to those determined at the MP2 and DFT levels. The CAS-PT2 provides, in fact, an exothermicity of 37.25 kcal mol1 and a barrier of 6.43 kcal mol1. These results indicate that the DFT(B3LYP) is adequate to describe this class of reactions and confirm the importance of the dynamical correlation to obtain accurate structures and energies.

5 Polyenes, Carotenoids and Singlet-Oxygen Quenching


Long-chain polyenes and the study of their chemical reactivity and properties provide an illustrative example of a DFT-based investigation in organic chemistry. As it will be shown below, this case-study displays quite well both the extended applicability of DFT methodologies, as well as their limitations. Moreover, in particular conditions, it turns out that situations which are in general not described by standard DFT methods (e.g. excited states, etc.) may also be investigated, thus gaining significant information for topical bio-chemical and bio-physical problems (which, otherwise, could be hardly inferred). This is the result of a deep knowledge and proper control of the computational tools and algorithms, which may be suitably tuned to get nonstandard DFT results. Furthermore, these examples clearly illustrate how limited and dangerous could be the use of a DFT-based approach if not supported by a proper knowledge of quantum-chemistry and computational methods. Although nowadays DFT is gaining more and more popularity due to its geometric/energetic accuracy and applicability on large molecular systems, nevertheless its use as a black box deserves attention and care. Still, when complemented by proper technical skills, we can frequently treat tricky problems, getting results with impressive reliability. 5.1 Biological Activity of Carotenes Radical scavenging (i.e. antioxidant ability) and singletoxygen (1O2) quenching activity are one of the most important biochemical properties of carotenoid systems. In fact, carotenoids protect vital biological structures against free-radicals and 1O2 (a highly reactive and toxic form of oxygen) degradation, both in bound (e.g. in photosynthetic centers) and unbound (e.g. in biological tissues) conditions [35 44]. Though these properties may act in
Quant. Struct.-Act. Relat., 21 (2002)

Figure 8. A schematic representation of the structures of reactants, intermediate m1, transition state TS1 and product m2 for the insertion of one ethylene molecule into the Ti-C bond in the (Cl2TiCH3) species. The values of the reported geometrical parameters have been obtained at the B3LYP/MIDI4, (B3LYP/631G*), [MP2/MIDI4] and CAS-SCF/MIDI4 computational levels. The corresponding energies (kcal mol1) are relative to reactants i.e. the (Cl2TiCH3) species a non interacting ethylene molecule (see Ref. 34 for further details). Bond lenghts are in ngstroms and angles in degrees.

and CAS-SCF levels and CAS-PT2 single-point computations have been carried out on the CAS-SCF optimized structures. The CAS-SCF active space includes the ethylene p and p* orbitals, the s and s* orbitals associated with the Ti-C bond and the empty 3dz2, 3dxy and 3dyz orbitals on the metal atom. The results obtained with the MP2 approach are very similar to those provided by DFTwith the same basis set. At this level of theory the reaction is exothermic by 39.44 kcal mol1 and the insertion barrier is 8.73 kcal mol1. At the CAS-SCF level the topology of the surface is identical to that already described at the DFT and MP2 levels. This is in agreement with the fact that CAS-SCF wavefunction is mostly dominated by the SCF configuration (0.95 is the corresponding weight). However, even if the number and the nature of the critical points do not change, the differences observed for several geometrical parameters between 140

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

concert leading to a common protective effect, still they depend on very different processes. Thus, for example, carotenoids may quench 1O2 catalytically (i.e. carotenoids are regenerated) via a very efficient (almost diffusioncontrolled) physical pathway, i.e. an energy transfer (ET) process, followed by an intersystem crossing (ISC):
1

O2 1carotenoid ! 3O2 3carotenoid ! 3O2 1carotenoid ( heat)

(12)

noids could depend by some specific and reactive form of these systems. Finally, other competing chemical reactions could be responsible for alternative catalytic 1O2 quenching pathways. These processes, even if less efficient than energy transfer (Eq. 12), could be competitive with oxidation reactions. The biological importance of these processes represent the background and the motivation for a full mechanistic investigation. Obviously, the first task in this quest is to define a reliable model and to select a proper computational method. 5.1.1 The Diradical Intermediate Model: trans ! cis Thermal Isomerization Barriers in long linear Polyenes. Towards a Carotenoid Model. It has been suggested that the radical trapping action of bcarotene (and carotenes in general) might derive from a reactive (diradical-type) twisted intermediate, half the way along the path for thermal trans ! cis isomerization about the central double bond [47 50]. The existence of such a diradical intermediate might depend on a special resonancetype stabilization effect at the twisted region, which makes the system stable (i.e. long living) enough to trap other radicals. For this to happen, the diradical twisted structure must be an energy minimum (i.e. an intermediate) on the potential energy surface (PES), and the energy barrier for trans ! cis thermal isomerization must be small enough (< 30 kcal mol1) to be active at physiological conditions (37 8C), see Scheme 3. This could grant a small but constant amount of reactive (i.e. diradical-type) twisted species in the

In contrast to physical quenching, carotenoids may also chemically react with 1O2 (and radicals in general). Radical scavenging involves, in fact, real chemical reactions (chemical pathway) [45] leading to carotenoid oxidation. These paths result in the generation of stable radical and diradical species, in the destruction of carotenoids, and thus in the loss of antioxidant protection (i.e. these are not catalytic processes) [46]:
1

O2/radicals carotenoid ! chemical pathway

(13)

The specific reactions involved in the chemical pathway (Eq. 13) are still not completely understood. It is not clear whether the observed oxidation products (which include apocarotenal chain cleavage fragments) [46] are formed by direct addition of 1O2 to the carotenoid system or by reaction between the triplet-oxygen (3O2) and the triplet carotenoid (3carotenoid) produced by the main energy transfer process (Eq. 12). Moreover, the radical trapping ability of carote-

Scheme 3.

Quant. Struct.-Act. Relat., 21 (2002)

141

F. Bernardi, et al.

Figure 9. Typical carotenoid structures (n is the number of conjugated CC double bonds). The available experimental second-order rate constants (1010 Kq M1s1) for the quenching of singlet-oxygen are given in parentheses [44a]. The common 9 double bonds moiety is highlighted in bold.

environment, which immediately scavenges incoming radicals. Otherwise, if no twisted minimum exists on the PES and/or there is a too high-energy barrier, such a trapping effect would be much smaller, if not completely absent. To elucidate this hypothesis, trans ! cis thermal isomerization about the most central double bond for a series of eight all-trans conjugated polyenes (C2nH2n2, where n is the number of double bonds) of increasing chain length (n 1, 2, 3, 4, 5, 7, 9, 11) has been investigated by means of DFT/ UB3LYP [18] computations using the 6-31G* basis set [15], and the nature of the twisted structures (i.e. minima or transition states) has been inspected. Note that the last two longest terms of the series can be considered as violaxantin (n 9) and b-carotene (n 11) models, see Figure 9. This investigation has also allowed us to benchmark the accuracy of the DFT approach versus accurate high-level ab initio methods such as multireference MP2 computations (obtained using the CAS-PT2 methodology [22 23]). These results are summarized in Table 8 where the CAS-PT2 and DFT isomerization barriers are listed along with the zeropoint energy (ZPE) corrections (computed at the ab initio complete active space self consistent field (CAS-SCF) and DFT level of theory), and along with the available experimental barriers. The DFT barriers agree with the CAS-PT2 ones within 1 kcal mol1 and the same happens for the ZPE corrections, providing a very strong validation for the accuracy of the DFT approach. Moreover, the impressive agreement between experimental barriers and DFT ones (within 0.6 kcal mol1) indicates that this method provides a realistic description for the energetics of the isomerization 142

Table 8. CAS-PT2 and DFT computed rotational barriers a ( DECAS-PT2,DEDFT ) for TRANS!CIS isomerization of all-trans polyenes, zero-point vibrational energy correction at CAS-SCF b/ 6-31G* and DFT/UB3LYP/6-31G* levels ( DEzpe ), DFT zeropoint energy corrected rotational barriers ( DEcor ) and available experimental enthalpies of activation ( DH)c. nd 1 2 3 4 5 7 9 11
a

DECAS-PT2 62.8 54.2 43.9 38.9

DEDFT 62.9 53.2 44.2 39.8 35.6 30.8 27.7 25.5

DEzpe CAS-SCF 5.0 3.7 3.2

DEzpe DFT 4.5 3.9 3.2 3.1 2.9 2.8e 2.7e 2.6e

DEcor 58.4 49.3 41.0 36.7 32.7 28.0 25.0 22.9

DH 58.1 40.9[50c] 32.1 27.5 24.5 22.4f

All values in kcal/mol. b All CAS-SCF computations have been performed using the full active space of p-electrons and p-orbitals. c Experimental data refer to trans!cis rearrangements of semirigid polyenes as reported by Doering et al. [50b]. d n number of double bonds. e Zero-point DFT vibrational energy corrections ( DEzpe ) are computed (via analytically frequencies on twisted critical points and all-trans equilibrium geometries) only for n 1, 2, 3, 4, 5. For longer terms we assume DEzpe A B/(n C ), where A 2.42, B 2.50, C 0.20 are obtained by interpolations of the computed DFT zero-point energy corrections for the first three terms with an odd number of double-bonds (n 1, 3, 5); the asympthotic behaviour of the selected function is suggested by the first 5 computed terms (n 1, 2, 3, 4, 5); the choice of the first three odd terms (n 1, 3, 5) to interpolate is suggested by the necessity to obtain DEzpe values for the longer odd terms (n 7, 9, 11). f Extrapolated non-experimental value [50b].

Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

in polyenes and carotenoid systems. The computed barrier for the two longer terms (n 9, 11) shows that this process is already active at physiological temperature. The 22.9 kcal mol1 barrier computed for the b-carotene model agrees very well with the 22.4 kcal mol1 value reported by Doering et al. [50b] and with a recent kinetic estimate of 20 kcal mol1 [49]. Without entering too much into the details (the interested reader should refer to the full paper [51]), here we just say that, while the closed shell planar minima show the typical single/double CC bond alternation, the optimized central bond twisted structures (i.e. the transition state) are formed by two orthogonal non-interacting polyenil radicals which are characterized (for the longer terms of the series) by a central allyl-radical type system. Anyway, these structures do not display any special stabilization effect, except the one depending on radical delocalization, which is responsible for the lowering in the barriers as increasing the length of the chain. Indeed, all the points correspond to real transition states (see Figure 10), and no twisted intermediates are involved in the process. The involvement of a diradical-type radical scavenger was further investigated exploring the triplet (T1) PES. An analogous isomerization process was located for the bcarotene model (n 11), see Figure 10. In this case, both the optimized planar minima (experimental [52] and theoretical [53] evidences support a T1 PES with planar minima at the cis and trans positions) and transition structure (TS) have a similar diradical-type character, with the TS which is degenerate and identical to the one previously optimized on S0. If we suppose that T1 ! S0 ISC easily occurs near this point (where the singlet-triplet energy gap becomes very small) and we calculate the all-trans T1 lifetime from the computed energy barrier (11 kcal mol1), we get a value of

6 ms [54], which gives a surprising agreement with that observed for the all-trans b-carotene T1 state (5 mg) [55]. Therefore, if a very easy ISC between the S0 and T1 twisted structures occurs, thermal equilibration among the four minima (two closed-shell singlet and two diradical-type triplet minima), may generate diradical triplets with potential radical scavenging activity. Anyway, also in the case of efficient thermal equilibrium, the population of the more stable all-trans triplet minimum would be very low. In fact, the concentration ratio between the two all-trans singlet and triplet minima is about 1011, as we can calculate in first approximation by their related energy gap, suggesting that T1 is not responsible for the radical trapping action. All these considerations lead to the mechanistic hypothesis that it is the S0 closed shell minimum of carotenes to be responsible for the trapping action via formation of resonance-stabilized carbon-centered radicals or diradicals [38]. Thus, this simple investigation provides two very important pieces of information: (i) the DFT/B3LYP/6-31G* approach gives impressively good results both in the singlet and triplet states, and will be the elective method for these investigations; (ii) the diradical intermediate model for radical trapping activity must be ineffective. 5.1.2 A Further Insight: Bio-physical vs. Bio-chemical Paths The next step in the quest for understanding caroteniods biochemical properties involves the computational study of the reactions with 1O2 (using the previously validated DFT/ UB3LYP/6-31G* approach). The all-trans decaottanonaene (P9), a polyene with 9 conjugated double bonds, was selected as a carotenoid model. In fact, the 9 conjugated double bond moiety is a common feature in many carotenoids [36, 44a], while the substituents at the two ends of the chain may be

Figure 10. Simplified DFT singlet (S0) and triplet (T1) energy profiles for the isomerization about the central CC bond of the bcarotene model P11; q is the dihedral angle of rotation about the central double bond. MIN S0, MIN T1 represent the DFT optimized closed-shell S0 and diradical-type T1 (all-trans or cis) minima respectively, connected by the twisted diradical-type transition state TS (see Ref. 51 for further details).

Quant. Struct.-Act. Relat., 21 (2002)

143

F. Bernardi, et al.

different (see Figure 9). Moreover, several carotenoid systems with only 9 conjugated double bonds exist, such as violaxantin and neurosphorene, which show a singlet-oxygen quenching efficiency and a chemical reactivity comparable to that of b-carotene and longer carotenoids [44a] (Figure 9). Our computational results [56] indicate that carotenoids can be involved in different types of reactions, which include energy transfer, 1,2-addition, T1 dissociation and triplettriplet recombination. A summary of the corresponding reaction pathways is illustrated in Figure 11. It can be seen that the energy transfer process involves an almost barrierless path; therefore this catalytic physical quenching (Eq. 12) is the preferred route. Nevertheless, secondary but concomitant low energy barrier reactions (Eq. 13) occur via direct attack of singlet-oxygen upon the double bonds of

the carotenoid model. These processes lead to diradical systems where singlet and triplet states are degenerate. While ring closure reactions on S0 produce 1,2-addition dioxetane intermediates, (which may then decompose to the final observed carbonyl chain cleavage oxidation products [46]) an efficient and competitive S0 ! T1 ISC may also occur at the diradical minima configuration [56], and deactivated triplet oxygen, together with the starting singlet ground state carotene, is produced through a dissociation process on T1. Singlet diradical formation followed by ISC and triplet dissociation represent an alternative chemically mediated catalytic quenching of singlet-oxygen which seems to be more favored than oxidation reactions. This alternative process may act together with the more efficient physical pathway, reducing competitive oxidation which results in the loss of carotenoids and thus of antioxidant

Figure 11. Summary of the DFT energy profiles (spin-projected values in kcal mol1) of the main reaction paths computed for the longer model-system (O2 P9). Available experimental energies are given in frames. The relative positions of the triplet (T1) and singlet (S0) states of reactants, intermediates, products and transition states (TS1, TS2, TS3 and TS4) are shown. P9/Sing-Min and P9/Trip-Min represent the optimized all-trans planar minima in the S0 and T1 states respectively (see Ref. 56).

Scheme 4.

144

Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

protection. Carotenoid regeneration can in fact occurs also through a competitive chemical singlet-oxygen quenching path. Our results suggest the general reaction pattern shown in Scheme 4. Note the good agreement between the DFT-computed and the available experimental data (see Figure 11). Moreover, to assess the accuracy of the computations, we have also investigated the same process using a shorter polyene system, the all-trans hexatriene (P3). This model reaction (1P3 1O2) allowed both DFT [15] and multi-reference Moller-Plesset perturbation theory (CAS-SCF/CAS-PT2/631G*) computations [22 23], resulting in a reasonable agreement (see Ref. 56 for details).

6 Singlet Ground and Excited States via DFT in the Energy Transfer Process
Though, in general, standard DFT methods cannot describe excited states, still special situations exist that allow excited state computations, and the energy transfer problem presented above is one of these lucky cases. While the S0 wavefunction for the starting reactant refers to a singlet-

oxygen plus singlet-carotene coupling (1P9 1O2), the S0 wavefunction for the relaxed energy transfer product refers to a triplet-oxygen plus triplet-polyene coupling (3P9 3O2). Due to the abrupt change in the S0 wavefunction spin densities on going from a singlet-singlet to a triplet-triplet coupling description, it was possible to follow (all along the path) each coupling situation as a DFT-stable wavefunction. In this way, the two diabatic components of the process were computed, and consequently the singlet ground (S0) state and the singlet excited (S*) state surfaces along the energy transfer path [57] were estimated. Thus, both the singlet ground and the singlet excited states could be described, see Figure 12a. It is worth to say that energy transfer efficiency is not only a matter of energetics, but it also depends on vibronic interactions, short-distance interactions (such as overlap of the electron clouds), long-range antenna-type dipole-dipole interactions, according to the type (short range or long range) of energy transfer involved [58]. However, the fact that we have obtained an almost zero energy barrier is a demonstration of the efficiency of such a process (at least from an energetic point of view). This result is consistent with the observation of a very fast process (approaching the

Figure 12. Linear interpolated DFT energy profiles (values in kcal mol-1) for a) the energy transfer process in the long model-system (O2 P9), and b) the same hypothetical process in the small model-system (O2 P3). Dotted lines in a) represent the diabatic (singlet-singlet and triplet-triplet coupling) components of the energy transfer path. The relative positions for the singlet ground (S0) and singlet excited (S*) states of the hypothetical reactants and products are shown. P3/Sing-Min and P3/Trip-Min represent the optimized all-trans planar minima in the S0 and T1 states of the P3 system, respectively (see Ref. 56).

Quant. Struct.-Act. Relat., 21 (2002)

145

F. Bernardi, et al.

diffusion control limit [36, 44a], see Figure 9). Moreover, if we consider longer polyenes, which have a smaller vertical S0 ! T1 excitation energy, we should expect an even easier process because the crossing between the two diabatic surfaces should occur closer to the singlet-singlet reactant, due to the smaller energy gap separation between the ground and excited singlet states. In fact, long chain carotenoids such as dodecapreno b-carotene (19 double bonds) or decapreno b-carotene (15 double bonds) have higher singlet-oxygen quenching rate constants than short chain carotenoids such as violaxantin (9 double bonds) [44a] (see Figure 9). The short model system (1P3 1O2) allowed us to access the differences in chemical reactivity between long-chain polyene (such as carotenes) and shorter conjugated chains toward singlet-oxygen. Though the computed reaction paths and their trends are almost the same (see Ref. 56 for details), a striking difference appears: 1P3 does not allow an energy transfer path. The triplet-triplet product (3P3 3O2) of the hypothetical physical quenching is associated with the singlet excited state S*, and the two diabatic curves describing the singlet-singlet and triplet-triplet coupling never cross (see Figure 12b), in contrast to the behavior found in the longer system, which gives rise to curve crossing and to the energy transfer channel on S0, (see Figure 12a). This depends on the larger singlet-triplet energy gap separation in short conjugated polyenes such as P3, which prevents singlet-singlet/triplet-triplet curve crossing. As for the longer system, we have carried out computations for the triplet-triplet coupling (S* state) due to the intrinsic stability of the corresponding wavefunction (DFT wavefunction stability has been checked both in the singlet-singlet (S0) and triplet-triplet (S*) states). An interesting question may naturally arise from these results, i.e. how many double bonds are needed for the energy transfer process to exist and how many for it to become efficient with respect to the other chemical (oxygen addition) paths. An answer just based on the results for the investigated short (1P3) and long (1P9) systems is certainly not conclusive. Still, from the energy profiles shown in Figure 12b, we know that shifting the two diabatic curves (describing the singlet-singlet and triplet-triplet couplings) one to the other of only 9.5 kcal mol1 will result in curve crossing and possibly in the existence of an energy transfer (even if not energetically efficient) process. We guess this may happen from 5 or 6 conjugated double bonds polyenes. However, for a competitive physical quenching, the singletsinglet S0 and triplet-triplet S* states have to be very closed each other already at the reactant FC region. In fact, this is the condition to have a sudden crossing and to prevent a significant energy barrier (this simply means that the singlettriplet energy gap for the polyene has to be as close as possible to that of O2, which is exactly what happens in carotenes). Since a barrier (even if very small indeed) is still observable along the interpolated path of our P9 model system (see Figure 12a), we guess that 9 (or at most 8) is the minimum number of conjugated double bonds in a polyene to have energetically favored energy transfer. 146

7 Perspectives of DFT in Organic Chemistry


In this review we have reported the results of DFT investigations carried out in different fields of organic and organometallic chemistry. In particular, we have discussed examples of radical reactivity, structure and reactivity of organometallic compounds, and biochemical/biophysical properties of unsaturated (e.g. carotenoid) systems. We have demonstrated that the DFT approach represents a powerful tool, which can be used as a valid alternative to more traditional correlated methods such as Moller-Plesset perturbation theory, configuration interaction methods, coupled-cluster methods which require, when applied to molecules of chemical interest, strong and often untenable computational effort. The computational expedience, which characterizes DFT-based methods, makes this approach particularly interesting since it is possible to obtain an accurate description, which includes correlation energy, of medium and large size molecules as those involved in a reliable simulation of organic reactions or in the modelling of bio-organic systems. The examples selected in this review, which are representative of important organic, organometallic and bio-organic reactions, have been investigated using different DFT functionals. These functionals (B3LYP, BHLYP, BLYP, BP86) are within the most popular non-local corrected functionals, which can be easily found in many commercially available quantum chemistry packages. In all cases we have proved that they are capable of providing data which reproduce satisfactorily the experimental results or the data obtained at higher levels of theory.

References
[1] Parr, R. G., and Yang, W., Density Functional Theory of Atoms and Molecules, Oxford University Press, New York 1989. [2] Fulde, P., Electron Correlations in Molecules and Solids, Springer-Verlag, Berlin, Heidelberg 1991. [3] Hohenberg, P., and Kohn, W., Inhomogeneous Electron Gas, Phys. Rev. B 136, 864 (1964). [4] Kohn, W., and Sham, L., Self-Consistent Equations Including Exchange and Correlation Effects, J. Phys. Rev. A 140, 1133 (1965). [5] Dreizler, R. M., and Gross, E. K. U., Density Functional Theory. An Approach to the Quantum Many-Body Problem, Springer-Verlag, Berlin 1990, pp. 172 188. [6] Versluis, L., and Ziegler, T., The Determination of Molecular Structures by Density Functional Theory. The Evaluation of Analytical Energy Gradients By Numerical Integration, J. Chem. Phys. 88, 322 328 (1988); Ziegler, T., Fan, L., Tschinke, V., and Becke, A., Theoretical Study on the Electronic and Molecular Structures of (C5H5)M(L) (M Rh, Ir; L CO, PH3) and M(CO)4 (M Ru,Os) and Their Ability to Activate the C-H Bond in Methane, J. Am. Chem. Soc., 111, 9177 9185 (1989); Ziegler, T., Tschinke, V., Baerends, E. J., Snijders, J. G., and Ravenek, W., Calculation of Bond Energies in Compounds of Heavy Elements by a Quasi-Relativistic Approach, J. Phys. Chem. 93, 3050 3056 (1989). [7] Ziegler, T., Approximate Density Functional Theory as a Practical Tool in Molecular Energetics and Dynamics, Chem. Rev. 91, 651 667 (1991); Density Functional Methods in

Quant. Struct.-Act. Relat., 21 (2002)

Exploring Organic Chemistry with DFT: Radical, Organo-metallic, and Bio-organic Applications

Chemistry, Labonowsky, J. K., and Andzelm, J. (Eds.), Springer-Verlag, New York 1991. [8] Becke, A. D., Hartree-Fock Exchange Energy of an Inhomogeneous Electron Gas, Int. J. Quantum Chemistry 23, 1915 1930 (1983); Becke, A. D., Density Functional Calculations of Molecular Bond Energies, J. Chem. Phys. 84, 4524 4529 (1986). [9] Fan, L., and Ziegler, T., Nonlocal Density Functional Theory as a Practical Tool in Calculations on Transition States and Activation Energies. Applications to Elementary Reaction Steps in Organic Chemistry, J. Am. Chem. Soc. 114, 10890 10897 (1992). [10] Becke, A. D., Density-Functional Thermochemistry. I. The Effect of the Exchange-Only Gradient Correction, J. Chem. Phys. 96, 2155 2160 (1992). [11] Fan, L., and Ziegler, T., Optimization of Molecular Structures by Self-Consistent and Non-local Density Functional Theory, J. Chem Phys. 95, 7401 7408 (1991); Fan, L., and Ziegler, T., Application of Density Functional Theory to Infrared Absorption Intensity Calculations on Transition-Metal Carbonyls, J. Phys. Chem. 96, 6937 6941 (1992). [12] ADF Amsterdam Density Functional Program, Scientific Computing & Modelling NV, Vrije Universiteit, Theoretical Chemistry, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands. [13] Turbomole, version 5.3 Ahlrichs, R., Institut fr Physikalische Chemie und Elektrochemie, Universitt Karlsruhe, 76128 Karlsruhe, Germany. [14] Gaussian 92/DFT, Revision G.1, Frisch, M. J., Trucks, G. W., Schlegel, H. B., Gill, P. M. W., Johnson, B. G., Wong, M. W., Foresman, J. B., Robb, M. A., Head-Gordon, M., Replogle, E. S., Gomperts, R., Andres, J. L., Raghavachari, K., Binkley, J. S., Gonzalez, C., Martin, R. L., Fox, D. J., Defrees, D. J., Baker, J., Stewart, J. P., Pople, J. A., Gaussian, Inc., Pittsburgh PA, 1993. [15] Gaussian 94, Revision B.2, Frisch, M. J., Trucks, G. W., Schlegel, H. B., Gill, P. M. W., Johnson, B. G., Robb, M. A., Cheeseman, J. R., Keith, T., Petersson, G. A., Montgomery, J. A., Raghavachari, K., Al-Laham, M. A., Zakrzewski, V. G., Ortiz, J. V., Foresman, J. B., Peng, C. Y., Ayala, P. Y., Chen, W., Wong, M. W., Andres, J. L., Replogle, E. S., Gomperts, R., Martin, R. L., Fox, D. J., Binkley, J. S., Defrees, D. J., Baker, J., Stewart, J. P., Head-Gordon, M., Gonzalez, C., Pople, J. A., Gaussian, Inc., Pittsburgh PA. [16] Gaussian 98, Revision A.6, Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, E. G., Robb, M. A., Cheeseman, J. R., Zakrzewski, V. G., Montgomery, J. A., Stratmann, R. E., Burant, J. C., Dapprich, S., Millam, J. M., Daniels, A. D., Kudin, K. N., Strain, M. C., Farkas, O., Tomasi, J., Barone, V., Cossi, M., Cammi, R., Mennucci, B., Pomelli, C., Adamo, C., Clifford, S., Ochterski, J., Petersson, G. A., Cui, Q., Morokuma, K., Malik, D. K., Rabuck, A. D., Raghavachari, K., Foresman, J. B., Cioslowski, J., Ortiz, J. V., Stefanov, B. B., Liu, G., Liashenko, A., Piskorz, P., Kamaromi, I., Gomperts, R., Martin, R. L., Fox, D. J., Keith, T., Al-Laham, M. A., Peng, C. Y., Nanayakkara, A., Gonzalez, C., Challacombe, M., Gill, P. M. W., Johnson, B. G., Chen, W., Wong, M. W., Andres, J. L., Gonzalez, C., Head-Gordon, M., Replogle, E. S., Pople, J. A., Gaussian, Inc., Pittsburgh PA 1998. [17] Slater, J. C., Quantum Theory of Molecules and Solids. Vol. 4, McGraw-Hill, New York 1974. [18] Becke, A. D., Density Functional Thermochemistry. III. The Role of Exact Exchange, J. Chem. Phys. 98, 5648 5652 (1993). [19] Vosko, S. H., Wilk, L., and Nusair, M., Accurate SpinDependent Electron Liquid Correlation Energies for Local

Spin Density Calculations: a Critical Analysis, Canadian J. Phys. 58, 1200 (1980). [20] Lee, C., Yang, W., Parr, R. G., Development of the ColleSalvetti Correlation Energy Formula into a Functional of the Electron Density, Phys. Rev. B 37, 785 (1988). [21] Perdew, J. P., Density Functional Approximation for the Correlation Energy of the Inhomogeneous electron gas, Phys. Rev. B 33, 8822 (1986). [22] MOLCAS version 3, Andersson, K., Fulscher, M. P., Lindth, R.., Malmqvist, P., Olsen, J., Roos, B. O., Sadlej, A. J., University of Lund, Sweden, and Widmark, P. O., IBM Sweden (1991). [23] Andersson, K., Malmqvist, P., Roos, B. O., Sadlej, A. J., and Wolinski, K., Second-Order Perturbation Theory with a CASSCF Reference Function, J. Phys. Chem. 94, 5483 5488 (1990); Andersson, K., Malmqvist, P., Roos, B. O., Second-Order Perturbation Theory with a Complete Active Space Self-Consistent Field Reference Function, J. Chem. Phys. 96, 1218 1226 (1992). [24] Yamaguchi, K., Jensen, F., Dorigo, A., and Houk, K. N., A Spin Correction Procedure for Unrestricted Hartree-Fock and Moller-Plesset Wavefunctions for Singlet Diradicals and Polyradicals, Chem. Phys. Lett. 149, 537 (1988); Yamanaka, S., Kawakami, T., Nagao, H., and Yamaguchi, K., Effective Exchange Integrals for Open-Shell Species by Density-Functional Methods, Chem. Phys. Lett. 231, 25 (1994); Goldstein, E., Beno, B., Houk, K. N., Density Functional Theory Prediction of the Relative Energies and Isotope Effects for the Concerted and Stepwise Mechanisms of the Diels-Alder Reaction of Butadiene and Ethylene, J. Am. Chem. Soc. 118, 6036 6043 (1996). [25] Tedder, J. M., and Walton, C., The Kinetics and Orientation of Free-Radical Addition to Olefins, Acc. Chem. Res 9, 183 191 (1976); Heberger, K., Walbiner, M., and Fisher, H., Addition of Benzyl Radicals to Alkenes: the Role of Radical Deformation in the Transition State, Angew. Chem. Int. Ed. 31, 635 636 (1992); Giese, B., Formation of CC Bonds by Addition of Free Radicals to Alkenes, Angew. Chem. Int. Ed. 22, 753 764 (1983); Tedder, J. M., Which Factors determine the Reactivity and Regioselectivity of Free Radical Substitution and Addition Reactions?, Angew. Chem. Int. Ed. 21, 401 410 (1982); Sosa, C., and Schlegel, H. B., Calculated Barrier Heights for OH C2H2 and OH C2H4 Using Unrestricted Moller-Plesset Perturbation Theory with Spin Annihilation, J. Am. Chem. Soc. 109, 4193 4198 (1987); Wong, M. W., Pross, A., and Radom, L., Comparison of the Addition of CH3 ., CH2OH . and CH2CN .. Radicals to Substituted Alkenes: A Theoretical Study of the Reaction Mechanism, J. Am. Chem. Soc. 116, 6284 6292 (1994); Bottoni, A., Theoretical Study of the Addition of Alkyl and Halogenoalkyl Radicals to the Ethylene Double Bond: a Comparison between Hartree-Fock, Perturbation Theory and Density Functional Theory, J. Chem. Soc. Perkin Trans. 2, 2041 (1996) and references reported therein. [26] Sosa, C., and Schlegel, H. B., An ab-Initio Study of the Reaction Pathways for OH C2H4 ! HOCH2CH2 ! Products, J. Am. Chem. Soc 109, 7007 7015 (1987). [27] Bernardi, F., and Bottoni, A., Polar Effects in Hydrogen Abstraction Reactions from Halo-Substituted Methanes by Methyl Radical: a Comparison between Hartree-Fock, Perturbation, and Density Functional Theories, J. Phys. Chem. A 101, 1912 1919 (1997) and references reported therein. [28] Chatgilialoglu, C., Structural and Chemical Properties of Silyl Radicals, Chem. Rev. 95, 1229 1251 (1995) and references reported therein. [29] Bottoni, A., Theoretical Study of the Hydrogen and Chlorine Abstraction from Chloromethanes by Silyl and Trichlorosilyl Radicals: a Comparison between the Hartree-Fock Method,

Quant. Struct.-Act. Relat., 21 (2002)

147

F. Bernardi, et al.

Peturbation Theory, and Density Functional Theory. J. Phys. Chem. A 102, 10142 10150 (1998) and references reported therein. [30] Bernardi, F., Bottoni, A., Calcinari, M., Rossi, I., and Robb, M. A., Comparison between CAS-PT2 and DFT in the Study of Ni(C2H4)2 Complexes, J. Phys. Chem. A 101, 6310 6314 (1997) and references reported therein. [31] Bernardi, F., Bottoni, A., and Rossi, I., A DFT Investigation of Ethylene Dimerization Catalyzed by Ni(0) Complexes, J. Am. Chem. Soc. 120, 7770 7775 (1998) and references reported therein. [32] Mitsudo, T., Naruse, H., Kondo, T., Ozaki, Y., and Watanabe, Y., [2 2] Cycloaddition of Norbornenes with Alkyns Catalyzed by Ruthenium Complexes, Angew. Chem. Int. Ed. Engl. 33, 580 581 (1994) and references reported therein. [33] Bernardi, F., Bottoni, A., and Miscione, G. P., A Theoretical Study of the Homogeneous Ziegler-Natta Catalysis, Organometallics 17, 16 24 (1998) and references reported therein. [34] Eisch, J. J., Pombrik, S. I., and Zheng, G. Active Sites for Ethylene Polymerization with Titanium(IV) Catalysts in Homogeneous Media: Multinuclear NMR Study of Ion-Pair Equilibria and their Relation to Catalysts Activity, Organometallics, 12, 3856 (1993) and references reported therein. [35] Bensasson, R. V., Land, E. J., and Truscott, T. G., Excited States and Free Radicals, in Biology and Medicine; Oxford University Press, New York, 1993, pp. 201 227. [36] Light-Harvesting Physics Workshop, Bristonas, Lithuania, September 1997; Special issue of the J. Phys. Chem. 101, 7197 7360 (1997). [37] Blot, W. J., Li, J.-Y., Taylor, P. R., Guo, W., Dawsey, S., Wrang, G.-Q., Yang, C. S., Zheng, S.-F., Gail, M., Li, G.-Y., Yu, Y., Liu, B.-Q., Tangrea, J., Sun, Y.-H., Liu, F., Fraumeni, J. F., Jr., Zhang, Y.-H., and Li, B., Nutrition Intervention Trials in Linxian, China Supplementation with Specific Vitamin Mineral Combinations, Cancer Incidence, and Disease-Specific Mortality in the General-Population, J. Nat. Cancer Inst. 85, 1483 1492 (1993). [38] Burton, G. W., and Ingold, K. U., b-Carotene - An Unusual Type of Lipid Antioxidant, Science 224, 569 (1984). [39] Everett, S. A., Kundu, S. C., Maddix, S., and Willson, R. L., Mechanisms of Free-Radical Scavenging by the Nutritional Antioxidant b-Carotene, Biochem. Soc. Trans. 23, 230 (1995). [40] Ozhogina, O. A., and Kasaikina, O. T., b-Carotene as an Interceptor of Free-Radicals, Free Radic. Biol. Med. 19, 575 (1995). [41] Mathis, P., Organic Photochemistry and Photobiology; Horspool, W. M., Song, P.-S., Eds; CRC Press, New York, 1995; Chapter 16, pp. 1412. [42] Ames, B. N., Dietary Carcinogens and Anticarcinogens Oxygen Radicals and Degenerative Diseases, Science 221, 1256 (1983). [43] Tinkler, J. H., Tavender, S. M., Parker, A. W., McGarvey, D. J., Mulroy, L., and Truscott, T. G., Investigation of Carotenoid Radical Cations and Triplet States by Laser Flash Photolysis and Time-Resolved Resonance Raman Spectroscopy: Observation of Competitive Energy and Electron Transfer, J. Am. Chem. Soc. 118, 1756 (1996); Bohm, F., Edge, R., Land, E. J., McGarvey, D. J., and Truscott, T. G., Carotenoids Enhance Vitamin E Antioxidant Efficiency, J. Am. Chem. Soc. 119, 621 622 (1997). [44] Conn, P. F., Schalch, W., and Truscott, T. G., The Singlet Oxygen and Carotenoid Interaction, J. Photochem. Photobiol. B: Biol. 11, 41 47 (1991); Speranza, G., Manitto, P., and Monti, D., Interaction Between Singlet Oxygen and Biologically-Active Compounds in Aqueous-Solution 3. Physical

and Chemical 1O2-Quenching Rate Constants of 6,6'-Diapocarotenoids, J. Photochem. Photobiol. B. 8, 51 56 (1990). [45] Foote, C. S., in: Singlet Oxygen, Wasserman, H. H., Murray, R. W. (Eds.), Academic Press, New York 1979; pp 139 171. [46] Stratton, S. P., Schaefer W. H., and Liebler, D., Isolation and Identification of Singlet Oxygen Oxidation-Products of bCarotene, Chem. Res. Toxicol. 6, 542 (1993) and references cited therein. [47] El-Oualja, H., Perrin, D., and Martin, R., Kinetic-Study of the Thermal-Oxidation of All-Trans-b-Carotene and Evidence for its Antioxygen Properties, New. J. Chem. 19, 863 (1995). [48] El-Oualja, H., Perrin, D., and Martin, R., Influence of bCarotene on the Induced Oxidation of Ethyl Linoleate, New. J. Chem. 19, 1187 (1995). [49] Perrin, D., El-Oualja, H., and Martin, R., Antioxidant Properties of All-Trans-b-Carotene a-Tocopherol Mixtures in the Induced Oxidation of Ethyl Linoleate, Journal de chimie physique et de physico-chimie biologique 93, 1462 (1996). [50] Doering, W. Von E., and Kitagawa, T., Thermal Cis-Trans Rearrangement of Semirigid Polyenes as a Model for the Anticarcinogen b-Carotene An All-Trans-Pentaene and an All-Trans-Heptaene, J. Am. Chem. Soc. 113, 4288 (1991); Doering, W. Von E., and Sarma, K., Stabilization Energy of Polyenyl Radicals All-Trans-Nonatetraenyl Radical by Thermal Rearrangement of A Semirigid (4-1-2) Heptaene Model for Thermal Lability of b-Carotene, J. Am. Chem. Soc. 114, 6037 (1992); Doering, W. Von E., Roth, W. R., Bauer, F., Boenke, M., Breuckmann, R., Ruhkamp, J., and Wortmann, O., Rotational Barriers of Vinyl-Substituted Olefins, Chem. Ber. 124, 1461 (1991). [51] Bernardi, F., Garavelli, M., Olivucci, M., and Robb, M. A., Trans ! Cis Isomerizaion in Long Linear Polyenes as bCarotene Models: a Comparative CAS-PT2 and DFT study, Molec. Phys. 92, 359 364 (1997). [52] Kuki, M., Koyama, Y., and Nagae, H., Triplet-Sensitized and Thermal-Isomerization of All-Trans, 7-Cis, 9-Cis, 13-Cis, and 15-Cis Isomers of b-Carotene - Configurational Dependence of the Quantum Yield of Isomerization via the T1 State, J. Phys. Chem. 95, 7171 (1991). [53] Negri, F., and Orlandi, G., The T-1 Resonance Raman-Spectra of 1,3,5-Hexatriene and Its Deuterated Isotopomers An AbInitio Reinvestigation, J. Chem. Phys. 103, 2412 (1995). [54] The rate constant is estimated using the simple Arrenius equation (k A e-Ea/RT, where A is the pre-exponential factor) at the physiological temperature of 310 K. For unimolecular reactions the lifetime (t) is calculated as the inverse of the firstorder rate constant (t k1), and a standard unimolecular preexponential factor of 1013 s1 (see Benson, S. W.; Thermochemical Kinetics; Wiley, New York, 1976) has been used. [55] Hashimoto, H., Koyama, Y., Ichimura, K., and Kobayashi, T., Time-Resolved Absorption-Spectroscopy of the Triplet-State Produced from the All-Trans, 7-Cis, 9-Cis, 13-Cis, And 15-Cis Isomers of b-Carotene, Chem. Phys. Lett. 162, 517 (1989). [56] Garavelli, M., Bernardi, F., Olivucci, M., and Robb, M. A., DFT Study of the Reactions between Singlet-Oxygen and a Carotenoid Model, J. Am. Chem. Soc. 120, 10210 10222 (1998). [57] The diabatic components of the adiabatic electronic eigenfunctions describe the energy of a particular spin-coupling (see Olivucci, M. Ph. D Thesis, University of Bologna, 1988), while the adiabatic functions represent the surfaces of the real states (the singlet ground and excited states in this case). [58] Gilbert, A. and Bargott, S., Essentials of Molecular Photochemistry; Blackwell Scientific Pubblications, Oxford, 1991; pp. 167 181. Received on December 4, 2001; accepted on February 27, 2002

148

Quant. Struct.-Act. Relat., 21 (2002)

You might also like