Generating CuII-Oxyl/CuIII-Oxo Species from CuI- α-Ketocarboxylate Complexes and O2: In silico studies on ligand effects and C-H-activation reactivity

(1)

Article

Reference

Generating CuII-Oxyl/CuIII-Oxo Species from CuI- α-Ketocarboxylate Complexes and O2: In silico studies on ligand effects and

C-H-activation reactivity

HUBER, Stefan M., et al .

Abstract

The mechanistic details associated with the generation and reaction of [CuO]+ species from CuI-[alpha]-ketocarboxylate complexes, especially with respect to modifications of the ligand supporting the copper center, were investigated (see scheme). Theoretical models were used to characterize the electronic structures of different [CuO]+ species and their reactivity in C—H activation and O-atom transfer reactions.A mechanism for the oxygenation of CuI complexes with -ketocarboxylate ligands that is based on a combination of density functional theory and multireference second-order perturbation theory (CASSCF/CASPT2) calculations is elaborated. The reaction proceeds in a manner largely analogous to those of similar FeII-[alpha]-ketocarboxylate systems, that is, by initial attack of a coordinated oxygen molecule on a ketocarboxylate ligand with concomitant decarboxylation. Subsequently, two reactive intermediates may be generated, a Cu-peracid structure and a [CuO]+ species, both of which are capable of oxidizing a phenyl ring component of the supporting ligand.

Hydroxylation by the [CuO]+ species is predicted to proceed [...]

HUBER, Stefan M., et al . Generating CuII-Oxyl/CuIII-Oxo Species from CuI- α-Ketocarboxylate Complexes and O2: In silico studies on ligand effects and C-H-activation reactivity. Chemistry - A European Journal , 2009, vol. 15, no. 19, p. 4886-4895

DOI : 10.1002/chem.200802338

Available at:

http://archive-ouverte.unige.ch/unige:3742

Disclaimer: layout of this document may differ from the published version.

(2)

DOI: 10.1002/chem.200802338

Generating Cu

^II

–Oxyl/Cu

^III

–Oxo Species from Cu

^I

–a-Ketocarboxylate Complexes and O

₂

: In Silico Studies on Ligand Effects and C H-Activation

Reactivity

Stefan M. Huber,

^{[a, b]}

Mehmed Z. Ertem,

^[a]

Francesco Aquilante,

^[b]

Laura Gagliardi,*

^{[a, b]}

William B. Tolman,*

^[a]

and Christopher J. Cramer*

^[a]

Introduction

Copper enzymes that perform catalytic oxidation reactions involving molecular oxygen feature a variety of active-site structures and traverse multiple reaction pathways that remain the targets of intense study.^[1]Among the approaches used in such investigations, the characterization of synthetic copper–oxygen complexes that mimic possible reaction intermediates in the enzymatic processes has been particularly fruitful. For example, detailed structural, spectroscopic, and mechanistic insights have been obtained from extensive studies of systems featuring [Cu2(O2)]²⁺, [Cu2ACHTUNGTRENNUNG(m-O)2]²⁺, or [CuO₂]⁺ cores.^[2]

On the other hand, other proposed reactive intermediates in reactions catalyzed by copper enzymes remain less well characterized, the most notable example being mononuclear copper–oxo species [CuO]⁺that can be described as limiting Abstract: A mechanism for the oxy-

genation of Cu^Icomplexes witha-ketocarboxylate ligands that is based on a combination of density functional theory and multireference second- order perturbation theory (CASSCF/

CASPT2) calculations is elaborated.

The reaction proceeds in a manner largely analogous to those of similar Fe^II–a-ketocarboxylate systems, that is, by initial attack of a coordinated oxygen molecule on a ketocarboxylate ligand with concomitant decarboxylation. Subsequently, two reactive intermediates may be generated, a Cu–peracid structure and a [CuO]⁺ species, both of which are capable of oxidizing a phenyl ring component of the supporting ligand. Hydroxylation by the

[CuO]⁺ species is predicted to proceed with a smaller activation free energy.

The effects of electronic and steric variations on the oxygenation mechanisms were studied by introducing substituents at several positions of the ligand backbone and by investigating various N-donor ligands. In general, more electron donation by the N-donor ligand leads to increased stabilization of the more CuÎI/CuÎII-like intermediates (oxygen adducts and [CuO]⁺ species) relative to the more CuÎ-like peracid

intermediate. For all ligands investigated, the [CuO]⁺ intermediates are best described as Cu^IIO·species with triplet ground states. The reactivity of these compounds in CH abstraction reactions decreases with more electron- donating N-donor ligands, which also increase the CuO bond strength, although the CuO bond is generally predicted to be rather weak (with a bond order of about 0.5). A comparison of several methods to obtain singlet energies for the reaction intermediates indicates that multireference second-order perturbation theory is likely more accurate for the initial oxygen adducts, but not necessarily for subsequent reaction intermediates.

Keywords: CH activation · electronic structure · multiconfigurational quantum chemical methods· OO activation

[a] Dr. S. M. Huber, M. Z. Ertem, Prof. Dr. L. Gagliardi, Prof. Dr. W. B. Tolman, Prof. Dr. C. J. Cramer Department of Chemistry

Center for Metals in Biocatalysis and Supercomputing Institute University of Minnesota

207 Pleasant St. SE

Minneapolis MN 55455 (USA) Fax: (+1) 612-642-7029 E-mail: [email protected]

[email protected]

[b] Dr. S. M. Huber, Dr. F. Aquilante, Prof. Dr. L. Gagliardi Department of Physical Chemistry

University of Geneva 30, Quai Ernest Ansermet 1211 Geneva (Switzerland) Fax: (+41) 22-379-6518

E-mail: [email protected]

Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/chem.200802338.

2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eur. J.2009,15, 4886 – 4895

4886

(3)

valence bond CuÎII–oxo or CuÎI–oxyl complexes. While these species have been suggested as possible intermediates in enzymes, such as peptidylglycine a-hydroxylating monooxyge- nase,^[3] and some synthetic systems,^[4] to the best of our knowledge they have only been unambiguously identified as reactive species in the gas phase.^[5]Initial theoretical investigations^{[1, 5, 6]}suggest that [CuO]⁺ systems should be powerful oxidants, similar to or even surpassing the closely related and extensively studied FeÎV–oxo complexes.^[7]

One means for generating the analogous [FeO]²⁺ functionality is by reaction of molecular oxygen with an Fe^II–a- ketocarboxylate moiety in various enzymes and model complexes.^[8]A consensus mechanism for these systems is shown in Scheme 1. After formation of initial O₂adduct1, the ter-

minal oxygen atom of the former O2moiety attacks the a- keto carbon of the ketocarboxylate ligand, resulting in the loss of CO2 and the generation of peracid intermediate 2.

Subsequent OO cleavage yields Fe^IV–oxo intermediate 3, which is widely considered to be the active species in hydrogen atom abstraction reactions from CH bonds of various substrates.

Recently we reported experimental and theoretical results^[9]that seem to indicate that a similar reaction can be achieved by starting from CuÎcomplexes witha-ketocarboxylate ligands (Scheme 2). In these experiments, novel CuÎ–a- ketocarboxylate complexes 4 and 5(amongst others) were reacted with O2at low temperatures. The color of the solu- tion was observed to gradually change and after warming and removal of copper under basic conditions, a mixture of the original ligand (L or L^m-OMe) and its hydroxylated form (LÔH or LÔH,m-OMe) was identified spectroscopically. In addi-

tion, a mixture of benzoyl formic acid and benzoic acid was obtained after acidic workup, indicating that partial decarboxylation had taken place (about 40 % for R=H and 60 % for R=OMe). This latter observation, in combination with a more favorable ratio of hydroxylated ligand to the original ligand for5compared with4, suggests that the reaction may proceed more efficiently with more electron-donating ligands. On the basis of isotope labeling studies, density functional theory (DFT), and multireference second-order perturbation theory (CASSCF/CASPT2) calculations, a mechanism largely analogous to the Fe^IIcase (Scheme 1) was proposed (see below).

Herein we refine and report further theoretical investigations concerning this reaction mechanism, especially with regard to modifications of the ligand environment around the copper center. In particular, we introduce donor or ac- ceptor substituents into the original ligand framework and examine oxygenation reactions in complexes with different N-donor ligands. Furthermore, we investigate the influence of those different ligand systems on the electronic nature of the [CuO]⁺ species and their reactivity in CH activation and O-atom transfer reactions.

Results and Discussion

Fundamental mechanism: A simplified^[10] and slightly re- vised^[11]version of the mechanism we proposed for the oxygenation of4(and5)^[9]is presented in Scheme 3 based on a Scheme 1. General mechanism for the reactivity of Fe^II–a-ketocarboxy-

late sites in enzymes and model complexes.

Scheme 2. Oxygenation of Cu^I–a-ketocarboxylate complexes4and5.

Scheme 3. Proposed general mechanism for the oxygenation of Cu^I–a-ketocarboxylate complexes. The N-donor ligand is only depicted schemati- cally. See also Figures 1 and 2 and Tables 1 and 2 for free energies. Struc- tures in gray are only relevant for ligandsL⁰–L⁶.

FULL PAPER

(4)

combination of DFT and CASPT2 calculations (detailed de- scriptions of the theoretical models and a discussion of their relative accuracies are provided in a later section; we use DFT to mean the M06L functional throughout). We first considered ligandL⁰ (Figure 1), a slightly truncated version

of the ligand system used experimentally (cf. Scheme 2).

The Cu^I–a-ketocarboxylate complexes supported by that ligand can exist as three isomers 6 a–c (Figure 2), which

differ primarily in the mode of coordination of the ketocarboxylate ligand. The most stable isomer6 a(cf. Table 1) features a bidentate ketocarboxylate ligand which is coordinated by one oxygen atom of the carboxylate group and the oxygen atom of the keto group. An isomeric bidentate motif with two coordinating carboxylate oxygen atoms (6 b) is 2.5 kcal mol¹ higher in energy (according to our computations), but still considerably more favorable than the mono- dentate isomer (6 c). As expected, all three species have a singlet ground state. Oxygenation of6 a is roughly ergoneutral and leads to the formation of oxygen adducts 7 a–c, which differ in the mode of coordination of the ketocarboxylate ligand and O2(see Figure 2).

Isomer7 a(which features the same PhC(O)CO2moiety binding motif as 6 a) is the most stable, with the other two variants only 1–2 kcal mol¹ higher in energy. Though one might expect a more CuÎII–peroxo-like character for side-on isomer 7 c(i.e., a singlet ground state),^[12] all three isomers are of triplet multiplicity and are thus closer to CuÎI than to CuÎII.

Decarboxylation of the ketocarboxylate ligand occurs by approach of the terminal oxygen of the O2 ligand to the keto functionality of the former and involves a barrier of about 23 kcal mol¹; the reaction step itself is exergonic by more than 40 kcal mol¹. Although decarboxylation transition state (TS) 8 still has triplet multiplicity, the resulting peracid structure,9, is a ground-state singlet (and shows no restricted to unrestricted instability at the DFT level).

Two pathways leading to the formation of product 11 starting from peracid structure 9 were found. One of them involves direct hydroxylation of a phenyl ring of the N- donor ligand via TS 10 (which also has a singlet ground state) and has a free energy of activation of about 18 kcal mol¹ (the reaction is exergonic by about 20 kcal mol¹). Alternatively, cleavage of the OO bond via energetically low-lying TS 12 yields the [CuO]⁺ intermedi- ate13, in which the two N-donor atoms and one carboxylate oxygen occupy the equatorial sites of an overall trigonal-bipyramidal coordination sphere. For the oxo oxygen atom to be favorably oriented for the following hydroxylation step, an isomerization must occur. This process again features a relatively low free energy of activation (7 kcal mol¹) and results in square-planar [CuO]⁺ intermediate 15, in which both N-donor atoms, the Cu atom, and both coordinated O atoms lie in the same plane. Hydroxylation of the ligand phenyl ring by this intermediate via TS16involves an activation free energy of about 6 kcal mol¹. The pathway from oxo species13to the product11^[13]occurs on the triplet surface. The highest activation free energy associated with this Figure 1. Model ligandL⁰, used for evaluation of the mechanism of the

oxygenation reaction shown in Scheme 2, and various substituted deriva- tivesL¹–L⁶, used to study the influence of substitution effects on the individual steps of the mechanism in Scheme 3.

Figure 2. Isomers of the starting complexes and their oxygen adducts (compare with Scheme 3).

Table 1. Free energies [kcal mol¹] of starting complexes (6 a–c) and oxygen adducts (7 a–c)^[a] relative to the lowest respective isomer (see Figure 3). Structures of ligands L⁰–L⁶ and ofL⁷–L¹² are shown in Fig- ures 1 and 2, respectively.

6 a 6 b 6 c 7 a 7 b 7 c

L⁰ 0.0 2.5 6.8 0.0 (T) 1.2 (T) 1.5 (T)

L¹ 0.0 4.0 8.3 0.0 (T) 2.7 (T) 2.0 (T)

L² 0.0 2.6 6.3 0.2 (T) 0.0 (T) 0.2 (T)

L³ 0.0 2.1 5.7 0.0 (T) 1.2 (T) 1.1 (T)

L⁴ 0.0 3.2 8.1 0.0 (T) 1.8 (T) 1.0 (T)

L⁵ 0.0 2.5 6.6 0.0 (T) 0.8 (T) 0.0 (T)

L⁶ 0.0 2.7 7.4 0.0 (T) 2.5 (T) 2.1 (T)

L⁷ 0.0 3.2 8.5 0.8 (T) 0.0 (T) 4.1^[b]

L⁸ 0.0 n/a^[c] 0.4 0.7 (T) 1.5 (T) 0.0 (T)

L⁹ 0.0 n/a^[c] 0.7 2.9^[d] 1.8 (T) 0.0 (S)

L¹⁰ 0.0 1.9 n/a^[c] 0.5 (T) 0.0 (T) 5.4^[b]

L¹¹ 0.4 0.6 0.0 0.0 (T) 1.5 (T) n/a^[c]

L¹² 1.2 0.0 0.7 0.0 (T) 5.1 (T) n/a^[c]

[a] Ground-state multiplicity of oxygen adducts7 a–cin brackets; S=singlet, T=triplet. [b] No minimum could be located on triplet surface.

[c] Isomer could not be located. [d] No minimum could be located on the singlet surface.

www.chemeurj.org 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eur. J.2009,15, 4886 – 4895

4888

C. J. Cramer, W. B. Tolman et al.

(5)

multistep reaction path is 12 kcal mol¹above peracid 9. As direct hydroxylation from the peracid has an activation free energy of 18 kcal mol¹, we conclude that the stepwise pathway via intermediates 13 and 15will be the preferred one provided spin-orbit coupling permits efficient crossing from the singlet to triplet surfaces. With a somewhat reduced ligand, we previously computed spin-orbit coupling matrix elements of about 350 cm¹ for TS structures 10 and 12, which suggests that such spin crossing should indeed be reasonably efficient.

Considering for a moment a third possible ring-oxidation pathway, we note that the oxo functionality in 13 is quite close to the orthoposition of the phenyl ring that is to be hydroxylated and may even form a weak hydrogen bond to the hydrogen atom at this position (OH distance 2.46 ).

Thus, an alternative hydroxylation mechanism involving H- atom abstraction followed by OH rebound might be envi- sioned (Scheme 4).^[14] However, the associated activation

free energies indicate that this pathway is probably not competitive with the “stepwise” hydroxylation in Scheme 3 (for which the activation free energy relative to 13 is only 10.1 kcal mol¹).

Influence of substituents on ligand L⁰: To investigate the influence of electronic modifications to ligandL⁰on the reaction mechanism, electron-donating and -withdrawing groups were introduced at different positions on the aromatic rings of either the N-donor ligand or the ketocarboxylate ligand (Figure 1). Our primary focus was the effect of these modifications on the oxygenation thermodynamics and activation barriers for the individual reaction steps.

Introducing a p-electron-donating (L¹) or -withdrawing (L²) substituent at thepara position of the aromatic system of the ketocarboxylate ligand mainly affects the relative energies of the starting complexes andDG2°. Because electron donation increases the donor capacity of the keto oxygen atom, isomer 6 a of the starting complexes is favored with respect to the other alternatives, especially theh¹isomer. In addition, the electrophilicity of the keto center is reduced, increasing the decarboxylation barrier by 3 kcal mol¹ forL¹ (cf. the lowering by 3 kcal mol¹ for L²). Isomer 6 a is also stabilized by introducing an electron-withdrawing group at position R²of the ligand system (L⁴). The increased electrophilicity of the Cu center disfavors the h¹ isomer. By contrast, electron donation at this position (L³) provides more

electron density at the Cu center and thus facilitates the oxygenation step. Finally, electronic modification at position R¹shows no pronounced effects on the energetics of the reaction mechanism, as might be expected given the degree to which it is remote from the Cu.

Thus, the decarboxylation step is most influenced by substituents at R³, whereas modifications at R² mainly affect the oxygenation step. The experimental observation that electron donors at position R¹ slightly improve the overall product yield cannot be rationalized based on these calculations and are thus assumed to derive from other, nonelec- tronic reasons (as the experiments are far from achieving 100 % mass balance, we do not speculate further on this point). The steps following decarboxylation are not particularly sensitive to ligand substitution, except that compared with the other ligands the electron-withdrawing nitro group in L⁴ destabilizes all species relative to peracid 9 by 3 to 6 kcal mol¹, which is consistent with9having the lowest oxidation state (Cu^I) of all of these species.

Influence of changes to the N-donor ligand structure: One might expect that broader variations of electronic and steric effects could be achieved with N-donor ligands with back- bones fundamentally different from that in L⁰ toL⁶. To investigate this point, we chose a number of typical N-donor ligands (L⁷–L¹², Figure 3) with which to model the corresponding reactions of Scheme 5 (see Table 2).

In contrast to L⁰ (and its substituted variants), none of these ligands possesses a phenyl ring that can be hydroxylated by the oxo intermediate, so we consider only those steps Scheme 4. Alternative mechanism for the hydroxylation of the phenyl

ring of the N-donor ligand. The N-donor ligand is only depicted schemat- ically, and all values are in kcal mol¹.

Figure 3. N-donor ligandsL⁷–L¹²used to investigate ligand effects on the oxygenation reaction of Scheme 3.

Scheme 5. Model reactions to study the CH activation reactivity and the strength of the CuO bond (see also Table 4). The energetically lowest copper oxo isomer was used for each respective ligand.

FULL PAPER

Cu^II–Oxyl and Cu^III–Oxo Species

(6)

up to the generation and possible isomerization of that reactive intermediate. Indeed, we are particularly interested in the degree to which we might be able to design a ligand or ligands for which oxo intermediates would be the lowest- energy points along the reaction coordinates because such species might then be accessible experimentally. There is also the interesting question of whether the decarboxylation reaction proceeds via a peracid intermediate in all cases, or whether OO bond cleavage might occur concomitantly with loss of CO2.

ComparingL⁷withL⁰, the isomer energies for the various pre- and post-O2 adducts are similar, as might be expected given the similar electronic nature of the two ligands (featuring two sp² N donors). The other such N-donor ligand, L¹⁰, features less distinction betweenh²isomers 6 a and 6 b and, despite several attempts, an h¹isomer could not be located. With increasing donor strength of the supporting ligand (L⁸,L⁹, L¹¹, andL¹²),h¹isomer6 cbecomes more favorable and the different isomers are almost isoenergetic.

Apparently the electronic demand of the Cu center is effectively saturated by the two (or three) N donors and one O donor in these cases. ForL¹¹andL¹², the donor strength of the third N center seems to be just as good as that of a second O donor. As indicated in Table 1, not all isomers could be found for all ligand systems. This might at least partly be due to steric hindrance because the N- and O- donors are similar in strength.

The energetic differences between the three oxygen adduct isomers 7 a–care generally also quite small. For the less donatingL⁷andL¹⁰, the side-on isomer7 cis somewhat less competitive because the more electrophilic Cu center is less readily oxidized to a more Cu^III–peroxo-like side-on isomer. In contrast, the stronger donor ligands L⁸ and L⁹ tend to favor the side-on isomer. From an electronic point of view, the same might be true forL¹¹andL¹², but in these cases steric hindrance prohibits the formation of a side-on adduct. For all ligands, the end-on isomers 7 a and 7 b are

ground-state triplets. Despite various attempts, no triplet side-on minimum could be found in the cases ofL⁷andL¹⁰. Strong N-donor ligandL⁹seems to induce a more Cu^III-like character in the side-on isomer, which consequently possesses a singlet ground state.^[12]

With respect to the remaining steps in the oxo-generating mechanism, ligand L⁷ shows comparable energetics to the parent ligand L⁰, at least until the [CuO]⁺ isomerization. In contrast with all other ligand systems, singlet peracid struc- ture9shows a very slight^[15]restricted-to-unrestricted singlet instability in this case. This appears to render it nonstation- ary, as no OO cleavage TS (12) could be located on the singlet surface for this case; a corresponding triplet TS structure was found, leading directly to formation of the square-planar [CuO]⁺ isomer15. UnlikeL⁰, the latter is isoenergetic to the trigonal-bipyramidal variant 13, which might be due to lesser steric hindrance in the simple bis-

ACHTUNGTRENNUNG

(imine) ligand L⁷ (compared to the bulky L⁰), especially with regard to the N-Cu-N plane.

With more donating ligand L⁸, the oxygenation step becomes exergonic, in agreement with the trends observed for the substituted versions of L⁰. Decarboxylation via TS 8 occurs on the singlet surface and yields peracid9(the triplet TS structure is 1.9 kcal mol¹ higher in energy and yields oxo species 15directly). Generation of the [CuO]⁺ isomers is more favorable than with L⁷, and trigonal-bipyramidal isomer13is slightly preferred over square-planar form15.

Most of the trends observed when comparingL⁸withL⁷ are quantitatively enhanced for the still more strongly donating ligand L⁹, for example, exergonicity of oxygenation, preference for a singlet decarboxylation TS structure, and stability of the oxo species. In this case, however, the energetically less favorable (1.4 kcal mol¹) triplet decarboxylation TS structure leads to the trigonal-bipyramidal oxo species 13. The energetic preference for 13 relative to 15 is markedly enhanced with L⁹ compared with L⁸, probably owing primarily to the increase in ligand bite angle.

Table 2. Free energies [kcal mol¹] relative to the lowest isomer of6 a–cand O2(left side) or peracid9(right side). Selected free energy changes in pa- rentheses (see also Scheme 3). For species8–16, ground-state singlets are indicated by values initalics(starting complexes6are generally singlets, for the multiplicity of7 a–cfor various ligands see Table 2).

6^[a]+O2 7^[a] 8(DG^°₂) 9+CO2 9 10 11 12 13 14(DG^°₄) 15(DG₄) 16(DG^°₅)

L⁰ 0.0 0.5 23.3 (22.8) 41.8 0.0 17.8 19.9 12.4 11.3 4.5 (6.8) 6.9 (4.5) 1.2 (5.6)

L¹ 0.0 0.6 26.6 (26.0) 40.1 0.0 18.7 24.1 13.4 11.2 3.7 (7.5) 5.9 (5.3) 0.3 (5.6)

L² 0.0 1.3 21.2 (19.9) 42.3 0.0 17.2 26.5 16.4^[c] 11.5 5.3 (6.2) 8.1 (3.4) 2.7 (5.4)

L³ 0.0 0.8 21.9 (22.7) 41.7 0.0 17.4 26.4 12.8 12.3 6.7 (5.6) 8.9 (3.4) 3.1 (5.8)

L⁴ 0.0 2.2 25.1 (22.9) 43.5 0.0 20.4 17.4 13.8 8.3 0.7 (7.6) 2.8 (5.5) 2.3 (5.1)

L⁵ 0.0 0.3 23.2 (22.9) 42.8 0.0 18.0 21.4 15.7^[c] 11.5 3.6 (7.9) 5.9 (5.6) 0.6 (5.3)

L⁶ 0.0 0.1 23.9 (24.0) 42.7 0.0 18.2 25.3 16.2^[c] 11.7 3.3 (8.4) 5.3 (6.4) 1.9 (3.4)

L⁷ 0.0 1.1 19.3 (18.2) 45.3 0.0 – – 12.0^[c] 7.7 6.0 (1.7) 7.6 (0.1) –

L⁸ 0.0 2.1 12.7^[d](14.8) 44.3 0.0 – – 7.0 15.8 13.9 (1.9) 14.9 (0.9) –

L⁹ 0.0 11.3 8.2^[e](19.5) 43.7 0.0 – – 8.1 25.4 16.8 (8.6) 17.6 (7.8) –

L¹⁰ 0.0 4.1 13.6^[f](17.6) 44.4^[f] 0.0 – – ^[f] 12.4 11.5 (0.9) 13.6 (1.2) –

L¹¹ 0.0 8.9 6.9(15.8) 41.6 0.0 – – 0.9 21.4 ^[b] 17.1 (4.4) –

L¹² 0.0 6.0 7.6 (13.6) 48.1^[g] 0.0 – – 34.8^[c] 15.2 ^[b] 8.9 (6.3) –

[a] Lowest-energy isomer. [b] No TS structure could be located. [c] No TS structure could be located on the singlet surface; energy refers to triplet TS structure leading directly to15. [d] Analogous higher-energy TS structure on triplet surface leads directly to15. [e] Analogous higher-energy TS structure on triplet surface leads directly to13. [f] The lowest-energy decarboxylation TS structure is found on the singlet surface and leads directly to15; only a triplet peracid intermediate is stationary. [g] Analogous lower-energy TS structure on triplet surface leads directly to15.

4890

(7)

Although from an electronic point of view ligand L¹⁰ might seem comparable to ligands L⁰ and L⁷ (two sp² N- donor atoms), its influence on the oxygenation mechanism is different from the other two ligands in several respects.

The oxygenation step itself is more exergonic than with the bisACHTUNGTRENNUNG(sp³) N-donor ligandL⁸. Decarboxylation occurs via a singlet TS structure, but leads directly to the square-planar [CuO]⁺ intermediate 15without proceeding via peracid in- termediate9. In this case, it is the energetically unfavorable (1.1 kcal mol¹) tripletTS structure that yields a peracid intermediate. Ligand L¹⁰ also shifts the preference between the [CuO]⁺ intermediates back towards square-planar var- iant15, and indeed it is the only ligand for which this geometry is lower in energy than isomeric13. It is evident from comparison of these ligands that those regions of the singlet and triplet potential energy surfaces in the vicinity of their respective decarboxylation TS structure(s) must be fairly flat between competing reaction channels.

LigandsL¹¹andL¹²differ from those discussed above in- sofar as they are tridentate N-donor ligands. However, their oxygenation mechanisms resemble reasonably closely those of the strong N-donor ligandsL⁸andL⁹. ForL¹¹, we considered whether decoordination of one pyridine N donor might reduce the energy of the decarboxylation TS, but the resulting TS structure is about 7 kcal mol¹higher in energy than the fully coordinated alternative. In the case ofL¹¹, the peracid itself leads very readily (0.9 kcal mol¹activation barrier) to13, suggesting that the tridentate ligand is very effec- tive in stabilizing the CuÎI/CuÎIIoxidation state in this product relative to the CuÎ peracid educt. However, in the case of L¹², the decarboxylation on the triplet surface leads directly to the formation of15whereas the higher-energy decarboxylation TS structure on the singlet surface produces the CuÎ peracid educt. The TS structure from the CuÎ peracid educt to 13could not be located on the singlet surface and the barrier on the triplet surface is relatively high (34.8 kcal mol¹).

In summary, details of the oxygenation of the copper complexes supported by the various N-donor ligands are only modestly sensitive to ligand structure. More interesting- ly, [CuO]⁺ intermediates13and15are predicted to be energetic minima in all cases and should thus be accessible for a variety of N-donor ligands. With the exception of one ligand (L¹⁰), the peracid structure9always occurs as an intermediate product of the decarboxylation step. LigandsL⁷toL⁹il- lustrate the degree to which more electron-rich Cu centers facilitate initial oxygenation. In addition, the decarboxylation step becomes less exergonic and [CuO]⁺ generation becomes more favorable. These trends are consistent with the change of oxidation state of the central Cu atom, from CuÎI/ CuÎII in the oxygen adducts to CuÎ in the peracid structure and back to CuÎI/CuÎIIin the oxo species (see below). In the absence of significant steric constraints in the ligand, trigonal-bipyramidal oxo isomer 13 is preferred over square- planar isomer15.

Electronic structure of [CuO]⁺ intermediates: Isomers 13 and 15may be described by two alternative, limiting, electronic valence bond structures, namely, a (singlet) Cu^III–oxo structure or a (triplet) Cu^II–oxyl species. As has been reported previously^{[1d, 5]} for related systems, we predict the latter variant to be the more accurate for [CuO]⁺ complexes.

Apart from the triplet ground state of 13and15(for all ligands investigated), this is also evident from orbital analyses and natural bond orbital (NBO)^[16]calculations. For instance, in the case ofL⁰one SOMO (singly occupied molecular orbital) is identical to an O p orbital, whereas the other SOMO is mainly located on a Cu 3d orbital (mixed with ligand p orbitals). An NBO analysis confirms this view by localizing approximately 9 d electrons on Cu and assigning five electrons as lone-pair types on O. NBO analysis thus considers the CuO bond to have only ascomponent, the covalency of which may be quantified by its relatively low Wiberg bond order of 0.4. Similar bond orders were obtained for ligand systemsL⁷toL¹², indicating a consistently weak CuO bond. Within the series of ligands investigated, only slight variations in several key parameters of the CuO core were observed (see Table 3).

Although the differences are small, they do exhibit a trend, with more electron-donating ligands resulting in 1) longer CuO bonds, 2) smaller singlet–triplet splittings (from increased stabilization of the Cu^III–oxo singlet), 3) less spin localization on O (but more on Cu), and 4) more electronic charge on O.

To investigate the relevance of these trends to the chemical behavior of the various oxo species, the model reactions depicted in Scheme 5 were considered. For the CH abstraction reaction (Scheme 5, top), the three substrates 1,4-cyclo- hexadiene (DG^diss_CH=305 kcal mol¹),^[17] acetonitrile (DG^diss_CH=393 kcal mol¹),^[18] and methane (DG^diss_CH= 439 kcal mol¹)^[18] with the CH bond strengths indicated were used.

The activation and reaction free energies presented in Table 4 indicate that [CuO]⁺ complexes with more electron- donating ligands (L⁹, L¹¹) are less reactive in CH abstraction reactions compared with complexes with less donating ligands (L⁰,L¹⁰). Based on these calculations, isomer15with Table 3. Bond lengths [], energies [kcal mol¹], charge [a.u.], and spin [a.u.] for trigonal-bipyramidal species13with different N-donor ligands.

Ligand d (CuO) S–T^[a] Charge on Cu^[b]

Charge on O^[b]

Spin on Cu^[c]

Spin on O^[c]

L⁰ 1.806 7.2 1.23 0.63 0.53 1.23

L⁷ 1.802 5.9 1.23 0.62 0.53 1.22

L⁸ 1.806 6.2 1.25 0.65 0.55 1.19

L⁹ 1.811 6.3 1.26 0.67 0.56 1.17

L¹⁰ 1.800 8.5 1.22 0.62 0.54 1.23

L¹¹ 1.804 6.9 1.27 0.65 0.58 1.21

L¹² 1.809 1.6 1.25 0.66 0.52 1.18

[a] Singlet–triplet state energy differences. [b] NBO charges. [c] Mulliken spin densities.

FULL PAPER

(8)

weak N-donor ligandL¹⁰is the most reactive, leading to an almost ergoneutral reaction with methane.

These findings are consistent with the trends evident in Table 3, in which more electron-donating ligands lead to slightly less O-radical character in the [CuO]⁺ species13in favor of more basic Cu^III-oxo character. Overall, of course, the differences between the various ligands are rather small, with all [CuO]⁺ species being very reactive, which highlights the experimental challenges that will be associated with isolation of a stable supported [CuO]⁺ moiety.

The weak nature of the CuO bond in 13, as evidenced from the NBO analyses, is reflected in the very exergonic reaction free energies predicted for oxygen atom transfer to trimethylphosphine (Table 4, last row). Following the rea- soning already presented above, more electron-donating ligands will lead to increased bond orders associated with Cu^III–oxo valence-bond character and reduce the transfer free energy. Such behavior is indeed observed, for example, when comparing theDGOvalues forL⁷ toL⁹, in which the former complex is more reactive by 15 kcal mol¹.

It is evident that stable [CuO]⁺ functionality will be diffi- cult to achieve without substantial isolation of the species from both internal and external reactive partners. Further experimental investigations in this direction are underway.

Spin-state energies as a function of theoretical protocol:

With the exception of starting complexes 6 a–cand peracid structures 9, the Kohn–Sham (KS) determinants for all other stationary points are characterized by restricted to unrestricted instabilities at the DFT level. Expectation values of the total spin operatorS² over unrestricted broken-symmetry (BS) singlet solutions in most instances were near 1.0, suggesting roughly equal weights of singlet and triplet spin states in the determinant (high-spin triplet KS determinants showed negligible spin contamination in all instances).

As described in more detail in the computational methods section, we considered three different approaches to estimate the energies of singlet species compared with their triplet analogues, assuming the latter to be well described by high-spin KS DFT calculations. In the first model, we simply took the BS energy to be the singlet-state energy, ignoring the value of <S²>. In the second model, we used the sum method to purify the BS energy of its triplet component.^[19]

In the third and last model, we computed the difference in

energy between the singlet and triplet spin states at the CASSCF/CASPT2 level^[20] and added this to the high-spin triplet energy of the optimized BS DFT geometry. This model effectively assumes that 1) the BS DFT geometry is a good approximation to the true singlet geometry, 2) the high-spin triplet energy of this geometry relative to the optimized triplet minimum will be accurately determined at the DFT level, and 3) the multiconfigurational CASSCF/

CASPT2 model will predict singlet–triplet state energy splittings with similar accuracy.

As an example, we may consider molecular oxygen as a test case. The diatomic O2 has a triplet ground state and a

1D_gstate 22.6 kcal mol¹higher in energy.^[21]As expected, BS DFT calculations for the singlet state lead to a KS determinant with <S²>=1. By using protocols 1–3 outlined above, and a full-valence active space for the CASSCF/CASPT2 model, we predict singlet–triplet splittings of 14.9, 29.1, and 24.0 kcal mol¹, respectively, for O2. It is evident that the hybrid DFT-CASSCF/CASPT2 approach provides the best results in comparison with experimental data. This test rep- resents a worst-case scenario for the KS DFT model because the proper ¹Dgwave function requires a minimum of two equally weighted determinants. Thus, while spin purification improves upon the uncorrect BS prediction, the error still remains relatively large. In the discussion above, we refer exclusively to singlet energies computed according to protocol 3, but it is of technical interest to examine the per- formance of the three different protocols for the various structures that were surveyed, to assess their relative utility in different circumstances.

Table 5 provides singlet–triplet electronic energy splittings at the uncorrected BS DFT (DEuncorr), spin-purified (DEcorr), and CASSCF/CASPT2 (DE_CAS) levels for all BS DFT structures showing spin contamination on the singlet surface with ligands L⁰andL⁷toL¹². Table 5 also lists the high-spin triplet DFT energy difference between the optimized triplet and BS singlet geometries (DET), in this case including thermal contributions computed at the DFT level for the respective optimized geometries/states. Thus, our best estimate of the

“proper” singlet–triplet splitting can be obtained by sum- mingDE_TandDE_CAS.

It is apparent that the DFT predictions vary most from the CASSCF/CASPT2 predictions for the oxygen adducts 7 a–c. For all structures 7, the mean unsigned error (MUE) between uncorrected DFT and CASSCF/CASPT2 is 4.8 kcal mol¹ and the MUE for purified DFT is 9.5 kcal mol¹. Evidently, the biradical character of these complexes is significant and thus similar to molecular oxygen to the extent that large DFT protocol errors are observed. This suggests a formal Cu^II–superoxide character with substantially separated spins that lead to small splittings predicted at the CASSCF/CASPT2 level but significant errors at the DFT level owing to the very multideterminan- tal character of the singlet state. For8and the following reaction intermediates, on the other hand, the three protocols provide more similar estimates, especially after cleavage of the OO bond. In particular, the MUE between uncorrect- Table 4. Activation and reaction free energies [kcal mol¹] for the reac-

tions shown in Scheme 5.

R=

DG^°_CH DGCH DG^°_CH DGCH DG^°_CH DGCH DGO

L⁰ 5.7 30.5 8.8 8.4 17.3 2.6 77.2

L⁷ 7.5 30.0 9.8 8.0 17.7 3.0 77.7

L⁸ 7.1 30.5 9.9 8.4 17.8 2.6 70.5

L⁹ 7.1 29.6 9.2 7.5 17.6 3.5 62.8

L¹⁰ 5.0 32.4 8.4 10.4 15.3 0.6 74.9

L¹¹ 8.3 27.6 9.4 5.5 19.5 5.5 67.5

L¹² 7.0 30.2 10.9 8.2 18.3 2.8 72.2

4892

(9)

ed DFT and CASSCF/CASPT2 is 2.5 kcal mol¹ and the MUE for purified DFT is 3.7 kcal mol¹. These values are within a conservative error estimate on the CASSCF/

CASPT2 predictions. Thus, we conclude that, for reaction intermediates other than7, DFT calculations either with or without spin purification may provide reasonable estimates of singlet–triplet energy separations. Because of the greater rigor of the CASSCF/CASPT2 model, it is certainly to be preferred in those instances where it may be conveniently applied. However, the DFT approach may be a useful approximation in those cases in which application of the CASSCF/CASPT2 model is not straightforward, for example, because of uncertainties in active space construction or size limitations. Moreover, comparison between the multiconfigurational and DFT models can be particularly helpful for identifying erroneous predictions that might otherwise not be obvious; this proved true in several instances during the course of the present investigation.

Conclusion

We have detailed mechanisms for the oxygenation and subsequent reaction of Cu^I complexes with a-ketocarboxylate ligands that are based on a combination of density functional theory and multireference second-order perturbation theory (CASSCF/CASPT2) calculations. Our findings suggest that the reaction proceeds in a fashion largely analogous to that of similar Fe^II–a-ketocarboxylate systems, that

is, by initial attack of coordinated oxygen on the ketocarboxylate ligand and concomitant decarboxylation. Subse- quently, two reactive intermediates can be generated, peracid structure 9 and [CuO]⁺ species 15(and its isomer 13), both of which are capable of oxidizing a phenyl moiety at- tached to the ligand backbone. Hydroxylation by the [CuO]⁺ species is predicted to be energetically more favorable.

The effect of electronic and steric variations on the oxygenation mechanism were studied by introducing substituents at several positions of the ligand backbone and by investigating several other N-donor ligands. In general, more electron donation by the N-donor ligand leads to stabilization of the more CuÎI/CuÎII-like intermediates (oxygen adducts, [CuO]⁺ species) relative to the more CuÎ-like peracid intermediate.

For all ligands investigated, the [CuO]⁺ intermediates are best described as Cu^II–oxyl species with triplet ground states. The reactivity of these compounds in CH abstraction reactions decreases with more electron-donating N- donor ligands. The latter also increase the CuO bond strength, although the CuO bond is generally predicted to be very weak (with a bond order of about 0.5). A comparison of several methods to obtain singlet energies for the reaction intermediates indicates that multireference second- order perturbation theory methods are more accurate than DFT for initial oxygen adducts7 a–c, but not necessarily for the further reaction intermediates.

In short, our results suggest that the generation of [CuO]⁺ intermediates like 13and 15 should be feasible for Cu^I–a- ketocarboxylate complexes with a variety of N-donor ligands. These species are, however, predicted to be very reactive, posing some challenges to their experimental isolation.

Computational Methods

All geometries were fully optimized at the M06L level of density functional theory^[22]by using the Stuttgart pseudopotential basis set on Cu^[23]

(including three f functions having exponents of 5.10, 1.275, and 0.32) and the 6-31G(d) basis set^[24]on all other atoms. All singlet geometries other than those for the peracid complex and the starting complexes were obtained by using broken-spin-symmetry calculations; the peracid was predicted to have a stable restricted KS wave function. The nature of all stationary points was verified by analytic computation of vibrational frequencies, which were also used for the computation of molecular parti- tion functions (with all frequencies below 50 cm¹ replaced by 50 cm¹ when computing free energies) and for determining the reactants and products associated with each TS structure (by following the normal modes associated with imaginary frequencies). To correct for biradical character in some of the singlet structures, multireference second-order perturbation theory (CASSCF/CASPT2)^[21]computations were also per- formed. Singlet energies were computed by taking M06L triplet energies and adjusting them by singlet–triplet splittings computed at the CASSCF/

CASPT2 level for structures otherwise identical to broken-symmetry singlets computed at the M06L level (also referred to as protocol 3 in the Discussion section). The CASSCF/CASPT2 method has proven to be successful in the study of many analogous transition-metal problems.^[25]

Table 5. Singlet–triplet energy splittings [kcal mol¹] from different theoretical models.

7 a 7 b 7 c 8 12 13 14 15

L⁰ DET 1.8 1.1 7.1 1.3 10.4 1.2 1.2 1.3 DE_CAS 3.2 2.2 4.1 0.2 7.1 6.0 5.8 4.8 DEcorr 21.4 21.5 5.2 0.0 2.8 8.4 8.9 7.7

DE_uncorr 10.6 10.7 3.7 0.0 2.0 4.2 4.4 3.9

L⁷ DE_T 3.1 2.0 n/a 2.2 n/a 0.9 2.1 2.0

DECAS 2.3 1.2 n/a 0.3 n/a 5.2 8.4 6.0 DE_corr 13.6 12.7 n/a 4.9 n/a 8.7 7.2 7.4

DEuncorr 6.7 6.3 n/a 3.1 n/a 4.4 3.6 3.7

L⁸ DET 4.3 0.0 3.8 1.1 n/a 0.2 1.6 1.8

DE_CAS 0.3 0.1 4.6 3.5 n/a 5.8 7.2 3.8 DEcorr 10.4 13.6 8.2 3.1 n/a 7.7 8.9 7.3

DE_uncorr 5.2 6.8 4.1 1.8 n/a 3.8 4.4 3.6

L⁹ DET 1.6 1.8 7.8 2.6 n/a 1.3 0.3 0.9

DECAS 0.4 0.7 10.3 4.3 n/a 4.5 4.0 5.8 DE_corr 18.0 11.7 2.1 4.8 n/a 7.4 15.9 9.6

DEuncorr 10.0 5.8 1.7 3.0 n/a 3.7 7.9 4.8

L¹⁰ DE_T 6.8 2.0 n/a 3.6 n/a 2.5 1.1 1.6 DECAS 1.1 8.0 n/a 4.7 n/a 5.8 6.8 6.8 DEcorr 2.0 11.6 n/a 6.7 n/a 6.2 9.9 8.7

DE_uncorr 1.0 5.8 n/a 4.8 n/a 3.1 4.9 4.3

L¹¹ DET 1.3 1.7 n/a 0.3 13.1 1.9 n/a 2.1 DE_CAS 2.1 2.5 n/a 14.4 9.7 5.2 n/a 2.4 DEcorr 11.8 12.0 n/a 2.1 9.2 7.1 n/a 7.4

DEuncorr 5.9 6.0 n/a 1.4 7.9 3.6 n/a 3.9

L¹² DE_T 1.7 5.6 n/a 2.5 n/a 2.6 n/a 1.5 DECAS 2.1 0.9 n/a 2.4 n/a 1.0 n/a 1.0 DEcorr 12.8 5.9 n/a 3.3 n/a 5.1 n/a 7.4

DE_uncorr 2.9 6.4 n/a 1.9 n/a 2.6 n/a 3.8