2.1 Origin and scope of “through-space” coupling

Since Roberts et al. reported in the sixties the observation of “long-range” spin–spin coupling constants apparently transmitted via spin polarization of nonbonded-electron pairs (named later “through-space” couplings, TS) many papers were published reporting a wide variety of magnetic coupling situations related to this phenomenon [1]. It was clearly established that organic compounds containing two fluorine atoms separated by four or more bonds, but spatially close, exhibit large FF nuclear spin coupling constants J(F,F) [2]. This phenomenon, best designated as nuclear spin–spin coupling via nonbonded interactions, has been intensively studied for over thirty years by Mallory and co-workers, who produced a useful combination of experimental and theoretical works [3]. These “through-space” couplings first identified from FF J-constants in small organic molecules, were shown to be a general phenomenon encountered for nuclei as different as ¹⁴N, ³¹P [4], ⁷⁷Se, ¹³C and even ¹H when involved in hydrogen bonding. The studies reviewed by Contreras et al. provided a more theoretical and predictive background for the observed phenomenon. These are based on modelling and calculation approaches, which take into account the four terms contributing to J(X,X′), namely the Fermi contact (FC), spin dipolar (SD), paramagnetic spin–orbit (PSO) and diamagnetic spin–orbit (DSO) interactions [5,6].

2.2 Geometrical approach and simplified modelling

The studies conducted by Mallory showed that “through-space” couplings result from overlap interactions between lone-pair orbitals on the two crowded atoms. For instance, in the particular constrained geometry of the compounds sketched in Fig. 1 (top), the C–F/C–F, or C–F/C–N bonds, are coplanar and approximately parallel. As a consequence, the nonbonding distances d F⋯F, or d F⋯N, are short. In this orientation, the two lone-pair orbitals experience a σ-directed overlap. As displayed in Fig. 1 (bottom), the overlap between these lone-pair orbitals is expected to afford an in-phase and out-of-phase combination. As both orbitals are occupied (two-orbitals, four-electron interaction) no stabilization is observed. However, it provides an adequate pathway for transmitting spin–spin information between the coupled nuclei. The magnitude of J(F,F) depends on the extent in which the two lone-pair orbitals interact as a result of their overlap.

This model, which initially was mainly qualitative [2,3a,3b], eventually allowed the so-called overlap interaction to be quantitatively estimated for a number of 1,8-difluoronaphthalene derivatives. This became possible once an exponential relationship between the ab initio calculated internuclear distances and the observed J(F,F) coupling was established. This work confirmed previous studies by Ernst and Ibrom, who came up with a similar valuable semi-quantitative relationship in a number of difluorocyclophanes [7].

2.3 General requirements for the occurrence of “through-space” nuclear spin–spin coupling

In the general approach, the nuclear spin–spin coupling constant J(X,X′) is expressed as the sum of three terms (Eq. (1)):

Empiric requirements for the detection of “through-space” coupling constants were detailed in studies dealing with FF [3a,3e,3g], NF [3d,3f] HF or CF [3b,3c], and occasionally with PX spin–spin couplings [3g,8a–d] (with X = P, Se, F). They can be expressed as follows: (i) chemically identical nuclei in interaction need to be anisochronous. In other words, the molecule must be in some way dissymmetric as regards the environment of the observed nuclei; (ii) the nuclei must be in close proximity in solution, as in Mallory's fluoroaromatic compounds (Fig. 1) and in related phosphinonaphthalenes, which both contain rigid backbones [4,9]. The same is also true for the P-containing polymetallic catenates reported by Rheingold and Fountain [10] as well as for some bis(ferrocenyl)diphosphanes [11]; (iii) finally, each of the interacting nuclei should usually bear a lone-pair for mutual overlap. However, recent findings contradict this last assertion (see below).

3.1 Ferrocenyl tri- and tetraphosphanes and their coordination complexes

A number of polyphosphanes with interesting structural and catalytic properties appear every year in the literature. Among them the ferrocenyl tetraphosphane (1,1′,2,2′-tetrakis(diphenylphosphino)-4,4′-di-tert-butylferrocene (Fc(P)₄tBu, Fig. 2 left)) was recently shown to promote various palladium-catalysed C–C cross-coupling reactions [12,13]. This was attributed to a conformationally blocked ferrocene backbone, which allows this ligand to bind the metal centre by three instead of only two phosphorus atoms, incidentally increasing the electron density on the metal. This behaviour contrasts with that of related ferrocenyl tetraphosphanes in which the Cp rings rotate freely (Fig. 2) [14].

The conformational rigidity of Fc(P)₄tBu was inferred from the corresponding ³¹P NMR spectrum which shows an AA‘BB’ system (Fig. 3). The occurrence of nonbonded nuclear spin couplings between phosphorus atoms was clear from the intense J(P,P) coupling constant (60 Hz) found between the heteroannular P atoms corresponding to A and A′ (see Fig. 2). From X-ray structure analysis a quasi eclipsed conformation is observed for these nuclei, with lone-pairs orientated towards each other (Fig. 3, internuclear distance determined in the solid state from X-ray measurements was d P⋯P′ = 3.728(2) Å). In keeping with Mallory's model, the phosphorus lone-pairs of Fc(P)₄tBu undergo a strong overlap which is responsible for the observed direct “through-space” spin–spin interaction.

In the course of NMR-mechanistic studies on transition metal complexes derived from the tetraphosphane ligand Fc(P)₄tBu [15], unusual features that can only be explained by the presence of TS spin couplings were detected in their ³¹P NMR spectra (See Fig. 4). In order to gain more in depth information on the nuclear spin–spin interactions in these coordinated tetraphosphanes, the palladium complexes [PdX₂{Fc(P)₄tBu}], [Pd₂X₄{Fc(P)₄tBu}] (X = Cl or Br), as well as their nickel analogs [NiX₂{Fc(P)₄tBu}] and [Ni₂X₄{Fc(P)₄tBu}] were synthesized and characterized by X-ray crystallography in the solid state and ³¹P NMR in solution [16]. The coordination complexes of Fc(P)₄tBu keep the conformational rigidity of the free ligand in solution. As a result, the same type of spin–spin coupling seen in the ligand was detected in the complexes. Data presented in Table 1 clearly establish a correlation between the P⋯P′ separations between non-adjacent phosphorus atoms (determined from crystallographic data) and the corresponding coupling constants.

The collected data allowed to correlate geometric features found in the solid state with J(P,P) coupling constants (solution spectra), so that an expression of J(P,P) vs. d P⋯P′ in coordination complexes could be devised (see details in Section 5).

TS spin couplings were also observed for other members of this family of ligands and related to their structure in solution. The ³¹P NMR pattern of 1,1′,2-tris(diphenylphosphino)-4-tert-butylferrocene (Fig. 5, right) displays a triplet at −19.3 ppm and a doublet at −23.0 ppm in CDCl₃, which were assigned to the 1′-PPh₂ and to the 1,2-PPh₂ groups, respectively (A₂B spin system). The observed spin–spin coupling constant was found to be larger in C₆D₆, J(P,P) = 5.0 Hz, than in CDCl₃ (J(P,P) = 2.8 Hz). The J(P,P) value of 5.0 Hz appears to be too high for a classical “through-bond” ⁴J(P,P) coupling constant (usually in ferrocenyl polyphosphane ⁴J(P,P) between anisochronous heteroannular phosphorus atoms is null); such a strong dependence on solvent is also unexpected for a “through-bond” coupling. Thus, only a “through-space” spin–spin coupling mediated by the ferrocenyl ring rotation and/or by changes in the relative orientation of the phosphino groups can explain such unusual values. To account for this TS coupling, the relative rotation of the cyclopentadienyl rings (which is not blocked in this case) has to be restricted to a certain degree around the conformation of higher symmetry (Cs) pictured in Fig. 5 [17], and possibly occurs rapidly to facilitate via space–spin interaction of phosphorus nuclei.

The interpretation of the NMR spectrum of another ferrocene derivative, namely triphosphane 1,1′,2-tris(diphenylphosphino)-3′-(tris-phenyl)methyl-4-tert-butyl ferrocene [18], also required a nonbonded coupling to be considered (see Fig. 6). The signals corresponding to the phosphorus atoms of the chelating pair, 1-P and 2-P, are centred at −18.8 and −24.9 ppm. They appear as doublet of doublets, due to a strong, but classical coupling constant of ³J(P,P) = 41 Hz, as well as to another coupling constant with the 1′-P phosphorus atom (in the following P′ designates a P atom belonging to the other Cp ring). The signal of the third phosphorus atom is a pseudo-triplet, thus reflecting nearly identical ^TSJ(P,P) coupling constants (^TSJ(P,P) = 11 and 12 Hz, respectively). These intense spin–spin couplings between heteroannular phosphorus atoms most probably operate “through-space” and therefore result from a spatial proximity of the spin-active nuclei. Consequently, in the absence of X-ray, we assigned to this triphosphane the conformation depicted in Fig. 6. This assumption was confirmed by the NMR spectrum of the coordination complex which readily formed with [Pd(allyl)Cl]₂. In this complex, all three phosphorus atoms are involved in binding, as deduced from the observed ³¹P chemical shifts (Fig. 7) [18]. Very interestingly, one of the phosphorus atoms (presumably 1′-P) is less strongly bonded, its signal appearing at −10 ppm, instead of an expected value near 25–30 ppm.

This example is the first in which NMR analysis allowed both to identify the existence of “through-space” J(P,P) coupling constants and, incidentally, to anticipate the tridentate coordination behaviour of a triphosphane [18]. In view of the ^TSJ(P,P) coupling constants inferred from the spectral analysis (10–12 Hz), we can reasonably assume that for each pair of phosphorus atoms the d P⋯P′ distance in solution is smaller than 4.3 Å.

3.2 Calixarene P(III)-podands

The ^TSJ(P,P) coupling constants observed in the aforementioned polyphosphanes and related to short spatial separations between phosphorus nuclei (d < 5.5 Å) are not restricted to ferrocene-derived P(III)-ligands. The successful preparation of a series of regioselectively phosphinated calix[4]arenes led Dieleman et al. to investigate the calixarenes 1–6 in which two proximal phenoxy rings are substituted by a –CH₂PPh₂ podand [19,20]. In these molecules the phosphorus atoms are separated by ten chemical bonds (Fig. 8). Calixarenes 2–6 are chiral, and were isolated either as optically pure compounds (2–4) or as a racemic mixture (5 + 6). Due to the difference between R¹ and R² groups the ³¹P NMR spectra of the latter display two distinct phosphorus signals. However, while the spectra of 2–4 are each characterized by clear singlets (with no J(P,P) couplings detected), the ³¹P NMR spectrum of the benzoyl-substituted enantiomers 5 and 6 displays a typical AB spin system with J(P,P) = 8 Hz, the corresponding doublets being centred at −20.1 ppm and −22.2 ppm, respectively [21].

The spin coupling existing between the two phosphorus atoms, separated by ten bonds, in 5/6 was evidenced by measuring the spectrum in CDCl₃ at different fields and also by carrying out a 2D homonuclear ³¹P COSY experiment. When the experiments were repeated in CD₃OD, the magnitude of the coupling did not significantly change. The observed coupling constant was intuitively assigned to a “through-space” interaction rather than to a “through-bond” coupling since the phosphorus atoms of 5/6 are linked through a non-conjugated bond system. It is worth noting, that in a highly conjugated naphthyl-derived diphosphane, a ⁹J(P,P) of barely 4.0 Hz has been reported [22]. The solid state structure of the enantiomeric mixture (5 + 6) was determined by an X-ray diffraction study (Fig. 9).

In the solid state, the separation between the two phosphorus atoms was found to be of 5.333(1) Å, with an approximate alignment of the two phosphorus lone-pairs. It appears likely that in solution the P⋯P′ separation becomes even temporarily shorter. As revealed by a 2D ROESY experiments, the calixarene core of 6 displays a high flexibility, the unsubstituted phenolic ring flipping rapidly through the calixarene annulus. This fluxional behaviour combined with the mobility of the –OCH₂PPh₂ arms increases the probability of interaction of the phosphorus lone-pairs. The J(P,P) coupling constant value increases upon lowering the temperature (from 8.0 Hz at 283 K to 10.5 Hz at 183 K), an observation that is consistent with a statistically longer period of overlap of the lone-pairs as a result of slower molecular dynamics. The spatial proximity in solution of the phosphorus atoms of 6 was further substantiated by the easy formation of cis-chelate coordination complexes. Finally, the “through-space” J(P,P) coupling constant observed for the 5/6 couple was rather unexpected, and to date, there is no other example of chelating diphosphane with a ten bond P⋯P′ separation for which such an intense spin coupling is directly observable.

4.1 ³¹P¹³C spin couplings in ferrocenyl polyphosphanes

The proximity of the two cyclopentadienyl rings in the ferrocene unit, which is in part responsible for the PP “through-space” couplings discussed in Section 3, also gives rise to P⋯C spin couplings between atoms connected to distinct Cp rings. The ferrocenyl triphosphane shown in Fig. 11 provides a remarkable example of ligand in which such a coupling occurs [24].

The ¹³C NMR spectroscopic data revealed a neat PC spin–spin nuclear coupling (J(P,C) = 5.5 Hz) between the methyl carbon atoms of the tBu group (fast rotating methyl groups) and the PiPr₂ phosphorus atom, as unambiguously demonstrated by selective ³¹P decoupling experiments (see Fig. 12). This means that the corresponding phosphorus atom is located close to the tBu group, and accordingly no J(P,P) coupling constant is observed from the A₂X ³¹P NMR spectrum of this triphosphane (the lone-pair is involved in J(P,C) coupling). The five-bond separation between the P and C nuclei of concern cannot lead to a detectable ⁵J(P,C) if only “through-bond” spin coupling is operating. Further proof for a “through-space” coupling mechanism came from the absence of any other ⁴J(P,C) or ⁵J(P,C) coupling constant. A single crystal X-ray diffraction study confirmed the spatial proximity of the tBu and PiPr₂ groups (see Fig. 11 left; shortest P⋯C separation: 3.64 Å). The staggered conformation imposed by the ferrocene backbone favours the observed P,C interaction. The P lone-pair is directed towards the carbon atoms of the tBu group, thus supporting the existence of an overlap between this orbital and a filled Csp₃ orbital (or with the C–C bond of tBu group since no J(P,H) is observed).

4.2 A trans-spanning cyclodextrin-diphosphane

Nonbonded PC spin–spin couplings are naturally not limited to ferrocenyl phosphanes. Such a phenomenon was also revealed in the rigid, C₂-symmetrical cyclodextrin-derived TRANSDIP (Fig. 13). Here no PP spin--spin coupling can be simply detected in the ³¹P NMR spectrum, due to the magnetic equivalence of phosphorus nuclei [25]. However, the C-4, C-5 and C-6 atoms of the P-bridged A, B, D, E glucose units display spin--spin couplings with both phosphorus atoms in the ¹³C NMR spectrum, an observation that can only be rationalized by the existence of P,P orbital overlap. Moreover, a neat spin–spin coupling J(P,C) = 3.5 Hz between the C-6^C atom (equivalent to C-6^F) and both phosphorus atoms was also detected (see Fig. 14), thus indicating that both C-6 atoms lie in close proximity to the P atoms. The triplet observed reflects the equivalence of the two phosphorus atoms regarding the carbon atoms, and also evidences an unexpected extent of the orbital overlap.

The “through-bond” contribution to PC spin-spin coupling is likely to be negligible given the eight-bond separation existing between the C₆ atoms of glucose units C and F (Fig. 13) and the two phosphorus atoms. Molecular mechanics calculations (MM2), as well as X-ray crystal structures of TRANSDIP metal complexes [26], confirmed that the C-6^C,F atoms are close to both phosphorus atoms (averaged P⋯C separation: 4.8 Å, see Fig. 15).

The J(P,C-6^C,F) coupling constant of 3.5 Hz is somewhat smaller than the “through-space” J(P,C) value (5.5 Hz) observed for the aforementioned ferrocenyl triphosphane, suggesting a longer P⋯C separation in the case of TRANSDIP. The fact that only the C-6^C,F carbon atoms of the cyclodextrin give rise to such a direct interaction is a further indication of the endo orientation of the P lone-pairs as well as of the ligand rigidity. Overall, the ¹³C{¹H} NMR spectrum gives valuable information on the structural features of this rigid diphosphane.