Synthesis and investigation of donor–porphyrin–acceptor triads with long-lived photo-induced charge-separate states

The powerful electron donor tetraalkylphenylenediamine (TAPD) facilitates photo-induced electron transfer, even in a frozen solvent at 10 K, generating a long-lived spin-polarized charge separate state which can be observed by EPR.


Introduction
Photo-induced intramolecular electron transfer can generate a charge-separated state (CSS) consisting of a hole and an electron with a spatial separation of 1-3 nm. 1a In most cases, the hole and electron recombine rapidly (from ps to ns) to regenerate the ground-state. However, in some cases, the CSS can have a longer lifetime 1 (from ms up to possibly hours 1h,i ), allowing chemical, physical or biological processes to exploit its high energy and unusual electronic structure.
Long-lived photo-excited CSSs are important for a variety of applications. They are studied to understand and mimic electron transfer in natural photosynthesis, in which energy from sunlight is converted into chemical potential. 2 In the area of quantum information processing, control of the spin dynamics of a CSS may allow the manipulation of a nuclear or electronic spin, to encode or transfer information. 3 It is also thought that some birds, such as the European robin, use the magnetic elddependence of the recombination rate of a CCS to orient themselves in the earth's magnetic eld. Mimicking this avian compass may enable small magnetic elds to be detected. 4 In the high-temperature limit, where solvent dynamics and nuclear motions can be treated as classical harmonic oscillators, recombination rates of CSSs are given by the Marcus equation 5 (eqn (1)), (1) where V DA is the donor/acceptor coupling matrix element, DG 0 is the free energy change, l is the global reorganization energy, T is the temperature, ħ is the reduced Planck constant and R is the ideal gas constant. In order to reduce the rate of backelectron transfers and create long-lived CSSs, chemists have attempted to reduce the V DA coupling term by increasing the distance between the photo-generated charges, by designing molecular dyads, triads and pentads. 6 Wasielewski et al. 7 designed a donor-porphyrin-acceptor triad TNQ-ZnP-TAPD ( Fig. 1) with three desirable features. First, the strong electron-donating and -accepting behavior of the tetraalkylphenylenediamine (TAPD) and triptycenenaphthoquinone (TNQ) moieties make charge separation favorable, even in a frozen solvent where the solvent cannot reorganize to stabilize the photo-generated zwitterion. Secondly, the produced electron/hole pair has no through p-bond electronic coupling (because the donor and the acceptor moieties are separated by isolating methylene bridges), and weak through-sbond coupling, due to the near-orthogonal porphyrin core.
Finally, the charges are rigidly separated by a distance of 2.3 nm, limiting through-space coupling. These last two features yield a very small V DA and consequently extend the lifetime of the CSS. TNQ-ZnP-TAPD showed a CSS with a lifetime of 4 ms, together with a spin-polarized radical-pair that closely mimics the bacteriochlorophyll cation-quinone anion pair found in photosynthetic reaction centers. 8 The initial objective of the work presented here was to synthesize porphyrin-based triads exhibiting long-lived CSSs, such as TNQ-ZnP-TAPD, so that they could be used for quantum information storage experiments. Since we also wanted to modulate the properties of our triads, we sought a versatile synthetic route with few steps from accessible precursors, which would tolerate a wide range of functional groups.
In order to avoid aggregation and enhance the solubility of 5,15-diarylporphyrins, two positions are available to introduce solubilizing groups (Scheme 1). Porphyrins can be substituted on the b-pyrrole positions (R 1 and R 2 on Scheme 1), generally with aliphatic chains, providing a locked 80-90 dihedral angle between the porphyrin and the aryls groups. However, bsubstituted pyrroles are less readily available than pyrrole, and the steric hindrance at the 5 and 15 meso-positions prevents cross-coupling strategies from being used to introduce the donor and acceptor moieties.
Another well-established strategy to avoid aggregation and enhance solubility in 5,15-diarylporphyrins is to introduce bulky aryl groups in the two remaining 10 and 20 meso-positions (R 3 ¼ Ar; R 1 ¼ R 2 ¼ H on Scheme 1). This substitution pattern avoids steric hindrance around the 5 and 15 positions.
TAPD derivatives, also known as Würster blue, 9 are highly electron-rich (oxidation potential: À0.24 V vs. Fc/Fc + , see later) Scheme 1 Two synthetic routes towards 5-acceptor- 15donorporphyrins. and are therefore promising electron donors. On the other hand, their low oxidation potentials makes them reactive towards oxygen and other oxidants, such as those used in porphyrin synthesis.
One retro-synthetic route to 5-acceptor-15-donor disubstituted porphyrin triads is the statistical condensation of a dipyrromethane with two aldehydes substituted with the donor and acceptor moieties (Scheme 1, top, Route A), and subsequent oxidation of the porphyrinogen, typically with DDQ or chloranil. This route is incompatible with the use of oxidation-sensitive donors such as TAPD. Indeed, isoindolines are known to be oxidized to isoindoles, which then react further via dimerization or Diels-Alder reactions. 10 Therefore the synthesis of TNQ-ZnP-TAPD required masking of the terminal dimethylamine as a nitro-group, resulting in a less convergent route (Scheme 2a). 7a Alternatively, a symmetrical 5,15-diaryl-10,20-dibromoporphyrin 11 can be synthesized, and functionalized with two different moieties via successive cross-coupling reactions (Scheme 1, bottom). Suzuki-Miyaura coupling is widely used as a mild, non-toxic, and efficient approach for the convergent synthesis of aromatic molecular materials, 12 including porphyrin derivatives. 13 In this case, Suzuki-Miyaura coupling allows not only an efficient synthesis of asymmetrical donor-porphyrin-acceptor triads, but also the introduction of the sensitive donor moiety in the very last step, as the mild conditions do not require the donor to be protected from oxidation.
Here we present a short and efficient synthesis of useful 2-(4dialkylaminophenyl)isoindoline electron-donating moieties (Scheme 2b), and their use in the Suzuki pathway as a convenient route to oxidation-sensitive acceptor-porphyrin-TAPD triads (Scheme 3). We also used this versatile approach to synthesize a triad with a C 60 acceptor moiety, which was predicted, and found, to have a longer-lived CSS than TNQ-ZnP-TAPD.

Synthesis
At the start of this project, we attempted to synthesize TNQ-ZnP-TAPD as reported by Wasielewski et al. (Scheme 1,Route A and Scheme 2a). 7 We developed an efficient three-step synthesis of aldehyde 3 (Scheme 2b): rst, 4-nitroaniline was condensed with 1,2,4-benzenetricarboxylic anhydride to yield 1. The imide and carboxylic acid functions were then simultaneously reduced with borane to give 2 and this benzyl alcohol was reoxidized using activated manganese dioxide to yield aldehyde 3, in 50% over 3 steps. However, in our hands, the condensation of aldehydes 3 and 4 with tetraalkyl-dipyrromethane 5 did not give the desired porphyrin. Therefore, we decided to explore Suzuki coupling routes to the closely related triad TNQ-ZnP Ar -TAPD (Scheme 1, Route B, and Schemes 3 and 4).
The boronic ester substituted donor moiety was synthesized in good yield using the phthalimide route developed for the synthesis of aldehyde 3: 4-(N,N-diethylamino)aniline was condensed with 4-bromophthalic anhydride in reuxing acetic acid to yield 6, which was then reduced to the isoindoline using borane in reuxing tetrahydrofuran to obtain 7. Its borylated equivalent 8 was subsequently obtained via palladium-catalyzed borylation in an overall 45% yield. The acceptor boronic ester 10 was synthesized (Scheme 3) from 2-bromoanthracene 9 (ref. 14) through Diels-Alder reaction with 1,4-naphthoquinone, 15 and subsequent palladium-catalyzed borylation 16 of the bromotriptycenequinone.
The donor and acceptor moieties were linked to the central porphyrin core via a two-step Suzuki cross-coupling. First, the reaction of 3 equivalents of dibromo-porphyrin 11 11 with 1 equivalent of the acceptor boronic ester 10 gave 12 in 69% yield, which was reacted in a second step with 1.1 equivalents of the donor boronic ester 8 to yield TNQ-ZnP Ar -TAPD in 72% yield. Performing the Suzuki couplings in two successive steps gave better yields and made purication easier than the simultaneous statistical one-pot coupling of both the donor and the acceptor to the porphyrin. To test the versatility of the Suzukibased route to porphyrin triads, we synthesized another triad, C 60 -ZnP Ar -TAPD (Scheme 4), using the second Suzuki route (Scheme 1, Route C). When we performed the Suzuki coupling of 13 (ref. 17) with 4-iodobenzaldehyde, substantial deborylation of the porphyrin was observed. Therefore, the reaction was carried out with an excess of 4-iodobenzaldehyde and stopped as soon as formation of the bis(p-benzaldehyde) porphyrin adduct was detected by thin layer chromatography (TLC), to obtain 14 in 35% yield. Using an excess of 8, in the second Suzuki coupling yielded 15 in 88%, which was then transformed into C 60 -ZnP Ar -TAPD using 2-((3,4-bis(dodecyloxy) benzyl)amino)acetic acid and C 60 in a fast Prato reaction. 18
Electrochemical analysis. The change in free energy associated with the rst electron transfer step, DG 1 , can be predicted, under solution-phase conditions, from the energy of the rst singlet excited state S 1 of the porphyrin, the oxidation potential of the porphyrin, E ox (P), and the reduction potential of the acceptor, E red (A), using the Rehm-Weller equation (eqn (2)), 19 where N A is the Avogadro constant, e is the elementary charge, 3 0 is the vacuum permittivity, 3 r is the dielectric constant of the solvent (8.9 for dichloromethane) and d 1 is the distance of charge separation in CSS 1 (1.0 nm for all three triads). Similarly, the energy changes for the second and third electron transfer processes can be calculated from eqn (3) and (4): where d 2 is the distance of charge separation in CSS 2 (2.3 nm in TNQ-X-TAPD and 2.4 nm in C 60 -ZnP-TAPD, from molecular mechanics calculations).
The redox potentials E ox (D) and E red (A) were measured using squarewave voltammetry, and E(S 1 ) values were estimated from absorption spectra (Table 1), using reference compounds 7 and those in Fig. 3 as models for the isolated donor, porphyrins and acceptor units. Electrochemical measurements on the complete triads gave very similar redox potentials to their isolated components, for example C 60 -ZnP Ar -TAPD shows oxidation waves at À0.29 V (TAPD) and 0.32 V (ZnP Ar ) and reduction waves at À0.95 V (C 60 ) and À1.84 V (ZnP Ar ). The values of DG for electron transfer for steps 1-3 ( Fig. 2) derived from the electrochemical potentials in Table 1 according to eqn (2)-(4) are listed in Table 2.
For the rst electron transfer in solution, DG 1 is negative for all triads (À1.01 eV for TNQ-ZnP-TAPD, À0.81 eV for TNQ-ZnP Ar -TAPD and À0.97 eV for C 60 -ZnP Ar -TAPD). The second electron transfer step is also exergonic for all the triads because E ox (TAPD) ( E ox (ZnP Ar ) < E ox (ZnP). The strong favorability of electron transfer is clear from the total free energies of electron transfer (DG 1 + DG 2 ). The aim of this project was to create triads that would give long-lived CSSs at low temperatures, in a frozen solvent glass, for EPR quantum information experiments. This makes the huge driving force for charge separation important, because it enables electron transfer to be favorable even at low Table 1 Electrochemical and optical measurements 2.17 2.12 --a Measured by squarewave voltammetry vs. Fc/Fc + in dichloromethane with 0.1 M NBu 4 PF 6 as electrolyte. b l max of the longest absorption band in dichloromethane. c E(S 1 ) ¼ hc/l max . Structures of reference compounds ZnTPP, ZnP 0 , C 0 60 and TNQ 0 are displayed in Fig. 3.

Fig. 3
Reference compounds used to measure redox potentials. Table 2 Experimentally and theoretically derived free energies for each electron-transfer processes for the three triads (eV) temperatures, under the conditions of a frozen solvent glass, when solvent dipoles cannot reorient in response to the new charge distribution in CSS 1 and CSS 2 , as discussed below. The third electron transfer, corresponding to the charge recombination, is also exergonic for all three triads because TAPD is not a strong enough electron donor to reduce the ground-state acceptors TNQ and C 60 . Molecular geometries. The b-alkyl substituents in TNQ-ZnP-TAPD enforce a strictly orthogonal conformation between the porphyrin and the aryl substituents linking the donor/acceptor moieties, while this torsion is less constrained in the X-ZnP Ar -TAPD compounds. 20 To estimate the resulting changes in dihedral angles between the porphyrin plane and the mesolinked benzenes plane, we performed a statistical analysis of meso-aryl zinc porphyrin crystal structures using the Cambridge Structural Database (CSD, ESI †). We analyzed 343 structures with b-alkyl substituents and 1032 structures without bsubstituents. For b-unsubstituted porphyrins, the distribution of dihedral angles is quite broad; the population density peaks at 68 and is greater than 50% of this peak value in the range 90 AE 28 . On the other hand, when there is a CH 2 at the b-position next to the meso-aryl group, the population density peaks at 90 and is greater than 50% of this peak value in the range 90 AE 4 .
The geometries of all three triads were calculated at the B3LYP/6-31G(d) level. Calculated dihedral angles between the porphyrin and the donor unit, as well as between the porphyrin and the acceptor moiety (Table 3) agree well with the distributions from our CSD analysis. The less orthogonal geometries in ZnP Ar , compared to ZnP, lead to stronger electronic coupling, as discussed below. The Boltzmann distribution of dihedral angles in TNQ-ZnP Ar -TAPD at the temperature relevant to our EPR studies (223 K, the freezing point of xylene) is presented in Fig. 4a.
Calculated energy levels. Frontier molecular orbital distributions for geometries optimized at the B3LYP/6-31G(d) level are shown in Fig. 5 (for HOMOÀ2 to LUMO+2, see ESI †). Despite the differences in the porphyrin-aryl dihedral angles, the shapes of the frontier orbitals are not signicantly affected and the HOMO and LUMO are strongly localized on the respective donor and acceptor moieties. The sp 3 carbons present in the porphyrin-donor and porphyrin-acceptor bridges act as insulators, regardless of the conformation, conrming that the design of TNQ-ZnP Ar -TAPD and C 60 -ZnP Ar -TAPD remains valid for stabilizing a long-lived CSS.
The calculated donor (TAPD) and acceptor (TNQ and C 60 ) orbital energies (HOMO and LUMO) show little variation between the three compounds (Table 4, see ESI † for other levels of theory). However, the tetraaryl porphyrin core ZnP Ar has a smaller gap (2.84 vs. 2.97 eV; HOMOÀ1 to LUMO+1) and is a somewhat weaker electron donor (lower HOMOÀ1) than alkyl ZnP (À4.92 vs. À4.79 eV), due to the lack of electron-donating alkyl groups. TD-DFT calculations conrm a 0.05/0.06 eV lower optical gap, with S 1 at 2.37/2.32 eV in ZnP Ar compared to 2.42/ 2.38 eV for alkyl ZnP at B3LYP/6-31G(d) and CAM-B3LYP/6-31G(d), respectively. Ionization potential calculations (IP, see below and ESI †) suggest a 0.05-0.20 eV difference in oxidation potential. This is in good agreement with the 0.15 eV difference in oxidation potential between the two ZnP and ZnP Ar porphyrins as measured by squarewave voltammetry, and the 0.05 eV difference in absorption energy from UV-visible spectroscopy ( Table 1).
The reaction free energies DG 1 , DG 2 and DG 3 were estimated for solution-phase ET using DFT. Since the singlet-triplet gap for long-range charge-transfer systems such as these is negligible, we modeled the CSSs as the lowest unrestricted DFT triplet. The effect of the solvent was included using the Integral Table 3 Angle between the mean plane of the benzene ring in donor and acceptor and the mean plane of the porphyrin a q (donor-porph) q (acceptor-porph) TNQ-ZnP-TAPD 82 86 TNQ-ZnP Ar -TAPD 70 68 C 60 -ZnP Ar -TAPD 68 71 a B3LYP/6-31G(d) equilibrium geometry.

Equation Formalism
Polarizable Continuum Model (IEFPCM) as implemented in Gaussian 09 with Universal Force Field (UFF) radii and default parameters. Energies calculated using the SMD model 21 were within 0.05 eV of these values. Calculations in liquid solution were carried out using 3 r , the static (or zerofrequency) dielectric constant of the solvent, which includes the effect of electronic and dipolar polarization. Given their small effect in the donor-acceptor charge transfer energies, the bis-3,5-tert-butylphenyl side groups in ZnP Ar were substituted by hydrogen atoms to reduce the number of nuclear and electronic degrees of freedom when the calculations did not involve states located on the porphyrin. The results of these calculations, using three different computational methods (CAM-B3LYP/M062X/B3LYP), are compared with DG values from electrochemical measurements in Table 2. The calculated free energies conrm that the intermediate CSS 1 is systematically shied up in energy by the structural modication in ZnP / ZnP Ar , making DG 1 less exergonic and DG 2 more exergonic.
Our UDFT calculations for the CSS 1 C 60 À -ZnP Ar + -TAPD did not converge to the diradical state, but rather to a localized triplet, preventing calculation of DG 1 and DG 2 . However for C 60 -ZnP Ar -TAPD, we can still calculate the total free energy change for electron transfer (DG 1 + DG 2 ), as shown in Table 2. All DFT functions tested incorrectly predicted TNQ to be a more powerful acceptor in solution than C 60 (between 0.15 to 0.05 eV, with both the 6-31G(d) and 6-311G(d,p)). In vacuum, however, this trend was reversed and the correct behavior was recovered, with a difference of 0.2-0.4 eV, as is the case when the effect of the frozen solvent is taken into account (see below). This discrepancy may be attributed to the failure of continuum solvent models to treat cavitation-dispersion interactions. Electron transfer in frozen solvents. Below the freezing point, the solvent molecules cannot reorient their dipoles in response to local changes in charge of the solute. This lack of dipolar polarization is equivalent to the outer-sphere reorganization energy for electron-transfer reactions in solution, where the reorientation of the solvent molecules is much slower than the response time of their electron clouds. We modeled this lack of dipolar polarization by performing non-equilibrium IEFPCM calculation. When including the frozen solvent effect in butyronitrile (optical dielectric constant: 3 N ¼ 1.9) we obtained the free energies for electron transfer in the frozen solvent glass shown in Table 1. In TNQ-ZnP-TAPD and TNQ-ZnP Ar -TAPD, the rst and second charge separated states, CSS 1 and CSS 2 , are shied up in energy by about 0.75 and 1.32 eV, respectively, which makes charge-separation scarcely favorable. If we compare this increase in the energy of CSS 2 (1.32 eV) with the experimental values of DG 1 + DG 2 in TNQ-ZnP-TAPD and TNQ-ZnP Ar -TAPD (À1.33 and À1.28 eV, respectively), it is evident that there is almost no driving force for charge separation in the frozen state. In C 60 -ZnP Ar -TAPD, the energy of CSS 2 is increased by about 1.23 eV, which is slightly less than in the other triads, and the experimental value of DG 1 + DG 2 in is slight more negative (À1.44 eV) so charge-separation is expected to be more exergonic.
These calculations were carried out for butyronitrile as the solvent, for consistency with earlier studies by Wasielewski and co-workers, 7 however other solvents have quite similar optical dielectric constants (3 N z n 2 , where n is the refractive index) so that the free energies changes are expected to be similar in other frozen solvents.

Calculated rates of charge recombination
Rates of electron transfer are governed by a combination of the reaction thermodynamics (DG), the stiffness of the potential energy surface (l) and the coupling of the initial and nal states (V DA ), as discussed above (eqn (1)).
The electronic coupling between the ground state and the CSS 2 (HOMO / LUMO excitation) was estimated using twostate approximation schemes: the Generalized Mulliken-Hush (GMH) 22 and the Fragment-Charge Difference 23 (FCD) methods. Both approaches produced very similar V DA values for the three compounds, with TNQ-ZnP Ar -TAPD showing the largest coupling in the series ( Table 5).
The electronic couplings listed in Table 5 were calculated by considering only the lowest energy conformation of each molecule. In the case of TNQ-ZnP Ar -TAPD, we also calculated the coupling V DA (GMH) as a function of the dihedral angles to the donor and acceptor (Fig. 4b). This plot, together with the Boltzmann distribution of dihedral angles (Fig. 4a), shows that the range of conformations populated in frozen xylene (q z 90 AE 30 ) show modest variation in coupling (V DA z 3 AE 2 cm À1 ). The energy DG 3 of the CSS 2 of TNQ-ZnP Ar -TAPD is also insensitive to the dihedral angle (ca. 1% variation for q ¼ 90 AE 30 , see ESI Fig. S2 †) which indicates that it is reasonable to consider only the lowest energy conformation of this molecule. a P, A and D indicate the location of the orbital on the porphyrin, acceptor or donor. b The LUMO of C 60 is triply degenerate, but the saturation at the pyrrolidine linking breaks the symmetry. Table 5 Electronic couplings (V DA ) and inner-sphere reorganization energies (l) for CSS 2 from TD-DFT B3LYP/6-31G(d) calculations The treatment of low-frequency, thermally accessible vibrational modes in Marcus theory is fairly straightforward via a harmonic potential with a recombination energy. However, in the low-temperature regime where reactions progress almost exclusively through tunneling, electron-vibration coupling must be treated more explicitly. We modeled vibronic coupling under the Franck-Condon approximation by determining the Huang-Rhys factors (S) for the electron transfer. 24 We used several approaches to estimate S (see ESI † for details), using both the ground state and CSS 2 vibrational modes. The results from these two methods were consistent with each other. At each level of theory, there is one S q for each vibrational mode q, of frequency u q . For ease of comparison, these can be rolled into classical reorganization energies l ¼ X q S q u q . Table 5 reports vibrational (inner-sphere) reorganization energy for the two donor-acceptor combinations. The system with a C 60 acceptor has a lower reorganization energy by around 0.15 eV, in keeping with the known low l of fullerenes in general. 25 Electron transfer rates (k r , Fig. 2), in the low-temperaturelimit, were estimated using eqn (5)

TNQ-ZnP-TAPD TNQ-ZnP
which is derived from the molecular crystal model. 26 It combines the parameters described above and can be solved approximately using the steepest-descent method, in combination with a saddle-point time chosen for optimally sharing the exoergicity among the vibrational modes; u 12 is the frequency corresponding to DG 3 , S q is the Huang-Rhys factor associated to vibrational mode q of frequency u q , and n q ¼ (e ħu q /kT À 1) À1 is the Bose Einstein occupation factor for mode q.
Combining the different levels of theory (CAM-B3LYP/6-31G(d), M062X/6-31G(d) and B3LYP/6-31G(d)) and the various estimates of S we obtained a range of values for these rates in frozen butyronitrile at 4 K. Very little temperature dependence of the results was observed in the 0-10 K regime. The predicted recombination lifetimes of TNQ-ZnP-TAPD range between 0.2 and 6.2 ms, with a geometric mean of 1.6 ms, in remarkable agreement with the experimentally reported 4 ms. Since the only difference between TNQ-ZnP-TAPD and TNQ-ZnP Ar -TAPD in our models is V DA (increased by a factor of 10, averaging between GMH and FCD) we predicted a lifetime 0.02 ms for TNQ-ZnP Ar -TAPD. Since C 60 -ZnP Ar -TAPD has V DA close to TNQ-ZnP-TAPD, but a much lower reorganization energy, the vibronic coupling between initial and nal state is lower. Thus, using eqn (5) we estimated a much longer lifetime (geometric mean prediction 260 ms, range between 6 and 2700 ms).

Experimental characterization of CSS 2 by EPR
We measured the time-resolved EPR spectrum of the photoexcited triad TNQ-ZnP Ar -TAPD under similar conditions to those reported by Wasielewski and co-workers in three different solvents (butyronitrile, 2-methyltetrahydrofuran and xylenes). Disappointingly, we were not able to detect any trace of a longlived photo-excited charge-separate state. Instead we only detected the signal of the porphyrin triplet excited state. This was conrmed by comparing with zinc tetraphenylporphyrin in the same solvent and concentration, which gave an identical transient EPR spectrum of the zinc-porphyrin triplet state (Fig. 6). The failure to detect a long-lived CSS for TNQ-ZnP Ar -TAPD, whereas one was observed for TNQ-ZnP-TAPD, can be explained by the greater electronic coupling, V DA , which arises from the orthogonal dihedral angle between the porphyrin unit and the benzene rings linking the donor and the acceptor (Fig. 4b). The slight differences in the thermodynamics of electron transfer between these molecules, also makes charge separation less favorable in TNQ-ZnP Ar -TAPD ( Table 2). The singlet excited state of the tetraaryl porphyrin (ZnP Ar ) is lower than that of the diaryl porphyrin (ZnP) by about 0.05 eV which slightly reduces the total driving force (DG 1 + DG 2 ) for formation of CSS 2 in TNQ-ZnP Ar -TAPD. However the main difference between these two systems is probably the lower oxidation potential of ZnP Ar which makes DG 1 less favorable for charge separation. This subtle change in thermodynamics appears to be enough to prevent charge separation in a frozen solvent matrix.
Field-sweep photo-EPR experiments on C 60 -ZnP Ar -TAPD showed the expected signal of the spin-polarized long-lived CSS 2 (Fig. 7a, bottom, central emission/absorption feature (A)) on the top of a polarized 3 C 60 triplet spectrum. The spectrum of the reference compound C 60 -ZnP Ar -H, recorded under identical conditions (Fig. 7a, top), shows only the 3 C 60 triplet signal (B). The CSS 2 signal was observed for solutions of C 60 -ZnP Ar -TAPD in xylene and in 2-methyltetrahydrofuran (but could not be investigated in butyronitrile due to limited solubility). We performed pulsed EPR experiments in which the integrated Hahn-echo intensity at a particular magnetic eld position in the spectrum was recorded as a function of the time aer laser excitation, to explore the time-evolution of the transient species probed at the chosen eld position. These echo-integrated ash delay experiments were carried out at two different eld positions corresponding to the fullerene triplet state (Fig. 7b, curve B) and the radical pair state (Fig. 7b, curve A) of C 60 -ZnP Ar -TAPD. We were able to simulate both decay curves using a kinetic model, considering the 3 C 60 triplet, the singlet CSS 2 and the three sub-levels, T 0 , T + and T À of the triplet CSS 2 (see ESI †). The 3 C 60 signal (curve B) follows a mono-exponential decay, with a lifetime of 300 AE 3 ms, which is consistent with previous reports. 27 On the other hand, the radical pair signal (curve A) displays two different decay processes. In the ms regime, the emission/absorption signal becomes inverted to an absorption/emission pattern. This is attributed to the spinallowed decay of the S-T 0 sublevels of the radical pair. We tted this decay to a Gaussian distribution of rate constants (mean: 1.8 Â 10 5 s À1 ; standard deviation: 0.7 Â 10 5 s À1 ) corresponding to a mean lifetime of 5.6 ms.
In the ms regime, we observe the decay of a positive signal (Fig. 7b, curve A and insert), corresponding to decay of the remaining T + -T À sublevel population of the CSS 2 radical pair. This decay process was modeled with a Gaussian distribution of rate constants (mean: 64 AE 6 s À1 ; standard deviation: 32 s À1 ). This distribution of rate constants for both decay processes, from S-T 0 and T + -T À sublevels, probably reects the spread of molecular conformation in the frozen solution. The mean lifetime of the charge-separate state of 16 ms is among the longest reported in the literature, and is about four times longer than for TNQ-ZnP-TAPD, 1,7 in keeping with the trend predicted by our computational studies.
Comparison of the integrated intensity of the EPR signal from CSS 2 with that of the C 60 triplet, at early times aer excitation, indicates that the quantum yield of formation of CSS 2 in C 60 -ZnP Ar -TAPD is about 0.1. This is a rough estimate, based on the assumptions that formation of CSS 2 and the C 60 triplet are the dominant decay channels, and that these species have similar polarizations. It is difficult to accurately integrate the CSS 2 signal as it has overlapping emissive and absorptive bands (see ESI † for details).

Conclusions
In this study, we have reported a versatile and convergent route to donor-zinc tetraphenylporphyrin-acceptor triads via Suzuki cross coupling reactions, which is compatible with oxidationsensitive moieties. This approach was applied to synthesize two tetralkylphenylenediamine/zinc porphyrin/acceptor triads. We explored the kinetics and thermodynamics of charge separation in both systems. Triads exhibiting long-lived CSSs at low temperatures in frozen solvents are needed for experiments in the area of quantum information processing, yet it is difficult to achieve charge-separation under these conditions, because the frozen solvent molecules cannot reorient to stabilize the new charge distribution. For the rst triad, TNQ-ZnP Ar -TAPD, our calculations indicated that there would be almost no thermodynamic driving force for charge-separation in a frozen solvent. In keeping with this prediction, we were unable to detect a CSS in this system by EPR. For the second triad, C 60 -ZnP Ar -TAPD, the greater electron-affinity of C 60 was expected to make electron transfer favorable, while the weak coupling between the donor and the acceptor, V DA , and the small reorganization energy, l, were predicted to result in an exceptionally long-lived CSS. These predictions were conrmed by the observation of a long-lived CSS in the solid state by EPR spectroscopy. The characteristic lifetime of the triplet CSS of this system is 16 ms, in xylenes at 10 K, which is among the longest reported. 1 Changing the acceptor from TNQ to C 60 has three important consequences: (1) it increases the driving force for electron transfer, making charge-separation favorable, even in a frozen solvent, (2) it reduces the coupling term V DA , resulting in a slow rate of charge recombination, and (3) it reduces the reorganization energy, l i , also contributing towards a slow recombination rate. Further spectroscopic studies on triad C 60 -ZnP Ar -TAPD are in progress, and will be described in a future report.
This work illustrates the value of quantum mechanical modeling for guiding the synthesis of electron-transfer systems. It also demonstrates the power of Suzuki coupling methodology, using a porphyrin core with either bromine or boronic acid substituents, for building triads with sensitive donor groups. The strategy developed here should provide a route to preparing advanced molecular materials with long photoexcited CSS lifetimes, for applications such as photo-voltaic devices, and optically gated molecular wires. 28

General information
All chemical reagents were used as received. 2-Bromoanthracene 14 and [5,15-bis-(3,5-bis-tert-butylphenyl)-10,20-bisbromoporphinato]zinc(II) 13 were synthesized following a literature procedures. Dichloromethane (DCM) and tetrahydrofuran (THF) were dried over activated alumina prior to use. Anhydrous N,N-dimethylformamide (DMF), acetic acid, nitrobenzene, pyridine, toluene, xylenes and anhydrous 2-methyltetrahydrofuran (MTHF) were supplied by Aldrich and used without further purication. Purge gas was high purity argon. Chromatography was performed on silica (200-400 mesh). 1 H NMR spectra were acquired on a 400 MHz (Bruker AVII 400), 500 MHz (Bruker AVII 500) or 700 MHz (Bruker AVIII 700) spectrometer. Chemical shis (in the ppm scale) were determined versus TMS using the residual solvent peak as the internal reference (CHCl 3 , d ¼ 7.26 ppm). The 1 H NMR spectra of the nal triads, TNQ-ZnP Ar -TAPD and C 60 -ZnP Ar -TAPD were fully assigned by comparison with the spectra of reference compounds, in combination with 2D techniques (COSY and HSQC); see ESI. † Deuterated chloroform was stored over potassium carbonate to avoid any acid trace. UV/Vis absorption spectra were recorded using a Perkin Elmer Lambda 20 UV-Vis Spectrometer. The absorption wavelengths are reported in nm with the extinction coefficient in M À1 cm À1 . Infra-red spectra were recorded in the solid state (neat) using a Bruker Tensor27 FT-IR spectrometer. Mass spectroscopy was performed either on ESI-TOF (Waters LCT Premier) or MALDI-TOF (Waters MALDI Micro MX) spectrometer or using the Bruker Ultraextreme MALDI-TOF/TOF spectrometer from the EPSRC National Mass Spectrometry Service (Swansea). Preparative scale size exclusion chromatography (SEC) was carried out using BioRad Bio-Beads S-X1 with toluene as eluent. ESR samples were prepared in 3.8 mm quartz tubes, sealed under vacuum and kept at 77 K in the dark.