Promoting photocatalytic CO2 reduction through facile electronic modification of N-annulated perylene diimide rhenium bipyridine dyads

The development of CO2 conversion catalysts has become paramount in the effort to close the carbon loop. Herein, we report the synthesis, characterization, and photocatalytic CO2 reduction performance for a series of N-annulated perylene diimide (NPDI) tethered Re(bpy) supramolecular dyads [Re(bpy-C2-NPDI-R)], where R = –H, –Br, –CN, –NO2, –OPh, –NH2, or pyrrolidine (–NR2). The optoelectronic properties of these Re(bpy-C2-NPDI-R) dyads were heavily influenced by the nature of the R-group, resulting in significant differences in photocatalytic CO2 reduction performance. Although some R-groups (i.e. –Br and –OPh) did not influence the performance of CO2 photocatalysis (relative to –H; TONco ∼60), the use of an electron-withdrawing –CN was found to completely deactivate the catalyst (TONco < 1) while the use of an electron-donating –NH2 improved CO2 photocatalysis four-fold (TONco = 234). Despite being the strongest EWG, the –NO2 derivative exhibited good photocatalytic CO2 reduction abilities (TONco = 137). Using a combination of CV and UV-vis-nIR SEC, it was elucidated that the –NO2 derivative undergoes an in situ transformation to –NH2 under reducing conditions, thereby generating a more active catalyst that would account for the unexpected activity. A photocatalytic CO2 mechanism was proposed for these Re(bpy-C2-NPDI-R) dyads (based on molecular orbital descriptions), where it is rationalized that the photoexcitation pathway, as well as the electronic driving-force for NPDI2− to Re(bpy) electron-transfer both significantly influence photocatalytic CO2 reduction. These results help provide rational design principles for the future development of related supramolecular dyads.


Introduction
The adverse effects on climate change related to increased anthropogenic CO 2 emissions has inspired the utilization of excess CO 2 as a sustainable feedstock for value-added chemicals and fuels. 1,2 While the activation of CO 2 is kinetically unfavorable, it can be readily accomplished electro-/photocatalytically via proton-couple multielectron chemical reductions. 3,4 Consequently, the development of capable molecular electro-/ photocatalysts has mainly focused on improving the efficiency and selectivity of the CO 2 conversion process. [5][6][7][8][9] Among the many comprehensively studied molecular catalyst systems, Re(2,2 0 -bipyridine)(CO) 3 Cl [Re(bpy)] is notable for its highly selective CO 2 -to-CO conversion. 10 The versatile bpy ligand has been modied with a variety of substituents to change both the electronic properties and/or the second-sphere H-bonding character of the catalyst. [11][12][13][14][15][16][17][18] And while Re(bpy) alone can be used as an effective CO 2 reduction photocatalyst, 19-24 the photocatalytic CO 2 reduction performance is greatly enhanced via the direct functionalization of Re(bpy) with photosensitizing (PS) units. [25][26][27][28][29][30][31] The development of ruthenium(II) diimine photosensitized Re(bpy) supramolecular dyads has been extensively reported by the Ishitani group. 5,[25][26][27][28][29][30][31] These Ru II -Re I dyads make use of a Zscheme architecture whereby the photoexcited electrons of the Ru II -moiety are reductively quenched and subsequently transferred to the Re I catalyst center to enable CO 2 reduction. To facilitate efficient electron-transfer (eT) and CO 2 photocatalysis, several supra-molecular dyad design principles have been established. First, the photoexcited electron should be localized near the tethering portion between the PS and the catalyst. 5 Second, the tether between the PS and catalyst moieties should be as short as possible (without being through-conjugated) to enable rapid intramolecular eT. [27][28][29][30] Third, increasing the molar absorptivity of the PS-moiety (i.e. by incorporating multiple PS units) can improve the quantum efficiency capabilities and the ensuing eT dynamics of the supramolecular dyad. 31 Restricted by the rst two design principles, attempts to improve the quantum efficiency of these Re(bpy) dyads has been made by using more strongly absorbing PS units, such as porphyrins, [32][33][34][35][36][37] naphthalimide, 38,39 naphthalene diimide, [40][41][42] and perylene diimide (PDI). 43,44 Although the photophysical dynamics of these dyads appear fundamentally well-understood, only a handful have been properly evaluated as CO 2 reduction photocatalysts (see ESI, Table S1 †). [35][36][37][38][39] Recently, we reported on four N-annulated perylene diimide (NPDI) functionalized Re(bpy) dyads as CO 2 reduction electrocatalysts. 45 Our investigation of these Re(bpy)-NPDI dyads revealed that the PS unit (NPDI) functions as an electronreservoir for Re(bpy), enabling efficient CO 2 reduction at an overpotential 300 mV lower than conventional Re(bpy)-type electrocatalysts. Moreover, it was also elucidated that the tether length between Re(bpy) and NPDI governs which CO 2 reduction mechanism is preferred for the supramolecular dyad(s), where the ethyl-linked Re(bpy)-NPDI dyad possessed the greatest degree of electronic communication. These promising results from our initial Re(bpy)-NPDI dyads led us to hypothesize that eT from the electron-reservoir to the Re(bpy) catalyst could be improved by electronically modifying NPDI in two different ways. It was theorized that the introduction of electron withdrawing groups (EWGs) on NPDI may inductively stabilize the entire dyad, thus enabling more efficient eT by increasing the overall electron affinity of Re(bpy)-moiety. Alternatively, the use of electron donating groups (EDGs) on NPDI could make the electron-reservoir more electron-rich and thus more willing to transfer electrons to the Re(bpy)-moiety. To determine which of these two opposing hypotheses was correct, a series of electronically modied ethyl-linked Re(bpy)-NPDI dyads [Re(bpy-C2- were designed (Fig. 1, le) and, for the rst time, their photocatalytic CO 2 -to-CO reduction performance was evaluated. It was revealed that installing an EDG, such as -NH 2 , led to a four-fold enhancement in turnover numbers of CO (TON co ), with respect to the benchmark Re(bpy-C2-NPDI-H) dyad. A mechanism based on molecular orbital (MO) energy levels is proposed to explain the observed differences in photocatalytic CO 2 reduction performance for these dyads caused by the installation of EWGs and EDGs on NPDI.
The electrochemical properties of these Re(bpy-C2-NPDI-R) dyads were next evaluated using cyclic voltammetry (CV). CV analysis was rst performed in CH 2 Cl 2 ( Fig. 2A and S65 †), with all reported redox events being referenced to the Fc +/0 internal standard. Under an atmosphere of argon, all Re(bpy-C2-NPDI-R) dyads exhibited four reduction and two oxidation redox processes. The rst two reversible reductions may be assigned to the NPDIc À/0 and NPDI 2À/ c À redox couples while the third and fourth reductions correspond to the quasi-reversible bpyc À/ 0 and the irreversible Re 0/I redox events, respectively. [11][12][13][14][15][16][17][18]45 With respect to the oxidation events, the irreversible Re II/I redox process remains consistently near E p z +1.0 V. The quasireversible NPDI-based oxidation event, on the other hand, underwent dramatic shis in potential depending on the nature of the electronic substituent.
Based on the presented CV data ( Fig. 2A, S64 and S65 †), it was observed that the installation of EWGs (i.e. -Br, -CN, and -NO 2 ) on NPDI caused both the redox events assigned to the lowest unoccupied molecular orbital (LUMO) and the highest occupied molecular orbital (HOMO) energy levels of NPDI to shi to more positive potentials. In all cases, the electronic bandgap was narrowed because the LUMO energy level was more signicantly perturbed than the HOMO energy level. Conversely, the installation of EDGs (i.e. -OPh, -NH 2 , and -NR 2 ) on NPDI caused the redox events associated with NPDI HOMO and LUMO energy levels to shi to more negative potentials. Once again, the electronic bandgap was decreased mainly due to the more substantial effects experienced by the HOMO energy level. The observed shis in E 1/2 for rst NPDI-R based reduction and oxidation events ( Fig. 2B and C), relative to -H, show good Hammett parameter correlation. 46 Similar correlations were previously reported for electronicallysubstituted Re(4,4 0 -R-bpy) complexes by Kubiak et al., 11 where EWGs shied redox events more positively and EDGs shied redox events more negatively. In principle, combining these two relationships could assist with proper energy level matching when designing future supramolecular dyads based on the Re(bpy-C2-NPDI-R) architecture.
Next, the electrochemistry of these Re(bpy-C2-NPDI-R) dyads were assessed in DMF, where only the reduction processes were measured due to solvent window effects. 59 Under argon, all Re(bpy-C2-NPDI-R) dyads exhibited the same four previously assigned reduction events (vide supra), where solvent effects caused the potential of most redox events to shi positively by $50-100 mV, relative to CH 2 Cl 2 (Fig. S66 †). 60 Upon measuring CVs at variable scan rates, each dyad displayed a diffusionlimited current response when tted to the Randles-Sevcik equation (see ESI, eqn (i) †), with a calculated diffusion Upon subjecting these Re(bpy-C2-NPDI-R) dyads to an atmosphere of CO 2 (in DMF), a moderate CV current enhancement was observed underneath the fourth redox couple (Fig. 3, red traces with E cat/2 z À2.1 V). When a catalyst is supplied with the appropriate combination of substrates at a sufficient applied potential, a CV current enhancement is oen observed as a result of initiating an electron-consuming catalytic process. 61,62 Moreover, these CV current enhancements may be monitored as a function of catalyst or proton-source concentration to gather preliminary mechanistic data on the catalytic process (see ESI, eqn (ii) †). 63,64 When the Re(bpy-C2-NPDI-R) concentration was varied, a linear CV current enhancement response was also observed for most dyads (Fig. S68 †). Similarly, when proton-source 2,2,2-triuoroethanol (TFE) was incrementally added to a CO 2 -saturated Re(bpy-C2-NPDI-R) dyad solution, it induced a CV current -CN (yellow), and -NO 2 (green). All measurements were recorded at 100 mV s À1 , under argon in CH 2 Cl 2 with 0.1 M TBAPF 6 supporting electrolyte (WE ¼ glassy carbon, CE ¼ Pt-wire, RE ¼ Ag/AgCl, and Fc +/0 as internal reference standard). Correlation diagrams plot the observed shifts in E 1/2 (relative to R ¼ -H) for the first reduction (B) and oxidation (C) as a function of R-group Hammett parameter. 46 enhancement increase under the fourth reduction event (Fig. 3). We further note that incremental addition of TFE also caused the NPDI 2À/ c À redox couple to gradually shi to more positive potentials. This behavior has been modeled for similar rylene diimide materials as a 2H + /2e À proton-coupled electrontransfer process, 65 whereby the NPDI imide oxygens are protonated by the proton-source. 66,67 While the collected CV data was not obtained under steady-state conditions (i.e. plateau current), 68 modelling these measured CV current enhancements as CO 2 reduction variables can qualitatively describe the effects of catalyst and proton-source as rst-order and second-order rate-dependent variables, respectively. In other words, these data imply that one Re(bpy-C2-NPDI-R) dyad, with the assistance of two proton-source molecules, can enable electrocatalytic CO 2 conversion. 45 The obvious exception to these generalized electrochemical trends is the -NO 2 dyad (Fig. 3G). An in-depth evaluation of the electrochemical behaviour of -NO 2 under argon and CO 2 is provided in the ESI (see Section VIII, Fig. S84-S86 †). Thorough analysis of this data strongly suggests that -NO 2 is converted in situ to -NH 2 under reducing conditions; as such, it is difficult to establish meaningful trends for -NO 2 (vide infra).

CO 2 electro-/photocatalysis
The electrocatalytic CO 2 reduction abilities of these Re(bpy-C2-NPDI-R) dyads were evaluated using controlled potential electrolysis (CPE). All experiments were performed in DMF (with 2 M TFE) using our previously described two-compartment Hcell. 45,69 At an applied potential (E appl ) of À1.8 V (Fig. S75 †), all Re(bpy-C2-NPDI-R) dyads (except -CN) achieved comparable turnover numbers of CO (TON co ¼ 21-25) and faradaic efficiencies (FE co ¼ 87-99%) aer 6 hours of electrocatalysis. The -CN derivative, on the other hand, attained about half the performance (TON co ¼ 13) at a slightly lower FE co (87%). This drop in performance is consistent with decreased efficacy of the electron-reservoir effect due to the increased electron-affinity of NPDI-CN (Fig. S77 †). Unlike the electronically modied Re(4,4 0 -R-bpy) series reported by Kubiak et al., where electrocatalytic CO 2 -to-CO conversion efficiency was highly dependent on the nature of the R-group, 11 all Re(bpy-C2-NPDI-R) dyads achieved high FE co at an overpotential that is $300 mV lower than the measured E cat/2 z À2.1 V (Fig. S67 †). It should be noted, however, that when an E appl ¼ À1.7 V was used for CPE ( Fig. S76 †), electrocatalytic CO 2 reduction was essentially shutoff for all Re(bpy-C2-NPDI-R) dyads (TON co # 6). This could indicate that altering the electronic properties of the electronreservoir may not be the most effective strategy towards further lowering the overpotentials required to enable electrocatalytic CO 2 reduction.
Over the 24 h testing period, the Re(bpy-C2-NPDI-R) dyads all showed good activity for $9 hours, aer which CO production would level-off for the remainder of the experiment (Fig. S78 †). The benchmark dyad, -H, achieved a TON co of 57 AE 1 with a selectivity for CO of 97%. The -Br (TON co ¼ 61 AE 5) and -OPh (TON co ¼ 59 AE 6) dyads achieved the same performance with roughly the same CO selectivity ($95%). The -CN derivative was completely inactive for CO 2 photocatalysis under these conditions. Interestingly, despite being the most EWG, the -NO 2 dyad achieved the second best TON co (134 AE 15) with a very high CO selectivity of 98%. When EDGs were functionalized on NPDI, the CO 2 -to-CO production and selectivity was improved to (TON co ¼ 86 AE 8) for -NR 2 and (TON co ¼ 234 AE 13) for -NH 2 . We note that the photocatalytic performance of -NH 2 was not greatly improved over the 24 h testing period by replenishing both BIH and CO 2 in 6 h intervals (TON co ¼ 294; Fig. S79 †), suggesting that depletion of substrate was not the limiting factor for TON co .
To conrm the importance of each component in the photocatalytic CO 2 reduction setup, various control experiments were conducted. As expected, the omission of Re(bpy-C2-NPDI-R) dyads or CO 2 from the setup stopped the production of CO. When sacricial reducing agent BIH was excluded (Table S2 †), the TON co was decreased by at least two-fold for all dyads. This result implies that while TEOA alone can simultaneously act as the proton-source and the sacricial electron-donor, 20 BIH is more efficient at reductively quenching the photoexcited dyads. When TEOA was replaced by TFE (Table S3 †), the production of CO for all dyads was decreased almost four-fold, except for -NH 2 (TON co ¼ 143). Overall, this result points towards the utility of TEOA as a sacricial electron-donor, as well as the importance of forming Re(bpy)-adducts during CO 2 reduction, as seen in previous literature examples. 5,[26][27][28][29][30][31] When the irradiation source was switched to a green light LED array (l ¼ 525 AE 32 nm; 1.9 mW cm À2 ), the measured TON co decreased at least ve-fold for all Re(bpy-C2-NPDI-R) dyads (Table S4 †). This signicant drop in performance could be the result of either inefficient photoexcitation pathways in the dyad [40][41][42][43][44] and/or the elimination of a photo-assisted CO cleavage process. 70,71 Lastly, the importance of tethering the NPDI-moiety to the Re(bpy)moiety was demonstrated by combining N 3 -C2-NPDI-R with Re(bpy), where it was observed that all samples obtained the same performance as Re(bpy) alone (Table S5 †). We further showed that all N 3 -C2-NPDI-R precursors were essentially inactive for CO 2 conversion under the optimized photocatalysis conditions (TON co < 3; Table S6 †).
While these control experiments clearly highlight the necessity of each component in the photocatalytic CO 2 reduction process, it does not fully account for the performance differences of each Re(bpy-C2-NPDI-R) dyad. Therefore, based on the photocatalytic CO 2 reduction results, the Re(bpy-C2-NPDI-R) dyads may be grouped together in three categories: (i) standard catalysts (weak EWGs/EDGs ¼ -H, -Br, and -OPh), (ii) inactive catalysts (strong EWGs ¼ -CN), and (iii) top-performing catalysts (strong EDGs ¼ -NR 2 and -NH 2 ). Note, the -NO 2 dyad can also be classied as a top-performing catalyst because -NO 2 undergoes an in situ transition to -NH 2 under photocatalytic CO 2 reduction conditions (see ESI Section VIII for more details †).

Mechanistic investigation
During electro-/photocatalytic CO 2 reduction testing of these Re(bpy-C2-NPDI-R) dyads, a series of dramatic colour changes were observed. Prior to light irradiation, the absorption prole of all Re(bpy-C2-NPDI-R) dyads were unchanged by the addition of both TEOA and BIH, conrming that neither reagent reduces Re(bpy-C2-NPDI-R) immediately. Aer sparging with CO 2 , the samples were then subjected to blue light and the progression of colour changes was monitored periodically by UV-vis-nIR spectroscopy (Fig. S80 †). Each Re(bpy-C2-NPDI-R) dyad underwent a transition from their initial colour to either a green (-H, -Br, and -OPh), a dark blue (-CN), or a beige (-NO 2 , -NR 2 , and -NH 2 ) colour that was found to persist throughout the remainder of catalysis. Photoluminescence spectroscopy revealed that the photoexcited state of all Re(bpy-C2-NPDI-R) dyads can be reductively quenched by both TEOA and BIH (Fig. S81 †). Thus, to gain further insight into the photoelectrochemical processes that were occurring during CO 2 reduction catalysis, UV-vis-nIR and FTIR spectroelectrochemistry (SEC) experimentation was conducted. UV-vis-nIR and FTIR SEC data was collected by monitoring an air-free Re(bpy-C2-NPDI-R) dyad solution that was held at a constant E appl (where Red1, Red2, and Red3 correspond to the NPDIc À/0 , NPDI 2À/ c À , and bpyc À/0 reductions, respectively). The UV-vis-nIR SEC data of each Re(bpy-C2-NPDI-R) dyad (Fig. 4) correlates very well with what was observed when we periodically monitored our photocatalytic experiments. When using an E appl ¼ Red1, the l max of all Re(bpy-C2-NPDI-R) dyads was bathochromically shied, the relative molar absorptivity of this l max was stronger, and some new vibrational ne-structure was also observed between 800-1000 nm. While the nature of installed R-group inuences the position of the l max and the absorption ne-structure, all these absorption features are consistent with selective formation of NPDIc À . 43,44 When the E appl is switched to Red2, the spectral features of NPDIc À are rapidly depleted and replaced by shied broad-band absorption peak(s). These spectral features are commonly associated with NPDI 2À , 67 where the vibronic structure of the NPDI 2À absorption prole is once again inuenced by the nature of installed Rgroup. With respect to -H (Fig. 4D), the incorporation of HOMO-modifying EDGs (i.e. -NR 2 , -NH 2 , and -OPh) resulted in minimal changes to the NPDI 2À absorption prole shape (Fig. 4A-C). Conversely, when LUMO-modifying EWGs (i.e. -Br and -CN) are used, the vibronic structure of NPDI 2À is signicantly different (Fig. 4E and F). When the E appl was changed to Red3, no other signicant spectral changes were observed. This result is unsurprising given the differences in molar absorptivity of the NPDI and Re(bpy)-moieties. 57,72 Except for -NO 2 (Fig. 4G), all Re(bpy-C2-NPDI-R) dyads displayed very similar spectral transitions. In-depth analysis of the UV-vis-nIR SEC data for -NO 2 (Fig. S86 †) shows that -NO 2 is in situ converted to -NH 2 under the conditions necessary for CO 2 reduction catalysis. These results help conrm why -NO 2 served as an efficient CO 2 reduction photocatalyst despite having a stronger EWG than the totally inactive -CN dyad derivative. Moving onto the FTIR SEC data, the behavior of all Re(bpy-C2-NPDI-R) dyads were nearly identical (Fig. S58-S63 †). While no spectral changes were detected at E appl ¼ Red1 and Red2, signicant shis in v co were observed at E appl ¼ Red3. At Red3, the Re(bpy)-moiety of these dyads is formally reduced by oneelectron [Re I (bpyc À -C2-NPDI 2À -R)]. The added electron density at the Re(bpy)-moiety results in a lowering of v co from 1895, 1915, and 2019 cm À1 to roughly 1865, 1885, and 1995 cm À1 , respectively. Over time, an equilibration process occurs whereby electron-density is shied from bpyc À to Re, and leads to Re-Cl dissociation [Re 0 (bpy-C2-NPDI 2À -R)]. 58 This crucial process generates a 5-coordinate Re metal-center and can be characterized by a Dv co to 1843, 1862, and 1978 cm À1 . Although this Cldissociation process was more readily observed for dyads bearing EWGs (-CN, -NO 2 , and -Br) than it was for dyads bearing EDGs (-NR 2 , -NH 2 , and -OPh), it was still detected to some degree for all Re(bpy-C2-NPDI-R) catalysts.
By CV, it was elucidated for all Re(bpy-C2-NPDI-R) dyads (except -NH 2 and -NR 2 ) that the HOMO is Re(bpy)-based and the LUMO is NPDI-based (Fig. 5A). Consequently, the rst two photoreductions most likely result from a Re-p* intersystem crossing (ISC) process; however, a direct NPDI-based p-p* transition may also be possible. 43,44 The third photoreduction likely occurs either via direct Re(bpy) 3 MLCT 19 or eT from a photoexcited state of NPDI 2À . 43,44 Following the formal three electron reduction of the Re(bpy-C2-NPDI-R) dyads (Fig. 5B), the next step is Re-Cl dissociation which generates a 5-coordinate Re metal-center whose axial position subsequently forms an adduct with TEOA. 20,74 Another consequence of the chlorodissociation step is that the Re metal-center undergoes a rehybridization process that lowers the overall energy of the Re(bpy)based MOs. 45,75 Following the formation of the TEOA-Re(bpy) adduct, it is possible for CO 2 insertion to occur without direct eT from the catalyst center (Fig. 5D). 20,[27][28][29][30] Protonation of the resulting carbonate intermediate induces a reorganization process that releases TEOA and forms a Re-CO 2 c À species (Fig. 5E). 22 Due to the presence of 13 CO and H 13 CO 3 À in the 13 C { 1 H} NMR spectrum aer blue light irradiation (Fi. S83 †), it is postulated that these Re(bpy-C2-NPDI-R) dyads operate via a BIH-mediated disproportionation reaction between Re-CO 2 c À and another equivalent of CO 2 that liberates HCO 3 À , 25 rather than a proton-coupled electron transfer process from NPDI 2À to the Re(bpy)-moiety that liberates OH À . 18 From there, reductive quenching of a photoexcited electron restores NPDI 2À (Fig. 5G) and the ensuing transfer of this electron to the Re(bpy)-moiety produces CO, as well as opens a coordination site for TEOA (Fig. 5H). 21 The photocatalytic cycle is completed by the photoexcitation and reductive quenching of the electron to regenerate NPDI 2À . Based on photocatalytic CO 2 reduction performance, the catalysts were loosely grouped into three categories: (i) standard catalysts (-H, -Br, and -OPh), (ii) inactive catalysts (-CN), and (iii) top-performing catalysts (-NR 2 and -NH 2 ). Looking at the proposed photocatalytic CO 2 reduction mechanism, it is also possible to map out the effects of EWGs and EDGs on the provided MO description of these Re(bpy-C2-NPDI-R) dyads (Fig. 5, highlighted in yellow). In the case of -Br and -OPh, the overall inuence of these R-group does not appear to change the eT dynamics of the Re(bpy-C2-NPDI-R) dyad (with respect to -H). The HOMO-LUMO transition (Re-p*) is identical and the relative shi(s) of the NPDI 2À energy level does not signicantly alter the electronic driving-force of eT between NPDI 2À and Re(bpy). In the case of -CN, the xation of that EWG on NPDI served to lower the energy of the NPDI-based HOMO and LUMOs. While the net result of this transformation retains the original HOMO-LUMO transition (Re-p*), it appears to lower the energy level of NPDI 2À enough to effectively prevent eT from NPDI 2À to the Re(bpy)-moiety, thus shutting down catalysis. Conversely, the installation of strong EDGs (-NR 2 and -NH 2 ) causes the energy levels of the NPDI-based HOMO and LUMOs to increase. The overall result of this transformation not only changes the HOMO-LUMO transition to an exclusively NPDIbased process (p-p*), but it also increases the driving-force for eT from NPDI 2À to Re(bpy). The sum of these two effects together lead to improved dyad eT dynamics, thereby enhancing photocatalytic CO 2 reduction (with respect to -H).
It should be noted that another feasible explanation for the improved performance of the amino-functionalized dyads, in particular -NH 2 , is the possibility of second-sphere H-bonding effects. 15,18 Previously we calculated the optimized geometries of various intermediates during CO 2 catalysis for the -H dyad. 45 It was shown in this study that, although the NPDI was initially folded over the Re(bpy)-moiety, reduction of the -H dyad caused the two moieties to extend away from one another (likely due to coulombic repulsion effects). By analogy, the aminofunctionalized NPDI-R bay position would most likely also be extended away from the Re(bpy)-moiety during CO 2 photocatalysis. While in this case the distance between the catalyst center and the amino-groups of -NR 2 and -NH 2 make it unlikely that second-sphere H-bonding effects are involved in CO 2 photocatalysis, they can't be conclusively ruled out at this time.
If nothing else, the synthetic versatility of the NPDI chromophore means that future iterations of the Re(bpy-C2-NPDI-R) motif could incorporate proximal second-sphere H-bonding groups as a means to further improve CO 2 conversion performance.

Conclusions
In conclusion, we present the synthesis and full characterization of six new Re(bpy-C2-NPDI-R) supramolecular dyad materials (where R ¼ -Br, -CN, -NO 2 , -OPh, -NH 2 , or -NR 2 ). The installation of R-groups on NPDI altered the optoelectronic properties of these dyads, as well as impacted the photocatalytic CO 2 reduction performance. Relative to the benchmark Re(bpy-C2-NPDI-H) dyad (TON co ¼ 57), the incorporation of EDGs (i.e. -NH 2 ) led to an over four-fold improvement in photocatalytic CO 2 reduction performance (TON co ¼ 234) while strong EWGs (i.e. -CN) resulted in complete deactivation of the dyads. Despite being the most electron-withdrawing, the -NO 2 functionalized NPDI was among the top performing CO 2 reduction photocatalysts (TON co ¼ 137), making it an outlier to the proposed trend. Through CV and UV-vis-nIR SEC experimentation, it was elucidated that -NO 2 undergoes an in situ conversion to -NH 2 , thereby forming a different dyad that is responsible for catalysis. A photocatalytic CO 2 reduction mechanism is proposed for these dyads, where EDGs served to accelerate CO 2 reduction rates by simultaneously changing the HOMO-LUMO excitation pathway and by increasing the electronic driving-force of intramolecular electron transfer from NPDI 2À to Re(bpy). Conversely, EWGs shied the LUMO energy levels of NPDI to the point where photocatalysis is shut down because there is no electronic driving force for eT between NPDI 2À and Re(bpy). This study clearly highlights the importance of evaluating structure-property relationships to develop and optimize the future design of new supramolecular dyad photocatalysts.

Author contributions
JDBK performed all experimental work and data analysis and prepared the manuscript. GCW directed the project and provided resources. WEP co-directed the project and provided resources.

Conflicts of interest
There are no conicts to declare.