Integrated Multistep Photochemical and Thermal Continuous Flow Reactions; Production of Bicyclic Lactones with Kilogram Productivity

Integrated Multistep Photochemical and Thermal Continuous Flow Reactions; Production of Bicyclic Lactones with Kilogram Productivity. Rowena A. Howie,a Luke D. Elliott,b Surajit Kayal,a Xue-Zhong Sun,a Magnus W. D. Hanson-Heine,a Jonathan Hunter,a Charlotte A. Clark,a Ashley Love,a Christopher Wiseall,a Darren S. Lee,a Martyn Poliakoff,a Kevin I. Booker Milburnb and Michael W. George*a aSchool of Chemistry, University of Nottingham, University Park, Nottingham, NG7 2RD, UK bSchool of Chemistry, University of Bristol, Cantock’s Close, Bristol, BS8 1TS, UK.


Introduction
There is increasing use of continuous chemical synthesis in both academia and industry. In this context, one unmet goal is the ability to efficiently integrate continuous flow photochemistry with thermal chemistry to construct complex molecules in a more sustainable manner particularly for pharmaceutical, specialty chemical, and flavor and fragrance applications. 1 This need for integration has been driven by efficient and innovative syntheses and has been accompanied by developments in continuous flow reactors which simplify the problems of scaling up photochemical reactions. 2 Whilst the recent explosion in the use of visible light mediated photoredox reactions has offered new pathways for cross coupling and late stage functionalization, the structures generated can often be accessed by traditional means. Excited state photochemistry on the other hand, offers rapid access to sp3 rich scaffolds often unobtainable by thermal and catalytic routes. Although the basic reactions have been known for many years, the products obtained are starting to be appreciated as novel scaffolds in their own right. 3 The focus on photochemistry has, in part, been facilitated by the increasing application of continuous flow chemistry, additionally, the unique reactivity of the often strained photochemical products can be exploited to further diversify the molecular space and complexity. 4 There is considerable interest in applying the release of strain within organic molecules to create versatile building blocks and complex scaffolds. This has proved to be a powerful and widely used approach in thermal chemistry, where releasing the high ring strain in, for example, cyclopropanes and cyclobutanes lead to many possibilities for ring-opening in synthesis. 5,6 This interest in ringopening has coincided with a renaissance in organic photochemistry, partly prompted by the ability of photochemistry to promote transformations that are not accessible by more traditional synthetic routes, 7 and particularly relevant to this paper, to easily introduce strain into organic molecules.
The scale-up problems of photochemistry are principally concerned with ensuring that every molecule in a large volume of reaction mixture is exposed to sufficient light to promote the desired transformation. Flow chemistry can be particularly useful for photochemistry because it enables reaction mixtures to be irradiated in a relatively narrow pathlength with much better light penetration than in a larger tank reactor. Such flow reactors can be used to produce kilogram quantities of compounds either by scaling out (i.e. several reactors in parallel) or by scaling up with highly efficient reactors in combination with high power light sources. 8 Of course, continuous flow chemistry also offers advantages in thermal reactions, such as those that release ring strain mentioned above. These advantages include (i) improved temperature control because of enhanced heat transfer, (ii) the opportunity to link the separate stages of a multi-step synthesis into a single, cleaner process by 'daisy-chaining' a number of different reactors in series and (iii) access to new operating windows (e.g. operating with traditional solvents well above their boiling points) which can widen the scope of known transformations. Superheating of solvents has particularly striking effects in the case of water because, at higher temperatures, H-bonding is weakened with a drop in dielectric constant and increased solubility of organic compounds (e.g. at 300 °C, H2O has solvent properties similar to those of acetone); at the same time, the ionic product of water increases giving enhanced concentrations of both [H] + and [OH] -, opening up the possibility of acid catalysis in the absence of added acid. 9 Thus, there is a considerable attraction in the concept of creating a single continuous flow process to link photochemical and thermal reactors which first create strained structures and then release that strain to generate complex molecular architectures. The challenge, however, is how to match the productivity rates (g/h) of the photochemical and thermal processes so that the reactors can be successfully integrated into a single chain, smoothly converting starting materials into desired product(s) with high selectivity and good productivity. Scheme 1. Synthesis of 8,5-bicyclic lactone 7a, Photochemical addition of 1 and 2 results in a mixture of non-bridged and bridged photoproducts (4 and 5) via a common intermediate, biradical 3. Hydrolysis and subsequent ring opening and lactonization of the bridged photoproduct leads to the desired product 7. The size of the two rings in 7 can be altered by adjusting size of both starting materials; the larger ring from the carbocycle of 1 (x) and the smaller ring by the altering the length of the chain in 2 (y).
In this paper we use the [2+2] cycloaddition / fragmentation of 3,4,5,6-tetrahydrophthalic anhydride (THPA, 1a) and propargyl alcohol (2a) and analogous compounds as a case study to show how photochemistry and thermal chemistry can be combined using continuous flow techniques to create complex structures, found in natural products 10 and of potential use for drug discovery, 11 on a relatively large scale. We show how we have taken a series of photochemical and thermal reactions leading to a bicyclic lactone, 7a in scheme 1, and have integrated them into a single process in continuous flow. The reaction was first reported over 25 years ago, 12 and then used in model studies towards the natural product pachylactone 13 and to study the thermal electrocyclic ring opening scope. 14 Together these papers established that the yields of the final ring-expanded product varied considerably with different sized rings; that is from 84% for 7a, the 8/5 compound, to as little as 6% for the 7/5 analog and, as might be expected, the rate of converting the diester of 6a to 7a increased with temperature (e.g. maximum yield of 7a was achieved in 28 hours in refluxing xylene (bp 139 °C) but only 7 hours in diglyme (bp 162 °C).
Although excellent yields of 7a were obtained in these early studies, the ability to scale-up and hence to fully exploit the methodology was severely limited by the efficiency of the initial photochemical step and the sluggishness of the final thermal step. In this paper we demonstrate that to successfully address this challenge requires a combination of engineering, fundamental chemistry and reactor modelling, which results in a short reaction time in a flowing high-temperature water reactor to increase the productivity transforming a 24-hour acid reflux to a ca. 10-minute residence time. Our multi-faceted approach to this reaction has the following aims: (i) to study the photophysics of the reaction to gain greater insight into the excited state behavior of the substrate; (ii) to optimize the photochemical cycloaddition and thermal fragmentation for optimal productivity and sustainability and (iii) to combine-photochemical and thermal steps into a single streamlined continuous flow process. Our initial optimization focused on the parent 1a/2a reaction but we also demonstrate how other thermally fragmented products inaccessible by traditional means can be accessed using high temperature flow techniques. This study develops a multi-step approach to achieving productivities of kg/day. Our initial strategy was to study the photophysics of the photochemical reaction and hence to accelerate the reaction by use of a thioxanthone photosensitizer, then to identify a solvent that it is compatible with all the different reactions and lastly to optimize the process to maximize its productivity. The final process has enabled us to avoid a solvent change and, by exploiting the unusual acidity of solvent mixtures containing high temperature water, to eliminate the need for concentrated HCl as a catalyst, and, hence, to combine two reactions steps into one. Finally, we demonstrate that our approach can transform the yields of other substrates, e.g. 7b and 7c, where previous, more traditional batch processing has failed to give usable yields of the final product.

Results and Discussion
The initial photochemical reaction which ultimately leads to 7a has been known for some years, but the overall productivity was somewhat low. 13 The overall process starting from 1a to the target lactone 7a involves three stages, (see Scheme 1) (i) the initial [2+2] photochemical addition, (ii) hydrolysis of the resulting anhydride and (iii) a combined electrocyclic ring opening and lactonization, which itself occurs via several intermediates, to form 7a. In the original publication these three steps were carried out in different solvents, which clearly would be problematic for a continuous integrated process.
In order to improve the efficiency of these reactions particularly targeted at scale-up we have investigated a range of photosensitisers (see ESI) and found that isopropylthioxanthone (ITX) to be an efficient triplet sensitizer giving ca. tenfold increase in productivity compared to the unsensitized reaction. We then screened a range of solvents and identified EtOAc as the optimal solvent for photochemistry in terms of productivity and yield (ESI) and sustainability. Using this solvent the photochemistry was scaled-up with the continuous flow Firefly reactor 8c at 3 kW, (EtOAc, 0.5 M), 1 mole scale in 1.1 hrs, then 5 mole scale in 5.5 hrs with no reactor fouling.
Addition of water to EtOAc allows the hydrolysis to be carried out rapidly as an emulsion and results in aqueous solution of 6a after separation. The thermal electrocyclic ring opening of 6a → 7a was found to proceed slowly at reflux in H2O (3 days) but could be catalyzed by 20 mol% HCl, reducing the time to 19 hrs. The resulting product could be isolated in high purity by filtration of the cooled reaction mixture, giving 510 g from a 5 mole photochemical reaction (see ESI). A 24 hour continuous run with the photochemical reactor would give over 2.2 kg of product but with this method would require the use of a 25 L batch reactor for the aqueous thermal reaction. These results, however, opened up the possibility of using flow chemistry to exploit the inherent acidity of superheated H2O to eliminate the need for HCl with the added advantage of accelerating the reaction by running it at higher temperatures. Unfortunately, EtOAc is incompatible with H2O at high temperatures as it hydrolyses rapidly. By contrast, acetonitrile (CH3CN) only reacts slowly with H2O even >300 °C. [15][16] In order to implement a fully continuous process it is preferable to avoid solvent change or biphasic mixtures and, although EtOAc gave the highest productivity for the photochemical step, CH3CN proved more desirable particularly as it has only marginally worse productivity for the photochemistry but is fully miscible with H2O. Therefore, we decided to build the process around using aqueous CH3CN for the subsequent steps i.e. carrying out the photochemistry in pure CH3CN and then adding an equal volume of H2O for the subsequent thermal stages since H2O inhibits the photochemistry if added from the outset. Figure 1 summarizes the key data from a detailed Time Resolved IR (TRIR) 17 study of the reaction of 1a with 2a in CD3CNthe deuterated solvent was used to avoid IR bands of CH3CN masking these regions of the spectrum. Initial experiments with 1a in the absence of other reagents allowed us to identify the IR bands of the singlet and triplet excited states, 1 [1a] and 3 [1a], Figure  1A and 1B. This experiment also showed that ca. 60% 1 [1a] decays back to the ground state of 1a rather than to 3 [1a], thereby confirming why direct excitation of 1a is not the most efficient approach to promoting the reaction of 3 [1a] with 2a. Similar experiments with solutions of the photosensitizer thioxanthone, TXO, (not illustrated) allowed the bands of the excited states of the TXO to be identified. Then, using 1a and TXO together, the rate of energy transfer from 3 [TXO] to 1a was measured ( Figure 1C and 1D). Finally, TRIR data from a mixture containing 1a, 2a and TXO confirmed, for the first time, the presence of proposed intermediate diradical 3a and shows that there is no reformation of 1a. This indicates that the quantum yield of formation of the diradical 3a is considerably higher than in the absence of TXO. The spectra ( Figure 1E and 1F) also showed that both of the photoproducts 4a and 5a are derived from the same intermediate, 3a, making it unlikely that one could easily change the ratio of 4a: 5a by simple variation of the reaction parameters. . The global analysis of these TRIR spectra reveal that irradiation initially led to bleaching of TXO and production of 3 [TXO] which decays ( = 15 ± 3 ns) as the bands of 1a bleach and 3 [1a] is formed. 3 [1a] subsequently decay ( = 12.5 ± 2 ns) to form the bi-radical intermediate 3a which then decays and the TRIR data also confirms that both photoproducts, 4a and 5a, are derived from 3a.
The development of the integrated process was carried out in three stages. First, using a small autoclave, we studied the batch ring opening and lactonization reaction of 6a to 7a in high temperature water (see ESI) to check that (i) the ring opening did occur and (ii) hydrolysis of the lactone did not take place. These experiments were successful with optimal conditions of 200 °C for 10 mins. Transferring this reaction to a continuous pressurized flow reactor revealed a problem, namely that, although the reaction took place, the desired product 7a precipitated when the water cooled down causing blockages, prior to the release of pressure. Fortunately, the lactone 7a is soluble in aqueous CH3CN, and the reaction still took place with near quantitative yield, when using the CH3CN/H2O mixture as the solvent, although, the optimal temperature required for maximum throughput (260 °C) was somewhat higher that than in the batch reaction in H2O. Then to investigate whether the hydrolysis of 5a could also be carried out in a high temperature aqueous CH3CN, we pumped a solution of 5a through thermal reactor and found that both reactions could be carried out as a combined process in a single reactor. The reaction worked equally well with a mixture of 4a and 5a obtained directly from the photochemical reaction, as opposed to pure 5a, Thus, it was not necessary to separate the photoproducts 4a and 5a prior to hydrolysis. In addition, the reaction in CH3CN/H2O was carried out with a residence time of < 2.5 mins. These results opened the way for a single integrated process from 1a→ 7a based on aqueous CH3CN.  Figure 2A is a schematic diagram of our apparatus, broadly it consists of a fluorinated ethylenepropylene (FEP) tubular photoreactor with a 400 W mercury lamp, filtered by borosilicate glass. Then, after a reservoir to prevent any back pressure reaching the photochemical reactor, the solution is fed into a HPLC pump to generate the pressure needed for the thermal stage. A second pump delivers the H2O and the resulting CH3CN/H2O reaction mixture passes directly into a heated coiled tubular reactor which serves as both preheater and reactor. This is followed by cooldown and pressure release, yielding a solution of the target product 7a, the hydrolysis product of 4a and residual TXO. Evaporation of CH3CN from this aqueous solution resulted in precipitation of 7a, in high purity, which could be then separated by filtration from the solution of any remaining diacid 6a and the hydrolysis products of 4a.
Using this system, 1a (0.1 M) with 1.5 eqs of 2a and TXO (1 mol%) in CH3CN were flowed through the photoreactor at 10 ml/min, giving greater than 95% conversion and a 68% yield of 5a, 41 mmol/h (plus 15% 4a). This solution was then mixed with water in a 1:1 ratio by volume to give a total flow rate of 20 ml/min and thermally treated at 240 to 260 °C, as summarized in Table 1. At the highest temperature, near quantitative yields of 7a were obtained for the thermal steps, giving an overall yield of 67% for the daisy-chained process. This corresponds to a productivity of 39 mmol/hr and is a marked improvement over the 44% yield of 7a obtained via separate batch processing, particularly considering the reduction in overall reaction time from greater than 24 hours to a residence time of just 10 minutes in flow. From these results, it was clear that the maximum throughput of the combined reactor was being limited by the photochemical rather than the thermal reactors. Therefore, we installed two additional identical photoreactors in parallel with the first, thereby at least tripling the productivity of the photochemical reaction. Running these combined reactors together with small changes to the reaction conditions and an increased concentration of substrate to 0.5 M allowed the overall productivity of the daisy-chained sequence to be increased to 225 mmol/hr. For this test run the reactors were successfully held at the desired reaction conditions for 1 hour after equilibration with no evidence of reactor fouling or precipitation of reactants. Extrapolating from this 1 hr period would give an equivalent rate of formation of 7a of greater than 1.1 kg/day, using lab scale equipment.
Work using a novel tubular quartz reactor, with a 3 kW Hg lamp, gave very promising results for the further scale up of the photochemical step. This so-called 'Firefly' reactor 8c was used to carry out the photochemical step on 1 mole of 1a at a 0.5 M concentration in ethyl acetate, using a 30 ml/min flow rate and a 3 kW lamp power. This resulted in full conversion, giving a 70% yield of 5a at a productivity of 630 mmol/hr. This reaction was subsequently repeated on a 5 mole scale, over the course of 5 hours, with no evidence of fouling within the reactor. These results represent an approximately three-fold increase in productivity and flow rate over the parallel, triple 400 W FEP reactors, taking the photochemical productivity beyond the current capabilities of our thermal reactor. Process modelling of the thermal reactor using the Process Systems Enterprise (PSE) gPROMS software predicts 82% conversion using similar conditions as reported above in the current reactor, which rises to full conversion and >95% yield if the reactor volume were doubled. 19 Construction of a larger scale reactor train was not carried out during this project, due to limitations in terms of space and equipment as well as the safety restrictions presented by a University research laboratory. Nevertheless, these results are very promising because they suggest that only a modest increase in scale for the thermal reactor would be required to match the output of the Firefly reactor and reach the equivalent of 1 tonne/year, an industrially relevant milestone/benchmark for many pharmaceutical products. Thermal flow chemistry is particularly suited to reactions involving challenging substrates which may decompose at conditions close to those required for their formation. Faster heating, compared to batch reactors of similar productivity, allows the desired conditions to be achieved rapidly, minimizing reaction times, while improved cooling rates help to avoid thermal degradation of unstable products. Therefore, the process intensification that we have achieved with 7a suggests that we might be able to increase the previously reported low yields for related reactions to a more useful level, see examples in Scheme 1, because the yields for the photochemical stage of those reactions were similar to that in the 7a system, with the loss of yield occurring mainly in the thermal steps. Thus, we have focused our investigations on thermal flow reactions, while performing the preceding photochemical steps in batch. Some of the starting materials required for these reactions are economically prohibitive for scale-up, so we decided to build a smaller scale thermal reactor, to reduce waste and to run these reactions at a reasonable cost. The performance of this smaller reactor was validated using the synthesis of 7a and monitored in real time using Raman spectroscopy, Figure 2. This allowed the scalability of the reaction to be investigated, with good agreement between results obtained previously in the larger reactor to those predicted from a gProms model developed using results from the small-scale reactor. This smaller reactor is described in detail in the ESI but, in brief, it consists of a similar tubular heated reactor as used previously but, with a 0.8 mL heated volume, ca. 50 times smaller. The same aqueous CH3CN was used as the solvent for these reactions. The thermal steps in reaction of 1a with 3-butyn-1-ol (2b), have previously required longer reaction times due to less favored formation of the 6-membered lactone 7b but, in our reactor, a 61% yield of the ring opened product, 7b, was obtained after only a 2 min residence time at 250 °C. This contrasts favorably to the previously reported 36 hour reflux in xylene.

Table 2. Optimisation of difficult substrates 5a-d
We then investigated the reaction sequence starting with 1c and 2a which ultimately leads to the 7/5 bicyclic ring-opened product 7c, Scheme 1. When this reaction was originally reported, the methyl ester of this analog was obtained via a series of batch processes over several days with an overall yield of just 4%. Again, good yields had been achieved for the photochemical and esterification steps, with the problem lying in the thermal ring opening steps (6% and later 25%), Reaction of 1a with alcohol, 2d, gave quite different results, because the steric bulk of the Me group in 2d appears to seriously hinder the ring expansion sequence. Tolerance of such methyl groups has been shown for analogs containing a larger ring system, although in the case of 5d batch processing was reported to lead to recovery of the starting material or decomposition. Reacting the photoproduct 5d in our continuous flow reactor in aqueous CH3CN led to the formation of the expected diacid 6d up to 68% at 250 °C. Higher reaction temperatures appear to promote unusual further reactivity, leading to the formation of two novel compounds, 8 and 9, which were isolated in reasonable yields (12% and 34% respectively), Scheme 2. The formation of 8 and 9 could occur via a decarboxylation and subsequent cyclisation. Investigations are ongoing to confirm the proposed mechanism and to establish whether this unusual reactivity could be exploited to carry out a range of similar reactions in high temperature water.

Scheme 2.
Photochemical addition of 1a and 2d to 5d followed by hydrolysis to 6d and thermal decarboxylation and rearrangement to products 8 and 9.

Conclusions
We have successfully integrated photochemical and high temperature water flow reactors to carry out a three-step reaction sequence as a single integrated and continuous process. The addition of a thioxanthone photosensitizer and exploiting the enhanced acidity of high temperature water/acetonitrile mixtures has enabled the reaction time to be reduced from >24 hrs to 10 minutes. This reaction was demonstrated on the equivalent productivity of a > 1 kg/day productivity using lab scale equipment. Although we were operating at the maximum practical scale for a university laboratory, our approach should be simple to scale-up in an appropriate factility, for larger scale production of chemicals. Process analytical technology and modelling were used throughout to support the reaction development, while UV and IR time resolved spectroscopy have been used to provide a deeper understanding of the reaction mechanism. Extending the approach ot other substrates has allowed us to increase previously low yields to levels high enough to make those reactions potentially useful for multi-stage synthesis.

Electronic Supporting Information
Experimental details including details of the reactors, time-resolved infra-red studies, reactor modelling, compound characterisation data and spectra are included in the ESI. 18. The ps-TRIR spectra obtained 1 ps following direct irradiation of 1a at 266 nm clearly show the parent bands are bleached and a new transient produced with bands at 1763 cm -1 which is assigned to 1 n* excited state of 1a, Figure1. The presence of a relatively strong (C=O) is perhaps slightly surprising for a 1 n * excited state but is consistent with the excited state being localized mainly on one of the carbonyls and this assignment was supported by DFT calculations (see ESI). The bands of 1 n* excited state decays at the same rate (= 15 (± 2) ps) as the parent partially reforms and a new band grows in at 1645 cm -1 assigned to formation of the 3

General Experimental
Reagents, solvents and gases were purchased from commercial suppliers and used without further purification, unless otherwise described.
Proton nuclear magnetic resonance ( 1 H NMR) spectra and proton-decoupled carbon nuclear magnetic resonance ( 13 C NMR) spectra were recorded at 25 °C (unless stated otherwise) using Bruker AV400 (400 MHz) and AV(III)400hd (400 MHz) spectrometers. Chemical shifts for proton are reported in parts per million downfield from tetramethylsilane and are referenced to residual protium in the NMR solvent according to values reported in the literature. Chemical shifts for carbon are reported in parts per million downfield from tetramethylsilane and are referenced to the carbon resonances of the solvent. The solvent peak was referenced to 7.26 ppm for 1 H and 77.16 ppm for 13 C in CDCl3.
Yields and conversions for flow experiments were determined by 1 H NMR using biphenyl as an internal standard.

Sensitizer Screen in Batch
A screen of commercially available sensitizers identified isopropylthioxanthone (ITX) as the optimal sensitizer for the [2+2] cycloaddition of THPA with propargyl alcohol (Figure S1). As previously reported for kilogram scale synthesis examples, [1] ITX is preferred over the parent thioxanthone for synthesis and process applications due to its increased solubility in organic solvents. Compared to the unsensitised reaction, productivity was increased by 10× using ITX, representing a large increase in energy efficiency for the reaction. Under these conditions, benzophenone is only marginally less productive than ITX but the greater extinction coefficient of ITX over BP allows it to be used at lower concentrations with minimal loss of UV transmission and hence productivity.

Solvent Screen with ITX in Batch
Having identified isopropylthioxanthone as the optimal sensitizer in acetonitrile, a solvent screen was carried out at the increased concentration of 0.5 M and ITX loading of 1% ( Figure S2). EtOAc was identified as the optimal solvent for the reaction giving a superior productivity to the standard MeCN solvent. Its lower cost, toxicity and environmental impact are also attractive features which fit well within the aims of the current work.

Screening Concentration and Lamp Power
The optimized solvent (EtOAc) and sensitizer (ITX) conditions were used to screen batch reactions with an increased lamp power of 400 W and concentration of 1.0 M ( Figure S3). As expected, the use of a 400 W lamp tripled the productivity of the reaction in comparison to the 125 W results. The reaction tolerates the higher concentration of 1.0 M but a minor decrease in productivity was observed.

Transfer to Large Scale Continuous Flow Reactor with a 3 kW Lamp -Predicting the Flow Rate From Batch Results
The flow reactor (Firefly) used here has been described before and was used without any deviation from the description in the reference.

Homogenous Lactone Hydrolysis Investigation in Batch
The addition of 1 M HCl aq. (10% vol) to a crude MeCN (0.5 M) reaction mixture led to the slow hydrolysis over a period of 3 days (Scheme S2). The hydrolysis rate did not appear to be significantly increased by simply heating the reaction. The diacid product precipitated out and could be filtered off in reasonable yield. The addition of DMAP to 10% water in MeCN resulted in a faster hydrolysis but the product did not precipitate out, even after acidifying with HCl. Prior to hydrolysis, the reaction mixture was concentrated from 2 L to 800 ml before the addition of 1M HCl. The product was isolated, along with the hydrolysed by-product of the photochemical step, by filtration ( Figure S4).

Biphasic Lactone Hydrolysis / Aqueous Ring Opening Sequence
As an alternative approach to hydrolysis, the crude EtOAc solution was stirred as an emulsion with water. Unlike the MeCN / water hydrolysis, the reaction rate could be increased with heating such that it was at full conversion within 1 hr at 60°C. Since the hydrolysed diacid product was formed as an aqueous solution, this method relies on the electrocyclic ring opening to be carried out in water. Before further optimization, the thermal ring-opening step was investigated.
No significant reaction was observed on heating an aqueous solution at reflux for 2 hours. The addition of catalytic HCl to the aqueous solution successfully accelerated the reaction at reflux and pleasingly, the product precipitated out in high purity after the solution was cooled (Scheme S4).

Scheme S4: Novel acid catalyzed ring opening of cyclobutene diacid in aqueous solution
In further trial reactions it was found that any significant residual EtOAc dissolved in the aqueous was hydrolyzed to AcOH and EtOH and had the effect of inhibiting the reaction and making purification more difficult. The best results were obtained when excess EtOAc was distilled from the aqueous solution using a rotary evaporator.

1 Mole Scale Photo/Thermal Reaction Sequence
The optimized reaction sequence was run through on a mole scale to demonstrate the potential for further scale-up (Scheme S5). The photochemistry proceeded at full conversion to give about 70% of the cyclobutene. The addition of hexane (200 ml) to the concentrated photosylate (700 ml) assisted with the separation and allowed the hydrolysed diacid product to be extracted into the aqueous solution more efficiently.

Scheme S5: Trial 1 mole scale flow [2+2] photochemistry / hydrolysis / aqueous electrocyclic ring opening
After heating at reflux in acidified aqueous solution, a clumpy suspension of product formed which was filtered to give a soft yellow solid. This was filtered with diethyl ether to give pure product as a free-flowing colourless powder 104 g, 50% -(68% photochem, 74% thermal)

5 Mole Scale Photo/Thermal Reaction Sequence
With increased confidence in the optimized conditions, the reaction sequence was repeated on a 5 mole scale to investigate the robustness of the procedure (Scheme S6).

Scheme S6: 5 mole scale flow [2+2] photochemistry / hydrolysis / aqueous electrocyclic ring opening
The photochemical reaction solution was prepared (10 L, 0.5 M, EtOAc) in two 5 L bottles. Each bottle contained THPA (380 g, 2.5 mol) and propargyl alcohol (175 ml, 3 mol) and made up to 5 L volume in EtOAC. The solvent was used as purchased with minimal degassing prior to use by stirring under reduced pressure during THPA addition. Full conversion to product was observed for the duration of the flow reaction, which took just over 5.5 hours at 30 ml/min (Figure S5, left) with the Firefly parallel tube flow reactor with Pyrex filter and Hg lamp at 3 kW. This indicates there was no fouling of the reaction tubing due to deposition of degradation by-products.
The photosylate was concentrated to 2.5 L and hydrolyzed by forming an emulsion with water (1.4 L) and hexane (700 ml) at 60°C for 20 mins with an overhead stirrer in a 5 L Radleys Reactor-Ready double jacketed vessel (Figure S5, right). Separation and re-extraction with water (1 L) gave an aqueous solution of the diacid from which dissolved EtOAc was removed by distillation under reduced pressure to give a volume of approximately 3 L.
The aqueous solution was further diluted with water (500 ml) and heated under reflux conditions with the 5 L Radleys Reactor-Ready with 1 mole (83 ml, 12 M) of HCl. Conversion to the ring-opened product was measured by 1 H-NMR to be 92% after 18 hours. The mixture was cooled after 19 hours before filtering the precipitated product and washing with water, MeOH and Et2O to give the product as a powder (510 g, 49% overall, see Figure S6).

Large Scale Thermal Flow Reactor
Large scale continuous flow reactions were carried out using a custom-built flow system constructed primarily from stainless steel Swagelok tubing and fittings, a schematic of which can be seen in Figure  S7, below. The system in brief is made up of two HPLC pumps, a heated coil reactor with an internal volume of 43.7 ml, a pipe-in-pipe cooling jacket and a manual back pressure regulator (BPR). The heated reactor consists of a 6 m coil of 1/4" OD Swagelok tubing formed around an aluminium heating block, which contains a cartridge heater and is surrounded by an additional band heater. An optional third pump, which can act as a quench to rapidly cool and dilute the reactor outflow is also present. The system temperature is controlled by a Eurotherm heater controller, attached to the reactor and monitored by four K-type thermocouples, two within the heating block and two in flow, downstream of the reactor and the upstream of the BPR. Reagents are introduced to the system via up to three Gilson 305 HPLC pumps, fitted with 5, 10 or 25 SC pump heads depending on the desired flow rate. The system pressure is set by a TESCOM manual BPR (26-1700 series) and monitored by three RDP electronics pressure transducers located in flow, just downstream of each pump. A trip system connected to the equipment which isolates power to the pumps and heaters in case of an over--pressure or over-temperature, being recorded in the system and a Swagelok sprung relief valves provides additional overpressure protection.
Care must be taken when designing or working with high pressure continuous flow systems and solvents above their boiling points due to the high temperature, pressure and the potential for high temperature vapours to be released potential causing severe burns. The temperature and pressure ratings of rig components must not be exceeded. The flammability and reactivity of solvents used should also be considered particularly at elevated temperature and pressure, the autoignition temperature must not be exceeded and reactions carried out below these conditions with a safe margin.

Small Scale Thermal Flow Reactor
To allow a broader range of substrates to be investigated, a smaller scale thermal reactor and set-up were constructed (Figure S8). This design of this system was similar to that described above, but scaled down to suit a 3 m long coiled reactor constructed from 1/16" OD Swagelok tubing, with an internal volume of 0.77 ml. This gave a total rig volume of less than 2 ml, including reagent feed lines, of which around 1.2 ml is pressurised. Jasco PU-980 HPLC pumps and BP-2080 Back pressure regulator were used in the small scale set-up to suit the lower flow rates required, while needle valves fitted upstream of the pumps allow them to be primed without depressurising the entire system.

Standard Operating Procedure for Thermal Flow Reactors
A typical continuous flow thermal reaction procedure follows. The standard operating procedures of each size of reactor are the same, with the exceptions of the equilibration periods, priming techniques and method of setting back pressure. For the larger reactor, the minimum equilibration period was taken as the time for three reactor volumes worth of solution to be pumped into the reactor, however this was increased to two total rig volumes for the smaller reactor (>5 reactor volumes), to allow for the increased proportion of non-reactor volume within the system. To account for minor fluctuations in rig conditions, all samples were collected in triplicate. Once collected, a known volume of each sample was dried under reduced pressure, before biphenyl was added as an external standard and the samples submitted for 1 H NMR in deuterated DMSO. S14

Batch Thermal Reactions
Initial thermal screening reactions were carried out in a series of stainless steel autoclaves with an internal volume of 10 ml, the design of which has been published previously. [2] This reaction set-up allows small scale reactions to be carried out at above the boiling point of the solvents, as the sealed reactors are pressurised autogenously, while attached temperature and pressure monitoring allow the reaction conditions to be monitored throughout. The reaction set-up in brief composes of two wells, one heated and the other water cooled, and the autoclaves can be moved between the wells without interrupting the pressure or temperature monitoring.
The desired reactants and solvent are loaded into a stainless steel autoclave, and the autoclave is sealed, before being placed into either well as required. Short reaction times are possible, as the combination of pre-heating the thermal well and efficient heat transfer to the relatively small reaction vessels allow the reactants to reach the desired temperatures (typically 200-300 °C) in around 7 minutes. The reactants are then held at the desired temperature for 10 minutes (timer started once the reaction temperature is within 5 °C of target), before the autoclave is transferred to the water cooled well, and the temperature rapidly decreased. Full cooling of the reactors takes around 10 minutes, although the internal temperature typically decreases by half within the first 2 minutes.

Photochemical Set-up
FEP coil reactors, [3] similar to those previously described, were used to carry out the photochemical step of these reactions in flow. These reactors consist of one or more layers of thin walled FEP tubing coiled around a central light source. For the initial photochemical flow reactions, FEP tubing (2.7 mm ID 3.1 mm OD, supplied by Adtech) was coiled in a single layer around a borosilicate cylinder into which a quartz immersion wells sized to fit 400 W medium pressure mercury arc lamps, was placed (all glassware and lamps supplied by Photochemical Reactors LTD.). This reactor had an irradiated volume of 75 ml. A Masterflex peristaltic pump with PTFE pump head was then used to flow the photochemical reagents in acetonitrile through the reactor.
To ensure safe containment of the UV lamps, each reactor was wrapped in aluminium foil to minimise stray light, and opaque shields were placed in front of the reactors. Due to the possibility of ozone generation, photochemical reactors were only run inside appropriately extracted boxes. Lamp cooling was provided by a recirculating cooling bath, filled with water and operated at 15 °C. A low-flow cutoff device, supplied by Photochemical Reactors LTD, was used to ensure sufficient cooling flow was maintained throughout. Thin K type thermocouples inserted between the immersion well and the inner layer of FEP tubing allowed reaction temperatures to be estimated.
To ensure that any samples collected were representative of equilibrium conditions, the UV lamps were allowed to reach temperature (with cooling flowing) for 30 minutes before the photochemical reactions were started. Likewise, a minimum of two reactor volumes worth of reagents were pumped before samples were taken. All samples were collected in triplicate and a known volume of each sample was dried under a flow of nitrogen, before biphenyl was added as an external standard and the samples submitted for 1 H NMR in deuterated acetonitirile.
To allow scale-up of the daisychain process, three double layer reactors, coiled around quartz immersion wells sized to fit 400 W medium pressure mercury arc lamps, with an inner borosilicate filter placed between the lamp and the immersion well in each case reactors were used. These reactors had an average irradiated volume of 132 ml, and were connected in parallel, with a separate pump feeding each reactor. The combined outflows of which provided the inlet stream of the thermal reactor.

Linking of the Photochemical Flow reactor with the HTW Reactor
To allow the photochemical and thermal steps of this reaction to be combined into a single continuous flow operation, a daisy-chained system was developed, see main text Figure 2. Up to three FEP coil reactors, operating in parallel, were fitted upstream of the thermal system described above, with the outflow of each photoreactor combined to provide the reagent feed for the thermal reactor.

Batch Photochemical Reactions
Batch photochemical reactions were carried out in a commercially available 100 ml borosilicate immersion well reactor, surrounding a 125 W medium pressure mercury arc lamp.

Standard Procedure
Diacid 6a (57 mg, 0.25 mmol) and solvent (water or water/acetonitrile mixture, 5 ml) was added to stainless steel autoclave. The autoclave was then sealed, heated and held at the reaction temperature for 10 minutes, before being rapidly water cooled. Once fully cooled the contents of the reactor were transferred to a measuring cylinder to check for any loss of solvent, which may indicate a reactor leak, before sampling. For reactions in water where precipitation of the product occurred, samples were either filtered to obtain the pure product or made up to a known volume with acetone before a subsample was dried under reduced pressure and biphenyl added as an external standard to allow semi-quantitative 1 H NMR analysis (Table S2). Using pure water as the solvent, full conversion and near quantitative yields were achieved for the 10 minute reaction at 200 °C, the reaction did not reach completion within the 10 minute reaction time below this temperature and some decomposition of the product was observed at 220 °C. Even trace quantities of ethyl acetate within the solvent mixture, lead to an appreciable drop in yield, and increase in reaction pressure. Near equivalent yields to pure water were observed at 200 °C when using a 50:50 v/v mixture of acetonitrile and water, and the reaction products remained in solution even after cooling.

Continuous Flow Ring Expansion/Lactonization of 6a in High Temperature Water/Acetonitrile Mixtures
Continuous flow synthesis of 7a ((E)-1-oxo-1,3,6,7,8,9-hexahydrocycloocta[c]furan-5-carboxylic acid) from 6a (7-(hydroxymethyl)bicyclo[4.2.0]oct-7-ene-1,6-dicarboxylic acid) in high temperature water and high temperature water containing solvent mixtures. These reactions were carried out in the larger of the thermal reactors, and followed the standard operating procedure described above (Table  S3). For the reactions carried out in water, the reactor temperature was set to 200 °C and pressure to 100 bar, a 0.05 M starting solution of diacid in water was introduced to the reactor, at the flow rates shown and an additional water feed at the same flow rate was introduced via the quench pump downstream of the reactor, to dilute the outflow and minimise the accumulation of solid within the reactor due to precipitation of 7a although a large volume of solid was subsequently removed from the reactor post reaction, indicating product accumulation had occurred and potentially explaining the lower than predicted yields.
For reactions in acetonitrile water mixtures, the system pressure was 100 bar, reagent concentrations of 0.06 M of diacid made up in the appropriate solvent mixture were used at the flow rate shown, and no quench pump was required, as reagents and products remained in solution. Isolated yields were obtained for two sets of conditions, via removal of the acetonitrile under reduced pressure and filtration, showing good agreement with yield obtained by NMR, and excellent purity.
Combined hydrolysis and ring opening reaction in batch indicated full conversion of the photoproduct 5a (0.05M in 5 ml solvent) and at least 95% yield of 7a were obtained in a 10 minute reaction at 200 °C for solvent mixtures containing a minimum of 25% water. Below this percentage diacid 6a was recovered in addition to 7a.
Similar yields were obtained when using crude photoproduct (approx. 0.06 M of 5a in acetonitrile) diluted with the appropriate volume of water. In this case additional peaks believed to correspond to the hydrolysis products of non-bridged photoproduct 4a and any remaining propargyl alcohol were also observed via NMR.
For reactions carried out in continuous flow, crude photochemical solutions were used (containing approx. 0.06M of 5a in acetonitrile) and the optimal solvent ratio was found to be 50:50 acetonitrile to water, which was found to increase the rate of hydrolysis. A higher pressure of 150 bar was used to minimise expansion of the higher volatility solvent mixture. Two pumps were used to generate the required solvent mixture in situ, with matched flow rate of water introduced by pump 1 and photoproduct in CH3CN introduced via pump 2.

Batch Photochemistry of Additional Substrates
Batch photochemical reactions were carried out in a 100 ml borosilicate immersion well photoreactor, fitted with a water cooled 125 W Hg arc lamp, all obtained from Photochemical Reactors LTD. Prior to irradiation, solutions were degassed by passing a flow of nitrogen through the solution for 15 minutes.

Optimisation of Small Scale Thermal Reactor for the Ring expansion of 5b and 5c.
Continuous flow synthesis of (E)-1-oxo-3,4,7,8,9,10-hexahydro-1H-cycloocta[c]pyran-6-carboxylic acid (7b) from purified and crude photoproduct (5b) These reactions were carried out in the smaller of the thermal reactors, and followed the standard operating procedure described above. Reactions carried out at 150 bar using either purified 5b (70 mM in acetonitrile) or crude photochemical solutions (containing approx. 70 mM of 5b in acetonitrile). Similar results were obtained in either case. Two pumps were used to generate the required solvent mixture in situ, with matched flow rate of water introduced by pump 1 and photoproduct in acetonitrile introduced via pump 2. Flow rates quoted are the total flow rate, calculate as a sum of all water and acetonitrile flow rates. Yields were obtained via 1 H NMR against a biphenyl external standard and calculated based on an average of two samples, normalised against a measured starting material sample, and were corrected for dilution, pump efficiency and sample evaporation during collection. > 95% conversion of the starting anhydride was observed in each case. These reactions were carried out in the smaller of the thermal reactors, and followed the standard operating procedure described above. Reactions carried out at 150 bar using crude photochemical solutions, which were diluted with acetonitrile to improve solubility giving 33mM concentrations of 5c. Two pumps were used to generate the required 1:1 solvent mixture in situ, with matched flow rate of water introduced by pump 1 and photoproduct in acetonitrile introduced via pump 2. Flow rates quoted are the total flow rate, calculate as a sum of all water and acetonitrile flow rates Yields were S21 obtained via 1 H NMR against a biphenyl external standard and calculated based on an average of two samples, normalised against a measured starting material sample, and were corrected for dilution, pump efficiency and sample evaporation during collection. Greater than 95% conversion of the starting anhydride was observed in each case. * indicates evidence of acetonitrile hydrolysis during reaction or conditions likely to result in sure hydrolysis. Continuous flow thermal reaction of 8-(hydroxymethyl)-9-methyl-4,5,6,7-tetrahydro-1H,3H-3a,7aethenoisobenzofuran-1,3-dione (5d) These reactions were carried out in the smaller of the thermal reactors, and followed the standard operating procedure described above. Reactions carried out at 150 bar using crude photochemical solutions(containing approx. 64 mM of 5d in acetonitrile). Two pumps were used to generate the required 1:1 solvent mixture in situ, with matched flow rate of water introduced by pump 1 and photoproduct in acetonitrile introduced via pump 2. Flow rates quoted are the total flow rate, calculate as a sum of all water and acetonitrile flow rates Yields were obtained via 1 H NMR against a biphenyl external standard and calculated based on an average of two samples, normalised against a measured starting material sample, and were corrected for dilution, pump efficiency and sample evaporation during collection. * indicates evidence of acetonitrile hydrolysis during reaction or conditions likely to result in such hydrolysis.

Description of Time Resolved Infrared Spectroscopy (TRIR) setup
The TRIR spectroscopy apparatus at the Nottingham has been described previously. [7] Briefly, fundamental pulses (800 nm, 100 fs, 80 MHz) are generated with a commercial Ti:Sapphire oscillator (Spectra-Physics MaiTai) and fundamental pulses are amplified in a Ti:Sapphire amplifier (Spectra-Physics SpitfirePro) to produce 800 nm, 100 fs, 1kHz, 2 mJ pulses. Half of the output is used to pump a TOPAS-C (Light Conversion) to produce tunable IR pulses using a difference frequency generator and the other part of the output is used to pump a harmonic generator (

Kinetic rate constants for the sensitized reaction:
The ns TRIR studies on the TXO+1a mixture after 355 nm photoexcitation allows us to isolate the reaction kinetics for the generation of 3 [1a] and its return to the ground state. We performed a global kinetic analysis to obtain quantitative information on the kinetics. From the above discussion, it follows that at least three kinetic components -3 [TXO], 3  where τ is the triplet lifetime in the presence of quencher and τ0 is triplet lifetime in the absence of quencher. For bimolecular quenching experiments, in the low concentration limit of the quencher (dynamic quenching), the kinetics is often pseudo-first order. The plot of the quenching rate constants S38 ( Figure S11(B)) versus 1a concentrations confirms a linear relationship indicating that in this concentration regime the quenching is controlled by diffusion.
Next, we explored the kinetics of the [2+2] photocycloaddition reaction by studying the kinetics of the TXO+1a+2a solution. We have to consider many different chemical species for the kinetic analysis.
Therefore, kinetics of the IR bands associated with the individual species are considered for the detailed kinetic analysis. Selected kinetic traces associated with the various species are presented in Our calculation yields Φdiradical-non sensitized ≈ 0.32 for the quantum yield of the formation of 3a by direct photoexcitation. Using equation (2)

Photochemical step
Raman spectroscopic measurements of the photochemical step were performed using a Kaiser Optical Systems Inc. RXN2 spectrometer, equipped with a sapphire tipped MarqMetrix BallProbe®. The photolysis was performed in an immersion well reactor using a 125 W medium pressure mercury lamp from Photochemical Reactors LTD. The Raman probe was inserted within an external sampling loop which was fed with the circulating reaction mixture via peristaltic pumps, in order to avoid exposure of the spectrometer detector to the mercury lamp light. The Spectra were collected using an excitation wavelength of 785 nm with a 400 mW laser power and a detector exposure time of 30 s with 10 averaged scans. The spectra collected during the course of the reaction are displayed in Figure S13. From these spectra the kinetics of the reaction could be monitored by tracking the normalised intensity of the THPA starting material vibration at 665 cm -1 and the bridged product at 607 cm -1 as a function of time ( Figure S14). Figure S14 -Kinetics of the photochemical step obtained from the normalised intensity of the THPA starting material vibration at 665 cm -1 (blue) and the bridged product at 607 cm -1 (red).

S41
During the course of the reaction the carbonyl region of the Raman spectrum also exhibited this behaviour with the disappearance of the THPA vibrations at ca. 1677, 1773 and 1845 cm -1 and the appearance of the bridged product peak at 1642 cm -1 . This is demonstrated in Figure S15.

Thermal steps
Raman spectroscopic measurements of the thermal reaction was again achieved using a Kaiser Optical Systems Inc. RXN2 spectrometer, equipped with a sapphire tipped MarqMetrix BallProbe®. These spectra were collected with the same detector exposure and laser power (30 s, 10 scans and 400 mW power) as the photochemical step, however the sampling was obtained via the submerging of the probe in a sample collection vessel of the outlet of the thermal flow reactor. The thermal reaction was run at 225 o C and 150 bar, at residence times ranging from 0.125 to 8 mins. Normalised Raman spectra of the product and starting material are shown in Figure S16. Figure S16 -Normalised Raman spectra of the both the starting material (blue) and product 7a (red) of the thermal step of the reaction.

S42
The Spectra in Figure S17 show the effect of the residence time on the yield of the product, as indicated by the height of the characteristic 7a product modes. This is particularly evident for the 1650 cm -1 vibration. Using the height of this mode and plotting it as a function of residence time, it can be seen that the yield of the product increases as the residence time within the thermal reactor increases from 0.125 to 7 mins. However as this residence time increases past this to 8 mins, the yield starts to decrease due to the formation of side products (Figure S18). Figure S17 -Normalised Raman spectra of the progression of the thermal reaction with increasing residence time (blue to red) displaying the growth of the characteristic 7a product modes until 8 min (dark red) where the product yield decreases. This is reflected in a lower intensity of the product's 1650 cm -1 vibration. Figure S18 -The intensity of the 7a product vibration at 1650 cm -1 from the normalised Raman spectra of the thermal reaction. This shows the increase in the product yield with increasing residence time until after ca. 5 mins where the yield starts to decrease as side products form.

gProms Modelling
Process modelling of the thermal reaction of 5a to 7a was carried out using gPROMS modelling software to allow the effect of changing the reaction scale on the performance of the reaction to be investigated, such as would be the case should the Firefly photoreactor be combined with the thermal rig, which would require elevated flow rates.

Model Description
A schematic of the reactor model is shown in Figure S19, showing a coiled reactor that is placed inside a heating block, depicted in red. The model of the high temperature thermal coil reactor has been developed within Process Systems Enterprise's gPROMS © ProcessBuilder 1.3.1. The inner diameter of the reactor tube is 0.305 cm, with a coil diameter of 7 cm and length of 600 cm. The temperature of the heating block, as measured by a K-type thermocouple inside the heating block, is assumed to be the fixed temperature of the reactor wall. The initial concentration concentrations are calculated at standard temperature and pressure, before being introduced into the reactor where it can change depending on conditions.
The physical properties, such as density, viscosity or thermal conductivity, of the reaction mixture were calculated using the Peng-Robinson Advanced 1978 (PR-78A) equation of state (EoS). [8] Due to the low concentration of reagents, the physical properties calculated for this system were assumed to be equal to that of the solvent. The mathematical relationships of the model have been developed from engineering principles so that, with parameter estimation, a semi-empirical model can be developed to predict behaviour outside of the experimental space. The following two reaction scheme (Scheme S7) was initially used, where reaction 1 is hydrolysis of the anhydride motif in 5a and reaction 2 is the electrocyclic ring opening of 6a, which occurs via multiple fast-lived intermediates.
Scheme S7 -The two steps of the thermal reaction, where the bridged product (5a) of the photochemical step is hydrolysed to 6a, which then undergoes a rearrangement to form the 8 membered ring compound (7a).

Density Functional Theory Calculations
Density functional theory (DFT) calculations were carried out for 1a, 1 [1a], and 3 [1a] using the Q-Chem quantum chemistry software package. [9] Kohn-Sham DFT calculations of the lowest energy electronic states with singlet and triplet multiplicity (1a and 3 [1a]) were carried out using the B3LYP exchangecorrelation functional and the 6-311G(d,p) electronic basis set. [10,11] Geometry optimizations, harmonic vibrational frequency calculations and molecular orbital calculations were performed both in vacuo and including an explicit solvent model consisting of three MeCN molecules initially positioned around the carbonyl groups. The 1 [1a] excited state was calculated using an excited state DFT self-consistent field procedure during both the geometry optimization and harmonic frequency calculations, in which an electron was moved from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO) and the maximum-overlap-method was used to prevent variational collapse to the ground state. [12] Time-Dependent DFT (TDDFT) calculations of electronic excitations from the ground state were carried out using the CAM-B3LYP exchangecorrelation functional, [13] both in vacuo and in explicit solvent with a state-specific polarizable continuum model (PCM) solvent with a dielectric constant of 37.5 and an optical dielectric constant of 1.81.
TDDFT calculations indicate that the lowest singlet and triplet electronic excitations of 1a that correspond to the formation of 1 [1a] and 3 [1a], have nπ* and ππ* orbital excitation character, respectively, when solvated by MeCN (see Table S8 and S9). Orbital analysis in Figure S20 then shows that the n-orbital involving the carbonyl oxygen atoms becomes increasingly localized onto only one of the two carbonyl groups as the 1 [1a] geometry relaxes . This in turn leads to a change in the bonding of the excited state where the double bonding character in only one of the carbonyl bonds is depleted. This observation is consistent with the presence of the experimental peak at 1762 cm -1 in the carbonyl stretching region of the TRIR of 1a, and a corresponding single carbonyl stretching mode is predicted from the harmonic vibrational analysis of the nπ* 1 [1a] species with a vibrational frequency of 1759 cm -1 . S47 Figure S20 -The highest energy occupied oxygen n-orbital in the ground state electronic configuration shown for the solvated 1a geometry (left) and 1 [1a] geometry (right).