Accessing Cyclobutane Polymers: Overcoming Synthetic Challenges via Efficient Continuous Flow [2+2] Photopolymerization

Abstract

We report an improved and efficient method to prepare well-defined, structurally complex truxinate cyclobutane polymers via a thioxanthone sensitized solution state [2+2] photopolymerization. Monomers with varying electron density and structure polymerize in good to excellent yield to afford a library of 42 polyesters. Monomers with internal olefin separation distances of greater than 5 Å undergo polymerization via intermolecular [2+2] photocycloaddition readily as opposed to the intramolecular [2+2] photocycloaddition observed in monomers with olefins in closer proximity. Use of a continuous flow reactor decreases reaction time, increases polymer molecular weight, and decreases dispersity compared to batch reactions. Furthermore, under continuous flow, the polymerization is readily scalable beyond what is possible with batch reactions. This advancement ushers truxinate cyclobutane based polyesters, which have been historically limited to a few examples and only research scale quantities, to the forefront of development as new materials for potential uses across industry sectors.

Graphical Abstract

INTRODUCTION

The [2+2] photocycloaddition reaction is ubiquitous in synthetic organic chemistry as one of the most frequently utilized photoreactions.^1–5 Extensive literature surrounds [2+2] photocycloadditions for the preparation of small molecules, including the truxinate scaffold which is of interest to us as a building block for cyclobutane-based polymers.^3–5,7 Cyclobutane-based polymers have been a topic of interest dating back to 1966⁶ due to increasing interest in the development of sustainable polyester and polyamide materials (Figure 1). Non-truxinate cyclobutanes polymers are also of significant interest and are prepared via topochemical polymerization strategies of non-cinnamic acid monomers.^8,9,12 Synthetic strategies to prepare truxinate cyclobutane-based polymers from cinnamates using solution-state [2+2] photopolymerization are limited in scope and number to only a few notable examples.^10–11 As such, there is an unmet need for methods accessing more structurally diverse truxinate cyclobutane polymers as well as improved scalable synthetic routes. Current examples are limited to only milligram quantities of these materials restricting their application and use as functional materials.

Figure 1.

(a) Intermolecular [2+2] photocycloaddition mechanism resulting in cyclobutane polymers (b) Intramolecular [2+2] photocycloaddition mechanism resulting in small molecule cyclobutanes.

Historically, truxinate cyclobutane polymers were synthesized using: 1) conventional condensation polymerizations;¹⁰ 2) solid state [2+2] photocycloadditions;^13–15 or, 3) solution state [2+2] photocycloadditions^11,16 with each method exhibiting distinct limitations. Condensation polymerizations are highly functional group dependent, relying on the presence of carboxylic acids or amines, and exhibit poor molecular weight (M_w) and dispersity (Ð) control. Similarly, intermolecular solid state [2+2] photocycloadditions highly depend on the molecular structure of the monomer: thus, requiring careful monomer selection to permit precise crystal packing of the reactive olefins in the correct orientation and in close proximity for polymerization.^14,17–20 This preorganization of the monomers guarantees regio- and stereoselective, an attractive feature of the solid-state reactions. The monomer design requirement limits the scope and functionality of cyclobutane polymers. While gram scale synthesis of highly crystalline polymeric products is possible, further advances are likely needed for industrial scale.^21,22 In contrast, solution state intermolecular [2+2] photocycloadditions are a promising method for the preparation of cyclobutane polymers due to enhanced monomer tolerance and potential for scalability. Unlike their solid-state counterparts, solution state [2+2] photocycloadditions do not suffer from the same geometric constrains, allowing for the synthesis of more diverse products. However, this method still yields polymers of relatively low molecular weight and broad dispersity due to the competing intramolecular [2+2] photocycloadditions resulting in small molecule dimers and depletion of the monomer feedstock (Figure 1b).

Synthetic methods toward small molecule cyclobutanes via [2+2] photocycloadditions are established and provide insight for methodology optimization of cyclobutane polymers. In fact [2+2] photopolymerizations are gaining increased attention today.^11,12,23 Solution state intramolecular [2+2] photocycloadditions utilize dilute substrate reaction conditions, employ electron rich olefins such as cinnamic esters^19,24–26, and heavily rely on olefin distances less than 4–5 Å to favor small molecule dimerization. For covalently tethered cinnamic esters, the linker defines the absolute configuration of the cyclobutane products resulting in high diastereoselectivity.²⁷ Generally, intramolecular syn head-to-head cycloadditions afford the β-truxinate isomers,^1,28 while intermolecular anti head-to-head cycloadditions yield δ-truxinate isomers.²⁴ Additionally, these reactions require an appropriate triplet-sensitizer to initiate photopolymerization. Iridium based photosensitizers are the most common,^11,29 however thioxanthone derived photosensitizers are attractive organic, low cost alternatives ^30–34 which have been previously employed in [2+2] photopolymerizations of cinnamates. Liao et al. describes a photopolymerization strategy utilizing thioxanthone derivates, all of which sufficiently promote the [2+2] photopolymerization.²³ Building upon the prior small molecule findings and photopolymerization strategies, we present an efficient synthesis of a library of functionally diverse cinnamic ester monomers for novel cyclobutane polymers (Figure 2a). Specifically, we report a library of functionally diverse cyclobutane polyesters from both electron rich and poor olefins, 36 of which have not been previously synthesized to the best of our knowledge. The use of a continuous flow reactor compared to traditional batch polymerization affords increased molecular weight and narrower dispersities at shorter reaction times. Furthermore, we synthesize multigrams of a cyclobutane polymer using continuous flow polymerization which highlights its use as a scalable and continuous process.

Figure 2.

(a) General monomer and polymer synthesis scheme. (b) Cinnamic acid starting materials. (c) Diol linkers. (d) Table of selective representative polymers.

RESULTS AND DISCUSSION

From a design perspective, we envisioned a facile, scalable, and efficient methodology to maximize yield and, thus considered the following five primary design parameters: 1) cinnamic acid structure; 2) linker structure; 3) reaction molarity; 4) photosensitizer selection, and 5) batch versus continuous flow methodologies. We selected six cinnamic acids (1-6) with varying electron densities and seven diol linkers (a-g) of varying length and rigidity to create a library of 42 diverse monomers for subsequent [2+2] photopolymerization (Figure 2b). We hypothesized that electron rich cinnamic acids such as 1 will readily undergo [2+2] photocycloadditions compared to electron poor cinnamic acids such as 6. Second, linkers which place the reactive olefins at a distance of 5 Å or more from one another will favor intermolecular [2+2] photocycloaddition facilitating polymerization as opposed to intramolecular [2+2] photocycloaddition (Figure 3a), avoiding formation of small molecule dimers and depletion of the monomer feedstock. For example, monomers containing bisphenol-A (g) or hydroquinone (f) linkers which position the olefins outside of the 4–5 Å distance (Figure 3b), will readily polymerize with decreased termination events (i.e., larger molecular weights) compared to monomers containing ethylene glycol (a) or catechol (e) which place the internal olefins in an advantageous position (4.9 Å) for the intramolecular [2+2] photocycloaddition (Figure 3a). Monomers that exhibit a preference for the intramolecular photocycloaddition as a result of their linker will be polymerized at sufficiently high monomer concentrations (> 0.7 M) to favor the intermolecular polymerization. Finally, photopolymerizations performed in a continuous flow reactor will afford larger molecular weight polymers compared to batch polymerizations facilitated by a smaller optical reaction pathlength allowing for more uniform irradiation of the reaction mixture. Additionally, we will use 1 mol% thioxanthone as our photoinitiator rather than the commonly utilized Ir(ppy)₃, to reduce cost of the polymerization and eliminate the use of heavy metals, given that our application interests are in biodegradable and environmentally friendly polymers.^11,13,35 Specifically, we synthesized 42 bis-cinnamate monomers 1a – 6g (see S2 for experimental details and S3 for detailed characterization data) via a Steighlich esterification from the corresponding cinnamic acids and diol linkers (Figure 2). We first determined the optimal irradiation time in a Lucent360 photoreactor equipped with 365 nm UV-LEDs for 6, 12, 18, 24 and 36 hours at 12–18 °C using monomer 4a (100–200 mg, 1 M, 1 eq., λ_max = 278 nm) and 1 mol % thioxanthone (λ_max = 384, 266 nm) in degassed (15 min) anhydrous DMF. In accordance with a traditional step-growth polymerization mechanism, the molecular weight logarithmically increases over time with broad polymer dispersity (Đ) at all time points. We selected 24 h as the optimal reaction time for all subsequent polymerizations as it provides a molecular weight greater than 30 kDa and Đ less than 3 for 4a’ (Figure 4b). Washing the products with 5% w/v LiCl followed by precipitation in methanol removed the DMF to afford pure polymer ranging in molecular weights from 12.2 kDa (2a’) to 142.9 kDa (3a’) and Đ less than 2 for most polymers apart from 3b’, 4a’ and 4g’ which exhibit higher dispersity (detailed yield as well as molecular weight and dispersity data as determined by light scattering analysis are available in Table S2 for all polymerizations). We also determined the dispersity of the polymers by RI and the values vary between the two measurements (Figure 2d, see Table S2 in the supporting information for complete dispersity data). In general, monomers containing linkers f and g, with separated olefins, polymerize readily in good yields (> 64%) and decreased intramolecular cyclobutane byproducts, as determined via NMR and GPC-SEC analysis, compared to monomer with linkers a – b and e (Figure 5). Monomers possessing linkers of varying lengths and flexibility separating the olefins show a similar trend in reactivity. Monomers linked by d, generally polymerize in higher yields with reduced amounts of small molecule cyclobutane byproducts. Additionally, the degree of polymerization also increases in most substrates with increasing linker length from a to d, of approximately 1.4 – 4x. It is noteworthy to mention, many monomers, particularly those containing linkers f and g initially exhibit limited solubility in DMF and other common organic solvents at 1 M (see supporting info for detailed solubility information), however following 15–30 minutes of irradiation the particulate monomer is not present, and polymerization proceeds in solution state.

Figure 3.

(a) Crystal structure of monomer 2e depicting olefin distance of 4.9 Å. (b) Crystal structure of monomer 4f depicting olefin distance of 9.9 Å

Figure 4.

(a) Polymerization scheme of monomer 4a (b) Optimization of batch polymerization conditions of monomer 4a (c) Continuous flow polymerization of monomer 4a (d) Direct comparison of molecular weight of 4a’ in batch and continuous flow indicating continuous flow facilitates increasing molecular weights

Figure 5.

(a) GPC spectra of 5a’ – 5d’ indicating that as linker length increases from a – d, small molecule dimerization decreases (b) GPC spectra of 5e’ – 5g’ indicating that as linker length increases from e – g, small molecule dimerization decreases.

Despite multiple polymerization attempts, both monomers 2g and 6g do not polymerize, nor show any unproductive reaction pathways such as E/Z isomerization or decomposition determined via GPC-SEC and ¹H NMR analyses. Upon analysis of the UV-Vis spectrum, monomers 2g and 6g exhibit blue-shifted absorbance maxima (λ_max = 312 nm and 273 nm respectively), suggesting that thioxanthone (λ^abs_max = 384, 266 nm and λ^em_max = 425 nm) is not an appropriate triplet sensitizer for these two monomers. To circumvent this issue, a shorter, higher energy wavelength lamp may be used for direct excitation of 2g and 6g, however other unproductive side reactions such as the Fries rearrangement^36,37 would also be possible, leading to an unclean polymerization pathway or complete material decomposition. As such, we screened alternative triplet sensitizers such as benzophenone and Ir(ppy)_3. Ultimately, Ir(ppy)₃ (λ^abs_max = 376, 430, 500 nm and λ^em_max = 519 nm)³⁸ in conjugation with increased reaction temperatures (50 °C) and lower reaction molarity (0.04 M) promotes photopolymerization of 6g (137.5 kDa, 1.14 Đ (LS)). Monomer 2g polymerizes under similar conditions, however extended reaction times (48 h), and increased molarity (0.078 M) are needed. Unfortunately, E/Z isomerization occurs under these conditions, resulting in decreased yield, and a bimodal polymer distribution by GPC-SEC (9.7 kDa, 11.2 kDa). Implementation of 450 nm blue LEDs circumvents E/Z isomerization with all other conditions remaining the same resulting in a clean polymerization pathway (10.2 kDa, 1.12 Đ). Similarly, monomer 5b also undergoes E/Z isomerization in addition to polymerization, however polymerization proceeds with the use of Ir(ppy)₃ and 450 nm blue LEDs without side reactions (21.4 kDa, 1.01 Đ (LS)). Monomer 1f exhibited very limited solubility in common organic solutions at 1 M as a result the polymerization was run in dilute conditions of 0.02 M resulting in a slightly decreased molecular weight compared to other polymers containing linker f (6.2 kDa, 1.01 Đ (LS), see S4).

Given the low polymer molecular weights and broad dispersity obtained in batch, we explored the use of a continuous flow reactor for the preparation of cyclobutane polymers, as we hypothesized it would improve both molecular weight and dispersity control as well as facilitate access to higher molecular weights with reduced reaction times.^39–43 Photochemical batch reactions traditionally suffer from poor irradiation of the reaction mixture and to overcome this limitation, low catalyst loading is often utilized to improve light transmission, consequently resulting in reduced polymerization rates.⁴⁴ Continuous flow allows for improved irradiation due to the increased surface area to volume ratio and shorter path length along with improved reaction temperature control, which facilitates an efficient and controlled photo-polymerization. Additionally, in contrast to traditional, batch photoreactors which are often plagued by scalability issues, continuous flow is a viable alternative for efficient, large-scale photo-polymerizations. Continuous flow strategies have been previously employed for [2+2] photocycloadditions of small molecules^24,27,45 and as post-polymerization ring-closing mechanisms of polymer reactive end-groups,⁴⁶ however this report describes the first continuous flow [2+2] polymerization. We constructed an in-house flow photoreactor comprised of 1/8-inch O.D. fluorinated ethylene-propylene (FEP) tubing (Idex Health & Sciences) wrapped around a steel container fitted to the Lucent360 Photoreactor (see reactor design scheme in Figure S1). Utilizing a 20 mL photoreactor, we varied the flow rate between 0.009 mL/min – 0.05 mL/min to access residence times of 6, 12, 18, 24 and 36 hours while remaining in laminar flow. Similar to the batch trials, a logarithmic increase in molecular weight occurs with time, however, molecular weights are substantially larger (7.9 – 181.3 kDa vs 5.5 – 64.3 kDa, Figure 4). Additionally, dispersities are remarkably smaller when compared to their batch counterparts indicating a more controlled polymerization event (Figure 4c). To demonstrate the capability for scale-up and continuous production, we conducted a steady-state experiment utilizing 4a, achieving 6.5 g per day production of 4a’ with a residence time of 24 hours, indicating a remarkable improvement in scalability when compared to the batch preparation of 4a’ (100–200 mg per polymerization).

Due to the nature of our tethered bis-cinnamate monomers, two main stereoisomers are accessible: β or δ with the δ isomer being preferred due to the intermolecular polymerization mechanism. The δ isomer, as reported by Jiang et. al, is the sole isomer in trans-cinnamic acid-based substrates.¹¹ Utilizing our method, we observed the δ isomer in all substrates, however for many polymers we observed a β:δ mix. β isomers are typically observed at ~4.4 and 3.8 ppm while δ appears at ~3.7 and 3.4 ppm. NMR analysis of polymers 4a’ and 4b’ indicates primarily δ isomers (approximately 85% and 80%, respectively), a result consistent with the observations made by Liao et. al. In order to confirm the β:δ ratio, we hydrolyzed polymers 4a’ and 4b’ using LiOH and THF/H₂O, and NMR analysis reveals a ratio of approximately 1:4 (see Figure S4.3 for ¹H NMR spectra of the hydrolyzed polymers). Polymers 4c’ (1:2 β:δ), 4d’ (1:2 β:δ), 4f’ (1:1.7 β:δ) and 4g’ (1:1.7 β:δ) contained increased β isomers as determined by NMR. Polymers 1a’, 1b’, 1c’, and 1g’ show sufficient aromatic proton peak separation in the NMR spectra to determine stereoisomer composition with all exhibiting approximately 1:1 β:δ. Additionally, polymers 5f’, 6d’, 6f’ all exhibit 1:1 β:δ, while 3f’ and 3g’ both exhibit 1:1.5 β:δ. However, for the remaining polymers, discerning stereoisomers via NMR spectroscopy proved to be difficult due to poor peak separation and shape. With our methodology, β isomers as well δ isomers can be accessed which has yet to be shown in cyclobutane polymers, thus allowing for more control over material composition and potential applications in the biomedical field.

Next, we performed thermogravimetric analysis on all polymers 1a’ – 6g’ (see S4.5 for detailed spectra and tabulated data) to determine the influence of electron density and linker type on decomposition temperature. Decomposition temperatures range between 109 °C – 375 °C indicating moderate to high thermal stability. Additionally, there is minimal impact of linker length or electron density on the decomposition temperature. We then performed differential scanning calorimetry on polymers 3a’ – 3g’ and 5a’ – 5g’ (see S4.6 for detailed spectra and tabulated data) to elucidate the influence of electron density and linker type on T_g. Electron density and aryl substitution patterns minimally impact the physical properties of the polymers. Unexpectedly, some polymers (3e’, 3f’ and 3g’) exhibit two glass transition (T_g) temperatures. This phenomenon is likely a result of a [2+2] reversion, which are known to occur at higher temperatures,⁴⁷ causing scission of the polymer chains and multiple T_g temperatures. Multiple T_g temperatures can also arise from difference in topologies (linear vs cyclic polymers), however we have no additional direct evidence for the formation of cyclic polymers in these studies. Increasing linker length from 3a’ to 3d’ increases main chain flexibility thus decreasing the T_g of 3d’ (−4.77 °C) compared to 3a’ – 3c’ (27.09 °C, 10.92 °C, 2.54 °C respectively). Linker change from e to f results in higher T_g of 3f’ (57.24 °C, 82.65 °C) when compared to 3e’ (44.36 °C, 73.26 °C) due to the increased rigidity of f.⁴⁸ A further increase in rigidity with linker g in 3g’ affords a higher T_g (84.76 °C, 238.45 °C) than both 3e’ and 3f’ (Figure 6a). Polymers 5a’ – 5g’, exhibit similar trends in T_g with increasing chain flexibility from linker a to d decreasing T_g. As rigidity increases from 5f’ to 5g’, T_g increases, however, increasing rigidity from 5e’ to 5f’ results in a decrease in T_g due to greater degree of polymerization of 5e’ comparatively (Figure 6b). Many of the polymers exhibited endothermic melting peaks in the first heating cycle, however, a T_m was not observed again in the second heating cycle. When repeated, the endothermic peaks were again present in the first heating cycle, indicating that the polymers do melt, but with the removal of the thermal history, crystallinity is no longer present in the second heating cycle. Finally, we performed a degradation study with polymer 4a’ in pH 4.1, 7.4 and 11.0. As expected, polymer 4a’ degrades more rapidly in both moderately acidic and basic conditions with a 75% loss in mass. GPC analysis reveals a decrease in molecular weight to 12–14 kDa for the samples at day 17 (see S6 for detailed degradation data).

Figure 6.

(a) Glass transition temperatures of 3a’ – 3g’ (b) Glass transition temperatures of 5a’ – 5g’.

CONCLUSIONS

In summary, we report an optimized synthesis of functionalized cyclobutane polymers via a solution state intermolecular [2+2] photocycloaddition, utilizing thioxanthone as the photosensitizer. Monomers of varying different electron density and linker compositions with monomers with olefin distances of greater than 5 Å polymerizing more readily with decreased termination events. All synthesized polymers exhibit high decomposition temperatures regardless of electron density or linker and DSC of polymers 3a’ – 3g’ and 5a’ – 5g’ show decreased T_g as linker flexibility increases (a – d) and increased T_g as linker rigidity increases (e – g). Continuous flow provides improved molecular weight and dispersity control and access to ultra-high molecular weights. This advancement ushers cyclobutane based polyesters, which have been historically limited to a few examples and only research scale quantities, to the forefront of development as new materials for potential use across industry sectors. We envision uses where biodegradation, be it in vivo applications or landfills, and high thermal properties are key design constraints.

ABBREVIATIONS

CBP

cyclobutane polymer