Pamela V. S.
Nylund
a,
Baptiste
Monney
b,
Christoph
Weder
b and
Martin
Albrecht
*a
aDepartment of Chemistry, Biochemistry & Pharmaceutical Sciences, University of Bern, Freiestrasse 3, 3012 Bern, Switzerland. E-mail: martin.albrecht@unibe.ch
bAdolphe Merkle Institute, University of Fribourg, Chemin des Verdiers 4, 1700 Fribourg, Switzerland
First published on 5th January 2022
Poly(lactic acid), PLA, which holds great promise as a biodegradable substitute of fossil resource-derived polyolefins, is industrially produced by the ring-opening polymerization of lactide using a potentially harmful tin catalyst. Based on mechanistic insights into the reaction of N-heterocyclic carbene (NHC) iron complexes with carbonyl substrates, we surmised and demonstrate here that such complexes are excellent catalysts for the bulk polymerization of lactide. We show that an iron complex with a triazolylidene NHC ligand is active at lactide/catalyst ratios of up to 10000
:
1, produces polylactide with relatively high number-average molecular weights (up to 50 kg mol−1) and relatively narrow dispersity (Đ ∼ 1.6), and features an apparent polymerization rate constant kapp of up to 8.5 × 10−3 s−1, which is more than an order of magnitude higher than that of the industrially used tin catalyst. Kinetic studies and end-group analyses support that the catalytically active species is well defined and that the polymerization proceeds via a coordination–insertion mechanism. The robustness of the catalyst allows technical grade lactide to be polymerized, thus offering ample potential for application on larger scale in an industrially relevant setting.
Currently, tin(II)octanoate (Sn(Oct2)) is the prevailing catalyst used for the industrial production of PLA by lactide ROP.9 However, Sn(Oct2) is cytotoxic and requires a high purity of the monomer, since water, lactic acid, and other impurities reduce the control of the polymerization considerably.7 In the search for environmentally friendly, non-toxic catalysts for the ROP of LA, complexes based on Earth-abundant iron have shown some promise and are attracting increasing interest. Pioneering work in the late 1990's revealed catalytic activity of simple iron salts such as halides, acetates, or alkoxides, yet these systems require unfavorably low monomer (LA) to initiator (I) ratios, elevated temperatures of around 180 °C, and/or very long reaction times (>100 h) to achieve high monomer conversions.10–13 Effective polymerizations with higher [LA]/[cat] ratios have been reported for iron alkoxide catalysts in toluene,14,15 as well as for melt polymerization at 130 °C, leading to a monomer conversion of ca. 90% within 24 h (I, Scheme 1).16 Initial attempts to enhance the catalytic performance of iron complexes by introducing salen-type or other multidentate N,N or N,O-based ligands were met with little success; ROP with these catalysts generally required very low monomer to initiator ratios ([LA]/[cat] 100:
1) and in many cases long reaction times.17–28 The Fe(N)4 system (II, Scheme 1) represents a notable exception, as this catalyst functions efficiently, even when used in low concentration.23 Recently, Thomas and co-workers developed a catalyst system that is based on an aminophenolate ligand and enables the stereoselective polymerization of LA at a [LA]/[cat] ratio of up to 800
:
1.29 The best performing iron catalyst to date was reported by Herres-Pawlis et al. in 2019 and is comprised of a guanidine-derived N,O-bidentate chelate. This complex (III, Scheme 1) shows a catalytic activity that is better than the industrially used Sn(Oct)2 and surpasses the activity of previously used iron catalysts by far.30
We reasoned that the use of N-heterocyclic carbenes (NHC)31–35 as spectator ligands might further enhance the robustness of the catalytically active species and hence improve the LA polymerization performance. Moreover, piano-stool iron NHC complexes36,37 have shown excellent catalytic activity in the conversion of carbonyl groups, e.g. in hydrosilylation reactions.38–40 Mechanistic studies indicated that the catalyst activation pathway of these complexes involves the initial interaction of the metal center with the carbonyl group, likely through CO dissociation.41 We suspected that a similar interaction of the iron center with the carbonyl group of LA might lead to substrate activation and cause ring-opening and polymerization. A classic coordination–insertion mechanism may then be accessible due to the presence of iodide or fortuitous water at the initiation stage, and subsequently the alkoxide of the growing polymer chain during propagation to promote the ring-opening of the coordinated LA.42,43 Here we demonstrate that iron complexes IV containing either an imidazolylidene or a triazolylidene44–46 as NHC unit constitute excellent catalyst precursors for LA polymerization with activities that exceed one of the industrially used Sn(Oct)2 and also considerably surpass the best iron catalysts reported so far.
![]() | ||
Fig. 1 NHC iron complexes 1–4 used for the polymerization of lactide as well as carbene precursors trz-H and trz-CO2. |
The catalytic activity of complex 1 for the ROP of rac-LA was initially tested by attempting the polymerization of technical grade monomer with 1,2-dichloroethane as solvent at a reaction temperature of 60 °C, and a 250:
1 molar ratio of [LA]/[cat] (Table 1, entry 1). However, 1H NMR spectroscopic monitoring of the diagnostic resonances for the methine and CH3 protons, which appear at δH = 5.1 and 1.6, respectively for LA, and downfield shifted at δH = 5.2 and 1.7, respectively, for PLA, revealed that no polymerization took place under these conditions (Fig. S4†). When the solvent was removed and the temperature was increased to melt the monomer, polymerization took place (entries 2, 3), even though the [LA]/[cat] ratio was increased twenty-fold to 5000
:
1. The reaction proceeded much faster at 150 °C, where a conversion of 82% was observed after 2 h, than at 130 °C, where the conversion remained modest (23%), even after 24 h.
Entry | Pre-catalyst | [LA]/[cat] | m(LA) (mg) | cat (μmol) | Time (h) | Conversionb (%) | M n (kg mol−1) | M w (kg mol−1) | Đ |
---|---|---|---|---|---|---|---|---|---|
a General conditions: bulk polymerization of rac-LA at 150 °C, unless noted otherwise; n.d. = not determined. b Determined by 1H NMR spectroscopy. c Determined by size-exclusion chromatography and corrected by the Mark–Houwink–Sakurada correction factor for PLA. d The polymerization was carried out in 1,2-dichloroethane (2.5 mL) at 60 °C. e The polymerization was carried out in bulk at 130 °C. f 29% conversion after 2 h. g 29% conversion after 2 h. | |||||||||
1d | 1 | 250![]() ![]() |
220 | 6.1 | 2 | 0 | n.d. | n.d. | n.d. |
2e | 1 | 5000![]() ![]() |
1600 | 2.2 | 24 | 23 | 24 | 42 | 1.8 |
3 | 1 | 5000![]() ![]() |
1600 | 2.2 | 2 | 82 | 48 | 89 | 1.9 |
4 | 1 | 5000![]() ![]() |
4800 | 6.6 | 2 | 57 | 38 | 63 | 1.6 |
5 | 1 | 1000![]() ![]() |
1600 | 11.1 | 2 | 91 | 21 | 53 | 2.5 |
6 | [FeCp(CO)2I] | 1000![]() ![]() |
1600 | 11.1 | 24 | 86f | 7.0 | 22 | 3.1 |
7 | trz-CO2 | 1000![]() ![]() |
1600 | 11.1 | 24 | 68g | 1.0 | 2.0 | 2.2 |
8 | trz-H | 1000![]() ![]() |
1600 | 11.1 | 24 | 3 | n.d. | n.d. | n.d. |
9 | — | — | 1600 | — | 24 | 2 | n.d. | n.d. | n.d. |
The observed activities, appreciable number-average molecular weights (Mn = 24 and 48 kg mol−1), and moderate dispersities (Đ = 1.8–1.9) established by size-exclusion chromatography (SEC) suggest that the catalytically active site is quite robust towards impurities that may be present in the technical grade LA, both at initiation and propagation stages. A small scale-up of the reaction (entry 4) led to a lower yield and a slightly reduced Mn under the same conditions, presumably due to less efficient stirring. When the [LA]/[cat] ratio was reduced to 1000:
1 (entry 5), the rate of conversion was increased and Mn was reduced in comparison to the reaction carried out with the higher [LA]/[cat] ratio (entry 3), suggesting that (a fragment of) the iron complex may also act as initiator.
In order to investigate the function of the catalyst in more detail, control experiments with a carbene-free iron complex ([FeCp(CO)2I]) as well as with the carbene ligand (without the metal center) were performed. When the melt-polymerization was carried out with the carbene-free iron complex [FeCp(CO)2I] at 150 °C, the rate of polymerization was low, with only 29% of monomer conversion within 2 h (86% after 24 h, entry 6) compared to >90% conversion with complex 1 under otherwise identical conditions (entry 5). Moreover, the molecular weight was significantly reduced (Mn = 7 vs. 21 kg mol−1) and the dispersity higher (Đ = 3.1 vs. 2.5 with complex 1), suggesting a prominent role of the NHC ligand in keeping a well-defined and active catalytic site. Since it is known that free carbenes can initiate the polymerization of LA,49 the CO2-protected triazolylidene, trz-CO2 (Fig. 1),41 was evaluated as catalyst to probe whether complex 1 acts as a carbene releasing agent. Under the same conditions as applied for 1, trz-CO2 displayed some activity, but a reaction time of 24 h was required to reach a modest 68% conversion (entry 7). These results suggest that the catalytically active species in complex 1 is indeed an NHC–iron complex, rather than just the free carbene or the carbene-free iron center. Hardly any polymerization took place when the free triazolium salt trz-H (ref. 39) was used (entry 8), and no polymerization occurred in the absence of any catalyst (entry 9).
Further insights into the polymerization process were obtained by monitoring the Mn and the Đ of the PLA produced by the bulk polymerization of rac-LA at 150 °C catalyzed by complex 1 as a function of reaction time (Fig. 2a, Tables S1 and S2†) and monomer conversion (Fig. 2b). The Mn values saturate after an initial increase in the first minutes and scale linearly with the monomer conversion, up to a conversion of ca. 70%. Up to this point, the dispersity remains relatively low for melt conditions (Đ = 1.6).48 At higher conversion the dispersity increases and the Mn does not further grow, suggesting that the reaction kinetics change, perhaps due to the high viscosity and inhibited stirring, which favors side reactions such as chain transfer, termination and transesterifications. Similar effects have been noted when using Sn(Oct)2 as catalyst.50 The dispersity is comparable for reactions carried out with [LA]/[cat] ratios of 1000:
1 and 5000
:
1 until about 50% conversion, while the molecular weights achieved with the higher [LA]/[cat] ratio are consistently higher by a factor of roughly two. The observed molecular weights correspond to only ca. 40% of the theoretical values for the reactions carried out with a [LA]/[cat] ratio of 1000
:
1 and only about 20% of the theoretical values for the reactions carried out with a higher [LA]/[cat] ratio of 5000
:
1 (Tables S1 and S2†). This discrepancy reflects that the polymerization is clearly not living, and suggests that either early termination is at play, possibly due to reactions with impurities in the technical grade monomer, or that the molecular weight is governed, at least to some extent, by the concentration of initiating species that are independent of the catalyst.
To investigate how impurities in the technical grade LA impact the polymerization and gain further insights regarding the nature of the initiator, recrystallized LA was used for ROP with and without BnOH that was added as an auxiliary initiator (Tables S3 and S4†).30 These experiments were carried out with [LA]/[1] = 1000. Intriguingly, the polymerization of recrystallized LA in the absence of BnOH was slower than that of the technical grade monomer, while the addition of 1 eq of BnOH with respect to iron brings the conversion rate to a similar level (Fig. 3). These data allow the conclusion that iron complex 1, BnOH, and impurities in the technical grade monomer all serve as effective initiators for the iron-catalyzed LA polymerization. This is further supported by molecular weight data. When recrystallized LA was polymerized without BnOH, the Mn was close to the theoretical value calculated from [LA]/[1] up to a conversion of about 50% (Table S3†), consistent with initiation by the catalyst. As discussed above, under otherwise identical conditions, the polymerization of technical grade LA afforded polymers with an Mn of slightly less than half the theoretical value (Table S1†), suggesting that additional initiating species are in play. The deliberate addition of 1 eq BnOH as initiator afforded polymers whose Mn was in between those of the other series (technical grade or recrystallized LA), supporting the conclusion that BnOH and 1 initiate the reaction (Table S4†). Under all conditions, the polydispersity is generally higher than 1.5, except at short reaction times. At high monomer conversion the molecular weights drop and the distributions broaden, suggesting that depolymerization involving a back-biting mechanism and trans-esterification reactions become important.
In order to produce low-molecular weight polymers whose end groups can be analyzed by matrix-assisted laser desorption ionization time-of-flight (MALDI-ToF) spectrometry, crystallized rac-LA was polymerized in the melt at 150 °C with 1 as a catalyst and a [LA]/[cat] ratio of 50:
1. In the absence of any auxiliary initiator, OH end-groups were identified by a signal at 903.10 Da (calcd. for (LA)6–OH–Na+ 904.25 Da), suggesting the presence of residual water as initiator (Fig. S5†). Notably, signals that can be attributed to cyclic oligomers (LA)n were also identified, e.g. at 887.78 Da, (calcd. for (LA)5Na+ 887.24 Da). These oligomers were present in the MALDI-ToF spectrum of a polymerization experiment with a higher monomer/catalyst ratio of 1000
:
1 at 1 min (Fig. S6†) and became dominant after prolonged reaction times (Fig. S7†). When BnOH was used as initiator at a 50
:
1
:
1 [LA]/[cat]/[I] ratio, BnO end-groups were indicated for most signals, e.g., 995.51 Da for H–(LA)6–OBn–Na+ (calcd. 995.84 Da; Fig. S8†). While no major signals indicative of macrocyclic products were detectable, the MALDI-data reveal signal separations for the oligomers of 72 Da, viz. one lactide monomer, hinting to transesterification reactions.20,27,51,52 Such post-polymerization modifications may be rationalized by the high activity of the catalyst as well as the elevated reaction temperatures.53
The MALDI-ToF spectra also reveal signals that can be attributed to small oligomers containing complex 1 without CO and I (666.55 Da, calculated for H–(LA)2–Fe(Cp)(trz) 666.25 Da) and without CO, I and Cp ligands (e.g. 890.07 Da, calcd. for H–(LA)4–Fe(trz) 890.30 Da; Fig. S9 and S10†). These data indicate that complex 1 or water can indeed initiate the polymerization, although there is no support for iodide as initiator. Notably, also signals attributed to (LA)2–trz-Na+ were detected (568.17 Da, calcd. 568.26 Da, Fig. S10†), which may indicate the contribution of free carbene as initiating species.54
A selected polymer sample produced by catalysis with complex 1 according to the conditions of entry 3, Table 1 was analyzed by thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC). The thermal decomposition of the polymer starts at 250 °C, well within the expected 230–260 °C range (Fig. S11†), while the DSC trace exhibits the typical features of amorphous poly(rac-lactide) with a glass transition temperature of around 52–55 °C and no melting or crystallizing events (Fig. S12†).55 Moreover, the polymer films do not show any significant coloration due to residual iron complex (Fig. S14†).
Because of the high activity and robustness of the polymerization catalyst derived from complex 1 with technical grade LA, related complexes 2–4 with different ligand architectures were investigated under comparable conditions (Fig. 1). Complexes 2–4 all feature the same iPr wingtip substituents but differ either in the heterocyclic carbene ligand (triazolylidene vs. imidazolylidene in 2vs.3), or in the arene ligand (Cp vs. Cp* in 3vs.4). The pertinent time-conversion profile indicates that the initial polymerization rates are lower for complex 2 than the parent complex 1 containing an aryl wingtip group, while the two imidazolylidene-based complexes 3 and 4 show similar initial activity (Fig. 4a). Notably, all three complexes 2–4 reach lower final conversions than 1 (62 ± 5% vs. 82%; Table 2). These different activities are also reflected in the apparent polymerization rate constants kapp, which were determined by linear regression of the polymerization data in a semilogarithmic plot (Fig. 4b, Tables S5–S7†).56 Complexes 1, 3, and 4 show all similar high rates with kapp as high as 8.5 × 10−4 s−1 for 1 (Table 2, entry 1). In comparison, exchanging the (Mes)(Bu) wingtip pattern in 1 for two iPr groups in complex 2 reduces kapp to 2.5 × 10−4 s−1 (entry 2). Introducing an imidazolylidene recovered activity (kapp = 7.3 × 10−4 s−1, entry 3), while the substitution of the Cp ligand with Cp* had no significant effect (kapp = 8.0 ± 0.6 × 10−4 s−1, entry 4). Notably, the rate differences observed as a function of ligand architecture reinforces the conclusion that a NHC iron species serves as initiator and polymerization catalyst.
Entry | Catalyst | k app (10−4 s−1) | Conversionb (%) | Conditions |
---|---|---|---|---|
a Conditions A: rac-LA (1.6 g, 11.1 mmol), catalyst (2.2 μmol), [LA]/[cat] = 5000![]() ![]() ![]() ![]() |
||||
1 | 1 | 8.5 ± 0.6 | 82 | A, 2 h |
2 | 2 | 2.5 ± 0.2 | 57 | A, 2 h |
3 | 3 | 7.3 ± 0.3 | 63 | A, 2 h |
4 | 4 | 8.0 ± 0.6 | 67 | A, 2 h |
5 | II | 0.13 | 75 | A, 30 hc |
6 | 1 | 85.3 ± 3.3 | 82 | B, 4 min |
7 | III | 43.5 ± 3.5 | 68 | B, 1.6 mind |
In order to benchmark the catalytic activity of complex 1 against state-of-the-art catalysts (cf.Scheme 1), polymerizations were performed neat at 150 °C, yet at a lower [LA]/[cat] ratio of 500:
1. Under these conditions, the apparent polymerization constant kapp for complex 1 is one order of magnitude higher than under more dilute conditions (kapp = 85 × 10−4 s−1; Table 2, entry 6, see also Fig. S15, Table S8†). This activity is twice as high as that of complex III, which has been recognized as the fastest and most robust iron catalyst for the LA polymerization to date (kapp = 44 × 10−4 s−1, entry 7).30 The other iron complex reported to be highly active, complex II, displays an almost 100 times lower activity than 1 (kapp = 0.13 × 10−4 s−1, entry 5 vs. entry 1).23 These data underpin the considerable potential of complex 1 in LA polymerization.
Kinetic studies were performed with complex 1 by carrying out the polymerization or rac-LA at constant monomer concentration and varying catalyst loadings (Table 3, Fig. 5a). Under these conditions, the polymerization is clearly not controlled and the dispersity rather high. The apparent polymerization rate constant, kapp (Fig. 5b, Table S9†) displayed values ranging from 53.9 to 0.9 10−4 s−1 for [rac-LA]/[cat] ratios between 1000:
1 and 10
000
:
1. The rate constant of propagation kp was obtained by plotting kappvs. the catalyst concentration (Fig. 5c, Table S9†).30 The linear fit suggests that the polymerization is first-order with respect to complex 1, in agreement with a molecularly well-defined active species that operates according to a coordination–insertion mechanism.42,57 An activated monomer mechanism would principally be conceivable as well, although the coordination insertion mechanism seems more likely based on the high monomer
:
initiator ratio, the occurrence of side reactions, the absence of a strong protic acid, and the insertion of single lactide units as shown with the 72 amu difference of polymers in MALDI-ToF spectrometry.58 Such a mechanism is also supported by the MALDI-ToF results, which reveal the iron NHC species as end group (Fig. S9 and S10†). However, the observed molecular weights are much lower than the ones calculated on the assumption that the iron complex serves as precursor for both initiation and polymerization (Table 3), which is in agreement with initiation by residual water or other impurities in the technical grade LA (see above). The propagation rate constant for the polymerization of LA with complex 1, kp = 0.74 (±0.04) M−1 s−1, exceeds that of the fastest iron catalyst known to date (viz.III, kp = 0.55 (±0.01) M−1 s−1) under identical reaction conditions.30 Importantly, this activity is also almost one order of magnitude higher than that of the industrially used Sn(Oct)2 catalyst (kp = 0.084 (±0.002) M−1 s−1). Moreover, the absence of induction period and the high conversions (25–70%) within the first 4 minutes of the reaction suggests that both initiation and propagation are fast.
Entry | [LA]/[cat] | Conversion/yieldb (%) | k app (10−4 s−1) | theor Mnc (kg mol−1) | obs Mnd (kg mol−1) | M w (kg mol−1) | Đ |
---|---|---|---|---|---|---|---|
a General conditions: bulk polymerization of rac-LA (1.6 g, 11.1 mmol), cat (1.1–11.1 μmol), at 150 °C, 2 h. b The conversion was determined by 1H NMR spectroscopy, isolated yields gravimetrically after precipitation with ethanol; all values are averages of two independent polymerizations. c The theoretical number-average molecular weight was calculated from the monomer to catalyst ratio and the conversion of the reaction on the assumption that the catalyst initiates the polymerization. d Determined by size-exclusion chromatography and corrected by the Mark–Houwink–Sakurada correction factor for PLA. | |||||||
1 | 1000![]() ![]() |
91/69 | 53.9 ± 4.1 | 130 | 21 | 53 | 2.5 |
2 | 1500![]() ![]() |
86/80 | 33.0 ± 0.4 | 190 | 26 | 59 | 2.3 |
3 | 2775![]() ![]() |
74/64 | 12.3 ± 0.4 | 300 | 30 | 60 | 2.0 |
4 | 5000![]() ![]() |
82/67 | 8.5 ± 0.6 | 590 | 48 | 89 | 1.9 |
5 | 10![]() ![]() ![]() |
42/n.d. | 0.9 ± 0.1 | 600 | 17 | 49 | 2.8 |
The end group analysis was performed by MALDI-ToF on a Bruker Autoflex III Smartbeam equipped with a 337 nm Smartbeam laser in linear mode. The instrument was calibrated with the MSCAL1 calibration kit from Sigma Aldrich. Data were acquired by the FlexControl software and evaluated by the FlexAnalysis software (Bruker). A THF solution of trans-2-[3-(4-tert-butylphenyl)-2-methyl-2-propenylidene]malononitrile (DCTB) (10 mg mL−1) was used as matrix. The sample was dissolved in THF (100 μL, approx. 1 mg mL−1), of a solution of NaI (5 μL solution in iPrOH/H2O 1:
1 at 1 mg ml−1) was added. A 0.5 μL sample solution was mixed with 0.5 μL of the matrix solution directly on the target spot. The spotted sample target was left to dry at room temperature. Spectra were acquired using 30% laser power, adding up 4–5 × 200 shots. The laser repetition rate was 100 Hz.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1cy02143e |
This journal is © The Royal Society of Chemistry 2022 |