Computational and Experimental Evaluation of Peroxide Oxidants for Amine–Peroxide Redox Polymerization

Abstract

Amine–peroxide redox polymerization (APRP) is the prevalent method for producing radical-based polymers in the many industrial and medical applications where light or heat activation is impractical. We recently developed a detailed description of the APRP initiation process through a combined computational and experimental effort to show that APRP proceeds through S_N2 attack by the amine on the peroxide, followed by the rate-determining homolysis of the resulting intermediate. Using this new mechanistic understanding, a variety of peroxides were computationally predicted to initiate APRP with fast kinetics. In particular, the rate of APRP initiation can be improved by radical and anion stabilization through increased π-electron conjugation or by increasing the electrophilicity of the peroxy bond through the addition of electron-withdrawing groups. On the other hand, the addition of electron-donating groups lowered the initiation rate. These design principles enabled the computational prediction of several new peroxides that exhibited improved initiation rates over the commonly used benzoyl peroxide. For example, the addition of nitro groups (NO₂) to the para positions of benzoyl peroxide resulted in a theoretical radical generation rate of 1.9 × 10⁻⁹ s⁻¹, which is ~150 times faster than the 1.3 × 10⁻¹¹ s⁻¹ radical generation rate observed with unsubstituted benzoyl peroxide. These accelerated kinetics enabled the development of a redox-based direct-writing process that exploited the extremely rapid reactivity of an optimized redox pair with a custom inkjet printer, capable of printing custom shapes from polymerizing resins without heat or light. Furthermore, the application of more rapid APRP kinetics could enable the acceleration of existing industrial processes, make new industrial manufacturing methods possible, and improve APRP compatibility with biomedical applications through reduced initiator concentrations that still produce rapid polymerization rates.

Graphical Abstract

INTRODUCTION

Approximately 45% of all polymer products are manufactured using radical polymerization (RP), of which a significant portion is initiated by redox initiators.^1–3 Redox RP uses the chemical energy stored in a reductant–oxidant pair to produce initiating radicals. This provides several important advantages including energy-efficient room-temperature (RT) activation, an unrestricted product shape and size, and the ability to activate polymerization without light.^1–3 Most redox initiators contain metal salts with metal cations such as Fe²⁺, Cr²⁺, V²⁺, and Ti³⁺ that activate the weak O–O bond in various peroxides such as peroxydisulfates (PDSFs), peroxydiphosphates, and permanganates. Fenton’s reagent is the most common example in which Fe²⁺ activates peroxide through the reaction Fe²⁺ + HOOH → Fe³⁺ + OH⁻ + OH^•. The benefit of metal salt-containing initiators is their wide variety, easy tunability, and sufficiently fast kinetics. However, the solubility and toxicity of metal ions remain a major disadvantage that generally limits their use to emulsions and suspensions in aqueous media.^1–3

Amine–peroxide redox polymerization (APRP) uses metal-free amine reductant–peroxide oxidant pairs as initiators to achieve additional benefits over metal-containing initiators, including biocompatibility⁴ and facile implementation in solvent-free bulk polymerizations. This has led to the wide application of APRP and its extremely active development since its discovery in the 1950s, as evidenced by the over 45,000 APRP-related patents filed since 1975.⁵ Despite their popularity, some of the common amine reagents have been found to be toxic.^6–8 To mitigate their potential toxicity, amines have been modified with reactive groups that can bind to oligomers or with long alkyl chains to reduce leaching from the polymer.^7,9,10 Amine-free redox systems have also been developed to avoid amine toxicity by using novel reagents that are likely to be more biocompatible.^11–14 However, none of these systems have achieved a level of success similar to amine–peroxide initiators for various reasons including their use of metals and unwanted coloration. APRP’s success is especially evident in biological applications such as biomaterials synthesis, which has in particular exploited APRP’s unique advantages in dental and orthopedic applications, where restricted conditions render thermal and photoinitiation processes impractical.^6,15–21 These advantages have also motivated APRP’s investigation for emerging applications in photoactivated APRP initiator development,²² frontal polymerization,²³ self-healing materials,²⁴ and interfacial polymerization.²² Despite the significant impact of APRP, no fundamental study examining the molecular features that determine initiation rates had been reported until 2019 and that study focused on the amine reductant.²⁵ This gap presents an opportunity to investigate how the choice of the amine–peroxide pair might be leveraged to optimize the rate of polymerization.

We previously conducted an extensive study of APRP via a combined computational and experimental investigation (J. Am. Chem. Soc. 2019, 141, 6279–6291).²⁵ In this study, we proposed and validated an APRP initiation mechanism that allowed us to accurately predict new amines for accelerated APRP. Using this approach, we identified, synthesized, and tested N-(4-methoxyphenyl)pyrrolidine (MPP), which generated radicals ~20 times faster than the previous state-of-the-art tertiary aromatic amines, making it the fastest amine implemented in APRP to date. While this previous study delved into the reactivity of structurally varied amine reductants with benzoyl peroxide (BPO), it did not examine the potential of pairing amines with various peroxides of different reactivities.

Previous studies regarding the reactivity of peroxide oxidants within the APRP framework have been limited and primarily observational. Moore and co-workers evaluated five peroxides (BPO, lauroyl peroxide, methyl-ethyl-ketone peroxide, tert-butyl peroxide, and tert-butyl peroxybenzoate)²⁴ within the APRP framework for use in self-healing composites, discovering that when added to the amine, three out of the five peroxides had either no effect or adverse effects on the polymerization rate and activation temperature, the exceptions being BPO and methyl-ethyl-ketone peroxide. He et al. investigated four peroxides [BPO, tert-butyl peroxybenzoate, dicumyl peroxide (DCPO), tert-butyl peroxide] and found that polymerization efficiency was inversely correlated with thermal stability during photopolymerization.²⁶ However, the amines used in this study were the tertiary, nonaromatic diazabicyclo[5.4.0]undec-7-ene (DBU) amine and secondary amines, which were found to be either poor initiators or even inhibitors of APRP.^22,25 Moore’s study was less focused on the peroxides’ reactivity and instead aimed at investigating self-healing properties of polymers that incorporated separately encapsulated amine–peroxide pairs. He’s study utilized APRP-inhibiting amines, making their conclusions regarding peroxide performance less relevant to identifying superior amine–peroxide pairs for APRP. Additionally, these two studies combined only examined five distinct peroxides, which are insufficient for characterizing the peroxide chemical space. Ultimately, the assessment of peroxide performance for APRP has been limited. The lack of a fundamental understanding regarding the effects that govern peroxide reactivity within APRP motivated this structure–activity relationship study of the peroxide chemical space. Understanding the molecular features that dictate initiation kinetics may enable the design of superior peroxide initiators for accelerated APRP.

In this contribution, we utilized the fundamental understanding elucidated by our previous study²⁵ to investigate peroxides that could further improve kinetic rates to expand the scope of APRP. Our computational analysis of 15 peroxide oxidants elucidates how structural variations dictate radical generation kinetics and peroxide stability, the former of which is supported by real-time Fourier-transform infrared (FTIR) spectroscopy. All the peroxides examined performed in polymerization experiments according to our computational predictions except one, phthaloyl peroxide, whose polymerization behavior significantly deviated from our computational prediction. Although this anomalous peroxide was predicted to generate radicals faster than all other peroxides, we observed only a negligible amount of polymerization along with significant coloration of the resin. This inconsistency between our predicted and measured results for this peroxide was investigated and explained by a new mechanism involving an outer-sphere electron-transfer (ET) step as opposed to our previous finding that the operative APRP mechanism proceeded through inner-sphere ET for other amine–peroxide pairs.²⁵ Our results demonstrate a unique instance where the identities of the reductants and oxidants determine which of these two branches in the mechanistic path of an inner- or outer-sphere ET step is taken, while also providing researchers valuable guidance toward the development of novel APRP chemistries. Last, the results of our two studies of APRP guided the development of a redox-based direct-writing process that leveraged the extremely rapid reactivity of an optimized APRP pair with a custom inkjet printer.

METHODS

Computational Methods.

All DFT calculations were performed with Gaussian 16.03.A using the MN15 density functional with the 6–31+G(d,p) basis set. In our previous study, this level of theory performed the best among combinations of 13 commonly used functionals when benchmarked against the highly accurate but computationally expensive CCSD(T) and CBS-QB3 methods, resulting in a root-mean-square error (RMSE) of 1.4 kcal/mol for activation energies.²⁵ Gas-phase frequencies were computed at the experimental temperature (298 K) and used to predict zero-point energies, enthalpies, and entropies. These predicted frequencies were also used to confirm local minima for reactant, intermediate, and product states, as well as saddle points for transition states (TSs). To include solvation effects, we applied the SMD implicit solvent model with solvent parameters describing ethyl acetate, which mimic the dielectric properties of the (meth)acrylate bulk resin. In order to more accurately capture the bulk resin environment in which librational modes are hindered by neighboring molecules, we removed translational and rotational entropies from the free-energy expression and reported free energies G_v(T) that only include entropic contributions from the vibrational states. This correction was previously shown to produce more accurate reaction and activation free energies for reactions in solvents in comparison to experimental results, while computed changes in free energies without this correction overestimated entropic contributions.²⁵ More comprehensive computational details are provided in Section S1.1 of the Supporting Information.

Experimental Methods.

Materials.

N,N-Dimethylaniline (DMA), BPO, tert-butyl peroxybenzoate, and DCPO were purchased from Sigma-Aldrich (Milwaukee, WI). tert-Butyl peroxide was received from Fisher Scientific. Ammonium PDSF, dilauryl peroxide, di-tert-butyl peroxide, and 2,3-dimethyl-2,3-diphenylbutane were obtained from Acros Organics (Geel, Belgium). All peroxides were obtained with the highest purity available of at least 98% and used as received except for water-stabilized hydrogen peroxide (30% in water, Fisher Scientific) and tert-butyl hydroperoxide (70% in water, Acros Organics). The monomers used are di(ethylene glycol) ethyl ether acrylate (DEGEEA, Sigma-Aldrich, 90% purity, 1000 ppm MEHQ) and bisphenol A ethoxylate diacrylate (BisEDA, Sigma-Aldrich, Quality Level 200, M_n 512, EO = 2, 1000 ppm MEHQ). All other chemicals were of reagent grade and used without further purification. Phthaloyl peroxide and MPP were synthesized according to the respective reported procedures.^25,27,28

Fourier Transform Infrared Spectroscopy.

The conversion of the monomer by ambient bulk redox-initiated polymerization was monitored in real time with an FT-IR spectrophotometer (Nicolet Magna-IR Series II, Thermo Scientific, West Palm Beach, FL) by monitoring the C=C stretching absorption band at 1637 cm⁻¹ for 20 min in OMNIC software. We chose DEGEEA monoacrylate to reduce a confounding factor of autoacceleration as opposed to cross-linking multifunctional monomers that exhibit autoacceleration. The spectrophotometer was equipped with an MCT/A detector, and the parameters on the FT-IR were set to 2 scans, a resolution of 16, an optical gain of 1, an optical velocity of 1.8988 cm/s, and an optical aperture of 15. We studied the polymerization kinetics initiated with two separate batches of the resin that respectively contain 3 mol % peroxide oxidant and DMA reductants. On a horizontal NaCl salt plate, 15 μL of the peroxide resin was first deposited, to which another 15 μL of the DMA resin was added. This solution was then mixed with a micropipette tip. Another NaCl salt plate was then placed on top of the mixture, after which data acquisition began immediately. A clear baseline of zero conversion during an induction period of ~60 s preceding each polymerization was observed for every sample.

Nuclear Magnetic Resonance.

¹H NMR and ¹³C NMR spectra were recorded on a Bruker 400 MHz NMR spectrometer. Proton chemical shifts are expressed in parts per million (δ) using tetramethylsilane as an internal standard. The δ scale was referenced to deuterated solvents, as indicated in each respective measurement.

UV–Visible Spectroscopy.

The UV–visible (UV–vis) spectra of redox pairs in dimethylformamide were recorded in PMMA cuvettes with a 1 cm optical path length in a UV–vis spectrophotometer (Thermo-Fischer Scientific) after baseline correction with a blank run.

Direct-Writing System.

Our custom direct-writing system includes a Raspberry Pi 3+ that controlled two 12 V DC peristaltic pumps (AE1207) to pump the resins and two stepper motors (SM15L) to enable X- and Y-axis directional control of the writing nozzle. A 3 mm inner diameter × 5 mm outer diameter silicone tubing was used to transport the resins to the mixing nozzle. Custom python software was developed to enable APRP direct writing with the Raspberry Pi.

RESULTS AND DISCUSSION

Amine–Peroxide Redox Mechanism.

The APRP initiation mechanism is detailed in Scheme 1. Our previously derived kinetic model for the APRP mechanism correlated the computationally calculated radical generation rate constant (k_r) and the experimentally determined polymerization rate (R_p) with good agreement (R² = 0.80). The governing rate equation $k_{\text{r}}=k_2\left(k_1^{1/2}/k_{-1}^{1/2}\right){[\text{AM}]}^{1/2}{[\text{PO}]}^{1/2}$ is first-order in the hemolysis (HM) rate (k₂) and half-order in the overall S_N2 rate (defined as k₁/k₋₁), thus supporting the claim that HM is the rate-determining step. For this model, S_N2 is assumed to be reversible because of its endergonic nature, and HM is assumed to have no distinct TS based on its monotonic potential energy surface (vide infra). Because these assumptions resulted in a good correlation of R² = 0.80 for various amines, the kinetic model was expected to be sufficient for predicting the reactivities of peroxides reacting within the APRP mechanism. Herein, the reactivities of various peroxides are evaluated by applying our kinetic model and comparing computational predictions to experimental data. Kinetic constants are calculated from computed free-energy activation barriers of each elementary step using Eyring’s kinetic equation.

Scheme 1. APRP Mechanism When Using DMA as the Amine Begins with S_N2 Nucleophilic Attack of the Peroxide (PO) by the Amine (AM)^a.

^aThe resultant TA and OA intermediates mediate an inner-sphere ET where AM becomes singly oxidized and the leaving OA becomes singly reduced. Following S_N2 attack of PO by AM, the metastable TA intermediate undergoes HM to produce an ARC and an OR, which is the first initiating radical generated from the amine–peroxide reaction. We will show below that these two reactions (S_N2 and HM) govern the overall initiation kinetics in the APRP mechanism, where HM is the rate-determining step. Meanwhile, reactive intermediates resulting from the S_N2 and HM reactions react with each other; OA and ARC may undergo a barrierless proton transfer at the α-carbon of ARC to produce a weak acid (OH) and an initiating α-aminoalkyl radical (AAR).²⁵

BPO was chosen as the reference peroxide for this study as nearly all pre-existing APRP applications incorporate BPO as an oxidant. DMA was selected as the reference amine reductant as it was determined to be a relatively efficient redox initiator and its small molecular size permits efficient computation. While this study examines APRP kinetics for different peroxides, the identity of the amine (DMA) remains constant in every reaction studied unless otherwise noted.

Acyl Peroxide.

The reaction between the BPO and DMA reference redox pair (Figure 1) was computed to have a forward S_N2 free-energy barrier ($\Delta G_{{\text{S}}_{\text{N}}2}^{‡}$)of 16.2 kcal/mol and a free energy of reaction ($\Delta G_{{\text{S}}_{\text{N}}2}^0$) of 13.5 kcal/mol and thus a small reverse barrier of 2.7 kcal/mol (Table 1). The HM reaction exhibits a monotonic potential energy surface so that the HM free energies of reaction and activation are equal ($\Delta {G_{\text{HM}}}^0=\Delta G_{\text{HM}}^{‡}$) and calculated to be 25.6 kcal/mol. These activation free energies lead to a k_r of 1.3 × 10⁻¹¹ s⁻¹ for BPO. The major contribution to the S_N2 barrier arises from torsion about the O–O bond to enable backside attack of the O–O antibonding ${\sigma }_{\text{O}-\text{O}}^{*}$ orbital (O–O*).²⁵

Figure 1.

Free-energy profile of the APRP initiation between DMA and BPO, leading to the generation of the first initiating radical. Free energies are in units of kcal/mol and calculated at the MN15/6-31+G(d,p)/SMD-ethyl acetate level of theory.

Table 1.

MN15 Free-Energy Barriers, Radical Generation Rate Constants (k_r), and BDEs with the Corresponding Decomposition Temperatures (T_d) for Peroxides Investigated in This Study^a

category	peroxide	$\Delta G_{{\text{S}}_{\text{N}}2}^{‡}$	$\Delta G_{\text{HM}}^{ǂ}$	k r	BDE (Td)b
acyl	BPO	16.2	25.6	1.3 × 10−11	40.0 (80–85)30,31
	MeBPO	17.3	25.0	1.4 × 10−11	39.3 (85)31
	MeOBPO	18.3	26.5	9.8 × 10−13	40.5
	ClBPO	14.9	24.8	1.8 × 10−10	39.8 (75)c31
	NiBPO	12.6	24.7	1.9 × 10−10	41.7
	NathPO	16.3	24.6	4.5 × 10−11	38.8
	AcPO	21.4	23.9	4.3 × 10−12	38.1
	PhthPO	8.8	33.8	9.4 × 10−6	30.2 (90–110)27
alkyl	DCPO	45.8	26.2	1.9 × 10−22	38.8 (75–130)31
	ClDCPO	44.5	25.5	2.8 × 10−15	39.2
carbonate	BEPC	14.3	24.4	1.7 × 10−9	37.5 (35–40)d30
	BPPC	9.2	22.2	1.8 × 10−6	36.7
inorganic	PDSF	32.6	40.0	5.0 × 10−21	21.3c
	HOOH	28.6	36.5	9.5 × 10−30	51.0

^aFree-energy barriers and BDEs are provided in kcal/mol, the rate constants in s⁻¹, and T_d in °C.

^bT_d is estimated based on self-accelerating decomposition temperature (SADT) and safe processing temperature (SPT).

^cT_d of ClBPO is estimated from the SPT of di(2,4-dichlorobenzoyl) peroxide.

^dT_d of BEPC is estimated from the SADT of diacetyl peroxydicarbonate and dimyristyl peroxydicarbonate.

First, peroxides resulting from the functionalization of BPO were examined. We exclusively modified the para groups of the phenyl rings of BPO as these positions would affect the peroxide reactivity more significantly than modification at the meta position and without inducing steric clashing through Pauli repulsion in contrast to modification at the ortho position. To maintain the symmetry of the peroxides, para substitutions were applied to both phenyl rings. We selected methyl and methoxy groups (MeBPO and MeOBPO) as electron-donating groups (EDGs) and chloro and nitro groups (ClBPO and NiBPO) as electron-withdrawing groups (EWGs). These four distinct substitutions provide a range of variations that allow the evaluation of EWG/EDG contributions to peroxide reactivity (Figure 2). We expect that methyl and chloro groups primarily affect the electronic structures of the peroxides via an inductive effect, while methoxy and nitro groups do so via a resonance effect.

Figure 2.

Diacyl peroxides investigated in this study. Groups shown in magenta denote electron-withdrawing groups, those in red denote electron-donating groups, and those in blue indicate aromatic groups.

Our computations show that EDG substitutions increase the S_N2 barrier by 1–2 kcal/mol and decrease the stability of the intermediates by ~1 kcal/mol, resulting in a lower k₁ with a similar k₋₁. Specifically, MeBPO had forward and reverse S_N2 barriers of 17.3 and 3.0 kcal/mol, respectively, while MeOBPO had forward and reverse barriers of 18.3 and 3.6 kcal/mol, respectively (Table 1). This results because electron donation from the EDG reduces the electrophilicity of the oxygens of the peroxy bond, making S_N2 attack on the electron-rich oxygens by the amine more difficult. The atomic charges calculated via atomic polar tensor (APT) population analysis were −0.64, −0.66, and −0.66 e for the oxygens of the O–O bond of BPO, MeBPO, and MeOBPO, respectively. This confirms the hypothesis that addition of an EDG to the peroxide increases the electron density on the O–O bond to inhibit attack by the amine lone pair. Electron donation also results in less favorable thermodynamics for S_N2 relative to BPO (by 0.8 kcal/mol for MeBPO and 1.2 kcal/mol for MeOBPO) by increasing localized cationic charge on the nitrogen of trialkyl ammonium (TA) (0.16, 0.17, and 0.19 e for BPO, MeBPO, and MeOBPO, respectively) and by increasing localized anionic charge on the oxygen of the oxy anion (OA) (−1.23, −1.24, and −1.25 e for BPO, MeBPO, and MeOBPO, respectively). As a result, EDG substitutions disfavor S_N2 reactions to a small but non-negligible extent. The effect of an EDG on HM is less straightforward. Relative to BPO, the HM barrier decreased by 0.6 kcal/mol for MeBPO and increased by 0.9 kcal/mol for MeOBPO, the difference being even smaller than the changes to the S_N2 reaction barriers. Additionally, this more subtle trend was not captured by the APT population analysis; in all relevant intermediates [OA, oxy radical (OR), and TA], both MeBPO and MeOBPO exhibited a greater electron density on the oxygens from the dissociated O–O bond relative to BPO. Because MeBPO and MeBPO exhibit the same charge density trend (more anionic oxygens), it might be expected that both would have larger HM barriers, contrary to what was calculated; this indicates that charge density is not the primary descriptor for HM. Ultimately, these calculated barriers are within the error of the MN15 DFT functional, despite it achieving an impressive 1.4 kcal/mol RMSE relative to CBS-QB3, making a chemical explanation ambiguous. Regardless, radical generation rates were similar for MeBPO (k_r = 1.5 × 10⁻¹¹ s⁻¹) and lower for MeOBPO (k_r = 9.8 × 10⁻¹³ s⁻¹) relative to unsubstituted BPO. Hence, these results suggest that EDG substitutions are not effective for achieving greater radical generation rates.

In contrast to EDG substitutions, EWG substitutions lowered both S_N2 and HM barriers while stabilizing the intermediates between these two reaction steps, which invariably increased k_r. Chloro and nitro groups lowered the computed S_N2 barriers by 1.3 and 3.6 kcal/mol and stabilized the TA and OA intermediates by 1.5 and 4.3 kcal/mol relative to BPO, respectively. EWG substitutions render the peroxy bond more electron-deficient, which elicits a greater electrophilic reaction response while substantially stabilizing the leaving OA. The APT oxygen charges on the OA were computed to be −1.23, −1.23, and −1.22 e for BPO, ClBPO, and NiBPO, respectively. Interestingly, EWG functionalization resulted in an increased electron density on the oxygen in the TA intermediate, yielding a polar N–O bond. The oxygens’ APT charges are −1.06, −1.07, and −1.08 e in BPO, NiBPO, and ClBPO, respectively, while the nitrogens’ APT charges are 0.17, 0.16, and 0.19 e for BPO, NiBPO, and ClBPO in TA, respectively. For both ClBPO and NiBPO, an increased negative charge on the oxygen resulted in a more polar N–O bond, which in turn leads to more facile HM to dissociate the N–O bond. The increased bond polarity might be expected to provide greater Coulomb attraction and thus a shorter, stronger N–O bond.²⁹ However, the N–O bond lengths of CIBPO and NiBPO are not shorter, suggesting that Coulomb attraction does not lead to any significant stabilization. Further examination shows that the increased bond polarity creates a charge gradient, which leads to a greater wavefunction curvature at the bond midpoint, ultimately increasing the electronic kinetic energy that destabilizes the N–O bond. The calculated HM barriers of NiBPO and ClBPO are 24.7 and 24.8 kcal/mol, both of which are lower than BPO’s HM barrier of 25.6 kcal/mol. As a result, both ClBPO with a k_r of 1.8 × 10⁻¹⁰ s⁻¹ and NiBPO with a k_r of 1.9 × 10⁻⁹ s⁻¹ are predicted to perform more efficiently relative to unsubstituted BPO with its computed rate constant k_r of 1.3 × 10⁻¹¹ s⁻¹.

In addition to phenyl ring functionalization at the para positions of BPO, we expanded our evaluation to include other diacyl peroxides such as dinaphthyl, diacetyl, and the cyclic analogue of BPO, phthaloyl peroxide (Figure 2). Dinaphthyl peroxide (NathPO) is predicted to perform similar to BPO with nearly identical energetics of $\Delta G_{{\text{S}}_{{\text{N}}^2}}^{‡}=16.3\text{ kcal}/\text{mol}$ and $\Delta G_{\text{HM}}^{‡}=24.6\text{ kcal}/\text{mol}$, where the S_N2 barrier is 0.1 kcal/mol higher and the HM barrier is 1.0 kcal/mol lower than those of BPO. The naphthyl groups slightly lower the electron density of the peroxide oxygens relative to the phenyl group (−0.63 with naphthyl vs −0.64 with phenyl), while their increased steric bulk imposes a larger distortion energy penalty to expose the O–O antibonding orbital for S_N2 attack. The effects of the increased distortion energy and decreased O–O electron density mostly cancel out each other, leading to a minimal change of the S_N2 barrier. However, the extended aromatic system produces a more polar N–O bond in the TA intermediate; the oxygen charge in TA is −1.08 e with the naphthyl substitution, relative to −1.06 e with the phenyl substitution, leading to higher electronic kinetic energy and thus a destabilized N–O bond and a lower HM barrier by 1.0 kcal/mol. Regardless of the slightly lower predicted HM barrier, NathPO exhibits no significant improvement over BPO.

In the case of diacetyl peroxide (AcPO), replacing the phenyl groups with methyl groups resulted in slower predicted kinetics for AcPO relative to BPO, indicating that aromatic systems facilitate redox reactions through resonance stabilization of radicals within the APRP mechanism. The inductive electron donation from alkyl groups leads to an increase in AcPO’s S_N2 barrier of 5.2 kcal/mol, while its HM barrier decreases by 1.7 kcal/mol relative to BPO. This result again emphasizes the need and challenge of simultaneously optimizing the kinetics of both the S_N2 and HM reactions for efficient redox initiation, despite the APRP rate being more sensitive to HM than to S_N2. Although the barrier for the rate-determining HM step was lowered, the larger increase in the S_N2 barrier leads to an overall lower rate of radical generation of k_r = 4.3 × 10⁻¹² s⁻¹.

Finally, BPO’s cyclic analogue—phthaloyl peroxide (PhthPO)—was predicted to possess a low S_N2 barrier of only 8.8 kcal/mol and a high HM barrier of 33.8 kcal/mol. The low S_N2 barrier results from unrestricted access to the ${\sigma }_{\text{O}-\text{O}}^{*}$ antibonding orbital and thus the absence of the O–O bond torsion penalty required of other peroxides to expose their ${\sigma }_{\text{O}-\text{O}}^{*}$ antibonding orbital. In addition, PhthPO is the only peroxide that was predicted to produce intermediates with an exergonic S_N2 free energy ($\Delta G_{{\text{S}}_{{\text{N}}^2}}^0=-19.1\text{kcal}/\text{mol}$) despite the greater entropic penalty that results from the anion group not leaving. The unique energetics predicted for PhthPO result in k_r = 9.4 × 10⁻⁶ s⁻¹, which is orders of magnitude faster than that of all other peroxides examined according to our kinetic model based on an inner-sphere ET APRP mechanism.

Alkyl, Carbonate, and Inorganic Peroxides.

In addition to diacyl peroxides, other common peroxide classes (Figure 3) were computationally evaluated to further explore potential candidates for accelerated APRP. DCPO is a non-acyl peroxide with two phenyl rings, used primarily as a thermal initiator that enables chain branching and cross-linking within polymer production.³² DCPO is similar to BPO but has tertiary carbons in place of carbonyl groups adjacent to the peroxide O–O bond. Because DCPO is an alkyl peroxide, the orbitals of its peroxy bond do not overlap with the π-space of its phenyl rings, unlike BPO. This precludes aromatic stabilization of its peroxy radicals (see Figure 4 for orbital structures of BPO and DCPO). Although we compute the modest HM barriers of 26.2 and 25.5 kcal/mol for DCPO and its para-chloro-substituted counterpart (ClDCPO), respectively, both have predicted S_N2 barriers of ~45 kcal/mol, which are inaccessible at RT and thus render APRP initiation with DCPOs infeasible at practical temperatures. We validated these predictions using FTIR (Figure 6), which showed that 3 mol % DCPO does not initiate polymerization for 20 min when added to 3 mol % DMA in the DEGEEA monomer. This indicates that DCPO and likely ClDCPO are insufficiently electrophilic because of their lack of acyl groups. Even AcPO, which possesses no phenyl rings, has a lower S_N2 barrier than DCPO and Cl-DCPO; this indicates that the carbonyl groups neighboring the peroxide bond are more essential for accelerating APRP than decoupled phenyl rings that do not overlap with O–O orbitals. In contrast to BPO derivatives in which EWGs significantly increase reactivity, Cl substitution of the phenyl groups does not lower the S_N2 barrier of ClDCPO because similarly, the tertiary carbon disrupts the π-conjugation between the O–O bond and the aromatic rings, thus obstructing favorable orbital interactions.

Figure 3.

Non-acyl peroxides investigated in this study, including dicarbonate peroxides (green), dialkyl peroxides (purple), and the inorganic disulfate and dihydrogen peroxides (gold).

Figure 4.

MN15 lowest unoccupied molecular orbitals (LUMOs) of (A) BPO, (B) AcPO, (C) DCPO, and (D) PhthPO. The LUMO is illustrated because upon S_N2 attack, the leaving OA becomes singly reduced with the additional electron occupying what was previously the LUMO. The LUMO for BPO is large and delocalized over the entire molecule. In contrast, AcPO’s LUMO is smaller but still delocalized over the molecule. The LUMO of DCPO is large but with no orbital amplitude localized on the tertiary carbons, indicating no conjugation between the O–O bond and the phenyl rings. The LUMO of PhthPO is large and delocalized over the entire molecule.

Figure 6.

Experimental kinetic profiles of various peroxides. BPO was predicted to efficiently initiate APRP, while all other peroxides examined were predicted to be relatively poor redox initiators, which was confirmed by the experimental kinetic results shown here. Dilauryl peroxide was used in place of AcPO. 3 mol % of the amine and the peroxides were used in the DEGEEA resin. Individual kinetic profiles are reported in S4 of the Supporting Information.

Peroxycarbonates (dicarbonate peroxides) were also examined (Figure 3). Peroxycarbonates possess ester groups that contain additional oxygens that bridge the carbonyls to the alkyl groups of bis(ethyl) peroxycarbonate (BEPC) and phenyl groups of bis(phenyl) peroxycarbonate (BPPC). We previously determined that acyl peroxides are superior to alkyl peroxides because of the π-conjugation between the peroxy group and the carbonyls regardless of the presence of phenyl groups. This π-conjugation delocalizes the electron density of the peroxy oxygen lone pairs, thus enhancing the electrophilicity of the O–O bond. The carbonates of dicarbonate peroxides were predicted to provide both increased π-conjugation and electron-withdrawing character, rendering the O–O bond more electrophilic. This increased O–O electrophilicity should lead to lower S_N2 barriers while retaining sufficient radical and anion stability through improved electron delocalization. Peroxycarbonates should therefore exhibit improved kinetics over acyl peroxides. To examine this hypothesis, we investigated BEPC (Figure 3) and BPPC (Figure 3) within our APRP model. Our results predict that BEPC has comparable kinetics and BPPC has improved kinetics relative to Ni-BPO, which is predicted to be the most efficient noncyclic BPO derivative for APRP operating through the APRP inner-sphere ET mechanism (vide supra). DFT predicted an S_N2 barrier and an HM barrier of 14.3 and 24.6 kcal/mol for BEPC and 9.2 and 22.2 kcal/mol for BPPC, respectively. The low S_N2 and HM barriers for both peroxycarbonate species suggest that they should indeed possess improved kinetics over BPO, agreeing with our previous conclusion that aromatic systems and EWG groups facilitate APRP reactions. However, despite these promising predictions for APRP initiated by peroxycarbonates, we excluded them as viable candidates to accelerate APRP because of their low decomposition temperatures (T_d = 35–40 C for BEPC; Table 1).

Last, we evaluated two inorganic peroxides (Figure 3): PDSF and hydrogen peroxide (HOOH). In PDSF, each oxygen of the peroxy bond is bound to a tetrahedral sulfur which contains two sulfinyl bonds (S=O) and a thiolate bond (S–O⁻). The abundance of electronegative oxygens bound to the sulfurs is expected to increase peroxy electrophilicity through the inductive effect and correspondingly decrease the S_N2 barriers of PDSF such that it may perform as an efficient peroxide within the APRP framework. On the other hand, steric hindrance from the tetrahedral geometry and substantial Pauli repulsion from the oxygen lone pairs may increase the S_N2 barrier. These two competing effects resulted in an intermediate S_N2 barrier of 32.6 kcal/mol, which falls between the predicted barriers of acyl and alkyl peroxides and is too high for PDSF to efficiently initiate RT APRP. Additionally, PDSF’s HM barrier of 40.0 kcal/mol indicates that the tetrahedral geometry precluded orbital overlap between the radical oxygen and the three sulfate oxygens and resulted in the absence of radical stabilization. Consequently, our results predict that PDSF would perform poorly for APRP initiation, which was confirmed by FTIR experiments (Figure 6). For the case of hydrogen peroxide (HOOH), the lack of functional groups excluded the potential for enhanced electrophilicity or radical/anion stabilization. However, HOOH has a lower S_N2 barrier (28.6 kcal/mol) than DCPO, primarily because of reduced steric hindrance as opposed to increased electro-philicity. Conversely, HOOH’s HM free-energy barrier of 36.5 kcal/mol is higher than DCPO’s free-energy barrier because HOOH lacks the ability to stabilize radicals, while DCPO’s tertiary carbons stabilize radicals. Ultimately, with a computed k_r of 9.5 × 10⁻³⁰ s⁻¹, HOOH was predicted to perform the worst of the peroxides considered in this study for APRP initiation.

In-Depth Investigation of PhthPO.

DFT calculations predicted that PhthPO will outperform all 15 peroxides studied here, including various BPO derivatives and new peroxycarbonate structures. PhthPO has superior thermostability (T_d > 90 C) and is employed primarily in synthetic chemistry.²⁷ With the promise of unprecedented APRP efficiency and relatively high thermal stability, we synthesized PhthPO and monitored its initiation behavior by FTIR (Figure 6). Contrary to our prediction, negligible polymerization was observed. Instead, upon addition of the PhthPO resin to DMA, the resin changed from colorless to brown, a behavior distinct from the other peroxides that exhibited no noticeable color change (Figure 5). The UV–vis spectrum of the mixture of PhthPO added to N,N,4-trimethyl-aniline (TMA) absorbed in the visible range with peaks at ~400 and ~550 nm. This color change suggests that PhthPO undergoes a reaction that involves a disparate mechanism from the inner-sphere ET APRP mechanism that we have heretofore assumed as the operative mechanism for APRP.

Figure 5.

UV–vis spectra of redox initiators. PhthPO was compared to BPO by adding it to TMA, a proxy for DMA. When BPO is combined with TMA, no significant change occurred, while the addition of TMA to PhthPO resulted in an obvious change in the UV–vis spectrum. Analytes were dissolved in dimethylformamide at a concentration of 6.1 mM.

To explore alternative reaction mechanisms for PhthPO with DMA, we calculated the activation barrier for outer-sphere ET to directly produce the amine radical cation (ARC) of DMA and PhthPO radical anion products. Such an ET would result in the dissociation of the peroxy bond, which can be theoretically understood as dissociative ET by Saveant’s model.³³ For the outer-sphere ET from amine to peroxide, Saveant’s model (eqs 1 and 2) requires three parameters, the peroxide bond-dissociation energy (BDE), the total reorganization energy (λ_tot) comprising solvent reorganization and molecular reorganization, and the net free energy of reaction (ΔG⁰).

$$\Delta G^{‡}=\Delta G_0^{‡}{\left(1+\Delta G^0/4\Delta G_0^{‡}\right)}^2$$ (1)

$$\Delta G_0^{‡}=\left(\text{BDE}+{\lambda }_{\text{tot }}\right)/4$$ (2)

For outer-sphere ET from DMA to PhthPO, DFT predicted a BDE of 30.2 kcal/mol, λ_tot = 0.25 kcal/mol, and a G⁰ of 14.7 kcal/mol. Hence, the predicted dissociative ET barrier is only 16.7 kcal/mol, which produced an outer-sphere rate constant (k_OS) of 3.4 s⁻¹, as opposed to PhthPO’s predicted inner-sphere ET rate constant of 9.4 × 10⁻⁶ s⁻¹. Consequently, although the kinetic constant of inner-sphere ET with PhthPO was predicted to be the largest of all peroxides investigated, the kinetic constant of outer-sphere ET with the same molecule was found to be 6 orders of magnitude larger, indicating that the DMA–PhthPO redox pair will invariably undergo unproductive outer-sphere ET. A similar analysis was conducted for a more nucleophilic amine that may permit facile S_N2 attack and initiate the desirable APRP reaction cascade through inner-sphere ET, but we concluded that regardless of the identity of the amine reductant partner, PhthPO remained incompetent as an oxidant for APRP (see Section S2 of the Supporting Information for a discussion of PhthPO reaction with a more nucleophilic amine).

Summary of Peroxides Studied.

The compiled redox kinetic constants of all peroxides considered to react with DMA are reported in Table 1. When these peroxides were examined within our FTIR polymerization study (Figure 6), those that were predicted to have lower rate constants than BPO were observed to exhibit no significant polymerization including HOOH, PDSF, and so forth. Conversely, BPO rapidly progressed to full conversion within 20 min. We expect that peroxides with rate constants larger than those of BPO exhibit rapid polymerization behavior based on the APRP mechanism unless like PhthPO, they undergo outer-sphere ET or are thermally unstable. However, these likely involve synthetic challenges and potential hazards, precluding them from being considered in our current experimental study. Therefore, BPO is reaffirmed as a good compromise of rapid redox kinetics and sufficient thermal stability.

Direct-Writing Application of APRP Based on an Efficient APRP Pair.

From the present study of peroxides and our previous investigation of amine initiators,²⁵ we have determined that the BPO and MPP redox pair may accelerate the rate of APRP to an extent that enables the direct writing of resins. The redox pair of BPO and MPP has a predicted k_r of 1.9 × 10⁻² s⁻¹ in comparison to k_r = 1.3 × 10⁻¹¹ s⁻¹ of the BPO and DMA pair. This prediction was experimentally confirmed by the observation of an ~20 times faster polymerization rate with this new redox initiator in comparison to other state-of-the-art tertiary aromatic amines. Considering the relationship between the rate of polymerization and the radical generation rate constant, R_p ∝ k_r, the radical generation efficiency is ~400 times higher.²⁵ The potential benefits of an APRP-enabled direct-writing system include a low cost of the initiator system relative to photoinitiators, high filler concentrations for structural or electronic applications, proven low toxicity for use in medical applications, and high interlayer strength due to curing across layer boundaries.^34–37 We designed our direct-writing system based on an inkjet printer using a redox chemistry of MPP and BPO (Figure 8). This system has two pumps which cause the flow of a moderate-viscosity amine resin and peroxide resin into a mixing nozzle which is connected to a motor that moves along the horizontal X-axis (Figure 7a). Polymerization begins in the mixing nozzle where the two resins combine, and the resulting amine–peroxide resin is then pumped onto a horizontal support platform that is moveable along the Y-axis by a second motor to provide 2-dimensional control. The addition of other resin reservoirs would enable multimaterial printing to fabricate printed objects with variable material properties, and the addition of a Z-axis motor could enable 3D printing. We used this writing system (Figure 7b) to print a spiral pattern with an average width of approximately 5 mm with some segments near a 1 mm resolution (Figure 7c). The large variability in printing width could be improved by incorporating more consistent pumps and a method to better control the pump rate; furthermore, a narrower nozzle may enable a resolution of <1 mm. Gelation was achieved in under 10 s with some oxygen-inhibited layer with the BisEDA resin at 1.5 wt % amine and peroxide concentrations. Although substantial work remains to more fully realize the APRP direct-writing technology, this proof-of-concept demonstration indicates that APRP can be successfully implemented as the initiation method in a direct-writing system and suggests that if the potential benefits of this technology are achieved, it could become a 3D printing system for a range of applications including structural, electronic, and medical devices where high filler concentrations and low toxicity could enable state-of-the-art printed objects. Further details are reported in S5 of the Supporting Information.

Figure 8.

Reaction mechanism and barriers for the APRP initiation of BPO and MPP. We previously showed that MPP could generate initiating radicals ~20 times faster than other state-of-the-art tertiary aromatic amines.

Figure 7.

(a) Schematic of our custom direct-writing system depicting the two separate pump lines from the amine and peroxide resin tanks that flow into a mixing nozzle, enabling the in situ combined amine–peroxide resin to be printed through the nozzle onto a support platform. (b) Image of the direct-writing system. (c) Polymer spiral written via APRP. BisEDA (M_n = 512, EO = 2) with 1.5 wt % BPO and MPP was used.

CONCLUSIONS

In this work, we have presented an extensive computational investigation of 15 peroxide oxidant candidates for initiating APRP through a mechanism that we previously reported to accurately predict experimental APRP rates. We found that acyl groups not only increase the electrophilicity of peroxides to enhance the S_N2 reaction but also stabilize anion and radical intermediates generated from the parent peroxides. As such, functionalization of acyl groups either through addition of para-EWGs (to increase peroxy electrophilicity) or through addition of phenyl rings (to improve resonance stabilization) is imperative for high redox efficiency. Alkyl peroxides have low HM barriers because of their tertiary carbons that stabilize radical intermediates, yet their reduced electrophilicity at the peroxy bond increases the S_N2 barrier and lowers alkyl peroxides’ overall reactivity. While peroxycarbonate structures appeared to be promising in terms of their predicted initiation kinetics resulting from their higher electrophilicity and stronger resonance, their low thermal stability may hinder their practical use for accelerated APRP. Inorganic peroxides such as PDSF and hydrogen peroxide are predicted to perform poorly in APRP initiation. PhthPO, which appeared to be the most promising candidate from our analysis, does not proceed through the inner-sphere ET APRP mechanism, in contrast to the other peroxides examined. We predict that it instead proceeds through an outer-sphere ET to produce noninitiating radicals, rendering PhthPO ineffective for APRP. Nevertheless, the ability to switch between different ET mechanisms based on the amine–peroxide pair is novel and provides additional chemical and mechanistic insight into amine–peroxide redox chemistry; this discovery could enable an entirely new class of redox pairs that generate initiating radicals through outer-sphere ET. Ultimately, we find that BPO remains the gold standard peroxide for APRP because of its rapid APRP kinetics and high thermal stability. Furthermore, we demonstrated an APRP direct-writing system that leverages an optimized redox pair, where automated mixing of two resins on a movable platform enabled printing of a spiral-shaped polymer. This amine–peroxide direct-writing technology has the potential to be the pre-eminent form of 3D printing where high filler concentrations and low toxicity are imperative.