A Kinetic Study of a Photo-Oxidation Reaction between α-Terpinene and Singlet Oxygen in a Novel Oscillatory Baffled Photo Reactor

1. Introduction

Heterogeneous photo-aided reactions and photocatalysis are initiated by the absorption of light. Scalable uniform mixing in handling multiphases and scalable uniform light distribution in providing light sources are two most critical requirements in photoreactors, so much so, the scalability of photoreactors has been identified as a key challenge in photochemical transformations in the chemical industry [1]. In terms of light provision, UV lamps/bars are traditionally placed outside or within reactors, suffering non-uniform distribution of light intensity, low power-to-photon efficiency and scale up ability [2]. LEDs, on the other hand, offer much narrower spectrum of light for targeted reactions and are sufficiently small to be installed within reactors, enabling uniform light distributions. LEDs have thus become the main light source in most of the lab scale photo reactors.

In terms of mixing, the century old problem remains where mixing gets worse with increasing scales: simply scaling up stirred tank reactors do not meet the above critical needs. Microreactors are seen as the viable alternative in solving the mixing issues, allowing safe exploitation of hazardous materials, while enabling optimal irradiance due to their narrow dimensions. Subsequently, a wide range of photochemistry applications using a variety of flow reactors have been reported in recent years [1,3].

While lab-scale flow photoreactors are sufficient to produce small quantities of targeted molecules [4], applications in industrial scale photochemical processes are comparatively low due to the generally poor economics, complex scalability and safety issues of many photochemical transformations [5]. There are currently three main strategies in scaling up photo-microreactors: numbering up, lengthening the channel or increasing channel diameter [6]. Numbering up is a method of adding reactor systems in parallel until a desired productivity of chemicals is achieved. However, high inventory costs are a major drawback as multiple pumps, mass flow controllers and reactors are required [7]. Lengthening the flow path while keeping the channel diameter constant enables faster flow velocities while maintaining the same photon flux, but leads to issues with large pressure drops over the reactor [8]. Increasing diameter is the least viable method in scaling up microchannel reactors, as it weakens/diminishes diffusion-driven mixing in the reactor [9], providing poorer scale up than the former methods.

In this study, we present a novel batch Oscillatory Baffled Photo Reactor (OBPR) that provides uniform mixing and solid suspension via the succession and cession of eddies. Orifice baffles are one of the key intrinsic elements of the reactor set up, planting LEDs on the surfaces of orifice baffles [10] delivers uniform and controllable light distribution/intensity. The scale up of this device is through continuous oscillatory baffled reactors (COBR) that have a number of unique offerings: a) plug flow mixing is generated in laminar flows, minimizing pressure drop and allowing longer residence times [11]; b) uniform mixing is scalable [12]; c) COBR is good in handling solids [13] and have been implemented in industrial applications [14], meeting both critical requirements in photoreactors. The work to be presented in this paper is only associated with the batch OBPR. Using a model photo-oxidation reaction between α-terpinene and singlet oxygen (¹O₂), we report our kinetic analysis and assessment of the reaction. Polymer-supported rose Bengal beads are used in the work, because the OBPR has the excellent capability of suspending solids and enabling uniform light absorption.

2. The Reaction Scheme and Experimental Set Up

2.1. Reaction Scheme

The photo oxidation reaction is shown in Scheme 1 with polymer-supported Rose Bengal (Ps-RB) porous beads as the heterogeneous photosensitizer. The beads were synthesised in house and characterised according to known literature methods [15]. The first step of the reaction is that the photo absorption (+hν) by the Ps-RB chromophore promotes an electron to a higher-order singlet electronic excited state (¹[Ps-RB]*); ¹[Ps-RB]* converts to a triplet excited state (³[Ps-RB]*) via intersystem crossing (ISC), which is the 2^nd step; the energy transfer (EnT) from ³[Ps-RB]* to ground state triplet molecular oxygen (³O₂) occurs then via triplet-triplet annihilation (TTA), to return Ps-RB to its initial ground state and produces ¹O₂; the next step is that ¹O₂ spontaneously decomposes to ³O₂ via non-radiative decay (NRD) through vibronic energy transfer with solvent molecules, ¹O₂ encounters α-terpinene in solution and undergoes a productive [4 + 2] cycloaddition to yield the endoperoxide product, ascaridole, known as a Diels-Alder-type reaction (Scheme 1). 1,3-Dienes commonly undergo [4+2] cycloaddition reactions with singlet oxygen ¹O₂ to yield endoperoxides. This reaction has been extensively studied on naphthalene derivatives [16]. Other chemical transformations such as [2+2] additions or ‘ene’ type reactions are also possible, forming dioxetanes and hydroperoxides, respectively [17]. 1O2 exists as a gas, but is dissolved in the reaction mixture or solvent [18]. Singlet oxygen is unstable, and its lifetime changes dramatically in different solvents, ranging from 4 to 628 μs [19], depending on the energy-transfer efficiency from electronic to intramolecular vibrational states, for instance, the lifetime in water is one order of magnitude lower than that in deuterated water. Chloroform (CHCl3) provides the longest 1O2 lifetime (~90 μs) of all common, non-deuterated laboratory solvents [20], hence was chosen as the solvent in this study. It should be noted that the amount of ascaridole produced does not equal to the amount of ¹O₂ formed due to the non-radiative decay, but what is sure is that the formation of ascaridole is the direct evidence of the existence of singlet oxygen [2a].

Heterogeneous photosensitizers have many advantages over homogeneous counterpart, for example, they are readily prepared, are easily removed post-reaction, when suspended in solution, do not accumulate or clog reactors. Heterogeneous photosensitizers are generally comprised of organic dyes bearing a (hetero)aromatic core, e.g. Rose Bengal (RB). Because RB suffers from extensive photobleaching/degradation under prolonged irradiation, the leakage is usually difficult to be removed from reaction effluents [21], modifications in synthesis have led to various robust solid photosensitizers. Previously, a versatile organic dye, 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene (BODIPY), with a polymer-support was used as the photosensitizer in the same reaction, where BODIPY has nitrogen and boron containing heterocycles [4] with high fluorescence quantum yields and significant visible-light absorption, the ideal light wavelength fell between 500 and 540 nm. In this work, polymer-supported Rose Bengal (Ps-RB) porous beads are used, have high absorption coefficients in the visible spectral region and are stable after photosensitization, its optimal light wavelength is 530 nm [22]. The key advantage of Ps-RB over BODIPY is that the maximum absorbance wavelength is red-shifted and lower in energy, reducing the likelihood of photodecomposition of reactants and products [23].

2.2. Materials

α-terpinene was purchased from the Division of Tokyo Chemical Industry (TCI) (>90% purity), all organic solvents and reagents were sourced from Fisher Scientific at SLR grade, were used as received without further purification unless otherwise stated.

2.3. Reactor Setup

The OBPR consists of a glass column of 50 mm in diameter and 480 mm in height as shown in Figure 1. The volume of the reactor is 600 ml, with the working volume of 500 ml. Orifice baffles have an outer dimeter of 46 mm, a hole diameter of 26 mm and a baffle spacing of 60 mm. Each set comprises of three orifice baffles, each baffle has 6 evenly spaced green LEDs (Cree 5-mm Round LEDs) planted on the lower surface (can also be both sides), as shown in Figure 1, for the purpose of providing light. The leads of LEDs are covered by epoxy resin that provides the insulation, material compatibility and corrosion resistance. The wavelength and light intensity of the green LEDs are 530 nm and 0.756 watts respectively. An air sparger is located at the base of the OBPR for introducing air at a controlled rate.

Based on a previous work in a microchannel reactor [6,7b], about 390 mL CHCl₃ was mixed with 4.225 mL of α-terpinene in the OBPR in the presence of Ps-RB beads of 1600 mg. A constant liquid volume of 400 mL was maintained for all experiments. Oscillation frequency and amplitude were switched on once all chemicals were charged into the OBPR, the reaction was initiated when LEDs were turned on, the duration of the reaction was 120 mins. Runs were repeated for the data repeatability and reliability (see Table 1), averaged values were used in data presentation later.

2.4. Analytic Method

Samples were taken regularly and analysed using proton nuclear magnetic resonance (¹H NMR) to determine the composition/concentration of the mixture. The procedure of treating samples was performed as follows:

• Each 2 mL sample containing CHCl₃ + α-terpinene + ascaridole was injected to a dark vial to stop the reaction (reaction stops when light is off)
• The sample was placed into a round bottom flask and the solvent (CHCl₃) removed under reduced pressure on a rotary evaporator (40°C, 365 mbar for ~5 minutes)
• The oily residue was dissolved in 0.5 mL of deuterated chloroform (CDCl₃) and a ¹H NMR was obtained (300 MHz Bruker AVIII spectrometer)
• Peaks in the region between 6.70 and 5.40 ppm were used to determine the conversion of α-terpinene and the appearance of ascaridole

Figure 2 is a typical ¹H NMR spectra in chloroform-d, showing the alkenyl proton resonances of α-terpinene as a multiplet at 5.61 ppm, and ascaridole as a doublet-of-doublets centered at 6.45 ppm, consistent with previous reports [24], the resonance signals count for two alkenyl protons of the respective molecules. While p-cymene is a byproduct in the reaction, it is not detected by the 1H NMR, which is consistent as other studies [25]. At the reaction stochiometric ratio of 1:1, the integrals of these signals are directly proportional to the relative concentrations of the two species. The concentration of α-terpinene at time t (C_αT) is the ratio of the integrated NMR area of α-terpinene at time t (Area_αT) over its corresponding area at time t=0 (taking as 100), multiplied by its initial concentration at t = 0 (C_αT0 = 0.051917 mol/L). The concentration of ascaridole (C_As) is then the ratio of the integrated area of ascaridole over the total area at time t multiplied by the initial concentration of α-terpinene, since the initial concentration of ascaridole at t = 0 is zero as:

$$C_T=\left(\frac{{Area}_T}{{Area}_{\alpha T0}}\right)C_{\alpha T0} C_{As}=\left(\frac{{Area}_{As}}{{Area}_{As}+{Area}_{\alpha T}}\right)C_{\alpha T0}$$

(1)

Since the byproduct is not detectable, Area_As + Area_αT = 100, making the concentration of ascaridole a reaction pair of that of α-terpinene.

3. Results and Discussion

3.1. Kinetic Assessment

The overall reaction in Scheme 1 can be described by a sequence of two reactions:

Following the reaction Scheme 1, some singlet oxygen (¹O₂) decomposes to molecular oxygen (³O₂) via non-radiative decay (NRD), the first part of the reversible reaction counts for this process. At a fixed oscillatory frequency (f) and amplitude (x₀) in the OBPR with a known quantity of polymer-supported Rose Bengal beads at continuous controlled light irradiation at steady state, a net amount of ¹O₂ (resulting from the 1^st part of the reaction) is generated and reacts with α-terpinene to produce ascaridole (the 2^nd part of the reaction). For one equivalent each of ascaridole and α-terpinene without any byproduct, the data analysis makes the consumption of α-terpinene as a reaction pair of the generation of ascaridole. This effectively assumes that the concentration of singlet oxygen is kinetically either in excess or constant. To the authors knowledge, no research publications have looked into this matter. In order to testify and verify this assumption, a number of planned experiments were carried out using different amounts of α-terpinene and chloroform while keeping the same reaction volume, see Table 1. Note that Run 2 was the default condition, based on previous studies [4,26].

Figure 3 compares the concentration profiles of ascaridole with different amounts of α-terpinene, an increase of the concentration of ascaridole as a function of time can be seen with the increase of α-terpinene in Figure 3a on the left (the filled symbols), this indicates that ¹O₂ supply in the OBPR is indeed in excess. Further increases in the reagent however have brought decreases in the product (see the open symbols) in Figure 3b on the right, implying that under these experimental conditions, ¹O₂ has become the limiting reagent, the concentration of ascaridole would solely be dependent upon the concentration of ¹O₂, instead of α-terpinene, this stipulates that the widely used method of treating α-terpinene and ascaridole as a reaction pair in eq. (1) would no longer be valid under these conditions. Noted that chloroform is the solvent that does not participate in the reaction and is in significant excess over α-terpinene, its effect on the concentration of ascaridole should be of minimum, while not tested.

Under the conditions where ¹O₂ is in excess (Figure 3a), the 2^nd order reaction of ¹O₂ + α-terpinene $\stackrel{k3}{\to }$ascaridole is then reduced to a pseudo first order of α-terpinene $\stackrel{k_3}{\to }$ascaridole, it is then justified to treat the two as a reaction pair. The rate equation and the integral form are given below:

$$-\frac{{dC}_{aT}}{dt}={\mathrm{k}}_3{\mathrm{C}}_{\mathsf{\alpha }}{\mathrm{T}}_{ } \ln \left(\frac{C_{T0}}{C_T}\right)=k_{3}t$$

(3)

By plotting $\ln \left(\frac{C_{T0}}{C_T}\right)$ vs time for α-terpinene for 2.12, 4.23 and 7 ml, approximate straight lines can be seen in Figure 4, verifying the pseudo first order kinetics with the averaged rate constant of 0.0020 /min.

At the 8.5 ml of α-terpinene, the conversion of ascaridole has reached the maximum, implying that neither singlet oxygen nor α-terpinene is in excess. Using the proportionality of ascaridole with respect to its maximum at 8.5 ml in Figure 3a, the excess ¹O₂ concentrations for 7/4.23/2.12 ml of α-terpinene can be estimated using the equation below taking 7 ml as an example:

$$C_{O_2^1}=[1+\frac{{\left(C_{As}\right)}_{8.47ml}-{\left(C_{As}\right)}_{7ml}}{{\left(C_{As}\right)}_{8.47ml}}] \times C_{\alpha }\mathrm{T}$$

(4)

Figure 5 shows the concentrations of all species for three volumes of α-terpinene. Note that there is no data of ¹O₂ at t = 0, this is due to the concentration of ascaridole being zero at the start, ¹O₂ concentrations were estimated from 20 mins onwards. We see that the concentrations of singlet oxygen are greater than that of α-terpinene, demonstrating that ¹O₂ is in excess under those conditions. We also note that the concentration of ¹O₂ is more or less unchanged at each amount of α-terpinene, this is consistent with what we stated earlier that a net amount of ¹O₂ is generated in the OBPR at steady state.

By feeding the data of α-terpinene, ascaridole and k₃ (from Figure 4) into the reaction kinetic model in COPASI (Complex Pathway Simulator), the concentrations of singlet oxygen can be reconstructed as shown in Figure 5 (filled symbol). Note that COPASI is a versatile modelling tool that can be used for model development, model simulations (such as stochastic simulations and differential equation simulations), numerous sorts of analysis (such as sensitivity analysis, parameter estimation, and metabolic control analysis) as well as kinetic extraction. [27] We see that the estimated ¹O₂ concentrations using the proportionality of ascaridole at different amounts of α-terpinene (open triangles) match reasonably well with these determined using the kinetic model (filled triangles). This further validates our proposed methodologies. We note that zero order reaction kinetics were reported by other studies for this reaction [25,28], the zeroth order kinetics indicate that the reaction rate is a sole function of the reaction rate constant, does not depend on the concentrations of reagents. When we vary the concentrations of α-terpinene, the dependence of the reaction rate on the concentrations of the reagent is clearly demonstrated in our work.

Thanks to the estimated ¹O₂ data, we can also evaluate the efficiency of ¹O₂ utilization using the definition below, even when it is in excess as:

$$Efficiency of singlet oxygen utilization \left(\%\right)=100\%-\frac{Singlet oxygen-ascaridole}{Singlet oxygen}$$

(5)

The efficiencies of ¹O₂ utilization are tabulated in Table 2 for α-terpinene from 2.12 to 7 ml.

3.2. The 1^st Step of the Reaction

The process from molecular oxygen to ¹O₂ in the first part of the overall reaction in eq. (2) relies on photo elevation/energy transfer. Due to a significant energy barrier (94 kJ/mol), ³O₂ molecules cannot directly be excited to ¹O₂ due to it being a forbidden electronic transition by spin selection rules, and therefore requires a triplet photosensitiser to enable the triplet-triplet annihilation energy transfer mechanism [29]. When photons of the correct wavelength (<600 nm) irradiate the Ps-RB beads, the supported Rose Bengal chromophore is electronically excited and converts to a triplet-state via intersystem crossing (ISC), which enables it to undergo an energy transfer, resulting in the triplet-triplet annihilation process with oxygen to produce singlet oxygen [30]. The very short lifetime of singlet oxygen depends on temperature and solvent environment as vibronic energy transfer with solvent is a competitive, non-productive process which returns ¹O₂ to its ground state, ³O₂ [31], referring to the 1^st part of the overall reaction. Does the molecule oxygen refer to oxygen gas molecule [32] or dissolved oxygen? Air/oxygen is generally sparged into the reactor, of which a significant amount of gas bubbles exits the reactor without participating in the reaction, due to the density difference, the buoyance effect and critically the lack of mixing mechanism of holding bubbles within traditional reactors. This alludes that ³O₂ is associated with the dissolved oxygen. The process of converting oxygen gas to dissolved oxygen in chloroform is a mass transfer-controlled process, is mainly affected by mixing, air flow rate together with novel reactor designs. The properties of mixing in the OBPR are governed by the oscillatory Reynolds number [11c] (Re_o) and Strouhal number (St), defined as:

$${Re}_0=\frac{2\pi f{x}_0\rho D}{\mu } St=\frac{D}{4\pi x_0}$$

(6)

Where D is the tube diameter (m), ρ the fluid density (kg m^–3), x₀ the centre-to-peak amplitude (m), μ the viscosity (Pa s) and f the oscillatory frequency (Hz). The oscillatory Reynolds number is related to the intensity of mixing, the Strouhal number indicates the degree of eddy propagation. [11c,13b] The effects of mixing (oscillation amplitude or St) and air flow rates on the conversion of ascaridole are given in Figure 6.

We see that generally the higher the oscillation, the higher the conversion (Figure 6A). As mixing intensity increases in the OBPR, the intensive and uniform mixing breaks up bubbles, leading to a decrease in bubble sizes and increase in bubble surface area; these bubbles are trapped in the vortices for longer periods of time, enhancing residence time for mass transfer within the reactor, [33] the direct result is more dissolved oxygen concentration in the solvent [34] and more excited molecular oxygen, in turn more ¹O₂. The amount of improvement in conversion is however rather limited for all amplitudes. The effect of air flow rate on the conversion (Figure 6B) tells a similar story, the higher the air flow rate, the higher the conversion, but there is an operational limit for further increasing air flow rate, as channelling takes place, bringing out the beads from the OBPR. The highest flow rate was used in the experiments of kinetics studies.

By comparing the total amount of air/oxygen sparged into the OBPR with the estimated amount of ¹O₂ every 20 mins, the efficiency of photo conversion of ¹O₂ can be evaluated as

$$Efficiency of photo conversion of singlet oxygen \left(\%\right)=100\%-\frac{Oxygen input-Singlet oxygen}{Oxygen input}$$

(7)

The oxygen input is calculated using the air flow rate (21% oxygen) and the time duration, the ¹O₂ concentration is then converted into mass for the same time duration. Table 2 summarizes both efficiencies.

The limited conversion to ascaridole in Figure 5 and Figure 6 is the direct result of the combined low efficiency of ¹O₂ utilization with very low efficiency of converting ³O₂ to ¹O₂, the latter is perhaps the number 1 barrier for this reaction, indeed for most photo-aided reactions and photocatalysis. The low conversion to ascaridole is also consistent with previous work, albeit with BODIPY as the photosensitizer [35]. We are in the process of expanding this work to determine the pressure dependence of ¹O₂ by varying the partial pressures.