Catalytic Oxidation of Flax Shives into Vanillin and Pulp

Abstract

This research deals with a process of catalytic oxidation of flax shives to vanillin and pulp. Catalytic oxidation of flax shives with molecular oxygen allows two main products to be obtained—vanillin with a yield of up to 12 wt.% of lignin, and pulp. Final forms of the catalyst particles (Cu₂O and CuO) are agglomerates or monocrystals 0.5–5 µm in size. Acid prehydrolysis of the shives does not affect the oxidation, in contrast to pine-wood oxidation. Lignin prehydrolysis and oxidation was suggested as illustrate this difference. The dependence of the vanillin formation rate on stirring speed was studied. Minimum alkali consumption in the process was attained with a mild stirring speed of the reaction mass.

1. Introduction

Today, vanillin is one of the most popular food and perfume flavors. It can be produced by two main methods, glyoxylic synthesis and lignin oxidation. The first has dominated industrial-scale production since the 1970s, and the second is the main method used to produce vanillin in 50–70 s [1,2,3].

Glyoxylic synthesis of vanillin starts from electrophilic substitution of 4th position in benzene nuclei of guaiacol by glyoxylic acid in alkaline media:

(1)

The reaction of two negative ions, guaiacolate and glyoxylate, causes very high selectivity of attack in the para-position to ionize the hydroxy group of guaiacol. On the other hand, repulsion of the negative charges decreases the equilibrium constant in Equation (1), and this is one of the major disadvantage of the process. The obtained guaiacyl glycolic acid is then oxidized with oxygen over a CuO catalyst [1]:

(2)

One more advantage of the glyoxylic process is the low capital costs due to high reagents concentration and low operating pressure [1].

Oxidation of lignins by nitrobenzene—or nitrobenzene oxidation (NBO)—is a long-known [2,3] and highly selective process; its results qualitatively—and even quantitatively—were used to determine the structure of lignins. Vanillin yields from softwood lignin NBO attains 25–28 wt.%, which is usually accepted as the theoretical maximum yield [2,3]. Another selective oxidant of lignins into vanillin is copper oxide CuO [4,5].

In the first half of the 20th century, it was discovered that the processes of lignosulfonate oxidation into vanillin by oxygen are catalyzed by oxides of copper, manganese, cobalt, and silver [2,3], and these catalysts exhibit similar effectiveness.

Stoichiometry oxidation of pine or aspen wood by copper requires 0.8–1.0 wt/wt of CuO on the wood [4,5]. CuO as a catalyst is used in the quantity of 0.01–0.1 wt/wt while oxidizing lignosulfonates [6], pine wood [5] and others [7], but such catalyst quantity did not catalyze the process under flow conditions [8]. Copper oxide of 0.14 wt/wt was used in the catalytic oxidation of aspen wood enzymatic lignin [9]. Copper oxide (0.1 wt/wt) produced from CuSO₄ in situ increased vanillin yield by a factor of 2 in the process of lignosulfonate oxidation with oxygen [10]. Copper catalysts (CuO, 0.5 g, wood 1.0 g) increase vanillin yield while oxidating a softwood Japanese cedar with H₂O₂ at 180 °C for 80 min [11]. It is likely that hydrogen peroxide is rapidly converted to molecular oxygen under these conditions by the catalyst, and after that the process of oxidation with oxygen catalyzed by CuO occurs. Despite the large 1:2 ratio of CuO:wood, copper oxide is not a stoichiometric oxidant, but rather, a catalyst, because both CuO and H₂O₂ in the reaction mass gave the most yields. To summarize, using copper catalysts increases vanillin yield in oxidation of lignins by oxygen by a factor of 1.5–3, and properly organized catalytic processes are, at most 10–15% inferior to nitrobenzene oxidation in terms of vanillin yields [3]. However, some studies have observed no vanillin yield increase in processes catalyzed by copper (II), cobalt (II), Pd/TiO₂, or Ce/MgO [12,13,14,15].

Vanillin is produced by Eurovanillin (approx. 1000 tones annually) by catalytic oxidation of lignosulfonates or Kraft lignin, the main wastes of pulp industry. A simple scheme of the vanillin formation from lignin is shown below. The main disadvantages of the process are the low concentration of vanillin in the reaction mass formed and the low vanillin yield of lignosulfonates (6–12 wt.%) and of the Kraft lignin. Any pulping process leads to condensation of native lignin, and these reactions causes low yield of vanillin from technical lignins, including lignosulfonates and Kraft lignin. The main advantage of the process is very low (close to zero) cost of the raw material.

To overcome problems of lignin condensation, the concept of “lignin first” was suggested: the first step of wood treating should be lignin conversion [16]. Indeed, catalytic oxidation of wood gives the highest yields of vanillin based on softwood lignin (up to 20–27 wt.%) together with cellulose [17].

Wood and even wood wastes are obviously more expensive compared to lignosulfonates but less expensive in comparison with guaiacol and glyoxylic acid. Principally, different agricultural wastes may be the resource of cheaper native lignins for obtaining vanillin.

The world production of vanillin is about 15–20 thousand tons per year at prices of more than USD 15 per kilogram [18]. The most promising method for vanillin and syringaldehyde production is catalytic oxidation of native lignins (contained in wood or plants) with oxygen, and cellulose (up to 80% of the initial content) is a by-product suitable for further chemical processing [3,16,17].

The possibilities of processing renewable plant raw materials are being actively studied at the present time and are focused on replacing oil and other fossil resources. Agricultural waste is a huge renewable resource of lignocellulose materials but is little used in chemical processing.

The production of flax in Russia is 600 thousand tons per year, and flax shives are the main waste product (up to 70%) of the manufacture. Shives are the lignified parts of the flax stem, mainly in the form of small straws, remaining after the threshing stage. Usually, flax shives remain in the form of huge dumps or are burned to produce heat. At the same time, the reserves for increasing the efficiency of processing flax are directly related to the rational usage of flax shives [19,20]. Recently, it has been shown that vanillin and cellulose can be obtained by oxidation of flax shives [21], and their hydrogenation gives guaiacyl propane derivatives [22,23,24,25,26].

Flax shives contain 18–28% of lignin. By this index flax shives surpass hardwoods and are close to that of softwood. The maximum yields of vanillin per lignin are 15–16 wt.% and 11–12.5 wt.% in the processes of flax shive oxidation with nitrobenzene and oxygen, respectively [21]. The yields of syringaldehyde from the shives are 4–5 wt.%, and para-hydroxybenzaldehyde is the minor byproduct.

Study of the quantitative regularities of the mass-transfer-intensity effect on flax shive oxidation to vanillin showed that under the studied conditions this process is diffusion-controlled without a noticeable contribution of kinetic limitation. Under these diffusion-controlled conditions, the maximum yields of vanillin per lignin in the shive oxidation processes with nitrobenzene (15–16%) and oxygen (11–12.5%) differ insignificantly, by 17–30 relative percent, and this difference is interpreted as the result of process transition from kinetic to diffusion limitation.

Excessive consumption of the main reagents—alkali and molecular oxygen—is a quite important, but rarely discussed, problem of lignin-containing raw material oxidation into vanillin [3]. Usually, the oxygen consumption is an order of magnitude higher than that stoichiometrically required for the observed vanillin yield. The oxygen excess produces acids, which leads to drastic additional alkali consumption. A significant reduction in the alkali consumption was attained by removing part of the hemicelluloses by acid prehydrolysis of wood, slight hydrolysis of lignin, and removal of the formed vanillin during oxidation [17]. Some prerequisites for the possibility of reducing the alkali consumption during the oxidation of flax shives by varying the intensity of mass transfer were obtained in [21].

The goal of this work was to study and improve the catalytic oxidation process of flax shives into vanillin and syringaldehyde, and the pulp:catalyst composition and distribution in the reaction mass, the effect of mass transfer intensity, and the influence of alkali load and acid prehydrolysis.

2. Results and Discussion

2.1. Catalyst Composition and Distribution in the Reaction Mass

In the oxidation processes of plant raw materials to vanillin, the copper catalysts are usually input into an aqueous-alkaline solution of the reaction mass in the form of a copper sulfate solution; the latter turns into hydroxide, and when heated, into copper oxide. Any indications on conversion of copper(II) oxide in copper(I) oxide or copper(0) in the process are lacking in the literature. The obtained X-ray diffraction data (Figure 1) showed that the catalyst is a pure Cu₂O phase at low-mass-transfer intensity (stirring speed 400 rpm). Increasing the stirring speed up to 500 rpm leads to appearance of the impurity of CuO phase, and the latter forms the main phase with impurity of Cu₂O at 700 rpm. Hence, the catalyst oxidizes products of alkaline destruction of flax shives and is reoxidized by oxygen. The rate of the latter depends on the mass transfer intensity and stirring speed; therefore the CuO phase, the oxidized form of the catalyst, dominates at high stirring rates.

Most of the Cu₂O is deposited on both the outer and inner surfaces of the oxidized flax shive particles (Figure 2). This makes it difficult to separate the catalyst from the resulting pulp by sedimentation methods. Catalyst particles have a wide size range from 0.5 µm (Figure 2, left) to 5 µm (Figure 2, right). These are mainly agglomerates (Figure 3a,b); there are also large monocrystals of 3–5 µm size (Figure 3c). The latter probably grow randomly during the process for tens of minutes, and this indicates a slow redistribution of copper between oxide particles through the liquid phase of the reaction mass.

To catalyze the process, we used not quite much catalyst, 0.24 wt/wt of CuO on the flax shives, and this require comment. Stoichiometry oxidation of pine or aspen wood by copper oxide requires 0.8–1.0 wt/wt of CuO on the wood [4,5]. CuO, as a catalyst, is used in the quantity of 0.01–0.5 wt/wt while oxidizing different lignins [6,7,8,9,10,11,12,13,14,15]. The catalyst load of 0.24 wt/wt used in this study is closer to the catalyst loads [3,5,6,7,8,9,10,11,12,13,14,15], than to a stoichiometry reagent [4,5].

Figure 4 and

Table 1. Influence of the catalyst on the vanillin yield and the reagent consumption.

Reagents		Vanillin Yield per Lignin, wt.%	Oxygen Consumption, mol	Oxygen Consumption, mol/mol of Vanillin	m (NaOH)/m (Vanillin)
1	O2, CuO, 5% NaOH	12.5	154	51.1	24.9
2	O2, CuO, 3.75% NaOH	11.6	89.5	32.0	17.4
3	O2, 3.75% NaOH	12.6	121	42.5	16.2
4	CuO, 3.75% NaOH	6.26	-	-	32.6

show that without oxygen the process produces less vanillin (Line 4), and the rate of its formation is lower than that of the catalytic process with oxygen under the used conditions (Lines 1, 2). In other words, stoichiometric oxidation of the shives by CuO does not play the main role in the catalytic process under the study.

Several papers have shown the lack of activity of copper [8,12,13] and other catalysts [14,15], and one more experiment to confirm the catalyst role is the process without copper. The result was quite unexpected: vanillin yields with and without the catalyst coincided (Figure 4, Lines 2, 3, Table 1). This result does not formally conradict the data of Figure 1 showing the catalytic properties of copper and the transition of CuO ⇋ Cu₂O during the process. This transition may be caused by participating the catalyst in side reactions.

We assume that the copper catalysts increase vanillin yield in most cases of lignin oxidation. The lack of the catalytic activity that we found may have been caused by some specific properties of flax shives, for example, by impurities of catalytically active metals in the shives.

2.2. Influence of Flax Shive Acid Prehydrolysis on Its Subsequent Oxidation

Previously, it was shown that acid prehydrolysis of pine wood greatly affects the consumption of alkali during the oxidation of the formed lignocelluloses (LC) to vanillin: it decreased it by about two times while going from the wood to the slightly prehydrolyzed LC (5% weight loss of the wood after prehydrolysis) [17]. This result was accounted for by the partial destruction of lignin due to its slight hydrolysis, and this increased the reactivity of the lignin in the process of vanillin formation. In this work we studied the possibility of using this approach for the oxidation of another lignin-containing raw material, flax shives. Figure 5 shows the curves of vanillin accumulation during the oxidation of the initial and prehydrolyzed Belarusian flax shives (weight loss of 10%). The curves were obtained at various alkali concentrations. Decreasing the alkali concentration obviously reduced the yields of vanillin; this drop was most significant when switching from 3.75 to 2.5%. All the results presented in Figure 5 show that acid prehydrolysis of flax shives did not lead to an increase in the yield of vanillin and, accordingly, a decrease in the consumption of alkali.

The different effect of acid prehydrolysis on the efficiency of subsequent pine-wood and flax-shive oxidation may have been due to differences in the structure of lignins in coniferous and herbaceous plants. Unlike coniferous lignin, which consists mainly of guaiacylpropane structural units, herbaceous lignins are characterized by the presence of phenolpropane fragments unsubstituted by methoxy groups. Free positions 3 and 5 of the phenol ring of herbaceous lignins play a significant role in the processes of acid-catalyzed condensation, while only one fifth position of the ring is free in coniferous lignins. It should be noted that in the lignins of hardwoods most of the fifth positions in the rings are substituted by a methoxy group, and this explains the low tendency of these lignins for condensation in various technological processes compared to coniferous and herbaceous ones. As a result, hardwood lignins give higher yields of monomeric products compared to coniferous and, moreover, herbaceous lignins in both oxidation and hydrogenolysis processes [24].

The differences in the reactivity of herbaceous and coniferous lignins can be illustrated by Scheme 1. In the process of mild acid hydrolysis of lignins, acid-catalytic cleavage of simple ether bonds (Scheme 1A,B), for example, α-O-4 or β-O-4 occurs [17,24]; in parallel with it, condensation of lignins occurs along the 3,5-unsubstituted C-H positions of the phenol ring of the phenylpropane (Scheme 1C,D), and these processes lead to a decrease in the yield of aromatic aldehydes during subsequent oxidation. A greater number of condensable positions 3 and 5 of the herbaceous lignin ring can lead to a higher rate of their condensation. As a result, for woody lignins there is a small area of acid hydrolysis depth, in which splitting dominates over condensation (W_3Destr > W_4Cond) and leads to an increase in the yield of vanillin, while for herbaceous lignins it does not appear (W_1Cond > W_2Destr).

2.3. Influence of Stirring Speed on the Vanillin Accumulation Kinetics and Its Yield

It was previously shown [21] that the catalytic oxidation under study is limited by mass transfer and diffusion processes. The relations between the stirring speed of the magnetic stirrer and the initial rate of oxygen consumption or vanillin accumulation (estimated as the reverse time of attaining the maximum vanillin concentration) were found. They are described quantitatively by the linear dependences in logarithmic coordinates Equations (3) and (4)

$${\mathit{lnW}}_{O_2}=\mathrm{}1.88\mathit{lnN}+\mathit{const}$$

(3)

$$\mathit{ln}(1/t_{\mathit{max}})=1.86\mathit{lnN}+const$$

(4)

with the slope tgφ = b = 1.86 ± 0.06, which coincides with the index of the known model of mass transfer [27,28]. The physical sense of this tangent may be connected with efficiency of consuming the stirring energy to intensify mass transfer in the reactor [27].

Figure 6 shows similar dependences of vanillin formation on time and its formation rate on stirring speed in logarithmic scales. Conditions of the process differ from the data [21] only by alkalinity of the reaction mass. Change of NaOH concentration in the reaction mass from 5 wt.% [21] to 2.5% (Figure 6) increased this tangent of the logarithmic dependence to the value tgφ = 2.61 ± 0.27. CuO—Cu₂O transition (Figure 1) does not change the line character of the logarithmic dependence of the vanillin formation rate on the stirring speed (Figure 6b). This dependence was not previously investigated and requires further study. It should be noted that classic estimates of the maximum values of this tangent give the value of tgφ = 3 [29,30].

2.4. Influence of Alkali Loading on Vanillin Yields and Oxygen Consumption

Within the framework of these experiments, alkali was a reagent for creating a strongly alkaline medium necessary for the retro-aldol cleavage of the phenolpropane chain [3,31]. From this viewpoint, one can assume an extreme dependence of the alkali consumption per vanillin formed on the alkali load into the reactor. The results presented in

Table 2. Influence of the stirring speed and the alkali loading on the product yield and the consumption of reagents during the oxidation of Belarusian flax shives.

	Original/Pre-Hydrolyzed Flax Shives
	5(100)% NaOH * [19]	3.75(75)% NaOH *			2.5(50)% NaOH *
	500 rpm	300 rpm	500 rpm	700 rpm	300 rpm	400 rpm	500 rpm	600 rpm	700 rpm
Vanillin yield, wt.% on lignin	12.5/10.5	7.7/7.5	11.6/11.2	10.2	4.7	5.2	5.3/3.9	4.8	3.8
Oxygen consumption, mol	154.3/166.5	87.4/93.7	89.5/108.2	93.6	62.4	66.6	64.5/77.1	73.3	79.1
Oxygen consumption, mol/mol of vanillin	51.1/78.4	63.9/48.3	32.1/37.1	37.9	59.5	27.2	63.5/74.6	63.2	86.2
m (NaOH)/m (van), g/g	24.9/27.6	26.5/26.8	17.4/17.6	20.4	22.5	20.4	19.9/24.7	22.1	27.9

* Weight concentration, (weight by weight of flax shives, %).

demonstrate just such a dependence: the consumption of alkali per vanillin obtained was minimal with an average alkali load of 75% by weight per shive, and increased by 12–40% with an increase or decrease of the mass of alkali in the reactor. A similar minimum of oxygen consumption per the vanillin yield was observed at alkali load of 75% due to different dependences of decreasing the oxygen consumption and the vanillin yield on the value of alkali load.

With constant loading of alkali, insignificant but reliable extreme dependences of the maximum vanillin yields on the intensity of mass transfer appeared during the oxidation of both Belarusian shives (Table 2) and prehydrolyzed shives from the Tver region [21]. Consequently, the alkali consumption per vanillin obtained at an average (500 rpm) stirring speed of the agitator was minimal. Therefore, extreme dependences on oxygen consumption were quite obvious, both absolute and based on the obtained vanillin, with an optimal (75% by weight per shive) loading of alkali (Table 2). With an increase or decrease in alkali loading in the reactor, the extremum was smoothed out and the dependence of oxygen consumption on the rotation speed increased monotonically. The obtained results confirm the possibility of decreasing the alkali consumption in the process by a using slow stirring speed of the reaction mass.

3. Materials and Methods

All reactants were of the “pure for analysis” grade, supplied by Reachem (Moscow, Russia). Air-dried powdered (≤1 mm) flax shives (Linum usitatissimum, Rosinka type [32], harvest of 2020, Belarus origin, lignin and cellulose content 24.5 and 39.6%, respectively) were used for the experiments.

Prehydrolysis of flax shive powder is described in [21]. A sample of air-dried raw powder was mixed with HCl (28%, 10 mL per gram of the shives) and kept at 20 °C for 1 h. The powder was filtered, washed with distilled water to obtain the neutral reaction, dried at room temperature, and the mass loss of the flax shives was approximately 10%. Lignin and cellulose content in the prehydrolyzed shives was 24.6 and 42.9%, respectively.

Oxidation of Flax Shives

Experiments on the oxidation were conducted in a stainless steel autoclave [5,8] (volume 1 L, internal diameter 95 mm, height 180 mm) at 160 °C and oxygen partial pressure of 0.2 MPa) equipped with a magnetic stirrer (stir-bar diameter 10 mm and length 60 mm) at 300–700 rpm for 60 min, unless otherwise noted [21]. A reaction mass was prepared by successive addition of required amounts of catalyst (CuSO₄ 5H₂O, 37.5 g/L), sodium hydroxide (50 g/L), water (300 mL), and shives (50 g/L), unless otherwise noted, into a reactor under stirring all the while.

Oxygen was fed into the reactor from a calibrated buffer volume through a valve to maintain the constant pressure in the reactor. The amount of the consumed oxygen was calculated from the change in pressure inside the buffer volume. The reaction mass was taken from the cooled reactor after the oxidation. The liquid phase from the reactor was acidified by hydrochloric acid to pH 2. The precipitated tars were filtered, and the filtrate was sequentially extracted by three chloroform portions.

The vanillin content in the extract was then determined by gas–liquid chromatography (Chromos Engineering GH1000 chromatograph, auto sampler DAG-23, column 30 m × 0.32 mm, stationary phase 25% trifluoropropyl polysiloxane). The internal standard (anthracene) method was used to calculate the vanillin mass in the reactor. Vanillin yield was calculated as the ratio of vanillin mass obtained in an experiment to lignin mass fed into the reactor.

Distribution of catalyst particles (CuO and Cu₂O) on the surfaces of the treated flax shives was determined by SEM (Hitachi scanning electron microscope TM-4000, magnification ×500–3000, resolution 30 nm).

For phase-composition determination, X-ray diffraction analysis (XDR) was used. A diffractometer DRON-3 (Russia) was used, with a graphite monochromator, to scan from 5 to 70° (2θ) with an increase of 0.02°. The time constant was fixed at 1.0 s. A Kα1 radiation of Cu was used with a wavelength of 1.5406 Å. Phase content of the samples was determined by comparing the obtained spectra with the JCPDS Database.

4. Conclusions

Catalytic oxidation of flax shives with molecular oxygen has two main products—vanillin with a yield of up to 12 wt.% of lignin, and pulp. This yield exceeds the efficiency of Kraft lignin oxidation.

The obtained X-ray diffraction data showed that after the process the catalyst was in the pure Cu₂O phase at low mass-transfer intensity, and increasing the stirring speed of the reaction mass led to oxidation of Cu₂O mainly to the CuO phase. Hence, the catalyst oxidized the products of alkaline destruction of flax shives and was reoxidized by oxygen. The rate of the latter depended on the mass transfer intensity and stirring speed; therefore the CuO phase, the oxidized form of the catalyst, dominated at high stirring rates.

The particles of the catalyst, copper oxides, obtained in situ during the oxidation process had characteristic sizes of 0.5–5 μm and were distributed over the outer and inner surfaces of the shives particles. Cu₂O single crystals with a size of 3–5 μm have also been found. They are probably formed randomly as a result of the enlargement of copper oxide crystals by migration of copper ions through the aqueous phase, despite its high alkalinity.

The influence of acid prehydrolysis on the efficiency of subsequent pine-wood and flax-shive oxidation was compared. On the one hand, prehydrolysis of pine wood led to an increase the yield of vanillin in the subsequent process of its oxidation. On the other hand, a similar prehydrolysis of flax shives does not increase the yield of vanillin during the oxidation process. This difference is accounted for by the well-known variations in the structure of lignins of conifers and herbaceous plants. Unlike coniferous lignin, which consists mainly of guaiacylpropane structural units, herbaceous lignins are characterized by the presence of phenolpropane fragments unsubstituted by methoxy groups, which are more prone to the acid-condensation processes. As a result, for woody lignins there is a small area of acid hydrolysis depth, in which splitting dominates over condensation and leads to an increase in the yield of vanillin, while for herbaceous lignins this does not appear.

It was shown that the tangent of the previously established linear dependence of the vanillin accumulation rate on the stirring speed depends on the alkalinity of the reaction mass. This dependence was not previously investigated and requires further study.

It was also shown that stirring speed variation in the reactor optimizes the consumption of alkali and oxygen during the catalytic oxidation of flax shives into vanillin and cellulose.