**The Relationships between the Working Fluids, Process Characteristics and Products from the Modified Coaxial Electrospinning of Zein**

**Menglong Wang 1, Tao Hai 1, Zhangbin Feng 1, Deng-Guang Yu 1,\*, Yaoyao Yang <sup>1</sup> and SW Annie Bligh 2,\***


Received: 3 April 2019; Accepted: 11 July 2019; Published: 1 August 2019

**Abstract:** The accurate prediction and manipulation of nanoscale product sizes is a major challenge in material processing. In this investigation, two process characteristics were explored during the modified coaxial electrospinning of zein, with the aim of understanding how this impacts the products formed. The characteristics studied were the spreading angle at the unstable region (θ) and the length of the straight fluid jet (*L*). An electrospinnable zein core solution was prepared and processed with a sheath comprising ethanolic solutions of LiCl. The width of the zein nanoribbons formed (*W*) was found to be more closely correlated with the spreading angle and straight fluid jet length than with the experimental parameters (the electrolyte concentrations and conductivity of the shell fluids). Linear equations *W* = 546.44*L* − 666.04 and *W* = 2255.3θ − 22.7 could be developed with correlation coefficients of *R*wl<sup>2</sup> = 0.9845 and *R*wθ<sup>2</sup> = 0.9924, respectively. These highly linear relationships reveal that the process characteristics can be very useful tools for both predicting the quality of the electrospun products, and manipulating their sizes for functional applications. This arises because any changes in the experimental parameters would have an influence on both the process characteristics and the solid products' properties.

**Keywords:** coaxial electrospinning; length of straight fluid jet; spreading angle; nanoribbons; linear relationship

### **1. Introduction**

For polymer processing engineering, a key requirement is to be able to accurately interrelate the experimental conditions and the properties of the final products [1–3]. This is particularly challenging when the final products are nanoparticles or nanofibers [4–9]. Both of them can be generated by electrohydrodynamic atomization (EHDA) using electrostatic energy [10–16], and while there are numerous reports of such fabrication processes, it remains the case that it is extremely difficult to predict the outcome of a given experiment.

Figure 1 presents a schematic depicting the experimental parameters that exert a significant influence on the final polymer nanofibers generated in the simplest electrospinning experiment, which involves a single working fluid. In electrospinning, the working fluid and electrostatic energy are brought together at the nozzle of the spinneret [17–21]. After exiting the spinneret, the working fluid is forced to bend and whip, and during this process it is solidified into nanofibers at an extremely rapid speed [22–27]. Based on this working procedure, the key experimental parameters can be divided into three categories (see Figure 1). Correspondingly, the resultant nanofiber diameter (*D*) is a function of

the working fluid properties (*w*), the operational conditions (*o*), and the environmental parameters (*e*): i.e., *D* = *f* (*w*,*o*,*e*).

**Figure 1.** A diagram showing the single-fluid electrospinning process and the experimental parameters exerting influence on the diameters of the polymer nanofibers generated.

Over the past two decades, electrospinning has developed very rapidly, with potential applications of polymer nanofibers having been proposed in a myriad of scientific fields [28–31]. In addition, the simple single-fluid electrospinning process has advanced to two-fluid coaxial and side-by-side processes, and tri-fluid coaxial and combined coaxial/side-by-side processes. It has proven to be possible to perform the electrospinning process even when one or more of the working fluids cannot on its own be processed: For instance, in modified coaxial electrospinning, a spinnable core solution is partnered with an unspinnable sheath fluid. These novel processes permit the production of nanofibers with increasingly complicated nanostructures [32–37]. As a result, in the literature, there are numerous publications that explore the influence of a single parameter on the nanofibers or nanostructures prepared by electrospinning, elucidating relationships for manipulating the products (mainly in terms of diameter, but also for other properties such as morphology, surface smoothness and the details of the nanostructure) [38–43]. However, there are a number of experimental parameters that can simultaneously exert an influence on the final products [44–47]. For example, the properties of working fluid include polymer concentration (*C*), viscosity (η), surface tension (δ), and conductivity (σ); the operational conditions include the applied voltage (*V*), the fluid flow rate (*F*), and the fiber collection distance (*L*); and the environmental parameters comprise of temperature (*T*), humidity (*H*) and the possible vacuum (*U*) (Figure 1).

Thus, although a lot of effort has been expended to predict and manipulate the diameters of electrospun nanofibers, the results are typically far from satisfactory [48,49]. During electrospinning, almost all the experimental parameters can influence the working process, and furthermore, they are not independent variables, and can also influence each other. For example, the flow rate of the working fluid and the applied voltage need to be matched, or droplets of working fluid may fall directly onto the fiber collector. Thus, although many mathematical models have been put forward for a particular working fluid [48], they often fail to be applicable to other situations. Although the experimental parameters have drawn intensive attention, it is strange that the detailed steps in the process of electrospinning have received very limited attention. These include the formation of the Taylor cone, the ejection of a straight fluid jet, and also the bending and whipping region [50,51]. These individual steps are influenced by all the experimental parameters, and thus can be directly controlled by researchers. It is hypothesized that the nature of each of these stages of the spinning process should have a distinct relationship with the final nanofiber properties, particularly their sizes.

Here, for the first time, we design a method to elucidate the interrelated relationships between the working fluids, the electrospinning process characteristics, and the nanoribbons formed during the modified coaxial electrospinning of zein. Zein is one of the best understood plant proteins. Extracted from maize, it has been widely used as a coating for candy, nuts, fruit, and pharmaceuticals. Zein can be processed into resins and other bioplastic polymers, which can be extruded or rolled into a variety of plastic products [52]. Zein has good processability using both these traditional technologies and advanced technologies such as electrospinning and electrospraying [53–55]. Here, it was selected as a model system for a detailed exploration of the individual stages in the electrospinning process. A series of modified coaxial electrospinning processes were carried out and several types of zein nanoribbons were fabricated. The working processes were digitally recorded and quantitatively described in terms of length of the straight fluid jet and the spreading angle of the unstable region, and these were interrelated with both the initial conductivity and the final zein nanoribbon widths.

### **2. Materials and Methods**

### *2.1. Materials*

Zein (98% purity) was purchased from Shanghai Hewu Biotechnology Co., Ltd. (Shanghai, China). Anhydrous ethanol and lithium chloride were obtained from the Husheng Reagent Co., Ltd. (Shanghai, China). Water was double distilled just before use.

### *2.2. Modified Coaxial Electrospraying*

The core fluid consisted of 28 g zein in 100 mL of a 75%/25% (*v*/*v*) ethanol/water mixture, which showed a yellow color. Four LiCl solutions in ethanol (at 0, 5, 10, and 20 mg/mL) were utilized as the sheath fluids, and the resultant nanoribbons were labeled as Z1, Z2, Z3, and Z4, respectively. The conductivities of the sheath fluids were assessed using a DDS-11 digital conductivity meter (Shanghai Rex Co-perfect Instrument Co., Ltd., Shanghai, China).

A homemade system was employed to conduct all the electrospinning processes. This consisted of two syringe pumps (KDS100 and KDS200, Cole–Parmer, Vernon Hills, IL, USA), a power supply (ZGF200, 60 kV/2 mA, Wuhan Huatian Corp., Wuhan, China), a homemade concentric spinneret, and an aluminum-coated flat piece of cardboard as the collector. The ambient temperature and humidity were 21 ± 4 ◦C and 53 ± 6%, respectively. All the working processes were recorded using a digital camera (PowerShot A490; Canon, Tokyo, Japan). Following optimization, the collection distance and voltage were fixed at 20 cm and 17 kV, respectively.

### *2.3. Morphology of the Prepared Nanoparticles*

The surface morphologies of the electrospun products were observed by scanning electron microscopy (SEM; Quanta FEG450, FEI Corporation, Hillsboro, OR, USA) at 20 kV acceleration voltage. Before observation, the samples were sputter-coated with gold under vacuum. The images were analyzed using the ImageJ software (National Institutes of Health, Bethesda, MD, USA), with measurements taken at over 100 different places to determine the average ribbon diameter.

### **3. Results and Discussion**

### *3.1. Implementation of Modified Coaxial Electrospinning*

Traditionally, coaxial electrospinning is carried out using an electrospinnable sheath fluid to encapsulate either a core liquid which may be spinnable or unspinnable [18,24]. Some years ago, Yu and co-workers expanded this concept to develop the modified coaxial process, with an unspinnable liquid as the sheath fluid (Figure 2).

**Figure 2.** The modified coaxial electrospinning process, which permits a range of novel structures to be obtained through the unspinnable sheath fluid.

The homemade concentric spinneret and the electrospinning apparatus used in this work are shown in Figure 3. The spinneret (Figure 3a) consists of a narrow metal capillary (inner diameter 0.3 mm, wall thickness 0.1 mm) nested into an outer capillary (inner diameter of 1.2 mm, wall thickness 0.2 mm). Two syringe pumps were employed to drive the core and shell liquids to the spinneret (Figure 3b). The yellow zein solution was directly guided to the inner needle of the spinneret through a plastic syringe, whereas the sheath LiCl solution was pumped to the spinneret through elastic silicon tubing. An alligator clip connects the spinneret to the power supply and carries electrostatic energy to the working fluid (Figure 3c).

**Figure 3.** The apparatus used for modified coaxial electrospinning: (**a**) The home-made concentration spinneret; (**b**) the arrangement of apparatus; and (**c**) the connection of the power supply and working fluids with the spinneret.

### *3.2. The Working Processes and the Resultant Zein Nanoribbons*

The electrospinning process consists of three successive stages: The formation of a Taylor cone, the straight fluid jet emitted from the Taylor cone, and the unstable region, which is composed of numerous bending and whipping loops. The formation of the Taylor cone is a balance between the electrical force exerted on the droplets exiting the spinneret, and the surface tension of the working fluids. When the conductivity of the working fluid increases, the electrical forces should increase correspondingly. Thus, an increase in the LiCl concentration in the sheath fluid is expected to result in a stronger electrical force being applied to the working fluids. Under the same applied voltage and spinneret-to-collector distance, this force will greatly change the behaviors of the working fluids. Digital photographs of these are given in Figure 4. As the LiCl concentration increased from 0 to 5, and from 10 to 20 mg/mL, the length of the straight fluid jet decreased from 3.3 ± 0.4, to 2.9 ± 0.3, and from 2.4 ± 0.3 to 2.2 ± 0.2 mm, respectively. Meanwhile, the spreading angles of the unstable region increased from 51 ± 4◦ to 59 ± 5◦, and from 68 ± 4◦ to 77 ± 6◦.

**Figure 4.** The changes of spreading angle and the length of straight fluid jet with the increase of LiCl in the sheath solution (mg/mL): (**a**) 0; (**b**) 5; (**c**) 10; (**d**) 20.

SEM images of the resultant nanoribbons and their diameter distributions are shown in Figure 5. All the ribbons have a linear morphology. No beads or spindles are found in the ribbons, suggesting the core zein solution has good electrospinnability. Nanoribbons Z1, Z2, Z3, and Z4 have an estimated width of 1.12 ± 0.14, 0.91 ± 0.12, 0.58 ± 0.09, and 0.52 ± 0.07 μm, respectively.

**Figure 5.** SEM images of resultant zein nanoribbons, with their width distributions. (**a**) Z1; (**b**) Z2; (**c**) Z3; (**d**) Z4.

### *3.3. The Influence of Conductivity on the Behavior of the Working Fluids*

Although a single-step and straightforward process for creating nanoribbons, the electrospinning process is in fact very complicated. This complexity is reflected in two ways. First, the process involves the overlap of multiple disciplines such as hydrodynamic science, polymer science and rheology, and electric dynamics. Second, a small change in the working fluid properties can greatly influence the process and its products.

As the concentration of LiCl increased, the conductivity of the sheath solution also rose (Figure 6a). This increase in conductivity will make the solution subject to stronger electrical forces, which in turn alter

the behavior of both the sheath and core working fluids. The length of the straight fluid jets gradually decreased with conductivity in a linear fashion, as shown in Figure 6b: *L* = 3.38 <sup>−</sup> 5.25 <sup>×</sup> 104 <sup>σ</sup>*,* with a correlation coefficient of *R*Lσ<sup>2</sup> = 0.9761. Similarly, the spreading angle of the unstable region gradually increased with conductivity (Figure 6c). A linear equation θ = 48.775 + 0.011 σ can be fitted to the data, giving a correlation coefficient of *R*θσ<sup>2</sup> = 0.9296. The clearly linear nature of the plots in Figure 6b, c and the high *R*<sup>2</sup> values obtained show there are clear causal relationships here.

**Figure 6.** The influence of the sheath fluid conductivity on the behavior of the working fluids: (**a**) The relationship between LiCl concentration and solution conductivity; (**b**) the decrease in the length of the straight fluid jet with an increase of conductivity; (**c**) the increase of spreading angle with rising conductivity.

In literature, numerous publications have investigated the electrospinnability of a certain working fluid, which is mainly determined by the type of filament-forming polymer, its concentrations in the working fluid, and the applied voltage. After the past two decades' effort, near 200 polymers can be processed into fibers using electrospinning. However, few efforts have been focused on the behaviors of working fluids within their electrospinnable windows. Knowledge about the adaptability of working fluids under the high electrical field should be useful for manipulating the fluid processing process in a more intentional manner.

### *3.4. The E*ff*ect of Sheath Working Fluid Properties on the Width of Zein Nanoribbons*

A wide variety of experimental parameters have been investigated in terms of their effect on the properties of electrospun fibers, and the solution conductivity of working fluid is recognized as being of major importance [56]. In this study, the electrolyte LiCl was added only into the sheath working fluid, because charges are always concentrated on the surface of the Taylor cone. The width of the zein ribbons produced is clearly correlated with the LiCl concentration in the sheath fluid, with a good fit to the data obtained with the linear equation *W* = 1033.2*C* − 26.9 (*R*<sup>1</sup> <sup>2</sup> = 0.9297; Figure 7a). A similar linear equation is observed when plotting ribbon width as a function of sheath solution conductivity (*W* = 11177.86 − 0.29σ; *R*<sup>2</sup> <sup>2</sup> = 0.9639). These linear equations suggested that the LiCl concentration and conductivity of sheath working fluid directly influenced the width of the zein nanoribbons fabricated. These equations can hence be exploited to predict the size of the products from this electrospinning process, and provide useful information for optimizing the working processes.

It is a common strategy to optimize the experimental conditions through simultaneous investigations on several levels of an experimental parameter, just as here with the LiCl concentration. However, only a small part of the related publications has taken a further step to disclose the inherent relationship between the vital properties of the working fluid with the final product's quality. Here, the conductivity of sheath LiCl solution showed a better linear relationship with the width of zein ribbons than the LiCl concentration. Thus, among many other solution properties such as surface tension, viscosity, and rheological properties, conductivity is the most important property of LiCl solution that exerted influences on both the working processes, and also the resultant ribbons' quality.

**Figure 7.** Correlations between the width of electrospun zein nanoribbons with: (**a**) The LiCl concentration; and (**b**) the conductivity of the sheath fluid.

### *3.5. Correlations between the Width of Electrospun Zein Nanoribbons and the Detailed Steps of Electrospinning*

The detailed observations of the electrospinning process discussed in Section 3.2 have a very close relationship with the size of the zein ribbons produced (Figure 8). A linear equation relates the width of the ribbons to the length of the straight fluid jet (*W* = 546.44*L* <sup>−</sup> 666.04; *R*wl<sup>2</sup> = 0.9845). Similarly, a linear equation *W* = 2255.3θ − 22.7 connects the width of the ribbons with the spreading angle (*R*wθ<sup>2</sup> = 0.9924). These relationships show that these parameters can be very useful tools for predicting the properties of the ribbons fabricated.

**Figure 8.** The correlations between the width of electrospun zein nanoribbons and: (**a**) The length of the straight fluid jet; and (**b**) the spreading angle of the unstable zone.

Right from the rebirth of electrospinning, a wide variety of experimental parameters have been studied to disclose their potential roles during the electrospinning processes. These parameters can all be manipulated by the researchers directly and changed within a certain range, which are concluded in Figure 1. However, these parameters often result in interrelated influences. For example, an increase of LiCl concentration resulted in a larger conductivity, but also changed the working fluid's surface tension, viscosity, and exerted on the effect of applied voltage. Thus, although many publications have reported the relationships between a certain experimental parameter and the final nanoproducts' size. It is difficult to disclose their relationship in an accurate manner. In contrast, the process characteristics, similarly as the final product to be influenced systematically from all the experimental parameters, have the essential advantages over the processing parameters in predicting the final nanoproducts' size, and in providing useful information for accurate and robust manipulation of the processing process.

### *3.6. The Role of Process Characteristics*

A schematic diagram of the modified coaxial electrospinning process is presented in Figure 9. Initially, the sheath LiCl solution surrounds the core zein solution to form a compound Taylor cone. The two fluids come through the straight fluid jet and enter the unstable region together. During the early stages of the unstable region, the sheath solution will be evaporated, and then later, the core zein solution will be gradually dried during the drawing processes. A series of different forces

will be exerted on the working fluids, such as the force between the two electrodes (*F*1) and gravity (*G,* which can often be neglected). Within the bending and looping fluid jets, repulsive forces will include those between different loops (*F*2) and those within the different parts of a single loop (*F*3). It is the *F*<sup>3</sup> forces that draw and narrow the working fluids. The spreading angle will be a parameter that reflects the combined actions of *F*1, *F*2, and *F*3. An increase in sheath solution conductivity will increase all three forces. An increase in *F*<sup>1</sup> would act to decrease the spreading angle. However, an increase in *F*<sup>2</sup> would make the fluid travel time increase during the drawing process, and thus provide a trend of enlarging the spreading angle. In addition, an increase in *F*<sup>3</sup> would make the loops larger, and correspondingly increase the spreading angle. Thus, the combined effects of *F*<sup>2</sup> and *F*<sup>3</sup> appear to have a more marked influence on the electrospinning process than *F*1, and as a result, the greater the conductivity of the sheath fluid, the larger the spreading angle observed. Similarly, another process characteristic, i.e., the length of straight fluid jet, has received the influences of LiCl concentration directly and comprehensively.

**Figure 9.** A diagram showing the formation mechanism of electrospun nanoribbons through the modified coaxial electrospinning.

In the biomedical applications of electrospun nanofibers, whether for tissue engineering or advanced drug delivery systems, the accurate manipulation of nanofiber diameter is very important for the fibers' functional performances [57–59]. This work reveals that the process characteristics (the length of straight fluid jet and the spreading angle of unstable region) have a close linear relationship with the final nanoribbon width, and can provide useful information for manipulating the working processes, and developing products with the desired physical properties.

### **4. Conclusions and Perspectives**

Using an electrospinnable zein solution as the core fluid and LiCl solutions as the sheath working liquids, a series of modified coaxial electrospinning processes were performed, and a number of zein nanoribbons successfully prepared. The nanoribbon width (*W*) was found to be directly correlated with the concentration of LiCl (*C*) and the conductivity of the sheath fluid (σ), with linear relationships of the form *W* = 1033.2*C* − 26.9 (*R*<sup>1</sup> <sup>2</sup> = 0.9297), and *W* = 11177.86 <sup>−</sup> 0.29<sup>σ</sup> (*R*<sup>2</sup> <sup>2</sup> = 0.9639) determined. Further, the width of the zein nanoribbons (*W*) were found to have still closer linear relationships with the spreading angle in the unstable region (θ), and the length of the straight fluid jet (*L*) (*W* = 546.44*L* − 666.04 and *<sup>W</sup>* <sup>=</sup> 2255.3<sup>θ</sup> <sup>−</sup> 22.7; *<sup>R</sup>*wl<sup>2</sup> <sup>=</sup> 0.9845 and *<sup>R</sup>*wθ<sup>2</sup> <sup>=</sup> 0.9924, respectively).

Today, electrospun nanofibers are rapidly approaching commercial applications in several fields such as drug delivery, food packaging, water treatment, and air filtration [60,61], and the production of electrospun nanofibers on a large scale is now possible [62]. Two important issues will require attention for accelerating nanofiber-based commodities to the market. One is the accurate and robust control of the processing process during electrospinning. The second is the prediction and maintenance of the nanofiber quality. For resolving these two issues, the characteristics of the working process itself offer a powerful source of information, and have advantages over the processing parameters (i.e., those that can be manipulated directly by the operator). This is because these working process characteristics manifest the simultaneous influence of all the processing parameters, as do the fibers produced. Thus, it is anticipated that they can act as a useful tool for stabilizing the working process, for systematic manipulation of the processing parameters, and for accurately predicting the resultant nanofiber size. Similar observations have been noted for electrospraying, an alternative electrohydrodynamic atomization process [63].

**Author Contributions:** Conceptualization, D.G.Y.; data curation, M.W.; formal analysis, Y.Y. and T.H.; funding acquisition, D.G.Y.; investigation, M.W., T.H. and Z.F.; methodology, D.G.Y.; project administration, D.G.Y. and S.W.A.B.; writing—original draft, M.W. and D.G.Y.; writing—review and editing, D.G.Y. and S.W.A.B.

**Funding:** The National Natural Science Foundation of China (No.51373101 and 51803121) and USST college student innovation projects (SH10252194 & 10252324/330) are appreciated.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **E**ff**ects of Fiber Surface Grafting with Nano-Clay on the Hydrothermal Ageing Behaviors of Flax Fiber**/**Epoxy Composite Plates**

### **Anni Wang 1,2,3, Guijun Xian 1,2,3,\* and Hui Li 1,2,3**


Received: 14 June 2019; Accepted: 11 July 2019; Published: 31 July 2019

**Abstract:** Flax fiber has high sensitivity to moisture, and moisture uptake leads to the decrease of mechanical properties and distortion in shape. This paper attempts to graft flax fabric with nano-clay, with assistance from a silane-coupling agent, in order to improve hygrothermal resistance. The nano-clay grafted flax fabric reinforced epoxy (FFRP) composite produced through vacuum assisted resin infusion (VARI) process were subjected to 80% RH chamber for 12 weeks at 20, 40 and 70 ◦C, respectively. Moisture uptake, dimensional stability, and tensile properties was studied as a function of humidity exposure. Through SEM and FTIR, the effects of hygrothermal exposure was elucidated. In comparison to control FFRP plates, nano-clay grafting decreases saturation moisture uptake and the coefficient of diffusion of FFRP by 38.4% and 13.2%, respectively. After exposure for six weeks, the retention rate of the tensile modulus of the nano-clay grafted flax fiber based FFRP increased by 33.8% compared with that of the control ones. Nano-clay grafting also reduces the linear moisture expansion coefficient of FFRPs by 8.4% in a radial direction and 10.9% in a weft direction.

**Keywords:** flax fiber; nano-clay; water uptake; hygrothermal properties

### **1. Introduction**

Flax fiber is a natural fiber that is biodegradable, renewable and environment-friendly compared to traditional carbon fiber and glass fiber. Flax fiber possesses relatively higher tensile strength compare to other natural fibers, which are considered as a high performance natural fiber. Although its water resistance properties are not very good compared to thermoplastic polymer, epoxy resin has been widely used as a resin matrix for polymeric composites. Epoxy resin has good wettability with flax fiber, which provides good interface properties of flax fiber reinforced epoxy polymer (FFRP) composites. FFRPs are widely used in decorative materials, automobiles and other fields due to their high specific modulus. However, due to the high hydrophilicity of flax fibers, their poor durability limits the development of FFRPs.

The special chemical composition and structure of flax fiber lead to high hydrophilicity. Flax fiber is composed of plant cells, whose main structure are cell walls [1,2]. Like most plant cells, flax fiber cell walls consist mainly of cellulose, hemicellulose and pectin [3–5]. Cellulose, hemicellulose and lignin are made up of macromolecular chains of glucose and contain a large number of hydroxyl groups, which can adsorb water molecules [6]. Cellulose, called micro fibrils, is wrapped by hemicelluloses and lignin, and glued together or linked by hydrogen bonds [7]. Crystalline cellulose cannot store water, but water molecules can store them inside the amorphous hemicellulose and lignin [8,9]. In

addition, flax fiber cells contain cell cavities that can store water. Thus, compared to traditional fiber such as carbon fiber, flax fiber shows high water absorbability.

Owing to the water absorption of fibers and the storage of water molecules at the interface between fiber and polymer, FFRPs also exhibit high water absorption [10,11]. Researchers have done a lot of research on the water absorption process of nature fiber reinforced polymer composites (NFRPs). The water absorption process of NFRPs is consistent to Fick law at lower temperature [12–14]. Both saturated water absorption and rate of water absorption of NFRP composites samples increase as the fiber volume fraction increase [14,15]. At the same time, the diffusion rate of water molecules in NFRP is related to temperature. The higher the temperature, the faster the diffusion rate. The researchers also studied the deterioration of mechanical properties of NFRP in a hygrothermal environment. On the other hand, the absorbed moisture results in more detrimental effects on the mechanical properties of NFRPs since the water not only interacts with fiber and polymer matrices, physically, i.e., plasticization, and/or chemically, i.e., hydrolysis, as in the unfilled system, but it also attacks the fiber–matrix interface [10]. Thus, the decrease of mechanical properties of NFRP caused by water molecules entering composite materials. Hongguang Wang et al. put the ramie fiber reinforced composites at 20 °C and 40 ◦C under 100% RH, and found that only after 1 day, both of the flexural strength and modulus were reduced dramatically and the deterioration rate of strength and modulus slowed down with the extension of immersion time [16].

Surface treatment of fiber is a good way to improve the properties of FRPs by improving the interface properties [17–19]. In order to promote the application of natural fiber, its hygrothermal resistance properties need to be improved by fiber treatment. At present, the main methods to improve the hygrothermal properties are as follows: removing the active hydroxyl groups on the surface of the fibers by chemical reaction, reducing the adsorption sites of water molecules; coating hydrophobic coatings on the surface of the fibers, hindering the diffusion of water molecules in the fibers. H. Alamri et al. used n-SiC fill cellulose fiber reinforced epoxy eco-nanocomposites and found that saturated water absorption of the composites decreased with the increase of n-SiC content [20]. Anna Dilfi K.F. et al. studied the durability of jute fiber reinforced epoxy composites treated by alkali and silane coupling agents and found that the deterioration of mechanical properties of jute fiber reinforced composites after chemical treatment are less than that of untreated ones [21]. Gao Ma et al. found that both alkali and silane treatments of jute fiber reduced water absorption and enhanced the tensile strength of the resulting jute fabric/epoxy composites [22].

The hydrophobicity and lamellar structure of nano-clay makes it possible to improve the durability of flax fibers after grafting onto the surface of flax fibers. Nano-clay is a kind of special nano-material with a large specific surface area, which is composed of two tetrahedral silicon atoms and eight sides of aluminum or magnesium hydroxide [23]. Nano-clay exhibits a hydrophobic lamellar structure, so polymer-clay nanocomposites have received much attention due to significant increase in mechanical properties, and a moisture barrier [24]. Polymer nanocomposites contain relatively small amounts (typically less than 5 wt. %) of nanometer-sized filler particles, which, if properly dispersed, have been found to cause significant reductions in both gas and water vapor permeability [25]. Neetu Malik et al. mixed biodegradable polymer polycaptalactone (PCL) and organic modified montmorillonite clay (OMMT) and found that with an increase in weight percentage of OMMT within the bio polymer films, the moisture absorption value of bio-nanocomposite films reduced rapidly from 34.4% to 22.3% [26].

In this paper, the method of improving the durability of flax fiber reinforced composites by nano-modification of flax fiber is studied. The effects of nano-clay grafting on the water absorption process as well as deterioration of mechanical properties of the related FRPs in hygrothermal environment were investigated. The mechanism of surface grafted nano-clay on durability of FFRPs was studied.

### **2. Experimental**

### *2.1. Materials*

The organic nano-clay, belonging to a high purity montmorillonite organic ammonium derivative, was purchased from Lingshou Huarun Mineral Factory (Shijiazhuang City, China). The nano-clay used in the present study is an organically treated montmorillonite (OMMT) with ammonium. Bi-directional flax fiber fabrics was purchased from Harbin Flax Textile Co., Ltd., (Harbin, Heilongjiang Province, China). The density of fabric is 1.5 g/cm3 and nominal thickness is 0.16 mm. The silane coupling agent used in the current work is 3-Triethoxysilylpropylamine (APTES, KH550), purchased from Chengong Silicon Company (Nanjing, China) with purity of 98%. The epoxy resin used is room temperature impregnating adhesive (TS), purchased from Shandong Dagong Composite Material Co., Ltd (Linyi, China).

### *2.2. Surface Grafting of Flax Fabric*

According to the author's previous work [27], the preparation process of nano-clay grafted flax fiber/epoxy resin composite is shown in Figure 1. The details can be found in Ref. [28]. Nano-clay were dispersed into a solvent (ethanol: distilled water = 4:1 by weight) with an ultrasonic bath. The organic nano-clay content in the dispersion medium is 1.3 wt. %. After 1 h of ultrasonic treatment at room temperature, flax fabric and 1wt % KH550 (to the dispersion medium) was added to the solution and sonificated for 15 min at room temperature. Finally, the fiber was washed in distilled water for 5 min. The composites were prepared by vacuum-assisted resin infusion process.

**Figure 1.** Methods of nano-clay grafted onto flax fiber and preparation of flax fiber reinforced polymer composites (FFRPs).

Figure 2 shows the SEM pictures of untreated fiber and nano-clay grafted flax fiber. When compared to the untreated one, lamellar nano-clay can be seen on the surface of the flax fiber. Due to its hydrophobicity, the presence of nano-clay enhances the barrier properties of the materials by creating tortuous pathways for water molecules to diffuse into flax fiber, which leads to a reduction in absorbed water and the coefficient of diffusion. Lamellar and hydrophobic nano-clays inhibit the diffusion of water molecules more obviously than other nano materials.

**Figure 2.** SEM photos of flax fiber: (**a**) untreated fiber (**b**) nano-clay grafted fiber.

### *2.3. Hydrothermal Environment Conditions*

In the experiment, three different humid and hot environments were prepared using a saturated salt solution (Table 1). At 20 °C and 40 ◦C, the saturated potassium bromide solution can have ambient humidity of 80 RH%. At 70 ◦C, the humidity is 80% with saturated potassium chloride solution. As shown in Figure 3, the humidity chamber is prepared with a saturated salt solution in glass boxes at different temperatures. The humidity and temperature in the sealed glass box are monitored. When the temperature and humidity are stable, FFRP samples are placed in the sealed box. The condensed water droplets are prevented from falling on to the surface of the sample during the experiment. Before placing into the hygrothermal environment, the FFRP samples were dried in an oven at 70 ◦C for 48 h.

**Table 1.** Preparation of a hydrothermal environment with a saturated salt solution.

**Figure 3.** Schematic diagram showing preparation of a hydrothermal environment with a saturated salt solution.

### *2.4. Characterization*

### 2.4.1. Moisture Uptake

According to ASTM D5229, the sample for moisture uptake is 76 mm × 25 mm, more than 5 g. Each group contains eight samples. Moisture uptake was detected by periodically recording the mass of the sample. Samples taken out of the humidity chamber were weighed using an electronic balance with accuracy of 0.01 mg. The presented data are an average for eight coupons. The immersion periods were set as 4 h, 8 h, 12 h, 1 day, 2 days, 4 days, 1 week, 2 weeks, 3 weeks, 4 weeks, 5 weeks, 8 weeks and 12 weeks.

### 2.4.2. Fourier Transform Infrared Test

Flax fiber fabrics were cut into powder—about 2 mg—to make fiber specimens. Fourier transform infrared (FTIR) spectra of the fiber specimens (control and treated flax fiber yarn) were recorded on a spectrometer (Spectrum 100, Perkin Elmer Instruments, Boston, MA, USA) at a range of 400–4000 cm<sup>−</sup>1.

### 2.4.3. FBG Monitoring

Fiber Bragg Grating (FBG) Demodulator is manufactured by Shanghai Qipeng Engineering Materials Technology Co. (Shanghai, China) The FBG sensor for temperature monitoring was encapsulated in a steel capillary. With protection of the steel capillary, the FBG sensor was in a strain-free condition and affected only by the temperatures. To measure the internal stress during the aging, the FBG sensors were embedded in the interlayers of the FRP wet layups. The FBG sensors were carefully located in the fiber direction or perpendicular to the fibers and placed in the middle of the plate along the depth of the layer and near the central part in the planar direction. The fabricated resin samples were cured with the same curing procedure as the FRP samples. In this work, FFRPs were embedded with two FBG sensors—one along the fiber direction and one in the perpendicular direction. The fellow equations describe the strain measurement with the FBG sensors:

$$
\Delta\lambda = \Delta\lambda\_1 + \Delta\lambda\_2 \tag{1}
$$

$$
\Delta\lambda\_1 = \alpha\_\ell \Delta\varepsilon \tag{2}
$$

$$
\Delta\lambda\_2 = \alpha\_T \Delta T \tag{3}
$$

where Δλ is FBG wavelength change; Δλ<sup>1</sup> is strain-induced FBG wavelength change; Δλ<sup>2</sup> is temperature-induced FBG wavelength change; αε is strain sensitivity coefficient, which is 1.2 pm/με for the FBG used in this experiment; α*<sup>T</sup>* is temperature sensitivity coefficient, which is 10 pm/ ◦C for FBG used in this experiment; Δε is strain value; ΔT is temperature change.

The wavelength change obtained by the FBG can be converted into strain:

$$
\Delta \varepsilon = \frac{\left(\Delta \lambda - \Delta \lambda\_2\right)}{a\_\ell} \tag{4}
$$

### 2.4.4. Mechanical Property Test

Tensile properties of the FFRP plate samples were tested according to ASTM D3039. The dimensions of the specimens were 250 mm × 15 mm. The crosshead speed is set as 2 mm/min. The samples were removed from the hygrothermal environment chamber at regular time intervals, i.e., 2, 4, and 6 weeks. Five samples were repeated for one condition.

### 2.4.5. Scanning Electron Microscope (SEM) Test

For the SEM test, all the specimens were sputter coated with gold for 15 min (Gatan Model 682 Precision etching coating system) before SEM analysis to improve their electrical conductivity. The control and treated FFRP samples were observed through a scanning electron microscope with accelerated voltages of 20–30 V.

### **3. Results and Discussion**

### *3.1. Moisture Absorption of FFRPs*

The water absorption percentage of FFRPs in hydrothermal environments can be calculated by the following equation:

$$
\Delta M(t) = \frac{m\_t - m\_0}{m\_0} \times 100\tag{5}
$$

where Δ*M* is moisture uptake, *m*<sup>0</sup> and *mt* are the mass of specimen before and after exposure time of *t*.

Figure 4 shows the percentage of weight gain of untreated (C), silane coupling agent treated (S) and 1.3 wt. % nano-clay grafted flax fiber reinforced epoxy composites (O) at 70 ◦C under 80% RH from 0 to 3 months. Fick's diffusion model describes the dynamic equilibrium of the water absorption process when the material reaches a certain degree of water content. In the initial stage, the material's water absorption is proportional to the square root of exposure time. With the passage of time, the moisture absorption rate decreases dramatically, and finally reaches a dynamic balance. As shown in Figure 4, the moisture uptake process for all samples is linear in the beginning, then levels off, indicating a Fick diffusion process. R<sup>2</sup> represents the deviation between the experimental data and fitting results with the Fick's model.

**Figure 4.** Water absorption process of FFRPs at 70 ◦C under 80% RH. (C: untreated FFRPs; S: silane coupling agent treated FFRPs; O: nano-clay grafted FFRPs).

According to Fick's law, the diffusion of water in materials is controlled by the concentration gradient of diffused substances. To obtain the water uptake and diffusion parameters, the curve fitting method was adopted with two-stage water uptake models:

$$M\_t = M\_m \left\{ 1 - \exp\left[ -7.3 \left( \frac{Dt}{h^2} \right)^{0.75} \right] \right\} \tag{6}$$

where *Mm* is the maximum weight gain, *Mt* is the weight gain at time *t* and h is the half of thickness of the composite.

Using Equation (2), the water diffusion parameters of the FFRPs are obtained as shown in Table 2. The nano-clay grafted flax leads to a remarkable decrease in saturated moisture uptake by 38.4% and 15.4% compared with the control and silane-treated ones. Similarly, nano-clay grafting also results in the reduction of the diffusion coefficient by 13.2% and 56.6% than that of the control and silane-treated ones. It is worth noting that silane-treated FFRPs show higher coefficient of diffusion and lower water uptake compared with the untreated ones. The lower water uptake of silane treated FFRPs is attributed to the reduced hydroxyl groups of flax fibers, which were reacted with silane. Note that the silane reaction is not complete, and some silane coupling agents did not form a network. Those relatively low molecules existing between the fiber and resin matrix accelerate water absorption. As a result, the FFRPs with silane-treated FFRPs show a higher coefficient of diffusion.

The water absorption process of nano-clay grafted FFRPs is also affected by temperature. Figure 5 shows the water absorption process of nano-clay grafted FFRPs at 20, 40 and 70 ◦C under 80% RH, which all followed a Fickian diffusion process. After aging for three months, the samples at 40 and 70 ◦C were saturated with water and had the same saturated water absorption, while the samples at 20 ◦C under 80% RH reach saturated water absorption and had lower saturated water absorption. As shown in Table 3, with the increase of temperature, the diffusion rate of water molecules increases. As

the temperature increases, the higher the activation energy of water molecules is, and the faster is water saturation of FFRPs. Ana Espert et al. got the same results by immersion of wood fibers/polypropylene composites in water at three different temperatures [14].


**Table 2.** Maximum water uptake and diffusion coefficient (D) of FFRPs at 70 ◦C under 80% RH.

**Figure 5.** Water absorption process of nano-clay grafted FFRPs at 20 °C, 40 °C and 70 ◦C under 80% RH.

**Table 3.** Maximum water uptake and diffusion coefficient (D) of nano-clay grafted FFRPs at 20, 40 and 70 ◦C under 80% RH.


### *3.2. FTIR Observation*

As shown in Figure 6, FTIR test results display the fundamental OH stretching vibration of untreated fiber and silane treated fiber [20]. In the FTIR results of untreated flax fibers, the peak at 3295 (cm−1) indicated hydroxyl groups adsorbed by hydrogen bonds and the peak at 3359 (cm−1) indicated free hydroxyl or amino groups. Because the untreated flax fibers contain no amino group, this peak completely represents the vibration of hydroxyl group [28,29]. FTIR results of silane coupling agent treated fiber showed that the peak at 3320 (cm−1) represented hydroxyl groups adsorbed by hydrogen bond, and the peak at 3419 (cm<sup>−</sup>1) represents free hydroxyl or amino groups [28,29]. Because silane coupling agent treated flax fibers contain a large number of amino groups, and the absorption peaks become wider, indicating that there are overlapping groups, free hydroxyl and amino groups are represented here. Because the hydroxyl absorption peaks of the two fibers are similar, it can be considered that the hydroxyl content of silane treated flax fibers is lower than that of untreated flax fibers. The saturated water absorption of silane-treated FFRP decreases due to the decrease of hydroxyl content and the decrease of water adsorption sites. This phenomenon also can be explained by the chemical reaction between flax fibers and nano-clay. Based on previous research by the author, the interface properties of flax fiber and resin was improved using this chemical treatment. Reduction of water absorption could be attributed to the complete adhesion and wettability between the flax fibers and the polymer matrix, which may have less gaps and flaws at the interface [30]. The increase of

interfacial bonding reduces the amount of water molecules stored at the interface of FFRP [10]. This chemical reaction was presented in a previous research work by the author [27].

**Figure 6.** FTIR test results of flax fiber (C-fiber: untreated fiber; S-fiber: silane treated fiber; O-fiber: nano-clay grafted fiber) [24].

Figure 7a, b show the mechanism of improving the hydrothermal ageing behaviors of nano-clay grafted FFRP. Water molecules enter FFRPs along the fiber direction and are perpendicular to the fiber direction. As shown in Figure 7a, water molecules exist in untreated FFRPs in two forms: (1) free water stored in the cell compartment of fiber, fiber and interfacial space, and micro cracks; (2) bound water adsorbed on the fiber surface and cell wall. As shown in Figure 7b, after the nano-clay grafted, the interfacial adhesion properties of FFRP are improved. Therefore, water molecules in the interfacial and cracks are reduced. The nano-clay is grafted onto the surface of the fiber, making the surface of the fiber hydrophobic, increasing the diffusion path of water molecules. In Figure 7c, water molecules enter the interface between the fibers and epoxy, and form the first layer of water molecules by hydrogen bonding on the surface of unmodified flax fibers. Then the water molecules enter the first layer of water molecules in the form of the second layer of water molecules. As shown in Figure 7d, chemical grafting occupies the hydroxyl groups on the surface of the fiber, reducing the adsorption of water molecules on the cell wall of the flax fiber. However, with the increase of aging time and temperature, water molecules will gradually destroy the chemical bonds between the silane and flax fibers and adsorb on the surface of the fibers.

**Figure 7.** *Cont.*

**Figure 7.** Diagram of water molecule existence and diffusion path in FFRPs (**a**) untreated FFRPs; (**b**) nano-clay grafted FFRPs; model of water molecule interaction on flax fiber-epoxy interface (**c**) untreated flax fibers (**d**) nano-clay grafted flax fibers.

### *3.3. FBG Monitoring*

Figure 8 shows the radial and latitudinal strain values of FFRP calculated by the above formulas, which vary with the time it is placed in the 70 ◦C under 80% RH environment. Because FBG is sensitive to strain and temperature changes, there are fluctuations in the test results. In order to facilitate the analysis of the results, FBG is used to process the collected data, and the results are simplified to 100 data points without changing the trend. In Figure 8, O/W and C/W represent the weft strain (with less fibers) change process of nano-clay grafted and untreated FFRP; O/R and C/R represents radial strain test results of FBG in nano-clay grafted FFRP and untreated FFRP. The maximum strain in the radial and weft directions of the nano-clay grafted FFRPs are smaller than that of the untreated FFRPs.

**Figure 8.** Effect of water absorption on the internal strain of untreated FFRP and nano-clay grafted FFRPs at 70 ◦C under 80% RH (**a**) the weft strain (**b**) the radial strain.

As the FFRP is placed in a hygrothermal environment for extended periods of time, water molecules gradually enter the fibers, causing them to expand. With the increase of water absorption, the water molecules diffuse into the specimen through the resin, fibers and the interfaces between them. The water ingress plasticizes the resin matrix, and even the fibers, leading to relaxation of the internal strain [31]. Consequently, the internal tension strain gradually reduces. When FFRP reaches saturated water absorption, the FFRP internal strain tends to balance after a certain fluctuation. As shown in Figure 8, the nano-clay-modified FFRP causes a decrease in the swelling amount of the fiber due to a decrease in saturated water absorption. Nano-clay grafted FFRPs show better dimensional stability.

According to the definition of the linear moisture expansion coefficient, when the laminate absorbs moisture, it produces line strain in the main direction of the material, as per the following formula [32,33]:

$$
\beta\_{\mathbb{X}} = \frac{\varepsilon\_{\mathbb{X}}}{\mathbb{C}} \tag{7}
$$

$$
\beta\_{\mathcal{Y}} = \frac{\varepsilon\_{\mathcal{Y}}}{\mathbb{C}} \tag{8}
$$

where β*<sup>x</sup>* is linear moisture expansion coefficient in the x direction; ε*<sup>x</sup>* is strain in the x direction; β*<sup>y</sup>* is linear moisture expansion coefficient in the y direction; ε*<sup>y</sup>* is strain in the y direction; *C* is water absorption concentration.

As shown in Table 4, the radial and latitudinal linear moisture expansion coefficient of untreated FFRPs and nano-clay grafted FFRPs are calculated by the maximum strain and saturated water absorption, respectively. The linear moisture expansion coefficient of nano-clay-grafted FFRP in two main directions is smaller than that of untreated FFRP. This is because the presence of the silane coupling agent film and the nano-clay constrains expansion of the fiber.

**Table 4.** Linear moisture expansion coefficient of FFRPs.


### *3.4. Tensile Properties*

Figure 9a,b show the degradation of tensile properties of FFRPs. The degradation of tensile strength is smaller and the degradation of the tensile modulus is larger. The tensile strength of FFRP is affected by the flax fiber strength and the interfacial bond strength between the fiber and resin. The mechanical properties of flax fibers are influenced by the composition, structure and number of defects in a fiber. Under stress, tensile failure occurs by intercellular and/or intracellular modes [34]. Cellular stress is mainly determined by cellulose content and the angle between cellulose microfibers and the axis. When water molecules enter the fibers, moisture in fiber influences the degree of crystalinity and the crystalline orientation of fibers whereby it results in higher amounts and better orientation of crystalline cellulose in fibers. The absorption of water in the pores and amorphous regions of the fibers serves to reduce interfibrillar cohesion and to relieve internal fiber stresses [35]. Cellulose microfibers are embedded in hemicellulose, wax, etc. Hydrogen bonds play a key role in their combination. Water ingress deteriorates the hydrogen bonds, leading to higher elongation and strength, but lower modulus. Besides, increase in tensile strength of flax fiber is due to the availability of free water molecules, providing a plasticizing effect, which is advantageous to the strength of cellulose fibers [15]. However, the plasticization effect of water weakens the fiber/matrix bonding, resulting in interfacial failure [36]. Therefore, when water molecule acts on the composite, the tensile strength of the composite decreases slightly.

As shown in Figure 9a, after a six-week immersion in 70 ◦C under 80% RH environment, the tensile strength of untreated FFRPs (C) reduced by 13.5%, and that of silane treated FFRPs (S) and nano-clay grafted FFRPs (O) decreased by 15.8% and 15.6% respectively. The tensile strength retention rates of C, S and O were 88.0%, 84.1%, and 84.3%. As mentioned above, the tensile properties of FFRP depend on the tensile strength of the fibers and the strength of fibers depend on the content and angle of cellulose. Thus, surface modification has little effect on the degradation of tensile strength of FFRP in a short time. On the other hand, the interface is the medium of stress transfer between fibers, and also affects the tensile strength of FFRP to some extent. Cellulosic fibers can absorb water from the environment and can swell. This causes shear stress at the interface, which favors ultimate debonding of the fibers, which in turn causes a reduction in tensile strength [36]. The silane coupling agent forms a thin layer of macromolecule at the interface between the fiber and the resin, but the coupling agent is highly sensitive to water molecules [22]. When water molecules enter the early stage of the composite, under the action of high temperature, the Si–O–C bond between fiber and silane is not stable towards hydrolysis [37]. As a result, some of the coupling agent molecules that do not form macromolecules are easily hydrolyzed [38]. After partial hydrolysis of silane coupling agent molecules, the interfacial properties of modified flax fiber composites decreased before those of unmodified composites. Therefore, in the early stage of aging, the tensile strength of the untreated FFRPs decreased less than that of the silane treated FFRPs. For the same reason, the nanoclay grafted FFRP has also been treated with silane coupling agent, so the short-term degradation of tensile properties shows the same regularity as the silane treated FFRP. However, many papers show that the tensile strength of NFRP modified by the silane coupling agent is less than that of the unmodified ones after prolonged soaking time in a hygrothermal environment [22,39,40]. It can be predicted that the tensile properties of the nano-claymodified flax fibers are less degraded under the long-term action of water molecules.

**Figure 9.** Effect of water absorption on the tensile properties (**a**) tensile strength (**b**) tensile modulus for FFRPs at 70 ◦C under 80% RH (C: untreated FFRPs; S: silane treated FFRPs; O: nano-clay grafted FFRPs).

Figure 9b shows the degradation of tensile modulus at 70 ◦C under an 80% RH environment. The tensile modulus degrades more than the strength. The tensile modulus of untreated FFRPs (C) reduced by 69.2%, and that of silane treated FFRPs (S) and nano-clay grafted FFRPs (O) decreased by 67% and 59.4% respectively. The tensile modulus retention rates of C, S and O were 30.7%, 32.9%, and 41.1%. After the water molecules enter flax fibers, because water molecules can exist in the amorphous structure [8,9], it makes the amorphous structure soften, resulting in a decrease in the modulus of the fiber. Another reason for the decrease in the modulus of the wet sample can be explained by the weakening of the cellulose structure of the natural fiber by the water molecules in the cellulose network structure, where water acts as a plasticizer and allows the cellulose molecules to move freely. Therefore, the quality of the cellulose is softened and the dimensions of the fiber can be easily changed by application of force [41]. The decrease of the absorption of water molecules reduces the plasticizing effect of the composite, so the decrease of the modulus of FFRP grafted by nano-clay is reduced.

The stress-strain curve of the FFRP tensile test is shown in Figure 10. In this picture, C, S and O represent various FFRP samples. 0W and 6W represent FFRPs before aging and placed at 70 ◦C under 80% RH environment for 6W, respectively. Curves extended into the nonlinear region in all cases. The stress-strain curve can be divided into three parts: (1) the first linear part, which is the deformation of each cell wall; (2) the second non-linear part, which is the elastic-plastic deformation of the fibers, is the rearrangement of the amorphous part (mainly made of pectin and hemicellulose) in the thickest cell wall (S2); and (3) the final approximately linear part, which is elastic response of cellulose microfibers to applied tensile strain [1]. The elastic linear area, where the damage is irreversible, reduces as a function of the water ageing [42]. The ultimate strain of untreated and nano-clay grafted FFRPs are increased after being placed in hygrothermal environment for 6 weeks. Water molecules can combine with hydroxyl bonds to act as plasticizers, which makes the material more ductile [15]. In addition, the significant increase in failure strain is due to the decomposition of the cellulose structure after the aging process, resulting in increased ductility of the flax fibers [43]. On the other hand, after the water molecules enter the FFRP, they occupy the pores and defects inside, which increases the ultimate strain of FFRP. At the same time, this increase is attributed to the lubrication of the water molecules, which may slide against each other during loading, resulting in more deformation and elongation [44]. As shown in Figure 10, the ultimate tensile strain of the nano-clay grafted FFRP after being placed at 70 ◦C under 80% humidity environment for six weeks was smaller than that of untreated FFRP. The reason is obvious. The saturated water absorption of the nano-clay grafted FFRP is lower than untreated FFRP, and the plasticization of FFRP by water molecules is reduced, so the increase in ultimate strain is reduced.

**Figure 10.** Effect of the water absorption on the stress-strain curves of different FFRPs at 70 ◦C under 80% RH.

Figure 11a,b show the degradation of tensile strength and modulus of nano-clay treated FFRPs in 20, 40 and 70 ◦C under 80% RH. Similar to the former, FFRPs also show the same degradation law at different temperatures, that is, tensile strength degradation is less, while tensile modulus degradation is greater. With the increase of temperature, the degradation of tensile properties of FFRPs increase, which is due to the increase of temperature, accelerating the movement of water molecules, increasing the diffusion rate of water, accelerating the aging of FFRP. Figure 11a shows the change in tensile strength of nano-clay-grafted FFRP over a six-week period in three different temperatures under 80% RH environments. After six-week immersion in 20, 40 and 70 ◦C under 80% RH environment, the tensile strength of nano-clay grafted FFRPs (O) decreased by 3.0%, 10.0%, 15.7%. When the nano-clay grafted FFRP is placed in an environment of 20 ◦C under 80% RH, the tensile strength increases during the first two weeks. Because of the slower diffusion rate of water molecules in a lower temperature, there are a few of water molecules inside the FFRP, and this part of the water molecules enhances the flax fiber without breaking the interface bonding between flax fiber and epoxy. Therefore, the FFRP tensile strength increased during the first two weeks. Subsequently, due to the prolongation of time, the water molecules gradually entered the FFRP, causing the debonding of the interface and damage inside the fiber. Therefore, after four weeks aging, the strength of the FFRP decreased. In addition, due to the difference in the diffusion rate of water molecules and the deterioration of the FFRP at high temperatures, the tensile strength of FFRP does not degenerate after four weeks at 20 and 40 ◦C under 80% RH, and the FFRP at 70 ◦C under 80% RH no longer degenerates in two weeks. It is worth

noting that although the saturated water absorption of FFRP is approximately the same under different temperature environments, the effects on the tensile properties of FFRP are different. The higher the temperature, the more severe the aging of the composite material. This is because the high temperature accelerates the movement of the water molecules and also increases the deterioration of the composite material. Figure 11b shows the same results, which are the higher the temperature, the greater the modulus drop. After six-week immersion in 20, 40 and 70 ◦C under an 80% RH environment, the tensile modulus of nano-clay grafted FFRPs decreased by 36.9%, 47.6%, 59.5%.

**Figure 11.** Effect of the water absorption on the tensile properties: (**a**) tensile strength (**b**) tensile modulus for nano-clay grafted FFRPs at different temperature under 80% RH.

The stress-strain curve of the nano-clay grafted FFRP tensile test after immersion in different environments is shown in Figure 12. Here, 20 ◦C/0W, 40 ◦C/0W and 70 ◦C/0W represent the stress-strain curve of nano-clay grafted FFRP before immersion; 20 ◦C/6W, 40 ◦C/6W and 70 ◦C/6W represent the stress-strain curves of nano-clay grafted FFRP subjected to exposure at 20, 40 and 70 ◦C under 80% RH for 6 weeks. The result also shows the same result. When FFRP is placed in the hygrothermal environment with a higher temperature, the greater the ultimate strain of FFRP.

**Figure 12.** Effect of water absorption on the stress-strain curves of nano-clay grafted FFRPs at different temperatures under 80% RH.

Elongation at break of FFRPs show the same results. Table 5 shows the elongation at break of FFRPs. After immersion in 70 ◦C under an 80% RH environment for six weeks, the elongation at break of FFRPs increases. Among them, the nano-clay grafted FFRP has the smallest change in elongation at break. On the other hand, as the temperature increases, the greater the increase in elongation at break.


**Table 5.** Tensile properties of FFRPs.

### *3.5. SEM Observation*

Figure 13a–d show the SEM test results of tensile fracture of FFRPs after aging. Figure 13a showed tensile fracture of untreated FFRP before aging in which the fibers were pulled out of the resin but the gap between the fibers and the resin was small. The tensile fracture picture (Figure 13b) of untreated FFRP after six weeks of aging showed that the direct gap between fiber and resin was larger, which indicated that the interface between unmodified fiber and resin was destroyed by water molecules. Figure 13c showed tensile fracture of nano-clay grafted FFRPs before aging. The interfacial adhesion between modified fibers and resins increased. Thus, the interfacial debonding of fiber and resin is less, and most of them are broken by fiber fracture. While after six weeks immersion at 70 ◦C under 80% RH, nano-clay grafted fibers (Figure 13d) are pulled out of the resin, but there is still a small amount of resin attached to the surface of the fiber, and the gap between the fiber and the resin is small. For the nano-clay grafted FFRPs, the entry of water molecule has a certain effect on the bonding between fiber and resin, but the damage is weaker than that of the untreated ones. Because plant fibers have a multi-stack structure, overall swelling results from the local swelling of each component and each cell-wall layer. As each component of S2 layer has a different swelling behavior, differential swelling stresses may induce structural damage of fibers and thus degrade the mechanical properties.

**Figure 13.** SEM photos of nano-clay grafted FFRPs tensile fracture: (**a**) untreated FFRP before aging; (**b**) untreated FFRP immersion in 70 ◦C under 80% RH for six weeks; (**c**) nano-clay grafted FFRP before aging (**d**) nano-clay grafted FFRP immersion in 70 ◦C under 80% RH for six weeks.

### **4. Conclusions**

In this article, the effects of grafting of nano-clay on the hydrothermal resistance of the flax fiber reinforced epoxy composite plate were investigated. The moisture uptake, dimension change and tensile properties of the composite plates were tested. The following conclusions can be drawn based on the testing results and analysis:


**Author Contributions:** G.X., A.W., and H.L. conceived and designed the experiments; G.X. and A.W. performed the experiment, analyzed the data and wrote the paper; G.X. and H.L. monitored the experimental process.

**Funding:** This research was funded by Chinese MIIT Special Research Plan on Civil Aircraft with grant No. MJ-2015-H-G-103 and the National Natural Science Foundation of China with Grant No. 51878223.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Flax, Basalt, E-Glass FRP and Their Hybrid FRP Strengthened Wood Beams: An Experimental Study**

### **Bo Wang 1, Erik Valentine Bachtiar 2, Libo Yan 1,2,\*, Bohumil Kasal 1,2 and Vincenzo Fiore <sup>3</sup>**


Received: 27 June 2019; Accepted: 25 July 2019; Published: 29 July 2019

**Abstract:** In this study, the structural behavior of small-scale wood beams externally strengthened with various fiber strengthened polymer (FRP) composites (i.e., flax FRP (FFRP), basalt FRP (BFRP), E-glass FRP ("E" stands for electrical resistance, GFRP) and their hybrid FRP composites (HFRP) with different fiber configurations) were investigated. FRP strengthened wood specimens were tested under bending and the effects of different fiber materials, thicknesses and the layer arrangements of the FRP on the flexural behavior of strengthened wood beams were discussed. The beams strengthened with flax FRP showed a higher flexural loading capacity in comparison to the beams with basalt FRP. Flax FRP provided a comparable enhancement in the maximum load with beams strengthened with glass FRP at the same number of FRP layers. In addition, all the hybrid FRPs (i.e., a combination of flax, basalt and E-glass FRP) in this study exhibited no significant enhancement in load carrying capacity but larger maximum deflection than the single type of FRP composite. It was also found that the failure modes of FRP strengthened beams changed from tensile failure to FRP debonding as their maximum bending load increased.

**Keywords:** flax FRP; basalt FRP; glass FRP; wood beam; bending; hybrid FRP

### **1. Introduction**

With an increasing concern on the energy conservation and environment protection, wood as a natural and sustainable construction material has returned to the spotlight after a long time flagging [1]. Compared with other conventional construction and building materials, wood has several shortcomings, e.g., relatively low tensile stiffness and strength compared to steel and low compression stiffness and strength compared to concrete. Wood is also susceptible to biological degradations, such as from fungi, bacteria and insects [2], which weaken its mechanical properties. To overcome the inferior mechanical properties of wood elements, fiber reinforced polymer (FRP) composite [3–5] can be one of the solutions. FRP has been widely utilized in the past two decades for rehabilitation and reinforcing of existing structures. FRP materials such as glass or carbon FRP have high strength-to-weight ratio, corrosion-resistance and provide design flexibility [6–8].

The commonly utilized FRP composites as reinforcement for wood beams are carbon FRP (CFRP), E-glass FRP (GFRP) and aramid FRP (AFRP) [3–5,9–12]. However, the production processes of these fibers are energy-intensive and the initial costs are still high. Recently, mineral-based natural FRP, such as basalt FRP (BFRP), has been introduced. BFRP has low material cost, high fire resistance, good thermal, electrical and sound insulating properties [13–15]. Furthermore, basalt fiber also has high tensile properties (e.g., tensile strength of 1850–4800 MPa) [14]. However, similar to glass fiber, the production of basalt fiber also requires a large amount of energy because of the high melting point of basalt rocks (1300 ◦C–1700 ◦C) [13].

As an alternative to glass, carbon and basalt fiber materials, the ecological and economical plant-based FRPs (e.g., flax or jute FRP) have been introduced in civil engineering. Various investigations on plant-based fibers (e.g., flax) have shown that as a single fiber, they have comparable specific mechanical properties (e.g., specific tensile strength and stiffness) compared to those of man-made E-glass fiber [6]. However, this is somewhat misleading since the length of natural fibers are limited, while carbon or glass fiber can be manufactured to have an endless length. The natural fibers are used in the forms of yarns, which will generally have lower mechanical properties compared to the ones of individual fibers.

Nevertheless, several investigations using the natural fibers in FRP as a reinforcement in civil engineering application have been carried out. Huang et al. [16] investigated flax FRP (FFRP) strengthened reinforced concrete (RC) beams. Their results revealed that the FFRP increased the ultimate load and maximum strain as well as the ductility of RC beams significantly. It also showed a better interfacial compatibility with the RC beams compared to GFRP and CFRP strengthened RC beams. Yan et al. [17] investigated the flexural properties of plain concrete beams externally strengthened with FFRP. It has been shown that the bending load capacity of plain concrete beams increased by 100%, 230% and 327% and their fracture energy were increase by 3500%, 4200% and 8160% with two-, four- and six-layer FFRP reinforcement [17]. In addition, FFRP has been used as external confining materials of natural aggregate concrete [18], recycled aggregate concrete [19] and fiber reinforced concrete [20,21].

In literature, a large number of studies have investigated FRP as an external reinforcement of wood structures, but only very few have considered plant-based FRPs. For example, Speranzini et al. [22] investigated solid wood beams externally strengthened with carbon, glass, basalt, hemp and flax FRP under a four-point bending test. No significant difference was observed on the loading capacity of the different FRP composites (i.e., the increase of the bending strength were 42.3%, 24.6%, 23.2%, 24.0% and 35.4% for carbon, glass, basalt, hemp and flax FRP, respectively) although there was a large difference in the tensile strength of these FRPs (i.e., 479, 142, 245, 36 and 25 MPa for carbon, glass, basalt, hemp and flax FRPs, respectively). According to the author, flax and hemp fibers may have better adhesion to wood compared to other FRPs. Borri et al. [23] investigated flax and basalt FRP strengthened low-grade (bending strength of 18.4 MPa) and high-grade (bending strength of 41.3 MPa) wood beams. The tensile strengths of FFRP and BFRP in the study was 240 MPa and 1880 MPa, respectively. The results showed an increase of bending strength of 38.6% and 65.8%, and maximum mid-span deflection of 58.2% and 40.2% respectively by two-layer FFRP and BFRP strengthened low-grade wood beams. Moreover, the strength increases were 29.2% and 25.9%, the increases of maximum mid-deflection were 9.1% and 14.5% respectively for two-layer FFRP and BFRP strengthened high-grade wood beams. This study concluded that both BFRP and FFRP provided the beams with higher strength and better ductile behavior. Similar results can be found in another research by Borri et al. [24] for flax and basalt FRP. André et al. [25] applied FFRP and GFRP with similar fabric density (i.e., 230 g/m<sup>2</sup> for flax and 250 g/m2 for glass) perpendicular to grain on wood beams. It is reported that the maximum bending load of the entire specimen strengthened with GFRP (45.1 kN) was 23% higher than that one strengthened with FFRP (36.0 kN).

Realizing the advantages and disadvantages of using different types of fibers in FRP, hybrid FRP (HFRP) was proposed in the literature. Hybrid FRP, which consists of two or more combinations of strengthened fibers or fabrics, was designed to inherit the advantages and minimize the disadvantages of the combined fibers. Kim et al. [26] investigated HFRP made of carbon and glass fabrics to retrofit RC beams. The results showed that the HFRP contributed to higher ultimate bending strength and ductility of the RC beams compared to the single type of CFRP or GFRP. The maximum load in bending of RC strengthened with GFRP–CFRP (G GFRP attached at the tension surface of the RC beam) specimens was 6.6% and 3.9% higher than the one strengthened with two-layer CFRP (CC) and two-layer GFRP (GG), respectively. Moreover, the maximum mid-span deflection was also 27.4% and 18.5% higher than that of CC and GG specimens.

Compared with man-made fiber/fabric materials in conventional FRP composites (e.g., E-glass and carbon), plant-based fiber/fabric has a lower price and positive ecological impact [27], but it has lower mechanical properties as it has been mentioned before. In order to balance the performance and the cost for proper material design, several studies have investigated the hybridization of a plant-based fabric with a man-made one in FRP composite [28,29]. Gupta et al. [29] have summarized the mechanical properties of this hybrid material reinforcing thermoset polymers. It was concluded that the tensile, flexural and impact strengths of hybrid FRP were higher than those of the single type natural fabric FRP. However, the application of the hybrid FRP with natural fabric for reinforcing wood beams have been scarcely investigated before. Throughout the literature, only very few studies have investigated HFRP strengthened wood beams. Yang et al. [30] strengthened wood beams with hybrid carbon and glass FRP. Compared to the wood beams strengthened by GFRP or CFRP alone, the HFRP provided a larger energy dissipation for wood beams.

In this study, the flexural behavior of flax FRP strengthened wood beams were investigated. The results were compared with man-made E-glass and mineral-based basalt FRPs. Additionally, hybrid flax/glass/basalt FRPs were also investigated and compared with single type of FRPs (i.e., FFRP, BFRP and GFRP). Various different FRP materials (i.e., FFRP, GFRP and BFRP), FRP thickness (i.e., one-, two- and three-layer) and the arrangement of FRP in the HFRP were considered as experimental variables. As complementary initial investigations, tensile and bending test of flat coupon single type fiber FRPs were also carried out. Furthermore, since the interfacial bonding of fiber/epoxy and FRP/wood are also critical points for the flexural behavior of beams, the microstructures of these interfaces from the fractured specimens were examined under light and scanning electron microscopes.

### **2. Materials and Methods**

### *2.1. Materials*

Flax, basalt and E-glass were selected to represent the plant-based, mineral-based and conventional man-made fiber/fabric material for FRP composites, respectively. Among plant-based fibers, flax has comparable specific tensile properties with a lower unit price compared to those of E-glass fiber [6]. In addition, flax has a short growing cycle (harvested within 100 days after sowing the seeds). It also has a large annual production, which is required due to its broad applications, e.g., for household textiles, sails or tents, etc. [6]. For mineral-based fibers, basalt is generally used as a replacement of dangerous asbestos fibers and probably the only mineral-based fiber type that is available on the market [27]. Furthermore, basalt fiber also has tensile properties close to those of carbon fibers (e.g., for tensile strength, basalt fiber: 1850–4800 MPa and carbon fiber: 3000–5000 MPa) [14]. E-glass is one of the most widely used fibers as it is cheaper than carbon or aramid fibers and it has relatively high tensile strength (1800–3500 MPa).

In this study, bidirectional woven flax fabric (FlaxPly BL 550 from Lineo, Valliquerville, France, seven single-strand yarn threads per cm in the fabric weft and warp directions) (Figure 1a), unidirectional E-glass fabric (S15EU910, Saertex GmbH & Co. KG, Saerbech, Germany) (Figure 1b) and randomly distributed basalt mat (HG Europe, Milano, Italy) (Figure 1c) were investigated as FRP fabric materials. Based on the supplier data sheets, the areal density of flax, E-glass and basalt fabrics are 550 g m−2, 600 g m−<sup>2</sup> and 220 g m<sup>−</sup>2, respectively. The nominal fiber thicknesses for one layer of flax, basalt and glass fabrics were 1.2 mm, 0.7 mm and 0.9 mm, respectively. However, it has to be mentioned that these nominal fiber thicknesses were only rough approximations as they are highly dependent on the pressure applied during measurement and the weaving structure of the fabrics. The FRPs were manufactured with a two-component epoxy polymer PRIMETM 20LV epoxy resin and Prime 20 Slow hardener by Gurit Company, Zullwil, Switzerland. The tensile strength, tensile modulus and strain at failure of the cured epoxy were 73 MPa, 3.5 GPa and 3.5%, respectively. Although, some other

adhesives (such as phenolic [31,32] or melamine [33] based adhesives), which are commonly used as adhesives for wood or other cellulosic materials, can be used as a matrix. Epoxy resin was selected in this study since it has been proven to have higher mechanical properties and chemical resistance than the other adhesives [6,34]. Epoxy is also the most commonly used polymer in FRP composites [7,8,20]. The structural wood beams, which were strengthened by the FRPs, were manufactured from Douglas Fir (*Pseudotsuga menziesii* Mirb.) with a dimension of 600 mm (length) × 40 mm (width) × 35 mm (height). The length direction of the beam was along the fiber direction of the wood (Figure 1d). The average density of the wood beams was 577 <sup>±</sup> 33 kg·m<sup>−</sup>3.

**Figure 1.** Photos of testing materials: (**a**) flax fabric, (**b**) glass fabric, (**c**) basalt mat and (**d**) wood beam.

### *2.2. Manufacture of FRP and FRP–Wood Specimens*

The FRP manufacture process in this study was conducted through hand wet lay-up process and two kinds of specimens were produced: (1) FRP laminates for tensile and bending test and (2) FRP strengthened wood beams for bending tests. Initially, the epoxy resin and hardener were mixed with a ratio of 1:0.26 by weight for five minutes. The first layer of the fabric was placed on a flat and water-proofed plastic foil surface. It was then saturated with the epoxy mixture by using a brush. To avoid excess epoxy resin on the fabric, the saturation process was conducted slowly and directly stopped as soon as the fabric reached the saturation point. After that, the next layer was laid on the top of the first one and slowly saturated again with the epoxy. This process was repeated until the targeted number of layers was reached. Similar steps were used for the hybrid FRP. The fabrics were laid one by one in the intended order. All the epoxy-impregnated FRP composites were then cured at a room temperature (20 ± 3 ◦C) for seven days before they were cut to laminates for the flat-coupon tensile and flexural tests. No external pressure was applied on the FRP composites during the curing process. For tensile and bending tests, the FRP was cut into the appropriate size after curing. For the production of FRP strengthened wood specimens, the fabrics were cut firstly into strips with the size of 600 mm × 40 mm and the surface of the beams were coated by epoxy. Then, the strips were applied directly on the wood beams. While the basalt mat was arbitrarily applied on the wood beam due to its random orientation, the main fiber direction of the glass fabric and the warp direction of the flax fabric were always applied along the grain of the wood.

### *2.3. Test Matrix*

A total of 39 small-scale wood beam specimens (three wood beams and 36 FRP strengthened wood beams) were tested under a three-point bending test according to DIN 52186 [35]. Table 1 shows the test matrix of the specimens used in this study. In the specimen name for each specimen type, W indicates wood, while F, B and G denote flax, basalt and glass as the type of the fabric for the FRP composites, respectively. The number of the FRP layers are denoted by 1L, 2L and 3L, i.e., one-, twoand three-layer. For hybrid FRP composite strengthened wood, the combination of F, B and G denotes the sequence of the arrangement of the FRP composite, i.e., 3L-GBF indicates the arrangement of the FRP, which is the outer layer (glass), middle layer (basalt), and the inner layer (flax) attached to the wood beams.


**Table 1.** Matrix of the specimens.

1. W for wood; L for layers; B, G and F for basalt, glass and flax, respectively.

The mechanical properties of the different FRP composites were determined before the bending test of FRP-wood beams. Flat coupon tensile and bending tests were carried out for the FRP laminates according to ASTM D 3039 [36] and ASTM D 790 [37], respectively. For both tests, FRP composites with three different fabric materials (i.e., flax, glass and basalt) and three different layers (i.e., one-, two- and three-layer) were tested. For each specimen type, 10 specimens were prepared with the size of 250 mm in length × 25 mm in width and 150 mm in length × 25 mm in width for tensile and bending tests, respectively. The final thicknesses of the FRP laminates were determined by averaging the thickness of the laminates at three different locations. These thicknesses are presented as results in Table 2.


**Table 2.** Testing result of flat coupon tensile test and standard three-point bending test of FRP laminates.

1. L for layer; B, F, G for basalt, flax and glass, respectively; Te and Be for tensile and bending, respectively. 2. Approximated nominal fiber thicknesses. The values depend on the pressure applied during the measurement and the different weaving structures of the fabrics.

### *2.4. Test Instrumentation*

Zwick 1474 Test Machine (from ZwickRoell GmbH & Co. KG, Ulm, Germany) with a load cell capacity of 100 kN was used for flat coupon tensile test (Figure 2a), bending test (Figure 2b) for FRP laminates and three-point bending test for FRP strengthened wood beams (Figure 3). The testing machine was equipped with a standard extensometer (with an initial distance of 140 mm) to record the displacement of the sample during the test. The tensile tests were carried out with a displacement-controlled rate of 2.5 mm/min. The bending tests on FRP laminates were performed with a span distance of 100 mm and based on the standard, the testing rate was calculated as:

$$R = ZL^2 / 6d \tag{1}$$

where,


**Figure 3.** Test setup of bending test for FRP–wood beams (unit: mm).

All tests for FRP laminates were conducted until failure or the maximum strain of 5% was reached.

The span of FRP strengthened wood beams tested under bending loading was 550 mm. The load was applied at the middle of beams with a loading rate of 12 mm/min until failure. The apparent flexural elastic modulus of the FRP–wood beams was calculated through the following equation, which is adapted from DIN 52186 [35]:

$$E = \frac{L^3}{4bd^3} \cdot \frac{\Delta F}{\Delta D} \tag{2}$$

where


Δ*F* difference of force between 20% to 40% of the maximum bending loading, kN

Δ*D* difference of mid-span displacement at the corresponding bending loading, mm

After the mechanical tests, the fracture areas of the FRP–wood beams were observed with a light microscope (ZEISS 47 50 57 from Carl Zeiss Jena GmbH, Jena, Germany) and a scanning electron microscope (SEM, JSM-6700F, JEOL LTD, Tokyo, Japan). The specimens for the SEM were vacuum-coated with gold by evaporation process in BAL-TEC SCD 050 sputter coater.

### *2.5. Data Analysis Method*

During the analysis and the interpretation of the data, the results were only compared based on the average value. The readers must be cautioned that these comparisons were only preliminary in character due to the comparing of the average values. No statistical analysis of the data was possible due to the limited number of specimens. Matching of specimens (for a pairwise comparison) is impossible for wood samples due to the variability within the material itself as well as variability between the specimens.

### **3. Results and Discussion**

### *3.1. Tensile and Bending Tests for FRP Laminates*

The results of the tested FRP laminates under tensile and bending loadings are presented in Table 2. For each specimen type, eight to ten specimens were successfully tested, except for 3L\_F\_Te and 2L\_G\_Te, where six and five specimens were successfully tested, respectively. The averaged value and the standard deviation of these successfully tested specimens are presented in the table. Furthermore, Figures S1 and S2 show the tensile and flexural stress–strain curves of the specimens during the tests, respectively. In these table and figures, indices Te and Be refer to tensile and bending tests, respectively.

Under bending loading, the maximum strengths of BFRP (79.6–156.8 MPa) were in general higher than FFRP (60.3–94.6 MPa) at any number of investigated fabric layers. Under tensile loading, however, FFRP (41.7–76.8MPa) had a comparatively similar strength than those of BFRP (49.6–61.1 MPa). Based on previous studies, the tensile strength of BFRP can be reached at around 1000 MPa (e.g., 707 MPa by Reyes-Araiza et al. [38] and 1282 MPa by Quagliarini et al. [39]). The low strength of BFRP obtained in this study was suspected due to the thin nominal fabric thickness, which led to a low areal density, and the random distribution of the basalt fibers in the mat. When compared with GFRP, FFRP presented significantly lower tensile and bending properties, and lower strain at peak load. This was expected since flax yarn consists of multiple bundles of short fibers, while glass yarn may have continuous fibers. Flax fibers may also contain natural defects [6], which cannot be avoided. Similar results were reported by Zhang et al. [40]. Their results showed that 10-layer FFRP had tensile

strength of about 220 MPa and tensile failure strain of 0.85%, which was much lower than 10L-GFRP with tensile strength of about 700 MPa and tensile failure strain of 1.41%.

The number of fabric layers also influenced the mechanical properties of the overall FFRP. A relatively similar tensile strength was observed for one-layer and two-layer FFRP (41.7 and 48.2 MPa, respectively). However, the three-layer FFRP provided distinctly higher tensile strength (76.8 MPa). Under bending, on the other hand, 1L-FFRP (60.3 MPa) had a lower strength compared to the 2L- and 3L-FFRP (94.6 and 90.3 MPa, respectively). The strains at the peak load of the FFRP specimens also followed the same pattern. Under tensile loading, 3L-FFRP was observed to have a higher maximum strain (1.69%) compared to 1L- and 2L-FFRPs (1.29% and 1.30%, respectively). In contrast, the 1L-FFRP specimens had the highest strain at failure under bending load (2.26% compared to 3.37% and 3.23% for 2L- and 3L-FFRP, respectively). Besides the number of layers and the type of loading (tension or bending loadings), the inconsistency of the produced fiber volume fraction of the FRP using hand lay-up method may have contributed to the current finding. Moreover, under bending loading the thickness of the specimen strongly influenced the results. When it was bent in a same span length, a thicker specimen produced more internal shear, thus, it was stiffer and failed faster compared to a thinner specimen.

### *3.2. Bending Tests for FRP Strengthened Wood Beams*

### 3.2.1. Effect of FRP Thicknesses on the Bending Behavior of FRP Strengthened Wood Beams

Figure S3 shows the representative load–displacement curves of wood beams strengthened with a different number of layers of B-, F- and GFRP. The results together with the calculated improvement of the properties due to the FRP reinforcements (unstrengthened wood beams as the reference) are also presented in Table 3.

The load capacity improvement increased with an increasing number of FRP layers for all wood beams strengthened with a single type of FRP. FFRP strengthened wood beams had maximum bending load capacities of 4.5, 5.5 and 6.2 kN for one, two and three layers, respectively. These corresponded to 60.7%, 96.4% and 121.4% load capacity improvement compared to unstrengthened wood beams, which had an average maximum load capacity of 2.8 kN. However, the improvement of the load bearing capacity was not linearly proportional to the increasing number of FRP layers. Similar to the FFRP, the load capacity improvements of one, two and three layers of GFRP were 71.4%, 117.9% and 132.1%. As the number of FRP layers increased, the gradient of the capacity improvement declined. However, surprisingly it was observed that the gradient of the capacity improvement of BFRP increased, i.e., 14.3%, 50.0% and 107.1% for one, two and three layers of BFRP. The reason could be due to the change of the failure mode, which is often governed by the weakest components in the FRP–wood composite beams. Under bending, the load is transferred to the compression and tension loadings. The compression loading was on the top part of the specimen, which was carried by the wood, while tensile loading was on the bottom part carried by the FRP. The tensile rupture of FRP may have initiated the overall failure of the composite if the FRP laminates were too thin (e.g., the 1L-FRP) or did not have enough strength (e.g., flax and basalt). With the increasing number of layers in the FRP, the tensile capacity of the FRP increased, which may have led to a shifting of the failure mode. The FRP–wood composite may then have failed due to the yielding failure of the wood in the compression zone or the delamination of the FRP–wood interface due to the induced internal shear loading. Further discussions of the different failure modes are given in Section 3.3.



**Table** 

Table 3 also presents the elastic modulus of the investigated FRP–wood composites. The elastic modulus may have increase up to 66% as the woods were strengthened with the FRP. The influence of FFRP and BFRP thickness on the elastic modulus of the overall beam was less pronounced compared to the one from GFRP. By using one-, two- and three-layer GFRP under the wood beams, the elastic modulus was increased from 9.0 GPa to 9.2, 13.3, and 15.1 GPa, respectively. Among all the specimens, wood beams strengthened with three layers of GFRP had the highest elastic modulus. Compared with BFRP (10.0–11.0 GPa), FFRP strengthened wood beams had a higher elastic modulus (12.6–12.9 GPa). However, these results do not fully follow the results of the tensile tests and bending tests of FRP laminates showed in Section 3.1. FFRP laminates had the lowest tensile and bending modulus (i.e., 4.8–5.6 GPa in tensile and 3.7–5.1 GPa in bending) compared to BFRP (6.0–6.2 GPa in tensile and 5.8–6.3 GPa in bending) and GFRP (19.3–23.3 GPa in tensile and 8.0–18.1 GPa in bending). The reason of these findings was suspected to be due to the different thickness of FRP beams and the compatibility between the fabric and wood.

Based on the cross-section inertia of the beams and also presented in Equation (2), the height of the beam to the power of three highly influences the elastic modulus. The actual thickness of each specimen was considered in the calculation. However, the different thickness of the FRP led to the different height of the FRP–wood specimens. Thus, the cross-sectional FRP–wood ratios were varied between specimens. This may have led to a different stress distribution during bending loading. Higher thickness of the FRP–wood may also have resulted in stiffer beams due to the more pronounced influence from the internal shear loading of the specimen under bending loading.

In addition to that, as a cellulosic natural material, flax has the same chemical components as wood (i.e., cellulose, hemicellulose and lignin). Therefore, similar bonding behavior is expected between flax/epoxy and epoxy/wood. On the other hand, the bonding behavior of glass/epoxy and basalt/epoxy are different. The similar bonding behavior was suspected to give a positive impact of the overall mechanical properties of the FRP–wood beams. This was also supported by the results from HFRP, the highest stiffness was reached when flax connected directly to the wood (14.6 GPa for W\_3L-BGF). This reason, however, is only a theory based on the results obtained in this study. Further investigations have to be conducted to support this theory.

### 3.2.2. Effect of FRP Materials on the Bending Behavior of FRP Strengthened Wood Beams

Figure 4 shows the representative bending load–displacement curves of the unstrengthened wood beam and all types of three-layer FRP strengthened wood beams. Their maximum load, maximum deflection and flexural elastic modulus are presented in Table 3. All the three-layer FRP reinforcements increased the maximum load of wood beams remarkably. The average load capacity of W\_3L-F, W\_3L-B and W\_3L-G were 6.2, 5.8 and 6.5 kN with increments of 121.4%, 107.1% and 132.1%, respectively in comparison to the average load capacity of unstrengthened wood beams. The hybrid FRPs showed similar enhancement in load capacity. The maximum bending load of W\_3L-BFG, W\_3L-BGF and W\_3L-GBF were 5.6, 5.8 and 5.9 kN, respectively. Among these tested FRPs, the best performance based on the maximum mid-span deflection was observed from HFRP strengthened wood beams. The W\_3L-BGF had the highest maximum strain increment by 142.5%, followed by W\_3L-GBF (134.6%) and W\_3L-BFG (121.3%), which were higher than that of FFRP (66.9%), BFRP (63.0%) and GFRP (111.0%).

When comparing the FFRP to BFRP and GFRP, it was found that FFRP laminates had higher ultimate strain than BFRP under tensile loading. Therefore, FFRP provided a larger enhancement in deflection than BFRP for FRP strengthened wood beams. FFRP had only a slightly lower tensile strength to BFRP (41.7 and 49.6 MPa for one-layer FFRP and BFRP, respectively), which was already enough to carry the tensile loads on the tensile area at the bottom of the wood beams. Moreover, FFRP laminates were also thicker than BFRP. As a result, FFRP provided larger enhancement than BFRP in FRP strengthened wood beams. It should be, however, kept in mind that the basalt fabric mat used in this study had a low areal density with short fibers that were orientated randomly. Furthermore, the bending results of FRP strengthened wood beams also showed that FFRP provided similar maximum strength and maximum deflection enhancement with GFRP for wood beams (especially with higher number of FRP layers), although FFRP laminate had much lower tensile strength than GFRP laminates (i.e., 76.8 MPa for three-layer FFRP vs. 449.1 MPa for three-layer GFRP). This was primarily because, at a high number of FRP layers, the failure of the interface between wood and epoxy would have been more decisive on initiating the whole failure of the FRP–wood beams. Thus, having a stronger FRP material such as glass, may not necessarily increase the overall performance of FRP–wood composite. The interface debonding will always initiate failure of the whole composite systems and the maximum capacity of GFRP cannot be fully utilized.

**Figure 4.** Load mid-span displacement curves of three-layer FRP strengthened wood beams.

### *3.3. Failure Modes and Microstructure of FRP–Wood Beam System*

### 3.3.1. Failure Modes

The typical failure modes of the reference wood beams and FRP strengthened beams are shown in Figure 5. The reference beams showed a typical tension failure (Figure 5a). The crack was initiated at the mid-span of the tensile zone and then propagated until the complete failure of the beam. For FRP strengthened wood beams, two kinds of failure were observed, i.e., tensile failure and debonding of FRP. The tensile failure in FRP–wood beams (Figure 5b) was initiated at the middle of FRP strips followed by the failure of the tensile zone of wood beams. The debonding of FRP took place at the interface between wood beams and FRP and occurred in either mid-span of the beam (Figure 5c) or at the edge (Figure 5d).

**Figure 5.** Typical failure modes of FRP–wood beams under three-point bending with different FRP laminates: (**a**) tensile failure for reference wood beam, (**b**) tensile failure (W\_1L-F), (**c**) debonding at mid-span (W\_3L-F) and (**d**) debonding at edge for FRP strengthened wood beam (W\_1L-G).

Table 4 presents the general failure modes for all specimens tested in this study. It can be observed that most FRP strengthened wood beams with low maximum bending loading (e.g., less than 5.8 kN) showed a primarily tensile failure mode. At a higher bending loading, interface debonding was observed. The reason has already been discussed previously that weakest parts (between wood, FRP and the FRP–wood interface) of the composite will decide the failure mode. Thus, by changing of the number of layers the failure mode may be also changed. Exception can be found in W\_1L-G with maximum bending load of 4.8 kN, where debonding failure was observed and W\_3L-B with maximum bending load of 5.8 kN, where tensile failure was observed. W\_3L-GBF had also a slightly higher maximum bending load of 5.9 kN, but tensile failure was observed This may have been due to the relatively low manufacturing quality and repeatability through the hand wet lay-up process. Further investigations on the relation between the failure mode and the tensile strength of FRP strengthened beams should be carried out in the future


**Table 4.** General failure modes for control and FRP strengthened wood beams.

### 3.3.2. Microstructure

Figure 6 shows the light microstructures of FFRP, GFRP, and BFRP as well as a hybrid FRP strengthened wood beams (W\_3L-GBF). The interface between epoxy/wood, flax yarn, glass and basalt fiber structure can be clearly observed under the light microscope. As can be seen, no gaps were found in the interfaces between epoxy and wood. Such interface eased the transfer of the bending load from wood to the FRP fabric. However, several air bubbles were observed in the FRP. These air bubbles might be regarded as defects which may result in stress concentration at the FRP/wood interface. This should be further identified in a future study. The presence of the air bubbles may explain why W\_1L-G had a lower bending load of 5.4 kN with debonding failure compared to W\_3L-B and W\_3L-GBF.

The scanning electron microscope analysis (SEM) was used for the observation of the interface of fabric/epoxy (or fiber/epoxy). Figure 7 shows the example of fracture surface from basalt FRP after tensile failure in the mid-span. In Figure 7a, no obvious gap between the fiber and the matrix was observed, which indicated a good interfacial bond between the fiber and the matrix. The close-up image of the fiber/epoxy interface in Figure 7b shows that only a small amount of epoxy remained on the basalt fiber after the tensile failure of BFRP strengthened wood beams. This indicates that the fiber was pulled out from epoxy matrix during the test. The reasons can be the smooth surface of the basalt fiber or the low wetting behavior between epoxy and basalt fiber. Similar pull-out failure can be also found in FFRP and GFRP. Therefore, methods to increase the surface roughness of fiber (e.g., with alkali solution for flax [7]) or to improve the wetting behavior between fiber and polymer are possibilities that could improve the interface bond between fibers and polymer in FRP composites.

**Figure 6.** Light microstructure of (**a**) flax (F)FRP–wood, (**b**) E-glass (G)FRP–wood, (**c**) basalt (B)FRP– wood and (**d**) hybrid FRP–wood.

**Figure 7.** The fracture surface from SEM of (**a**) the interface between epoxy and basalt fiber, (**b**) a close-up of the interface.

### **4. Conclusions**

This study presented the structural behavior of wood beams externally strengthened with various FRP composites. The effects of fabric materials, FRP thicknesses and the sequence of arrangement of the FRP laminas on the flexural behavior of FRP strengthened wood beams were investigated through three-point bending tests. It was shown that the load bearing capacity of the beam under bending was increased as the number of FRP layers increased. The beam strengthened with HFRP had an average higher maximum deflection before failure, yet relatively similar maximum bending loading and elastic modulus compared to the ones strengthened with single type FRPs. It was also observed that the failure modes of FRP strengthened wood beams changed from tensile failure to FRP debonding as the number of layers and maximum bending load increased. This was an indication that the interface between epoxy and wood became more decisive as the FRP became stronger. Under the light microscope, air bubbles were observed in the FRP, which may create inhomogeneity and stress concentration in the cross section of the FRP and could have led to the premature failure of the FRP and the whole beam structure. Under scanning electron microscope, fiber pull-out failure was observed at fracture area of the FRP. The failure was suspected mainly due to the combined smooth surface and the low wetting behavior of the fiber. Improvement can be made by increasing the surface roughness and by improving the wetting properties of the fiber.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4360/11/8/1255/s1.

**Author Contributions:** L.Y. and B.K. managed the project and provide critical comments on the experiments and paper writing. B.W. and E.V.B. drafted manuscript. L.Y., B.W. and V.F. designed the experimental works. All the authors discussed the results and edited the manuscript.

**Funding:** The research was financially supported by Fachagentur Nachwachsende Rohstoffe e. V. (FNR, Agency for Renewable Resources) founded by Bundesministerium für Ernährung und Landwirtschaft (BMEL, The Federal Ministry of Food and Agriculture of Germany), under the Grant Award: 22011617 and by Bundesministerium für Bildung und Forschung (BMBF, Federal Ministry of Education and Research of Germany) (Grant No.: 01DS18023).

**Acknowledgments:** The authors acknowledge support by the German Research Foundation and the Open Access Publication Funds of the Technische Universität Braunschweig. The first author also acknowledges the PhD scholarship awarded by the Chinese Scholarship Council (CSC).

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Polymers* Editorial Office E-mail: polymers@mdpi.com www.mdpi.com/journal/polymers

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18