Experimental investigation on the mechanical and interfacial properties of fiber-reinforced geopolymer layer on the tension zone of normal concrete

This study aimed to understand the mechanical and interfacial properties of fiber-reinforced geopolymer (FRG) layer on the tension zone of normal concrete. A trial mix of FRG using local materials obtained in Hokkaido was investigated. Compression and splitting tensile tests on the FRG test pieces were conducted under different mixing procedures, constituents, volume fractions of the polyvinyl alcohol fiber, curing conditions, and ages. Flexural tests using FRG-normal strength concrete (NSC) composite specimens with different FRG layer thick- nesses were carried out. By placing the FRG layer on the tension side of the composite specimen, the flexural strength and energy absorption were significantly increased. The flexural strength for the NSC alone was 3.5 MPa, while the FRG-NSC composite specimens showed higher flexural strengths of 6.1 – 6.6 MPa. Also, XRD, FTIR, and SEM analyses were carried out for the FRG samples.


Introduction
Geopolymers [1][2][3], also known as alkali-activated binders, have attracted attention as a substitute construction material for ordinary Portland cement (OPC) concrete [4,5]. Basically, geopolymers are synthesized from aluminosilicate-based raw materials, alkaline activators, water, and other ingredients [6]. Geopolymers are cement-free materials and can be manufactured from industrial waste or by-products, including fly ash and blast furnace slag [7,8]. Lower CO 2 emissions were estimated for geopolymer concrete than for OPC concrete [9], and geopolymer concrete has the advantages of energy savings and environmental protection [10]. Geopolymer technology is expected to contribute to development and application of environmentally friendly construction materials. Due to these advantages, recently, geopolymers have been used as a substitute material for OPC concrete [11].
As with OPC, geopolymers have brittle characteristics with low tensile strength and suffer from cracks under mechanical loadings [12]. Use of fibers can improve the flexural strength and absorbed energy of geopolymers [13]. The applicable fibers for fiber-reinforced geopolymer (FRG) include steel fibers [14,15], natural fibers [16,17], and polyvinyl alcohol (PVA) fibers [18,19]. PVA fibers have high tensile strength and bonding characteristics, as well as non-toxicity and acid and alkaline resistance [20]. Ohno and Li [21] studied strain hardening behavior of fly ash (FA) based FRG with PVA fiber by cubic compressive and dogbone tensile tests. Nematollahi et al. [22] investigated the compressive and fracture properties of FRG with PVA fiber by mechanical testing. Masi et al. [23] studied the mechanical and microstructural properties of FRG with PVA fiber. Tanyildizi and Yonar [24] experimentally investigated the high-temperature resistance of geopolymers with PVA fiber. Also, Celik et al. [25] examined high-temperature and mechanical behaviors of FRG with PVA fiber. Zhang et al. [26] tested compressive and flexural strength, elastic modulus, and fracture properties of FA and metakaolin (MK) based FRG with PVA fiber. Wang et al. [27] experimentally evaluated the effects of PVA fiber on compressive strength and durability of MK and FA based FRG. Zhang et al. [28] conducted microand macroscopic analyses on MK and FA based FRG with PVA fiber. Zhang et al. [29] evaluated the effect of PVA fiber and exposure temperature on the behavior of FRG with PVA fiber after exposure to high temperatures. These studies [21][22][23][24][25][26][27][28][29] demonstrate that the addition of PVA fibers is a promising approach for enhancing the mechanical properties of geopolymers. However, the mechanical characteristics of PVA-FRG members have not been fully clarified.
To date, experimental and analytical research has been conducted to understand and evaluate the structural behavior of composite members consisting of normal strength concrete (NSC) substrate and strengthening elements made of fiber-reinforced concrete such as ultra-highperformance concrete (UHPC). The flexural response [30], flexureshear resistance [31], and punching shear resistance [32] were investigated using analytical cross-sectional models for UHPC-reinforced concrete (RC) composite members. Safdar et al. [33] experimentally investigated the flexural behavior of RC beams repaired with UHPC. Loading tests [34], finite element modeling simulation [35], and flexural capacity analysis [36] were conducted for RC members strengthened with UHPC layer. These studies [33,34] show that the placement of the UHPC layer is effective for strengthening RC members. However, research on the mechanical performance of concrete members strengthened with FRG layer has been limited. Menna et al. [37] experimentally investigated the effectiveness of FRG-based systems for strengthening RC beams. Gao et al. [38] studied the interfacial bond properties of PVA-FRG and concrete substrate by a double-interfaced shear testing. The nonlinear mechanical response, including the flexural strength and energy absorption behavior, of FRG-NSC composite structural members is still unclear.
The present study aimed to understand the mechanical and interfacial properties of FRG layer on the tension zone of normal concrete. This study applied PVA fiber for FRG in consideration with the abovementioned advantages of PVA fibers. A series of trial mixes, compression tests, and splitting tensile tests of FRG, named Series A, was investigated under different mixing procedures, mix constituents, fiber volume fractions, curing conditions, and ages. Flexural tests, named Series B, were conducted using FRG and FRG-NSC composite flexural specimens with different FRG layer thicknesses. This study contributes to advancement of strengthening methods for existing NSC structural members by an FRG layer based on geopolymer technology. In addition, the present paper adds experimental results to significantly augment previous reports [39,40].

Trial mixes and methods for Series A
Raw materials and mix constituents of the geopolymer used in the trial mix was determined based on the literature [18,41]. The raw materials included fly ash (FA), blast furnace slag fine powder (BS), sodium hydroxide (NaOH), sodium silicate or water glass (WG), fine aggregate (S), silica fume (SF), sodium gluconate (GNa), water (W), and polyvinyl alcohol (PVA) fiber. For the FA, fly ash for concrete (JIS A 6201) class II was used, whose main component was SiO 2 (45.0 % or more) and whose specific surface area was 2500 cm 2 /g or more. The BS was blast furnace slag fine powder for concrete (JIS A 6206), whose specific surface area was 4050 cm 2 /g and main components were SiO 2 , Al 2 O 3 , and CaO. The NaOH had a purity of 97.0 % or more and included Na 2 CO 3 of 1.0 % or less. For the SF, EFACO silica including SiO 2 of 95.2 % (representative value) was used. The WG was grade No.1, whose main components were SiO 2 (36-38 %) and Na 2 O (17-18 %). For the GNa (C 6 H 11 O 7 Na), the purity was 98.0 % or more. For W, either well water (potable) or industrial pure water was used. The FA, BS, S, and well water were local raw materials produced in Hokkaido. Tables 1 and 2 show the mix proportion of the FRG used in Series A and the specifications of the PVA fibers (Kuraray) [42] mixed into the FRG, respectively. Four PVA fiber volume fractions, 0, 1.5, 3.0, and 4.5 vol% were investigated. Cylindrical test pieces (Ø50 mm × 100 mm) were manufactured in six batches, cases A1 through A6. The compression tests were conducted on test pieces from cases A1-A6, while the splitting tensile tests were performed on test pieces from cases A4-A6. Hereafter, the mix proportions of the FRG for cases A1-A2, case A3, and cases A4-A6 are referred to mixes 1, 2, and 3, respectively.
The three FRG mixing procedures shown in Fig. 1 were adopted and studied. The first was based on a mixing method shown in the literature [18], which is referred to as mixing procedure I. The second and third were also based on a mixing method shown in the literature [41] and were modified from the procedure I to pre-dissolve the SF in the NaOH and diluted WG solution and then add the alkaline silicate solution to the powders (the FA, BS, and S). The method where WG was used is referred to mixing procedure II, and the method where WG was not used is referred to as mixing procedure III. The mixed geopolymer was cast in molds and then sealed with a wrap. After sealing, some of the test pieces were heat cured (at 50 • C or 60 • C for 6 h or 24 h) using an incubator, followed by room temperature curing in a container. The other test pieces were cured at room temperature in a container after sealing. The mold was removed approximately 2-3 days after casting for each test piece. The daily average room temperature during the room temperature curing fell within approximately 10-25 • C in Series A.

Methods of compression and splitting tensile tests for Series A
Compression testing was performed on test pieces aged 7, 28, and 84 days for cases A1-A6 of Series A. The compression tests were conducted in reference to the procedure in JIS A 1108. An unbonded rubber cap was used on the upper surface of each test piece. The mean value of the compressive strength results for mainly-three test pieces under the same conditions was adopted as the compressive strength.
Splitting tensile strength tests were performed on the test pieces for cases A4-A6 after aging for 7, 28, and 84 days for Series A. The splitting tensile tests were carried out in reference to the procedure in JIS A 1113. The mean of the test results of mainly-three test pieces made under the same conditions was adopted as the splitting tensile strength. The tensile FA: fly ash, BS: blast furnace slag, NaOH: sodium hydroxide, WG: sodium silicate, S: fine aggregate, SF: silica fume, W: water, GNa: sodium gluconate, PVA: polyvinyl alcohol. *1: Well water for cases A1-A4 and industrial pure water for cases A5-A6. *2: 0, 19.5, 39.0, and 58.5 kg/m 3 for cases with volume fraction of 0, 1.5, 3.0, and 4.5 vol%, respectively.
where f t is the tensile strength, P max is the peak load, d is the cylinder diameter, and L is the cylinder length.

Fiber-reinforced geopolymer (Series B)
Three cases (B1-B3) of flexural tests were studied for Series B. In case B1, to investigate the effects of PVA fiber content on the flexural performance, three flexural specimens with different fiber content ratios (0, 1.5, and 3.0 vol%) were tested. In case B2, three flexural test specimens with the same conditions were made to evaluate the variation in the flexural strength of the FRG. In case B3, six different flexural specimens with various configurations of FRG and NSC layer thicknesses were made for examining the bending behavior of the FRG specimens and the FRG-NSC composite specimens. For the FRG-NSC composite specimens, the FRG layer was placed on the tension side. Thus, a total of 12 specimens were prepared and used in the flexural tests. Table 3 gives the specifications of the flexural specimens for Series B. Three test pieces were prepared and used for each of the compression and splitting tensile tests for each of the representative flexural specimens (i.e., cases B1-1, B1-2, B1-3, B2-1, B3-1, and B3-6).
For the FRG in Series B, a mix proportion (mix 3) similar to that used in cases A5 and A6 in Series A was adopted, as shown in Table 4. The same raw materials (FA, BS, NaOH, S, SF, GNa, W, and PVA fiber) as those in Series A were used for the FRG, and the FA, BS, S, and W (well water) were local materials produced in Hokkaido. Three volume fractions of the fiber (0, 1.5, and 3.0 vol%) were studied. The FRG was mixed by mixing procedure III in Series A [ Fig. 1(c)]. After the mixing of the FRG was completed, it was cast in cylinder molds (Ø50 mm × 100 mm) and formworks of the specimens for the flexural tests. The geometry of the specimens was 100 mm deep × 100 mm wide × 400 mm long for all the specimens, except for specimens B3-4 and B3-5, which had the same length but were 50 mm deep × 100 mm wide. For the NSC, a premixed concrete with a design strength of 18 MPa was used. No reinforcing steel rebar was emplaced and no fibers were added to the NSC. For the FRG-NSC composite specimens, the NSC portion was first cast and cured for 5 days, after which the surface of the interface was lightly scraped off and the FRG layer was cast. After the first and second casting, each of the test pieces and flexural specimens was placed in a container and cured at room temperature without heat curing. Demolding was carried out 1-4 days after casting.

Methods of compression, splitting tensile, and flexural tests (Series B)
The compression tests in Series B were performed using the same methods as for Series A. In addition, Young's modulus was obtained using an attached strain gage. The age of each test piece was 38 days. The mean of the results for the three test pieces in each case was used for NSC: normal strength concrete, FRG: fiber-reinforced geopolymer. Young's modulus and compressive strength.
The splitting tensile tests for Series B were performed using the same method as that for Series A. The age of each test piece was 38 days. For the tensile strength, the average for three test pieces in each case was adopted.
Flexural strength tests, where the specimen is subjected to four-point loading, were performed. The loading in the flexural testing was conducted in accordance with JIS A 1106. The age of FRG for each specimen was 38 days. Fig. 2 is a photograph of the front of a FRG-NSC composite specimen (B3-3) for flexural testing. The flexural strength was calculated as follows: where F b is the flexural strength, P max is the peak load, l is the span (=300 mm), b is the width, and h is the height. Moreover, based on the flexural test results, hysteretic energy in the loading direction was calculated as the area of each load-deformation relationship (PD curve) from the origin to (1/2)P max point after the peak load, where P max is the peak load. Here, the deformation was the vertical displacement at the loading point. The deformation of the PD curves was adjusted such that the zero-load point based on the secant stiffness between the (1/2)P max and (1/3)P max points was the origin point. Also, in consideration of the deflection of jigs used in the flexural tests, the deformation in the PD curves was modified by assuming a virtual series spring between the specimen and the jigs such that the effect of jig deflection is removed.

Methods of XRD, FTIR, and SEM analyses (Series B)
X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FTIR) analyses were conducted for FRG samples. Also, a microstructure observation was carried out using a scanning electron microscope (SEM) with an energy dispersive X-ray spectroscopy (EDS) function. The FRG samples for the XRD, FTIR, and SEM analyses were taken from the B3-3 specimen after the flexural test. The XRD, FTIR, and SEM analyses were performed for the samples 4-5 months after the flexural test.
The XRD analysis investigated a powdered FRG sample using an Xray diffractometer (SmartLab, Rigaku).
In the FTIR analysis, the FRG sample was powdered and investigated by KBr plate method in a vacuum condition using a Fourier transform infrared absorption spectrophotometer (FT/IR-660Plus, JASCO).
A SEM (JSM-6510LA, JEOL) was used in the SEM imaging and EDS analysis. Two samples, S1 and S2, were used in the analysis. Both samples were taken from the specimen B3-3 after the flexural test such that they included the NSC-FRG interface portion, as shown in Fig. 3. The sample S2 was taken from the vicinity of the failure zone (i.e., main flexural crack) in the flexural test, while the sample S1 was taken from a portion at the end of the specimen. Fig. 3. Location of the samples S1 and S2 taken from the specimen B3-3 after the flexural test for SEM images and EDS analysis.

Compression test results for Series A
The results of the compression testing for cases A1-A6 are listed in Table 5. The obtained compressive strength with respect to the curing conditions is plotted in Fig. 4. The test pieces cured under higher heat and longer duration (24 h at 50 • C) showed a higher compressive strength than those cured with the same heat but a shorter duration (6 h at 50 • C). However, the rate of increasing compressive strength due to the longer heated curing duration was reduced for the cases without the addition of GNa. Overall, when GNa was added, the compressive strength increased. It is inferred that the addition of GNa promoted the fluidity of the FRG, thus the workability and uniformity were improved, resulting in an increase in the compressive strength. Fig. 5 depicts the relationship between the compressive strength and the fiber addition ratio. The compressive strength tended to be higher in the cases without PVA fibers (0 vol%) than in the cases with fibers (1.5, 3.0, and 4.5 vol%). For cases cured at room temperature (28 days), the compressive strength was roughly 50 MPa without fibers and approximately 40-45 MPa with fibers (1.5-4.5 vol%). These results showed that the addition of fibers contributed to no increase in the compressive strength and tended to slightly or somewhat decrease the compressive strength. For each fiber volume fraction, the case with curing at 50 • C (24 h) followed by room temperature curing (age 7 days) showed a higher compressive strength than the case cured under room temperature only (age 28 days) without heat curing. Also, when comparing the same age of 7 days (Table 5), the compressive strengths for 0 and 1.5 vol% (A5-1 and A6-1) with curing at 50 • C (24 h) followed by room temperature curing became approximately 1.9-2.1 times those for 0 and 1.5 vol% (A5-2 and A6-2) cured under room temperature only. These results showed that heat curing at 50 • C is effective for developing compressive strength. The relationship of the compressive strength versus aging at room temperature curing is shown in Fig. 6. The compressive strength increased by approximately 1.7-2.0 times after aging for 28 days and by approximately 2.0-2.3 times after aging for 84 days, compared to the strength after aging for 7 days, for both volume fractions of the fiber (0 and 1.5 vol%). Fig. 7 plots the compressive strength with respect to the mix constituent, showing a higher compressive strength for mix 3 than that for mix 2. Table 6 lists the splitting tensile test results for cases A4-A6. Fig. 8 plots the obtained splitting tensile strength with respect to the age under room temperature curing. The tensile strength tended to increase as the age increased from 7 to 28 and 84 days. Because the difference in the    tensile strength was small between cases A4 and A6 with 1.5 vol% fiber, it was found that the effect on the tensile strength of the difference between the water used (well water or industrial pure water) was not significant. Fig. 9 depicts the relationships of the splitting tensile strength and the content ratio of the fiber. The tensile strength increased as the volume fraction of the PVA fiber increased. Under room temperature curing (age 28 days), the tensile strength was 4.66 MPa for the case with 1.5 vol% fiber (i.e., 196 % compared to that without fiber of 2.38 MPa) and 5.70 MPa for the case with 3.0 vol% fiber (i.e., 239 % compared to that without fiber). This is attributed to the action of the PVA fibers transmitting the stress in the FRG. Unlike the compressive strength (Fig. 5), with increasing the fiber amount from 0 to 4.5 vol%, the splitting tensile strength significantly improved (Fig. 9). Similar to the results for compressive strength, the case of the curing at 50 • C (24 h) followed by room temperature curing (age 7 days) showed the highest tensile strength, indicating that heat curing is effective for improving the tensile strength. Table 7 lists the compression test results for Series B. Here, the curing temperature is the average room temperature during curing at room temperature. Fig. 10(a) plots a comparison of the compressive strength with the fiber addition ratio. When the fiber addition ratio in the FRG became larger, the compressive strength tended to decrease, but the degree of decrease was slight compared to that without fiber. Fig. 10(b) compares the compressive strength between the NSC and FRG. In case B3, the compressive strength of the FRG was lower than that of the NSC without fiber. It is inferred that this was because the geopolymer used in this study has a larger dependence on the curing temperature, and the strength development was delayed in the FRG with the low average room temperature of 4.0 • C (Table 7). In addition, cases B1-2, B2-1, and B3-1 in Table 7, which had the same fiber volume fraction (1.5 vol%), showed different compressive strengths. This was likely due to the different average temperatures during the room temperature curing for cases B1-2, B2-1, and B3-1. Although coarse aggregate was used in the Table 6 Results of the splitting tensile tests for Series A (cases A4-A6, mixing procedure III, mix 3, with GNa). GNa: sodium gluconate, PVA: polyvinyl alcohol, WW: well water, IPW: industrial pure water, RTC: room temperature curing throughout, 50 • C (24 h): heat curing at 50 • C (24 h) followed by room temperature curing.   NSC: normal strength concrete.

Results of splitting tensile tests (Series B)
The results of the splitting tensile tests are shown in Table 8. Fig. 11 (a) compares the splitting tensile strength with the fiber addition ratio. As the addition ratio of the fiber increased, the tensile strength clearly increased. The tensile strengths between the NSC and FRG are compared in Fig. 11(b). The FRG with fiber (1.5 vol%) showed tensile strength higher than that of the NSC without fiber. From Table 8, cases B1-2, B2-1, and B3-1 with the same volume fraction of PVA fiber (1.5 vol%) showed different tensile strengths. The tensile strength decreased as the average room temperature during curing decreased. It is inferred that this was due to the temperature dependence of curing. Table 9 lists the obtained flexural test results. Fig. 12 compares the obtained flexural strengths. From Fig. 12(a), the flexural strength increased when the addition ratio of the PVA fiber increased. This is mainly owing to the stress and load transmission ability of the PVA fibers mixed in the FRG. The increase in the flexural strength by the PVA fiber corresponds to results in the literature [23,24]. In contrast to the compressive strength, which was not increased by PVA fiber addition [ Fig. 10(a)], the flexural strength clearly rose as the fiber amount increased from 0 to 3.0 vol% [ Fig. 12(a)]. From Fig. 12(b), the flexural strength of the FRG was significantly larger compared to that of the NSC without fibers, for each of the specimens with depths of 50 or 100 mm. From Fig. 12(c), the flexural strength of the NSC alone (specimen B3-6) was approximately 3.5 MPa, while the FRG-NSC composite specimens (B3-3 and B3-2) showed higher flexural strengths of approximately 6.1-6.6 MPa, which were almost equal to or slightly higher than the flexural strength of the FRG (specimen B3-1). In addition, Fig. 12(c) shows that the flexural strength of the FRG (specimen B3-1) was lower than that of the FRG-NSC composite specimen with the 50-mm FRG layer (specimen B3-2). It is inferred that this was because the NSC with higher Young's modulus and compressive strength compared to those of the FRG was placed at the compression side of the FRG-NSC composite   specimen (B3-2), while the FRG with lower Young's modulus and compressive strength was placed at the compression side of the FRGonly specimen (B3-1). Fig. 12 demonstrates that the placement of the FRG layer can enhance the flexural strength of the concrete composite specimens. Fig. 13 depicts a comparison of the hysteretic energies obtained from the flexural tests in Series B. Overall, a tendency was observed in the consumed energy (Fig. 13) that was roughly similar to that observed in the flexural strength (Fig. 12). From Fig. 13(a), the consumed energy increased as the PVA fiber amount increased, and the tendency is clearer than that shown in the flexural strength [ Fig. 12(a)]. This is attributed to the ductility enhancement [21] by the PVA fiber as well as the tensile    strength increase shown in Fig. 11(a). The results from Fig. 13(c) show that the placement of the FRG layer can significantly enhance the energy absorption performance compared to the NSC alone without fibers.

XRD analysis results (Series B)
Fig. 14 shows the obtained XRD pattern. From the XRD analysis result, quartz and mullite were mainly identified in the FRG sample. This result corresponds to the results of a past study on FA based engineered geopolymer composites [43]. Fig. 15 depicts the obtained FTIR spectrum, whose horizontal axis is from 4000 to 450 cm − 1 . From the FTIR spectrum, a pronounced band appeared at 1200-950 cm − 1 . This band may be attributed to the stretching vibrations of Si-O-Si and Al-O-Si [44] and considered a characteristic for the geopolymer formation in alkali activated systems [43]. Fig. 16(a) and (b) show the obtained SEM image (magnification × 1500) and the EDS analysis result for calcium (Ca), respectively, for the sample S1. In the SEM image [ Fig. 16(a)], the upper half was identified as NSC because it contained relatively more Ca [ Fig. 16(b)], while the lower half was identified as FRG. It is inferred that spherical objects seen in the FRG were unreacted or partially reacted FA [ Fig. 16(a)]. The boundary between NSC and FRG was confirmed in Fig. 16(b), and no clear damage was observed at the NSC-FRG interface in Fig. 16(a). Similarly, the SEM image for the sample S2 and the corresponding EDS result are shown in Fig. 16(c) and (d), respectively. From Fig. 16(c) and (d), no obvious damage was observed at the NSC-FRG interface. These results indicate that the NSC-FRG interface was not clearly damaged even near the failure zone (main flexural crack) when subjected to the flexural loading. This implies that a good bonding behavior at the NSC-FRG interface was achieved in the composite specimens.

Conclusions
In this study, trial mixes of FRG using local raw materials obtained in Hokkaido were investigated under different mixing procedures, mix constituents, PVA fiber volume fraction, curing conditions, and ages. The compression and splitting tensile strengths of the various FRG test pieces were obtained. Also, flexural tests using the FRG-NSC composite specimens with different FRG layer thicknesses were conducted. XRD, FTIR, and SEM analyses using the FRG samples were carried out. The following conclusions were obtained: 1) The compressive strength of the FRG (1.5 vol% and age 7 days) cured at 50 • C for 24 h followed by room temperature curing became  approximately 1.9 times that cured at room temperature only. Also, the addition of GNa improved the compressive strength, and differences in the mixing procedure affected the compressive strength.
2) The splitting tensile strength of the FRG significantly depended on the PVA fiber volume fraction; and the tensile strength for 1.5 and 3.0 vol% (age 28 days under room temperature curing) increased to 196 % and 239 %, respectively, compared to the cases without fibers.
3) The flexural strength and the energy absorption performance of the flexural specimens of the FRG alone were clearly enhanced as the addition ratio of the PVA fiber increased. 4) Placing the FRG layer on the tension side of the NSC specimens significantly increased the flexural strength and the energy absorption performance compared with those of the NSC-only specimens. The flexural strength for the NSC alone was approximately 3.5 MPa, while the FRG-NSC composite specimens showed higher flexural strengths of approximately 6.1-6.6 MPa (age 38 days under room temperature curing). 5) The SEM-EDS analysis results indicate that the NSC-FRG interface in the composite specimen was not clearly damaged even near the failure zone (main flexural crack) when subjected to the flexural loading, implying a good bonding behavior at the NSC-FRG interface.
This study contributes to advancement of effective and lowenvironmental-impact strengthening methods for existing concrete members using FRG as a strengthening structural element.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.