SIMULTANEOUS TREATMENT WITH OIL HEAT AND DENSIFICATION ON PHYSICAL PROPERTIES OF Populus × Canadensis WOOD

Samples of wood from Populus × canadensis (9,5 % moisture) were treated with olive oil at 195 °C simultaneously with 15 % or 30 % compression densification, and the results were compared with samples subjected to oil heat treatment without densification, and control samples. The density of the treated samples increased by 18 %, 43 % and 1,5 % respectively, and barely changed over the six subsequent months stored inside the laboratory room (at approximately 65 % RH, 20 °C). This was due to the fact that the slight weight increment caused by the additional moisture content was offset by the increase in volume from the springback effect. When subjected to atmospheres with different relative humidities, the treated samples stabilised at the same time as the control samples, although the treated samples had a significantly lower moisture absorption than the control samples. It was also observed that the hygroscopic shrinkage in oil heat densification treatment samples was approximately half those of the control samples. The initial densification was partially lost as a result of springback: approximately 3 % in the first springback at a relative humidity of 65 % RH, and an additional 4 % in the second springback to a relative humidity of 85 % RH. Once this latter relative humidity had been attained, no new losses in densification were observed. The ageing of the oil used in the treatment caused a slight loss of densification in the densest samples.


INTRODUCTION
Wood is a renewable material that offers undoubted advantages but also has its drawbacks. Various researchers have developed techniques for modifying wood to avoid or reduce its unfavourable properties. The objective of modified woods is to improve certain properties, for example, by increasing its dimensional stability, resistance to bases, acids and ultraviolet radiation and decreasing biodeterioration and hygroscopicity, or improving its mechanical performance (Forest Products Laboratory 2010).
One of the modification techniques is heat treatment, which consists of maintaining the wood at high temperatures for some hours (Kamke 2006, Kutnar and Sernek 2007, Forest Products Laboratory 2010. In Finnish ThermoWood and PLATO wood the treatment occurs by heating the air around the wood, while in rectification wood the process is carried out in an atmosphere with a high nitrogen content and less than 2 % oxygen (Rapp 2001). In oil heat treatment (OHT), the oxygen is separated from the wood by immersion in hot oil (Dubey et al. 2012b). The OHT samples presented a significant reduction in equilibrium moisture content and water ab-sorption compared to the untreated samples (Bak andNemeth 2012, Lee et al. 2018). Bak and Nemeth (2012) reported that water absorption is not blocked in samples treated with OHT, but merely reduced as it decreases the amount of places where the water can connect. This reduction is due to the lower number of hydroxyl groups in which the water can bind and to the increased crosslinking in the lignin (Lee et al. 2018). Equilibrium moisture is achieved at the same time in the OHT and the untreated samples (Bak and Nemeth 2012).
The shrinkage in the wood also improve with the OHT content. Dubey et al. (2012a) observed that this treatment causes a reduction in the volumetric swelling percentage and increases anti-swelling efficiency (ASE). The improvement in the dimensional stability of the wood is greater when the oil temperature and treatment time increases (Bak and Nemeth 2012). Rapp (2001) also observed a higher ASE in the samples treated at higher temperatures. Bak and Nemeth (2012) reported that in OHT wood treated with three vegetable oils the increase in ASE was greater in the tangential direction then in the radial direction. Although the anisotropy of the wood decreased, it is not completely removed. Okon et al. (2018) also observed greater reductions in shrinkage in the tangential than in the radial direction in wood treated with silicone oil as a heating medium.
One factor that can affect the OHT treatment is the type of oil used. The most commonly used oils are of vegetable origin. Some studies reported differences in the results based on the type of oil. Wang and Cooper (2005) found that palm oil was more effective than soybean oil for achieving dimensional stability. Lyon et al. (2007) considered that the most important characteristic of the oil is its unsaturation degree. Oils with a higher proportion of polyunsaturated fatty acids are more recommended for OHT treatments. Tomak et al. (2011) reached the same conclusion and used iodine values in their work to establish the unsaturation degree. However, studies by other authors observed no significant differences in equilibrium moisture or moisture absorption between OHT samples treated with sunflower oil, linseed oil and rapeseed oil (Bak and Nemeth 2012). Dubey et al. (2011) observed less moisture absorption in OHT samples treated with fresh oil than in samples treated with pre-heated oil, although there were no significant differences in volumetric swelling. In that work, the authors justified these differences as being due to the evaporation of volatile compounds and the heat polymerization of the oil.
Another common type of modification of the wood is densification, developed since the early 20th century (Kollmann et al. 1975). Densification consists of increasing density by applying pressure to the previously heated wood in such a way that the lumens partially collapse (Song et al. 2018). Wood is usually heated by means of liquid water or steam. Some authors have used other fluids, as in the case of Song et al. (2018), who successfully applied densification in a solution of NaOH and Na 2 SO 3 in their work. Densification is a technique that is particularly suited to low-density woods (Kamke 2006), because it enables them to be used for purposes where high density woods would normally be chosen (Kutnar and Sernek 2007). Kawai et al. (1992) indicated that during densification the cellulose crystals become reoriented, the hemicelluloses are partially hydrolysed and the lignin is partially degraded. The main application of densification is to increase the mechanical properties of the wood (Welzbacher et al. 2008, Gašparík et al. 2016, Laskowska 2020. Kutnar et al. (2008) obtained increases in MOE and MOR of between 35 % and 100 %, depending on the degree of densification. Sotomayor (2016) reported twice the MOE in densified wood than in the control samples. Kamke (2006) obtained MOE values in densified wood that were three times greater than in the control samples. It is important to highlight that the effect of densification may be reversible when the moisture content in the wood increases (Kamke 2006, Welzbacher et al. 2008), a phenomenon known as springback, or recovery from compression. This occurs when the internal stresses that appear in the densification are relaxed, and part of the wood's internal structure seeks to recover its original form (Morsing 1998). Springback is reduced through several mechanisms: by making the cell wall inaccessible to water, forming crosslinks between the wood components in the deformed state, and releasing stresses in the microfibrils during compression (Morsing 1998). Welzbacher et al. (2008) found that springback depends more on the temperature of the treatment then on the duration of the compression, and reported a particularly significant effect at temperatures over 180 °C.
To combine favourable effects, some researchers have analysed the behaviour of wood subjected to densification and subsequently to OHT. Fang et al. (2011) observed that in wood treated in this way the equilibrium moisture and the springback were reduced, and this reduction was greater in the samples treated with OHT at higher temperatures. Hsu et al. (1988) and Laborie (2006) indicated that in this combined treatment there was a joint occurrence of a reversible swelling due to the hygroscopicity of the wood and another irreversible swelling due to the compression-set recovery. Fang et al. (2011) quantified the incidence of the irreversible swelling by performing saturation tests on the samples, and observed that it decreased as the time and the treatment temperature increased.
In spite of the advantages obtained by performing the densification of the wood in combination with the OHT treatment, no works were found in which these two processes are carried out simultaneously to benefit of the heating caused by the oil treatment for wood densification. The aim of the present work was to establish the behaviour in regard to moisture content in low density wood of Populus × canadensis when the densification and OHT are carried out simultaneously.

Samples preparation
Ten logs were used from the I-214 clone of Populus × canadensis Moench, measuring 0,35 m long and with a diameter of 0,2 m to 0,3 m. They were obtained from a wooden fruit packaging company that uses trees from the Ebro Valley (Spain). Each log was identified with a letter (from A to J) to control the log factor. The largest possible quantity of strips with cross-sectional dimensions of 22 mm x 22 mm and 350 mm long was extracted from each log, with the rings parallel to the transversal edges. The strips were stored inside the laboratory until the samples attained a moisture content of near 12 %. They were subsequently processed with a thickness planer until they had a final section of 20 mm x 20 mm, when they were stored for another 15 days. Immediately before treatment, each 350 mm strip was cut transversally to obtain a sample of 275 mm in length, and a moisture sample of 70 mm which was only used to estimate the moisture content of the probe (Figure 1).

OHDT and OHT treatments
Four samples were randomly selected from each of the ten logs, one for each treatment type. The following treatments were performed: OHDT h : OHT treatment simultaneously with high densification, with an approximate 6 mm reduction in thickness.
OHDT l : OHT treatment simultaneously with low densification, with an approximate 3 mm reduction in thickness.
Control: samples without densification and without OHT.
The main samples of 275 mm were measured lengthwise with a 1 mm precision measuring tape and crosswise with a 0,01 mm precision micrometer. Both the radial and tangential crosswise measures of the samples were taken in three measurement points: two at 50 mm from the ends and one in the centre (Figure 1). Each measurement point was marked by circling the zone of contact of the micrometer with a pencil to be able to repeat the measures at exactly the same point. The sample was weighed on a 0,01 g precision scale.
The oil used in the treatment was virgin olive oil from Aceites Toledo S.A. (Los Yébenes, Spain). Fullana et al. (2004) indicated a smoke point of 210 °C for this type of oil, whereas Li et al. (2016) indicated a range between 205 °C and 215 °C. It was decided to use a temperature of 195 °C ± 4 °C in order not to reach the smoke point, which could lead to the premature degradation of the oil. The treatment was performed in a Nevir NVR-6522F fryer, and the temperature was checked with a Digiflex TP101 digital thermometer.
The densification of each sample was carried out individually, with a press formed by two steel plates with a thickness of 8 mm and 283 mm in length, joined with eight M10 cap screws (four each side of the probe). The time required for the interior of the sample to attain a temperature of 190 °C was calculated using MacLean's formula, based on Fourier' differential equation (Kollmann 1959). For the section of the samples of 20 mm x 20 mm, the estimated time was two minutes. To maintain a minimum temperature of 190 °C at all points of the sample for 60 minutes, a preheating phase was implemented, submerging the sample in oil at 195 °C for 62 minutes. In this phase the sample was loosely attached to the outside of the steel press with a steel wire with a section of 0,3 mm to prevent it from floating, and ensure that the oil was in contact with the entire surface of the wood. When the sample was introduced in the oil, an intense bubbling was observed caused by the expulsion of air and moisture from the interior of the wood. At the end of the preheating phase, the initial bubbling ceased.
After the preheating phase, the compression was performed by placing the sample between the two plates of the press, centring it between the two rows of cap screws. The plates were closed without tightening the cap screws, simply by adjusting the discs to the sample, and it was once again submerged in the oil. The press was removed from the hot oil every 15 minutes and each cap screw was tightened a quarter turn, four times (to avoid imbalances in the plates). The quarter-turn tightening sequence is similar to the one used in cylinder heads in combustion engines. In each extraction of the hot oil, the samples were tightened approximately 1,5 mm and it took one minute to become re-immersed in the hot oil. The compression was done in a radial direction to achieve plastic yielding and the gradual collapse of the wood cells (Reiterer and Stanzl-Tschegg 2001), locating the tangential faces of the sample in contact with the plates of the press. In the OHDT l treatment the plates were tightened twice, so a final densified thickness of approximately 17 mm was estimated (a decrease of 3 mm due to two tightenings of 1,5 mm each). The treatment time of the OHDT l samples was 92 minutes (62 minutes of preheating and two of heating before tightening periods of 15 minutes each).
In the OHDT h treatment the plates were tightened four times, so a final densified thickness of approximately 14 mm was estimated (decrease of 6 mm due to four tightenings of 1,5 mm each). The treatment time of the OHDT h samples was 122 minutes (62 minutes of preheating and four of heating before tightening periods of 15 minutes each).

Conditioning of samples
After the treatment, the samples were left to cool outside the oil and in the press for 20 hours. Once this time had elapsed, the samples were removed from the press, the radial dimensions were measured in the three measurement points, and they were weighed again. The treated samples and the control samples were stored inside the laboratory. The OHT and OHDT samples were treated simultaneously over three months. Once the treatment was finalised, all the samples were preconditioned for 15 days in a conditioning chamber at 65 % RH and 20 °C and subsequently stored for five months, grouping the ten samples of each treatment type in a plastic bag. The response to moisture content was analysed in five stable environmental conditions (SEC), alternating relative humidities 65 %, 85 %, 65 %, 85 %, and 65 % (SEC 1 to SEC 5 , respectively), all at 20 °C. At the start and end of each SEC the samples were weighed and measured in each measurement point. In the first four days of each SEC the samples were weighed every day, and on subsequent days the samples were weighed every four days. The moisture content of the wood was considered to be stable when the difference in weight taken four days apart was less than 0,2 %.
As the samples were not oven-dried in any part of the process, the moisture content of the samples was estimated from the 70 mm long moisture sample (Figure 1). The moisture content of the moisture samples was obtained from the difference in weights before and after oven drying at 103 °C, following EN 13183-1:2002EN 13183-1: (2002. Based on this moisture, the oven dry mass (m 0 ) of the treated and control samples was estimated using Equation 1. (1) Where m at is the mass of the sample before starting treatment and ω MS (in %) is the moisture content obtained from the 70 mm long moisture sample (Figure 1).
From the mass of each sample (m i ) in a SEC and from the estimation of its m 0 , the percentage of total sorption of oil and water (TS ow , in %) can be obtained by means of Equation 2. In the case of control samples, TS ow corresponds exclusively to water and is therefore equivalent to moisture content. In the samples treated with oil (OHDT h , OHDT l , and OHT), TS ow corresponds to a combination of water and oil.
Similarly, the partial sorption of water (PS w , in %) was calculated by means of Equation 3. This indicates the change in the mass of water that occurs between the start of a SEC i (m i ) and the end (m i+1 ).
To determine the recovery of form in the compressed samples, Fang et al. (2011) used the compression-set recovery variable. This was obtained by measuring the samples in oven-dry moisture content. In the present study, the compression-set recovery was not calculated because the aim was to simulate the behaviour of wood in natural conditions, and artificial oven-drying could produce irreversible alterations. This was also the reason that the anti-swelling efficiency was not used (Forest Products Laboratory 2010). The intensity of the densification of the samples was estimated by means of compression set (Welzbacher et al. 2008, Wehsener et al. 2018 Where CS is compression set (%), and R bt and R SECi are the mean of the radial dimensions in the three measurement points before treatment and after the SEC i stage respectively.

Statistical analysis
The differences between the groups were analysed by ANOVA, followed by Tukey's post hoc test. The densities of the samples before treatment, after treatment and before SEC 1 were compared by means of the paired samples T-Test. Linear regression was used to establish the influence of the age of the oil on the shrinkage. The statistical analysis was performed with the R program, version 3.6.1 (R Core Team 2019) with a significance level of 0,05.

RESULTS AND DISCUSSION
The treated samples darkened as a result of the treatment, and this change was maintained throughout the entire SEC. This effect coincides with the results of Dubey et al. (2011) in samples of Pinus radiata treated with linseed oil. Lee et al. (2018) justified the change in colour as being due to the formation of an oil layer on the wood surface and the caramelisation of soluble sugars produced from hydrolysed hemicellulose during heat treatment.
The densities of the samples according to the type of treatment are shown in Table 1. The mean density before treatment was 0,456 g/cm 3 (with 9,6 % moisture content), coinciding with the interval 0,420 to 0,480 g/cm³, included in AITIM (1997). Other authors obtained lower densities in their studies (Istok et al. 2016). This difference can be explained by the intrinsic variety in the properties of the wood, and is increased by the diversity of the development techniques that are applied to this species (plantation framework, pruning, watering, fertilisation, etc.). Table 1: Densities (in g/cm³) before treatment, after treatment and before conditioning. In brackets Coefficient of variation (in %). The lower part in each case shows the increase in density between stages. a b c Values with a different superscript letter in the same row present significant differences between treatments.
In the control samples, the density changed less than 0,001 g/cm 3 during the six months before treatment and before SEC 1 . The OHDT h treatment caused a significant increase in density of 43 %, and an increase of 18 % in OHDT l as a result of densification. A slight increase in density (1,5 %) was observed during the OHT treatment (without densification), representing less than 0,001 g/cm 3 . In the six months after the treatment, the treated probes underwent slight changes in density. In these months the moisture content only increased from 9,5 % to 12 %. There were two overlapping effects in this period, which modified the density of the OHDT samples in opposite ways: on the one hand there was a slight increase in moisture content, which led to an increase in density; and on the other there was a springback effect (Kamke 2006, Kutnar andSernek 2007) which increased the volume of the wood and therefore decreased its density. The springback was more significant in the OHDT h samples, which were more densified. A slight decrease in density was observed in these samples, as indicated by the negative increase detected in the density. In contrast, in the less densified OHDT l samples the effect of the moisture content was slightly greater than the springback, so the density increased slightly.
The changes in the amount of water and oil absorbed are shown in Figure 2. The stabilisation of the moisture content in each SEC of the treated samples and the control samples occurred at the same time, so the decrease in the moisture uptake rate in treated samples can be attributed to the lower water storage capacity (Bak and Nemeth 2012). The TS ow values corresponding to the control samples showed the usual hygroscopic behaviour for wood. In the control samples it was observed that in the different SECs of 65 % RH, a moisture content of between 11 % and 13 % was found; lower values were obtained when this was achieved from lower relative humidity (SEC 1 ) and higher values from higher relative humidities (SEC 3 and SEC 5 ). This difference in values is explained by sorption hysteresis (Forest Products Laboratory 2010). The treated samples followed parallel behaviours of TS ow , ordered according to the intensity of the densification. All the treated samples had the same TS ow before treatment, and separation occurred during treatment. From this point on the distances remained approximately constant. It is worth noting that the values of TS ow were higher in the treated samples than in the control samples. This appears to contradict the results of other authors (Jalaludin et al. 2010) who indicated that OHT wood exhibited a marked reduction in equilibrium moisture content. This discrepancy is explained because TS ow includes both water and the oil absorbed in the treatment, so PS w must be used when comparing only water contents. Another aspect that can be seen in Figure 2 is that the changes in TS ow between the 65 % SEC and 85 % SEC were always lower in the treated samples than in the control samples, both in adsorption and in desorption. This indicates a more mitigated exchange of moisture content in the treated than in the control samples. This result coincides with Bak and Nemeth (2012), who indicated that moisture uptake is not blocked in OHT samples but merely decreases. The explanation may be due to crosslinking caused by the polycondensation reactions in lignin and to diminishing amounts of water-affinity hydroxyl groups owing to the heat during the treatment and the crystallisation of the cellulose (Lee et al. 2018). The changes in PS w are shown in Table 2. At the end of SECs 3 to 5, the PS w values in the treated samples (approximately 3 %) were significantly lower than in the control samples (approximately 5 %). This confirms the mitigation of the moisture exchanges that were detected in the TS ow . Other authors have also found lower equilibrium moistures in samples treated with hot oil (Jalaludin et al. 2010, Bak and Nemeth 2012, Lee et al. 2018. The exception was the stage that elapses after treatment and before SEC 1 , in which the control samples practically maintained their PS w constant whereas the treated samples had a significantly greater increase of approximately 4 %. This initial increase that occurred over six months of stable relative humidity conditions reflected a partial recovery of the hygroscopic character of the treated samples. In this initial phase no significant differences were detected in the increases in PS w between the OHDT and OHT samples. A marked difference in PS w can be observed between SEC 2 and SEC 4 (with the same 85 % RH in both cases). This is because SEC 2 comes from an environment of 65 % RH obtained by adsorption, and SEC4 comes from an environment of 65 % RH obtained by desorption. This difference is explained by sorption hysteresis (Forest Products Laboratory 2010).Within the SECs, the OHT samples presented significantly higher PS w then the OHDT h samples, with some exceptions. The OHDT l samples with intermediate densification also had intermediate values. The moisture absorption was significantly lower in the samples that had been more densified.
The CS values are shown in Figure 3 and Figure 4. It can be seen that the OHT treatment (hot oil without densification) produced a 1 % shrinkage in thickness (Figure 3) due to the replacement of water by oil. The difference in thicknesses with regard to the control samples decreased to 0,4 % in the period prior to the SECs and remained more or less constant in all the SECs. The results showed some CS in the OHT samples which reached -0,7 % (equivalent to an increase in swelling of 0,7 %). The CS of the control samples were greater and reached -1,2 %, a less favourable value as it is farther from 0 %. In all SECs the control samples presented a CS of approximately double the one obtained in OHT samples. The decreases in shrinkage in the samples treated with hot oil coincided with the results of Dubey et al. (2012a) who detected improvements in the volumetric swelling percentage in samples of Pinus radiata treated with linseed oil. Several authors have also reported improvements in anti-swelling efficiency in samples treated with OHT (Rapp 2001, Dubey et al. 2012a, Bak and Nemeth 2012. Fang et al. (2011) reported the extreme case in which the compression recovery of aspen wood veneers was eliminated with OHT treatment. This could be explained by the thinness of this type of samples. The CS values in the densified samples ( Figure 4) were higher than those obtained in the OHT and control samples due to densification. The initial CS obtained with densification, 30 % in the OHDT h samples and 15 % in the OHDT l samples, decreased throughout the rest of the measurements. The main cause of this reduction is springback, which occurred in the initial phase at a relative humidity of 65 % (first springback) and in the phase of high relative humidity at 85 % (second springback). The first springback can be seen in the step between after treatment and SEC 1 which led to a 3,3 % loss of compression in the OHDT h samples and 2,6 % in the OHDT l samples. These values were similar to the 4,6 % and 1,7 % found by Dubey et al. (2016) in densified samples of Pinus radiata subsequently treated with oil at 180 °C and 210 °C. The second springback presented higher values of 5,0 % in OHDT h samples and 3,1 % in OHDT l samples. These correspond to the step between SEC 1 and SEC 3 : the two first SECs with 65 % relative humidity, separated by a SEC with 85 %, although part of this decrease was due to the fact that SEC 1 was caused by adsortion and SEC 3 was due to desorption, which could explain only a small part of the change. Kamke (2006) also found that the swelling of the densified wood occurred with greater intensity when the samples were submitted to high moisture content (second springback).
The presence of springback in the results of the present study coincides with the works of Navi and Girardet (2005) who observed that part of the densification is lost if the wood is re-wetted. As in the control and OHT samples, from the equilibrium in SEC 2 at 85 % RH, the changes in CS between the remaining SEC remained constant, indicating that springback had already totally occurred. The rate of swelling that occurred between environments of 65 % RH and 85 % RH differed depending on the treatment. The control samples had an oscillation in CS of 0,7 %, which was reduced to 0,5 % in the OHT samples. In the case of the densified samples, the oscillations rose to 1,6 %, and 1 % for OHDT h and OHDT l respectively. Spear and Walker (2006) also observed an increase in shinkrage in the densified samples, which they attributed to the increased ratio of cell wall mass to lumen volume. To verify this hypothesis, the CS values were compared in each SEC for the four treatments. No significant differences were detected between the CS of the control samples and the OHT samples in any of the five SECs. In contrast, the CS values in the densified samples were significantly greater than the control samples in the OHDT l samples, and even greater in the OHDT h samples. Table 3: Regression slope between the age of the oil (in minutes from previous treatments) and CS (in %). * = p-value between 0,01 and 0,05; ** = p-value between 0,001 and 0,01; ns = non-significant.
The influence of the age of the oil on CS can be seen in Table 3. Significant effects were only detected in the OHDT h samples. The negative value of the slope indicates that as the oil ageds, the capacity to maintain the densification decreaseds. This decrease remaineds more or less constant in all the SECs of the OHDT h samples, with a 0,12 % decrease in CS for each hour of prior use of the oil. In the OHDT l and OHT samples, the decrease was merely a trend, as all the slopes were negative even though the values were not significant. Significance was only found in the most densified samples, where the CS values are higher. However, in woods without densification or with slight densification, the influence of other effects may have masked the significance of the age of the oil. Figure 5 shows an example in SEC 5 . Dubey et al. (2011) detected significant differences in water absorption between woods of Pinus radiata treated with fresh linseed oil and pre-heated linseed oil. However, coinciding with the results of the present work, they found no significant differences in OHT samples when they compared volumetric swelling.

CONCLUSIONS
Once the samples had stabilised at 85 % RH, the OHDT samples presented significantly less water absorption than the untreated samples. This absorption was lower in the samples that were subjected to greater densification.
The springback in OHDT samples occurred in two ways: a first springback to constant relative humidity and a second springback when they were subjected to greater relative humidities. Once the balance had been attained at a higher relative humidity, the springback remained constant for equal or lower humidities.
In the more densified OHDT samples, the age of the oil used in the treatment had a significant influence on the loss of densification due to moisture. The OHDT treatment made it possible to obtain the advantages produced by the densification of the wood after an OHT process, but with a single treatment. This produces an energy savings as heat only needs be to be supplied in one of the phases. The treatment time is also reduced compared to when the sequential densification process is used for the OHT treatment.