Degradation of beech wood by Kretzschmaria deusta: its heterogeneity and influence on dynamic and static bending properties

2.1 Sample preparation

Small orthotropic bending samples (7 × 7 × 100 mm) of European beech (Fagus sylvatica L.) were crafted during January 2021 from four wood boards obtained from different fresh timber felled during the previous two weeks. Chosen wood came from the University Forest Enterprise Masarykův les – Křtiny (Czech Republic). These dimensions were chosen to fit the specimens to fungal culture in Kolle flasks and preserve the dimension ratio of bending specimens for mechanical testing according to the testing standards (20 × 20 × 300 mm). For samples’ crafting, only intact sapwood was used. To obtain oven-dry mass, samples were completely dried (moisture content (MC) = 0%) in a kiln for two days at 103 °C and weighed after approaching steady condition. Before vibroacoustic measurement and the samples’ inoculation, a higher moisture content (MC = 60 ± 10%) was established through vacuum impregnation with demineralised water (20 kPa, 3 min).

The moisture content (MC), green density ( ρ w ) and conventional density ( ρ c  – ratio of oven-dried wood mass and the volume of the same sample with a moisture range over the fibre saturation point) were calculated from sample dimensions and dry/wet (before and after degradation) masses.

In order to calculate a coefficient ratio to obtain the longitudinal dynamic modulus of elasticity ( E d y n l ) from the bending on M O E D , larger intact specimens with the same dimensions’ ratio (40 × 40 × 600 mm and 60 × 60 × 900 mm) were crafted and conserved in clean water to reach a moisture range over the fibre saturation point (MC = 60 ± 10%).

2.2 Fungal exposure

Each Kolle flask (400 ml) was filled with 70 ml of sterile malt extract agar base (malt extract = 30 g/l; mycological peptone = 5 g/l; agar = 15 g/l; final medium concentration = 5%). Flasks were inoculated with a K. deusta strain obtained in May 2020 from the wood of a mature standing beech tree located at the edge of a forest in Bilovice, Czech Republic. Before flask inoculation, K. deusta mycelium was cultivated on small beech wood blocks in Petri dishes to stimulate enzymatic activity. After two weeks of incubation at 22 °C and 70% relative humidity, mycelium completely covered the medium surface, and flasks were prepared for samples’ insertion. After the first vibroacoustic measurements using the frequency-resonance method, all the wood bending specimens were sterilised by steam (10 min, 120 °C). Both faces of half of the samples destined for fungal deterioration were covered with pure paraffin wax to try to limit the hyphal propagation in longitudinal direction (Figure 1A). All samples were laid on the radial face (Figure 1B). After 12 weeks of fungal exposure (Figure 1C), all specimens were removed from Kolle flasks, cleaned from superficial mycelium, and assessed with vibroacoustic techniques and static mechanical testing immediately afterwards.

Figure 1:

Specimen’s face covered with pure paraffin wax before inoculation (A), specimens in Kolle flask lying on the radial face (B), samples after 12 weeks of fungal exposure (C).

2.3 Vibroacoustic measurements

Resonant frequency technique was used to determine MOED. Mallet strikes on radial and tangential sides caused oscillations. Vibrations of longitudinal-tangential and longitudinal-radial bending modes (Figure 2A) and longitudinal modes (Figure 2B) were sensed using Doppler’s laser vibrometer PDV-100 (Polytec GmbH, Germany) and recorded using dynamic signal acquisition module DEWE-41-T-DSA with DEWESoft (DEWETRON, Inc., USA) at sampling frequency 20 kHz. Records of vibrations were transformed from time domain to frequency domain using the Fast Fourier Transform (FFT) processed using MATLAB^® (The MathWorks, Inc., USA). Obtained frequency was used for the calculation of MOED in tangential (t) and radial (r) directions using the following equation:

where f is the resonant frequency of the first bending mode, m is the sample mass, l is the sample length, and I is the moment of inertia. M O E D was calculated for both small degraded and larger intact samples. Due to the small length samples, it was not possible to measure the longitudinal mode frequency from which dynamic modulus of elasticity in the longitudinal direction ( E d y n l ) and c were calculated. Therefore, the conversion coefficient ratio from M O E D to E d y n l was determined on larger intact samples (40 × 40 × 600 mm and 60 × 60 × 900 mm), where their size allowed the measurement of longitudinal mode frequency (Figure 2B). E d y n l was calculated from the frequency of the first longitudinal mode shape of vibrations:

where ρ w is the green density, f is the resonant frequency of the first bending mode, and l is the sample length. According to data obtained from the measurement of the bigger specimens, the conversion coefficient was calculated through multiple linear regression (function ‘regress’ – MATLAB) from the mean of M O E D t and M O E D r to E d y n l . To determine c, dynamic modulus of elasticity in the longitudinal direction ( E d y n l ) was used:

where E d y n l is the calculated dynamic modulus of elasticity in longitudinal direction and ρ w is the green density. To evaluate the effect of the drying and sterilisation treatment process on M O E D , a group of 30 specimens was tested before and after the preparation process. After each operation, samples were acclimatised for one week (20 °C, 65% relative humidity) and tested. Results after the whole process were then compared with results obtained before treatment.

Figure 2:

Working set-up for the measurement of longitudinal-tangential and longitudinal-radial bending modes (A) used for M O E D calculation and longitudinal (B) for the calculation of E d y n l . Samples were supported at the nodes of the first bending mode (22.4 and 77.6% of the sample length).

2.4 Static mechanical testing

Static three-point bending tests were carried out on a universal testing machine (ZWICK^® Z050, 50 kN) (Figure 3A), where the force (F) applied to the sample in tangential direction was obtained. Bending stress in the tangential direction ( σ b ) was calculated according to the following equation:

where σ b is the bending stress, F is the applied force, l o is the distance between the two supports (12 times the sample’s height), and b , h are the base and the height of the specimen’s cross-section. The span/depth ratio was 12. For comparison, 30 reference intact specimens were also tested. Half of the reference group went through the same drying, sterilisation, and impregnation process (MC = 60 ± 10%); the remainder only underwent impregnation. Modulus of rupture ( M O R ) was calculated as the bending stress at the highest applied force ( F max ). Bending deflection ( ϵ b ) was evaluated by a point-probe optical displacement measurement based on the Digital Image Correlation (DIC) technique. Two cameras (AVT Stingray Copper F504B, cell size: 3.45 µm, resolution: 5 MPx, image-capture frequency: 2 fps) were used to acquire the images (Figure 3A). Images were processed in Mercury software (Sobriety s.r.o., Czech Republic), where the DIC technique allows the measurement of displacement at the point of applied force (Figure 3B). The static bending moduli of elasticity in tangential direction ( M O E t ) were established using a least-squares method fitting data from the zone of linear elastic behaviour from the stress-strain diagram of each sample. The data was processed using MATLAB^® _.

Figure 3:

Set-up of optical measurement based on DIC technique and universal testing machine during static testing (A), point-probes location for displacement measurement in Mercury SW (B).

2.5 Computed tomography assessment

After static mechanical testing, all samples were stored in a climatised room (20 °C, 65% relative humidity) for one week, together with 11 intact (reference) specimens. Afterwards, all samples were scanned with a computed tomography (CT) system DeskTom (RX Solutions, France) with a resolution of 80 μm. For each sample, 1000 cross-sections in the longitudinal direction were obtained. Because of the limited dimension of the chamber, all specimens were scanned in three different groups to reach a higher resolution (Figure 4A). Cross-sections of each sample were afterwards cropped in MATLAB^® (Figure 4B). During image cropping, approximately 0.5 mm from each side was cut off from each sample. To obtain the longitudinal density distribution of each specimen, all cross-sections were imported into MATLAB^® as numerical matrixes, and the mean value was calculated for each of them (Figure 4D). To eliminate the influence of cracks produced during static testing on the density distribution, mean values obtained from cross-sections where the crack was present were deleted and replaced with values computed through a linear regression between the first intact cross-sections before and after the crack. The density obtained by CT scanning ( ρ C T ) was calculated for each specimen as the mean of the numerical vector used for the longitudinal density distribution.

Figure 4:

Graphical representation data-elaboration process of CT results: cross-sections obtained from the CT scanning of a group of specimens (A), cropper cross-sections of one specimen (B), imported cross-sections in numerical form (1 pxl = 1 numeric value) (C), means of each matrix used for the comparison of longitudinal density distributions (D).

2.6 Scanning electron microscopy

A scanning electron microscope (SEM) was used to conduct microscopy analysis. Small samples (4 × 4 × 4 mm) were sectioned from three different specimens: one intact (reference), one degraded with paraffin wax, and one degraded without. Three small samples were prepared for each specimen: one taken from the centre and two taken 2 cm from the edges. The samples were softened with water, hand-cut with a razor blade to ensure a smooth surface, and dried at room conditions. The samples were sputter-coated with gold using a Luxor gold coater (Aptco Group, Belgium). The layer thickness was 15 nm. The cross-section and radial section were observed with SEM Tescan Vega 4 (TESCAN ORSAY HOLDING a.s., Czech Republic). The scans were prepared in RESOLUTION scanning mode in a high vacuum using a detector of secondary electrons (SE detector). The images were obtained with these settings: landing energy 7 keV, beam current 7 pA.

Relationships between measured parameters were statistically investigated in MATLAB^® (Spearman’s correlation coefficient (Sc) at \alpha=\ 0.05, and linear regression). Because of the higher variability of measured data (natural variability of wood material and high variability caused by fungal deterioration) and the sample size, the ANOVA significance level (\alpha) was set at 0.01, which is commonly used in biostatistics (Marusteri and Bacarea 2010; van Duong and Matsumura 2018).

3.1 Variability of density caused by fungal degradation

According to the results shown in Figure 5D, there was very high variability between density distribution of degraded specimens. According to ANOVA, no statistically relevant difference was proved among the variability of specimens with paraffin wax and without (p > 0.01). Nevertheless, compared with the degraded specimens, reference intact samples showed a very low variability and a statistically significant difference (p < 0.01) (Figure 5A). The graphical representation of each density profile obtained from CT scanning for both degraded and intact specimens supports this statement (Figure 5B). According to these results, the application of pure paraffin wax on the faces of specimens to limit the longitudinal propagation in the longitudinal direction through vessels (Rayner and Boddy 1988; Schmidt 2006) and consequent variability of degradation was not effective. The skewness of each sample group shown in Figure 5C shows a higher amount of lower local densities (calculated for each cross-section) influencing the mean.

Figure 5:

Densities and density profiles obtained from CT scanning ( ρ C T ) and comparison with conventional density ( ρ c ).

Observing the different density profiles showed in Figure 5A, there is no specific trend in differences suggesting that the deterioration was primarily concentrated in the extremities of the degraded specimens. The absence of any trend can be caused by the ununiform surface of the medium in the Kolle flasks caused by the growth of K. deusta’s mycelium. Each specimen was positioned on the cultivated mycelium only in some points and not along its whole length, so fungal growth and specimens’ degradation could have been influenced. Figure 5D shows a significant relationship between ρ C T and ρ c . According to their strong linear regression (r ² = 0.81), densities obtained from CT scanning can be considered representative. This result is supported by Fromm et al. (2001), who showed a strong relationship between densities assessed with CT scanning and gravimetric-volumetric method (r ² = 0.89 for spruce; r ² = 0.71 for oak). Freyburger et al. (2009) demonstrated a very strong linear relationship between CT scanning and gravimetric method (r ² > 0.99) using a calibration data set which consisted in tropical wood samples.

Figure 6 shows a different level of degradation in different samples obtained from degraded specimens (6B/C/D) and one reference intact sample (6A). Incipient decay shown in Figure 6B presents different stages of soft rot Type I with mostly isolated cavities in the cell walls of libriform fibres – typical for soft-rot fungi as K. deusta. Different authors have already described this kind of cell wall degradation (Levy 1966; Morrell and Zabel 1985; Schmidt 2006; Schwarze et al. 1995, 2004). After longer exposure to fungal deterioration, singular pits in the S₂ layer showed a significant thinning of the cell wall (6C). In a stage of progressed degradation (6D), the only intact part of the lignified cell wall was the compound-middle lamella. Schwarze et al. (1995) stated that its preservation is not only caused by its high lignin content but also by the high percentage of Guayacil lignin. These different levels of deterioration were found on the sides (2 cm from the edges) of the degraded samples and in the middle of the specimens. Obtained results from SEM scanning supported the results presented in Figure 5B, where no specific degradation trend among the specimens’ length was observed.

Figure 6:

SEM scans of samples obtained from intact and degraded specimens: intact beech wood (A), incipient degradation – isolated cavities in the S₂ layer (B), connection of the cavities of the S₂ layer (C), and progressive degradation – preserved middle lamellas (D).

3.2 Influence of mass loss on MOR

Figure 7A shows the mass losses for the degraded samples with paraffin wax and without. According to the Kruskal-Wallis test, no statistically significant difference between these two groups was proved (p > 0.01). The average mass loss was 31%, with the maximum and minimum 54.6 and 17.2%, respectively. Schwarze et al. (1995) presented the mass loss caused by K. deusta on beech and large-leaved linden wood. After 12 weeks of incubation, beech wood presented an average mass loss of 4.5%, while for inden wood was 8.3%. The lower mass loss was probably caused by the different specimen dimensions (45 × 20 × 20 mm), presenting a lower specific surface contact area. Deflorio et al. (2008b) showed that the degradation caused by K. deusta on standing beech trees felled 28 months after artificial inoculation caused a 12.9% mass loss. These results showed that the degradation in standing trees is relatively slower compared to in-vitro deterioration. This difference is caused by the presence of parenchyma cells and their task in the compartmentalisation of decay in standing trees, retarding the process of wood degradation (Schwarze et al. 2004).

Figure 7:

Mass loss of degraded samples (A) and comparison between M O R and ρ c (B).

According to Figure 7B, considering the biological origin of measured values and the even higher variability caused by the degradation process, a relevant moderate linear relationship between ρ c and M O R was proved (r ² = 0.54). A strong negative correlation was also proved between mass loss and M O R (Sc = −0.77). No difference in MOR was proved between sterilised and non-sterilised reference specimens (p > 0.01). Curling et al. (2002) showed the relationship between mass loss and M O R in southern pine caused by Gleophyllum trabeum and Rhodonia placenta, where a linear trend was observed till 20% mass loss. At higher mass losses (up to 40%), MOR reduction dropped. Xu et al. (2019) also proved a strong linear relationship (r ² = 0.75) between mass loss and compressive strength of Chinese poplar blocks exposed for 90 days to G. trabeum. Lundström et al. (2007) also showed a strong relationship (r ² = 0.92) between ρ c and M O R of green Norway spruce wood. According to our data, the variance of density among each specimen correlates with the average density (Sc = −0.69).

The weaker linear relationship proved in Figure 7B can be explained by the higher variability of density among the length of the degraded specimens, showing the influence of the heterogeneity of degradation caused by K. deusta on bending testing results. With a decrease in density (progression in the degradation process), an increase in the variability of the density profile in longitudinal direction occurred. The variation in density among each specimen also had a relevant correlation with MOR (Sc = −0.58), which can be explained by the strong correlation between density and coefficients of variance. According to these results, we can state that the variability of density caused by fungal degradation should be considered during the assessment of M O R reduction.

3.3 Relationship between static/dynamic elastic parameters and MOR

During the assessment of vibroacoustic properties, no significant influence of drying and sterilisation treatment processes was proved (p > 0.01). Therefore, all changes in mechanical and vibroacoustic properties were caused only by fungal degradation. Figure 8 shows the relationship between M O R and different elastic and dynamic properties ( M O E t , M O E D t , E d y n l and c l ). The coefficient ratio obtained from the vibroacoustic measurements of big bending specimens and used to calculate E d y n l from M O E D t / r was 1.1166. According to Figure 8B, a strong linear relationship between M O R and M O E t was proved (r ² = 0.79). The strong correlation between M O E t and M O R contradicts the conclusion proposed by Schwarze et al. (1995), who stated that wood degraded by K. deusta during incipient decay loses strength faster than stiffness.

Figure 8:

Comparison between M O E D t and M O R (A), M O E t and M O R (B), E d y n l and M O R (C), and stress-wave velocities in longitudinal direction ( c l ) and M O R (D).

Figure 8A shows a strong linear regression between M O E D t (calculated from the vibroacoustic measurement) and M O R , which is also confirmed by the strong correlation between M O E t and M O E D t (Sc = 0.92). These results concur with Baar et al. (2016), who proved strong linear relationships between M O E t and M O R (r ² = 0.73) and M O R and M O E D t measured with a flexural resonance method (r ² = 0.87) of different tropical wood samples at 8% M C . As shown in Table 1, values for static and dynamic measured and calculated properties for degraded samples with paraffine have higher variability than values obtained for degraded specimens without wax. Nevertheless, as for the variability of measured density for degraded samples shown in Figure 5A, these differences are not considered statistically relevant (p > 0.01).

Comparing the relationships of E d y n l and c l against M O R (Figure 8C and D), it is possible to see a relevant correlation of both parameters with M O R . Nevertheless, E d y n l had a higher r ² value (r ² = 0.72). To that, some degraded samples with similar c l to the reference group, showed a pronounced strength loss and the relationship between c l and M O R for intact specimens did not have the same trend as the degraded specimens (Figure 8D). On the other side, the linear regression calculated for E d y n l and M O R was also applicable to intact samples (Figure 8C). Obtained results agree with previous investigations, which could not detect soft rot decay by K. deusta with a laboratory stress-wave timer (Schwarze et al. 1995). Deflorio et al. (2008a) showed decreased stress-wave velocity in standing beech trees caused by the artificial inoculation with K. deusta. The authors showed a statistically proven difference (α = 0.05) comparing two measurements, 2 and 27 months after inoculation. Nevertheless, they did not consider the strength loss caused. The stronger relationship between M O E D t and M O R suggests a strong influence of ρ on strength-loss prediction. According to the above-mentioned results, utilization of calculated dynamic moduli of elasticity trough measured velocities and approximately known absolute densities can lead to an improvement in results interpretation of instrumental methods (e.g. acoustic tomography) used for non-destructive tree stability assessement. M C of degraded samples was highly variable ( M C min / M C max  = 58/149%; coefficient of variation = 27.89%). Nevertheless, no significant correlation between M C and M O R was proved (Sc = −0.15). This result supports the conclusions proposed by Niklas and Spatz (2010), who proposed mechanical testing for green wood above 50% M C to eliminate its influence on wood strength.