Wood Science and Technology

, Volume 47, Issue 3, pp 481–498

Stiffness of normal, opposite, and tension poplar wood determined using micro-samples in the three material directions

Authors

    • Ecole Centrale Paris, LGPM
  • Anh Tuan Dinh
    • AgroParisTech, LERFoB UMR1092
  • Carole Assor
    • INRA, UR BIA
  • Xavier Frank
    • INRA, LERFoB UMR1092, ENGREF
  • Gilles Pilate
    • INRA, UR0588 Amélioration, Génétique et Physiologie Forestières
Original

DOI: 10.1007/s00226-012-0511-x

Cite this article as:
Perre, P., Dinh, A.T., Assor, C. et al. Wood Sci Technol (2013) 47: 481. doi:10.1007/s00226-012-0511-x

Abstract

Mechanical tests on micro-samples were performed in the three material directions in normal, opposite, and tension wood collected from a poplar tree. Two custom micro-devices were designed and built in the laboratory to test samples under pure tension in the transverse direction and under 4-point bending conditions in the longitudinal direction. Both devices were designed to handle samples with a small transverse section (a few square mm), which allowed to select zones with homogenous anatomical features. The results indicate a very high longitudinal stiffness in tension wood (up to 35 GPa compared to an average of 18 GPa for normal wood). Considering wood density, the value represents a specific modulus that is nearly 70 % crystalline cellulose. However, tension wood is slightly less stiff in the tangential and radial directions (1,150 vs. 1,500 MPa for normal wood in the radial direction and 430 vs. 530 MPa in the tangential direction).

Introduction

Wood is an anisotropic, variable, and heterogeneous material. Accordingly, its mechanical properties depend on its material directions [Longitudinal (L), Tangential (T), and Radial (R)] and vary among and within species, and even with the position inside the tree, where different types of wood are classically described: sapwood/heartwood, earlywood/latewood, normal wood/reaction wood. This high wood heterogeneity explains the huge stiffness variability encountered in wood. Therefore, many works were published to understand which parameters at ultrastructure and anatomical levels are involved in the macroscopic stiffness, and how these parameters could be used to predict macroscopic values. In the longitudinal direction, the density proved to be an excellent indicator of the stiffness (Kollmann and Côté 1968; Guitard 1987). This results from the fact that most of the xylem cells (fibers and vessels) are oriented along the axial direction of the stem, with the exception of ray parenchyma cells, in order to ensure two of the xylem main physiological functions in living trees, namely mechanical support and water conduction. However, even if a rather good correlation exists between macroscopic density and macroscopic stiffness (properties determined at the cm scale), density fails to be a good predictor of wood stiffness due to the importance of mean microfibril angle (MFA), a highly variable ultra-structural parameter independent of wood density (Butterfield 1998).

The situation is more complex in transverse directions (radial and tangential). In the cross section, the intricate coupling between the anatomical features at different spatial scales is necessary for a comprehensive understanding of wood stiffness (Booker and Sell 1998; Bergander and Salmen 2002). These scales broadly include the macromolecular composition, the cell wall organization, the cellular structure, the annual ring anatomy, and the macroscopic heterogeneities such as heartwood versus sapwood, knots, and grain angle. In order to assess the coupling between these scales, two important and complementary fields of wood science were developed: (1) formulation and modeling of successive scaling in wood and (2) mechanical characterization at different spatial scales.

Modeling includes classical scaling approaches (homogenization) able to pass from one scale to the next one. In this strategy, scaling examples are found to pass from the macromolecular structure to the cell wall (Marhofer et al. 1996), from the cell wall to the cellular arrangement (tissue) (Gillis 1972; Gibson and Ashby 1988; Koponen et al. 1991; Farruggia 1998; Nairn 2007), or from the tissue to the annual growth ring (Perré and Badel 2003). The advancements in computer technology and the field of applied mathematics has resulted in reports on multiscale approach attempts during the last decade (Holmberg et al. 1999; Hofstetter et al. 2005; Qing and Mishnaevsky 2009; Simon 2009; Frank and Perré 2010; Perré 2010). However, all scaling approaches need rigorous experimental data, to supply the models with data at the lower spatial scale or to validate the properties predicted at the upper scales. Several experimental methods have been suggested to achieve such difficult characterizations at the microfibril level (Sakudara et al. 1962; Nakai et al. 2005; Cheng et al. 2009), at the cell wall level (Gindl and Schöberl 2004), or on individual fibers (Navi and Heger 2005).

Measurements at the tissue level have several requirements, which are difficult to merge. Some of the opposing requirements are as follows:
  1. 1.

    The sample must be small enough for the anatomical features to be homogeneous,

     
  2. 2.

    The sample must be large enough for the measured properties to be representative of the tissue,

     
  3. 3.

    The sample must be long enough to use St. Venant’s principle, and

     
  4. 4.

    The strain field must be determined without perturbation of the stress field.

     

It is possible to circumvent some of these requirements. For example, modern 3D experimental tools provide comprehensive geometrical sample description, initial porous medium morphology, and the comprehensive deformation field during the test (Forsberg et al. 2008). The information obtained combined with a comprehensive mechanical model may afford a rigorous determination of mechanical properties using the inverse method. However, this procedure remains too complex to be applied to a full sample set.

One has also to keep in mind that reaction wood, which always exists in trees, is one additional cause of structure heterogeneity at the cell wall and tissue levels. Reaction wood is a particular type of wood which is modified in its anatomical and chemical properties. It is formed on one side of the stem to make possible its reorientation in response to environmental cues (wind, snow, light, gravity, etc.). The wood formed on the other side, called opposite wood, has also peculiar features and cannot be considered as normal wood. In hardwood, the reaction wood is referred to as “tension wood,” which has much higher longitudinal maturation tensile stress than normal wood. Consequently, the formation of tension wood in one side of the stem allows reorientation toward this side (Jourez 1997; Thibaut et al. 2001).

The most outstanding characteristic of tension wood is the presence of a thick layer termed G-layer characterized by its high cellulose content. Indeed, the G-layer structure is composed of 80 % cellulose (Yamamoto 2004) compared to 40 % in an earlywood normal cell (Bergander and Salmen 2002). According to Marhofer et al. (1996) and Sakudara et al. (1962), the Young’s modulus of the crystalline cellulose structure is in the range of 134–137 GPa. The reported value is ca. 20 times higher than hemicelluloses, and 30–60 times higher than lignin (Neagu and Gamstedt 2007). In addition, the average MFA in tension wood is smaller than that in normal wood. Yamamoto et al. (1992) reported a MFA in the S2 layer of tension wood of 23.1° compared to 27.2° in normal wood. More specifically, the MFA in the G-layer is almost parallel to the cell axis (Yamamoto 2004). Based on these facts, the G-layer Young’s modulus values were estimated as ten times stiffer than a normal wood cell wall (Okuyama et al. 1990). Therefore, the mechanical behavior of tension wood in the longitudinal direction is reinforced. The Young’s modulus value is much higher than that of normal wood with an estimated value of 45.1 GPa in tension wood compared to an average of 13.3 GPa in normal wood (Yamamoto et al. 1992).

Concerning the transverse direction, it should be noted that the G-layer is weakly attached to the secondary wall (Kollmann and Côté 1968) and can be easily detached from the cell wall. Consequently, the wood behavior in the transverse plane is not reinforced by the presence of the G-layer but differs from normal wood due to the different S1 and S2 layer thicknesses and the lignin concentration in the layers (Keller 1994).

The main outcome of the present work is a full data set of stiffness values for poplar in the three material directions and for well-identified anatomical features, including tension and opposite wood. The tree used in this work was artificially bent during its life to force it to produce reaction wood on one side of the stem. Another interesting consequence of bending the tree is a quasi-absence of tension wood in the rest of the stem section.

In spite of this careful selection of material, the goal remains quite challenging, as the sample section in the transverse plane has to be very small to obtain a homogeneous cellular structure. A custom machine was conceived and built to perform tensile tests both in radial and in tangential directions. The good image quality of carefully polished samples allows the strain to be directly determined on the sample using the anatomical pattern as texture. However, this machine is not suitable for longitudinal tests, as (1) the MOE is much higher in this direction (2) the absence of clear anatomical pattern in this direction does not allow the strain field to be determined with enough accuracy.

For this reason, a second micro-device was conceived in order to test the longitudinal samples in pure bending (4-point bending).

Experimental procedure” section presents in detail the sampling procedure and the two micro-devices, whereas “Results and discussion” section is devoted to the result presentation and discussion.

Experimental procedure

Materials

In this study, the selection of the plant material was of the upmost importance, as the goal was to prepare samples containing only normal, tension, or opposite wood. A study performed on 10 poplar cultivars in commercial plantations reported a proportion of tension wood ranging from 5 % to almost 25 % (Marchal et al. 2009), depending not only on the genotype but also on the position in the tree and on the growth conditions (wind, quality of soil…).

In another experiment on three Populus deltoides x Populus nigra hybrid clones (I214, Robusta and Blanc du Poitou) grown in a green-house, the average proportion of tension wood, measured on three one-year-old trees per clone, ranged from 0 to 3 % (Perré et al. unpublished data). The low proportion observed in this case probably comes from the growth conditions, which preserves the plants from the wind.

When one-year-old trees from the same clones grown in similar conditions were intentionally bent at 30°, the proportion of tension wood fluctuated between 15 and 45 % of the wood formed after bending. As a comparison, tension wood proportion in one-year-old trees from the clone used in the present study (INRA # 717-1B4, Populus tremula × Populus alba) grown in comparable conditions, raised, up to 41–53 %. When trees were bent over several years, this proportion remained roughly around 50 %. This demonstrates the decisive advantage of using artificially bent trees in order to obtain high proportions of tension wood with a precise localization.

The study was performed on a 10-year-old female hybrid poplar clone (INRA # 717-1B4, Populus tremula × Populus alba). Planted in 1996, this tree was artificially bent in 1998 to force it to produce a well-defined sector of tension wood at the upper side of the stem. Because of the small diameter (approximately 15 cm; Fig. 1) and young age of the tree, samples were all in the juvenile wood. Considering the tension wood zone, the cross section of the stem was divided into three different zones: tension wood (T), opposite wood (O; 180° from the tension wood zone), and normal wood (N; 90° from the tension wood zone) (Jourez et al. 2001).
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig1_HTML.jpg
Fig. 1

Material tested in this study a Artificially bent poplar tree and b sampling zones in disk

Accordingly, the samples were cut from these three zones: tension wood, opposite wood, and normal wood (Fig. 1). Each sample was numbered according to its angular and radial position in the disk: Ti, Oi, and Ni, where the letters T, O, and N stand for tension, opposite, and normal wood, respectively. The index “i” is the sample number from pith to bark. Due to the small annual ring width in the opposite wood zone, sampling was done from two successive disks in the longitudinal direction.

Four-point bending test

To determine the longitudinal stiffness, a specific four-point bending micro-machine was designed and built in the laboratory. The device takes advantage of the load system utilized in old laboratory microbalances. In the microbalance systems, the mass was determined by balancing the unknown mass using a collection of calibrated masses. A clever system of cams and handles allowed the controlled mass to gradually increase or decrease by different increments to seek equilibrium. This original system was kept to apply various weight loads to the sample. A four-point bending frame was designed and adapted to the original balance (Fig. 2). Small cylinders mounted on minute ball bearings provided a pure four-point bending, without any axial forces despite the sample bending. This ensured that a constant bending moment acted on the sample in the portion between the inner supports (Fig. 3). The sample deflection was measured without contact using a laser beam micrometer (Bullier M5L/2, ±1 mm full scale, 0.5 μm resolution including noise).
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig2_HTML.jpg
Fig. 2

Micro-machine used for the 4-point bending test; this machine was designed and developed by the authors’ team

https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig3_HTML.gif
Fig. 3

Geometrical and mechanical configuration used in 4-point bending tests

The material direction was controlled by taking a block of wood from the disk using a splitting method. Then, a micro-circular saw (Struers Accutom-5) fitted with two parallel blades was used to obtain a flat piece of uniform thickness in the tangential direction. The sample was then carefully cut in the radial direction from the plate using a craft knife guided by a ruler. This procedure was chosen to avoid any loss of wood in the radial direction. Finally, a special support was conceived to handle and polish the radial faces of the sample at the desired and constant thickness (ESCIL, manual polishing machine using glass disks to ensure excellent flatness). The final sample dimensions were ca. 35 × 1.8 × 0.9 mm3 in the longitudinal, tangential, and radial directions, respectively. The actual dimensions were carefully measured at different positions with a caliper before each test. The averaged values were accepted as being representative of the sample geometry. The samples were soaked in water for 24 h under several vacuum cycles. A first series of tests was then performed using the fully saturated samples. After the first series, the samples were left under ambient conditions for several weeks and the polishing phase was repeated to eliminate geometrical defects due to drying. A second series of tests was performed in the air-dried state (MC close to 8 %). The dimensions of the dried samples after polishing were about 35 × 1.7 × 0.7 mm3 (L, T, and R directions, respectively).

In the protocol, the applied force ranged from 0 to 1.5 N, in increments of 0.1 N. For the dried samples, each step could last for 30 s without significant creep. The tests performed on the saturated samples were much more difficult because they required reduced duration of each step at the minimum (seconds) to limit the high creep rate due to the water plasticizer role on the cell wall. The comparison of loading and unloading curves enabled confirmation of the measurements in the linear elasticity domain.

In the linear domain (linear elasticity and small deformations), the deflection, f, at mid-length, increases linearly with the load, F. To avoid sensor offsets, the sample stiffness, E, was determined using the slope, α, in the deflection versus load curve:
$$ E = \frac{{l^{2} (L - l)}}{32I\tan \alpha } $$
(1)

In Eq. (1), L is the distance between the two external flexion points, l is the distance between the two internal flexion points, and I is the section inertia \( \left( {I = \frac{{{\text{width}} \times {\text{height}}^{3} }}{12}} \right) \)(Fig. 3).

Tensile test

The tensile tests were performed on a micro-tensile device designed and developed by the authors’ team several years ago (Badel and Perré 1999; Farruggia and Perré 2000). The experimental device comprises a rigid metallic structure, a load gauge, a micro-positioning device, and a system of two jaws designed to tighten the sample. The whole system was placed under an optical microscope in reflection mode (Fig. 4). The force was measured with a resolution of 0.1 N with a maximum load of 100 N. The tests were performed on small samples with typical dimensions of 4 (height along the L direction) × 1.5 (width) × 12 mm3 (length).
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig4_HTML.jpg
Fig. 4

a General view of the tensile device placed on the stage of the optical microscope and b typical image obtained under reflected light with a polished surface

The camera was placed on the optical microscope which captured digital images of the sample surface at each load increment. To obtain a clear image, the observed surface of the prismatic specimen requires careful polishing using a micro-abrasive disk. Decreased grain sizes to 3 μm were of suitable quality. A 20× lens was used which produced a field width of ca. 0.5 mm.

At certain cellular levels, strain field measurements can be obtained without perturbations only without contact. The good image quality of carefully polished samples allows the strain to be directly determined by image correlation without artificial markers using the anatomical pattern as texture. The method was based on a global comparison of the gray levels between a deformed image and the initial image. The comparison required two assumptions:
  1. 1.

    The gray level of one specific sample location does not change from the initial state to the deformed state, only its position changes.

     
  2. 2.

    The deformation is constant over the field of view which implies that only the average deformation of the entire image can be determined.

     

With these assumptions, a “virtually” deformed image was calculated from the initial image using the values of six independent parameters (for example two independent parameters for the translation and four independent parameters to define the deformation gradient tensor in 2D). The whole procedure was implemented using the custom image processing software, MeshPore (Perré 2005). The present procedural detail was improved upon (Perré and Huber 2007) utilizing the inverse method to determine the deformation gradient tensor instead of the strain tensor which allows the assumption of small deformations to be discarded.

The entire procedure was CPU-intensive because the objective function was computed on most image pixels (thin bands along the border were not considered to be sure that the computed image remains inside the initial frame). However, in return, there was a gain in accuracy, with a quality image, such as the one depicted in Fig. 4. The strain field calculated on the image of one object simply translated is zero to within 10−5. The accuracy combined with the strain measurement without contact explains the quality of the data which is further supported by the linear relationship between variables having an excellent correlation coefficient (Fig. 5). An interesting feature of the inverse method relies on its ability to produce full 2D strain tensor values. The strain value in the load direction is directly related to the stiffness, and the strain value in the perpendicular direction provides Poisson’s ratio. Finally, the shear strain was useful if the sample had a defect or was not properly cut along the material direction: if the value did not remain close to zero, the test was rejected.
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig5_HTML.gif
Fig. 5

Typical strain/stress curves obtained without contact for sample T1 using image correlation. Note the quality of the linear regressions

A specific sampling was prepared for the series of transverse measurements. Samples located next to each other were cut along the longitudinal direction in the three zones, respectively, named: Ti, Oi, and Ni, for tension, opposite, and normal wood, respectively. The sample section was about 5 × 2 mm2 (width × thickness) for samples tested in the radial direction and 3.5 × 2 mm2 (width × thickness) for samples tested in the tangential direction. Long samples, especially in radial direction, were difficult to obtain, and therefore, the wood samples were glued in disposable U-shape plastic supports (Fig. 6). The plastic parts provided grip between metal jaws, rather than the wood sample. The use of the clamps simplified sample handling and better fulfilled St. Venant’s conditions.
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig6_HTML.jpg
Fig. 6

Disposable U-shape plastic supports used to extend the samples cut for radial and tangential tensile tests

The test protocol utilized the following protocol: every 20 s, a displacement increment of 20 μm was imposed, the indication of the load cell was recorded, and an image of the sample was captured. These images were subsequently treated using the image analysis software to determine the strain tensor in the microscope focal plane. The sample modulus of elasticity (MOE) was calculated from the linear part of the experimental stress/strain curve. The Poisson’s ratio was simply calculated as the ratio of the slopes obtained from the tensile and perpendicular strain values.

Density

The density was systematically determined for the longitudinal samples because their geometrical shape was representative of the tested tissue (small section in the R-T plane). However, this was not the case for samples cut in the disk for R or T tensile tests: indeed, due to the sample length (either in R or T), tissues of various features were present in the sample volume. The density measured in this case was therefore not representative of the type of tissue tested and was not reported here.

The basic density of the saturated samples was determined using conventional methods:
  1. 1.
    The green volume was determined using Archimedes’ principle, which calculates the sample volume by sinking a saturated wood sample to a fixed grid placed in a container filled with water (Fig. 7).
    https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig7_HTML.gif
    Fig. 7

    Simple device designed to measure the green volume of fully saturated samples

     
  2. 2.

    The dry mass was weighed on an electronic balance after stabilization in a dry chamber set at 103–105 °C.

     
  3. 3.

    The oven-dry mass was then divided by the green volume.

     

For such small samples, great care was required to determine the green volume, as the surface over volume ratio increases as the sample size decreases. The difficulty lies in the surface water removal when taking the sample from the container just before applying Archimedes’ principle. It is important to gently clean the sample using a wipe that was nearly water saturated with water before use to avoid sample water removal through capillary suction. The averaged value of three repetitions was accepted as the green volume.

After drying under room temperature conditions, the samples were polished again to ensure geometrical shape. During mechanical measurements under air-dried conditions, the samples were systematically weighed. The air-dried density was subsequently determined using the mass and volume calculated by the multiplication of the three averaged dimensions, assuming a perfect geometrical shape. In spite of the small size of the samples, a strong correlation was found between the air-dried density (MC ≈ 8 %) and the basic density:
$$ {\text{Air-dried}}\;{\text{density}} = 1.152 \times {\text{Basic}}\;{\text{density}}(R^{2} = 0.92) $$
(2)

Results and discussion

Longitudinal direction

The full data set obtained in the longitudinal direction is plotted as a bar graph in Fig. 8. The air-dried samples exhibit a MOE value higher than the saturated samples. The observed result confirms the effect of moisture content, which is well reported (Kollmann and Côté 1968). On average, air-dried samples display stiffness values 39 % higher than saturated samples: +19 % for normal wood, +53 % for opposite wood, and +39 % for tension wood. Note, however, that the MOE determined on saturated samples seems to be more scattered than for air-dried samples, which may be an effect of the difficulty to use this device on very small and saturated samples due to the time required for a full test. Therefore, the remainder of the data discussion will focus on air-dried samples.
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig8_HTML.gif
Fig. 8

Longitudinal MOE determined for the three types of wood: tension (T), normal (N), and opposite (O). Air-dried state (left bar) and saturated state (right bar)

For the air-dried samples (ca. 8 % MC), the average MOE values are equal to 18.2, 15.5, and 34.3 GPa for normal, opposite, and tension wood, respectively. The values obtained for normal poplar wood are larger than those available in other literature data (Guitard 1987; Marchal et al. 2009), which reported a longitudinal poplar MOE value at 12 % MC ranging from 8 to 10 GPa. This difference is certainly due to the effect of sample size on the MOE, which is quite well known in materials science. The micro-samples here are highly homogeneous and defect-free compared to samples of structural size.

The most noticeable data set is the much higher MOE obtained for tension wood, which is roughly double that of normal wood. The G-layer present in tension wood partially supports the larger MOE value through the effect of density; however, the high cellulose content with crystallites closely aligned along the cell axis in this layer is the more likely explanation.

For a better overview, the MOE of all samples is plotted as a function of density (Fig. 9). This graph is of high interest because two data sets are easily distinguished. The first one includes both normal and opposite wood with a positive correlation, as expected, between MOE and density. This unique correlation tells us that the slightly lower MOE obtained for opposite wood compared to normal wood is the effect of lower density values for opposite wood (14–17 GPa in opposite wood compared to 16–22 GPa in normal wood). The data set emphasizes the stem reorientation efficiency with an organized structure of tension wood and a larger annual ring width on one side and a high level of maturation stress level and a thin annual ring of less dense wood on the opposite side.
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig9_HTML.gif
Fig. 9

Longitudinal MOE values for the three types of wood plotted versus air-dried density (at ca. 8 % MC)

In comparison to the first set, the data obtained for tension wood exhibit higher MOE values, but, surprisingly, these values are independent of density. One tentative explanation is based on the limited ability of the multilayered cellular structure to transfer the G-layer stiffness to the tissue (shear effect between the G-layer and S2+ shear effect between the cells). This would explain why a deeper expression of tension wood features (thicker G-layer which results in higher density) would not affect directly the stiffness of the tissue. A careful, but bold result analysis was evaluated on saturated samples (Fig. 8) and provided confirmation that possible artifacts due to drying could intensify the trend.

Transverse directions

Table 1 summarizes the comprehensive data set obtained for the radial and tangential directions. The careful sampling and experimental protocol provides data with low scattering regarding the small size of the samples, with the exception of opposite wood. As previously mentioned, the annual ring width is very small in the opposite wood region of the disk. The narrowing of the rings causes a greater heterogeneity of the samples in this zone and a greater difficulty to sample exactly in the desired area. In both directions, normal wood is more rigid than opposite wood and tension wood, which present similar stiffness values, regardless of contrasted anatomical features. For all types of wood, the anisotropy ratio approaches a factor of three. Contrary to what is often stated, the main reason for this anisotropy ratio is not the presence of ray cells, but the cellular pattern in the cross section, which presents walls aligned in the radial direction due to cambium activity. This statement was proved by homogenization results, in which the sole cell shape explained the observed anisotropy ratios (Perré 2002). The Poisson’s ratio ranges from 0.3 to 0.45 for the tangential tensile load and close to unity for the radial tests. This is consistent with a cellular and anisotropic medium, as was observed by others in spruce (Farruggia and Perré 2000).
Table 1

Summary of all data obtained for tensile tests in tangential and radial directions

Tangential

Radial

Sample number

MOE (MPa)

Mean (standard deviation)

Poisson’s ratio

Mean (standard deviation)

Sample number

MOE (MPa)

Mean (standard deviation)

Poisson’s ratio

Mean (standard deviation)

Normal wood

1

459

527 (66)

0.53

0.44 (0.056)

1

1,551

1,501 (75)

0.87

0.99 (0.087)

2

459

0.39

2

Rejected

Rejected

3

539

0.40

3

1,458

1.03

4

588

0.45

4

1,419

1.07

5

592

0.44

5

1,577

1.00

Opposite wood

1

Rejected

420 (55)

Rejected

0.42 (0.051)

1

1,457

1,170 (269)

0.81

0.86 (0.133)

2

361

0.45

2

1,188

0.92

3

481

0.37

3

887

0.87

4

389

0.48

4

1,413

0.68

5

448

0.39

5

906

1.04

Tension wood

1

497

432 (43)

0.31

0.32 (0.049)

1

1,022

1,144 (111)

0.96

0.97 (0.064)

2

379

0.28

2

1,174

1.01

3

Rejected

Rejected

3

1,030

1.00

4

436

0.25

4

1,277

1.06

5

401

0.34

5

1,004

1.03

6

448

0.4

6

1,290

0.91

7

395

0.35

7

1,200

0.89

8

470

0.3

8

1,083

0.98

9

Rejected

Rejected

9

1,213

0.88

Stiffness values determined at ca. 8 % MC

Table 2 summarizes all values obtained in the present work. The most apparent data set is the large MOE value obtained in the longitudinal direction for tension wood, 34,300 MPa, compared to 18,200 and 15,500 MPa obtained for normal wood and opposite wood, respectively. Even if this measured value remains smaller than the theoretical value of 45,100 MPa proposed by Yamamoto et al. (1992), it is twice that of normal wood, which is impressive.
Table 2

Young’s modulus and specific modulus of tension (T), opposite (O), and normal (N) poplar wood in radial (R), tangential (T), and longitudinal (L) directions

Wood type

MOE (MPa) (standard deviation)

Basic density (standard deviation)

Air-dried density (standard deviation)

Specific MOE (MPa.m3 kg−1)

T

R

L

(kg m−3)

(kg m−3)

T

R

L

Normal

527 (66)

1,501 (75)

18,200 (2,600)

469 (50)

554 (58)

1.12

3.2

38.8

Opposite

420 (55)

1,170 (269)

15,500 (2,000)

449 (42)

517 (45)

0.94

2.61

34.5

Tension

432 (43)

1,144 (111)

34,300 (1,400)

545 (21)

642 (37)

0.79

2.1

62.9

Stiffness values determined at ca. 8 % MC

Alternatively, the values measured in the transverse plane (R-T) are higher for normal wood, even though the differences are less contrasted than the longitudinal direction (roughly speaking, in the tangential direction, 500 MPa for normal wood and 400 MPa for tension and opposite wood, with stiffness being always larger by a factor of three in radial direction). The values for normal wood are in agreement with Brancheriau et al. (2002) in tangential direction (400–700 MPa with a density ranging from 460 to 690 kg/m3) and Watanabe et al. (2002) in the radial direction (1,200–1,500 MPa).

In the transverse plane, the smaller stiffness of opposite wood is explained by its smaller density, but this is not the case for tension wood, where a clear decrease in both radial and tangential stiffness values is observed in spite of a higher density (Fig. 10). This may result from two cumulative effects, the small stiffness value of the G-layer in the transverse plane as a result of the microfibril alignment along the longitudinal direction, and the lack of G-layer cohesion with the rest of the cell wall (Keller 1994; Clair et al. 2005). In this case, the real part of the tension wood participating in the mechanical strength would be thinner than the normal wood (i.e., S1 + S2 instead of S1 + S2 + S3) (Kollmann and Côté 1968; Yamamoto 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs00226-012-0511-x/MediaObjects/226_2012_511_Fig10_HTML.gif
Fig. 10

Averaged MOE values obtained experimentally plotted versus basic density (stiffness values determined at ca. 8 % MC)

The wood anisotropy is much higher in tension wood (EL/ET = 79 and EL/ER = 30) than in normal or opposite wood (EL/ET = 34 and 37 and EL/ER = 12 and 13, respectively). Indeed, the presence of the G-layer and its crystalline cellulose mainly oriented along the cell axis reinforce the structure in the longitudinal direction by a factor of 2, but weaken the structure in the transverse plane. Determining the stiffness of air-dried samples certainly emphasizes the weak cohesion of the G-layer with the rest of the cell wall.

Finally, it was interesting to compare the calculated specific MOE of tension wood (MOE divided by density) to the value obtained in the crystalline cellulose (134–137 GPa for a density equal to 1,500 kg/m3) (Marhofer et al. 1996; Sakudara et al. 1962). This ratio is indicative of tension wood being near 70 % the crystalline cellulose MOE, which is remarkable, even with a cellulose content of 80 % in the G-layer (Yamamoto 2004). The measured values tend to prove that nearly 100 % of the crystalline cellulose in the G-layer participates in the stiffness at the tissue level.

Conclusion

A full data set of MOE values determined on poplar wood samples was proposed under careful sampling conditions using two custom mechanical devices. All samples tested in this work were from the same tree, intentionally bent during growth to force production of tension wood in high quantity in a well-defined area. Three types of wood were distinguished (normal, tension, and opposite) and the designed device allowed the stiffness to be measured in the three material directions.

The results clearly emphasize the role of the G-layer present in tension wood, which reinforces the longitudinal stiffness, but weakens the structure in the transverse plane. The reported work should be considered as an experimental contribution to the understanding of wood behavior at the tissue level. Further efforts on compression tests for large deformation and modeling of the tissue behavior using the MPM (Material Point Method) are ongoing.

The authors believe that this data set might be useful for wood scientists, either as validation data to test theories regarding the relationships between ultrastructure features and cellular properties, or as input data of scaling approaches able to predict/explain the differences between the properties of the so-called “perfect” wood (cellular/annual ring level) and those of structural timbers.

Copyright information

© Springer-Verlag Berlin Heidelberg 2012