Wood Science and Technology 36 (2002) 487499 Springer-Verlag 2002 DOI 10.

1007/s00226-002-0151-7

Microfracture in wood monitored by confocal laser scanning microscopy

tze S. Aicher G. Dill-Langer S. Lu

487 Abstract The paper deals with the experimental characterisation of damage evolution within the radial (R)tangential (T) growth plane of softwood loaded in tension perpendicular to the grain. The reported investigations comprise in-situ monitoring of crack propagation by means of Confocal Laser Scanning Microscopy (CLSM) and evaluations of crack patterns of broken specimens. Three types of notched specimens, representing different crack propagation systems, were tested; for all congurations, both, loading and crack propagation direction were located within the RT plane of wood. The CLSM pictures of broken specimens show distinct differences among the regarded congurations with respect to crack paths. Two different damage mechanisms were identied being rupture of earlywood cell walls in the case of crack propagation in tangential direction and debonding of wood bers, i.e. rupture of the interface zone between adjacent tracheids, in case of crack progression in radial direction. In the case of an intermediate crack system with an angle of 45 between initial notch direction

Received 20 May 2000 G. Dill-Langer (&) Otto-Graf-Institute, University of Stuttgart, Pfaffenwaldring 4, 70569 Stuttgart, Germany e-mail: dill-langer@po.uni-stuttgart.de Tel: +49-711-6852280 tze S. Lu r Kunststoffkunde und Kunststoffpru fung, Institut fu University of Stuttgart, Pfaffenwaldring 32, 70569 Stuttgart, Germany e-mail: luetze@ikp.uni-stuttgart.de Tel.: +49-711-6853015 S. Aicher Otto-Graf-Institute, University of Stuttgart, Pfaffenwaldring 4, 70569 Stuttgart, Germany e-mail: simon.aicher@po.uni-stuttgart.de Tel.: +49-711-6852287 The nancial support of German Science Community (DFG) via grant to Sonderforschungsbereich 381 Characterisation of damage development in composite materials using nondestructive test methods and hereby to sub-project A8 Damage and NDT of the natural ber composite material wood is gratefully acknowledged.

and radial direction the crack evolution was monitored in-situ during the tension test, whereby the combined action of both basic fracture mechanisms was observed.

488

Introduction The natural ber composite wood exhibits strongly anisotropic properties with respect to stiffness and strength. Tension stresses from external loads or induced by transient ambient conditions such as moisture and temperature gradients are most unfavourable for the directions perpendicular to grain, i.e. perpendicular to direction of the wood bers. Crack initiation and propagation within the weakest plane of the material wood caused by tension stresses perpendicular to grain is therefore a major issue of research activities in the eld of wood (fracture) mechanics. Considerable efforts have been undertaken to obtain fracture mechanics parameters of RL and TL fracture congurations or crack systems, where crack propagation is parallel to the longitudinal (L), i.e. ber direction, while load direction resp. normal vector of crack surface is parallel to radial (R) or tangential m 1992). The present (T) growth direction. (e. g. Valentin et al. 1991; Bostro investigation, however, focuses on the characterisation of RT, TR and intermediate crack systems, whereby both, the load and the crack propagation directions are located within the end-grain face. The latter crack systems are of special practical relevance in the case of tension stresses perp. to grain in curved or tapered glulam beams. Earlier studies (Aicher et al. 1998, Dill-Langer et al. 1999) have shown, by means of numerical calculations and by empirical evidence that Mode I cracks in glulam due to tension perpendicular to grain are most often initiated near mid-width of the lamellas. This fact results from the pronounced stress peaks due to polar anisotropy within the RT-plane (Aicher and Dill-Langer 1996). The initial cracks most often propagate globally parallel to the wider sides of the lamellas towards the outer edges of the boards whereby the crack system is varying from RT (initial) to intermediate systems and in extreme cases nally to TR orientation. Complex macroscopic crack patterns were observed in the intermediate range where the annual ring structure obviously interacts with the propagating crack yielding in some cases crack-stop behaviour accompanied by global load recovery. Therefore the knowledge of crack-propagation properties as related to the meso-structure at annual ring level as well as to the micro-structure at tracheid level is an important prerequisite to assess the global damage behaviour of structural sized solid wood or glulam members loaded by transverse tension stresses. The inspection of fracture surfaces by means of microscopic imaging is a common method in wood science in order to characterise damage features and mechanisms. Most frequently fracture surfaces of failed specimens have been investigated by Scanning Electron Microscopy (SEM). Most literature known studies deal with tension parallel respectively perpendicular to grain; in the latter case predominantly RL and TL crack systems are regarded (e.g. Debaise et al. and Hanna 1983; Zink et al. 1994; Zimmermann et al. 1966; Borgin 1971; Cote 1994). However, the analysis of fracture surfaces provides only limited insight in the damage mechanisms. Therefore in some investigations specimens were not loaded completely until failure but to a certain stage of damage; then the inspection of the surface was performed (e. g. Job and Navi 1996). The monitoring of crack growth in-situ during application of the load, yielding the most complete representation of the damage mechanisms, is reported very scarcely in literature

(Mindess and Bentur 1986, Bodner et al. 1997, 1998). This may be explained by the specic risk of artefacts for in-situ SEM monitoring ( e. g. Dinwoodie 1966; Hoffmeyer and Hanna 1989) and by several technical problems related to the installation of an apt testing machine inside a vacuum chamber. In order to avoid the mentioned problems, but utilising the advantages of in-situ testing, a special kind of light microscopy, the confocal laser scanning microscopy (CLSM), was employed in the presented investigation.

Material, specimen built-up and preparation The selected softwood species was spruce (picea abies) being most widely used for structural load-bearing elements in European timber engineering. The tested material was clear wood, i.e. incorporated no visible defects as the microscopic investigations aimed at the typical behaviour of the basic composite built-up of wood without interference of growth bound defects such as knots or vessels. All specimens were cut from one board with an average density q12 of 500 kg/m3; the wood moisture content was 12%. The mean annual ring width varied between 1 mm for RT specimens to 2.5 mm for RT45 and TR-specimens due to different locations in the cross-section (the different specimen types are discussed below). The built-up of the necked specimen shown in Fig. 1 consisted of three parts being the notched spruce test volume and two lateral wooden on-gluings adhered to the centre piece by a one-component polyurethane adhesive. The ongluings, made of beech wood, were used in order to transfer the applied tension load to the test volume smoothly with minimised stress concentrations at the boundaries. The total specimen dimensions, length depth width, were 60 mm 4.2 mm 10 mm; two holes for load application with a diameter of 4 mm and a distance of 40 mm were drilled into the beech on-gluings. The length of the actual test volume of spruce wood was 12 mm; with respect to width the specimen was necked from 10 mm to a minimal width of 5 mm at the centre with a radius of curvature of 20 mm. The subsequently stated specimen preparation details have to be considered in connection with the aim of the presented investigation being the in-situ inspection of the crack-propagation within the radial-tangential (end-grain) plane, i.e. within the plane perpendicular to ber direction. Therefore the end-grain plane

489

Fig. 1. Built-up and dimensions of the investigated notched tension specimen with schematic view of the three regarded load/annual ring congurations

490

had to be prepared in a special way to yield a clearly cut surface of minimally deformed ber cross-sections. The difculties of a proper preparation of endgrain surfaces, especially of the earlywood part, arise from the built-up of the wood bers being hollow tubes with rather thin walls in the early- and transitionwood mainly consisting of the polycrystalline material cellulose. As the specimens had to be prepared for in-situ mechanical tests at ambient climate conditions no physical (shock freeze, water storage) or chemical treatment as recommended in literature for end grain cuts (e.g. Exley et al. 1977) could be adopted. As conventional cuts by means of commonly used microtome blades resulted in collapsed ber cross-sections a new cutting technique had to be developed, consisting of three steps as outlined following: the specimen was cut with respect to thickness by a conventional saw. The area at the centre (5 mm 5 mm) was left 0.1 mm thicker than the rest of the specimen, the projecting centre area of the specimen was equalised almost entirely using a conventional microtome blade, while the specimen was kept xed to the microtome apparatus, the conventional blade was changed to a special one-way blade, normally used for human bones for medical purposes, which was then used for the nal cut of the surface. A one-sided initial notch of 0.2 to 0.5 mm length was applied at mid-length of the specimens at the smallest cross-section of the necked part by means of a hand-held razor blade. In order to realise different congurations with respect to annual ring orientation and load direction three types of specimens as sketched in Fig. 1 were manufactured: conguration RT with radial (R) growth direction parallel load axis and initial notch parallel to tangential (T) growth direction, conguration TR with tangential growth direction parallel to axis of applied load and initial notch parallel to radial growth direction and conguration RT45 with an angle of nominally 45 between radial (resp. tangential) growth direction and load axis; initial notch at nominally 45 to radial direction. In total 30 specimens were manufactured, whereof 17, excelled by the best surface quality, were tested. The test series comprised 4 and 5 specimens with RT and TR congurations, respectively, and 8 specimens with RT45 conguration. The paper presents results of representative specimens: After description of the experimental procedure, the fracture paths of representative broken specimens of each conguration (RT, TR and RT45) are presented. Subsequently the in-situ monitored crack propagation of one representative RT45 specimen, not identic with the specimen discussed in the preceding section is presented and analysed.

Test equipment and experimental procedure The mechanical tests were performed by means of a small screw driven loading device adjusted for the use with a Confocal Laser Scanning Microscope (CLSM); the load of the machine was transferred to the specimen both sided pin-jointed via bolts. The loading procedure consisted in a stepped application of global deformation (cross head displacement) at a machine bound fairly high displacement rate of about 0.5 mm/sec within the displacement steps. The time lag

between two subsequent steps normally about 20 seconds was bound to the time necessary to adjust the microscope and to store the pictures of the previous load step. At certain load stages near ultimate failure the deformation was kept constant in order to observe visco-elastic crack growth behaviour. The major aim of the presented investigation was the damage monitoring resp. the observation of crack growth at a microscopic level in-situ, i.e. during load application. Thus a microscopic method had to be applied allowing observation and picture storage at sufcient resolution without inducing any additional damage or change of the wood properties. The Scanning Electron Microscopy (SEM) method widely used to characterise the microstructure of wood has the disadvantage that, besides conductive coating of the specimen, operation at vacuum conditions, i.e. at nearly zero percent of wood moisture content, is necessary. As all mechanical properties of wood are most sensitive to moisture content variations, tests at vacuum conditions are by no means representative for tests at normal conditions. Furthermore, the electron beam may eventually cause additional damage to the wood structure; beam damage induced artefacts in SEM imaging are reported in (Hoffmeyer and Hanna 1989). The Confocal Laser-Scanning Microscopy (CLSM), however, needs no specimen coating and works at ambient air pressure with normal relative humidity. Moreover, as no vacuum chamber is necessary, it is relatively easy to use an apt testing machine for in-situ experiments. Unlike normal light microscopy the CLSM method uses a laser beam which illuminates a small spot on the inspected surface. The laser wave is focused into a spot on the specimen. The reected light is then focused by an operationally separate optical system into a plane with an special lter (pinhole) onto a detector. Light reecting from other planes than the focal plane is discriminated. An image from the specimen is built up by scanning the spot in a square raster pattern (512*512 pixel). The axial and lateral resolution depends on the pinhole size, the objective and the wavelength of the laser. The main advantages of this technique are sharp pictures of rough surfaces and insensitivity vs. stray light. The principle of CLSM microscopy and some current applications in the frame of material science are reported in (Ritter et al. 1998). The specic microscope used in this investigations was of the type Zeiss LSM 410 being an invert type microscope with a very stable stand that allows to place the tension testing machine onto the microscope.

491

Crack patterns of broken specimens Figures 2a, b depict typical crack patterns of the RT and TR crack propagation systems, altogether with schematic views of the respective load/annual ring congurations. The photographs show a representative RT and TR specimen from the total of 4 resp. 5 specimens; the crack paths of all specimens within the respective conguration conformed well to the given pictures. In case of the RT conguration (Fig. 2a) the crack path completely within earlywood is generally parallel to the direction of the initial notch and perpendicular to load direction. However, in detail, the path consists of many steps with edges parallel and perpendicular to the main direction. There are parts of broken tracheids visible at the rough crack surfaces. Contrary, the crack path of the TR specimen (Fig. 2b) is nearly a straight line with only minor deviations from the direction perpendicular to the load. The latter holds true for both earlyand latewood. Besides one major bent in the crack propagation direction at the

492

Fig. 2a, b. Typical appearances of crack patterns of specimens with a RT conguration and b TR conguration

very beginning of the crack extension the crack surface is generally smooth with hardly any violations of cell walls. The different appearances of RT and TR crack surfaces and the underlying different damage mechanisms can be roughly described by a simplied morphological model. Hereby the softwood cross-section is represented by a twocomponent composite consisting of hollow bers with rectangular cross-sections bonded at the periphery with a very thin layer of matrix material to the adjacent bers. The approximation does not include effects of the cell wall ultrastructure, i.e. of its structural built-up consisting of outer (S1), main (S2) and inner (S3) cell wall layers and of their respective constituents such as microbrils and hemicellulose. In terms of ultrastructure the simplied two-component model merges the outer cell wall layer S1 with the middle lamella (quasi homogeneous matrix material) and the two inner cell wall layers (quasi homogeneous wall of the hollow bers). The actual slightly unsymmetric hexagonal ber cross-section is compatible with the conclusions from the approximate rectangular shape. Similar models of softwood cross-sections are discussed in (Persson 1997), whereby the anisotropy of elastic compliances in transverse direction could be derived from the structural or morphological built-up. Figure 3 gives a schematic representation of the two-component model of wood. The structural built-up at the regarded micro-structural level, represented by an assembly of a few tracheid cross-sections, can be compared to a brick wall: in radial direction the ber walls are smoothly aligned; in tangential direction, however, the cross-sections are shifted with respect to each other yielding a stepped path of cell walls. The damage mechanisms based on the described morphological micro-structure are also illustrated schematically in Fig. 3: Tension stresses acting parallel to the radial direction (RT crack system) result predominantly in cell wall rupture

493

Fig. 3a, b. Simplied morphological model of softwood cross-section and derived tension damage mechanisms for a RT conguration and b TR conguration

(Fig. 3a) with rough and stepped crack surfaces as obtained empirically (see Fig. 2a). Although both, matrix material and cell walls are damaged, the strength at this loading mode is governed by cell wall thickness so that RT crack paths tend to run solely within the thin-walled earlywood. Contrary, tension stresses in tangential direction (TR crack system) yield mainly inter-cellular damage (ber debonding) with intact cell wall material and smooth crack surfaces as shown in Fig. 3b, compare also Fig. 2b. Thus, tension strength due to ber debonding mechanism may be assumed almost independent of cell wall thickness. In case of intermediate crack conguration RT 45, as for RT and TR systems, one exemplary specimen is shown subsequently which represents quite well the qualitative behaviour of all 8 tested specimens. The crack path of the RT45 specimen, shown in Fig. 4, exhibits a zigzag behaviour. First within earlywood the crack follows the direction of the initial notch perpendicular to load direction. Thereby the crack surface consists of both, ruptured and debonded cell walls. The more the crack approaches the earlywood/latewood transition interface the more it deviates from the initial direction until propagation coincides with the nominally 45-angle of the transition interface. At this stage the crack surface consists predominantly of ruptured cell walls comparable to the surface of RT conguration. Having reached half the specimen width the crack turns 90 anticlockwise propagating through late- and transition wood. Thereby the crack surface is smooth as in case of the pure TR conguration caused by the debonding of adjacent bers with hardly any cell wall damage. The close-up in Fig. 4 shows the crack path in the vicinity of the 90 crack turn revealing in a very transparent manner insight in the features of the described different damage mechanisms. Thus, three stages of crack propagation were identied in case of RT45specimens: Firstly combined action of damage mechanisms within early wood, secondly pure cell wall rupture near the earlywood latewood interface within the early wood section and nally predominantly ber debonding within the latewood layer.

494

Fig. 4. Crack path of a specimen with RT45 conguration with close-up of crack propagation at the growth-ring transition

In-situ monitoring of crack growth First, it should be stated that stable crack growth is prerequisite for in-situ monitoring of damage evolution. The in-situ CLSM observation in order to monitor the crack propagation during load application was done for all specimens of the three respective crack systems. For TR conguration, in case of all ve specimens, brittle behaviour was observed without visible stages of stable crack propagation. Similar ndings were obtained in case of RT conguration; for all 4 specimens only a very short period of stable crack growth in which the crack propagated about 0.1 mm could be monitored. Compared to RT and TR crack systems the RT45 conguration showed the most stable behaviour. This is presented in detail following, also revealing a correlation between microscopically observed damage evolution and global stress-strain behaviour. In total eight tests with RT45 specimens were performed whereby in two cases the tip of the initial notch ended in the latewood region. Those two specimens yielded unstable behaviour, i.e. crack opening without propagation and subsequent brittle failure. All six specimens with initial notches ending in the early-

wood yielded stable crack propagation until about 50% of specimen width; thus stable crack growth mainly within earlywood could be monitored in-situ until the occurrence of ultimate brittle failure related to unstable crack growth. In order to observe crack growth at cellular level a magnication factor of the microscope objective of about 20 is necessary. As the observation range of the used microscope is hereby limited to an area of about 0.5 0.5 mm2, the imaging window had to be shifted continuously in order to follow the crack tip within the deforming specimen. Therefore the microscopic pictures discussed below do not show the crack development from a constant view point but show detailed closeups of respective damage propagation steps; in order to enable spatial correlation among the pictures of each respective series, the pictures contain an orienting xed point, marked by a cross. The global stress strain curve of the exemplarily discussed RT45 specimen with a nominally 45 initial notch is shown in Fig. 5; additionally numbers denote the CLSM pictures taken at the respective strain stages given by dotted vertical lines. A selection of these pictures related to noteworthy steps within the damage evolution is shown in Figs. 6 a-d and Figs. 7a-d. Figures 6a-d give a survey of the rst stages of crack propagation. Figure 6a, showing picture No. 1, as indicated in Fig. 5, graphs the tip of the unloaded initial notch (marked by an arrow) which can be described as blunted; the latter statement is related to the presented microscopical level. Additionally the load conguration is sketched in Fig. 6a in an inserted drawing. During loading of the specimen to a deformation of about 90% of global failure deformation the stress strain curve exhibits a rather linear behaviour (Fig. 5). The slight tendency towards a global stiffness increase stems from the local non-linear compression behaviour at the load-application zones in beech wood remote from the actual test volume. The small unsteady deviations of the stress strain curve from steady behaviour result from stepwise application of the deformation load and subsequent stress relaxation of the test volume. It is important to note that within the linear stress strain domain until about 90% of ultimate strain the crack is opening but not propagating.

495

Fig. 5. Global stress-strain curve of representative RT45 specimen altogether with numbers of respective microscopic pictures; the strain value of the individual picture is marked by a dotted line

496

Fig. 6ad. Microscopic pictures 1, 12, 15 and 22 of representative RT45 specimen taken in-situ during loading (see Fig. 5); the cross marks a common xed point: a tip of the initial notch (marked by an arrow) before loading (picture 1); b, c, d progress of crack growth (pictures 12, 15, 22)

The rst signicant load drop of about 10% below the preceding stress level is accompanied by crack propagation as observed in picture No. 12, here shown in Fig. 6b; a combined damage of ber walls and ber debonding occurred. The crack at this stage is propagating parallel to the initial notch direction (very roughly at 45 with respect to the radial growth direction). Further crack propagation was observed in pictures 13 to 22 (pictures 1222 are termed series 1) although the applied deformation was held constant for about two minutes. The viscoelastic crack growth, as it is usually termed at the macroscopic level is exemplarily shown by pictures 15 and 22 (i.e. Figs. 6c and 6d) which have to be compared to the stage directly after stop of loading depicted in Fig. 6b. Both, cell wall rupture and damage of matrix interface between cell walls can be detected. A surprising observation was the absence of additional micro-cracks or large deformations in the vicinity of the main crack tip. After the pre-peak load drop the stress strain curve shows a load recovery for two more load steps. At 98% of peak load level the deformation is held constant again for about four minutes. During this time viscoelastic crack growth continues until ultimate failure which can be termed as creep rupture. The crack behaviour in this stage is documented in pictures 26 to 48 summarised as series 2. Figures 7a-d (pictures 33, 44, 47 and 48 of series 2) show the last stages of

497

Fig. 7ad. Microscopic pictures 33, 44, 47 and 48 of representative RT45 specimen taken in-situ during loading (see Fig. 5) at the last stages of stable crack propagation

crack propagation which could be observed by means of CLSM. While pictures 33 and 44 (Figs. 7a and b) represent mixed damage mechanisms and crack propagation with an angle of about 45 vs. radial direction, the last two pictures 47 and 48 (Figs. 7c and 7d) show a signicant deviation from the initial propagation direction: the crack tends to follow the radial growth direction. At this last stage of stable damage propagation the crack growth is governed by the ber debonding damage mechanism. Due to the increasing thickness of cell walls in the approached area of transition wood the ber walls exhibit higher resistance vs. the applied load and consequently the matrix interface between the cell walls becomes the weakest link of the composite structure. This leads to a deviation of the crack path now no more perpendicular to the applied resp. maximum principal stress. The delamination or debonding failure results in a more brittle behaviour with unstable crack propagation which could only be monitored at the presented early stage directly in advance of ultimate global failure of the specimen. In the fracture observations of series 2 (pictures No. 2648) equally as in case of series 1 no additional micro-cracks exceeding the length of a few cell diameters have been observed in the vicinity of the major crack. Moreover this statement can be generalised for all in-situ tests performed: there was no evidence of a zone of micro-cracks larger than about 100 lm.

498

Conclusions Crack patterns and crack growth within the radial-tangential growth plane of spruce wood bound to in-plane tension loading at different angles were monitored at microscopic level by means of Confocal Laser Scanning Microscopy. The observations revealed that the micro-structure of wood (here considered at a level of several tracheid diameters) as well as the meso-structure at annual ring level have to be recognised for a proper description and understanding of damage evolution in case of the investigated loading mode. Two different damage mechanisms, being cell wall rupture and ber debonding, were identied. Cell wall rupture was found to be the predominant mechanism in case of an RT crack system. The occurrence of this mechanism is bound to the meso structure related to annual ring level because of its dependency on cell wall thickness. The ber debonding mechanism was observed mainly in case of the TR crack system where the rather straight radial matrix bands are ideal delamination on-sets. This delamination type rupture is very brittle so that the resulting unstable crack propagation could be monitored in-situ during loading solely at the very rst stage and thus was essentially identied after ultimate failure. The intermediate crack system, RT45, showed a complex fracture behaviour with zigzag crack paths and different damage mechanisms at different stages of the damage history. The crack evolution of the RT45 system could be monitored in-situ and correlated to global stress-strain behaviour during periods of stable crack growth where frequent interactions of both described basic damage mechanisms occurred. The results of the presented study on phenomenological features of fracture will serve as a basis for mechanism based damage modelling approaches for tension perpendicular to grain. The separation of damage mechanisms and the used simplied morphological model are a prerequisite for implementation of an anisotropic damage law into, e. g. non-linear continuum mechanics type models. Thereby the empiric ndings provide reasons for the following assumptions on tension strength perpendicular to grain within one growth ring: tension strength in radial direction is assumed to vary proportional to density and tension strength in tangential direction may be taken as a constant. Furthermore the crack paths and crack propagation recordings of RT45 specimens can serve as calibration test results for modelling results. References
Aicher S, Dill-Langer G, Ranta-Maunus A (1998) Duration of load effect in tension perpendicular to the grain of glulam in different climates. Holz Roh- Werkstoff 56: 279284 Aicher S, Dill-Langer G (1996) Inuence of the cylindrical anisotropy of wood and loading conditions on off-axis stiffness and stresses of a board in tension perpendicular to grain. Otto-Graf-Journal, 7: 216242, Otto-Graf-Institute, University of Stuttgart ll G (1997) Fracture initiation and progress in wood specimens Bodner J, Schlag MG, Gru stressed in tension. Part I: Clear wood specimens stressed parallel to the grain. Holzforschung 51: 479484, Part II: Compression wood specimens stressed parallel to the grain. Holzforschung 51: 571576 ll G (1998) Fracture initiation and progress in wood specimens Bodner J, Schlag, MG, Gru stressed in tension. Part III: Clear wood specimens with various slopes of grain. Holzforschung 52: 95101 Borgin K (1971) The cohesive failure of wood studied with the scanning electron microscope. J. Microscopy 94: 111 m L (1992) Method for determination of the softening behaviour of wood and the Bostro applicability of a non-linear fracture mechanics model. Report TVBM-1012, Div. of Building Mat., Lund University

 WA, Hanna RB (1983) Ultrastructural characteristics of wood fracture surfaces. Wood Cote and Fiber Sci 15: 135163 Debaise GR, Porter AW, Pentoney RE (1966) Morphology and mechanics of wood fracture. Materials Research & Standards 6: 493499 fin L, Aicher S (1999) Fracture of solid wood and glulam in tension Dill-Langer G, Ho perpendicular to the grain monitored by acoustic emission. Proc. 1st RILEM Symp. Timber Eng.: 161170, Stockholm, Sweden Dinwoodie JM (1966) Induction of cell wall dislocations (slip planes) during the preparation of microscope sections of wood. Nature 212: 525527 Exley RR, Meylan BA, Buttereld BG (1977) A technique for obtaining clean cut surfaces on wood samples prepared for the scanning electron microscope. J. Microscopy 110: 7578 Hoffmeyer P, Hanna RB (1989) Electron beam damage during testing of wood in the SEM. Wood Sci. Technol. 23: 211214 Job L, Navi P (1996) Microscopic analysis of crack propagation in softwood, mode I and II. Proc. Int. Conf. on Wood Mechanics: 541554, Stuttgart, Germany Mindess S, Bentur A (1986) Crack propagation in notched wood specimens with different grain orientations. Wood Sci Technol 20: 145155 Persson K (1997) Modelling of wood properties by a micromechanical approach. PhD thesis, Lund University, Lund, Sweden Ritter S, Busse G, Press P (1998) In-situ observation of the stress eld of the failure process in single bre reinforced polymers. Rev. Progress Quantitative Nondestructive Evaluation 17B: 12011208 m L, Gustafsson PJ, Ranta-Maunus A, Gowda S (1991) Application of Valentin G, Bostro fracture mechanics to timber structures RILEM state of the art report. Research note 1262, VTT, Finland Zimmermann T, Sell J, Eckstein D (1994) Rasterelektronenmikroskopische Untersuchun chen von Fichtenholz. Holz Roh- Werkstoff 52: 223229 gen an Zugbrucha Zink AG, Pellicane PJ, Shuler CE (1994) Ultrastructural analysis of softwood fracture surfaces. Wood Sci. Technol 28: 329338

499