FRACTURE OF COMPOSITES AND WOOD-ADHESIVE JOINTS: A COMPARATIVE REVIEW Michael P.C. Conrad Graduate Research Assistant Department of Metals and Materials Engineering Gregory D. Smith² Assistant Professor Department of Wood Science

and GoÈran Fernlund Assistant Professor Department of Metals and Materials Engineering The University of British Columbia Vancouver, BC Canada V6T 1Z4 (Received April 2002)

ABSTRACT This paper presents a comparative review of the literature on fracture of wood composites and wood-adhesive joints. Wood-adhesive joints are two or more adherends joined with resin in a speci®c geometry. Wood composites can be considered an amalgamation of wood-adhesive joints, and the effect of geometry, adherends, and resin on fracture toughness for wood-adhesive joints is similar to that of wood composites. Parameters that have similar effects on the two systems are: slenderness ratio, a purely geometrical consideration; adherend surface preparation; and bond-line uniformity, where ideally, the bond-line is self-similar along the entire length. Applicability of fracture mechanics is also demonstrated as models based on the concept of intrinsic ¯aws have been found to adequately predict fracture behavior of these composites. Keywords: Fracture mechanics, wood/adhesive joints, wood composites, resin dispersion, resin dis- tribution, oriented strandboard.

INTRODUCTION amalgamation of wood-adhesive joints. As Over the past decades, the use of wood one moves to larger and larger furnish, from composites has increased. While the literature particles to strands to veneer, only the length on fracture of solid wood is extensive, limited to thickness ratio of the furnish changes, and research has been undertaken on fracture of each composite is essentially a series of ad- wood composites. The fracture behavior of hesive joints. Therefore, an examination of the solid wood is important when studying wood structural adhesive joint literature, which is re- composites as wood composites are identical lated to the furnish-resin (meso) scale of wood to solid wood up to the cellular level, with the composites, will allow greater understanding only difference being that at larger scales, the of the behavior of wood composites at the wood-adhesive bonds add further complexity macroscopic scale. to the wood composite not present in solid The literature can be divided into wood-ad- wood. Wood composites can be considered an hesive research and wood-composite research, with the majority of studies concentrating on ² Member of SWST and corresponding author. either the geometry of the joint or composite,

Wood and Fiber Science, 36(1), 2004, pp. 26±39 ᭧ 2004 by the Society of Wood Science and Technology Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 27 effects associated with the adherend, or effects At the macroscopic scale, similar observa- associated with the adhesive. Table 1 lists the tions are made. Lei and Wilson (1979) show wood-adhesive joint fracture toughness ge- that there is a reduction in fracture toughness ometries, adhesives, species, and effects ex- with increasing veneer thickness, a result also amined in the literature. Table 2 lists the rel- corroborated by Jung and Murphy (1983). Lei evant literature for wood composites. and Wilson also found that for LVL construct- The resin system used in wood composites ed with very thin veneers, 0.8 mm (1/32 in.), is typically phenol-formaldehyde (Barnes the directionality of solid wood disappears for 2000, 2001), but isocyanate resins have also the RL, TL, TR, or RT crack directions, and been studied (Lei and Wilson 1979, 1980; therefore crack direction has no effect on the Youngquist et al. 1987; Kamke et al. 1996; fracture toughness. This ®nding should be ap- Furuno et al. 1983). plicable to other wood composites since the Literature on fracture of wood composites thickness of the veneer is similar to that of relating to joint geometry, adherend character- OSB strands. Thus, the fracture toughness of istics, and resin will now be discussed in more OSB should not be affected by the orthotropic detail. The literature will be discussed in re- nature of solid wood. lation to the behavior of structural adhesive Of the two remaining dimensions for ad- joints, and the ®ndings summarized. Note that hesive joints, width and length, the one typi- this is not an exhaustive but a select review cally studied is lap length. Results of Leicester that compares reported fracture behavior of (1974), Komatsu et al. (1976), and Walsh et wood composites and wood-adhesive joints. al. (1973) indicate that the failure load of lap and strap joints increases with lap length. SPECIMEN GEOMETRY However, as shown in Fig. 3, increasing the overlap length leads to diminishing returns The strength and fracture toughness of and the normalized failure load plateaus above wood composites can be more easily inter- a characteristic length, for a given system. The preted by drawing on the large body of re- plot of normalized failure load, P␣/2Y, versus search on structural adhesive joint behavior. normalized lap length, 0.5 ␣L, in Fig. 3 allows Examples of common structural adhesive a dimensionless comparison of load and lap joints are shown in Fig. 1. Geometric param- length. eters examined in the literature include the ef- Komatsu (1984) examined Mode III frac- fect of lamination or veneer thickness and lap ture toughness by bonding lap shear samples length. at varying angles and loading the samples in An examination of the adhesive lap and compression. The ratio of Mode II to Mode strap joint literature reveals that a reduction in III is a function of the angle between the ad- lamination thickness reduces both the shear herends with a signi®cant Mode II component and peel stresses at the end of the bonded at large angles (pure Mode II at 180Њ). The overlap (Tong and Steven 1999). For wood- GIIIc value decreased as the angle increased adhesive joints, it has similarly been found from 90Њ to 180Њ. The fracture pattern also that a decrease in lamination thickness leads changed with lap angle. At low lap angles (90Њ to an increase in fracture toughness and failure and 120Њ), fracture was simple and brittle, oc- stress (Leicester 1973; Komatsu et al. 1976; curring through fracture in rolling shear at the Walsh et al. 1973; Leicester and Bunker 1969; glue lines. At angles of 150Њ and above, frac- Jung and Murphy 1983; River et al. 1989; ture was more complex, with some samples Scott et al. 1992). The effect of lamination failing within the wooden members them- thickness on failure stress is shown in Fig. 2, selves. with suf®ciently thin laminations having frac- For modelling purposes, it is often easier to ture loads similar to that of defect-free wood. describe the properties of wood composites in 28 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1) , Ebewele ϭ ϭ : (5, 6) LT Jung and Murphy P, P, DF: (9) P, DF: (9) ϭ Lawson Cypress, M ϭ Ebewele et al. 1979; 16 ϭ Sasaki et al. 1973; 8 : (11) ϭ BL C, M, East Indian Kauri, LC ϭ : River and Okkonen 1993; 15 BL, SC ϭ Leicester and Bunker, 1969; 7 Douglas , EIk ϭ D, T, (7) ϭ Scott et al. 1992. ϭ urea-formaldehyde. ϭ : (8) EIk, Cp, SIa, BT, Dillenia, DF River et al. 1989; 14 Resin ϭ ϭ Leicester, 1973; 6 : (21) ϭ LT, SC LT : (15, 17) : (16, 18, 20) DF, A: (14) M: (14±20) Bi, BL SC Campnosperma, D Mijovic and Koutsky 1979; 21 ϭ ϭ resorcinol-formaldehyde, UF ²: (1)³ Walsh et al. 1973; 5 ϭ : (12) Bi: (13) ϭ Yellow Poplar. BL ϭ BL Gagliano and Frazier 2001; 13 Cedar, Cp surface characteristics. ϭ ϭ ϭ YP, SC Taun, YP resorcinol, RF : ϭ ϭ Komatsu 1984; 4 : (2) : (2) ϭ Ebewele et al. 1986b; 20 BL, SC ϭ : (3) LT, LC LT, LC Brown Terminalia, C D, T, (7) LC ϭ White and Green 1980; 12 lamination thickness, LC, P, ϭ ϭ phenol-resorcinol, R ϭ LT LT, Solomon Island Albizia, T Komatsu et al. 1976; 3 ϭ ϭ Blackwood, BT Ebewele et al. 1986a; 19 C* E P PR R RF UF N/A ϭ ϭ : (4) : (10) White 1976; 11 phenolic, PR ϭ ϭ LC BL lap characteristics, Shinanoki, SIa ϭ , Bk ϭ ϭ epoxy, P LC ϭ , Bi bond-line, Ebewele et al. 1982; 18 Shimizu and Okuma 1981; 2 casein, E ϭ Quandong, Sh ϭ ϭ Wood-adhesive fracture specimen geometries, adhesives, species, and effects examined in the literature. ϭ ϭ ϭ BL Smith and Penney 1980; 10 1. ϭ Specimen geometry , Q lever Beam (CDCB) (DCB) ϭ ABLE ² Effects: * Resins: C ** Species: A ³ Authors: 1 Butt Elk, Cp, SIa, BT, Contoured Double Canti- Compact Tension (CT)Double Cantilever Beam P, Four Point Bend Butt Double Strap P, Bi, Q, Bk, Double Lap LC, T Single Lap Sh**, et al. 1980; 17 P 1983; 9 Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 29 Furuno Simpson ϭ ϭ : (11) : (4) D D (MDF) ®berboard U, Medium density Barnes 2001; 17 ϭ Kamke et al. 1996; 9 ϭ (PB) DF: (10) DF: (10) Particleboard Higgins 1990; 8 Laufenberg 1984; 16 ϭ ϭ : (2) ␳ : (16) : (17) board (OSB) TS BL, : (5, 8, 9) : (5, 8) : (15) Oriented strand : (5) : (5, 13) A: (6, 7) Ad, Bi, C, P: (7) ␳ ␳ BL BL A: (7, 15) Bi: (7, 16) DF: (16) Ad, C, P: (7) BL LC, LT DF: (5, 13) A: (6) U, Composite Youngquist et al. 1987; 7 ϭ River and Okkonen 1993; 15 ϭ unknown. : (3) U, D ϭ : (5) DF: (5) tensile strength. ␳ ϭ (5) : (5, 12) , U TS (LVL) ␳ BL, Laminated veneer ϭ Lei and Wilson 1980; 6 ϭ BL, D, LT density, Lei and Wilson 1981; 14 Pine, S ϭ ϭ ␳ϭ Sato 1988a; 5 , P ϭ ϭ **: (1)² C, D lamination thickness, Glued laminated timbers (Glulam) ϭ Lei and Wilson 1979; 13 , L LT ϭ ϭ Mihashi and Hoshino 1989; 4 Cedar, DF ϭ ϭ Sato 1988b; 12 lap characteristics, ϭ ϭ Birch, C LC ϭ Barnes 2000; 3 defects, ϭ ϭ , Bi D ϭ Specimen geometry bond-line, Ilcewicz and Wilson 1981; 11 Aspen, Ad Murphy 1986; 2 ϭ ϭ ϭ ϭ Wood composite fracture toughness specimen geometries, species, and effects examined in the literature. BL 2. ABLE ** Effects: * Species: A ² Authors: 1 Single Edge Notch (SEN) DF: (12) Compact Tension Shear (CTS) Contoured Double Cantilever BeamTension (CDCB) U: (14) P: (14) T Three Point Bend DF*, Compact Tension (CT)Internal Bond (IB) DF, C, S, L, DF, 1977. et al. 1983; 10 Unknown 30 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1)

FIG. 2. Effect of lamination thickness on failure load for laminated pine containing butt joints. ⅜ ϭ stress at ultimate load (psi), ⅷ ϭ stress at fracture (psi). Adapted from Leicester (1973).

FIG. 1. Some common structural bonded lap joints (a± Simpson Model: i) and butt joints ( j±l ). (Tong and Steven 1999) ␴ ϩ ␴ϭ w(r k) R , (3) terms of the ratio of furnish length to thick- r ϩ ku ness. This ratio is referred to as the slender- l ness ratio. Barnes (2001), Eqs. (1) and (2), and r ϭ , (4) t Simpson (1977), Eqs. (3) to (5), developed ␴ models based on the inherent properties of ϭ w wood and the slenderness ratio. Barnes uses a u ␶ , (5) modi®ed Hankinson equation, whereas Simp- ␴ son's model accounts for the shear strength of where R, is the tensile strength of the OSB, ␴ the adhesive used. (It should be noted that the w, the tensile strength of the strand in the di- ␶ symbols for the variables within these and rection of orientation, , the shear strength of subsequent equations have been changed for the adhesive bond between ¯akes, r, the slen- ease of comparison.) derness ratio, l, the ¯ake length, t, the ¯ake thickness, and k, the proportionality constant Barnes Model: relating the forces for tensile and shear failure

␴ϫ␴࿣ ⊥ ␴ϭ , R ␴ nnϩ␴ ࿣sin [arctan(2tb/l)] ⊥cos [arctan(2tb/l)] (1) ␳ ϭ ϫ a tbat ΂΃␳ , (2) b ␴ ␴ where R is the resultant strength (psi), ࿣, the strength parallel to the grain (psi), ␴⊥, the strength perpendicular to the grain (psi), n, the experimentally determined coef®cient (0.9 to 1.5), l, the length of strand (in), t , the initial FIG. 3. Effect of lap length on failure load for a double a strap butt joint where P ϭ failure load, ␣ϭf(geometry strand thickness (in), tb, the in situ strand ϭ ␳ and material properties), Y maximum allowable shear thickness (in), a, the initial wood density (lb/ stress, L ϭ lap length. Adapted from Tong and Steven 3 ␳ 3 ft ), and b, the product wood density (lb/ft ). (1999). Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 31

of the ¯akes to the number of strands that fail by each method (assumed to be 1). Both models predict that at higher slender- ness ratios the strength properties: modulus of elasticity (MOE) and modulus of rupture (MOR) in the Barnes model, and tensile strength in the Simpson model, increase. In- creasing the slenderness ratio of the furnish is analogous to moving from OSB to Parallam௢.

This supports the concept of a wood compos- FIG. 4. Effect of board density on Mode I fracture ite being an amalgamation of wood-adhesive toughness (resin spread rate ϭ 1.350 lb/1000 ft2). Adapted joints, as the trends of increasing failure load from Lei and Wilson (1980). and stiffness with decreasing thickness and in- creasing length are also seen at the wood-ad-

hesive scale. However, the Simpson model has ao, the intrinsic ¯aw size (in), Y(ao/W), the only been compared qualitatively to data in the fracture toughness geometry factor, W, the literature, as the author states that it would be width of the specimen (in), ␭, the character- dif®cult to test quantitatively. istic dimension of the material, particle thick- ␴ As has been shown, models based on wood- ness (in), c, the intrinsic strength (psi), and k, adhesive joint geometry have been used to the stress concentration factor equal to 0.73. predict the strength of wood composites. Mod- Applicability of the intrinsic ¯aw concept els based on fracture mechanics and intrinsic for wood composites is con®rmed by the work ¯aws have also been proposed. Ilcewicz and of Lei and Wilson (1980, 1981), who studied Wilson (1981) studied the fracture mechanics the fracture toughness of oriented ¯akeboard of particleboard using a nonlocal theory. Un- in the TL, RT, RL, and TR directions. A TL like theories based on continuum mechanics, ¯akeboard is equivalent to solid wood with the which considers only the behavior at a point crack propagating in the TL direction that has (locally), nonlocal theories take into account been cut into ¯akes and re-assembled. Lei and the behavior of a region when predicting fail- Wilson found that the fracture toughness is af- ure. The results from Ilcewicz and Wilson fected by inter¯ake void size and board den- (1981) show that the fracture toughness of par- sity and is unaffected by direction and the ticleboard can be predicted by Eqs. (6) and amount of resin applied to the ¯ake. The fre- (7), which includes the intrinsic strength, in- quency of inter¯ake voids re¯ects the unifor- trinsic ¯aw size, and a characteristic dimen- mity of the bond-line, with the voids them- sion of the material. For the speci®c example selves acting as sites for crack initiation. It is modelled by Ilcewicz and Wilson, particle- assumed that the size of inter¯ake voids will boards with a speci®c gravity of 0.70, this decrease with increasing board density and characteristic dimension is equal to the particle that LVL represents perfectly bonded OSB, thickness. Increased particle thickness at this e.g., OSB with no inter¯ake voids as the bond- speci®c gravity gives increased fracture tough- line is continuous rather than formed of dis- ness. crete spots. Therefore, the fracture toughness ϭ␴ ͙ of LVL can be taken as an upper bound to that Klc IBaY o(a o/W), (6) of OSB. Thus, one would expect fracture ␭ toughness to decrease as the number of inter- ␴ 22a ϭ␴, (7) IB o2k2 c ¯ake voids increases in agreement with work by Lei and Wilson (1980), as shown in Fig. 4.

where KIc, is the Mode I fracture toughness Lei and Wilson (1981) proposed a model 1/2 ␴ (psi´in ), IB, the internal bond strength (psi), for the prediction of fracture toughness of ori- 32 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1)

FIG. 5. Predicted and experimental retention of Mode I fracture toughness. Calculations for expected crack length are based on inter¯ake void length plus nonbonded length. Adapted from Lei and Wilson (1981). ented ¯akeboard that includes a term to ac- count for the in¯uence of inter¯ake voids, Eqs. (8) and (9). This model assumes that an FIG. 6. Schematic of Type A, crack lies across the induced crack will be extended by an existing veneers, and Type P, crack lies parallel to the veneers, fracture geometries. Adapted from Mihashi and Hoshino void or nonbonded region, and thus predicts (1989). the propagation fracture toughness of an es- tablished crack rather than the critical stress intensity factor required for crack initiation. where KЈ , is the fracture toughness of ori- Lei and Wilson have shown that fracture Ic ented ¯akeboard, K , the fracture toughness of toughness of the panel is affected by the av- Ic wood used to make strands, ⍀, the average erage size of the inherent ¯aws in the wood, size of inherent ¯aws, a, the initial crack ⍀, which is a material property, and the av- length for an edge-notched specimen, ⌬a, the erage void length, 1/␮, which is affected by expected increase in crack length due to inter- resination and other processing parameters. ¯ake voids, Y(a/W), the geometry factor for They combined these in a proportionality con- edge-notched specimen, 1/␮, the average void stant, e(⍀Ϫ1/␮). This constant accounts for the length, and 1/␭, the average distance between possibility of producing a composite that has voids. fracture toughness greater than that of clear, A comparison of model predictions and ex- solid wood. One feature of the model is that perimental data is shown in Fig. 5. The model it predicts fracture toughness identical to that predicts a K value equivalent to that of solid of the solid wood when the average void Ic Douglas-®r at an expected crack length of 2.5 length is set equal to the average inherent ¯aw mm (0.1Љ), in agreement with the measured size of solid wood. However, note that as the ¯aw size for Douglas-®r of 2.5 to 3.8 mm average inherent ¯aw size increases, an in- (0.1Љ to 0.15Љ) reported by Schniewind and crease in board fracture toughness over that of Lyon (1973). solid wood is predicted given a constant void Mihashi and Hoshino (1989) found that length. fracture mechanics, provided one used the a nonlinear J-integral method, accurately pre- aY1/2 ΂΃ Ј W dicted the experimental results on LVL. Linear K Ic ⍀Ϫ ␮ ϭ [e( 1/ )] , (8) fracture mechanics, KIc and GIc, underestimat- KIc a ϩ⌬a ed the fracture toughness. Two fracture ge- (a ϩ⌬a)1/2Y ΂΃W ometries were studied: one where the crack lies across (A) the veneers, and one where the 1 ␭ ⌬a ϭ , (9) crack runs parallel (P) to the veneers, both ␮␮ϩ␭΂΃ shown schematically in Fig. 6. They found Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 33

TABLE 3. Failure mechanisms in randomly (Laufenberg 1984).

Orientation of Failure type strand with respect Failure pattern Characteristics (%) to loading axis Transverse/Shear Failure along wood ®ber 81 10Њ to 90Њ Rolling Shear Fracture perpendicular to principal 670Њ to 90Њ stress, rotation of prismatic cross section of wood strand Tensile Long splintered strand end, wood ®- 5 Ͻ12Њ ber nearly parallel to load direction Disbonding Low strand-to-strand bond strength or 8 Ͻ45Њ high strand strength that the tensile strength of type A specimens creased with increasing angle of the strands to was less than that of type P. However, the frac- the applied load for both parallel (unidirec- ture toughness of type P specimens was less tional) and cross-angled (cross-ply) compos- than for type A. The type A specimen is an ites. The effect was less for cross-angled prod- example of crack divider geometry where the ucts due to a change in failure mode. specimen acts as a series of thin -stress Laufenberg also con®rmed the effect of dis- samples rather than one thick plane-strain bond length, which includes all bond-line fail- sample (Hertzberg 1996). Since fracture ures, delaminations, and any area that may not toughness is higher in plane stress, the type A have been in contact with other ¯akes, inter- specimen will exhibit greater fracture tough- ¯ake voids. The tensile strength is seen to de- ness than type P. crease with longer disbond lengths. Since ad- However, the applicability of fracture me- hesive failure at the bond-line between the res- chanics to wood composites is not universal. in and the wood typically occurs at a lower Sato (1988a, b) examined the Mode I fracture strength than cohesive failure in either the toughness using a compact tension (CT) spec- wood or resin and longer disbond lengths cor- imen and the mixed mode fracture of medium respond to an increase in the amount of bond- density ®berboard (MDF) and found that the line failure, the tensile strength should neces- fracture toughness decreased for deeper, sharp- sarily decrease. Longer disbond lengths can er notches. Fracture toughness, Kc, was also also represent an increase in the size of inter- found to increase with specimen width. Since ¯ake voids, which must decrease the tensile fracture toughness is a material property, there strength since these are zero strength regions. should be no size effect associated with notch Laufenberg found that the orthotropic failure depth. Therefore, based on this, the use of lin- criteria: the maximum stress criterion, the ear elastic fracture mechanics is suspect in the Tsai-Hill criterion, and Hankinson's Formula, case of MDF. provide reasonable upper-bound estimates of A better understanding of fracture at the the panel strength. Laufenberg attributed the macroscopic level can be obtained by exam- deviation between predicted and measured ining the underlying micromechanisms of strength to bond-line failures and recommends fracture. Laufenberg (1984) studied the frac- that the in¯uence of bond quality should be ture surface of OSB tested in tension, and examined using a fracture mechanics ap- found that the strands failed in four distinctive proach. patterns, listed in Table 3. Examination of the Work on specimen geometry for wood com- table reveals that the type of failure is related posites leads to the following conclusions. to the angle of strand orientation with respect Furnish should have a high slenderness ratio to the loading axis. This was con®rmed by (i.e., be as thin and as long as possible for a Barnes (2000), who found that strength de- composite product with high strength). In ad- 34 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1)

dition, fracture mechanics predicts the behav- The increase of fracture toughness and ior of wood composites with reasonable ac- strength with increasing density for solid curacy since the models for wood composites wood is well known. This trend has also been are based on intrinsic ¯aws within the product. observed in the strength of wood composites, Note that an increase in slenderness ratio leads speci®cally OSB. Barnes (2000) found an in- to a higher surface area to volume ratio for the crease in board properties with increased furnish, which will require higher resin con- board density, and developed the following sumption. Thus, there is a trade-off between model, to predict the strength properties (MOE mechanical properties and cost for the wood and MOR) of OSB: composite. ␳ x ϭ b FRiF ΂΃␳ , (10) ADHEREND a

Fracture behavior of wood-adhesive joints where FR is the resultant property, MOE or and wood composites is dependent on the MOR (psi), Fi, the initial property, MOE or ␳ fracture behavior of the solid wood as well as MOR (psi), a, the initial dry wood density 3 ␳ the surface preparation of the wood. Rele- (lb/ft ), b, the dry wood density in product vance of the fracture behavior of solid wood (lb/ft3), and x, the exponent for the desired is most evident for glulam. Since the size of property (1.0 for MOE, 1.2 for MOR). the wood component is relatively large, it be- Processing of wood-adhesive joints and haves similarly to solid wood. In a series of wood composites affects the behavior of the studies, Murphy (1986) compared the strength ®nal product. Surface preparation of the ad- reduction of notched beams to beams contain- herends is important in the formation of an ing a narrow slit parallel to the long axis of adhesive bond. Surface damage increases the the beam. It was found that slits had lower discontinuity of the bond-line, thus increasing bending strengths than a notch of the same the number of internal ¯aws (White and Green length. The magnitude of the strength reduc- 1980). Sasaki et al. (1973) shows that the frac- tion led Murphy to recommend that large, ture strength of a wood-adhesive bond is notched beams should be replaced with clear greatest for microtomed and planed surfaces beams equal to the net section of unnotched followed by ®ne-sawn, rough-sawn, disk-cut material in design. (planed with a disk planer), and sanded. Each Moisture content of the wood used to form of these preparation techniques leads to in- the joint also has an effect on adhesive joint creased damage of the wood ®bers. performance as the fracture energy increases The effect of the OSB strand surface char- with decreasing moisture content, down to ap- acteristics on the bond-line formed is similar proximately 10% (Ebewele et al. 1986a). This to the effect of surface roughness for wood- is similar to solid wood where the fracture adhesive bonds. Strands having a rough sur- toughness passes through a maximum at ap- face or containing large surface voids are more proximately 6 to 8% and decreases thereafter. likely to trap resin and form discontinuous The difference is due to the range examined. bonds after pressing. Strand size homogeneity Ebewele et al. did not study moisture contents is also important since increased homogeneity below 10%; however, it is likely that the max- should, theoretically, lead to a better resin dis- imum would be found below this value. In ad- tribution (Conrad et al. 2000). Therefore, im- dition, thermal effects, such as harsh drying, proved strand generation (i.e., reduction of the can compound moisture effects, and reduce size variability between strands and the sur- the overall fracture toughness of the wood it- face roughness) should improve the fracture self by introducing more and larger internal toughness of OSB. Damage of the wood ®bers ¯aws. can also occur through thermal degradation, Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 35 e.g., heating that results from machining with The studies discussed above show that the dull blades (Ebewele et al. 1986a). As above, behavior of solid wood is an important factor the increased surface damage will reduce joint in determining the overall behavior of wood- performance. Jung and Murphy (1983) have adhesive joints and wood composites. The ef- also shown that there is a reduction in fracture fect of notches, moisture content, and density toughness with increasing veneer thickness. on the fracture toughness of wood composites They speculate that as the thickness increases, is similar to that for solid wood. In the for- there is increased damage of wood ®bers. mation of wood-adhesive joints, surface prep- The adherend surface will also affect the aration is vital to ensure that appropriate con- crack path in a wood-adhesive joint. There are clusions are drawn from experiments. Improp- three possible fracture paths: the crack can er preparation may lead to deviations of the propagate in the wood alone, the adhesive crack from the desired location: in the wood, alone, or at the wood-adhesive interface. The in the resin, or at the wood-resin interface. highest fracture energies occur in the ®rst two cases (Ebewele et al. 1979) with a high per- RESIN centage of wood failure indicating a strong ad- hesive bond (Ebewele et al. 1986b). The per- The adhesive is the component used to bond centage of wood failure can be estimated by the joint or composite together. Numerous au- ASTM Practice D5266-99: Standard Practice thors (Shimizu and Okuma 1981; Sasaki et al. for Estimating the Percentage of Wood Failure 1973; White 1976; White and Green 1980; in Adhesive Bonded Joints (2000). Ebewele et al. 1979, 1982, 1986b; Gagliano In contrast, Ebewele et al. (1980), who and Frazier 2001) have studied parameters as- compared hand-sanding to machine-sanding, sociated with the bond-line, including bond- found that the hand-sanded surface gave high- line thickness and resin penetration (Shimizu er fracture toughness than a comparable ma- and Okuma 1981). Both Sasaki et al. (1973) chine-sanded specimen. This was despite the and Ebewele et al. (1979) have shown that fact that hand-sanding led to greater surface there is an optimum bond-line thickness for roughness. However, hand-sanding is done in fracture toughness. This is evident in the work a back-and-forth motion, whereas machine- by Ebewele et al. (1979) that compared the sanding is in one direction, the direction of the fracture toughness for crack initiation with grain. The back-and-forth motion of hand- that for arrest of the same crack over a range sanding leads to the creation of pre-failed in- of bond-line thicknesses for the hard maple/ terfaces between ®bers. These planes of weak- phenol-resorcinol system, shown in Fig. 7. It ness allow the crack to deviate from the ad- should be noted that optimums were found for hesive layer and arrest the crack, thus increas- two very different systems. Sasaki found an ing the fracture toughness. optimum of 500 ␮m for Kauri/epoxy, while A similar effect is seen when examining the Ebewele et al. (1979) found an optimum thick- angle the grain makes with the bond-line of ness of 85 ␮m for maple/phenol-resorcinol. the wood-adhesive specimen. Mijovic and Since the optimum thickness is speci®c to Koutsky (1979) found that minimum fracture each wood/resin system, each combination of toughness occurs at approximately 30Њ. Below wood species and resin type should be studied 30Њ the crack deviates into the wood, similar independently. Uniformity of the bond-line to the effect found by Ebewele et al. (1980) thickness and the area coverage are also key with hand-sanded substrates, giving a mea- parameters as Shimizu and Okuma (1981) surement of solid wood cohesive fracture found that increased uniformity leads to higher toughness. As the grain angle is reduced from strength bond-lines. For a bond-line to be con- 90Њ to 30Њ there is less surface area available sidered perfectly uniform, a sample from any for resin penetration. point along that bond-line will be identical to 36 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1)

no resin at the interface and available for bonding. Kamke et al. (1996) noted the phe- nomena in their studies of OSB. This also leads to the sharp decrease in toughness at thin bond-lines as seen in Fig. 7. Joint starvation due to over penetration of resin into the wood will result in reduced mechanical properties. However, some penetration is necessary to re- pair the adherend surface from damage caused during preparation. This is shown by White (1976), who found that the fracture toughness FIG. 7. Effect of bond-line thickness on Mode I frac- increased as the penetration depth of the resin ture toughness for hard maple/phenol-resorcinol system. into the wood increased. In contrast, Ebewele GIc is the critical fracture energy and GIa is the arrest load energy. If GIc is greater than GIa unstable crack growth et al. (1986b) state that a shallow penetration occurs. (Ebewele et al. 1979) depth can produce weak bonds, but at this time one cannot make a de®nitive statement on the correlation between penetration depth and me- any other sample taken at a different point. For chanical properties. example, continuous bond-lines will have the The effect of resin amount was also studied same thickness; discontinuous bond-lines will by Higgins (1990), who used a modi®ed Han- have the same pattern (e.g., droplet spacing kinson equation, similar to Barnes (2001), and size). In both cases, the bond-line is self- Eqs. (1) and (2), to model the strength (MOR) similar along the entire length. of oriented strand composites. In this case, the At the wood composites scale, fracture modi®cation is the use of a von Mises prob- toughness of oriented strandboard is directly ability distribution function (pdf), g(␪, m, k), affected by the resin dispersion and distribu- to account for the orientation of the strands. tion. These in turn affect the uniformity of the Inputs to the model are the concentration pa- bond-line and the probability of ®nding sec- rameter, k, which de®nes the degree of orien- tions that are not bonded. The resin dispersion tation of the board; the speci®c longitudinal, is the size distribution of the resin droplets L; and transverse, T, tensile strengths of the produced by the atomizer, and resin distribu- strands, Eqs. (11) to (13). As the concentration tion is the spatial distribution of the resin on parameter increases, the von Mises pdf nar- the strand. Because of the small amount of rows and the maximum value increases mean- adhesive applied in OSB manufacture, on the ing that a higher proportion of the strands are order of 2 to 3 weight %, the bond-line formed oriented parallel to the board axis. is discontinuous. This leads to stress concen- 1 trations at each resin droplet. Kamke et al. ␪ ϭ k cos 2(␪Ϫm) g( , m, k) ␲ e , (11) (1996) and Youngquist et al. (1987) have dem- Io(k) onstrated that the modulus of rupture (MOR) ␴Å ࿣ and internal bond (IB) strength increase with s(␪) ϭ , (12) 2 an increased uniformity of resin coverage. As 1 ϩ [(␴Å ࿣ /␴Å ⊥) Ϫ 1]sin ␪ well, Kamke et al. (1996) show that a reduc- ␲ /2 tion in droplet size also increases the mechan- ␪ ϭ ͵ ␪ ␪ ␪ Sm( ) g(m, k, )´s( ) d , (13) ical performance of the OSB. Kamke et al. Ϫ␲ /2 (1996) have also stated that the optimum dis- persion and distribution are unknown. where g(␪, m, k) is the grain angle pdf, ␪, the If a resin droplet is too small, it may be individual strand grain angle, with respect to completely absorbed into the strands, leaving the board axis, m, the angle between the prin- Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 37

cipal orientation axis and the axis of load, k, al. (1982) found that there is an optimum cure the orientation parameter of strand angular time for maximum fracture toughness. Again,

spread, Io(k), a modi®ed Bessel function of or- each system has its unique optimum, depen- ␪ der zero, Sm( ), a mathematical expectation of dent on the parameters used to form the joint composite strength, s(␪), a Hankinson expres- including the substrate, resin, and formation sion for the speci®c strength of a strand loaded temperature. Gagliano and Frazier report that

at angle ␪ with respect to its grain, ␴Å ࿣, the maximum fracture toughness for the yellow mean speci®c tensile strength of strands tested poplar/phenolformaldehyde combination was

parallel to grain, and ␴Å ⊥, the mean speci®c ten- achieved at a temperature of 175ЊC after 20 sile strength of strands tested perpendicular to min, whereas Ebewele et al. (1982) reports grain. maximums reached after 30 min to 4 h as the Higgins concluded that the model is accu- bonding temperature decreases from 150ЊCto rate when suf®cient resin is available to pro- 50ЊC for the hard maple/phenol-resorcinol sys- vide adequate stress transfer, and underlines tem. Both also show a reduction in fracture the importance of the bond-line in achieving toughness beyond this optimum time, which the maximum strength of oriented wood com- they speculated is the result of embrittlement posites. Bonding was increasingly critical as due to excessive cross-linking. the orientation level, the percentage of strands The effect of the bond-line on wood-adhe- oriented in the direction of applied load, in- sive joints and wood composites can be sum- creased. Experimentally, 5 to 7 times the marized as follows. Wood-adhesive joints with amount of resin typically used in industry was the greatest fracture toughness are theoretical- required to achieve the maximum speci®c ten- ly those with a uniform thickness of resin: op- sile strength predicted by the model. As well, timum thickness depends on the wood-adhe- Higgins also found that the internal bond sive combination used. Care should be taken strength of randomly oriented board was re- so that overpenetration of the resin into the duced by 25% to 30% when as little as 10% wood does not occur. This is a concern both of the surface was inadequately covered with for continuous and discontinuous bond-lines. resin. In addition, the model put forth by In the case of discontinuous bond-lines, the Barnes, Eqs. (1) and (2), also predicts increas- resin distribution should be uniform and com- ing stiffness and strength with increasing posed of small resin droplets. However, the strength perpendicular to the bond-line, and optimum distribution has not been determined. emphasizes the signi®cance of the bond-line. While work by Barnes and Higgins demon- It should be remembered that wood-adhe- strates the importance of the bond-line in sive joints and wood composites are used for achieving the desired macroscopic board prop- long periods of time, and thus time effects as- erties, none of the proposed models addresses sociated with the effectiveness of the bond- how the adhesive bond is affected by resin dis- line are of importance. River et al. (1989) persion and distribution. Therefore, an exam- found that the fracture toughness of UF resins ination of the effect of resin dispersion and changed with time. Material may initially ex- distribution (size variability and inter-droplet hibit stable fracture and have a constant frac- spacing) on the fracture toughness of wood- ture toughness, and at longer times, up to 2 adhesive joints and wood composites would weeks after bonding, the fracture toughness of be bene®cial. the adhesive joint can continue to increase. River et al. speculate that this is due to con- SUMMARY AND CONCLUSIONS tinued cross-linking and physical aging of the resin. This concept of optimum time can also This paper reviews the literature of fracture be applied to the curing of a bond-line. Both of wood composites and relates it to fracture Gagliano and Frazier (2001) and Ebewele et of wood-adhesive joints and solid wood. The 38 WOOD AND FIBER SCIENCE, JANUARY 2004, V. 36(1) review is not exhaustive and literature has ACKNOWLEDGMENTS been omitted for the sake of brevity. Based The authors would like to thank and ac- purely on geometrical considerations, the lit- knowledge Forintek Canada Corporation, the erature shows that the basics of structural Structural Board Association, and the National bonded joint behavior are readily applicable to Science and Engineering Research Council of wood-adhesive joints and wood composites. Canada Collaborative Research and Develop- Strength of both increases with increasing lap ment Grant Program for providing the ®nan- length and increasing slenderness ratio (i.e., cial support for this work. decreasing lamination or veneer thickness). Fracture mechanics is also applicable to wood REFERENCES composites since the models proposed are based on the concept of intrinsic ¯aws within AMERICAN SOCIETY FOR TESTING AND MATERIALS (ASTM). the product. 2000. Standard practice for estimating the percentage of Behavior of the adherends is important. All- wood failure in adhesive bonded joints. Annual Book of ASTM Standards. D5266-99, 15.06, 438±440. wood component sizes: lumber, veneer, BARNES, D. 2000. An integrated model of the effect of strands, particles, and ®bers have been stud- processing parameters on the strength properties of ori- ied; and the majority of researchers have con- ented strand wood products. Forest Prod. J. 50(11/12): cluded that fracture of wood-adhesive joints 33±42. . 2001. A model of the effect of strand length and should occur in wood alone and not at the strand thickness on the strength properties of oriented wood-resin interface for optimum perfor- wood composites. Forest Prod. J. 51(2):36±45. mance. Thus, an understanding of the fracture CONRAD, M. P. C., G. FERNLUND,G.D.SMITH, AND R. of solid wood at all levels is also essential to KNUDSON. 2000. Identi®cation of key parameters for the better comprehend the behavior of wood com- blending dynamics of OSB: Literature review and pre- liminary experimentation. Forintek Canada Corp., Van- posites. The location of fracture in a wood couver, BC, Canada. 18 pp. composite is dependent upon the surface prep- EBEWELE, R. O., B. H. RIVER, AND J. A. KOUTSKY. 1979. aration and grain angle of the wood substrates. Tapered double cantilever beam fracture tests of phe- For OSB in particular, fracture toughness can nolic-wood adhesive joints. Part I: Development of specimen geometry; effects of bondline thickness, wood be improved by improving strand generation. anisotropy, and cure time on fracture energy. Wood Fi- Improved strand generation leads to decreased ber 11(3):197±213. surface roughness and therefore, as is the case , , AND . 1980. Tapered double can- with wood adhesive joints, increased fracture tilever beam fracture tests of phenolic-wood adhesive joints. Part II: Effects of surface roughness, the nature toughness. of surface roughness, and surface aging on joint fracture Fracture toughness of OSB also increases energy. Wood Fiber 12(1):40±65. with increasing uniformity of resin distribution , , AND . 1982. Relationship between (spatial variability) and decreasing resin dis- phenolic adhesive chemistry, cure and joint perfor- persion (size variability). Both will lead to mance: Effects of base resin constitution and hardener on fracture energy and thermal effects during cure. J. greater bond-line uniformity and will increase Adhesion 14:189±217. the fracture toughness, as found in wood-ad- , , AND . 1986a. Wood processing hesive joints. Despite this knowledge on resin variables and adhesive joint performance. J. Appl. Po- dispersion and distribution, optimum condi- lym. Sci. 32:2979±2988. , , AND . 1986b. Relationship be- tions will depend on the wood species and res- tween phenolic adhesive chemistry and adhesive joint in system used. In addition, there are currently performance: Effect of ®ller type on fraction energy. J. no proposed models that address the effect of Appl. Polym. Sci. 31(7/8):2275±2302. resin dispersion and distribution on the adhe- FURUNO, T., C.-Y. HSE, AND W. A. COÃ TEÂ . 1983. Observa- tion of microscopic factors affecting strength and di- sive bond, such models are necessary to op- mensional properties of ¯akeboard. Pages timize the desired macroscopic board proper- 297±312 in Proc. 17th International Particleboard Sym- ties. posium, Washington State University, Pullman, WA. Conrad et al.ÐFRACTURE OF WOOD COMPOSITES 39

GAGLIANO,J.M.,AND C. E. FRAZIER. 2001. Improvements Fracture: Fracture Behaviour and Design of Materials in the fracture cleavage testing of adhesively-bonded and Structures. wood. Wood Fiber Sci. 33(3):377±385. MIJOVIC, J. S., AND J. A. KOUTSKY. 1979. Effect of wood HERTZBERG, R. W. 1996. Deformation and fracture me- grain angle on fracture properties and fracture mor- chanics of engineering materials. John Wiley & Sons, phology of wood-epoxy joints. Wood Science 11(3): Inc., Toronto, Canada. 786 pp. 164±168. HIGGINS, E. D. 1990. Strength of wood strand composites. MURPHY, J. F. 1986. Strength and stiffness reduction of Ph.D. Thesis, University of British Columbia, Vancou- large notched beams. J. Struct. Eng. 112(9):1989±2000. ver, BC, Canada. RIVER, B. H., AND E. H. OKKONEN. 1993. Contoured wood double cantilever beam specimen for adhesive joint ILCEWICZ,L.B.,AND J. B. WILSON. 1981. Fracture me- fracture tests. J. Testing Eval. 21(1):21±28. chanics of particleboard using nonlocal theory. Wood ,T.C.SCOTT, AND J. A. KOUTSKY. 1989. Adhesive Science 14(2):65±72. joint fracture behavior during setting and aging. Forest JUNG, J., AND J. F. MURPHY. 1983. An investigation of the Prod. J. 39(11/12):23±28. fracture of butt joints in parallel-laminated veneer. SASAKI, H., E. MCARTHUR, AND J. W. GOTTSTEIN. 1973. Wood Fiber Sci. 15(2):116±134. Maximum strength of end-grain to end-grain butt joints. KAMKE, F. E., E. KULTIKOVA, AND C. A. LENTH. 1996. OSB Forest Prod. J. 23(2):48±54. properties as affected by resin distribution. Pages 147± SATO, K. 1988a. In¯uences of specimen size, crack length, 154 in Proc. Fourth International Panel and Engineered- loading rate and notch acuity on medium density ®ber- Wood Technology Conference and Exhibition. board. J. Japan Wood Research Society 34(11):955± KOMATSU, K. 1984. Application of fracture mechanics to 958. the strength of cross-lapped glued timber joints. Forest . 1988b. Fracture of medium-density ®berboard Research Institute Bulletin No. 61. Forest Research In- under mixed mode loading. J. Japan Wood Research stitute, New Zealand Forest Service, Rotorua, New Zea- Society 34(10):828±833. land. SCHNIEWIND,A.P.,AND D. E. LYON. 1973. A fracture me- ,H.SASAKI, AND T. M AKU. 1976. Application of chanics approach to the tensile strength perpendicular fracture mechanics to strength analysis of glued lap to grain of dimension lumber. Wood Sci. Technol. 7:45± joints. Wood Research: Bull. Wood Research Institute, 59. SCOTT, T. C., B. H. RIVER, AND J. A. KOUTSKY. 1992. Kyoto University 61:11±24. Fracture testing wood adhesives with composite canti- LAUFENBERG, T. L. 1984. Flakeboard fracture surface ob- lever beams. J. Testing Eval. 20(4):259±264. servations and correlation with orthotropic failure cri- SHIMIZU, S., AND M. OKUMA. 1981. Fundamental studies teria. J. Inst. Wood Science 10(2):57±65. on the strength of the oriented particleboard: On the LEI, Y. -K., AND J. B. WILSON. 1979. Fracture toughness bonding strength between wood particles. Mokuzai of parallel-laminated veneer. Forest Prod. J. 29(8):28± Gakkaishi 27(1):8±13. 32. SIMPSON, W. T. 1977. Model for tensile strength of ori- , AND . 1980. Fracture toughness of oriented ented ¯akeboard. Wood Science 10(2):68±71. ¯akeboard. Wood Science 12(3):154±161. SMITH,F.W.,AND D. T. PENNEY. 1980. Fracture mechanics , AND . 1981. A model for predicting frac- analysis of butt joints in laminated wood beams. Wood ture toughness of ¯akeboard. Wood Science 13(3):151± Science 12(4):227±235. 156. TONG, L., AND G. P. STEVEN. 1999. Analysis and design LEICESTER, R. H. 1973. Effect of size on the strength of of structural bonded joints. Kluwer Academic Publish- structures. Pages 1±13 in Forest Products Laboratory, ers, Norwell, MA. 288 pp. Division of Building Research Technological Paper No. WALSH, P. F., R. H. LEICESTER, AND A. RYAN. 1973. The 71. Commonwealth Scienti®c and Industrial Research strength of glued lap joints in timber. Forest Prod. J. 23(5):30±33. Organization (CSIRO). Melbourne, Australia. WHITE, M. S. 1976. In¯uence of resin penetration on the , AND P. C. BUNKER. 1969. Fracture at butt joints fracture toughness of wood adhesive bonds. Wood Sci- in laminated pine. Forest Prod. J. 19(2):59±60. ence 10(1):6±14. . 1974. Applications of linear fracture mechanics , AND D. W. GREEN. 1980. Effect of substrate on in the design of timber structures. Pages 156±164 in the fracture toughness of wood-adhesive bonds. Wood Proc. 23rd Conference of the Australian Fracture Group. Science 12(3):149±153. Melbourne, Australia. YOUNGQUIST, J. A., G. C. MYERS, AND L. L. MURMANIS. MIHASHI, H., AND M. HOSHINO. 1989. Fracture toughness 1987. Resin distribution in : Evaluated by in- and tension softening properties of glued laminated tim- ternal bond strength and ¯uorescence microscopy. bers. Pages 799±804 in 8th European Conference on Wood Fiber Sci. 19(2):215±224.