Dental Materials 19 (2003) 700706 www.elsevier.

com/locate/dental

Viscoelastic properties of demineralized dentin matrix

David H. Pashleya,*, Kelli A. Ageea, John C. Watahab, Frederick Rueggebergb, Laura Ceballosc, Kousuke Itoud, Masahiro Yoshiyamad, Ricardo M. Carvalhoe, Franklin R. Tayf
a

Department of Oral Biology and Maxillofacial Pathology, School of Dentistry, Medical College of Georgia, Augusta, GA 30912-1129, USA b Department of Oral Rehabilitation, Medical College of Georgia, Augusta, GA, USA c Division of Dental Materials, University of Granada, Granada, Spain d Department of Conservative Dentistry, Okayama University, Okayama, Japan e Department of Operative Dentistry, Endodontics and Dental Materials, Bauru School of Dentistry, USP, Bauru, SP, Brazil f Conservative Dentistry, University of Hong Kong, Hong Kong, SAR, Peoples Republic of China Received 20 November 2001; received in revised form 27 June 2002; accepted 13 November 2002

Abstract Objectives. To evaluate the viscoelastic properties of demineralized dentin matrix. Stress relaxation studies were done on matrices in tension and strain elongation or creep studies were done in both tension and compression. Methods. Mid-coronal dentin disks were prepared from extracted unerupted human third molars. Disks were 0.5 mm thick for stress relaxation or tensile creep experiments and 0.2 0.3 mm thick for compressive creep studies. I beam specimens were prepared from dentin disks and the middle region was demineralized in 0.5 M EDTA (pH 7) for 4 days. The specimens were held in miniature friction grips in water and pulled at 100 mm s21 to strains of 5, 10, 15 or 20% and then held for 10 min to follow the decay of stress over time. Creep was determined on demineralized dentin immersed in water in tension and in compression. Compressive creep was measured using an LVDT contact probe with loads of 0.02 0.5 N. Strain data were converted to compliance time curves (strain/stress) and expressed as total compliance Jt ; instantaneous elastic compliance Jo ; retarded elastic compliance JR and viscous response t=h or creep. Results. The dentin matrix exhibits both stress relaxation and creep behavior. Stress relaxation and tensile creep were independent of strain but compressive creep rates were inversely related to compressive strain. Creep values were about 10% at low compressive strains, but fell progressively to 1% at high strains. Compliance time curves fell with stress and came closer together. However, tensile creep was about 3% regardless of the strain. Signicance. The dentin matrix exhibits viscoelastic properties, but is not linearly viscoelastic. The relatively high creep rates of the matrix under low compressive loads may cause viscous deformations in poorly inltrated hybrid layers in resin-bonded teeth under function. 2003 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
Keywords: Dentin matrix; Stressrelaxation; Creep; Hybrid layers

1. Introduction Many collagenous materials exhibit the characteristics of elastic solids and viscous liquids and hence are classied as viscoelastic materials. They all share common properties such as a strain-dependent modulus of elasticity, sensitivity to strain rate, creep and stress relaxation [1 3]. The stress strain curves of such materials are generally not linear. They have an initial region where stress does not rise much with initial strain. With more strain, the slope of the stress strain curve (i.e. stiffness) continues to increase.
* Corresponding author. Tel.: 1-706-721-2033; fax: 1-706-721-6252. E-mail address: dpashley@mail.mcg.edu (D.H. Pashley).

With such materials, one must dene the strain when reporting a modulus of elasticity, as well as the strain rate. Creep is another characteristic of viscoelastic materials. It is dened as an increase in strain over time under a constant stress [4], or strain relaxation. A related viscoelastic property is stress relaxation, the fall in stress at a constant strain. Stress relaxation measurements in mineralized dentin demonstrated that air-dried dentin was stiffer and exhibited decreased stress relaxation compared to wet mineralized dentin [5]. These viscoelastic properties have recently been measured in demineralized dentin matrices at low strains [6]. They may contribute to stress dissipation during dentin bonding because the bonds are formed by resin inltration of demineralized dentin matrices [7].

0109-5641/$ - see front matter 2003 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved. doi:10.1016/S0109-5641(03)00016-2

D.H. Pashley et al. / Dental Materials 19 (2003) 700706

701

Several reports have found that the bottom half of some hybrid layers contain little resin around demineralized collagen brils [8,9]. These naked collagen brils can undergo creep and stress relaxation during masticatory function, as well as during polymerization shrinkage. The purpose of this work was to evaluate the viscoelastic characteristics (i.e. creep and stress relaxation) of demineralized dentin matrix. The null hypothesis that was tested was that creep and stress relaxation are not dependent on strain.

2. Materials and methods Specimen preparation. Disks of mid-coronal dentin were prepared (Fig. 1A) from extracted, unerupted human third molars that had been stored in water at 4 8C for less than 1 month, using an Isomet saw (Buehler Ltd, Lake Bluff, IL, USA). Disks (0.5 mm thick) used to measure tensile creep and stress relaxation were cut into I beam congurations [10]. Both ends of the I beams (Fig. 1B) were protected with two layers of nail varnish, and the middle section (3 mm gauge length) was demineralized using 0.5 M EDTA (pH 7) for 4 days at 25 8C with stirring. Disks used to measure compressive creep were 0.2 mm thick and were also demineralized in EDTA as described above. Dentin disks

Fig. 1. Schematic illustrating the shape (A) and dimensions (B) of I beam specimens used to measure stress relaxation in water (C). Dentin disks 0.2 mm thick were used to measure compressive compliance time relationships using an LVDT (D). Disks that were 0.5 mm were used to measure tensile creep.

used for tensile creep measurements were made thicker (0.5 mm) because the 0.2-mm thick specimens often broke when they were placed in friction grips described below. The actual thickness of each demineralized specimen was measured on edge in three regions that were then averaged. This was done with the specimens immersed in water using a videomicrometer at 40X [11]. Stress relaxation measurements. Stress relaxation measurements were made in a Vitrodyne Model V-1000 universal tester (John Chatillon & Sons, Inc., Greensboro, NC, USA) while the specimens were immersed in water (Fig. 1C) at 25 8C. The mineralized ends of 10 I beams were held in miniature friction grips. All specimens were strained to 5, 10, 15 or 20% strain at a rate of 100 mm s21, and then held at that strain for 600 s as the fall in stress was recorded at a rate of one observation every 3 s. The decay in stress at a constant strain was analyzed by regression analysis. Least squares curves were calculated along with R2 values. Stress relaxation data were normalized for differences in strain-induced stress by expressing each trial as a fraction of the maximum value. Creep measurements. Tensile creep was measured using the same apparatus as was used to measure stress relaxation. I beams 0.5 mm thick (Fig. 1C) that were demineralized in EDTA were randomly pulled to 5 or 15% strain in water and held at that stress for 600 s while measuring small increases in strain, using the software of the Vitrodyne V1000 universal tester. The degree of tensile strain was calculated by dividing the measured cross-head displacement by the measured gage length. This was the distance between the mineralized ends of the I beams (Fig. 1B) as measured to the nearest 0.01 mm using the videomicrometer on demineralized specimens immersed in water [11]. Since specimen displacements were not directly measured with an extensometer, but were only measured by cross-head displacement, the displacements are only approximate. Load displacements were converted to stress strain curves. Compliance time curves were plotted from strain/stress curves as described below. Compressive creep was measured in a modied Thermal Mechanical analyzer (Perkin Elmer, Model TMS-2, Wellesley, MA, USA). The furnace was removed and an adjustable stage holding an aluminum well was placed below the quartz rod (Fig. 1D) connected to a LVDT (linear variable differential transformer). A 0.6 mm diameter glass 1 ml micropipette (Microscope, Fisher Scientic, Atlanta, GA, USA) that was heat-sealed and attened on the end was xed to the quartz rod to serve as the contact probe. The weight of the probe, rod and weight pan was 1.1 g as determined using a top-loading balance at the same position as the experimental specimens. Additional weights between 1 and 50 g (equivalent to 0.08 1.8 MPa) were applied to the pan during creep measurements. The reported loads included the weight of the pan, probe and extrinsic weight. Load displacement curves were generated by sequentially applying increasing loads. Load application was done for

702

D.H. Pashley et al. / Dental Materials 19 (2003) 700706

300 s, with 600 s elapsing between loads. Compressive strain was calculated as the ratio between the LVDT probe displacement under a compressive load and the specimen thickness, that was always measured in three places on the edge of the uniform disk of demineralized dentin using the videomicrometer microscope at 40X [11]. During all creep measurements, the specimens were completely covered with water at 25 8C. Prior to load applications for compressive creep measurements, the matrices were dried with dry N2 gas to determine the maximum possible linear shrinkage of dehydrated matrices. Addition of water resulted in complete re-expansion. Sequential loads were applied to the matrix (0.024, 0.043, 0.052, 0.062, 0.111, 0.307, 0.405, 0.503 N) by instantaneous application of weights to the pan on top of the contact probe of the LVDT (Fig. 1D). Additional experiments were done in reverse order to rule out order effects. Since the thickness of the specimen was known, the compressive displacement divided by the original specimen thickness gave a direct measure of negative strain. Data were collected at a rate of 1 point every 3 s. Creep was calculated graphically from the strain versus time curve according to the method described by Goldberg [14]. When creep was divided by the applied stress, the resulting creep compliance curves were analyzed according to previously described methods [15 17]: JT J0 JR t=h where Jt total compliance (MPa21) Jo instantaneous elastic compliance (MPa21) JR retarded elastic compliance (MPa21) t=h viscous response (MPa21) The stiffness was calculated by converting the load displacement curves to stress strain curves. Statistical analysis. Slopes of the stress relaxation curves were compared by one-way ANOVA seeking to identify any signicant differences among the slopes. Multiple comparisons were performed using Student Newman Keuls test at a 0.05 level of signicance. Similarly, signicant differences were sought between tensile and compressive creep at 5 and 15% strain using a one-way ANOVA. Signicant differences were isolated using Student Newman Keuls method at a 0.05 level of signicance.

Fig. 2. Stress relaxation curves of dentin matrix obtained when I beams were subjected to strains of 5, 10, 15 and 20% delivered 100 mm s21. Data were collected every 3 s but appear as a solid line. Brackets indicate ^ 1 S.D.

expressing them as a fraction of their maximum stress, the resulting stress relaxation curves were nearly identical regardless of strain (Fig. 3). Regression analysis revealed the best Least Squares t to be power curves (Fig. 3) with the exponent representing the slope of the line. One-way ANOVA analysis of these slopes failed to identify any signicant difference p 0:21 among them (Table 1). When the same strains were produced at a strain rate of 10 mm s21 (data not shown), the maximum stresses that developed were higher p 0:05; but the normalized stress relaxation curves were not signicantly different p 0:19 indicating that stress relaxation was independent of the degree of strain or strain rate (Fig. 3). Creep measurements. Dehydration of the specimens with dry N2 gas induced a rapid linear shrinkage of between 31 and 55% (Fig. 4), that was fully reversible when the specimens were covered with water. Application of increasingly heavier loads to specimens immersed in water resulted in greater compressive strain (Fig. 4). To rule out order effects, several specimens were exposed

3. Results Stress relaxation measurements. When specimens were subjected to tensile stress at a strain rate of 100 mm s21, to strains of 5, 10, 15 or 20%, they developed stresses that were proportional to their strains (Fig. 2). When the stresses developed by each specimen were normalized over time by

Fig. 3. Stressrelaxation curves of dentin matrix normalized as a percent of the maximum generated by each strain. Data were collected every 3 s but appear as a solid line. Upper line was obtained at 20% strain. Next line was obtained at 15% strain. Lowest line represents data obtained at 5% strain.

D.H. Pashley et al. / Dental Materials 19 (2003) 700706 Table 1 Slopes (%s21) of normalized stressrelaxation curves obtained with demineralized dentin matrices as a function of strain Specimen Strains 5% 1 2 3 4 5 6 7 8 9 10 Mean SD Signicance 0.032 0.028 0.013 0.046 0.055 0.046 0.064 0.073 0.012 0.041 0.021 a 10% 0.0337 0.0272 0.030 0.0373 0.027 0.040 0.039 0.053 0.053 0.029 0.037 0.010 a 15% 0.0388 0.0268 0.0331 0.0463 0.029 0.038 0.033 0.063 0.047 0.051 0.041 0.011 a 20% 0.060 0.024 0.036 0.050 0.030 0.042 0.031 0.093 0.051 0.050 0.047 0.020 a

703

Fig. 5. Single specimen of demineralized dentin alternatively exposed to 0.5 vs 0.05 N compressive loading. Note the slope of the strain elongation curve (i.e. creep) was always lower with the large load than it was with the small load (dotted lines).

All stressrelaxation curves were normalized by recalculating decay of stress as a percent of each specimens maximum value at that strain. Regression analysis revealed the best ts to be power functions, y x2s ; where 2s is the slope of the lines in %s21. All specimens were loaded at 100 mm s21 to the desired strain. Signicance groups identied by the same lower case letters are not signicantly different p , 0:05:

to the heaviest load (0.5 N) rst, followed by a lighter load second. Fig. 5 shows that the strain elongation (i.e. negative slope of strain vs time or creep) was consistently higher in the light load and consistently lower in the heavy load, ruling out any order effect. When compressive loads were applied, there was an immediate elastic compressive strain that was independent of time (Fig. 5), followed by a slower, time-dependent further elastic strain, followed by a still slower timedependent viscous strain (static creep). The stress strain curves of eight specimens are shown in Fig. 4 (insert). The insert shows eight specimens, ve of which were strained to 28% and two that were strained to 35%.

The compliance of the specimens was calculated by dividing the measured strains by the applied stress. The compliance time data of a single specimen is shown in Fig. 6. Note that the strain/stress compliance values are plotted as positive values even though the strain was actually negative. In this example, a load of 0.02 N was applied. The instantaneous elastic compliance (Jo) was 36% of the total. Where the slope of this line rst begins to deviate is the time-dependent retarded elastic compliance JR : This value was 56% of the total in this specimen with this load. The sum of Jo and JR represents the elastic response of the specimen (ca. 92%). The slope continued to fall to reach a steady-state that represented the viscous compliance t=h or creep that, in this example, was about 8% of the total. The range of values for static compressive creep was 1 10% and was dependent on load (Table 2). When the load was removed, there was no full recovery of the original dimensions. The difference was due to the viscous strain or creep that resulted in a permanent viscous deformation.

Fig. 4. Straintime response of single specimen of demineralized dentin matrix to compressive loads of 0.020.51 N. Note that the rst and last negative strains were produced by drying the matrix with dry nitrogen gas, followed by rehydration with water. The progressive fall in the recovery baseline represents cumulative plastic creep. Inset: stressstrain curves of eight specimens obtained from load displacement curves.

Fig. 6. Analysis of a single compressive loaddisplacement curve obtained by applied a load of 0.02 N to demineralized dentin. The negative strains were made positive. Individual data points are not shown but were collected every second. The curves were graphically analyzed according to Goldberg [14] where Jt Jo JR t=h (see text for details).

704

D.H. Pashley et al. / Dental Materials 19 (2003) 700706 Table 3 Comparison of tensile and compressive creep compliance of demineralized dentin matrix Strain (%) 5 15 Tensile creep (SD) (%) 3.5 (1.9)(10)a 2.7 (1.2)(10)a Compressive creep (SD) (%) 9.1 (2.2)(10)b 4.3 (1.1)(10)a

Table 2 Contribution of Jo ; JR and t=h to the total compliance of demineralized dentin at different loads Load (N) 0.02 0.03 0.04 0.05 0.06 0.11 0.21 0.31 0.41 0.51 Strains (SD) (%) 3 (1)a 5 (1)b 9 (2)c 11 (2)c 15 (3)c,d 19 (4)d 25 (5)d 28 (6)d 31 (6)d 34 (6)d Jo (SD) (%) 30 (14)a 26 (10)a 34 (10)a 28 (10)a 30 (13)a 40 (16)a,b 40 (23)a,b 34 (21)a,b 50 (14)a,b 59 (19)b JR (SD) (%) 60 65 59 67 65 57 59 64 48 40 (13)a (10)a (10)a (10)a (12)a (16)a,b (23)a,b (21)a,b (13)b (19)b t=h (SD) (%) 10 (2)a 9 (2)a 7 (2)b 5 (1)b 5 (1)b 3 (1)c 2 (1)c 2 (1)c 2 (1)c 1 (1)d

Groups identied with different superscript letters are signicantly different p , 0:05: N 10:

4. Discussion It is clear that the dentin matrix behaves as a viscoelastic material [6]. The elegant AFM studies of Balooch et al. [6], using small strains (i.e. 35 mm indentations into 1000 mm thick specimens, or compressive strains of 3.5%) of EDTAdemineralized human dentin in water, reported moduli of elasticity of 132 162 kPa. However, they did not use a wide range of compressive stresses, and hence, did not report a stress-related change in creep. The slopes of the lines of the stress relaxation curves obtained at strains of 5, 10, 15, or 20% in water, were all very similar indicating that stress relaxation of the demineralized dentin matrix is independent of strain. It is remarkable that a load of 0.51 N (stress 1.80 MPa) produced a compressive strain that was almost equivalent to that induced by drying the specimen with dry N2 gas (Fig. 4). That is, it is possible to physically displace water from interbrillar spaces using physical force while the specimen remained immersed in water. If the load is not applied very long to minimize creep, when the force was taken off, the matrix recoiled almost back to its original height in a rapid manner. Our results indicate an unusual, load-dependent decrease in compressive strain relaxation (i.e. creep, Table 2,Fig. 5). Usually, creep increases with load, at least in impression materials [15], orthodontic elastics [16] and resin composites [17] Although the current work was done on demineralized dentin, other reports comparing creep in mineralized and demineralized bone concluded that the collagen phase of bone is responsible for creep behavior of mineralized bone [18]. Thus, it is likely that mineralized dentin, demineralized dentin and resininltrated demineralized dentin may all exhibit similar amounts of creep [18] when compared at normalized stresses s=E: This speculation must be conrmed experimentally. Obviously, demineralized dentin with a modulus of elasticity of 0.1 10 MPa [6,18] will exhibit more creep under functional loads than will stiffer (ca. 15 20 GPa) mineralized dentin [19]. Our compressive strains ranged from 3 to 34% (Table 2). Although this represented a large strain range, these extremes were used to screen the viscoelastic behavior of the dentin matrix. This was done to determine if the compliance time curves for different loads all fell on the same line. Had this occurred, the dentin matrix could have been classied

Compliance (strain/stress) normalized to percent. Jo instantaneous elastic compliance; JR retarded or viscoelastic compliance; t=h viscous compliance or creep; Jt Jo JR t=h. Those groups not identied by different superscript letters are not signicantly different p . 0:05: N 10:

Fig. 7 summarizes the compliance components as a function of compressive load for ten specimens subjected to loads varying from 0.02 to 0.50 N. Note that the compressive creep values decreased with increasing load, and the compliance time curves came closer together. Table 2 summarizes these results as percent changes. Tensile creep measured in water at 5 and 15% strain were about 3% (Table 3) and were not signicantly different p 0:29 by Student t-test. Also shown in Table 3 are the mean compressive creep rates measured at the same compressive strains. The compressive creep at 5% strain was 2.5 times greater than the tensile creep p , 0:05: There was no detectable signicant difference between the tensile vs compressive creep of the dentin matrix when they were measured at 15% strain (Table 3), although the compressive creep was 1.6 times greater than the tensile creep.

Fig. 7. Summary of compliance time curves for six specimens of demineralized matrix at increasing loads. Half brackets represent one standard deviation. Note that the nal slopes of the curves (i.e. creep) decrease with increasing load, except for the 0.111 N load.

D.H. Pashley et al. / Dental Materials 19 (2003) 700706

705

as a linearly viscoelastic material [4]. Because these curves were different (Fig. 6), the matrix can only be classied as viscoelastic. The results of this study support the null hypothesis that stress relaxation is not dependent upon strain (Fig. 3), but do not support the null hypothesis that the degree of creep is not dependent upon applied stress, at least not when measured in compression. Since the tensile creep results (Table 3) were not strain-dependent, the results obtained in compression represent a special case. At compressive strains of 3 15% (Fig. 4), it is likely that the measured viscoelastic properties of dentin matrix represent the properties of the water-lled, collagen bril meshwork. Although there are dentinal tubules in the matrix, they probably only occupy 5 15% of the volume of mid-coronal dentin [20]. The remainder is composed of collagen brils oriented randomly, that have a mean diameter of 100 nm and interbrillar spaces of 20 30 nm, that are interconnected to create tortuous channels [21]. We speculate that when the water-saturated matrix received a compressive load, the rapid, instantaneous compression (Jo in Fig. 6) was due to the rapid loss of water from 2 3 mm diameter tubules as the matrix begins to collapse. The subsequent slower exit of water via the much smaller interbrillar channels may be responsible for the slower time-dependent collapse of the matrix (JR in Fig. 6). We speculate that the high creep rate seen at low strains is due, in part, to the viscoelastic properties of the collagen matrix and to slow water loss via tortuous interbrillar spaces. At higher compressive strains, most of the water was gone and all of stress was borne by the collagen matrix. It is at this point that the strain slowly decreases to a constant slope that represents creep. This explanation for the decrease in compressive creep with increasing loads must be regarded as speculative until further proof can be obtained. The viscous response of the dentin matrix (i.e. viscous strain or creep) was not recoverable when the load was removed, regardless of whether the stress was tensile or compressive. This is seen inFig. 4 as a fall in the baseline with each succeeding loading trial. Thus, creep creates a permanent deformation of the matrix. The molecular events that are thought to occur when collagen matrices are elongated under load, include molecular elongation of collagen helices (i.e. slight stretching of H-bonds), an increase in the size of the gap region between longitudinally adjoining molecules along the bril axis, and relative slippage of laterally adjoining molecules along the bril axis [3,18,22. The rst two events allow elastic recovery if the elongation was not excessive, but the molecular slippage would be permanent. We speculate that creep in dentin matrices is due to slippage of collagen molecular aggregates [23], whether in tension or compression. In compression, the random three-dimensional distribution of collagen brils would place a signicant fraction of the meshwork in tension. Although covalent cross-links bind many adjacent collagen molecules together [24,25], there may be regions in

the aggregations of collagen molecules that could permanently elongate if an external load were applied to perturb the aggregate. Obviously, it is desirable to have as little creep as possible, relative to the expected functional stress per unit time. The elongation of tendons and ligaments due to creep can be repaired over time by the slow metabolic turnover of collagen in those structures, but dentin collagen is not metabolically active. Therefore, creep in dentin matrices may slowly alter local stress concentrations in resin-bonded collagen brils under stress. If the resin in the hybrid layer creeps at the same rate as the collagen brils, no shear forces between the components should develop. However, if collagen brils creep less than the resin that surrounds them, then signicant shear forces might develop that could accelerate the degradation of the hybrid layer under function [26 28]. Several groups have measured compressive static creep in unlled resins as well as in composite resins [17,29,30]. Creep is much higher in unlled resins [12] or low lled resin composites [13] than in highly lled composites because of the greater volume fraction of polymer matrix [31]. Even in resin composites, creep can range from 1 to 6% depending upon the volume fraction of ller [17,31]. We have measured creep values of 1 14% in unlled adhesives, depending on the product (Pashley, unpublished observations). Thus, it appears that unlled resins exhibit more creep than does the collagen matrix at the same stress. We speculate that this may shift more load from the resin to the collagen brils under sustained force. It should be noted that during tensile creep measurements the entire 3 mm long 0.5 mm wide 0.5 mm thick matrix was undergoing bulk creep. During compressive creep measurements, only the matrix beneath the 0.6 mm diameter probe was undergoing creep. Thus, the two methods differ and sampled different amounts of matrix. It is unclear whether creep can occur during 1 2 s cyclic chewing. It may occur, but not be easily measured until hundreds of thousands of cycles have elapsed. This was the case for low-copper dentin amalgams. They underwent viscous deformation under routine occlusal function, but very slowly [32]. These ideas must be regarded as speculative until they can be tested. Clearly, more research on the viscoelastic properties of resins, dentin matrices (both normal and caries-affected) and resin-inltrated matrices is warranted.

References
[1] Larrabee WF. A nite element model of skin deformation. I. Biomechanics of skin and soft tissue. A review. Laryngoscope 1986;96:399405. [2] Pioletti DP, Rakotomanana LR. On the independence of time and strain effects on the stress relaxation of ligaments and tendons. J Biomech 2000;33:172932. [3] Sasaki N, Shukunami N, Matsushima N, Izumi Y. Time-resolved Xray diffraction from tendon collagen during creep using synchrotron radiation. J Biomech 1999;32:28592.

706

D.H. Pashley et al. / Dental Materials 19 (2003) 700706 [19] Marshall GW, Marshall SJ, Kinney JH, Balooch M. The dentin substrate: structure and properties related to bonding. J Dent 1997;25: 441 58. [20] Pashley DH. Dynamics of the pulpo-dentin complex. Crit Rev Oral Biol Med 1996;7:104 33. [21] Van Meerbeek B, Dhem A, Goret-Nicaise M, Braem M, Lambrechts P, Vanherle G. Comparative SEM and TEM examination of the ultrastructure of the resindentin interdiffusion zone. J Dent Res 1993;72:495501. [22] Sasaki N, Nakayama Y, Yoshikawa M, Enyo A. Stressrelaxation function in bone and bone collagen. J Biomech 1993;26:1369 76. [23] Sasaki N, Odajima S. Stressstrain curves and Youngs modulus of a collagen molecule as determined by the X-ray diffraction technique. J Biomech 1996;29:655 8. [24] Yamauchi M, Katz EP. The post-translational and molecular packing of mineralizing tendon collagens. Connect Tissue Rex 1993;29: 81 98. [25] Knott L, Bailey AJ. Collagen cross-links in mineralizing tissues: a review of their chemistry, function, and clinical relevance. Bone 1998;22:1817. [26] Sano H, Yoshiyama T, Pereira PNR, Kanemura N, Morigami M, Tagami J, Pashley DH. Long-term durability of dentin bonds made with a self-etching primer, in vivo. J Dent Res 1999;78:90611. [27] Hashimoto M, Ohno H, Kaga M, Endo K, Sano H, Oguchi H. In vivo degradation of resin dentin bonds in humans over 1 to 3 years. J Dent Res 2000;79:1385 91. [28] Hashimoto M, Ohno H, Endo K, Kaga M, Sano H, Oguchi H. The effect of hybrid layer thickness on bond strength: demineralized dentin zone of hybrid layer. Dent Mater 2000;16:40611. [29] Nielsen LE. Creep and dynamic mechanical properties of lled polymers. Trans Soc Rheol 1969;13:14166. [30] Ishai O, Bodner SR. Limits of linear viscoelasticity and yield of a lled and unlled epoxy resin. Trans Soc Rheol 1970;14:25373. [31] Ferracane JL, Matsumoto H, Okabe T. Time-dependent deformation of composite resinscompositional considerations. J Dent Res 1985; 64(11):13326. [32] Mahler DB, Adey JD. Factors inuencing the creep of dental amalgam. J Dent Res 1991;70:1394400.

[4] Craig RG, Powers JM. Restorative dental materials, 11th ed. St. Louis: Mosby; 2002. pp. 94 7. [5] Trengrove HG, Carter GM, Hood JAA. Stress relaxation properties of human dentin. Dent Mater 1995;11:30510. [6] Balooch M, Wu-Magidi I-C, Balazs A, Lundkvist AS, Marshall SJ, Marshall GW, Siekhaus WJ. Viscoelastic properties of demineralized dentin measured in water with atomic force microscope (AFM)-based indentation. J Biomed Mater Res 1998;40:53944. [7] Nakabayashi N, Pashley DH. Hybridization of dental hard tissue. Quintessence Publishing Co; 1998. [8] Spencer P, Swafford JR. Unprotected protein at the dentin adhesive interface. Quintessence Int 1999;30:5017. [9] Spencer P, Wang Y, Walker MP, Wieliczka DM, Swafford JR. Interfacial chemistry of the dentin/adhesive bond. J Dent Res 2000;79: 145863. r J, Carvalho R, Sacher B, Russell CM, Pashley [10] Zhang Y, Agee K, No DH. Effects of acid-etching on the tensile properties of demineralized dentin matrix. Dent Mater 1998;14:2228. [11] Pashley DH, Zhang Y, Agee KA, Rouse CJ, Carvalho RM, Russell CM. Permeability of demineralized dentin to HEMA. Dent Mater 2000;16:7 14. [12] Ruyter IE, Espevik S. Compressive creep of denture base polymers. Acta Odontol Scand 1980;38:16977. [13] Ruyter IE, ysaed H. Compressive creep of light cured resin based restorative materials. Acta Odontol Scand 1982;40:319 24. [14] Goldberg AJ. Viscoelastic properties of silicone, polysulde, and polyether impression materials. J Dent Res 1974;53:10339. [15] Duran RL, Powers JM, Craig RG. Viscoelastic and dynamic properties of soft liners and tissue conditioners. J Dent Res 1979; 58:1801 7. [16] Liu CC, Wataha JC, Craig RG. The effect of repeated stretching on the force decay and compliance of vulcanized cis-polyisoprene orthodontic elastics. Dent Mater 1993;9:3740. [17] Ruyter IE, ysaed H. Compressive creep in light cured resin based restorative materials. Acta Odont Scand 1982;40:31924. [18] Bowman SM, Gibson LJ, Hayes WC, McMahon TA. Results from demineralized bone creep tests suggest that collagen is responsible for the creep behavior of bone. J Biomed Engng 1999;121:253 8.