Acta Materialia 51 (2003) 365371 www.actamat-journals.

com

Compressive elastic modulus and its relationship to the structure of a hydrated starch foam
Z. Liu, C.S.L. Chuah, M.G. Scanlon
Department of Food Science, University of Manitoba, Winnipeg, Manitoba, Canada R3T 2N2 Received 10 May 2002; received in revised form 22 August 2002; accepted 22 August 2002

Abstract The elastic modulus of a hydrated starch foam was evaluated from indentation and compression tests. After using an appropriate correction factor, based on specimen aspect ratio and Poissons ratio, the indentation elastic moduli were found to be identical to the compressive ones. Image analysis showed that the foam possessed a non-periodic but self-similar texture, with a fractional dimension (FD) of 1.89 0.07 for cell numbers and 2.41 0.16 for areal mass of cell wall solids. The relative modulus exhibited a power law dependence on relative density. This dependence was corroborated from microstructural considerations of the scaling behaviour of the elastic properties according to a weak-link regime theory. The theory was deemed apt for starch foam, since the underlying microstructure of the cell walls of the foam is comprised of swollen starch spherulites sintered by a linear polymer (amylose) that is exuded from the spherulites during processing. 2002 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved.
Keywords: Foams; Mechanical properties-elastic; Image analysis; Microstructure

1. Introduction For engineering purposes, cellular solids represent an attractive set of materials with desirable attributes such as high strength/weight ratio for lightweight construction [1], and superior energy absorption for impact protection [2]. A variety of technologies have been developed for creating cellular structures out of almost all genres of materials, including metals, ceramics, polymers
Corresponding author. Tel.: +1 204 474 6480; fax: +1 204 474 7630. E-mail address: scanlon@cc.umanitoba.ca (M.G. Scanlon).

and composites [3], although in terms of annual production, biogenic cellular solids represent the largest single sector for cellular materials [4]. One group of biogenic cellular solids whose structure and properties can be manipulated by processing to attain certain desirable attributes are processed foods [5]; examples of these, where the relative density does not exceed 0.3 [4], include bread, cakes, meringues and processed fruits and vegetables [46]. Desirable attributes for these cellular solids include their physical properties: Kamyab et al. [7] list elastic modulus, fracture toughness, yield strength and coefcient of friction as critical to consumer satisfaction [8]. The recognition that the performance of cellular solids is dependent on

1359-6454/03/$30.00 2002 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved. doi:10.1016/S1359-6454(02)00400-7

366

Z. Liu et al. / Acta Materialia 51 (2003) 365371

cell structure (the size, shape, integrity and volume fraction of cells), and on the properties of the solid material [4], has been a cogent approach for understanding how process manipulation will produce cellular solids of specic end-use properties [9,10]. Establishing the link between a given measured mechanical parameter and the processing conditions necessary to create it, therefore requires a knowledge of the structure that the processing creates, and from which the mechanical performance of the cellular solid can be predicted [11,12]. This is not always a straightforward task due to the discrepancy between the methods used to probe the specic property [13] and the complexity of the porous structure itself [10,14]. The structure of cellular materials has been essentially classied into sponges and foams; the former having open cells with beams as cellular connections, and the latter having closed cells, where cell faces must be considered instead [4,15,16]. The porous structure has been modeled after regular honeycomb and irregular Voronoi tessellations in two dimensions [17,18] and by regular tetrakaidecahedral foams and Voronoi tessellations in three dimensions [12,19]. Finite element analysis (FEA) has been extensively and successfully employed to simulate the mechanical response of the modeled structure under tension or compression. It is a particularly useful technique for assessing the effect of imperfections in the structure, since non-uniformity or irregularity of the cell wall can be taken into account during computation [2023]. Non-periodicity in the cellular structure of foams has also been characterized by the fractional dimension (FD) [2]. For certain biogenic cellular solids, gas cell structure is produced through random physical and biochemical processes. Therefore, the existence of a fractional dimension that characterizes the structure of the cellular solid is apt for prediction of the elastic modulus, as has been observed for silica aerogels [24]. Indentation is a convenient means of assessing several mechanical properties of solids including the elastic modulus [2527]. It is particularly useful for probing the properties of soft materials where conventional mechanical tests may not be employed due to small specimen size or low stiff-

ness [28], and where minimal specimen preparation ensures that the native structure is undisturbed to prevent experimental artefacts [29]. With cellular solids, though, a number of factors have had to be explored to ascertain the conditions under which indentation will furnish the bulk mechanical properties, as would be the case from a standardized compressive or tensile test. For instance, it has been stated that the ratio of indenter diameter to cell diameter is a key factor affecting determination of the modulus of elasticity and the plateau stress of aluminum foams [1,30]; this ratio should be at least six to minimize discrepancies between indentation and compression measurements [31]. Also, radial and axial dimensions of the specimens should be a number of times greater than the indentation dimensions (for instance, the radius of cylindrical at-ended indenters), in order to discount complications arising from end-effects [2527]. The purpose of this study was to determine the conditions under which bulk measurements of the elastic modulus of a biogenic cellular solid could be extracted from indentation measurements, and then to relate the modulus to the characteristics of the structure formed during processing. The cellular solid was a hydrated starch foam (a portion of the crumb excised from bread loaves). This soft solid has a moisture content of approximately 40% [32] and consists of a network of partially melted starch spherulites [33] admixed with smaller amounts of other polymers and small molecules [34].

2. Materials and methods Hydrated starch foam specimens were prepared from bread loaves made from Canadian western red spring (CWRS) wheat our, according to a short dough process [35]. To obtain foams with a range of densities, proong times of 15, 35, 55, 70 and 85 min, and a baking absorption of 63.9% were used. Longer proof time permits the biological blowing agent (yeast) to generate more CO2 and expand the gas cell nuclei, although gas cell coalescence is an issue at longer proof times [36]. Baked bread was sealed in double polyethylene bags after cooling, and conditioned overnight at

Z. Liu et al. / Acta Materialia 51 (2003) 365371

367

ambient temperature to allow for moisture equilibration and stabilization. The baking order for different proong times was randomly applied and each proong time was replicated at least three times for each mechanical test, over a nine-week period using a new batch of compressed yeast each week. Just before the mechanical testing (1624 hr after baking), approximately 25 mm-thick slices were excised from the central portion of the loaf being investigated [32] using a serrated knife and digital calipers. Gentle and even hand movement was applied to minimize disrupting the structure. All slicing for one loaf was completed within 5 min, and the slices were immediately sealed in a zipped polyethylene bag. Just prior to indentation testing, a slice was removed and immediately indented; for compression testing the cylindrical specimen was excised using a sharp 30 mm-diameter stainless steel borer. The heights of the cylinders were measured by the calipers at four points; specimens were weighed on a digital balance immediately after cutting. Foam density (r) was obtained from the weight divided by specimen volume. Indentation and compression tests were performed on a computerized Lloyd testing machine (Model L1000R) at ambient conditions (23 2 C, ambient humidity). The 100 N capacity load cell was calibrated with a 2 kg standard weight. For indentation, cylindrical at-ended indenters with diameters of 3, 6, 12, and 36 mm, and a at aluminium bottom platen (120 mm in diameter) were used. Two at aluminium platens (120 mm in diameter) were used for compression. Compressive elastic modulus (E) was extracted from engineering stress-strain curves [37]. From indentation load (F)displacement (d) curves (Fig. 1), the apparent elastic modulus E was calculated by [27,38]: ia
Eia

Fig. 1. Indentation force-displacement curves of starch foams of different density indented by a 36 mm-diameter cylindrical indenter.

a(1 n2) 2R

(1)

where a dF / dd, denoting the slope in the linear elastic regime; R the indenter radius, and n the Poissons ratio, taken as 0.21 from experimental measurements at low strain [39]. All image acquisition was performed with a customized PC image analysis system in our depart-

ment [40]. The frame grabber provides 256 grey level (GL) digital images (640 480 pixels). Four incandescent tungsten-lament frosted envelope lamps in a ring conguration provided illumination. For normalization of the imaging systems GL response a white opal acrylic plastic plate (2 mm thick) was used as a working standard, and this was calibrated to reectance using a 12 step grey scale (catalogue #152 2267, Kodak, Rochester, NY). A central area of interest on the reectance standard was repeatedly captured with adjustment of the aperture until the computed mean GL corresponded to a predetermined target GL value (160), which corresponded to 79% reectance. To objectively determine a threshold for accurate delineation of cells and cell walls and subsequent quantication of cell features, image segmentation was employed, using a cluster analysis technique (K-means algorithm). The threshold GL for all our samples was almost identical: a mean value of 128 (coefcient of variation: 7.4%), which was taken as the intensity threshold for segmenting the cells and cell walls, using SigmaScan Pro 5 (SPSS Inc.). Images corresponding to the full range of starch foam densities were selected for fractal analysis, with values obtained for a narrow range in density being averaged. For each image, the number of cells was counted on an overlay (a monochrome image) of cells (intensity threshold 0128). The fractional dimension (FD) of cells was computed based on the box counting method [41]. The total number of cells (N) was counted on the overlay of

368

Z. Liu et al. / Acta Materialia 51 (2003) 365371

width (W) and height (W). A scale number of n was set. The overlay was scaled down (cropped) repeatedly by a xed step, i.e. (W/n) for width and (W/n) for height (to maintain a constant () ratio of height to width), and then N was recounted. According to fractal theory, from the n sets of W N values, N is a power law function of W with an exponent of FD: N WFD (2)
Fig. 2. Effect of indenter diameter (D) on the apparent elastic modulus (E ). ia

The slope of a plot of log(N) against log(W) was taken as the value of FD for cells. To determine the FD of the cell walls, N in Eq. (2) was substituted by the total area of cell walls (A). Area measurements of cell walls in an image were done on an overlay of cell walls (intensity threshold 128255). We coded macros using Visual Basic for Applications (VBA) working with the software itself to automate the computation.

3. Results and discussion 3.1. Deriving Youngs modulus from indentation tests Fig. 1 presents typical force-displacement curves, measured using a 36-mm indenter, which is the specied standard for quality evaluation of baked products [42,43]. Some of the curves showed a non-linear region up to a displacement of 1.0 mm (apparent for the higher density foams in Fig. 1); non-linearity arises from incomplete contact of the indenter with the foam surface. The indentation elastic modulus (E), which was i obtained from the slope of the linear elastic region (which typically occurred up to a displacement of 2.0 mm), increased with increasing foam density. This is consistent with observations of simple/unconstrained compression in other foams [4]. In general, the apparent elastic modulus, E , ia increases almost linearly with increasing indenter diameter, and the extent of the increase rises with increasing density (Fig. 2). The assumption of a semi-innite elastic medium associated with Eq. (1) is responsible for the rise in Eia with increase in indenter size [44], while the dependence of Eia on density is likely attributable to a change in Pois-

sons ratio as density changes with processing conditions [45]. Corrections for Poissons ratio and the ratio of indenter radius to specimen depth were performed by dividing the right hand side of Eq. (1) by the (n, a/h) values taken from a nite element simulation of indentation of a circular indenter into a soft material on a rigid substrate [27]. Conservative estimates for the change in Poissons ratio were 0.1. The comparison of compressive and corrected indentation moduli is shown in Fig. 3 for the two largest indenters. The indentation moduli for the smaller indenters are not shown due to their larger variability (mean CV is 33%, compared with 22% for the two larger ones). This variability is likely a result of cell size effects [12,30], since average cell size was 1.01 mm, and for these small indenters, the indenter to cell size ratio would not exceed 6, a value necessary to ensure good agreement between moduli from

Fig. 3. Comparison of the corrected indentation elastic modulus (E) and the compressive elastic modulus (E). i

Z. Liu et al. / Acta Materialia 51 (2003) 365371

369

indentation and compression tests [31]. Therefore, having established that valid modulus measurements for the starch foam can be obtained from a fairly common technique employed throughout the food industry, the next step was to understand how the modulus for the hydrated starch foam relates to its structure as dictated by variation in processing conditions. In the following, the compressive elastic modulus was used. 3.2. The modulus-structure relationship It has been shown previously that density scaling relationships based on beam bending theories [4] are applicable for predicting the modulus of hydrated starch foams in compression [16] and tension [46]. Based on the 42.0 weight% moisture content (MC) in the starch foam [47], the elastic modulus of the cell walls (Es) was assigned as 3.0 MPa, which falls within the range of elastic modulus ( 4.0 MPa) of thermoplastic starch with more than 25% MC [48] and freshly prepared starch gels (2.0 MPa) with 49.2% MC [49]. The density of cell wall (rs) was assigned as 1500 kg/m3 [50]. A power law dependence was observed (Eq. (3)) when relative elastic modulus (E/Es) was plotted against relative density (r/rs) (Fig. 4), as suggested by the theory of the behaviour of periodic cellular structures [4] and by percolation models of disordered foams [2]. E Es 1.05 r rs
2.54

Although the constant was found to be quite close to unity [4], the exponent (n 2.54 0.32) obtained by power law tting lay between the expected values for idealized regular open-cell cellular solids (n 2) [51] and prismatic closedcell cellular solids where solid material is distributed evenly throughout the cell wall (n 3) [51,52]. Because the starch foam arises from random natural processes, interpreting its structure on the basis of periodic cellular models may not be as instructive as using models that incorporate disorder [19]. The disorders include non-periodicity of cell arrays [18], uneven distribution of solids over the cell walls [52], cell wall corrugations [22,53] and thickness variations [54], coalescence of cells [55], and cell irregularity [20,23]. All these defects can signicantly alter the elastic properties of cellular solids, and therefore change the exponent and constant of the power law function of relative Youngs modulus against relative density. Modeling work on cellular structures has shown that the exponent will vary between 1 and 2 for closed-cell [19], and between 1.3 and 3 for opencell [12] solids, respectively. Non-periodicity in foam structure is apparent from a 2D section of the starch foam (Fig. 5). We obtained a FD value of 1.89 0.07 for gas cells, a value which is independent of foam density (Fig. 6), conrming the random distribution of cell sizes. This result concurs with the theoretical value of 1.896 for twodimensional representations of porous materials,

(3)

Fig. 4. Relationship between relative elastic modulus (E/Es) and relative density (r/rs).

Fig. 5. Digital image of starch foam structure (r g/cm3).

0.115

370

Z. Liu et al. / Acta Materialia 51 (2003) 365371

the elastic constant of the links between the microstructures [58,59]. This theory has been shown to apply to fat crystal networks where the properties are dominated by the fractal nature of the microstructure [41]. Assuming that the dominant microstructure in the cell walls of the foam is the sintered spherulties, then scaling theory can be used to relate the elastic properties of the foam to its volume fraction of microstructural elements: E(j) jm
Fig. 6. Relationship between foam density (r ) and fractional dimension (FD) of cell numbers and cell wall areal fraction for the structure of hydrated starch foams.

(4)

which has been observed for aluminium foams where foaming agents create a random distribution of cell sizes [2]. Therefore, because the models developed for symmetrical structures are not fully applicable here, the scale at which the actual microstructure effects (sic) the elastic properties [56] was examined. Recent phase-contrast microscopic evidence conrms that the starch spherulites have their long axes aligned parallel to the gas cell surface [34], but that they are only partially melted; leaching of the linear polymer, amylose, from them during processing causes a sintering of the spherulites. In such a structure, where the cell walls of the starch foam are predominantly composed of numerous swollen starch granules joined to each other by a thin layer of amylose, the elastic properties of the starch foam will be dominated by the compliance of the amylose regions between the spherulites. FD values of the cell walls for foams with various densities are shown in Fig. 6. As with the result for gas cells, and in agreement with theoretically derived results for the cellular backbone of aerogels [24], the FD values were independent of foam density. An average FD value of 2.41 0.16 for cell walls lies close to the FD value of 2.544 calculated from scanning laser microscope surface height measurements of white bread slices [57]. To relate this parameter, that characterizes the structure of the foam, to its elastic modulus we employed a weaklink regime theory which presumes that the elastic constants of a network system are dominated by

where j is the volume fraction (identical to r/rs), and, for the weak link regime [58], where the elastic properties of the foam are dominated by the modulus of the links between the spherulites, the exponent m is equal to (d-2)/(d-FD), or, 1/(3-FD) since the foam is three-dimensional. Eq. (4) can be rearranged to the same form as Eq. (3): E Es r rs
m

(5)

where the symbol denotes approximately proportional to. Applying the exponent value of 2.54 0.32 from the modulus-density relationship (Eq. (3)) into m of Eq. (5) gives an FD value in the range of 2.55~2.65, which overlaps the cell wall FD values (2.41 0.16 or 2.25~2.57) measured by image analysis. The good relationship between the exponents obtained from microstructural analyses and from mechanical analyses validates the applicability of a weak-link regime theory for relating the microstructure of the hydrated starch foam to its elastic modulus.

4. Conclusions Within experimental uncertainty, the corrected indentation elastic modulus (corrected as a function of Poissons ratio and aspect ratio) was identical to the compressive elastic modulus of starch foam. This suggests the applicability of the indentation test as an alternative to compression, and this is particularly advantageous when determining the properties of soft solids such as hydrated starch foams. The relative elastic modulus of the starch

Z. Liu et al. / Acta Materialia 51 (2003) 365371

371

foam was a power law function of relative density, with an exponent of 2.54. Fractal analysis of the images of the starch foam showed that it had a random but self-similar cellular structure independent of the foam density, with a fractional dimension for cell numbers of 1.89 0.07. Applying a weak-link theory to understand the density dependence of the elastic modulus permitted us to understand the relationship between elastic properties and the microstructure of the solid material. References
[1] Onck PR, Andrews EW, Gibson LJ. Int J Mech Sci 2001;43:681. [2] Kovacik J, Simanck F. Scripta Mater 1998;39:239. [3] Duarte I, Banhart J. Acta Mater 2000;48:2349. [4] Gibson LJ, Ashby MF. Cellular solids: structure & properties. Oxford: Pergamon Press, 1988. [5] Aguilera JM, Stanley DW. Microstructural principles of food processing and engineering. Gaithersburg, MD: Aspen Publishers, Inc, 1999. [6] Warner M, Thiel BL, Donald AM. Proc Natl Acad Sci 2000;97:1370. [7] Kamyab I, Chakrabarti S, Williams JG. J Mater Sci 1998;33:2763. [8] Weiss G. Science 2000;293:1753. [9] Simone AE, Gibson LJ. Acta Mater 1998;46:3109. [10] Pajonk GM, Venkateswara Rao A, Parvathy NN, Elaloui E. J Mater Sci 1996;31:5683. [11] Roberts AP, Knackstedt MA. Phys Rev E 1996;54:2313. [12] Roberts AP, Garboczi EJ. J Mech Phys Solids 2002;50:33. [13] Fusheng H, Zhengang Z. J Mater Sci 1999;34:291. [14] Roberts AP, Knackstedt MA. J Mater Sci Lett 1995;14:1357. [15] Keetels CJAM, Visser KA, van Vliet T, Jurgens A, Walstra P. J Cereal Sci 1996;24:15. [16] Keetels CJAM, van Vliet T, Walstra P. J Cereal Sci 1996;24:27. [17] Silva MJ, Gibson LJ. Int J Mech Sci 1997;39:549. [18] Silva MJ, Hayes WC, Gibson LJ. Int J Mech Sci 1995;37:1161. [19] Roberts AP, Garboczi EJ. Acta Mater 2001;49:189. [20] Zhu HX, Hobdell JR, Windle AH. J Mech Phys Solids 2001;49:857. [21] Fortes MA, Ashby MF. Acta Mater 1999;47:3469. [22] Grenestedt JL. J Mech Phys Solids 1998;46:29. [23] Zhu HX, Hobdell JR, Windle AH. Acta Mater 2000;48:4893. [24] Emmerling A, Fricke J. J Sol-Gel Sci Technol 1997;8:781. [25] Briscoe BJ, Sebastian KS, Adams MJ. J Phys D: Appl Phys 1994;27:1156.

[26] Hiller S, Bruce DM, Jeronimidis G. J Texture Stud 1996;27:559. [27] Zhang M, Zheng YP, Mak AF. Med Eng Phys 1997;19:512. [28] Goh SM, Charalambides MN, Williams JG. Zurich: Laboratory of Food Process Engineering, ETH, 2000. In: Proceedings of the 2nd International Symposium on Food Rheology and Structure. [29] Turner CH, Rho J, Takano Y, Tsui TY, Pharr GM. J Biomech 1999;32:437. [30] Andrews EW, Gioux G, Onck P, Gibson LJ. Int J Mech Sci 2001;43:701. [31] Miller RE. Int J Mech Sci 2000;42:729. [32] Zghal MC, Scanlon MG, Sapirstein HD. Cereal Chem 2001;78:1. [33] Butler MF, Donald AM. J Mater Sci 1997;32:3675. [34] Hug-Iten S, Handschin S, Conde-Petit B, Escher F. Lebensm Wissen Technol 1999;32:255. [35] Preston KR, Kilborn RH, Black HC. Can Inst Food Sci Technol J 1982;15:29. [36] Scanlon MG, Zghal MC. Food Res Int 2001;34:841. [37] Attenburrow GE, Goodband RM, Taylor LJ, Lillford PJ. J Cereal Sci 1989;9:61. [38] Timbers GE, Staley LM, Watson EL. Agr Eng 1965;46:274. [39] Rohm H, Jaros D, Haan M. J Texture Stud 1997;28:245. [40] Sapirstein HD, Roller R, Bushuk W. Cereal Chem 1994;71:383. [41] Narine SS, Marangoni AG. Phys Rev E 1999;59:1908. [42] American Association of Cereal Chemists. Approved Methods of the AACC, Method 74-09. The Association, St. Paul, MN, 2000. [43] Baker AE, Ponte JG. Cereal Foods World 1987;32:491. [44] Hayes WC, Keer LM, Herrmann G, Mockros LF. J Biomech 1972;5:541. [45] Arnold M, Boccaccini AR, Ondracek G. J Mater Sci 1996;31:1643. [46] Zghal MC, Scanlon MG, Sapirstein HD, J Cereal Sci 2002;36:167. [47] Liu ZQ, Scanlon MG. Cereal Chem 2002;79:763. [48] van Soest JJG, Benes K, de Wit D. Polymer 1996;37:3543. [49] Rolee A, Le Meste M. Cereal Chem 1997;74:581. [50] Liu ZQ, Yi X-S, Feng Y. Starch/Starke 1999;51:406. [51] Gibson LJ, Ashby MF. Proc R Soc Lond Ser A 1982;382:43. [52] Simone AE, Gibson LJ. Acta Mater 1998;46:2139. [53] Simone AE, Gibson LJ. Acta Mater 1998;46:3929. [54] Grenestedt JL, Bassinet F. Int J Mech Sci 2000;42:1327. [55] Guo XE, Gibson LJ. Int J Mech Sci 1999;41:85. [56] Brechet Y, Deschamps A, Verdier M. J Phys IV 1998;8:3. [57] Pedreschi F, Aguilera JM, Brown CA. J Food Process Eng 2000;23:127. [58] Shih W-H, Shih WY, Kim S-I, Liu J, Aksay IA. Phys Rev A 1990;42:4772. [59] Wu H, Morbidelli M. Langmuir 2001;17:1030.