Colloids and Surfaces A: Physicochemical and Engineering Aspects

Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 Contents lists available at ScienceDirect Colloids and Surfaces A: Physicochemical and Engineering Aspects journal homepage: www.elsevier.com/locate/colsurfa A unique microstructure of the fiber networks deposited from foam fiber suspensions Ahmad M. Al-Qararah a, Axel Ekman b, Tuomo Hjelt a, Jukka A. Ketoja a,, Harri Kiiskinen a, Antti Koponen a, Jussi Timonen b a VTT Technical Research Centre of Finland Ltd., P.O. Box 1000, FI-02044, VTT, Finland b University of Jyvaskyla, Department of Physics and Nanoscience Center, P. O. Box 35, FI-40014, University of Jyvaskyla, Finland h i g h l i g h t s g r a p h i c a l a b s t r a c t Foam and water deposited porous fiber structures differ at equal sheet density. Bubble-size distribution of the foam affects the mean pore size. The effect of bubble size on the porous structure is strongest for stiff fibers. Strength and other macroscopic sheet properties are affected by foam properties. a r t i c l e i n f o Article history: Received 30 April 2015 Received in revised form 3 July 2015 Accepted 6 July 2015 Available online 10 July 2015 Keywords: Fiber network Pore Foam Bubble X-ray tomography Structure Strength a b s t r a c t Fiber networks can be formed using aqueous foam as the suspending medium. The mean bubble size of the foam affects the resulting pore-size distribution of the fiber network. The foam fiber interactions cause in particular an increase in the proportion of large micropores of the network, in comparison with the fiber networks that result from traditional water forming at a similar material density. Experiments were carried out for two different types of cellulose fiber, and characterization of the resulting pore structure was based on X-ray microtomography of the resulting fiber networks. The unique pore structure obtained with foam forming was reflected in various macroscopic properties of the networks, which provides an intriguing opportunity to control the material properties of fiber networks via the selection of their forming. 2015 Elsevier B.V. All rights reserved. 1. Introduction Corresponding author. E-mail address: jukka.ketoja@vtt.fi (J.A. Ketoja). Aqueous foam [1,2] consists of a close packing of gas bubbles in a continuous liquid phase. With added fibers, such a system provides an interesting suspending medium, which forms when removing the liquid phase, i.e., by a deposition process, a fiber network simhttp://dx.doi.org/10.1016/j.colsurfa.2015.07.010 0927-7757/ 2015 Elsevier B.V. All rights reserved.

A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 545 ilar to those found in paper making. This provides excellent sheet homogeneity and enables the increase of solids content of the fibre furnish during the forming operation [3 6]. In addition to that, fibre materials with low density can be obtained [5 6]. The main drawback of this technology is a decrease in strength. To address the loss of strength in foam-formed sheets, it has been demonstrated that the strength can be regained by using micro-fibrillated cellulose (MFC) without losing the obtained bulky structure [7]. Foam-formed fibre structures have received an increasing interest recently as sustainable solutions for a wide range of material applications [5]. There is a long-lasting tradition in paper physics to investigate (when water is the liquid phase) the factors that affect the microstructure of the network [8,9]. These factors have ranged from properties of raw materials (e.g., fiber dimensions, fiber flexibility, their external fibrillation, and the fine-particle content of the furnish) to various process factors (e.g., concentration and rheology of the fiber suspension, fiber orientation, drainage, wet pressing, drying). In water forming, a great deal of effort has been put to creation of a turbulent suspension, before its deposition, which helps achieve a homogeneous fiber network in small (about a fiber length or a couple of millimetres) scale. Otherwise, the characteristics of the suspending medium as such has not been considered much in the paper-making literature. This paper goes beyond the current paradigm (water forming) in paper making and analyzes the role of the suspending medium in the structure of the resulting fiber network. To this end, porous structures obtained by foam forming are compared with those obtained using a water suspension. It has been already suggested that the porous structure obtained with foam forming could differ significantly from that obtained with a traditional water forming [10,11]. However, the two forming methods have been previously compared only as separate technologies, i.e., without any attempt to avoid the contribution of the density of the final (or targeted) structure. In general, density of the resulting fiber network will depend on several factors, like the type of fiber, the vacuum level in drainage, and the load used in wet pressing [8]. In this work, the structures resulting from the two forming methods were compared for a similar final density of the network. This required in fact that no vacuum was applied in the water forming, whereas a normal 0.5 bar vacuum level was applied in the foam forming. Moreover, no wet pressing was applied in either process so as to leave a maximal trace of the foam-fiber interactions in the resulting network structure. Here the word interaction means any mutual influence between the two phases (foam and fibers) [12 16]. This influence may also be of geometric origin so that the bubble structure limits the orientation and movement of the fibers and vice verse. Characterization of the structure of the fiber network has earlier been mainly based on (two-dimensional) visible-light microscopy or various indirect measurements. In recent years, however, modern three-dimensional (3D) imaging methods have opened a new avenue for the structural characterization of the network [17,18] together with a possibility to simulate material properties of the real network structure [19]. Here, characterization of the network structure was based on X-ray microtomography of the networks. In addition to determining the pore-size distributions of these networks, we measured several of their macroscopic properties, and various properties (e.g., bubble size) of the suspending medium. Because of absence of wet pressing and vacuum dewatering (drainage) in the case of water forming, the densities of the samples remained in the range 100 300 kg/m 3. In Chapter 2 we explain the experimental forming procedures and characterization methods. Chapter 3 reports the differences in the pore-size distributions for the foam-formed and water-formed samples. The effects of the microscopic structural parameters on macroscopic physical properties are discussed in Chapter 4. In Chapter 5, some general conclusions are drawn based on the results. 2. Experimental 2.1. Fiber materials Pre-refined bleached kraft pulp (Scots pine, coarseness 142 g/m, freeness 607 ml) and chemi-thermo-mechanical pulp (CTMP; Norway spruce, coarseness 220 g/m, freeness 570 ml) were obtained from Finnish paper mills. The coarseness gives the dry fibre mass per unit length, whereas the freeness of a pulp is an indication of its dewaterability. The average (length weighted) length of the kraft fibers, as measured with a common fiber-analysis device (an optical measurement), was 2.2 mm [4]. The average length of the CTMP fibers was 1.6 mm [4]. For other fiber properties, we refer to typical values found in the literature for similar fiber materials [8]. The cross-sectional shape of the pre-refined wet kraft fibers was quite circular with an average diameter in the range 35 40 m. For the CTMP fibers, with similar shape, the diameter was about 30 m. Thus, the overall dimensions of the fibers in the two furnishes were quite similar. The main difference between the qualities of these two types of fiber was the bigger conformability (a higher level of their shear deformation seen as more bending) of the kraft fibers, which gave rise to a higher density of the resulting fiber network [8]. 2.2. Preparation and characterization of the foam fiber suspension Pulp furnish (kraft pulp or CTMP) was diluted with distilled water to 0.33% consistency. Here, the term consistency denotes the mass fraction (or its percentage) of the solid and filterable materials in the suspension. The furnish was made such that three liters of a diluted furnish (10 g solids) was mixed with 0.60 g (kraft) or 0.90 g (CTMP) anionic surfactant (Sodium Dodecyl Sulphate (SDS), C 12 H 25 SO 4 Na). Thus, the ratio of added SDS was 0.20 g/l (kraft) or 0.30 g/l (CTMP). SDS was obtained from Sigma Aldrich and its purity was 90%. A different amount of the foaming agent was needed in order to achieve a high enough air content of the respective foams. At 0.20 g/l, the air content of the CTMP foam remained at a low level (c.a. 60%). A foam fiber suspension was generated by axially agitated mixing in a cylindrical vessel (diameter 200 mm) with three impellers Table 1 Mean bubble size, r 32, and air content, ϕ, for various rotation speeds of a pure foam (SDS concentration 0.3 g/l) and the foam fiber suspensions studied. Air content was determined from the final foam volume V f and the initial water amount V w, ϕ = (V f V w)/v f. The error in V w was very small as the initial water amount was measured by weighing. Thus, the absolute error in the measurement of the air content remained on the low level ±0.004 in all cases. Rotation speed (RPM) Pure foam CTMP Kraft r 32 ( m) ϕ r 32 ( m) ϕ r 32 ( m) ϕ 2000 95.6 ± 2.1 0.643 93.8 ± 3.5 0.655 99.3 ± 2.6 0.655 3500 72.5 ± 2.1 0.681 70.5 ± 2.0 0.688 56.6 ± 1.4 0.684 5000 61.2 ± 1.3 0.737 48.7 ± 1.0 0.709 48.8 ± 1.4 0.714 6900 39.2 ± 0.9 0.758 34.4 ± 0.8 0.639 39.6 ± 1.1 0.647

546 A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 Fig. 1. Manufacturing of a handsheet, in which wet foam was used as the suspending medium. (a) The foam fiber suspension was decanted into the handsheet mold and filtrated through a forming fabric using a vacuum chamber. (b) The sheet together with the fabric was detached from the mold, after which the sheet was pre-dried on the suction table (vacuum 0.2 bar). The size of the sheet was 38.5 cm 26 cm (0.1 m 2 ). (diameter 80 mm) [12]. Rotation speed was set to 2000, 3000, 5000, or 6900 rev/min (RPM) in order to vary the properties of the foam. When the foam reached its maximal air content, the top surface of the foam did not rise any more, and a steady-state bubble-size distribution was formed. Its air content was determined from the final volume of foam and the initial amount of water (see Table 1). The foam surface became even within a couple of seconds after the mixing was stopped, and the surface height of the foam was read immediately after that. The bubble-size analysis of the foam was based on foam samples taken from a controlled position about 2 cm below the foam surface after its air content had stabilized (after about 15-min mixing period). The foam samples were sucked into a narrow gap between two glass plates. An image of the foam was then taken with a CCD camera within 10 20 s from taking the sample. The resolution of these images was 2448 2048 pixels and their typical absolute size was 1.8 1.5 mm 2. The typical number of bubbles measured for their average size was approximately 250. Properties of the foam were sensitive to its mixing conditions. In earlier publications [12,13], a simple model was developed to describe the effect of air content, ϕ, surface tension,, and rotation speed, N, on the Sauter mean bubble size, r 32 : r 32 = ir 3 i i r 2 i N( 1 1) where r i are the bubble radii. The air contents of the foams were practically equal (their deviation was close to the measurement error) for the two types of fiber, when mixed with a similar speed, see Table 1. However, the drop in bubble size [13] between rotation speeds 2000 RPM and 3500 RPM was clearly bigger for kraft fibers than for CTMP fibers. For rotation speed 6900 RPM, the bubble size of the foam was less than a half of that for rotation speed 2000 RPM, for both types of fiber. The systematic behavior observed for the mean bubble size as a function of rotation speed helped analyze the effect of bubble size on the structure of the resulting fiber network (sheet). 2.3. Manufacturing of foam-deposited ( foam-formed ) and water-deposited ( water-formed ) laboratory sheets (1) The procedure of making the foam-formed laboratory sheets is shown in Fig. 1. The size of the sheet thus obtained was 38.5 cm 26 cm (0.1 m 2 ), and the target grammage was 100 g/m 2. When the steady state of the foam fiber suspension was achieved, it was quickly decanted into the handsheet mold (Fig. 1a) and filtrated through a normal papermaking forming fabric [20] using a vacuum chamber (vacuum level was 0.5 bar for the initial dewatering). A slightly prefered orientation of the fibers in the direction of the foam flow took place in this process. This orientation was visible as the ratio of tensile strength in the machine direction (direction of the flow) to the cross-machine direction deviating from unity (1.08 1.18 for kraft fibers). The filtrated sheet was detached together with the forming fabric from the mold (Fig. 1b) and pre-dried on the suction table. The suction table had a slit of 5-mm width, and air was sucked through the sheet using this slit with a 0.2-bar vacuum. Each sheet, while laying on the fabric, was slided over the slit for five times (sliding took about 5 sec). The pre-dried, but still wet sheets, were finally dried between a metal plate and a drying fabric for one night in an air tunnel (at 23 C, 50% RH). Water-formed sheets were manufactured in a similar fashion, but no vacuum was used during filtration. The furnish and target grammage were the same as for foam forming. These sheets were made with a standard handsheet mold. Their size was 16.5 cm 16.5 cm, and the initial consistency was 0.045%. During the filtration state, the same forming fabric was used as when manufacturing the foam-deposited sheets. Their consolidation was driven only by the hydrostatic pressure (initially about 0.02 bar) caused by the height of the suspension column (initially 20 cm). Because of absence of suction in the forming state, the consolidation time of water-deposited sheets was roughly twice longer than that of foam-deposited sheets. A uniform in-plane orientation of fibers was expected for this type of water-deposited sheets. Predrying on the suction table and final drying in the air tunnel were done in exactly the same way as for the foam-deposited sheets. We prepared two parallel sheets for each trial point so that one sheet was used for microscopic characterization and the other parallel sheet for macroscopic physical measurements. The sheet thickness was obtained as the average of 10 measurements from each sheet, see Table 2. The measurements were done with equal spacing in order to represent the whole area of the sheet. Table 2 shows the grammage (mass/area) and the density of the (initially) foam-deposited and water-deposited sheets used for the macroscopic measurements. Application of different vacuum levels during drainage of the two suspending media (foam and water) enabled us to get roughly equal sheet density in these cases. The water-deposited kraft sheets were slightly denser than the corresponding foam-deposited sheets. For CTMP, sheet density was almost independent of the forming method. For the foam forming, a small variation of the sheet density for a varying rotation speed was observed for both types of fiber.

A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 547 Table 2 Basic sheet properties for water-deposited and foam-deposited sheets. In foam forming, the surfactant concentration was 0.2 g/l for the kraft pulp and 0.3 g/l for the CTMP pulp. The thickness measurement was carried out with a Lorentzen and Wettre device (App 141, Type 1-1, Nr 474). Fiber type Rotation speed (RPM) Grammage (g/m 2 ) Thickness ( m) Density (kg/m 3 ) CTMP 2000 84 835 ± 21 100 ± 3 3500 97 846 ± 18 115 ± 2 5000 84 753 ± 16 111 ± 2 6900 100 864 ± 21 116 ± 3 Water 96 895 ± 26 107 ± 3 KRAFT 2000 98 559 ± 18 175 ± 6 3500 110 557 ± 20 197 ± 7 5000 104 537 ± 18 194 ± 7 6900 112 521 ± 18 215 ± 8 Water 101 420 ± 21 241 ± 12 2.4. Characterization of the porous sheet structure X-ray tomographic imaging was done of a 2 mm 3 mm sample area of the formed sheet. Before taking the image, it was checked that the studied region was representative in terms of homogeneity and density (formation) by comparing it visually with the rest of the sheet. One image for each experimental case was thereafter taken with a 1 m resolution (in all directions). Fig. 2 shows typical images obtained for water-formed and foam-formed samples. Differences between foam-formed and water- formed sheets were expected to be clearest in their respective bulk structures. Therefore, the analysis was carried out for a rectangular central region where sheet surfaces and edge voids were excluded. In other words, neither the wire patterning nor the top-surface roughness should have played any role in the pore structures finally analyzed. From the X-ray tomography images (reconstructions), we determined the number of voxels that belonged to a certain class of pore size. The pore-size distributions were obtained by determining the local thickness of the void space. The local thickness at a point is the diameter of the largest sphere that contains that point and is completely inside the void structure. The normalized frequency distributions of these diameters gave the volume-weighted pore-size distributions. The above analysis was carried out with three different threshold gray values. In a subsequent parametric distribution fitting, 95% confidence intervals of all parameters were determined. It turned out that the variation of fitting parameters with changed gray level threshold generally remained within the confidence interval for the central threshold. In Chapter 3, the results for the central gray level threshold are presented only. The error estimates are obtained from the respective confidence intervals. 2.5. Measurement of the macroscopic sheet properties In addition to sheet grammage, thickness, and density, we carried out measurements on macroscopic surface properties, optical properties, and strength of the sheet. The standard Bendtsen roughness was measured with a Lorentzen and Wettre SE114 device [21], and was based on the amount of air flow in the surface structure of the sheet. For very porous structures investigated here, we can expect a significant amount of air flow within the pore structure and surface topography (roughness) of the sheet. Thus, the result of this measurement will depend not only on roughness, but also on the pore structure of the sheet. Sample was placed on a glass plate under a weight, which had a hole in the circular measurement area, where the flow of (pressurized) air was recorded. Ten parallel measurements for each side of the sample were done for the average. The opacity measurements were performed with a Lorentzen and Wettre SE070R Elrepho spectrophotometer according to related standards [22]. In order to see the effect of microstructure of the sheet on its strength, the tensile-strength measurements of the sheet were performed with an Alwetron TH1 horizontal tensile tester in a standard way [23]. Twelve 15-mm-wide strips from each handsheet were tested. Six strips were cut in the machine direction (MD) and six in the cross-machine direction (CD). In the paper industry the tensile index, i.e., tensile strength (N/m) divided by the sheet grammage (g/m 2 ), is commonly used to get rid of the (main) variation depending on the basis weight of the sample. In a similar way, the elasticity is characterized by the tensile stiffness index (initial slope of the stress strain curve in units N/m divided by grammage). The tensilestrength measurements were carried out at 23 C and 50% RH. 3. Comparison of pore-size distributions for varying forming conditions In this Chapter, we compare the volume weighted pore-size distributions, or the probability density functions (PDFs), of waterformed and foam-formed sheets. In order to have a better picture of the characteristic features, each distribution was fitted by a linear combination of the Gaussian, g(r), and log-normal, lognorm (r), distributions with coefficient A: f (r) = Ag(r; 1, 1 ) + (1 A)lognorm(r; 2, 2 ) (2) here 1,2 and 1,2 are the fitting parameters that describe the mean and standard deviation of g(r) or similar parameters for the natural logarithm in the case of log-normal distribution. The log-normal distribution was included in Eq. (2) as this distribution is characteristic for deposited fibre networks [8,9,24,25]. On the other hand, bubble size distributions in mixing have been earlier found to be Gaussian [13], and the foam-fibre interaction could give rise to a similar component for the the pore sizes as assumed in Eq. (2). These distributions together with their fits for water-formed sheets are shown in Fig. 3 for both CTMP and kraft fibers. Waterformed sheets were dominated by the log-normal component as indicated by the low value of coefficient A (see Table 3). This was in agreement with earlier analysis of traditional paper structures [24,25], where the log-normal distribution of pore sizes has been found with a standard deviation proportional to the mean. A lognormal distribution follows from a random deposition process of fibers that takes place when the consistency is very low. However, it is interesting that, in addition to the the overall log-normal decay, there was a very small Gaussian component, at about 40 m for the CTMP fibers and at about 30 m for the kraft fibers, as indicated by Fig. 3. Although this size range was close to that of the average diameter of fibers, the exact location of the Gaussian component was higher for the narrower CTMP fibers, which prevents any easy explanation of this small component.

548 A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 Table 3 Parameters of the fitted pore-size distributions (see Eq. (2)) for both foam-formed and water-formed CTMP and kraft sheets. The last column gives the Sauter mean pore radius. The errors are given by the 95% confidence intervals. Fiber type Medium A 1 1 2 2 r 32 CTMP Foam, 2000 RPM 0.79 ± 0.03 66.3 ± 3.3 43.9 ± 1.8 2.9 ± 0.1 0.67 ± 0.06 63.7 ± 1.3 Foam, 3500 RPM 0.74 ± 0.03 80.0 ± 2.7 42.5 ± 1.9 2.9 ± 0.1 0.69 ± 0.04 57.5 ± 0.8 Foam, 5000 RPM 0.76 ± 0.03 70.8 ± 2.1 38.0 ± 1.5 2.9 ± 0.1 0.65 ± 0.04 51.9 ± 0.7 Foam, 6900 RPM 0.72 ± 0.03 65.2 ± 2.0 35.8 ± 1.3 2.8 ± 0.1 0.65 ± 0.05 48.7 ± 0.7 Water 0.25 ± 0.10 44.1 ± 4.6 21.5 ± 4.5 3.6 ± 0.1 0.94 ± 0.03 37.7 ± 0.6 Kraft Foam, 2000 RPM 0.52 ± 0.18 39.4 ± 3.0 18.0 ± 2.5 3.0 ± 0.2 0.83 ± 0.10 26.5 ± 0.6 Foam, 3500 RPM 0.50 ± 0.14 35.9 ± 2.2 16.7 ± 2.1 3.1 ± 0.2 0.86 ± 0.07 25.1 ± 0.6 Foam, 5000 RPM 0.52 ± 0.15 35.6 ± 2.2 16.6 ± 2.1 3.1 ± 0.2 0.86 ± 0.08 24.9 ± 0.6 Foam, 6900 RPM 0.53 ± 0.16 35.1 ± 2.4 16.6 ± 2.1 3.0 ± 0.2 0.86 ± 0.09 24.8 ± 0.7 Water 0.41 ± 0.10 26.7 ± 1.7 12.6 ± 1.6 3.2 ± 0.1 0.83 ± 0.04 22.1 ± 0.5 Table 4 Parameters of the Gaussian distribution used to fit the volume weighted bubble-diameter distributions in the foam fiber suspensions. The errors are given by the 95% confidence intervals. Rotation speed (RPM) CTMP Kraft BD ( m) BD ( m) BD ( m) BD ( m) 2000 134.2 ± 7.4 98.1 ± 6.5 133.9 ± 6.1 98.2 ± 5.4 3500 102.8 ± 7.8 70.3 ± 4.7 86.1 ± 3.8 57.3 ± 4.4 5000 86.4 ± 3.6 55.7 ± 2.2 78.3 ± 2.4 50.4 ± 3.2 6900 64.2 ± 3.8 40.3 ± 4.3 71.3 ± 4.1 45.2 ± 3.8 Figs. 4 and 5 show that the pore-size distribution of the foamformed sheets is much wider than those of the water-formed sheets in the corresponding cases. Moreover, there was a significant Gaussian component in the distributions of the foam-formed sheets with A = 0.72 0.79 for the CTMP fibers and A = 0.50 0.53 for the kraft fibers (see Table 3). The difference between the two forming methods appears particularly clear for the CTMP fibers. Eq. (2) fitted well the pore-size distributions of all foam-formed samples except the Fig. 2. X-ray tomographic images of the sheets obtained with (a) foam forming (rotation speed 5000 RPM, 0.5 bar vacuum in dewatering) and (b) water forming (no vacuum in dewatering) for CTMP fibers. Densities of the two types of sheet were almost equal, and their main difference was found in the microscopic structure of the sheet. Fig. 3. Volume-weighted pore-size distributions (probability density functions, PDFs) of the water-formed sheets obtained for (a) CTMP and (b) kraft fibers. The solid lines are fits by Eq. (2) to the distributions, whereas the dashed lines indicate its log-normal and Gaussian components.

A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 549 Fig. 4. Volume-weighted pore-size distributions of the foam-formed sheets for CTMP fibers at the rotation speeds (a) 2000 RPM, (b) 3500 RPM, (c) 5000 RPM, and (b) 6900 RPM. The solid lines are fits by Eq. (2) to the distributions, whereas the dashed lines indicate the log-normal and Gaussian components. sample obtained at rotation speed of 2000 RPM for the CTMP fibers (see Fig. 4a). This sample had a very low density (see Table 2), and it included many exceptionally large pores with a diameter of more than 200 m. The Gaussian component can be interpreted to be caused by the foam fiber interactions during the structure-forming process. Here the word interactions refers mainly to geometric restrictions in the location and orientation of fibers caused by the foam structure. However, interaction with bubbles can also bend fibers as shown in Fig. 6. During the forming process, bubbles can thus act as ghost particles, and can leave a considerable amount of large pores in the fiber network during its drainage. In Ref. [11], the area weighted bubble sizes followed a Gaussian distribution both for the pure foam and the foam fiber suspension. Table 4 shows the parameters of the Gaussian distributions that fitted these distributions. In all these cases, the bubble-size distribution was rather broad with a standard deviation exceeding a half of the average. Therefore, the distribution of large pores, affected by foam fiber interactions, can be expected to be broad as well. With monodisperse foam, a more peaked distribution of large pores could be expected. The pore-size distribution is compared with the volumeweighted bubble-diameter distribution as shown in Fig. 7 using the parameters of Tables 3 and 4. According to this Figure, changes in the pore-size distributions for varying rotation speed are not as big as the corresponding changes in the bubble size. However, we still see a clear systematic behavior of the Gaussian component for a varying bubble size. The only exception is the apparent location of the Gaussian component for rotation speed 2000 RPM for the CTMP fibers, but as explained above, the Gaussian component (any Gaussian function) describes poorly the largest pores in the sheet. The Sauter mean of the pore radius describes the overall behavior better in this case. As Fig. 8 shows, correlation between the mean pore radius and bubble radius is strong for the CTMP fibers. For kraft fibres, a similar trend is visible but the changes between the points are rather small compared to the relative error. However, the effect of bubble size seemed to be an order of magnitude stronger for the CTMP fibers, as shown by difference in the slopes of the two linear correlation functions. Extrapolating the observed trends to a vanishing bubble radius gave the limits 39.8 (CTMP) and 23.5 (kraft), which were very close to the mean pore radii obtained for water forming, 37.7 (CTMP) and 22.1 (kraft) (see Table 3). It is quite remarkable that, for the CTMP fibers, the mean pore size appeared to be determined by structural features of the suspending medium for such a broad window.

550 A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 Fig. 5. Volume weighted pore-size distributions of the foam-formed sheets obtained with kraft fibers for the rotation speeds (a) 2000 RPM, (b) 3500 RPM, (c) 5000 RPM, and (b) 6900 RPM. The solid linea are fits by Eq. (2) to the distribution, whereas the dashed lines indicate its log-normal and Gaussian components. Fig. 6. CTMP fibers (consistency 1.25%) in the foam made with SDS surfactant (concentration 1.0 g/l). The geometric restrictions caused by bubbles not only affect the location and orientation of fibers, but can also cause their bending. As indicated by these results, the mean pore size was always clearly smaller than the mean bubble size. Even if bubbles standed the forces caused by the flow of the foam fibre suspension, fibers would bend locally after removing the foam, and would thus form a denser sheet structure after drainage than indicated by the loose dynamic network of fibers in the foam. The stiff CTMP fibers were able to extend across voids left in the structure by bubbles much better than flexible kraft fibers. Even for ideally stiff fibers, the structure imposed by the foam would, however, undergo major deformations during drainage. The excess pressure inside a bubble is inversely proportional to its radius. For a typical 60 m radius and 55 mn/m surface tension of the bubble (see Fig. 9 of Ref. [13]), its pressure would be of the order of 0.01 bar, which should be compared with the 0.5 bar vacuum applied in the drainage. Thus, the forces during dewatering of the foam were probably large enough to reduce the effective bubble ( ghost particle ) size or even break up the largest bubbles. Thus, the location of the Gaussian component in the pore-size distribution below the measured bubble size (in equilibrium) is not surprising. Interestingly, in foam forming, the log-normal component of the distribution was quite similar for all rotation speeds for both types of fiber. This component reflects the distribution of small pores, whose size is smaller than the mean bubble size. Such pore sizes can thus be expected to be affected by the fiber geometry and flexibility rather than the foam properties.

A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 551 Fig. 7. Pore-size distribution in comparison with the volume weighted bubble-diameter distribution (mean BD of the Gaussian fitting) using the parameters of Tables 3 and 4 for the (a) CTMP and (b) kraft fibers. In Fig. 7a the open circle denotes the result for rotation speed 2000 RPM, for which the Gaussian component gave a poor description of the largest pores. This point is therefore an outlier in the general trend. Fig. 8. Sauter mean pore radius as a function of the mean bubble radius for (a) stiff CTMP fibers and (b) flexible kraft fibers. 4. Sheet properties for varying forming conditions 4.1. Surface characterization with the Bendtsen measurement In addition to a direct pore-size analysis based on the X-raytomography images, we characterized the surface porosity of the samples indirectly using the Bendtsen air-leakage measurement used normally to determine surface roughness. Because of its limited measuring range, the highly porous CTMP sheets could not be measured with the Bendtsen tester. Thus, these measurements were carried out for the kraft sheets only. The foam-formed kraft sheets manufactured with a 0.2 g/l surfactant content had all a clearly higher Bendtsen roughness than the corresponding water-formed sheets as shown in Fig. 9. The Bendtsen roughness of these sheets depended linearly on the Sauter mean radius (average volume/area ratio) of the pores. The R 2 - correlation of the Bendtsen roughness with the pore size of the foam-formed sheets was higher (98%) than with their density (81%). This is expected as the air flow is more sensitive to the contribution of large-permeability channels than to the average density. Moreover, the linear correlation extended quite nicely to the waterformed sample as well. The strong correlation between air leakage through the porous structure near the sheet surface and the mean pore size determined in the center region of the sample suggested that the same interaction mechanisms between bubbles and fibers apply throughout the forming process. 4.2. Opacity As shown in Fig. 10 for the kraft sheets, the measured opacity based on diffuse reflectance depended on the rotation speed used in foam forming. This is not surprising as the both sample density and pore-size distribution are systematically affected by the rotation speed. It is somewhat curious that the opacity was found to correlate best with 1 among the parameters that describe the porous structure, as 1 exceeded clearly the wave length of the light used. It is possible that 1 describes indirectly the non-uniform distribution of fibers which act as reflecting surfaces. For a large 1, the structure may include holes that enable an effective light transmission. 4.3. Strength properties The properties related to the mechanical strength (elastic modulus, tensile strength) of the water- and foam-formed sheets have earlier been found to be quite similar provided that their mate-

552 A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 Fig. 9. Bendtsen roughness of the foam-formed and water-formed kraft sheets for 0.2 g/l SDS concentration. The strong correlation with the pore size (Sauter mean) suggests that the pore structure near the sheet surface is created by the same mechanisms as the internal pore structure. rial density has been similar [4]. However, these investigations have been done at a higher density of the resulting fiber network, where the relative bonded area could have been even 50%, and the porous microstructure of the sheets have possibly played a minor role. For a decreasing network density, the deformations become non-affine affecting average mechanical properties [26,27]. Earlier work on well-bonded fiber networks has shown that the threshold of a non-affine behavior in the elastic modulus of wood fibers is about a density of 300 kg/m 3 [28]. In the present study, the network densities were clearly below this threshold density. Therefore, we can expect a non-affine behavior to have affected the measured strength properties of the fiber networks studied. Usually, a non-affine behavior is visible as a nonlinear behavior of the elastic modulus as a function of density of inter-fiber bonds [28]. Fig. 10. Opacity as a function of parameter 1 for kraft sheets made with a 0.2 g/l SDS concentration. For foam-formed kraft sheets, this density was systematically varied by varying the rotation speed. The measured values of the tensile strength and elastic modulus of these sheets correlated strongly with their measured density, with extrapolations close to the corresponding values of a water-formed sheet with the highest sheet density. The situation was quite different for the CTMP sheets, whose density varied much less with the rotation speed (see Table 2), and the corresponding water-formed and foam-formed sheets had a similar density. Now, correlation with density was completely lacking with R 2 = 0.01. Instead, the both tensile strength and elastic modulus depended strongly on the mean pore size as shown in Fig. 11. This dependence is not necessarily linear even though the foam-formed data points can be crudely extrapolated to the corresponding water point with a line. This result suggests that, an increasing non-affinity of the network, when its Sauter mean of pore radius increases and the microscale structure becomes more heterogeneous, results in a rapid decay of the both elastic modulus and strength of the network. Fig. 11. (a) Tensile stiffness index and (b) tensile index in the machine direction as a function of the pore size (Sauter mean) for CTMP sheets manufactured with a 0.3 g/l surfactant concentration. These data include results for foam-formed sheets of varying rotation speed and for a water-formed sheet.

A.M. Al-Qararah et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 482 (2015) 544 553 553 5. Conclusions In deposition processes the suspending medium has a significant role in determining the structural properties of the resulting fiber network. In particular, we showed that there is a clear difference in the microscopic pore structures of the foam-formed and water-formed sheets at a roughly equal density. Moreover, the pore structure was affected by the bubble-size distribution of the foam. This opens up a possibility to tailor the pore structure with foam properties. This tailoring would also include the structural homogeneity in various scales of the network, its overall material density, and possible layers of different types of fiber materials. These features were not studied here. The key parameters of the foam include, in addition to the bubble size, the air content, surfactant type and content, surface tension of the liquid, shear rate of the flow, and rheological properties of the foam. These parameters interact in a complicated fashion. Thus, there is a relatively wide playground for future, more detailed studies of foam-fiber interactions and the resulting microscopic structures of the network. However, it seems that fiber type is still the most essential factor that controls the properties of the resulting fiber networks. Flexible fibers such as in a kraft furnish lead to higher sheet density and smaller mean pore size. On the other hand, with stiff fibers like in a CTMP furnish, the resulting mean pore size can be controlled more effectively with the bubble size. Here we considered bubbles with a radius in the range of tens of microns. It would be interesting to consider other foam-production methods which would produce foams with clearly different bubblesize distributions. Acknowledgments This work was financially supported by Academy of Finland (Project Fibre-laden foams ), European Regional Development Funds (Project no A31918), and by the VTT Graduate School (A. A.-Q.). We would like to thank Dr. Atsushi Tanaka for assistance in the sheet making. References [1] D. Weaire, W. Hutzler, The Physics of Foams, Clarendon Press, Oxford, 1999, 246 p. [2] I. Cantat, F. Cohen-Addad, F. Elias, O. Höhler, F. Pitois, A. Saint-Jalmes, Foams: Structure and Dynamics, Oxford University Press, Oxford, 2013, 265 p. [3] B. Radvan, A.P.J. Gatward, The formation of wet-laid webs by a foaming process, Tappi 55 (1972) 748. [4] J. Lehmonen, P. Jetsu, K. Kinnunen, T. Hjelt, Potential of foam-laid forming technology for paper applications, Nordic Pulp Paper Res. J. 28 (2013) 392 398. [5] J. Poranen, H., Kiiskinen, J., Salmela, J., Asikainen, J. Keränen, E. Paakkonen, Breakthrough in papermaking resource efficiency with foam forming. Proceedings of TAPPI PaperCon 2013, vol. 1, Atlanta, GA, USA. P. 807 814. [6] A. Madani, S. Zeinoddini, S. Varahmi, H. Turnbull, A.B. Phillion, J.A. Olson, D.M. Martinez, Ultra-lightweight paper foams: processing and properties, Cellulose 21 (2014) 2023 2031. [7] J. Kinnunen, N. Lehmonen, P. Jetsu, T. Hjelt, Benefits of foam forming technology and its application in high MFC addition structures, in: Proceedings of the 15th Pulp and Paper Fundamental Research Symposium, Cambridge, Sept. 8 13, 2013. [8] Paper physics, in: K. Niskanen (Ed.), 2nd edition, Finnish Paper Engineers Association/Paperi ja Puu Oy, Helsinki, 2008, 360 p. [9] M. Alava, K. Niskanen, The physics of paper, Rep. Prog. Phys. 69 (2006) 669 723. [10] T. Hjelt, K. Kinnunen, J. Lehmonen, N. Beletski, E. Hellén, V. Liljeström, R. Serimaa, A. Miettinen, M. Kataja, Intriguing Structural and Strength Behavior in Foam Forming, Progress in Paper Physics Seminar, Graz, 2011, pp. 135. [11] A.M. Al-Qararah, T. Hjelt, K. Kinnunen, N. Beletski, J.A. Ketoja, Exceptional pore size distribution in foam-formed fiber networks, Nordic Pulp Paper Res. J. 27 (2012) 226 230. [12] A.M. Al-Qararah, T. Hjelt, A. Koponen, A. Harlin, J.A. Ketoja, Bubble size and air content of wet fiber foams in axial mixing with macro-instabilities, Colloids Surf. A Physicochem. Eng. Aspects 436 (2013) 1130 1139. [13] A.M. Al-Qararah, T. Hjelt, A. Koponen, A. Harlin, J.A. Ketoja, Response of wet foam to fiber mixing, Colloids Surf. A Physicochem. Eng. Aspects 467 (2015) 97 106. [14] I. Mira, M. Andersson, L. Boge, I. Blute, G. Carlsson, K. Salminen, T. Lappalainen, K. Kinnunen, Foam forming revisited. Part I. Foaming behaviour of fiber-surfactant systems, Nordic Pulp Paper Res. J. 29 (2014) 679 688. [15] T. Lappalainen, K. Salminen, K. Kinnunen, M. Järvinen, I. Mira, M. Andersson, Foam forming revisited. Part II. Effect of surfactant on the properties of foam-formed paper products, Nordic Pulp Paper Res. J. 29 (2014) 689 699. [16] A. Jäsberg, P. Selenius, A. Koponen, Experimental results on the flow rheology of fiber-laden aqueous foams, Colloids Surf. A Physicochem. Eng. Aspects 473 (2015) 147 155. [17] A. Goel, M. Tzanakakis, S. Huang, S. Ramaswamy, D. Choi, B.V. Ramarao, Characterization of the three-dimensional structure of paper using X-ray microtomography, Tappi J. 84 (2001) 72 80. [18] E.J. Samuelsen, O.W. Gregersen, P.J. Houen, T. Helle, C. Raven, A. Snigirev, Three- dimensional imaging of paper by use of synchrotron X-ray microtomography, J. Pulp Paper Sci. 27 (2001) 50 53. [19] M. Aaltosalmi, A. Koponen, A. Timonen, G. Lee, S. Ramaswamy, Numerical analysis of fluid flow through fibrous porous materials, J. Pulp Paper Sci. 30 (2004) 251 255. [20] Stock Preparation and Wet End, in: H. Paulapuro (Ed.), 2nd edition, Finnish Paper Engineers Association/Paperi ja Puu Oy, Helsinki, 2008, 516 p. [21] Paper and board - Determination of roughness/smoothness (air leak methods) - Part 2: Bendtsen method, ISO 8791-2. 1990. [22] Paper and board - Determination of opacity (paper backing) - Diffuse reflectance method, ISO. 2471.2008. [23] Paper and board - Determination of tensile properties - Part 2: Constant rate of elongation method, ISO 1924-2: 1994. [24] W.C. Bliesner, A study of the porous structure of fibrous sheets using permeability techniques, Tappi J. 47 (1964) 392 400. [25] C.T.J. Dodson, W.W. Sampson, The effect of paper formation and grammage on its pore size distribution, J. Pulp Pap. Sci. 22 (1996) J165 J169. [26] C. Heussinger, E. Frey, Stiff polymers, foams and fiber networks, Phys. Rev. Lett. 96 (2006) 17802. [27] C. Heussinger, E. Frey, Floppy modes and nonaffine deformations in random fiber networks, Phys. Rev Lett 97 (2006) 105501. [28] R.C. Picu, Mechanics of random fiber networks a review, Soft Matter 7 (2011) 6768.