Rheology and tribology of starch + κ‐carrageenan mixtures

Abstract In this study, we investigated the rheological and tribological properties of biopolymer mixtures of gelatinized corn starches (0.5 – 10.0 wt%) and κ‐carrageenan (κC) (0.05 – 1.0 wt%). Two different starch samples were used. The first starch (CS1), despite extensive heating and shearing contained “ghost” granules, while the second starch (CS2) had no visible ghost granules after the same gelatinization process as CS1. Apparent viscosity measurements demonstrated that κC + CS1 mixtures were shear thinning liquids, with viscosity values being lower than the corresponding weight average of the values of the individual equilibrium phases at shear rates < 50 s−1. Tribological results revealed that κC ≥ 0.5 wt% was required to observe any decrease in friction coefficients in the mixed lubrication regime. Starch (CS1) showed an unusual behavior at ≥ 5 wt%, where the friction coefficient decreased not only in the mixed regime but also in the boundary regime, probably due to the presence of the “ghost” granules, as the latter became entrained in the contact region. The CS1 + κC mixtures showed significantly lower friction coefficients than that of pure CS1 and κC in the mixed regime. However, the CS2 + κC mixture (i.e., containing no ghost granules) showed similar behavior to pure κC in the mixed regime, while lower friction coefficients than that of the pure CS2 and κC in the boundary regime. These findings illustrate new opportunities for designing biopolymer mixtures with tunable lubrication performance, via optimizing the concentrations of the individual biopolymers and the gelatinization state of the starch.

Oral processing includes both rheological and tribological phenomena (Sarkar, Andablo-Reyes, Bryant, Dowson, & Neville, 2019). In the early stages of oral processing, bulk rheological phenomena could be dominant where the food is considered to be in the continuum, which is measured at a fixed gap, equivalent to the size of the food material. However, with time, oral processing involves a range of deformation process, and the properties of food are not only driven by the viscosity, but also the friction between the surfaces of the oral tissues and the food as well as between the surfaces of the tissues Stokes, Boehm, & Baier, 2013). Therefore, the tribological properties become dominant in the later stages of oral processing involving surface interactions of the food driving the mouthfeel perception. Therefore, oral tribology has become an increasingly important part of oral processing studies of model and real foods (Laguna, Farrell, Bryant, Morina, & Sarkar, 2017;Sarkar, Andablo-Reyes, Bryant, Dowson, & Neville, ;Stokes et al., 2013;Stribiţcaia, Krop, Lewin, Holmes, & Sarkar, 2020).
There has also been a gradual increase in interest in understanding specifically the tribological properties of polysaccharides, since they can play an essential role in adjusting the mouthfeel of foods and beverages (Garrec & Norton, 2012Stokes, Macakova, Chojnicka-Paszun, De Kruif, & De Jongh, 2011;Torres et al., 2019;Zhang et al., 2017). Starch is the most widespread polysaccharide in foods that affects food structure and texture and starch textural properties tend to vary widely depending on their origin, for example, corn (maize), wheat, potato, tapioca, and rice. In addition, the textural properties of starches of one origin depend on the starch granule size, shape, degree of swelling, disruption, that is, gelatinization. (Blazek & Gilbert, 2011;Buléon, Colonna, Planchot, & Ball, 1998). The degree of gelatinization may determine the tribological properties, for example, stickness, slipperiness, and so forth (Ai & Jane, 2015;Evans & Lips, 1992), though there have been relatively few systematic studies on the lubrication properties of starch (Torres, Andablo-Reyes, Murray, & Sarkar, 2018;Torres, Tena, Murray, & Sarkar, 2017;Zhang et al., 2017). Zhang et al. (2017) studied the tribological properties of suspensions of cooked swollen starch granules (ghost suspensions) from maize or potato across a wide range of concentration. The coefficient of friction (μ) was shown to decrease in the boundary and mixed regimes with increasing concentration (0.01 -1.0 wt%) because the ghost particles became entrained in the contact zone at low entrainment speed (U, 40 mm s -1 ) as compared to that of water, where the latter was squeezed out of the hydrophobic tribological surfaces used in their study. Torres et al. (2018), on the other hand, investigated the tribological effects of wheat starch-based microgel particles. The magnitude of μ decreased on increasing starch content of the microgels from 15 to 20 wt%, the latter particles having been formed from a bulk gel with a higher shear modulus. It was postulated that these higher starch content microgels might therefore be slightly stiffer and better at keeping the hydrophobic tribological surfaces apart, enabling lowering the μ values.
The lubricating behavior of non-starch polysaccharides, such as locust bean gum, carrageenan, gellan, guar, pectin, xanthan gums, and so forth in the boundary and mixed lubrication regimes has also attracted recent research attention. In particular, the tribological properties of κ-carrageenan (κC) have been investigated Stokes et al., 2011). Garrec, Guthrie, and Norton (2013) showed that as the concentration of κC increased, the polymer was entrained in the contact zone, and μ values decreased at the site of the converging geometry formed between the tribopairs. Interestingly, the lubrication behavior with κC was consistent across smooth to rough PDMS substrates (mean asperity radius 10-400 nm), highlighting their ability to entrain and prevent direct PDMS-PDMS asperity contact, irrespective of the degree of surface roughness (Stokes et al., 2011).
Although some information on the tribological properties of biopolymer solutions is available in the literature, studies of combinations of biopolymers are very rare. On the other hand, there is extensive literature on the phase separation of biopolymer mixtures.
In this study, we used corn starch (CS) and κC as model biopoly- shown to exhibit elastic/plastic properties, therefore, they are likely to affect oral processing and even lubrication (Atkin, Abeysekera, & Robards, 1998

| Preparation of starch + κC mixtures
Gelatinized waxy corn starch (CS1 or CS2) (0.5 -5 wt%) was prepared by dispersing the starch powder in phosphate buffer at pH 7.0, followed by heating in a water bath at 90 C for 20 min with constant shearing using a magnetic stirrer, to gelatinize the starch. κ-carrageenan (κC) was similarly dispersed in phosphate buffer at pH 7.0 (0.05 -1.0 wt%) for at least 24 h at room temperature and then heated at 90 C for 20 min. To prepare the biopolymer mixtures, both solutions were prepared separately as above before mixing. Equal volumes of starch solution and κC solution of different concentrations were blended and homogenized at 21,000 rpm for 30 min using an Ultra Turrax T25 homogenizer (IKA-Werke GmbH &Co., Staufen, Germany) at room temperature (25 C).

| Apparent viscosity
Rheological characterization of the pure biopolymers and their mixtures was performed using a modular compact rheometer, model MCR 302 (Anton Paar, Austria) at shear rates ranging from 0.1 to 1,000 s −1 at 37 C. The biopolymer mixtures did not phase separate within the first 2 h of preparation but started to separate after 7 days of storage at ambient conditions (see Figure S1) and all the rheological experiments were carried out within 2 h of preparation of the biopolymer mixtures. A cone-and-plate geometry system (CP50-2, cone diameter 50 mm, cone angle: 2 ) with a gap of 1 mm was used for all measurements. For each measurement, 2 ml of sample were pipetted onto the plate, excluding any air bubbles. A temperature-controlled cover prevented evaporation during the measurements and helped to maintain the temperature at 37 C (i.e., representing oral processing temperature). Samples were left on the plate for $2 min to achieve a steady state, following which the apparent viscosities were measured.
High shear rate limiting viscosity (η ∞ ) was determined to scale the tribological data in the Stribeck curves (see later in Section 3).

| Tribology
Tribological measurements on the pure biopolymers and their mix-  (1): where, U B is the rolling speed of the ball and U D is the sliding speed of the disc. The temperature in the tribological experiment was set at 37 ± 1 C, matching that of the rheological measurements. In addition, the load-bearing ability of the biopolymers was also measured at loads (W) ranging from 1 to 5 N, at U = 0.005 m s -1 . The friction coefficients are reported as the mean and SD of 3 measurements carried out on at least triplicate samples prepared on different days.

| Microscopy
Optical microscopy (Nikon, SMZ-2 T, Japan) was used to observe the microstructure of the heated starch samples to identify any residual granule structures. Samples were diluted with buffer (1:10 wt/wt

| Statistics
All experimental results were reported as mean and SDs of three measurements on triplicate samples (n = 3 × 3). The statistical analyses were conducted for the rheological data at 50 s −1 simulating oral processing shear and tribological data at boundary regime (0.005 m/s) and mixed regimes (0.05, 0.1 m s -1 ) using one-way analysis of variance and multiple comparison test using SPSS software (IBM, SPSS statistics, version 24) and the significant difference between samples were considered when p < .05 using Tukey's test. Statistical results can be observed in Tables S1a-d  the pressure in the contact region between the PDMS tribopairs is sustained partly by the fluid and partly by the surfaces. Such a prolonged boundary regime with phosphate buffer has been seen previously (Sarkar, Kanti, Gulotta, Murray, & Zhang, 2017), where phosphate buffer was proposed to be squeezed out of the hydrophobic PDMS-PDMS contact zone.
The κC solutions (Figure 1ai) showed interesting concentrationdependent tribological properties, where the extents of the boundary regime were shortened, that is, the onset of mixed regime occurred at lower speeds, upon increasing the concentration of κC from 0.05 to 1.0 wt% with no changes in boundary friction observed from 0.05 to 0.5 wt% concentration (p > .05) (see Table S1a for statistical analyses).
Also, the μ values in the mixed lubrication regime decreased sharply as the concentration was increased from 0.05 wt% to 0.1 wt% (p < .05) but no significant decline was observed when the concentration increased from 0.5 wt% to 1.0 wt% (p > .05) (see Table S1a for statistical analyses). For instance, the μ values decreased by an order of magnitude at U = 0.1 m s -1 , when the concentration of κC was increased by order of magnitude, that is, from 0.1 to 1.0 wt%. Stokes et al. (2011) have already indicated that κC is a useful potential lubricant because it decreases friction coefficients efficiently in the mixed regime, compared to numerous other aqueous polysaccharides (pectins, locust bean gums). It is noteworthy that κC solutions ≥ 0.5 wt% also demonstrated a characteristic hydrodynamic lubrication regime behavior, where the μ values started to increase at the highest U.
Shifting our focus to starch, 0.5 -2.0 wt% CS1 (Figure 1bi) showed similar boundary behavior to that of the phosphate buffer (up to U = 0.1 m/s), but slightly lower μ values in the mixed lubrication regime as compared to that of the buffer. Interestingly, μ was more or less the same for all CS1 concentrations between 0.5 and 2.0 wt% irrespective of the regimes (p > .05) (see Table S1b for statistical analyses), unlike the behavior with κC (Figure 1ai Table S1b for statistical analyses), but converged on the same values as U approached the maximum applied (U = 1 m s -1 ).
Particularly CS1 at 5.0 wt% did not show any visible boundary regime.
This behavior might be due to the presence of higher concentrations of "ghost" granules at the higher CS1 concentration, which potentially might have granule-granule inter-molecular friction, flattened and providing sufficient hydrodynamic lift to separate the PDMS surfaces leading to the immediate onset of the mixed lubrication regime even at low speeds (U ≤ 0.005 m s -1 ).
The presence of ghost granules in the CS1 sample was evidenced via optical and confocal laser scanning microscopy, as shown in Figure 2a,b, respectively. It is clear from both these images that ghost granules remained in the CS1 sample even though the starch was subjected to high temperature (≈ 90 C) and shearing. In addition, the gelatinized rice starch granules in the afore-mentioned study formed a thicker continuous tribological film that separated the tribo-pair surfaces due to the increased viscosity. Therefore, the deformability of ghost granules, which are tens of microns in size, plus their potential ability to form a more continuous film, could explain the absence of the boundary regime and rapid onset of mixed lubrication regime seen (Figure 1bi) Tables S2a,b for statistical analyses) and both κC and CS1 showed shear thinning behavior. However, marked shear thinning only occurred for CS1 at higher concentrations (≥ 3.0 wt%) and the viscosity was significantly higher than those of 0.5 -2.5 wt% CS1 at the typical oral processing shear rate of 50 s −1 (see Table S2b for statistical analyses). In order to estimate the viscous contribution in the tribological data, the high shear rate limiting viscosity (η ∞ ) was obtained. The Reynolds equation for softelastohydrodynamic lubrication (EHL) was applied when the sample showed a clear hydrodynamic regime (iso-viscous-elastic lubrication regime), as observed for κC in Figure 1ai at ≥ 0.5 wt%. This model (De Vicente, Stokes, & Spikes, 2005) gives an arithmetic expression of the Reynold's equation for the soft EHL lubrication regime between sliding-rolling ball and plate contact under fully flooded conditions. F I G U R E 2 Optical (a) and confocal (b) micrographs of 1 wt % CS1 after gelatinization. The bright regions in (b) are due to CS1 labeled with Rhodamine Blue. Scale bar is 50 μm The friction coefficient in contact can be expressed in terms of the characteristic parameters: W, η and U, as follows (Equation 2) where Young's modulus and reduced radius of the contact, respectively.
Here E 0 and E 0 0 are the elastic moduli of the surface material and R 0 and R 0 0 are the radius of the surfaces.
The first term in Equation (2) (2) can be successfully fitted to the experimental data in Figure 1aii Figure 1aiii shows that 0.1 -0.5 wt% κC gave similar shaped curves with overlapping characteristics, which means that the tribological properties of these solutions were dominated by their η ∞ (Figure 1aiii). However, at concentrations ≥0.5 wt% κC, there was a lower μ in the mixed lubrication regime, which might be attributed to the coil-helix transition of κC as it transforms from a solution to a gel (Gabriele, Spyropoulos, & Norton, 2009;Rochas & Landry, 1987;Rochas & Rinaudo, 1982), allowing κC to entrain into the contact zone fully.
For CS1 (Figure 1biii), the Stribeck curve obtained using Equation (2) did not show good agreement with increasing concentration of CS1, the Stribeck curves being significantly shifted to the right of the data points. These results support the idea that the stickiness of starch granules in the dispersions might resist the relative motion of the tribo-pairs due to molecular adhesion as reported by previous studies (Liu et al., 2016;Wu, Gunaratne, Collado, Corke, & Lucas, 2015;Zhang et al., 2017), although this needs further investigation.

| Load-bearing abilities of the biopolymers
The load-bearing capacity of the lubricating film is an important parameter to help understand the lubrication performance of biopolymers. The friction force (F) of κC and CS1 as a function of W at low U (0.005 m s -1 ) is shown in Figure 3, aiming to test the load-bearing abilities of the biopolymers (see μ values of κC and CS1 as a function of U at different W in Figure S2). According to Stokes et al. (2011), μ for an adsorbed polysaccharide scales with W as μ $ W 2 3 . This dependency originates from Equation (3): where A = the contact area for a circular contact = π 3 WR Ã 4E Ã À Á2 3 ; τ i is the interfacial shear stress between the surfaces. The F versus W data in Figure 3 were thus fitted using Equation (4): The dashed lines in Figure 3 show the best fits and the slope of these lines gives τ i , which for κC and CS1 were 0.144 and 0.099 MPa, respectively, that is, lower than the value for PDMS/water/PDMS (τ i = 0.23 MPa). Thus CS1 lowers the interfacial shear stress between the surfaces even more than κC suggesting that CS1 is more likely to remain bound to the PDMS surfaces. This further explains the larger decrease in friction coefficient with CS1 and the almost non-existent boundary lubrication regime illustrated in Figure 1bi.

| Rheological versus tribological behavior of starch + κC mixtures
Both Figure 4ai,bi demonstrate that the CS1 + κC mixtures, irrespective of the total biopolymer concentration (1.65 or 2.75 wt%) have non-Newtonian behavior and the viscosity values at orally relevant shear rate of 50 s −1 were significantly different in comparison to either CS1 or κC (see Table S2c  (κC and CS1) at corresponding shear rates, except in case of the 1.65 wt % mixtures at ≥ ca. 50 s −1 (Figure 4ai). In other words, the viscosity behavior appeared to be inordinately affected by the κC in the mixture.
This is in line with the rheological properties of the biopolymer mixtures studied previously, where the viscosity of the mixtures tended to be higher than the value of κC alone (Fakharian et al., 2015;Huc et al., 2014;Lafargue et al., 2007).
Looking at the tribological data, the μ of the mixtures (Figure 4aii, bii at 1.65 and 2.75 wt%, respectively) were considerably lower than that of pure CS1 across the whole range of mixed regime (see Table S1c,d for statistical analyses). Most surprisingly, unlike the bulk rheology results (Figure 4ai,bi), the μ values of the mixtures were statistically similar to those of pure κC (p > .05) except for the 2.75 wt% mixture (p < .05) (see Table S1c, or 2.75 wt%) (Figure 4aii,bii). In addition, the swollen, more deformable gelatinized granules could entrain in the gap to support the load (Torres et al., 2018).
To understand further the role of these "ghost" starch granules, Figure 5 shows the μ values versus U of a mixture of 2.5 wt% CS2 + 0.25 wt% κC, that is, the starch which contained no starch granules.
Similar to the mixtures containing CS1 (Figure 4aii,bii), the 2.5 wt% CS2 + 0.25 wt% κC mixture did not show a reduction in friction in comparison to pure κC in the mixed lubrication regime (p > .05) (see Table S1d for statistical analyses). Comparing the data of the mixture containing CS2 in Figure 5 with the behavior of CS1 in Figure 4bii, the μ values of the CS2 + κC mixture appeared to be comparatively higher than those of the CS1+ κC mixture at higher entrainment speeds, though statistically this was not significant (p > .05) (see Table S1d for statistical analyses). Nevertheless, it was interesting to observe that the 2.5 wt% CS2 + 0.25 wt% κC mixture ( Figure 5) reduced the boundary friction significantly as compared to pure κC and the friction in the mixed regime with respect to CS2 (p < .05) (Figure 4bii and Table S1d for statistical analyses).
The inset to Figure 5 is a corresponding optical micrograph of the CS2 starch after gelatinization, clearly illustrating the lack of "ghost" granules, which would therefore appear to be the explanation of this stark difference in behavior compared to CS1. This suggests that the F I G U R E 4 (i) Apparent viscosity (η) as a function of shear rate (γ) and (ii) friction coefficients (μ) versus entrainment speed (U) of biopolymer mixtures at (a) lower biopolymer concentrations: (1.5 wt% CS1 + 0.15 wt% κC, ) and (b) high biopolymer concentrations (2.5 wt% CS1 + 0.25 wt% κC, ) plus the controls of 0.15 wt% κC ( ), 0.25 wt% κC ( ), 1.5 wt% CS1 ( ) and 2.5 wt% CS1 ( ) alone. The weight average values of the corresponding individual controls for the mixtures are also shown ( ). Values represent means and error bars represent the SDs for at least three measurements on triplicate samples (n = 3 × 3) micron-sized "ghost" granules present in the CS1 + κC mixtures were efficient in reducing the friction in the mixed lubrication regime. However, they were not able to get in between the PDMS surfaces in the boundary region, that is, at low U. On the other hand, the CS2 + κC mixture containing no "ghost" granules were beneficial in creating potentially nanometric sized boundary films creating smoother tribocontact surfaces. It is worth highlighting that CS2 apparently formed such boundary lubrication films only in presence of κCsuch behavior is not seen in pure CS2 ( Figure 5). This suggests some local interaction is occurring between CS2 and κC in the boundary region, which requires further investigation in the future.

| CONCLUSIONS
In this study, we investigated the rheological and tribological properties of κ-carrageenan and gelatinized CS and their mixtures. Both κC solutions and gelatinized starch dispersions were shear-thinning liquids, and one starch (CS1) still contained ghost granules while the other (CS2) did not. In tribological measurements, κC showed good lubrication performance by efficiently reducing the friction coefficient in the mixed lubrication regime, especially when the concentration was ≥ 0.5 wt% and also showed hydrodynamic behavior at higher entrainment speeds (U). On the other hand, the CS1 at ≥ 5 wt% immediately showed onset of a mixed lubrication regime at U < 0.01 m s -1 without any observed boundary regime. This was attributed to the presence of "ghost" starch granules that flattened in the confinement and enabled accelerated onset of the mixed lubrication regime, which also provided better load-bearing ability than κC. Mixtures of CS1 + κC (at 1:10 wt/wt ratio) showed that the viscosity values were lower than the weight average of the individual equilibrium phases (CS1 + κC) and the μ values of the mixtures containing the ghost starch granules (from CS1) were much lower in the mixed lubrication regime. The mixture CS2 + κC, lacking ghost starch granules, did not offer such lubrication benefits in the mixed lubrication regime. However, CS2 + κC did offer effective boundary lubrication with respect to κC, owing to the gelatinized starch + κC somehow forming a lubri-

DATA AVAILABILITY STATEMENT
The data presented in this article will be openly available from the University of Leeds data repository: https://doi.org/10.5518/927 ETHICAL STATEMENTS F I G U R E 5 Friction coefficient (μ) versus entrainment speed (U) of biopolymer mixtures using CS2 without the "ghost" granules, that is, 2.5 wt% CS2 + 0.25 wt% κC, ( ) and controls of 2.5 wt% CS2 ( ) and 0.25 wt% κC ( ). Values represent means and error bars represent the SDs for at least three measurements on triplicate samples (n = 3 × 3). The inset is an optical micrograph of CS2 starch after gelatinization, illustrating the lack of "ghost" granules. The weight average values of the corresponding individual controls for the mixtures are also shown ( ). Error bars represent SDs