Capturing the colloidal microplastics with plant-based nanocellulose networks

13 Microplastics accumulate to various aquatic organisms causing serious health issues, and they have 14 raised concerns about human health by entering our food chain. The recovery techniques for the 15 most challenging colloidal fraction even for the analytical purposes are limited. Here we show how 16 hygroscopic nanocellulose network acts as an ideal capturing material even for the tiniest 17 nanoplastic particles. We reveal that the entrapment of particles from the aqueous environment is a 18 result of the network’s hygroscopic nature - a feature which is further intensified with the high 19 surface area. We determine the nanoplastic binding mechanisms using surface sensitive methods, 20 and interpret the results with the random sequential adsorption (RSA) model. The microplastic 21 uptake does not rely on any specific interfacial interaction but rather on the water transport behavior 22 of nanocellulose. These findings hold potential for the explicit quantification of the microplastics and qualification of colloidal (ø . Here we show networks

To further elaborate the role of surface interactions between nanoplastics and the binding substrates, 148 the adsorption of anionic nPp particles was followed using Quartz crystal microbalance with 149 dissipation monitoring (QCM-D). With this approach, we aim to exclude the influence of network 150 porosity that generates water transport functions and amplify the role of direct surface interactions. 151 We focus our interfacial investigations on colloidal-sized nPp particles since the behaviour of 152 nanoscaled particles is mostly taking place at interfaces. In nature, nanoplastics tend to accumulate 153 e.g. toxins and therefore from the environmental point of view, pure particles do not exist. 9 154 Therefore, we used either stabilised or purified PS particles, all carrying a net negative surface 155 charge (Supplementary Table 1). 156 Our results show that anionic nPp particles -both stabilised and purified -adsorb on native CNF, 157 regenerated cellulose (RC), and on polystyrene (PS) (ΔfRC >> ΔfCNF > ΔfPS) ( Fig. 3a and   158 Supplementary Fig. 6 a,c) whereas no adsorption was detected on TEMPO-CNF. This result 159 indicates that the electrostatic repulsion between anionic domains prevents the direct binding of nPp 160 on the highly anionic TEMPO-CNF. 161 Finally, we introduce a systematic approach for explicit nanoplastic particle detection to bridge the 162 well-known methodological gap of detection and quantification of nPp from the environment 15 . We 163 qualified the substrate performance to bind colloidal plastics via surface interactions by comparing 164 the experimental surface coverage to the theoretical maximum coverage. This was carried out by 165 linking the adsorption data to comprehensive image analysis and by applying a random sequential 166 adsorption (RSA) model. In the RSA model, the jamming limit at which the density of adsorbed 167 particles (particles treated as geometrically restricted and fixed circular objects without 168 conformational and orientational changes) saturates on a 2D film is defined as a theoretical 169 maximum coverage (θ∞ = 0.547). Therefore, the saturation limit in RSA is significantly lower than 170 the optimum filling of the surface. 30,31 By fitting the QCM-D data (Fig. 3a) with the RSA model 171 (Fig. 3b), and by applying the image analysis (Fig. 3c, Supplementary Fig. 7-9) we gain access to the adsorption rate and the number of particles per unit area after the adsorption (dN/dt) 173 (Supplementary Table 4, Equation 2), which can be translated to surface mass density (Γ) and 174 adsorption rate (dΓ/dt) (Equation 6) since the nanoplastic adsorption process meets well the RSA 175 requirements (See Methods). QCM-D detects the adsorbed hydrated total mass by acoustic principle 176 showing simultaneously high changes in energy dissipation ( Supplementary Fig. 6 b,d and 10).  Table 1 collects the relevant experimental data on particle adsorption, a 184 factor describing water coupled to the adsorbed layer, and surface coverage (θmax) at the solid-gas 185 interface at the end of the irreversible adsorption process (t = ~1h). Table 1 also tabulates the 186 system-specific parameters from RSA fittings i.e. surface coverage (θmax) at the solid-liquid 187 interface, the adsorption rate coefficient (ka), and the occupied area (a) of single nPp including the 188 water, which is strongly interacting with the nanoplastic particle. It should be noted that RSA-189 derived θmax takes into account the particle diameter with coupled water resulting in higher surface 190 coverage values when compared to "dry systems". Our results show that the strongly bound water 191 layer does not prevent the particle packing and therefore the estimation of the true surface coverage 192 (θmax/θ∞) using dry system data is warranted. The full treatise of the adsorption data with the RSA 193 model is described in the Methods and Supplementary Fig. 11. 194 We extracted the key findings, and the discussion is supported by the schematic presentation shown  capturing. We show that by combining surface-sensitive methods with nanomicroscopy, image 235 analysis, and modeling, we are able to quantitatively assess the nPp behavior at interfaces. This type 236 of nPp adsorption data has not been previously collected, and it is essential when designing 237 materials for quantitative analysis purposes, and for collection and recovery from different 238 environments ranging from wastewaters to the sites where the nano-and microplastics are 239 produced. Nanocellulose originates from the natural sources, it is renewable and non-toxic, which 240 are key aspects when designing next-generation functional materials diminishing the dependency on 241 the synthetic counterparts. Today, nanocellulose can be produced and modified in various ways to 242 yield hydrogels, self-standing films, and porous aerogels and cryogels, which make it an ideal 243 material for many future solutions where the high hygroscopicity is an asset 21,37 .    Table 1 Experimental data, adsorption parameters and surface coverage estimations for different nPp systems.

277
Experimental data RSA fitting parameters    where area a is obtained from RSA fitting.

285
f Occupied area of a single nanoplastic particle including particle and the coupled water i.e. water strongly interacting     Fig. 12). Fluorescence studies were carried out by immersing the films in the 388 aqueous dispersion containing 0.1 g L -1 particles for 10 min without mixing (Supplementary Fig. 4).

389
The fluorescence of the solutions was recorded before and after the immersion. The amount of where Δm is the areal mass, n is the overtone number (n = 1, 3, 5, 7, 9, 11  particle capture and adsorption. The self-standing films and QCM-D crystals were dried after the 482 measurements and attached onto SEM sample holders using carbon tape. Samples on the holders were coated with gold by sputtering (2 nm thick gold surface) to improve sample conductivity. above categories the image was classified into. When the particles were mainly detached, the 509 particles were identified from the threshold image by their area (we know the diameter of particle 510 size in each image) and shape (circular objects). The size of the clusters observed was assumed to 511 be three particles. In samples where the clusters were large, and the particles were mainly in the 512 clusters, individual particles were not reliably identified. In this case, the area of each cluster was 513 determined, and the number of particles needed to achieve the same area was calculated. Finally, 514 the identified particles were presented by drawing a marker on the original SEM image (see 515 Supplementary Fig. 8, 9).

516
Theoretical maximum adsorption of particles -Fittings with RSA model 517 We used random sequential adsorption (RSA) model to evaluate the maximum adsorption capacity 518 of the ultrathin films. The thickness of the ultrathin film is well below 100 nm. Thus, the PS 519 particles cannot penetrate the film. If the goal is to determine the maximum number of PS particles 520 that can fit on the surface of ultrathin films, the question can be simplified to the packing of circular 521 disks in a plane. Adsorption of particles on solid, flat surfaces is often an irreversible process, as 522 was also verified by QCM-D measurements in this study (Supplementary Fig. 6). In addition, 523 particles usually do not adsorb one on top of each other, instead they form a monolayer 524 ( Supplementary Fig. 7). The basic RSA model assumes that only steric repulsion is present between 525 the circular disks. For circular disks of the same size, saturation occurs at a surface coverage θ∞ of 526 0.547. If there are disks of different sizes (particles of varying diameter) in the system, higher 527 surface coverage θ can be reached. In this study, all particles were the same size. If the viewing area 528 is one mm 2 and the PS particles are the same size, the maximum area covered by the particles is 529 0.547 mm 2 . In this case, a 1 mm 2 area can hold 5.76 × 10 7 circles (PS particles) with an effective 530 diameter (diameter that perceives also the estimation of electrical double layer and hydration shell 531 of the particle) of deff = 1.1dabs = 110 nm. 47 The cross-sectional area of one particle was calculated 532 using Aparticle = π(deff/2) 2 . Also the RSA model assumes that particles hit the surface at the same rate throughout the adsorption process 48 . Therefore, the concentration c must be high enough to form a 534 monolayer in the saturation regime. If there are not enough particles, the adsorption process may 535 stop before reaching the saturation surface coverage. Table 1 shows that the maximum surface mass 536 density (Γmax) was 1890 ng cm -2 for purified nPp adsorbed on RC. With a QCM-D sensor diameter 537 of 9 mm, there was ~1200 ng of nPp on the sensor surface. This corresponds to approximately 0.2% 538 of nPp (100 000 ng cm -3 nPp dispersion was introduced into the QCM-D chamber with a flow rate 539 of 0.1 ml/min for approximately 1 hour). Therefore, we can assume that nPp hit the surface at the 540 same rate throughout the adsorption process, and the requirements to utilize the RSA model are 541 met. In our system, all of the main RSA principles are valid, and therefore, the adsorption 542 behaviour of nPp particles can be described using the random sequential adsorption (RSA) model. 31

543
Assessment of particle adsorption kinetics and the amount of coupled water 544 In order to evaluate the kinetics of the PS particle adsorption process, we modeled the QCM-D data 545 (Fig. 3a, Supplementary Fig. 6) with the RSA model ( Fig. 3b)  (SEM imaging coupled with image analysis) to obtain the adsorbed dry mass. The amount of 575 adsorbed nPp was calculated from SEM images of dry ultrathin films after the adsorption measurements. SEM images were then analyzed as described above, and the number or particles 577 was determined. Since the mass of a single nanoplastic particle was known (5.26 ×10 -7 ng) the total 578 dry mass can be calculated (∆ ) which equals to the surface mass density Γ (ng cm -2 ). To 579 eliminate the influence of water on QCM-D measurement, the QCMD-D data m(t) were rescaled 580 utilizing the surface mass density (Γ) determined from the SEM images (Γ (t)=m(t)* (Γmax/mmax)) 581 where max refers to the maximum value determined by each measurement method. For CNF and RC surfaces and both particle types (purified and stabile) the frequency response was 597 a factor of 3 larger for the experimental frequency compared to the calculated frequency from 598 image analysis. This result indicates that 2/3 of the sensed mass by the QCM-D 5 th overtone 599 frequency corresponds to the coupled water. For PS ultrathin film, the result was that 3/4 of the sensed mass by QCM-D was coupled water. According to previous studies, the difference between 601 the hydrated and dry mass is typically a factor of 1.5-40 32 . The factor in the case of adsorbing nPp is 602 in the lower range since QCM-D is typically used to study protein adsorption, where the proteins 603 can be thought as soft spheres, including water in their structure. In contrast, in the case of nPp they 604 are hard spheres with water only on the surface. Thus, the amount of water that adsorbs along the 605 particles is smaller than for proteins.