To synthesize the photo-crosslinkable polymers, we first dissolved HPC in THF (tetrahydrofuran) and then brought it to reaction with fatty acid chlorides. To generate a hydrophobic polymer, HPC was treated with stearoyl chloride (6 equiv.) and 10-undecenoyl chloride (1 equiv.) in a one-pot synthesis to afford a mixed HPC ester 1. Details of the synthesis can be found in the Supplementary Material. The reaction led to degrees of substitution of 2.75 for stearoyl and 0.25 for undecenoyl moieties (DS were determined by NMR, see the Supplementary Material; Figure 1). In addition to this non-polar mixed ester, we synthesized a polar, hydrophilic polymer via the same method, which after cross-linking yielded a hydrogel rather than an organogel. To this end, we treated HPC with small amounts (0.5 equiv.) of 10 undecenoyl chloride so that hydrophilic reference ester 2 was obtained in quantitative yield, exhibiting a DS of 0.3 (Figure 1). Finally, the chemical structure of each polymer was characterized according to our recently published work (for details see Supplementary Material).

Note that stearoyl esters of cellulose are capable to form complexes with proteins such as bovine serum albumin (Niegelhell et al., 2017). For this reason, the formation of stearoyl esters, and tailored networks thereof is of particular interest.

Next, we were interested in designing organogels by attaching polymer 1 to the surface of a planar solid substrate in a spatially controlled fashion using a photo-initiator. Benzophenone derivatives, amongst others, are capable of generating radicals by illumination. These well-established photo-initiators provide various highly beneficial properties for the crosslinking of many systems: The activation and radical formation of benzophenone (BP) itself can be accomplished at relatively high wavelengths when compared to similar compounds (the upper absorption maximum of BP is located at λ = 340 nm; Allen et al., 1990). Consequently, it is not necessary to use high-energy light, which would otherwise be likely to damage sensitive molecules (Riga et al., 2017). Furthermore, radical formation with BP is a reversible process, implying high efficiency as an initiator (Dorman and Prestwich, 1994). If BP is functionalised with electron-donating or electron-withdrawing groups, the activation wavelength can be shifted, which allows the use of a broad range of light sources (Allen et al., 1990). To form the spatially controlled network of 1, we decided to use a commercial laser diode (λ = 405 nm) as the light source because it is readily available and inexpensive. In addition, 405 nm light can be produced by light emitting diodes (LEDs), which opens the opportunity for the high-efficiency implementation of this method in a larger scale setting, in contrast to the use of mercury vapor lamps to generate UV radiation. Therefore, Michler's ketone [4,4′-bis(diethylamino)-benzophenone, DEABP] was found to be the most suitable initiator (see the Supplementary Material). To cross-link the reactive groups within the polysaccharide with functional groups on the solid substrate, appropriate pre-functionalization of the planar surface was carried out. We treated a glass slide with a mixture of allyltriethoxysilane (All-TES) and tetraethyl orthosilicate (TEOS; Ratio 15:85) by the method of Andrieu-Brunsen and co-workers (Krohm et al., 2016). Extensive characterization of this surface was reported in the literature (Krohm et al., 2016). Thus, a stable and homogeneous surface providing allyl functionalities was obtained. Note, allyl groups of All-TES are predestined for radical reactions and thus enable the polymer to be attached to the surface by a radical reaction pathway (Burkhard, 1950).

A DEABP-containing polymer solution in chloroform was solvent-casted onto the allyl-modified glass slide followed by air drying and treatment with short pulses of laser light (laser-diode, 1,000 mW, 100 μm spot dia., λ = 405 nm). The actual spatial resolution of this setup is currently limited by said spot diameter. A scheme of this process is shown in Figure 2A. By installing the laser in a computer numerically controlled (CNC) x/y movement system (see Supplementary Material), the production of polymer networks with well-defined geometries was easily accomplished (resolution: 350 DPI, limited by the control system; see Figure 2B). Finally, any unbound polymer was removed by rinsing with CHCl₃, yielding a surface-bound swellable polymer network in well-defined regions (see Figure 2).

Next, we examined the influence of the illumination time on the generation of the polymer network by laser light. To this end, the pulse-lengths during illumination were altered between 5 and 40 ms, and the swelling behavior of the spatially confined and surface-attached polymer networks was characterized by analyzing the linear deformation parameter α (one-dimensional relative swelling degree: the thickness of the swollen polymer film divided by the thickness of the dry polymer film) in equilibrium swelling. While α refers to the linear deformation in general we introduce α_m to as an additional parameter to denote α in equilibrium state. Therefore, the film thicknesses were determined by confocal laser scanning microscopy (CLSM; for details of the analysis see the Supplementary Material, for still frames captured at different times see Figure 3A). Note that DEABP itself is non-fluorescent when excited at a wavelength of 488 nm (the wavelength used in the CLSM measurements, see below) but becomes fluorescent after the crosslinking reaction due to alteration of the electronic structure of the molecule. The use of CLSM not only allows a static determination of the film thickness in equilibrium swelling but also enables dynamic monitoring of the swelling process (resolution: 460 ms; Figure 3B). For the comparison of the different samples, the parameter α is always referenced to their respective dry film thickness. By illuminating the deposited polymer film with laser light for 5 ms per spot, sufficient network points can be generated to form a surface-bound polymer network. If the illumination time is increased to approximately 30 ms, the degree of equilibrium linear deformation α_m of the polymer network exponentially decreases from α_{m, 5ms} = 5.8 to α_{m, 30ms} = 2.2, to a value that is in good agreement with the theoretical value of a quantitatively crosslinked polymer network in bulk (for details of the calculation of this particular value, see Supplementary Material; Figure 3C, square symbols). If the illumination time is further increased to 40 ms, the polymer network detaches from the surface. The latter phenomenon may be caused by decreased network flexibility, which leads to high mechanical forces and cohesive failure during swelling.

Because of their differences in chemical network structure, the individual samples show distinctly different swelling kinetics: If the netpoint density n is low, the diffusion of the solvent molecules to the substrate near the polymer layers is essentially unhindered by diffusion through the gel. Thus, the swelling behavior appears to follow first-order kinetics, as discussed for a similar system in the literature (Figure 3B, black line; Schott, 2006). If the netpoint density is increased, then the local viscosity increases, and the swelling kinetics becomes a product of the actual thermodynamically controlled swelling and diffusion processes, leading to a more complex behavior that may no longer be described by simple first-order kinetics (Figure 3B, purple line; Schott, 2006).

To learn more about the network structure from the swelling behavior, the netpoint density n of the surface-attached polymer gel can be estimated by using the Flory-Rhener equation. Because swelling is hindered by surface linkages, only one-dimensional swelling can occur. Consequently, to determine n, the linear deformation α_m may be considered instead of using the volumetric degree of swelling q_m (Toomey et al., 2004). Rühe and co-workers demonstrated that with surface-attached polymer networks, q_m scales as αm9/5 and not as αm3, as could be expected, because the surface-attached networks can only swell in one direction, i.e., away from the surface (Toomey et al., 2004). Finally, the Flory-Rhener equation can be transformed into equation 1 for the determination of the netpoint density of our polymer (for further details, see Supplementary Material).

In equation 1, V₁ represents the molar volume of the solvent (i.e., 80.66 cm3/mol for chloroform), and χ is the characteristic Flory-Huggins polymer-solvent interaction parameter. In our case here, this parameter was estimated to be χ = 0.48 for 1 in chloroform at 22°C, according to incremental calculations by the method of Hoy (as shown in the Supplementary Material; van Krevelen and te Nijenhuis, 2009). As calculated from equation 1, the netpoint density of the polymer gel increases from n_5ms = 2.4 × 10⁻⁶ mol/cm3 to n_30ms = 1.8 × 10⁻⁴ mol/cm3 (Figure 3C, triangular symbols). Note that the latter apparently corresponds to the calculated maximum of n (n_theor.max = 1.1 × 10⁻⁴ mol/cm3, determined from the number of crosslinkable double bonds, see Supplementary Material) if the numerous approximations within the Flory-Rhener theory and Hoy's incremental calculation are taken into account (Valentín et al., 2008). Based on these findings, we are able to tailor the netpoint density to desired values simply by modifying the crosslinking parameters, without alteration of the polymer itself. This result demonstrates the versatility of our surface-attached cellulose-derived polymer network.

Because the prepared organogels can be attached to solid substrates, we were also interested in using the gels in a microfluidic demonstration device for the separation of an organic model pollutant from an aqueous solution. For this proof of principle a polymer patch, either made from polymer 1 or 2 was generated within a microfluidic channel comprising two glass slides held together by double-sided adhesive tape and with a capillary gap of 100 μm (Figure 4A). As a model substance for organic molecules (“pollutants”) in water, a saturated aqueous solution of pyrene was transported through the channel via capillary flow, and the gray values at different channel areas were determined during the experiment (Figure 4B; Supplementary Material). Note that the use of pyrene at the same time serves as a sensor due to its intrinsic fluorescence. If the solution comes into contact with the patch of crosslinked polymer 1, pyrene accumulates within the patch to a very large extent, and the gray value (i.e., fluorescence) of the polymer patch increases as a result. Consequently, the mean gray value in the same area before the solution has passed through the patch is significantly higher than the value afterwards (Figure 4B), demonstrating the absorption and upconcentration of the pyrene in the patch. Note that pyrene did not accumulate in the channels that were composed of hydrophilic network patches of polymer 2, because hydrophobic interactions are intrinsically prevented due to missing hydrophobic side chains as compared to the network composed of polymer 1 (see Figure 4C). Although the results reported here are qualitative, nonetheless, this simple experiment demonstrates that the surface-attached organogel is capable of upconcentrating non-polar organic molecules from solution.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.