Spatioselective surface chemistry for the production of functional and chemically anisotropic nanocellulose colloids

Maximizing the bene ﬁ ts of nanomaterials from biomass requires unique considerations associated with their native chemical and physical structure. Both cellulose nano ﬁ brils and nanocrystals are extracted from cellulose ﬁ bers via a top – down approach and have signi ﬁ cantly advanced materials chemistry and set new benchmarks in the last decade. One major challenge has been to prepare de ﬁ ned and selectively modi ﬁ ed nanocelluloses, which would, e.g. , allow optimal particle interactions and thereby further improve the properties of processed materials. At the molecular and crystallite level, the surface of nanocelluloses o ﬀ ers an alternating chemical structure and functional groups of di ﬀ erent reactivity, enabling straightforward avenues towards chemically anisotropic and molecularly patterned nanoparticles via spatioselective chemical modi ﬁ cation. In this review, we will explain the in ﬂ uence and role of the multiscale hierarchy of cellulose ﬁ bers in chemical modi ﬁ cations, and critically discuss recent advances in selective surface chemistry of nanocelluloses. Finally, we will demonstrate the potential of those chemically anisotropic nanocelluloses in materials science and discuss challenges and opportunities in this ﬁ eld

Dr Katja Heise is an Academy of Finland postdoctoral researcher currently working at Aalto University (Finland) in the group of Prof. Eero Kontturi. She received her PhD in 2017 from TU Dresden (Germany). Her current research lies at the interface between nanocellulose materials science and organic synthesis.
Dr Tetyana Koso is a postdoctoral researcher, currently working at the University of Helsinki (Finland) in the group of Prof. Dr Ilkka Kilpeläinen. She received her PhD degree in 2013 from Justus-Liebig University (Germany). Her current areas of research are cellulose modication fundamentals and NMR analytics of cellulose.

Introduction
The utilization of bio-based colloids in materials science has witnessed exponential growth during the past decade, 1 which has been largely amplied by the climate crisis, 2 hazardous issues associated with the lifecycle of synthetic thermoplastics, 3 and the global urge to shi to a bio-based economy. This renders the development of renewable functional nanomaterials and sustainable materials chemistry most timely and necessary. In this context colloids, such as cellulose nanocrystals and cellulose nanobrils, are at the forefront of materials science. Both are based on the structural building blocks of wood cell walls, i.e. the elementary brils (also called microbrils), which have a diameter of approx. 2-4 nm (Fig. 1) and extremely high tensile strength and modulus of up to 7 GPa 4,5 and 140 GPa, [6][7][8] respectively. This exceeds the corresponding properties of most metals, synthetic polymers, and even many ceramics (especially considering the low cellulose density), 5 and explains the strong and rising interest in nanocellulose-based materials.
In this review, we will explain the inuence and role of the multiscale hierarchy of cellulose bers in chemical modications, critically discuss recent advances in selective surface chemistry of nanocelluloses and demonstrate the potential of those chemically anisotropic nanocelluloses in materials science.
Cellulose has three available surface hydroxyl groups, the primary C6-OH as well as the two secondary C2-OH and C3-OH ( Fig. 1A and B) with each having specic reactivity. Besides bearing accessible hydroxyls, nanocelluloses have aldehyde (hemiacetal) groupsthe so-called reducing end-groups (REGs) on one side of the polymer chain, which is in the case of cellulose nanocrystals located at the end of the particle (Fig. 1C). The functional groups can be selectively reacted to yield chemically anisotropic nanocelluloses, e.g., C6-OH- (Fig. 1A), C2-OH and C3-OH- (Fig. 1B), or end group-functionalized nanocelluloses (Fig. 1C). These chemical anisotropies at the nanocellulose surface represent a hitherto barely exploited potential in materials chemistry, enabling highly selective molecular patterning to netune the surface chemistry and interactions of these nanoparticles.
Alistair is a 'Principal Scientist' at VTT -The Technical Research Center of Finland and a Docent in Organic Chemistry at the University of Helsinki. His main research area is in cellulose chemistry and technology, with the bulk of his career dedicated to the development of cellulose processing technologies and analytics, using ionic liquids as a platform technology. All reactions, either in ber or nanoparticle scale, occur at a solid interface, the respective particle surface, and it is important to emphasize that typical nanocelluloses are produced via a top-down approach, from the cellulose ber into a nanoparticle (Fig. 2). This is in contrast to most organic nanoparticles, which are generally assembled via a bottom-up approach from dened and soluble synthetic precursors, i.e., polymers or monomers. Despite the relatively simple chemical structure of cellulose chains based on b-O-1,4-linked glucopyranose repeating units, 9 the structure of the cellulose ber is complex and composed of hierarchically ordered rigid elementary brils (Fig. 2). We use cellulose ber as a term to describe a processed ber of high cellulose content.
The elementary brils represent the smallest particle subunit (further information in Section 2). During deconstruction into nanocellulose, the cellulose ber is broken down into Fig. 1 Nanocelluloses, i.e., cellulose nanofibrils and cellulose nanocrystals, feature a native anisotropic chemical structure due to the presence of surface functional groups of varying reactivity at their surface (A and B) and end (C). This offers a unique opportunity to modify selectively either at the positions of C6-OH (orange balls, A) or C2-, C3-OHs (yellow balls, B); and the reducing end-group (highlighted in purple, C); giving straightforward access to spatioselectively modified nanoparticles. The reducing end-group is in an equilibrium of the unreactive ring and the reactive chain form bearing an aldehyde group (inset in C). bundles of elementary brils ( Fig. 2A, cellulose nanobers with a diameter below 100 nm), individual cellulose nanobrils (CNFs, 2-4 nm diameter), or into shorter, more crystalline cellulose nanocrystals (Fig. 2B, CNCs). 1,10 Here, we will focus on the rather dened, individualized nanocelluloses, CNFs and CNCs.
The size and reactivity of nanocelluloses are ultimately dictated by the native hierarchical structure of the cellulose ber, e.g., the size of the microbril (Section 2), the composition of the cellulose ber, and the used chemical treatments during the deconstruction into nanocelluloses (Section 3.1). In the latter sections, we introduce the current state of the art in surface chemistry of nanocelluloses and add the perspective of regioselectivity (Section 3.2-3.4). Finally, we associate the impact of regioselectivity with the possibility to tune and control the properties of functional nanocelluloses (Section 4) for designing the next-generation cellulosic nanomaterials.

Cellulose microfibril structure and reactivity
Cellulose is produced in nature by plants and other organisms, including some bacteria and animals. It is synthesized by cellulose-synthesizing complexes (CSCs), which are enzyme complexes that polymerize hundreds or thousands of uridine diphosphate-glucose units into b-1,4-linked glucan chains. 13 The newly formed cellulose chains are extruded into the medium surrounding the cell membrane, in which they form semicrystalline aggregates called microbrils (or elementary brils). Due to a rotation of the glycosidic linkage for each glucose monomer unit, the chain obtains a two-fold helical symmetry leading to alternating functionalities of C6-OH, and C2-and C3-OH groups on the microbril surface ( Fig. 1A and B). 14 Cellulose in nature mostly exists in the cellulose I crystal structure, whereas cellulose II is formed upon mercerization or regeneration from dissolved cellulose. 15 Cellulose II is characterized by the anti-parallel arrangement of the cellulose chains. 16 In the native cellulose I, which exists in two forms called cellulose Ia and Ib, parallel chains are packed in layers with the pyranose rings oriented in the plane. 17,18 These layers are held together by hydrogen bonding and a smaller contribution from London dispersion forces (a type of van der Waals forces), whereas the stacking of the layers is dominated by the dispersion forces. 19,20 The morphology of the microbrils directly affects the hydrophobicity and hydrophilicity of the crystallite surfaces and the chemical groups available for modication.
The exact shape of the wood microbril cross-section has not been fully veried yet, whereas the most plausible candidates are based on an 18, 21,22 or 24-chain model (Fig. 3A). [23][24][25] The most common models of an average microbril expose crystal surfaces with both hydrophobic (200 in cellulose Ib) and hydrophilic characters (1-10 and 110 in cellulose Ib). Furthermore, the currently most promising candidate for the 18-chain microbril, i.e. one with 2, 3, 4, 4, 3, and 2 chains in the layers, offers two possible arrangements corresponding to different proportions of the 1-10 and 110 surfaces (one of which is shown in Fig. 3A). 26 The interaction of water molecules shown in Fig. 3C is hence selective to the hydrophilic 1-10 and 110 planes, 27 and nanocelluloses are considered amphiphilic particles. 28 The crystalline arrangement of cellulose chains in the microbril is susceptible to slight modications for instance due to mechanical stresses and interactions with moisture. [29][30][31] More drastic changes in the microbril structure may be induced by partially irreversible aggregation/cocrystallization of microbrils, e.g., during drying at high temperatures, or disintegration and partial decrystallization of microbril aggregates during the deconstruction of the cellulose ber into CNFs (Fig. 3B). 26,32,33 Therefore, the shape of a cellulose microbril cross-section and the number of chains forming it may vary due to both biological reasons and the processing history of the material.
Due to the small crystal size and the high ratio of surface to inner chains in the microbrils of land plants, their degree of crystallinity is relatively low. Here the degree of crystallinity should be understood as an indicator of the degree of order in the crystallites, which is different from synthetic polymers exhibiting separated amorphous and crystalline domains. In this regard, the plant cell wall may be regarded as mesocrystalline, with regions where there are varying degrees of disorder. In general, the cellulose chains located at the surface are chemically accessible and can be considered to be less crystalline, i.e., less ordered, whereas the chains in the Fig. 3 (A) Cellulose chains assemble in the form of microfibrils, shown exemplarily in the form of the 18-and 24-chain models. The microfibril size defined by the chain arrangement and number dictates the relative amount of accessible hydroxyl groups at the microfibril surface (access. OH in mol mol −1 ). (B) Interfibrillar interactions between microfibrils, e.g., in a microfibril bundle, may lead to aggregation decreasing the available OH at the surface and increasing the crystallinity. The disintegration of such aggregates, e.g., during the preparation of CNFs exposes OH-groups and decreases sample crystallinity. The chain arrangements in the microfibrils expose hydrophobic (200) and hydrophilic (110, 1-10) crystal surfaces and water molecules interact preferably with hydrophilic crystal surfaces, as shown in the modeled structure (C). (C) was adapted with permission from Paajanen et al. 27 Copyright 2019 Springer Nature. The plot in A was drawn according to Okita et al. 34 microbril core are chemically inaccessible and more crystalline, i.e., more ordered (Fig. 2). Heterogeneous reactions or surface reactions of cellulose are hence conned to the surface hydroxyl groups. 1 Larger crystallites are produced, e.g., by algae, 35 cotton, 36 and bacteria, 37 and give access to nanocelluloses of bigger bril diameter and different aspect ratios. 38,39 As shown in Fig. 3A, the amount of accessible OHs decreases with increasing crystallite size, simply because the relative number of surface chains is reduced. The size and structure of the microbril are crucial to predict the number of accessible hydroxyl groups ( Fig. 3A and B). In the case of woodderived cellulose (18 or 24 chains per microbril), 25-33% of the total hydroxyl groups in the microbril are at the surface and hence chemically accessible, which corresponds to a degree of substitution (DS) of 0.75-1.00 (total amount of OH-groups per monomer unit is 3, the maximum DS is 3). In comparison, aggregated cellulose microbrils, e.g., in a ber construct (Fig. 2), feature a lower reactivity as fewer hydroxyl groups are available at the surface (Fig. 3B). If a reaction proceeds into the core (containing inaccessible OH groups) of the microbril, for instance, due to harsher reaction conditions, it can be categorized as a bulk modication. Such modication alters the physical properties of cellulose, such as transparency, and mechanical or thermal properties. 40 This is exemplarily shown for acetylated cellulose in Fig. 4B, comparing the tensile strength and elastic modulus of a surface-acetylated sample (DS = 0.2) with a highly substituted cellulose acetate (DS = 2.5), the elastic modulus and tensile strength are signicantly reduced in the case of the cellulose acetate sample. Consequently, chemical modication in the course of the production of functional nanocelluloses should be conned to the surface, to preserve their intrinsic properties.
The reactivity of the hydroxyl groups decreases in the following order, C6-OH > C2-OH > C3-OH, although all three hydroxyl groups are in theory accessible at the surface of cellulose brils. Experiments have shown that the C3-OH is hardly reactive due to steric effects and intrachain hydrogen bonding. [41][42][43][44] Consequently, the DS threshold for surface modications (Fig. 3A), such as acetylation, is lower as only C6-OH and C2-OH are chemically accessible under heterogeneous conditions. This is also supported by the literature that discusses an upper esterication limit of approximately 66% in surface esterications. 43,45,46 Hence the extent of a surface modication largely inuences the mechanical properties of celluloses ( Fig. 4B and D), and the inuence of the modication on the properties has to be constantly monitored to reach an optimal level, even at relatively low DS values. 47 Common acetylation of cellulose as shown in Fig. 4C is not regioselective, however, one can assume that until surface acetylation of 30-40% most acetylation occurs at the primary C6-OH due to its higher reactivity. An improvement in mechanical properties for C6-OH modied samples has been also shown in the case of CNFs. 48 A surface DS above 40% and hence modications at C2-OH and C3-OH have been shown to reduce signicantly the tensile strength.
This can be most probably attributed to a partial modication of C 3 -OH, which will ultimately affect the crystalline domains and cause polymer degradation and occur below the theoretical limit of surface acetylation, thereby diminishing mechanical properties. 49,50 Similar observations were also made in the esterication of CNFs and CNCs, which conrm that a low DS (even below the limit of surface modication) is recommended to avoid fragmentation and degradation of nanocelluloses. 51,52 Controlling the extent of reaction and surface-connement is hence of utmost importance, and can be achieved through comparison of crystallinity before and aer modication, 42 but is still not very common in the literature.
Solvent interactions of cellulose particles inuence their dispersibility, swelling, and chemical accessibility in a given solvent, and are thereby also a crucial aspect inuencing the cellulose reactivity. The abundance of hydrophilic surface hydroxyl groups in cellulose explains its hygroscopicity and sorption-induced swelling. 55,56 The main contributors to the water interactions are the surfaces of the microbrils, 57,58 that are covered with a layer of water, under ambient conditions (Fig. 3C). 27,55 Due to the strong cellulose-water (or hemicellulose-water) interactions, this surface-bound or otherwise spatially conned water can be distinguished from bulk water by its properties and is referred to as bound water. [59][60][61][62] Unprocessed, native wood cell walls swell by water adsorption on the crystallite surfaces and in the hemicelluloses separating the cellulose microbrils. 30 Similarly, swelling in processed celluloses may be attributed to residual hemicelluloses or interfacial disorder between neighboring microbrils. Such disorder in pure celluloses could be caused by a mismatch of the crystallite orientations, for instance, due to twisting. 27 From the viewpoint of chemical modication in aqueous environments, the location of water in the cellulose nanostructure and its interactions with the microbril bundles are crucial. The swelling of the microbril bundles and the size of water clusters, the connectivity of their network, and the width of water channels control the accessibility of soluble chemical agents to the lower levels of the hierarchical structure. All of these are affected by the moisture content and the origin and processing history of the cellulosic sample, which modify the aggregation state and crystallinity of the cellulose microbrils. The amount of conned water in the ber structure is especially important in solid-state 43 or gas-phase reaction 63 and will be further discussed in Section 3.3.

Preparation of functional nanocelluloses
In this section, we will rst introduce general methods to isolate CNFs and CNCs from the ber source (Section 3.1) and basic avenues to produce chemically functionalized nanocelluloses (Section 3.2). We further introduce and discuss in detail pathways toward regioselectively modied nanocelluloses, which ultimately enable in a controlled manner spatially conned modication at the nanoparticle surface (Section 3.3). An overview of analytical methods to characterize those functional nanocelluloses is presented in Section 3.4.

Deconstruction: from bers to nanocelluloses
The production of nanocelluloses, i.e. CNF and CNCs, has been well reviewed 1,64-66 and will, therefore, be only briey discussed. In general, completely individualized CNFs are produced by a chemical pretreatment of the cellulose ber, modifying the ber surface ( Fig. 5A 1 ), and subsequent mechanical brillation. The purpose of the modication step is to decrease interbrillar interactions, either through the introduction of repulsive ionic charges (e.g., carboxylate or quarternary ammonium groups) or in general by decreasing the number of available hydroxyl groups for hydrogen bonding (e.g., aliphatic ester groups). 1 Consequently, a controlled modication signicantly reduces the energy required for mechanical brillation, which is usually conducted via high-pressure homogenization, 67-69 microuidization, [70][71][72] or high-intensity ultrasonication, 73,74 and facilitates the deconstruction of the ber into individualized cellulose microbrils. In addition, from a practical viewpoint, the chemical modication at the ber level has the advantage of a more straightforward and efficient purication, since remaining reactants or solvents can be easily removed by normal centrifugation or ltration methods, and reactions can be conducted at high consistency to improve their efficiency. 75 What we obtain through this combined isolation pathway, however, is essentially a cellulose derivative, at least on the nanobril surface, which has properties that signicantly differ from that of non-modied cellulose.
Similar to CNFs, the introduction of repulsive charges facilitates the isolation of individualized CNCs and provides them with their unique colloidal properties. The different production pathways and sources are reviewed in detail. [76][77][78] The most common pathway affording the rod-shaped nanocrystals is hydrolysis in 64-wt% sulfuric acid, which introduces sulfate half-ester (-OSO 3 − ) groups to the nanocrystal surface through esterication. Alternatively, aqueous HCl 79,80 or HBr 81,82 have been used. However, the resulting non-charged nanocrystals have poor dispersibility in water. Gaseous HCl also affords CNCs at a high yield (>80%) and has been combined with a subsequent TEMPO-oxidation and high-intensity ultrasonication to sufficiently disintegrate and disperse the hydrolyzed material. 83 In situ introduction of charges to the CNC surface during the top-down isolation, similarly to the H 2 SO 4 pathway but far less common, can be also achieved with phosphoric acid [84][85][86] or through hydrolysis with organic acids (e.g., oxalic or formic acid) 87,88 esterifying the nanocrystal surface with anionic -OPO 3 − or -COR groups (where R is any substituent), respectively. Ultrasonication or high-pressure uidization is in all cases an indispensable tool to ensure complete individualization of the CNCs and to disintegrate potential nanocrystal aggregates aer acid hydrolysis. An interesting pathway toward functionalized CNCs is the combination of acid hydrolysis and Fischer esterication, which has been used to attach a whole range of functionalities, from polymerization initiators, to double or triple bonds, to the CNC surface in situ during its production. 89

Preparation of functional nanocelluloses
Nanocelluloses decorated with chemical moieties, e.g., ester or oxidized groups can be obtained by conducting surface chemistry on a micrometer or nanometer-sized particles, i.e., the cellulose ber or the native nanocellulose. 1 Hence, the chemical modication is conducted either before (ber modication, Fig. 5A) 48,53,90 or aer the deconstruction of the ber into a colloidal state (nanoparticle modication, Fig. 5B). 44 Alternatively, there have been also endeavors to isolate modied nanocellulose directly from bers or even raw biomass ( Fig. 5C and C 1 ), by a combination of chemical modication and deconstruction in a single step. [91][92][93] Choosing the order of processing steps is highly dependent on the used starting material, the targeted functionality, and the type of nanocellulose, e.g., modications of CNCs are oen conducted starting from dried CNCs (Fig. 5B) since CNCs can be dispersed from the dry state due to their morphology. In addition, many functionalities, e.g., ester groups, would be removed during the acidic hydrolysis step of CNC production. Modication of CNFs with the same pathway (nanoparticle modication) is conducted in dispersion with low solid content (mostly below 5 wt%), which renders most chemical modications inefficient. The reactions are generally conducted in a never-dried state, since the large aspect ratio of CNFs and their high cohesive interactions, cause the formation of irreversible aggregates upon drying or solvent removal. 94 CNF reactions are either conducted directly in aqueous dispersion, 95,96 in which water can be disturbing as it might react with the used reactants (e.g., anhydride, acid chlorides), or in organic solvents, which require a solvent exchange. 97 Taking this into account, the ber modication ( Fig. 5A) is oen selected for the preparation of functional CNFs, 11,53,90 since it can be conducted at higher solid content to increase the reaction efficiency, 75 and offers a more straight-forward purication step due to the larger ber size, which does not require ultrane membrane ltration or dialysis steps.
The composition of the used cellulose is important to consider since residual non-cellulosic biopolymers can inuence the chemical modication, 48 or adsorb onto colloids inuencing greatly their properties, 98-101 and reactivity. 48 Further details on the extraction of cellulose from wood 1,102 and other sources 103 are accessible in the literature. For the simple reason of simplication and increasing the control of chemical reactions, most modications of (nano)celluloses are performed on cellulose bers (>90% cellulose purity), 90,95,102,104 socalled dissolving pulp, which contrast with raw or paper-grade cellulosic bers containing signicant amounts of lignin and hemicelluloses.

Regioselective modication of nanocelluloses
We encompass in this discussion selective chemical modications at a certain type of surface hydroxyl groups (primary C6vs. secondary C2-, C3-OHs, Fig. 1A and B, Sections 3.2.1 and 3.2.2) or the reducing end-group (Fig. 1C, Section 3.2.3). The focus lies on methods, which are straightforward, mild, and have none or only minor inuences on the physical properties of nanocelluloses (crystallinity, molar mass, etc.); and are hence suitable for their functionalization. This stays in contrast to traditional chemical pathways toward regioselectively modied celluloses, 105-107 which are not surface-conned, require multistep syntheses routes including protective groups (these groups are priorly introduced to ensure modication at a specic, free hydroxyl group), and/or are conducted under harsh conditions.
Concerning surface hydroxyl group modication, we distinguish between different reaction classes, (a) esterications as a reversible chemical substitution of the cellulose backbone (Section 3.2.1, schemes of esteried nanocelluloses are colored in blue), and (b) oxidations, which are irreversible and change the chemical structure of the cellulose backbone at the bril surface (Section 3.2.2, schemes of oxidized nanocelluloses are colored in orange or yellow).
3.3.1 Esterications (C6-OH selectivity). Esterication and mostly acetylation are one of the most successful commercial reactions to produce cellulose-based chemicals. Traditionally, it is performed under heterogeneous conditions using acetic anhydride in acetic acid, with sulphuric acid as catalyst and it is a very simple method to modify free hydroxyl groups on cellulose surfaces and, thus, reduce hydrogen-bonding. [108][109][110] This reaction itself is not regioselective, but in general follows the order of reactivity of the hydroxyl groups, as detailed in Section 2. Such reactions might be at the beginning more preferable at C6-OH, but acetylation of C2-OH or C3-OH is likely to occur to a certain degree in parallel, and the reaction proceeds into the bulk/crystalline regions of the cellulose. 111 In the industry, the focus still lies on the fabrication of cellulose with very high or complete bulk acetylation. Bulk acetylation completely changes the native microbril structure and properties, and hence those esterication methods are not suitable for CNFs or CNCs, where crystallinity and morphology should be preserved.
If the traditional H 2 SO 4 -catalysed acetylation conditions are applied to CNCs, practically all the original cellulose I crystallinity is removed and acetylation DS values reach approx. 2. 112 This process yields also lower DS acetylated CNCs of cellulose II crystal structure, which have a rod-like shape. 113 As acetylation beyond surface modication causes dissolution of cellulose, this phenomenon can be also benecial, e.g., to extract acetylated cellulose nanobers in an energy-efficient manner from a swollen, partially dissolved cellulose matrix. 114,115 Apart from supporting the brillation, it has to be taken into account that mechanical performance will be diminished (Fig. 4B), and signicant amounts of solvents are necessary during the process. Tailoring reactive agent quantities and reaction time can allow for maintaining low DS values where, we presume, the acetylation is conned to the surface, and enable the preparation of acetylated CNCs 116 and CNFs. 111 Reactive ballmilling in aprotic polar organic solvents, such as DMSO and DMF, in presence of anhydrides or acid chlorides without any additional catalysts, is surface-conned and enables energyefficient production of hydrophobic CNF esters with variable ester length. 117-119 Thereby wood-based CNFs were prepared with a DS of approx. 0.5; which is in the range of full surface modication of C6-and C2-OH. 43 Not many articles are paying attention to the spatio-and regioselectivity of esterications. The actual selectivity of a reaction is hardly determined and mainly predicted by comparing the actual DS with the theoretical one corresponding to surface modication, and by proving that the crystallinity of the sample was preserved during modication. 42 Hence, selectivity for esterication of C6-OH vs. C2-OH or C3-OH was only assumable, as the most popular cellulose analysis methods oen do not provide sufficient resolution. 13 C CP MAS NMR analysis has been used to provide proof of selective C6-OH esterication, 120 however, the resolution with this method is too limited to exclude modications of C2-OH and C3-OH at lower DS. This is very important as it has been recently shown that in presence of a catalyst or at elevated temperatures, the acetylation is favored at C6-OH of CNC, but is not highly selective, as also modication of secondary OHs (C2-OH and C3-OH) is occurring at signicant levels. 44 The more traditional method of perpropionylation, which is used for regioselectivity determination of esteried cellulose under homogeneous conditions 121 is also limited due to the rather low DS values in surface-modied nanocelluloses, poor signal-to-noise of the 13 C NMR carbonyl signals, and has to the best of our knowledge not been applied for regioselectivity determination on esteried CNCs or CNFs. However, quite recently it became possible to perform high-resolution solution-state NMR on (nano)celluloses by using the ionic liquid-electrolyte [P 4444 ][OAc] : DMSO-d 6 as solvent (Fig. 6C). 43,122,123 This has given the ability to run the more sensitive and relatively high-resolution quantitative 1 H NMR on CNCs 44,123 and CNFs, 43,53 in addition to even higher resolved 2D experiments and 1D quantitative 13 C experiments. This analytical method enabled the development of esterication methods of high regioselectivity via modication of cellulose either in solid, suspension, or gas states. The reaction of Nacetylimidazole enables modication of C6-OH at high selectivity, can be conducted in a solid state, and is promoted by conned water in the hydration layer of cellulose bers (Fig. 6A and D 1 ). 43 The high efficiency of this reaction was explained by the fact that the reaction takes place in the conned water layer of cellulose, which covers all microbril surfaces. The results indicate that these water clusters are connected continuously or allow at least the diffusion of solubilized reactant along internal brils. Moreover, it has been shown that increasing the thickness of this layer increases the regioselectivity towards C6-OH of cellulose and enables thereby highly selective reactions (Fig. 6D). This system can be also extended to the preparation of functional CNFs, cellulose bers have been in this case directly modied with N-acylimidazoles in a mixture of water and acetone. 43,48,53 The N-acylimidazoles can be either directly added or prepared in situ from the corresponding carboxylic acid anhydrides and imidazole. Upon esterication, the functional bers were brillated to yield CNFs decorated with either acetyl, iso-butyl, 53 or succinate groups. 90 Introduction of negative charge via succinate groups enabled complete individualization (through increased electrostatic repulsion) of succinylated CNFs, 90 comparable to the well-known TEMPO-oxidized CNFs. 124 Apart from esterications with acylimidazoles in presence of water, gas-phase acetylation of CNCs with acetic anhydride is highly regioselective for the surface C6-OH positions (Fig. 6B). 44 Further to this, a comparison between the uncatalyzed gas-and liquid-phase reactions yielded similar reaction kinetics. However, the gas-phase reaction was found to be more regioselective, supporting the concept that the correct choice of solvent or complete absence of solvent is required to yield high C6-OH regioselectivity. Moreover, density functional theory transition-state modeling on cellulose I fragment demonstrated that the activation energies for esterication of the three different OH groups using acetic anhydride are rather different, with the C6-OH being much more favored, which explains the high regioselectivity under mild gas-phase conditions (Fig. 6E). The CNC morphology was found unaffected by the gas-phase acetylation, whereas it was noted that the fully acetylated surfaces of CNCs can be peeled off in dipolar aprotic organic solvents, such as DMSO, which is also visible in Fig. 6. 44 Similar observation was also made for TEMPO-oxidized cellulose, 14 and is important to take into account that in case of C6-OAc-CNC, it only occurs in special organic solvents, and not in water, which is the standard solvent. When the liquid-phase reactions were catalyzed using organic bases pyridine and DABCO, a progressive conversion of the outer CNC surfaces, to CTA, towards the core was observed. Based on this knowledge one would expect that under mild and non-swelling conditions it should be possible to control the reaction to only surface C6-OHs. Nevertheless, the reactivity of the acylation agent and temperature are also important, e.g., gas-phase reactions of acid chlorides are non-surface-selective and proceed also in the core of CNC, 120,125 whereas surface-selectivity was achieved when dicarboxylic acid chlorides are used, such as suberoyl chloride. 126 The mechanistic understanding of the surface chemistry of CNCs and CNFs is developing to a point where the connement of reactions to nanocellulose surfaces can be controlled. Further studies of the regioselectivity of various cellulose modications are expected to lead to the establishment of more advanced and versatile methods for chemical modication, which do not have to be limited to esterication reactions but could also encompass other substitution reactions, such as etherications 127 or sulfonation. 128 3.3.2 Oxidations (C2-, C3-OHs, or C6-OH selectivities). Chemical and enzymatic oxidations are established for cellulose with the aim of oxidizing selected hydroxyl groups, to aldehydes and carboxylic groups, or via hydrolysis and oxidization of the glycosidic bond. In the case of enzymatic oxidation with, e.g., lytic polysaccharide monooxygenases (LPMOs), 129,130 oxidation is coupled with simultaneous cleavage of the glycosidic bond, usually at the C 1 and C 4 positions. 129 The cleavage and oxidation lead to products of oxidized solubilized molecules or oxidized cellulose substrates. 131 Enzymatic modi-cations are hence both hydrolyzing and functionalizing and can be applied to the various cellulose hierarchies. However, when it comes to the oxidative function, the balance between liberating functionalized fractions from the solid substrate and preserving the functionality of (nano)celluloses is still to be developed, especially when the latter is the target.
Chemical routes of oxidizing cellulose are dominated by TEMPO-mediated (Fig. 7A) 11,124,132,133 and periodate oxidation (Fig. 7B) 75,104,[134][135][136] as well as their combination. 137,138 TEMPOoxidation is considered selective to oxidize the primary hydroxyl group at the C6 position and proceeds through the Noxoammonium ion formation from TEMPO radical, which converts alcohol into aldehyde, followed by its further oxidation to carboxylic acid by oxidant, usually, hypochlorite. 11 When the aim of a modication via TEMPO-oxidation is to decorate nanocellulose with carboxylic groups at the C6 position, the oxidation can be performed on cellulose bers before disintegration into CNFs, or on the already liberated CNFs. Oxidation of cellulose bers leads to signicant swelling and loosening of the bril networks 133 and hence, aids the brillation process.
Periodate oxidation proceeds by the reaction with the C2-C3 diol, cleavage of this bond, and conversion to two aldehyde groups that can further form hydrates, hemialdals, and, with vicinal hydroxyl groups, hemiacetals. 135,139 Both TEMPO and periodate routes to modify cellulose are considered as regioselective reactions with regards to the position in the monomer unit, also referred to as anhydroglucose unit (AGU), that can be oxidized. It is important to consider that these oxidative conditions usually cause a decrease in the molar mass of the cellulose. 140,141 Nevertheless, TEMPO-mediated oxidation is commonly selected when surface modication of cellulose 34 is targeted and very popular to individualize cellulose bers into high-performance TEMPO-oxidized CNF (TO-CNF). 142 As discussed in Section 3.2, CNCs are conventionally prepared through acid hydrolysis to further deconstruct cellulose's elementary brils, alternatively, cavitation treatment of TO-CNFs enables an acid-free preparation of CNCs. 143 TEMPOoxidation seems to destabilize to a certain extent further defects or weaknesses of CNFs, which would be otherwise attacked in the typical hydrolytic deconstruction. TEMPOoxidation has been also used to modify CNCs obtained through HCl hydrolysis, which would be otherwise unstable due to their low surface charge. 144 In addition, the carboxylate groups introduced through TEMPO-oxidation can also be used to introduce further functionality through amidation reactions, 145 by ion-exchange of Na + with ammonium salts, 146,147 or through a three-component Passerini reaction (Fig. 7A). 148 The periodate oxidation, on the other hand, is principally considered as not highly selective to accessible surfaces and propagates also into crystalline domains. 140,149 However, this is still under debate since recent approaches report reactions that can be controlled to attack preferentially the less ordered regions. 91 Revelation of the simultaneous but different rates of oxidations on the ordered and disordered regions are evaluated to understand the oxidation fundamentals and to allow them for controlled functionalization purposes. When functionalization of CNCs or CNF is targeted, taking the degradation into account, oxidation of cellulose needs to be controlled to limit its inuence on the physical properties, similar to what has been achieved with the wet esterication method using acylimidazole that enable highly regioselective esterication of the C6-OH of cellulose while preserving the cellulose molecular weight and crystallinity. 43,48 Controlling the degree of oxidation to a range of accessible surface hydroxyl groups enables the preparation of celluloses decorated with aldehyde groups, which can be post-modied to yield C2, C3-carboxylated, 104,152 sulfonated, 153 or cationic CNFs in a straightforward manner (Fig. 7B). 154 These repulsive charges support, similar to carboxylate groups, the brillation of the bers into individualized CNFs. Apart from these reactions, the reactivity of aldehyde enables the introduction of a plethora of functional groups. 1,155 Mild periodate oxidation of CNCs and further oxidation to carboxylic groups with ozone have been demonstrated to cellulose Janus lms with side-specic chemical functionality (aldehyde vs. carboxylic acid). 136 Periodate oxidation to a higher extent is an efficient way of modulating cellulose hierarchies, 135 and leads, e.g., to the formation of dangling polymer ends attached to the CNCs, socalled hairy nanocrystals. 156,157 Moreover, it allows the direct preparation of CNCs from micron-sized cellulose or biomass, 158 even in complex systems, such as Pickering emulsions (Fig. 6D). 159 Combined TEMPO and periodate oxidation has been used to produce highly carboxylated and charged nanocellulose, 160 which can self-brillate into CNFs as a response to change in pH. 133 All in all, both periodate 135 and TEMPO-oxidations 11,34,124 have been reported to cause polymer degradation for oxidation degrees in the range of surface modications. Although this inuence on molar mass can be reduced under mild conditions, e.g., through TEMPO-oxidations at neutral/acidic pH, 161,162 it still limits the possible mechanical properties of oxidized CNCs and CNFs. 1,49 In addition, it is important to take into account that treatment of cellulose with such oxidizing agents might as well cause side reactions, e.g., the oxidation of secondary hydroxyl groups. In the case of TEMPO-oxidation, postoxidation is necessary to oxidize the remaining C6-aldehyde groups. 124 Moreover, occurring reactions are getting rather complex in presence of lignocelluloses, e.g., periodate degrades lignin and also oxidizes hemicelluloses, such as xylan. 163 This might be a desired side-effect, but reduces the efficiency of oxidations, and does not allow a selective modication of individual lignocellulosic polymers. In addition, the recycling of periodate is possible but requires a rather complex setup. 164 While TEMPO-oxidation has been recognized to be highly selective to the surface of cellulose brils without affecting the crystallinity of cellulose, 142,143 this is still under debate in the case of periodate oxidation, but unwanted reactions of crystalline, non-accessible regions can at least be kept to a minimum if reaction conditions are carefully chosen. 135 Consideration of the high efficiency of such oxidations, the fact that they are compatible with water as the solvent, and the many possibilities for post-modications, explain the high potential of these reactions and why they are widely used in the preparation of nanocelluloses.
3.3.3 Spatioselective modication of reducing end-groups. End-group modications are conducted exclusively on CNCs, which is due to (a) their lower length and hence the more dominant inuence of the end-group functionality on CNC properties, and (b) the spatial location of the chain end-groups at one side of the CNC. 165 However, the general labeling of endgroups can be also performed on bers or CNFs, and is performed, e.g., for analytical studies; 168 but their locations are not as dened as in the case of CNCs (as those might be situated as well along the ber axis and not only at their end).
Despite the increasing variety of functionalities being graed to CNC REGs, the basic chemistry used to approach the endstanding aldehyde groups is not very broad with only four prevailing mechanisms (Fig. 8A): (1) ligation of hydrazine analogous, [169][170][171][172][173] hydroxylamines, 174 or thiosemicarbazide, 166 (2) Pinnick oxidation forming carboxyl groups, [175][176][177][178] (3) reductive amination attaching primary amines, [179][180][181][182][183][184] and (4) Knoevenagel condensation with dicarbonyls. 174 The peculiarities and the different reaction conditions for each of these mechanisms are described in two recent reviews and will, therefore, not be discussed in detail here. 185,186 Given the nature of CNCs, including their excellent colloidal stability in water, most reactions are carried out in an aqueous environment. This also aids the activation of the REGs as the reactive open-ring aldehyde is in equilibrium with the nonreactive hemiacetal form. In theory, one water molecule 187 should be enough to sufficiently catalyze the proton transfer opening the REG hemiacetal and liberating the reactive aldehyde. However, to further shi the equilibrium to the aldehyde, mildly acidic (i.e., pH 4.5) 181,182 or alkaline (e.g., aqueous bicarbonate at pH 8.5) 174 conditions have been used. In the absence of water, the ring opening can be catalyzed by a mixture of organic acids and bases acting synergistically in a concerted mechanism to enable proton transfer at the REGs. 188 On CNCs, these catalytic possibilities are yet to be explored. 189 Besides the pH and solvent conditions, the reaction temperature, affecting the mutarotation rate, [190][191][192] and the reaction time are relevant for the cellulose end-group conversion. For sufficient conversion rates, the reaction times are usually very long, easily exceeding 48 hours, which might be a major challenge when up-scaling CNC REG-modication to industrial applicability. This limitation, however, might be surrendered with more efficient catalysis.
Especially in the realm of synthesizing Janus-type nanorods from CNCs, the aldehyde-specic modication of the REGs is only the initial, activating step for introducing, e.g., highmolecular-weight compounds like polymers, dendrimers, metal nanoparticles (Fig. 8B) or biomolecules to the CNCs. In general, one has to distinguish between native CNCs of cellulose I crystal structure or CNCs obtained through mercerization with NaOH, which changes their crystal structure into cellulose II (CNC-II). Dependent on the nature of CNCs, end-groups are either located anisotropically at one (CNC) or both ends (CNC-II). This difference is visualized using TEM and CNCs with metal nanoparticle-labeled end-groups (Fig. 8B).
Graing-to approaches are very common in the modication of CNCs, where 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide (EDC/NHS)-mediated amidation to previously oxidized REGs (i.e., via NaOCl 2 Pinnick oxidation) is probably the most popular pathway. [175][176][177][178]180 This also highlights that the rst aldehyde-specic step on the pristine REGs determines the selectivity of the later graing protocols. Zoppe et al., for instance, observed side reactions on the CNC surface aer oxidizing the REGs with sodium chlorite followed by graing polymerization initiators, using a two-step protocol, for subsequent atom-transfer-radical polymerization. 175,177 The result was a patchy distribution of polymer chains rather than a Janus-type architecture. The reason for this side reaction could be impurities of oxidized moieties on the CNC surface or defects in the nanocrystal structure, i.e., REGs that are situated at the nanocrystal surface and not exclusively at the end. In general, we have to remember that we are working with naturally sourced nanomaterials bearing both the imperfections of their top-down isolation (e.g., impurities, surface roughness) and heterogeneous size-and molecular-weight distribution.
Side-reactions during the REG modication can also negatively affect the later polymer gra density, as shown in a recent contribution by Delepierre et al. 181 They observed that signicant proportions of the amino-functionalized initiator were passivated during its attachment to the REGs via a reductive amination protocol, at 70°C and pH 4.5, in a 72 hours reaction. This again highlights how challenging especially multi-step protocols at REGs can be and how important it is to thoroughly understand and select the chemistry. In terms of selectivity, the use of enzymes acting specically at the reducing or non-reducing ends might be a promising and green future avenue toward a Janus-type CNC modicationa concept that is, to the best of our knowledge, still out of reach for modifying nanocelluloses.

Analysis of spatioselectively modied nanocelluloses
Spatial analysis of various functionalities on cellulose hierarchies is challenging and oen requires a combination of techniques delivering qualitative and, if needed, quantitative (DS) information. Visualization and mapping of functionalities spatially in particle hierarchies, such as nanocelluloses, usually requires microscopy and the introduction of bulky functionalities with sufficient contrast, or microscopy combined with spectroscopic techniques. Atomic force microscopy (AFM) is superior for the visualization of morphologies. It also allows the identication of chemical features on surfaces indirectly via adhesion 193 or directly via employing functionalized tips [194][195][196] that have a different affinity to the diverse chemical functionalities on the surface and this can be quantied by measurement of attractive or repulsive force. However, one of the bottlenecks of AFM techniques is the dimensions of the tips dening spatial resolution. For circumventing this, extensions to the traditional AFM techniques are developed. 197 Combining AFM with IR spectroscopy, for example, has been used for the mapping of wood cell wall components with nanometer resolution based on differences in chemical composition. 198 In addition, it is to be expected that further development of scanning electron microscopy (SEM) and transmission electron microscopy (TEM) methods coupled with energy-dispersive Xray spectroscopy (EDX) will be able to reveal further details of the microbril chemistry and allow more detailed chemical mapping. The main obstacles of electron microscopy are the limitations of electron beam radiation to low voltage, to avoid changes in the surface morphology and chemistry of the samples, which limits the resolution of this method. 199 For instance, progresses in SEM-EDX, which allows localized elemental analysis, are expected to increase in resolution, once the beam damage of the specimen can be reduced or avoided.
Alternatively, the self-assembly behavior of spatioselectively modied nanocelluloses, as in the case of end-group functionalized CNCs (Fig. 9B 1 -B 3 ) has been used to visualize and conrm selective chemical modication by microscopic means. For instance, star-shaped CNC assemblies (Fig. 9B 2 ), formed in response to increased temperatures 200 or through end-wise crosslinking, 178 have been visualized by AFM and TEM. Also, labeling with metal nanoparticles (Fig. 8B) is a well-established way to visualize functionalized REGs of CNCs. 166 For simultaneous visualization of structure and chemistry, the combination of microscopic and spectroscopic techniques is advantageous. Imamura et al. combined microscopy with FTIR spectroscopy and visualized carbonyl functionalities on cellulose bers. 201 However, the resolution of the microscopy still limits the resolution of the observation, and reaching the nanoscale is a challenge. CNC ordering in suspension and solids enables structural color and tuning of mechanical and optical properties. Kádár et al. 202 reviewed the combination of Fig. 9 Influence of spatioselective modification of nanocelluloses on nanoparticle and materials properties. C2, C3-dicarboxyl-CNFs (produced by sequential periodate and Pinnick oxidation) forms highly ordered CNF dispersions (A 1 ), which can be processed into aerogels with controlled porosity and alignment (C 1 ) to enable optimal thermal insulation behavior (C 2 ). Anisotropic and isotropic end-group grafting of CNCs can be used to influence their order in liquid crystalline films (A 2 ), and enable self-assembly into star-shaped (D 1 ) or network-like superstructures (D 2 ). Reversible succinylation at the C6-OH enables individualization into CNFs and is reversible to restore the native chemical interface of cellulose (B 1 ). This can be used to produce highly stiff nanopapers of assembled native CNFs (E 1 and E 2 ). TEMPO-CNFs and their regularly patterned carboxyl moieties enable, e.g., high ion conductivity (B 2 ), and allow straight-forward tuning of their degree of oxidation for optimal electricity generation (F 1 ), as well as to limit polymer degradation for the production of ultra-strong fibers (F 2 ). (A 1 ), (C 1 ), and (C 2 ) were adapted with permission from Plappert et al. 104  rheology methods with other techniques for revealing chiral nematic ordering; and combining rheology that probes macroproperties with for example X-ray analytics 203 is not far away from reaching visualization resolution of an individual CNCs. The combination of macroscopic, microscopic, and atomic methods in line would enable in situ analytics where spatial observations can be related to sample interactions during processing.
High-resolution solution-state 1D and 2D NMR techniques have molecular resolution and can be used to conrm spatioselective modication, as already mentioned above. The accessibility of solution-state NMR to nanocelluloses is not selfevident since crystalline celluloses do not dissolve in typical perdeuterated solvents. Tackling this bottleneck, King and coworkers have developed NMR methods that use the ionic liquid electrolyte [P 4444 ][OAc] : DMSO-d 6 to dissolve nanocelluloses. 204 This ground-breaking development has, for instance, enabled in-depth qualitative and semi-quantitative analysis of REG-modied CNCs (Fig. 8), 174,181 oxidized 122 or regioselectively esteried 205 nanocelluloses. In addition, this method can also take advantage of existing structural assignments to yield detailed regioselectivity information, 122,123,[206][207][208][209][210] to give insights on regioselectivity on a variety of chemical modications.
As another method of delivering molecular information, Xray photoelectron spectroscopy (XPS) and, specically, the high-resolution carbon (C 1s) region can be used to differentiate, e.g., O-C]O, O-C-O, C-O, and C-C linkages of oxidized celluloses. 211 However, the limited analysis depth of XPS to < 10 nm restricts this method to the surfaces, 212 strictly dening requirements for the sample preparation to obtain reliable information. In addition, it is so far not possible to distinguish a substitution at the different OH-groups of cellulose to prove for example, a regioselective modication. Moreover, experimental factors, including contamination before and outgassing or degradation of the sample during the measurement may bias the XPS data. 211 For reliable quantitative analysis of spatioselectively modi-ed nanocellulose, i.e., for determining the surface DS, the range of different, especially, wet-chemical methods is broad, including the classical charge titration for determining surface carboxylates, infrared spectroscopy, 43 or derivatization of surface groups and their later quantication by uorescence or UV spectroscopy. 174,213,214 In the case of REGs, there are multiple ways to determine the number of REGs. One possibility is the utilization of copper complexes (e.g., the Cu I bicinchoninate complex) coupled with spectrophotometric quantication. 175,181

Implications of spatioselective nanocelluloses on materials science
The impact of spatioselective modications of cellulose in biomass is multifold. At the building block scale, it enables the extraction of the elementary brils of cellulose, the control over colloidal stability, and ion interactions. Moreover, interfacial interactions can be nely tuned by changing the spatial distribution of functional groups, affecting as well their affinity for self-assembly and adhesive interactions with non-cellulosic interfaces. Finally, a selective modication will highly inuence the mechanics of the obtained nanocellulosic material and their optical properties as associated with the long-range order of the obtained bers. We showcase below some relevant examples.

Controlling interparticle interactions
For manufacturing of CNFs, TEMPO-mediated, and periodate oxidation are currently pivotal for controlling the colloidal properties. The introduction of charged groups facilitates electrostatic repulsion that aids colloidal stability. Regioselectively modied CNFs have been shown to possess extraordinary longrange ordering capability, even at low concentrations below 1 wt% in water, and yield anisotropic alignments with liquid crystalline behavior (Fig. 9A 1 ). 104,218 The high ordering capability of these CNFs can be attributed to their regular patterning of functional groups, e.g., in the case of C2, C3-dicarboxyl-CNFs, or TEMPO-CNFs. This structure can be also translated into ordered and mechanically robust aerogels, and enable strain hardening and pore size harmonization (Fig. 9A 2 ) by compression-induced alignment to reach optimal heat insulation properties (Fig. 9A 3 ). Other properties may be optimized with such considerations, for instance, the gas permeability that is a key parameter for cellulosic to form sustainable replacements for single-use plastics used as food packaging. Noteworthy, completely individualized CNFs have been obtained so far only from cellulose modied with self-repulsive charges (e.g., carboxylates, sulfonates, etc.), and the preparation of completely individualized hydrophobic CNFs is still a challenging task, especially using spatioselective chemistry, which does not partially solubilize or degrade the cellulose sample. Concerning CNCs, self-assembly and alignment play even a more dominant role and gives access to lms or coatings with structural color, 219,220 or allows the expansion of the upper limits in mechanical properties of CNC materials. 221,222 So far, most related efforts have been conducted with CNCs obtained from sulphuric acid hydrolysis, which have a small number of sulfate groups at their surface.
Colloidal properties can be also controlled by spatioselective modication of the CNC end groups. Polymer graing is one concept that has been described for CNC and CNC-II and that offers two main design parameters for tuning the properties of the semi-synthetic nanohybrid: the physicochemical nature of the polymer chains and their molecular weight. The graing of hydrophobic polystyrene from the REGs of CNCs, for instance, renders the particles amphiphilic and has improved their ability to stabilize oil-in-water emulsions. 183 Similar concepts might be also interesting for the compatibilization of immiscible polymer blends 223 while simultaneously enhancing mechanical properties and the dispersibility of the CNCs in the polymer matrix. The graing of hydrophilic polymers, on the other hand, has been described to improve the redispersibility of previously dried CNCs in aqueous media, 181,182,224 while maintaining, for instance, their ability to form liquid crystalline phases (Fig. 9A 2 ). 182,224 This has been shown for asymmetrically or symmetrically graed CNCs with thermo-responsive polymers at their end-groups enabling the formation of chiral nematic tactoids. 182 This behavior was especially interesting for the smaller CNC-II, where the polymer graing introduced longrange orientational order and allowed the CNC-II to selfassemble 40-times faster than its unmodied counterpart. Jean and co-workers graed temperature-responsive Jeffamine polyether amines to the end-groups of CNC and CNC-II (Fig. 9B 1 ) to control their assembly into dened superstructures. 167,225 The star shapes (Fig. 9B 2 ), observed for end-graed CNC-I above the LCST of the graed polymer, are an impressive example of controllability and reproducibility of CNC selfassembly. 225 CNC-II, symmetrically graed with Jeffamine chains, formed micron-sized networks of end-to-end connected CNC-II gras (Fig. 9B 3 ). 167 These complex 2D and 3D nanostructures cannot be achieved through uniform surface modi-cation of CNCs.
Apart from CNC-II, there has been also a recent effort in the preparation and characterization of so cellulose II nanospheres. 226,227 They have unique swelling capability due to their less crystalline particle shell, 227 which can be also used to enable the fabrication of surfaces with high protein loading for improved detection sensitivity, 228 e.g., in COVID-19 antigen tests. 229 Details on their exact morphology and structure are still under debate, and although regioselective modication has not been adopted on this type of nanoparticles, it might support improving their properties and establishing a better understanding of cellulose chain assemblies in different types of nanocelluloses.

Controlling surface chemistry
Apart from controlling self-assembly, selective surface modication can be also further tuned to vary, e.g., the type of functional group or the degree of surface modication. Typically, the interactions occurring at cellulose interfaces will signicantly impact the mechanical properties but also the adsorption isotherms, binding constants, and adsorbed conformation of other macromolecules (hemicellulose, lignin, proteins, etc. in non-processed cellulosic bers) and molecules (e.g., surfactants, ions, etc.). [230][231][232] In turn, beyond controlled mechanics, a range of key properties will be associated with these parameters such as water interactions, 216,233 biological responses, 234,235 colloidal stability, 236,237 ion interactions (e.g., conductivity), 238,239 or enzymatic activity. 240,241 These interactions of nanocelluloses can be tuned by varying, e.g., the type of functional group and the degree of surface modication. C6-O-succinylated-CNF has been shown to possess similar properties as TEMPO-CNF but offers apart from its higher molar mass another clear advantage as the ester linkage in C6-O-succinylated-CNFs can be cleaved to recover the intrinsic cellulose chemical interfaces (Fig. 9C 1 ). 90 This was also demonstrated at the materials level, nanopapers prepared from C6-O-succinylated-CNFs were treated under alkaline conditions to remove/hydrolyze the ester groups (Fig. 9C 2 ). Thereby, hydrogen bonding and brillar interactions can be strengthened signicantly (Fig. 9C 3 ), increasing 2.5-fold the tensile strength and reaching an elastic modulus of 19.4 GPa (in comparison to 7.9 GPa before ester hydrolysis).
The extent of a C6-O-carboxylation of CNFs can be modulated by varying, e.g., the reaction time of the TEMPO-oxidation. 11 A higher degree of C6-carboxyls introduced goes hand in hand with a lower molar mass, 11,141 as well as stronger repulsive interactions and water interactions, which may limit the performance of CNF assemblies. 216 Based on this fact, there is generally an optimal degree of oxidation, which is below complete surface oxidation, as shown in the dependence of the open-circuit voltage of TEMPO-CNF aerogels on the charge density of CNFs (Fig. 9D 2 ). 216 This was also well demonstrated during the preparation of highly aligned, ultra-strong cellulose microbers (Fig. 9D 3 ). 217 In the former case, CNFs with the highest carboxylate content produced less energy, explained by over hydration and structural collapse of respective aerogels, 216 whereas in the latter case the lower mechanical stability of highly charged CNFs is explained by their lower degree of polymerization. 217 Taking this into account, alternative preparation of CNFs, e.g., through selective succinylation, are noteworthy to be investigated and could potentially increase the performance of assemblies of charged CNFs, due to their higher degree of polymerization. Surface modication of CNCs by TEMPO-oxidation has been also shown to induce synergistic interactions in composites, as shown in the case of polypyrrole/ CNC lms. 242 In this case, it was demonstrated that the carboxylate groups at the CNC surface, increased aerial capacitance, energy density, and cycling stability in prepared supercapacitors.
Dependent on the application of the material, the use of mild gas-phase reactions can be of interest for post-processing, e.g., hydrophobization, to avoid the occurrence of swelling and/ or capillary forces during the reaction or the drying step, which would occur in liquid state reactions. The C6-OH selective gasphase acetylation was used to modify and thereby hydrophobize labile and complex chiral-nematic aerogels. 44 Regular treatment in organic solvents would disrupt this delicate and structurally colored assembly. This treatment can be applied also to other labile materials or potentially be used to produce regioselectively modied CNC model lms for fundamental studies.

Conclusions
Considering the dened structure of CNCs and CNFs, and the possibility to regularly pattern functional groups at their surface, there are many hidden opportunities in the chemistry of nanocellulosic materials. In this review, we summarized basic knowledge in this eld to give the reader the toolbox and background information to prepare such spatioselectively modied nanocelluloses, as well as to demonstrate its longterm potential in materials science.
Most focus on spatioselective modication has been set on endwise modication of CNCs, which was proven to signicantly alter the colloidal properties, as well as the interactions of CNCs with themselves and other polymers/particles, in the nano-and materials scale.
Concerning modication of the surface hydroxyls, it has been clearly shown that a restriction to a surface modication preserving crystalline, non-accessible cellulose regions, is crucial to improve the properties of nanocellulosic materials, such as transparency, mechanical and thermal properties. In addition, although clear evidence is rare and typically unavailable, there are indications that a regioselective patterning increases the tendency of nanoparticles to align and improve the mechanical properties of their assemblies. For example, in the case of TEMPO-oxidation, it is unclear if the inuence of carboxylate content on CNF properties is dominated by charge density, molar mass, hydration effects, or a combination of these. This is a good demonstration of how complex it is to give a clear statement on the role of selective patterning of nanocelluloses.
Moreover, there is yet a lack of analytical and non-invasive methods to analyze the regioselectivity of nanocellulose modications and to visualize a certain spatial distribution directly at the surface of those nanoparticles. Recent developments in the chemical analysis of nanocelluloses enabled direct determination of the regioselectivity of esteried CNFs and CNCs through highresolution NMR spectroscopy in an ionic liquid electrolyte. Based on this method, regioselective pathways using N-acylimidazoles or gaseous acetic anhydride were established, which are surfaceselective, and versatile (as different types of ester groups can be introduced). Initial research has already demonstrated their potential to surpass current limitations, e.g., in the mechanical properties of nanocelluloses. In this case, the motivation is clear, since the mechanical strength of nanocellulosic materials has a current maximum of approx. 1 GPa, 217,243 and is still far from the maximum theoretical value that ranges between 2 and 7 GPa. 244,245 There are still limitations in the available toolbox of spatioselective modications in terms of reaction efficiency, available chemistries, and slow reaction kinetics, e.g., in the case of endwise modication of CNCs. The sustainability of nanocellulose modications is currently given little attention, many reactions use a massive excess of reactants and solvents. In addition, the atom economy of reactions and the recyclability of reactants are oen not considered, rendering chemical modi-cations of renewable nanomaterials still problematic from an ecological and economical viewpoint.
Before functionalization, one should carefully consider the cellulosic precursor to be reacted (e.g., lignocellulosic bers, containing lignin and hemicelluloses, vs. cellulose ber). Most selective reactions are conducted using highly pure cellulose bers, which limits their feedstock to a low percentage of available biomass. Further development should also consider raw or more complex biomass, to enable selective modication of cellulose in presence of other natural building blocks. These selective modications have also the potential to ease the deconstruction of biomass into nanocellulose, which would be important to enable a more efficient and sustainable production on a bigger scale.
Apart from the huge potential of spatioselective chemistry to ne-tune interactions and surface properties of nanocelluloses, the development of new selective pathways of high atom efficiency economy will also tremendously increase the sustainability and efficiency of chemical reactions on nanocellulosic substrates, which is especially relevant in establishing functional nanocellulose as the future building block for highperformance materials.

Conflicts of interest
There are no conicts to declare.