C4-based CI value
The NMR spectra of the samples with different crystal sizes were recorded at various CTs (0.1–10 ms). The Scherrer crystal sizes of the acid hydrolysates of tunicin and cotton, DP, TO-DP, and TO-CNF were 10.0, 6.5, 4.0, 3.6, and 2.0 nm, respectively. For all the samples, the intensity of the C4 region changed depending on the CT (Fig. 1a). Furthermore, the spectra became sharper and split into several peaks as the crystal size increased, which is clearly visible in the spectra of the tunicate cellulose. These results can be interpreted as the crystallographic inequivalence of the glycosidic unit, and the resolution is improved when the crystal size is larger [8]. In the case of the crystalline atoms in small crystallites or the noncrystalline atoms, the structural heterogeneity might be too large to produce clearly resolved peaks in the spectrum.
The dependency of the CT to the signal intensities of both the C4 crystalline and noncrystalline regions was then analyzed. Figure 1b, c shows the CP curves for the C4 crystalline and noncrystalline regions, respectively. The signal intensities were normalized at the maximum intensity for each carbon. In both regions, all the signal intensities were rapidly increased at the initial stage before reaching a maximum. The maximum of these intensities was at approximately 1–2 ms CT, irrespective of the crystal size. This region was in accordance with the reported CT as a preferable time for evaluation of the CI value of cellulosic materials [16, 21, 22].
At the initial increase in the C4 crystalline and noncrystalline regions, there was no significant difference in the trend and slope (Fig. 1b, c). The initial increase mainly reflects the efficiency of magnetization transfer from the 1H spin reservoir to the 13C spin, described by the time constant of TCH. In general, a rigid structure has a smaller TCH than a nonrigid system, although the chemical structure may be the same. Therefore, this result indicates that the rigidity of the C4 carbons is similar in the time scale represented by the TCH relaxation, regardless of the crystal size.
In contrast, the decay in the C4 regions showed different behavior depending on the crystal size. The decay of each sample was steeper in the noncrystalline region than that in the crystalline region. In the crystalline region, the decay slope became steeper with a decrease in the crystal size (Fig. 1b). The decay at the longer CT reflects the relaxation of the 1H and 13C spins in the rotating frame [T1ρ (H) and T1ρ (C), respectively], which are related to molecular mobility. Thus, the decay trends shown in Fig. 1b indicate that the crystalline region of large crystallites, such as the tunicate cellulose, have lower molecular mobility. However, the smaller crystallites exhibited the higher mobility, even in the crystalline region, because of the large contribution of the crystallite surface. Furthermore, the similar trend was visible in the decay of the noncrystalline region; the decay was faster with a decrease in the crystal size (Fig. 1c). Indeed, the conformational heterogeneity is present in both the crystalline and noncrystalline regions. It has been reported that part of the molecules that exist close to the surface sit in a noncrystalline conformation even in the interior of the microfibril [23, 24]. Our results also suggest this.
Another possibility of the existence of heterogeneity is also conceivable. It was previously found by some of us that the crystalline C4 signal decreased when the bundled fibers dispersed as microfibrils and, correspondingly, the noncrystalline signal increased [18]. Indeed, the cellulose molecules at the interface between bundled fibrils or at the surface of the microfibrils are partially crystallized. This interface has been regarded as an inaccessible surface, where solvent molecules cannot penetrate. It has been proposed that the NMR signal due to the molecules at the inaccessible surface appear in the noncrystalline region [25]. Taking these phenomena into account, some of the inaccessible molecules at the bundled interfaces are probably occupied in the crystalline region of the NMR spectra rather than in the noncrystalline region, thus causing the structural heterogeneity.
The CP curve of the amorphous cellulose sample, whose molecular sheet stacking and the atomic conformation are disordered and noncrystalline, respectively [26], was also compared with other cellulosic samples (Fig. 1c). The initial increase and the maximum were similar to that observed for the C4 noncrystalline carbons in other samples. The decay slope was slightly steeper than that of TO-CNF, but this difference is somewhat minimal. These results suggest that the glycoside linkage may be restricted even in the randomly packed cellulose molecules.
The CI values for all the samples were then plotted against the CT (Fig. 1d). In the range of approximately 1–2 ms CT, where the signal intensities were at a maximum, the CI values of each sample were almost the same, and the difference in the CI values was within only 3%. At over 2 ms CT, the CI values gradually increased. This is because of the decrease in the intensity of the noncrystalline region resulting in overestimation of the CI value (see Fig. 1c). For a quantitative demonstration, the CI value of the mixture of the tunicate and amorphous cellulose was calculated from the NMR spectra recorded at 2 ms CT (Fig. 2). The calculated CI value was 51%, which is in good agreement with the ideal value (49%) determined gravimetrically. These results indicate that, although the CP/MAS NMR spectroscopy is generally not quantitative for polymers having crystalline and amorphous regions, it is possible to evaluate the CI value of cellulose with an accuracy < 3% from the NMR spectra at the optimized CT of 2 ms.
Carboxylate groups at the C6
In the process of nanocellulose production, cellulose is often surface modified in advance before wet disintegration. One example is by TEMPO-oxidation. By this method, the primary hydroxy groups exposed on the surface of the cellulose crystallite were converted to carboxyl groups, maintaining the crystallinity [27]. Figure 3a shows the NMR spectra of the C6 carbonyl region of TO-DP and TO-CNF at various CTs (0.1–10 ms). Both the shape and intensity of these NMR spectra changed depending on the CT, similarly to the C4 region.
The signal intensities of the carbonyl carbon were then plotted against the CT, see Fig. 3b. Interestingly, TO-DP having bundled microfibrils showed similar CP dynamics with TO-CNF. The signal intensities of the carbonyl carbon increased more slowly than the signals of the C4 carbon, and then reached a maximum at approximately 3 ms. In other words, the TCH of the carbonyl carbon was larger than that of the C4 carbons (see Fig. 1 for the dependency of the CT to the signal intensity of the C4 carbons). This is likely because of the longer distance between the proton and carbonyl carbon. Furthermore, the signal intensities of the TO-DP were the same as that of the TO-CNF. This result indicates that the mobility of the carbonyl carbon on the surface of cellulose crystallite is independent of the crystal size.
Figure 3c shows the change in the NMR spectra at 2 ms CT before and after TEMPO-oxidation. The signals corresponding to C = O, trans-gauche (tg), gauche-trans (gt), and gauche-gauche (gg) are centered at approximately 174.8, 65.2, 62.6, and 60.6 ppm, respectively [28]. By the oxidation, the signal corresponding to the noncrystalline gg or gt was remarkably decreased, whereas the signal of the crystalline tg ratio remained constant, and the signal of the carbonyl carbon appeared at approximately 175 ppm. As shown in Fig. 3d, there was a 19% decrease in the signal ratio of the noncrystalline hydroxy group, and this value is in good agreement with the increase in the signal of the carbonyl carbon (22%). This result is consistent with data reported by Montanari et al. [29], suggesting the conversion to carboxyl groups from the hydroxy groups can be evaluated from the NMR spectra under optimized conditions. Furthermore, the relationship between the CI value and the degree of oxidation of the samples, see Fig. 3e, shows that the degree of oxidation can be evaluated from the NMR spectra at the optimized CT, irrespective of the crystallinity of cellulose. Note here that the total C6 signal intensities were slightly smaller than the C1 signal intensity, and the degree of oxidation evaluated from the spectrum was lower by approximately 16% than that estimated from conductivity titrations (Fig. 3e). This is probably because the chemical structure and molecular mobility are different between the C6 noncrystalline carbons and C1 carbons [15, 30], and both the total C6 signal and the amount of the carboxy groups were underestimated in the present pulse condition.
Relationship between C4-based CI value and C6-based tg ratio
Based on a computational study, it has been reported that the conformation of the exocyclic groups at C6 significantly influences the C4 peak separation, and the signal at approximately 89 ppm, corresponding to the crystalline C4 carbon, is dominated by the tg conformation [31]. Figure 4 shows the relationship between the C4-based CI value and the C6-based tg ratio of samples with different crystal sizes. Interestingly, the CI value was found to be linearly correlated to the tg ratio, with a slope of 0.9 (R2 = 0.98). This result demonstrates that the change in the C4 crystalline signal is strongly affected by the tg conformation. The slight difference from 1.0 in the slope leads to the interpretation that the glycosidic linkage is partly restricted even when the C6 hydroxy groups sit in the noncrystalline gt conformation [31].