Quantum-Eciency Enhancement and Mechanical Responsiveness of Solid-State Photoluminescence Materials Based on Uniaxial Cellulose Nanocrystal Arrays in Flexible Polymer via Assembly-Induced Emission

Assembling cellulose nanocrystals (CNCs) can induce solid-state photoluminescence based on Stokes scattering. Such photoluminescence is free of photo-quenching and should have great potential in optical materials, whereas poor exibility of assembled CNC arrays limits its applications. Here, a co-assembly of binary components including 1D nanoparticles and long-chain polymers had been explored to introduce the uniaxial CNC arrays into a transparent poly (vinyl alcohol) (PVA) membrane, which enhanced the mechanical properties, especially the stretchable property. Besides, the CNC assembly was controlled by adjusting the volume ratio between CNC and PVA. The result indicated that co-assembly with PVA could improve the uniaxial orientation of assembled CNC arrays, which played a crucial role in enhancing the emission quantum-eciency (EQE) of CNC. Stretching the PVA/CNC membrane could furthermore induce an enhancement in EQE together with a gradual shift in emission wavelength. The mechanism study on that stimulation-response suggested that the enhancement and shift came from the change in the uniaxial orientation degree and periodicity of the CNC assembly, respectively. Since the stimulation-responsive enhancement in EQE (from ca. 40% to ca. 60%) can even be observed by naked eyes, we believe such cellulose-based materials can be widely used in optical sensors.


Introduction
Periodic assembly structures with structural colors have great potential in optical sensors (Wu et (Giese et al. 2014), which made CNC assembly potential in sensors. For example, CNCs could be co-assembled with poly (ethylene glycol) (PEG), whose structural color could be regulated by changing the fraction between CNC and PEG, and the obtained lms owned responsiveness on humidity (Zhang et al. 2013). Besides, CNCs could assemble in phenolformaldehyde resin, and then elastic membranes with structural colors that can make a response to strain and humidity were obtained via eliminating CNCs (Khan, Giese, Yu, Kelly, Hamad and MacLachlan 2013).
Although those studies showed that exible CNC-based membranes could be used in optical sensors, their CNC assembly structures were chiral instead of uniaxial. The reason was related to the strong interaction between CNC particles, the main of which was hydrogen bonds (H-bonds). Thus, we coassembled CNC in poly (vinyl alcohol) (PVA), which could also form H-bonds and was exible, and the assembly ratio between CNC and PVA was controlled to obtain a uniaxial structure of CNCs. With a morphological test from atomic force microscopy, we found that the uniaxial assembly degree of the CNC in PVA was even better than neat CNC, leading to an increase in emission quantum e ciency from 13.90-61.10%. Then, the obtained PVA/CNC membrane showed wavelength shifts in emission and excitation under strain, along with a change in photoluminescent intensity, which could be directly observed by naked eyes. The mechanism study in such a stimulation-response phenomenon suggested that the response mainly came from the change in the effective refractive index of the membrane, and the assembly ratio between CNC and PVA played a key role in increasing response sensitivity. Since the PVA/CNC co-assembly structure owned high quantum-e ciency and exibility, it could not only be applied in the optical sensor, but also in the eld of information security, like anti-counterfeiting labels and invisible patterns.

Extraction of cellulose nanocrystals (CNCs)
The CNC was extracted from cotton linters by the acid hydrolysis with H 2 SO 4 according to the previous work ). For the rst step, the linters were treated with a 2 wt% NaOH solution (generally 50 g bers for 2 L solution) for 12 h at room temperature. Then the alkali-treated ber was ltered and washed with distilled water for about 3 times to get a neutral supernatant. After that, the acid hydrolysis (12.5 g ber for 250 mL solution) was conducted at 45 °C using 65 wt% H 2 SO 4 with mechanical stirring for 60 min. Subsequently, the suspension was quenched by diluting with about 600 mL cold deionized water, washed several times, and centrifuged at a rotation rate of 4000 rpm. Subsequently, the suspension was dialyzed until neutrality. Finally, a high concentration (3 wt%) of the CNC suspension was obtained by rotary evaporation.
Preparation of the CNC/PVA uniaxial co-assembled membranes Firstly, 20 g of PVA was dissolved into 180 g deionized water under mechanical stirring at 95 °C for about 2 h to prepare an aqueous solution of 10 wt% PVA. Subsequently, the given amount of CNC suspension was added into the PVA solution with the magnetic stirring to gain a homogeneous suspension. After short sonication, 10 mL mixture suspension was poured into a 30 mL sample bottle, and a sliding glass was vertically inserted into the suspension, then the CNC/PVA membranes were obtained by evaporating the water at 30 °C. Based on the mass ratio of CNC and PVA in the resultant membrane, the samples were coded as CNC-X, where the number X represented the mass percentage of CNC in the total of CNC and PVA.

Characterizations
Ultraviolet and visual adsorption (UV-vis), zeta potential, photoluminescence (PL) spectra, and PL emission quantum yield analysis UV-vis tests were carried out on a Cary Lambda 750 S spectrophotometer (PerkinElmer). Zeta potential of CNC/PVA suspension (pH = 7) was measured by a NanoBrook Omni (Brookhaven). PL spectra of CNC/PVA membranes were tested on the 5JI-004 (Hitachi) with the solid-state mode. The measurement of PL emission quantum-e ciency (EQE) was conducted with F-4500 uorescence spectrophotometer (Hitachi). The measuring conduction was the solid-state mode.
Scanning electron microscopy (SEM) and atomic force microscope (AFM) The fracture plane morphologies of the CNC/PVA assembled membranes were characterized by SEM using a JSM-IT300 eld emission scanning electron microscope (JEOL),and the acceleration voltage was 10.0 kV. The morphologies of the membranes were carried on a Dimension Icon (Bruker). The characterization was tested by the tip of Scanasyst-air under the Scanasyst mode and by the tip of OTESPA-R3 under the Tapping mode.

Tensile tests
The tensile test was performed using a CMT6503 universal testing machine (SANS) with strain rate of tensile tests was 5 mm/min. For CNC-50 and CNC-40 membranes, they were strengthened to an elongation of 20%, 40%, 60%, 80% and till break to make the PL tests.
Thermogravimetric analysis (TGA) The TGA test was performed using TGA 4000 (PerkinElmer). The experiments were tested under dry nitrogen purge at a ow rate of 10 mL/min from room temperature to 800 ℃ at the heating rate of 40 ℃/min.

X-ray diffraction (XRD)
XRD was used to study the structure property of CNC/PVA co-assembly membranes before and after stretching, and it was carried out on X-Ray diffractometer (XRD-6000; Shimadzu, Japan). Cu Kα radiation was operated at λ = 0.154 nm from 5° to 50° with a step speed of 2°/min.

Results And Discussions
Luminescence properties and mechanical response of CNC/PVA co-assembly membranes Stable dispersion of cellulose nanocrystals (CNCs) in suspension is necessary for the uniaxial assembly of CNC. Table 1 shows the zeta-potential of CNC in the suspension with poly (vinyl alcohol) (PVA) with different ratios, which indicated that neat CNC suspension owned an extremely high zeta potential of -72.2 mV. As the PVA percentage increased, the zeta potential decreased but remained at a high level of > 30 mV, which indicated the suspensions were still highly stable (Kiprono et al. 2018). In order to eliminate the chiral structure, the zeta potential of the CNC dispersion should not be too high to achieve the uniaxial assembly with higher orientation. Thus, the CNC/PVA system should have better assembly potential than neat CNC.  Figure 1a shows the absorption spectra of CNC/PVA co-assembled membranes. Since both cellulose and PVA had no absorption in UV and visible range, the absorption peaks around 368 nm must come from the structural color of CNC assembly. The photoluminescence (PL) spectra Fig. 1b show excitation peaks at similar wavelengths and the emission wavelengths (Em) were around 368 nm. Those results suggested that CNCs could uniaxially assemble with PVA and emit blue light under UV light. Figure 2a and 2b proved the uniaxial assemble structure still existed in the CNC/PVA composites, but PVA took up a large volume, and it wrapped CNC around it. As seen in Fig. 2c and 2d, for the fracture surfaces of CNC-50 and CNC-40 displayed similar results, there were not apparent chiral structures, and they were both smooth. However, in Fig. 2e, CNC-100 showed a different format under the larger observation size than CNC-40 and CNC-50. It could be the large scale of chiral nematic structure though it had a high orientation in small size less than 1 µm. This was also the reason for the EQE of the CNC/PVA uniaxial co-assembly composites were much higher than that of the neat CNC membrane.
The percentage of CNC also affects the mechanical and tensile-response properties of the CNC/PVA membranes. CNC-80 and CNC-60 showed a low elongation at break, which was 4.2% and 7.0%, respectively, when that of neat CNC membrane was less than 0.1%. Those membranes displayed no tensile responsiveness. By contrast, the elongation at the break of CNC-50, CNC-40, and CNC-20 could be more than 90%. Furthermore, CNC-50 and CNC-40 exhibited shifts in excitation wavelengths (Ex) during stretching, while CNC-20 showed little responsiveness. In detail, the Ex of CNC-50 showed an apparent blue shift when the elongation reached 40%, and it continued to decrease to 346 nm at the break, compared with the initial Ex wavelength of 368 nm, as shown in Fig. 3. The Ex of CNC-40 sample displayed similar results, and the maximal decrement of Ex wavelengths was 22 nm and 17 nm for CNC-50 and CNC-40, respectively. Emission wavelengths (Em) of CNC-50 and CNC-40 also showed a slight blue shift overall when the elongation from 0 to 90% or 94%.
Besides, the luminesce intensity also increased during the stretching. As shown in Table 2, the emission quantum-e ciencies (EQEs) of CNC-50 and CNC-40 were 48.30% and 60.60%, respectively, both of which were much higher than neat CNC uniaxial assembled membranes (13.90%) (Gan, Feng, Liu, Zheng, Li and Huang 2019). During stretching, the EQEs of CNC-50 and CNC-40 could increase to 61.10% and 63.52%, respectively, and their photoluminescent lifetime also increased slightly. Figure 4 also illustrates that the luminous intensity of CNC/PVA membranes was higher after tensile treatment.

Responsiveness mechanism
The mechanisms of tensile response and photoluminescent enhancement were studied with the uniaxial assembly structure of CNC as the beginnings shown in Scheme 1.  In this work, the incident angle was 90°, so the value of sin 2 θ was 1. The refractive index of CNC and PVA were 1.534 (Shopsowitz et al. 2010) and 1.520-1.550 (Natarajan et al. 2017), respectively. Since the λ was 368 nm, the D should be 158.2 nm, which was slightly larger than the average length of CNC. The difference should be ascribed to the introduction of PVA between CNC.
After stretching, the Ex of both CNC-40 and CNC-50 rstly decreased a little at the elongation of 20%, and then showed a substantial decrement as the elongation increased from 20-40%, then steeply decrease until the break. The decrease of Ex wavelength might be from the change in D and n eff . However, we found the change was mainly attributed to the n eff . The n eff decreased with an increasing strain due to the vacuum, which owned a lower refractive index. In byte DHA of CNC in PVA, we regarded CNC as a cylinder with a radius of r (4.5 nm), a length of L (144 nm), and a minimum vertical distance between CNC particles of e. The edge of hexagonal was coded as R. According to the thermogravimetry analysis (TGA) results in Table 3 3 The R of CNC-50 and CNC-40 were 5.36 nm and 6.70 nm, respectively, which indicated CNCs might be tightly arranged in the PVA matrix. We rst assumed the D changed, which meant it must decrease during stretching to keep consistency with the decrease in λ. However, the decreasing D implied the material was shrinking, which was clearly contradicted to the tensile test conditions.
Then, we assumed the D was constant and the n eff decreased during stretching due to the vacuum generation, so we could calculate the n eff of stretched membranes with Eq. where n 1 and n 2 are the effective refractive index of PVA/CNC and vacuum, the value of n 2 is 1, and f 2 is the fraction of vacuum and air. Apparently, f 2 = 1 -f 1 .
The results of f 1 were shown in Table 4. The Δf 1 displayed the same tendency of the change by stretching the CNC-50 and CNC-40 membranes. Those results suggested that the membrane did not gain much vacuum or air at rst, so the λ nearly did not change when the elongation was smaller than 20%. By contrast, when the elongation increased further, the plastic strain of PVA led to the introduction of vacuum and air. Then the f 2 increased rapidly, and thus the Ex decreased obviously. Subsequently, with stationary stretching, the vacuum generation of CNC-50 and CNC-40 showed consistent results of Δf 1 with the change of elongation. Finally, the materials fractured and resulted in a massive change in n eff , which led to a large decrease in Ex wavelength again. The results were also coincident with SEM and XRD analysis. Compared with the unscratched membrane in Fig. 2c, the stretched membrane in Fig. 6a seemed to own better orientation, and the traces of vacuum during stretch could be found. Meanwhile, the CNC-40 fracture surfaces displayed similar results, but the orientation was not noticeable. The XRD results in Fig. 7a and 7b displayed a typical diffraction peaks at 2θ values of around 22.0° in all samples, which is coincident of diffraction planes of (200) in the cellulose I crystal (French 2013 EQE increase was greater than that of CNC-50. In addition, although CNC-20 was also tough, the CNC content was too low (15.0 wt%). Thus, the formation of the vacuum could be hard, and it showed little response to mechanical stimulation. These results also con rm our assumption that the decrease of the Ex wavelength and the luminescence enhancement of CNC/PVA membrane was due to the n eff change should be correct.

Conclusions
Cellulose nanocrystals (CNCs) were able to uniaxial assemble to form photoluminescent membranes, but the brittleness of the material limited the application. Herein, we combined the advantages of exible poly (vinyl alcohol) (PVA) and assembled abilities of CNCs to uniaxial co-assemble to form exible photoluminescent membranes. In addition, we regulated the volume fraction of the two components to control the tensile property and luminesce intensity of the co-assembled membranes. The obtained composites had not only high photoluminescence quantum yield (EQE) but also gain mechanical response capacity. With the introduction of PVA, the elongation at break of the material could increase to more than 90%, and the excitation (Ex) wavelength decreased with the increase of the elongation. In addition, the EQE increased a lot compared with the neat CNC membrane, and it could further increase after stretching. We studied the mechanism of such responsiveness and constructed a close-packed hexagonal model of CNC assembly. Through the calculating, we found the decrease of Ex wavelengths was attributed to the formation of vacuum during the tensile treatment.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupportingInformation.docx