Synergetic effect of carbon dot at cellulose nanofiber for sustainable metal-free photocatalyst

Metal-free photocatalyst was synthesized by attaching carbon dots (CDs) to cellulose nanofibers (CNFs) via simple in-situ synthesis. The graphitic core and functional groups of CD on CNF was controlled along the precursor concentration, while the fibrous structure of CNF was intact. The optical property of the samples was profoundly analyzed focusing on the electron recombination pathway. It revealed that the excited electrons in the CD transferred to the CNF and delayed the radiative recombination. The electron–hole pairs were efficiently separated in the composite where abundant amide group and sulfide bonding existed. The CDs on the CNF surface improved the morphological stability of CNF under UV irradiation as well. As a result, the composite showed superior photocatalytic performance to degrade 98% of MB molecules within 25 min. The aerogel synthesized from CDCNF had good re-usability without sacrificing its photocatalytic effect. The synthesized CDCNF composites showed superior possibility for applying cellulose nanofiber to sustainable metal-free photocatalyst.

Abstract Metal-free photocatalyst was synthesized by attaching carbon dots (CDs) to cellulose nanofibers (CNFs) via simple in-situ synthesis. The graphitic core and functional groups of CD on CNF was controlled along the precursor concentration, while the fibrous structure of CNF was intact. The optical property of the samples was profoundly analyzed focusing on the electron recombination pathway. It revealed that the excited electrons in the CD transferred to the CNF and delayed the radiative recombination. The electronhole pairs were efficiently separated in the composite where abundant amide group and sulfide bonding existed. The CDs on the CNF surface improved the morphological stability of CNF under UV irradiation as well. As a result, the composite showed superior photocatalytic performance to degrade 98% of MB molecules within 25 min. The aerogel synthesized from CDCNF had good re-usability without sacrificing its photocatalytic effect. The synthesized CDCNF composites showed superior possibility for applying cellulose nanofiber to sustainable metal-free photocatalyst.

Introduction
Sunlight, one of the natural energy resources, is regarded as promising alternative resources as it offers inexhaustible energy. Among several advances using solar energy, photocatalytic reaction is a highly encouraging technology as it could effectively degrade the environmental pollutants in air and in water. Wide range of materials were studied for photocatalyst, such as metal oxide, metal organic framework or covalent organic framework (Wang et al. 2020a(Wang et al. , 2020bHe et al. 2019;Zeng et al. 2020;Cheng et al. 2020). Of all materials, TiO 2 is one of the widely used photocatalyst because of its chemical inertness, high activity, low cost, and durability to chemical corrosion (Srisasiwimon et al. 2018;Lisowski et al. 2017). However, the energy absorption efficiency of bare TiO 2 is very low due to its large band gap and it needed to be supported by matrix for better recovery. To reduce the band gap, wide range of materials including metal and semiconductor were doped with crystalline TiO 2 (Guayaquil-Sosa et al. 2017;Gomes et al. 2011;Cai, Wu, Li, & Zheng, 2014). However, there are still critical disadvantages: most of the doping reaction needs high temperature accompanied by sintering process. Moreover, the doped metal ions can promote the electron-hole recombination and suppress the photocatalytic reactivity. Carbon-based materials such as graphite or CNT have been employed as matrix for metal oxide as well as for light absorption-assistant materials (Sharma et al. 2021;Idris et al. 2021;Ibrahim et al. 2020;Isari et al. 2020;Stan et al. 2019;Jung et al. 2018). The carbon materials can absorb light of lower energy and play the role of an electron reservoir in the system, leading to the increased efficiency of photocatalytic reaction.
Recently, there have been numerous reports of the use of carbon materials as photocatalysts. Especially carbon nitride is one of the candidates to replace metal oxide materials (Gholipour et al. 2017;Wang et al. 2020a,b;Zhang et al. 2017). It is synthesized from nitrogen-containing materials such as melamine and comprises only C and N, bonded by p-conjugation. Nitrogen in the structure plays as reactive site of photocatalytic reaction. The photocatalytic performance of carbon nitride can be tuned according to the synthetic condition. For example, Oh et al. controlled the temperature for carbon nitride synthesis with direct heating method and sequential heating method (Oh et al. 2018). They revealed the sequential heating of urea could increase the grain size and reactive site for photocatalytic reaction. Another method to increase the reactive surface area is to reduce the size of carbon materials. Nano-size carbon nitride dot was also applied to photocatalytic materials. Yadav et al. synthesized carbon nitride dots (g-CNQDs) using melamine and ethylenediaminetetraacetic acid as precursors by carbonization at 350°C (Yadav et al. 2019). Without metal supporter, the synthesized g-CNQDs exhibited high performance toward reactive oxygen species (ROS) generation. However, it still needs high temperature over 300°C to precisely control the structure of carbon nitride.
Compared to carbon nitride dot, carbon dot can be prepared with mild condition. Generally, various organic materials can be utilized as precursors and hydrothermally carbonized under 200°C. Carbon dot has excellent optical and electrical features, acting as light absorber and electron reservoirs. Many researchers reported carbon dot as highly efficient photocatalytic materials (Duarah, & Karak, 2017;Martindale et al. 2015;Sarma, Majumdar, & Sarma, 2019;Li et al. 2010). However, it is hard to handle and to re-use because of its nano size. For the recovery, carbon dot was supported by matrix such as metal particles (Zeng et al. 2018;Ryu et al. 2014).
In this study, metal-free photocatalyst were synthesized via green hydrothermal treatment. Nitrogen and sulfur doped carbon dots (CDs) were synthesized and attached to cellulose nanofibers (CNFs). The CNF played as trap for excited electron and delayed the electron-hole recombination in the system. CNFs with CDs on the surface were synthesized into aerogel and showed excellent photocatalytic power. It degraded * 98% of pollutant in 25 min without sacrificing the shape.

Materials
Tempo-oxidized cellulose nanofiber (CNF) suspension (1.8 wt% in water, carboxylate content: 1.7 mmolÁg -1 ) was purchased from ANPOLY (Republic of Korea). Glutathione (GT) and citric acid (CA) for the precursors of CD, Methylene blue (MB, 82% dye content), N, N-diethylaniline (DEA), tertbutanol (t-BA) and para-benzoquinone (p-BZQ) were purchased from Sigma Aldrich. Co (USA). CNF and GT were stored at 2°C before use and all the other chemicals were used without any treatment.

Synthesis of carbon dots (CD)
The carbon dots (CDs) in this study were prepared via hydrothermal carbonization of GT and CA. Briefly, 0.1, 0.3 and 0.5 mmol of each precursor (GT and CA) was dissolved in 30 ml distilled water, together. For the synthesis, the mixture was hydrothermally carbonized at 120°C for 4 h in a Teflon-lined autoclave. After the hydrothermal reaction, the obtained CDs were purified by dialysis using a membrane (0.5-1 kDa, Biotech CD Tubing) against water for 24 h. According to the precursor amount, the synthesized CDs were named as 1CD, 3CD and 5CD, respectively. For example, 1CD represents the carbon dot synthesized of 0.1 mmol of GT and CA.
In-situ synthesis of CD at CNF The CD was synthesized on the CNF's surface according to the previously reported protocol (Ahn et al. 2021). As same to CD synthesis procedure, 0.1, 0.3 and 0.5 mmol of each precursor was carefully dissolved in 35.5 g of CNF suspension. The mixture was poured in a Teflon-lined autoclave and reacted at 120°C for 4 h for the in-situ synthesis of CD at CNF. The prepared suspensions were diluted to 150 ml and stirred for over 3 h, and then filtrated by a nylon filter paper (0.45 lm) for purification. The synthesized composites and the CNFs were prepared in suspension state as well as in aerogel state via freeze-drying. The in-situ synthesized CD at CNF composites were named as 1CDCNF, 3CDCNF and 5CDCNF, as following the CD precursor concentrations. The CD in the composite was differently named as CD@x-CDCNF. For example, CD@1-CDCNF refers to the CD in 1CDCNF composite.
Photocatalytic activity measurements MB, a typical cationic dye was used as a model material to evaluate photocatalytic activities of the CDCNFs. For the test, 0.5 wt% of the catalyst suspensions (CNFs and CDCNFs, 5 g/L) and 5 ppm MB were mixed. Mercury lamp (1000 W) was used as the light source and the distance between the light and the sample was set as 3 cm. Before the test, the mixtures were kept at dark for 20 min for MB absorption on catalysts. The MB concentration was measured by UV-vis absorption measurement.

Characterizations
The morphologies of the prepared samples were observed using high-resolution transmission electron microscopy (TEM) images (JEM-F200, JEOL, Japan) at 80 kV. For chemical structure analysis, Fouriertransform infrared spectroscopy (FT/IR-4000, JASCO, UK) and X-ray photoelectron spectroscopy (K-alpha, Thermo Scientific, UK) were carried out. The photoluminescence (PL) properties of the samples were studied using UV-vis absorption spectroscopy (Cary 60 UV/vis spectrophotometer, Agilent Technologies, USA) and PL spectroscopy (FS-2, SICNCO, Republic of Korea). Fluorescence lifetimes were obtained by time-correlated single photon counting method (TCSPC, Fluo Time 200 instrument, Picoquant, Germany). A 342 nm pulsed LED with a repetition rate of 5 MHz was applied for excitation. The decay profiles were analyzed by FluoFit Pro software using exponential fitting models through deconvolution with instrumental response functions (IRF).

Results and discussion
CDs were synthesized by changing the precursor concentration and their structures were studied ( Fig. 1). In the TEM images, CDs were individually separated and their diameters became larger as the precursor concentration increased (Fig. 1 a). The diameter increase was dependent on the synthesizing mechanism of CD, especially nuclear growth procedure. Hydrogen bonding between the citric acid promoted the construction of the core of CD (Dong et al. 2012). The core grew by consuming the molecules in the system. The molecules in the thin solution (1 mmol of precursor) had very rare chance to contact with the other molecules. When the solution got thicker, the molecules had better chance to be grow, resulting in bigger size.
Regardless of the concentration, all the precursors were successfully reacted to each other into CDs, as seen in FT-IR result ( Fig. 1b & Fig. S1). All the CDs had peaks such as O-H, N-H bonding (3400 cm -1 ), carboxyl (1714 cm -1 ), amine (1670 cm -1 ), amide (1108 cm -1 ) and sulfonate/sulfide groups (1000-1400 cm -1 ), as coincident to our previous work (Ahn et al. 2021). The functional groups of CDs had no significant difference. However, the peak for sulfonate group of 5CD was smaller than those of 1CD and 3CD (colored in orange). The decrease of sulfonate group in 5CD will be discussed further in XPS analysis ( Fig. 1c- Difference in the number of the molecules in unit volume led to difference in chemical structure of CDs. The N1s and S2p speciations of CDs were shown in the figure. Most of the nitrogen was formed in amino N in all the CDs. Graphitic N was barely developed with 1 mmol of precursors. Nitrogen in the thin solution was hard to be conjugated into core structure since the contact between the molecules was hard. As the precursor amount increased as 3 mmol, nitrogen was combined into the graphitic structure. Nitrogen could be easily participated in core-growth owing to the increased contact chance as well as the similar atomic size with carbon. The graphitic N speciation of 5CD was slightly decreased compared to 3CD because of the competitive structure formation between nitrogen and sulfur. The big atomic size makes sulfur hard to be penetrated the core structure. Instead of penetration, sulfur was developed as sulfide bonding at the edge and sulfonate group on the surface. As the reacting solution got thicker, the content of sulfide bonding was gradually increased while sulfonate group was decreased. With more chance of contact, more sulfur was combined with carbon at the edge. Based on the CD development mechanism, the structure of CDCNF composites was further studied. In Fig. 2 (a), XRD profiles were shown to study the effect of in-situ synthesis on the CNF physical structure. All the samples had cellulose I crystal structure, indicating the hydrothermal treatment condition was not so severe as to destroy the crystalline structure of CNFs. However, the crystallinities of the CDCNFs were changed. The crystallinities of the CNFs were obtained as 62%, calculated by peak height method (Park et al. 2010). After the hydrothermal treatment, the crystallinity was increased. The crystallinities of CDCNFs were 75, 86 and 86% with 1, 3 and 5 mmol of CD precursors. Amorphous region can be more rapidly degraded compared to crystalline region resulted in the crystallinity increase. Citric acid in the reactor helped to break down more cellulose chain. The crystallinity did not increase over 3 mmol of citric acid. The citric acid promoted the degradation up to a certain amount, but the excessive amount did not affect the degradation significantly.
The morphologies of the CNFs and the CDCNFs were observed using TEM and shown in Fig. 2 (b). The CNFs were highly entangled because of its high aspect ratio. The CDCNFs became less entangled after  the in-situ synthesis treatment. The degradation of amorphous region during the hydrothermal carbonization reduced the CNF length, resulted in lower entanglements. Still the process conditions were not severe to destroy the fibrous structure. As shown in the high magnification images, CDs were developed on the CNF surface (Fig. S2). The CDs on the CNFs were individualized with spherical shape for the all-precursor concentrations. Via the in-situ synthesis, CDs were successfully attached to the CNFs, regardless of the precursor concentration.
The chemical structure of CDCNFs were characterized using FT-IR and XPS (Fig. 3 & Fig. S2.) The CNFs had fingerprint region of cellulose as well as carboxylate group peak at 1606 cm -1 , which are typical in tempo-oxidized cellulose nanofiber (Fig. 3  a). In the process of tempo-oxidization, the glucose unit is partially changed into glucuronate group, containing carboxylate group on the C6 position. All the CDCNFs had the distinct fingerprint region of cellulose in FT-IR spectra as like to the CNF. The detailed peaks from CDs such as sulfonate groups were not clearly observed due to its small amount. However, the carboxylate group of the CNF was remarkably changed after the in-situ synthesis. The peak at 1606 cm -1 was decreased while the peak at 1719 cm -1 (colored in orange) was increased in all the samples. The increased peak was originated from the functional groups of CDs, such as carboxyl groups and amino groups. Glucuronate groups can be decomposed easily by mild hydrothermal carbonization because it is less stable than glucose group. During the in-situ synthesis, the decomposed glucuronate fragments were participated in CD-core growth and attached to the CNF surface. The peak at 1719 cm -1 in 1CDCNF was smaller than 3CDCNF and 5CDCNF. As discussed above, more cellulose (including glucuronate group) were degraded with higher precursor concentration. More CDs were attached to the CNF surfaces in 3CDCNF and 5CDCNF. The shoulder peak at 1604 cm -1 (colored in green) was increased as well. The peak was driven from C = O bonding of amide bond, originated from the linkage between the CD and the CNF (Ahn et al. 2021). It was clear that CD was attached to the CNF by sacrificing the carboxylate group and linked by amide bonding.
XPS analysis was conducted to study the chemical structure of CD@CDCNFs ( Fig. 3 b & Fig. S2). Interestingly, the N1s and S2p speciations of CD@CDCNFs showed similar results to CDs except for CD@1-CDCNF. In the in-situ synthesis, the presence of CNFs certainly affected the CD structure development with 1 mmol of precursor. Graphitic N structure was more developed and sulfur was barely functionalized in the CD@1-CDCNF. Bonding formation between sulfur and the carbon core was disturbed by the CNFs. Without competing to sulfur, nitrogen penetrated the core structure easily and developed graphitic N structure. With equal or more than 3 mmol of precursors, the N1s and S2p speciations of CD@CDCNFs followed the tendency of CDs: the graphitic N contents was highest in CD@3- Fig. 3 Chemical structure analysis of CDCNFs a FT-IR spectra, b N1s and S2p spectra from XPS c nitrogen and sulfur speciations of CDCNFs versus the precursor concentration CDCNF and the formation of sulfide bond predominated over sulfonate groups in CD@5-CDCNF, as coincident to CD. In turn, the number of molecules per volume majorly influenced the chemical structure of CDs and the presence of CNFs minorly affected the CD formation. The chemical structures of the CD@CDCNFs were considered similar to that of CDs, except for CD@1-CDCNF. Figure 4 shows the optical properties of each sample, the UV absorption curves and the PL emission spectra. In Fig. 4 (a), all the UV absorption curves did not exhibit the absorption peak related to p-p* transition because of the low graphitization of CDs (Ahn et al. 2019). In 1CD curve, small shoulder absorption around 350 nm was observed. As the precursor concentration increased, the absorption of CDs at 350 nm became more distinct. The absorption was from n-p* transition arouse by the functional groups such as carboxylate, amine and amino groups. As shown in Fig. S3 (a), absolute nitrogen content of Fig. 4 Optical properties of the samples: UV spectra of a CDs and b CDCNFs, PL spectra of c 1CD, d 3CD, e 5CD, f 1CDCNF, g 3CDCNF and h 5CDCNF 1CD was much lower than the others, resulted in weak absorption. Higher n-p* absorption of 3CD and 5CD was achieved from nitrogen both in the core and surface. More amino N in 5CD led the stronger absorption at 350 nm. Figure 4 (b) shows the UV absorption of CDCNFs. Unlike CDs, all the curves had a smooth decreasing absorption without any apparent peak. The diverse functional groups of the CD@CDCNFs split the energy level under the p * level, leading to broad absorption (Woo et al. 2020). As well the long-tailed absorption was observed originated from the geometrical feature of the CNF (Ahn et al. 2021). The fibrous structure of CNF scattered the light. CD@1-CDCNF couldn't strongly absorb light because of their small amount and the incident light was scattered by the CNFs. With the increase of precursor amount, stronger absorption curves were observed, as coincident to CDs. Incident light was absorbed by the CDs on the surface more instead of being scattered by CNFs.
Figures 4 (c-e) shows the emission spectra of the CDs. All the CDs had emission maximum at 425 nm and showed weak emission dependency on the excitation wavelength. This was driven from the sulfurcontaining surface state. The sulfur functionalized surface can create electron trap site with high density, weakening the excitation wavelength dependency on PL property (Dong et al. 2013). The emission wavelength of the CDs was red-shifted when the excitation wavelength shifted from 320 to 380 nm. As explained in the previous reports of our groups, the emission wavelength changes at different excitation wavelengths depends on the chemical structure of carbon dot. The well-graphitized carbon core of carbon dot has strong influence on the recombination of the excited electrons (Ahn et al. 2019). When carbon dot has highly developed graphitic core, the emission wavelength is firstly blue-shifted with the increasing excitation wavelength until it reaches the emission maximum, and then red-shifted as excitation wavelength increases. In this study, no blue-shift was observed with increasing excitation wavelength. The cores of CDs were not fully developed and had not significantly affected the emission behavior. In turn, the radiative recombination of electron at functional groups, especially the sulfur containing group, dominantly managed the PL property of CDs. The emission behaviors of the CDs were affected by anchoring to CNFs (Fig. 4 f-h). The emission spectra of the CDCNFs became slightly broader and showed stronger emission wavelength dependency on the excitation wavelength. The anchor between the CD and the CNF led to energy level splitting. Still the emission wavelength was red-shifted along the excitation wavelength, showing that the functional groups managed the emission behavior. The changes in emission pathway will be further studied in lifetime analysis.
The effect of anchorage to CNFs on the electron pathways of CDs were investigated using lifetime decay. The obtained lifetime and the quantum yield (QY) of each sample were summarized in Table. 1. The average lifetimes were 7.47, 7.75 and 7.74 ns for 1CD, 3CD and 5CD, respectively (upper row). All the decay curves of the CDs were fitted with the short lifetime component s 1 , and the long lifetime component s 2 . s 1 for all the CDs had similar value to each other, as the core structure did not have significant impact on PL emission (Ahn et al. 2019). N-doped structure did not significantly affect the core state- Table 1 The quantum yield (QY), photoluminescence lifetime (s av , s 1 and s 2 ) of the samples emission pathway although the number of electrons in p system increased. s 2 were derived from the functional group emission (Ahn et al. 2019). Comparing to 1CD, 3CD and 5CD had significantly long s 2 . The increase of s 2 was derived from the increase of sulfide bonding contents. As many researchers described, oxidized sulfuric functional groups such as sulfonate group has strong electron withdrawn property (Xu et al. 2015). It effectively recombines the electron-hole via light emitting pathway. The abundant sulfonate group in 1CD rapidly recombined the electrons. On the contrary in 3CD and 5CD, unoxidized sulfide bonding held the electrons in the excited state of CDs and prolonged the lifetime s 2 .
The chemical structure managed the QYs of the CDs as well. 3CD had high QY of 42% compared to 1CD (35%) and 5CD (32%). The contents of sulfonate group and sulfide bonding served the major effect on QY changes. The sulfonate group contents gradually decreased as the precursor concentration increased while sulfide bonding was increased. The electrons in the excited state of 1CD could effectively recombined via light-emitting pathway owing to the large sulfonate group content. Despite of the decreased sulfonate group contents, 3CD exceptionally showed the highest QY value among CDs. Large number of electrons could be excited to p* owing to the developed graphitic N structure, resulted in increased QY (Ahn et al. 2019). 5CD had the lowest QY value as it had small sulfonate group content among the CDs. Instead of fast radiative recombination, the excited electrons in the p system moved to energy level endowed by functional groups via vibrational relaxation.
In spite of the similarity in chemical structure, the electron of in-situ synthesized CDs had different electron recombination pathways when attached to the CNFs. The average lifetimes of 1, 3 and 5CDCNF were 2.82, 2.88 and 3.50 ns, respectively. The decay curves were fitted with two components. CDs and CDCNFs did not have difference in s 1 , because the CNFs did not influence on the core state emission of the CD@CDCNF. It is noteworthy that s 2 of CDCNFs had huge decrease as CNFs were introduced. Regardless of the surface states (especially the sulfur speciation) differences, the s 2 values of CDCNFs were similar to each other. The electron recombination of the CD@CDCNFs was not controlled by the functional groups. As well, the QY of CDCNFs were dramatically decreased compared to CDs: The excited electrons were mostly recombined via non-radiative pathway. It is reasonable to assume that the decrease of s 2 was attributed to the electron transfer from the CDs to the CNFs. s 2 of CDCNFs were slightly increased with increasing precursor concentration originated from the different surface state. As studied in the XPS, more precursors produced more sulfide bonding in the CDs, i.e., the electrons in the p system remained for longer time in the excited state and had more chance to be transferred to the CNFs. CD@5-CDCNF had abundant sulfide groups resulted in more electrons transferred to the CNFs. In other words, the electron-hole pairs of CD@5-CDCNF could be efficiently separated.
The lifetime decay results indicates that the excited electrons from CD@CDCNFs were transferred to the CNFs, owing to its electron withdrawing property. To confirm this, the changes of PL intensities were studied after the strong electron donor N,N-diethyl aniline (DEA, 0.88 V vs. NHE) was added to the CDCNF suspensions (Fig. 5). Unlike common carbon dots, the emission intensities were barely changed as the concentration of DEA was increased (LeCroy et al. 2019; Wu et al. 2018). As an electron donor, DEA injected more electrons to the ground state of CD@CDCNFs (black dotted arrow) and excited more electrons to p* (Fig. 5 (d),). Via vibrational relaxation, the electrons were trapped to the CNFs (blue dotted arrow) and non-radiatively recombined, resulted in negligible change in PL emission. Coincident to the lifetime decay results, CNFs played an essential role in electron relaxation process under UV irradiation. The trap suppressed the electron radiative recombination, hence believed to enhance photocatalytic effect. Figure 6 shows the catalytic performance of the samples. For the test, Methylene blue (MB) was used as the model pollutant. The MB degradation rate along time were plotted in Fig. 6 (a), in the absence (blank) and presence of the catalysts. All the CNFs had catalytic effects against MB degradation. The CNFs could degrade the dye without help of CD due to the radical generation by sacrificing itself under UV irradiation, which will be further discussed in Fig. 7. With the CDs attached on the CNFs surface, MB was degraded more effectively by photocatalytic reaction of CD@CDCNF. Among the CDCNFs in this study, 5CDCNF showed the highest photocatalytic power. Most of the MB molecules (* 98%) were degraded within 25 min in 5CDCNF suspension. For the clear comparison, the photocatalytic kinetics were studied ( Fig. 6 (b)). The degradation performance of MB with each catalyst were plotted with the pseudo first-order degradation equation where C 0 and C are the initial concentration and the concentration of MB at time t, respectively. The apparent rate constants (k app ) of MB degradation were calculated as expressed in the figure. k 5CDCNF was almost 3 times higher than k CNF , exhibiting the effect of CDs on the CNF surfaces.
The photocatalytic mechanism of the CDCNFs can be explained based on the active centers of photocatalytic reaction. The photocatalytic mechanism was schematically expressed in Fig. 6 (c). Under the UV irradiation, the electrons in CD@CDCNFs were excited and electron-hole pairs were generated (Eq. (1)). The excited electrons were radiatively recombined with the holes or transferred to the energy trap induced by the CNFs (Eq. (2)). The trapped electrons were retarded to be recombined, which enhanced the photocatalytic efficiency. The separated electron and hole favored to form the active radicals: the electron generated superoxide radicals ( • O 2 -) with molecule oxygen (Eq. (3)). The hole reacted with the water absorbed on the CDCNF surface to form hydroxyl radicals ( • OH, Eq. (4)). The ROS interacted  5) and (6)).
Two types of active-species-trapping experiments were carried out to elucidate the role of the major active species in the photocatalytic process ( Fig. 6  (d)). Two radical scavengers (1 mM of tert-butanol (t-BA, as • OH scavenger) and para-benzoquinone (p-BZQ, as • O 2 scavenger)) were separately added to each suspension before the UV irradiation. Figure 6 (d) exhibits the results of photocatalytic effect in the absence (control) and presence of the radical scavengers. In CNF and 1CDCNF suspensions, the effects of the scavengers were less than * 5% indicating the role of • OH and • O 2 were not significant in the MB degradation. In 3CDCNF and 5CDCNF suspensions, the addition of t-BA had minor effect on the photocatalytic power. On the other hand, the addition of p-BZQ in 3CDCNF and 5CDCNF suspensions suppressed the degradation efficiencies to * 94% and * 76%, respectively, indicating that • O 2 radical made the major contributions to the photocatalytic reaction. As discussed in the lifetime analysis, 5CDCNF had the longest lifetime among the samples as more sulfide bonding existed. This indicated that more electrons remained in the excited state and could transferred to CNFs. The more electrons transferred to CNFs, the more electron-hole pairs were separated, resulting in high production of •O 2 radical under UV light. Thereafter, 5CDCNF effectively and rapidly degraded the MB molecules (Fig. 6 (c), Eq. (5)).
For easy handling, the composites were formed into aerogel via freeze-drying process. The photocatalytic reaction of the aerogels was tested (Fig. 7 (a)). The images of MB/aerogel solutions along the UV irradiation time were displayed in Fig. S4. With all the aerogels, MB degradation occurred faster than the suspensions in early 15 min. The movement of MB molecules in the suspensions were restricted as the entangled fibrous structure of CNFs caused viscosity. On the other hand, MB in the solution could be dispersed without restriction and continuously absorbed to the surface of the aerogels. As well known, the negative charged carboxylate groups of CNFs can highly interact with positive charged MB. Even after the CD attachment on the surfaces of CNFs, the MB absorption-ability was maintained, leading to the faster degradation (Ahn et al. 2021).
Though all the aerogels showed similar morphology regardless of the precursor concentration, the degradation results of the CNF and 1CDCNF aerogels after 10 min were hard to be obtained while 3CDCNF and 5CDCNF aerogels degraded most of MB within 25 min as coincident to the suspensions (Fig. S7). As shown in the images (Fig. 7 b, S5 & S6), the morphologies of CNF and 1CDCNF aerogels were severely damaged after 10 min of reaction. As mentioned, CNFs could be decomposed by radicals under UV irradiation (Ma et al. 2021). During the decomposition, ROS such as hydrogen peroxide was generated as side reactant, leading to the sequential decomposition of CNFs. The continuous ROS generation helped to degrade MB as well in the first 10 min. However, the further decomposition of the CNFs made aerogel lost its shape. The self-decomposition of CNFs was greatly suppressed in 3CDCNF and 5CDCNF aerogels, owing to the CDs on the surfaces Fig. 7 a Photocatalytic efficiency of the aerogels b digital images of the aerogels after photocatalytic reaction and c recycle test results of (red) 3CDCNF and (blue) 5CDCNF (Fig. S5). The UV light was absorbed by CD@CDCNF and the decomposition of the CNFs was inhibited. As shown in the images, 3CDCNF and 5CDCNF aerogels retained their shapes even after the reaction for 25 min. With simple wash with ethanol, the 3CDCNF and 5CDCNF aerogels were reusable while maintaining their photocatalytic performances (Fig. 7 c). 3CDCNF was not able to test over the 2nd cycle because the aerogel lost its shape, as like to the pristine CNF and the 1CDCNF aerogels. On the contrary, 5CDCNF aerogel maintained its shape and was reusable over 5th cycles owing to the high UV adsorption capability of CD@5CDCNF. The photocatalytic performance was slightly decreased as the cycles repeated, due to the accumulated MB adsorption on the surface. Still, the degradation efficiency was over 91% on the 5th cycle. The photocatalytic performance of 5CDCNF was compared with the reported results of carbon dot-related catalyst (Table S1). Without any use of heavy metals, 5CDCNF showed excellent performance, compared to the others.

Conclusion
In the study, CDCNF composite was synthesized for green photocatalyst. The comparison of the chemical structures of CDs synthesized with or without CNFs confirmed the CNFs did not have significant influence on the CD's structural formation. The increase of the precursor concentration increased the sulfide binding and amide group contents on CDs. The structural differences determined the photocatalytic performance. The sulfide bonding and amide groups extended the lifetime of excited electrons and provided the transfer pathway to the CNF's potential, which resulted in the suppression of fast electron-hole recombination. The CDCNF composite synthesized with the highest precursor concentration showed the highest photocatalytic efficiency. 98% of pollutant was degraded within 25 min by the CDCNF composite. When it was fabricated as aerogel, it was reusable without sacrificing the photocatalytic performance. The sustainable metal-free CDCNF composite showed the potential to be used as a high-performance photocatalytic material for sustainable development.