Carbon dots prepared by different bottom-up methods: a study on optical properties and the application as nanoprobes for metal ions detection

Abstract Carbon dots (CDs) have emerged in the last few years as new carbon nanostructures due to their excellent optical and physicochemical properties and thus promising applications in different areas. In this work, we describe the preparation of different CDs obtained from distinct bottom-up methods and also using different low molar mass precursors in order to study advantages and disadvantages of each experimental method used. Different techniques were performed to characterize the nanoparticles, such as Transmission electron microscopy, Potentiometric titration, Fourier transform infrared, Fluorescence and Ultraviolet-visible spectroscopy. The CDs obtained using conventional hydrothermal carbonization showed the highest values of quantum yield (QY), whereas the nanoparticles obtained by the heating reflux reaction method exhibited the lowest QY values mainly due to incomplete polymerization and/or carbonization reactions of the material. In addition, a study to evaluate the use of the CDs as nanoprobes for metallic ions showed that one of the samples exhibited high selective and sensitive response toward Fe3+, with a detection limit of 0.89 µM. Furthermore, the nanoparticles used as sensors were obtained with an environmentally friendly strategy, and the CD-based photoluminescence method is simple, has a short response time and present low-cost.


Introduction
Carbon dots (CDs) have attracted a lot of attention in recent years, as they represent a relatively new class of carbon nanomaterials with interesting properties and multiple potential applications in different fields. [1][2][3] These nanoparticles consist of spherical carbonaceous materials ranging in size from 2-10 nm, exhibiting sp 3 -hybridized amorphous carbon containing small regions of sp 2 -hybridized carbon atoms. [4] They also possess interesting optical properties, such as strong absorption in the UV region and photoluminescence. [1] In addition, CDs also stand out in terms of water dispersibility, good biocompatibility, resistance to photodegradation, and easy surface functionalization due to the presence of chemically reactive surfaces, thereby expanding the possible applications of these nanoparticles. [5] CDs also have advantages over semiconductor quantum dots (QDs), [6] which are known to be toxic and environmentally hazardous. In contrast, carbon-based nanoparticles are not toxic and can be prepared using environmentally friendly methods such as hydrothermal carbonization. [7] In addition, low-cost precursors can be used for the preparation of CDs, such as biomass materials, [8,9] wastes [10] and a plethora of inexpensive low molar mass reagents. [11,12] In general, the approaches used to prepare CDs can be divided into two main strategies: top-down and bottomup. [1] Historically, top-down approaches have been used more in the first works describing the preparation of these nanoparticles, often resulting in nanocarbons with low fluorescence quantum yield (QY) values. [13] In this case, CDs can be obtained by breaking down larger carbon structures such as graphite oxide, [14] carbon soot [15] or carbon nanotubes [16] using different techniques such as arc discharge [17] or laser ablation. [18] More recently, bottom-up strategies have been used more widely, as these methods are easier to perform, less expensive, and generally produce CDs with higher QY values. [19] Among the main bottom-up strategies currently used, conventional hydrothermal carbonization using Teflon-lined autoclave (CHC) and microwave-assisted hydrothermal carbonization (MHC) can be highlighted. Another possible bottom-up approach to prepare CDs is the heating-reflux reaction method. In this method, CD can be obtained using low temperature and thus low energy consumption. [20] It is known that, in addition to the different precursors that can be used for the preparation of CDs, the experimental approach also plays a fundamental role in the physicochemical characteristic of the nanoparticles, mainly on the surface properties and optical characteristics of the nanomaterials. The nanostructures of carbon nanoparticles obtained by the bottom-up preparation methods have been described in recent years as having a "core-shell" structure, with an inner carbon core surrounded by a shell formed by different functional groups on the surface. [21] The relative size of the core and shell structures depends on the precursors and, mainly, on the methods and temperatures used. A higher degree of carbonization increases the "core size" due to the greater extent of carbonization of the nanoparticle. [21] Among possible applications of the CDs, their use as sensors for different substances such as pollutants, drugs and metal ions can be highlighted. [22][23][24][25] The use of fluorescent sensors has advantages over other detection techniques such as atomic absorption spectroscopy because they are easier to use, have simple operation, short response times, and low cost. [22,26,27] Considering different substances that can be used as sensors, such as QDs and organic dyes, carbon dots are advantageous due to their high photostability (compared to organic molecules) and low toxicity (compared to QDs). [22] Regarding the use of CDs as nanoprobes for metal ions, various reports have been described in the literature detailing the use of these carbon nanoparticles (either obtained from biomass or from low molar mass precursor) for the selective and sensitive sensing of different ions, such as Hg 2þ , [28] Pb 2þ [29] and Fe 3þ , [25,30,31] mainly in aqueous environment. In addition, the versatile properties of the CDs allow the preparation of magnetic nanoparticles that can be used not only in the identification of target analytes but also as promising agents in separation science. [32] In this work we have studied the influence of different bottom-up methods on the final properties of CDs obtained from different low molar mass precursors. The obtained nanoparticles were studied for their physical-chemical and optical properties, and the fluorescence quantum yield was used as an important parameter to characterize the final properties of the CDs, as it is a crucial factor for the majority of the applications of these nanocarbons. The obtained CDs were investigated as sensors for different metal ions, and one of the prepared fluorescent carbon nanoparticles showed considerable selectivity as well as sensitivity to the presence of Fe 3þ ions.

Preparation of the carbon dots
The CDs were prepared using three different carbon sources (suberic acid, 1,4 butanediol and citric acid) and two different sources of nitrogen (L-alanine and ethylenodiamine). Also, different methods were carried out to compare and study the different optical and physicochemical properties of the obtained nanoparticles. A detailed description of each approach to obtain the CDs is provided in the next sections.

Hydrothermal carbonization in a Teflon-lined
autoclave In a typical experiment, 15 mmol of a carbon source and an equal amount of a nitrogen source were added to 30 mL of deionized water and mixed thoroughly at room temperature on a magnetic stirrer to form a clear solution. The solution was then transferred into a Teflon-lined stainless-steel autoclave and heated at 180 C for 12 h. Afterward, the system was cooled until room temperature and the CDs suspension was filtered through a 0.22 mm membrane. Lastly, the resultant material was dried in a vacuum oven. The samples obtained by conventional hydrothermal carbonization (CHC) in autoclave were namely as SuEtA (using suberic acid and ethylenodiamine as the carbon and nitrogen source, respectively), BuEtA (obtained from 1,4 butanediol and ethylenodiamine), CaLaA (prepared from citric acid and L-alanine) and BuLaA (obtained from 1,4 butanediol and Lalanine).

Microwave-assisted hydrothermal carbonization
The same carbon and nitrogen sources, as well as their respective quantities, were used to prepare the CDs using the microwave-assisted hydrothermal carbonization method (MHC), as described for the CHC method. After mixing the precursors in the appropriate amounts in 30 mL of deionized water, the resulting clear solution was heated in a Milestone Start D microwave reactor at 180 C for 40 min. Then, the system was cooled to room temperature, the final suspension was filtered using a 0.22 mm membrane and finally the material was dried in a vacuum oven.
The samples were named as described in the previous section, adding the suffix "M" to indicate the microwaveassisted reaction method: SuEtM, BuEtM, CaLaM and BuLaM.

Heating reflux reaction method
In this method, a 1:1 molar ratio between the carbon and nitrogen source was maintained. In a typical experiment, 15 mmol of the carbon source and an equal amount of the nitrogen source were added to a solution containing 0.72 g of NaOH and 30 mL of deionized water. The reflux was maintained for 6 h at 120 C, and, at the end of the reaction, the suspension exhibited a yellow-brown color. The final suspension was filtered through a 0.22 mm filter, and the obtained material was dried in a vacuum oven.
The samples were named as described in the previous sections, adding the suffix "R" to indicate the reflux reaction method: SuEtR, BuEtR, AcLaR and SuLaR.

Metal ions sensing
Those CDs which reached the highest values of quantum yield of fluorescence (QY) were used as probes for the presence of metal ions.
Firstly, the selectivity was evaluated, adding 100 mL of a metal ion solution (0.01 mol L À1 ) into a 2.5 mL of a CD suspension at 20 mg L À1 concentration. For stabilization purposes, the fluorescence intensity of the CDs in the presence of the metal ions were collected 2 minutes after the addition of the metals into the carbon nanoparticles suspensions.
To study the sensitivity of the CDs as probes for metal ions, different amounts of a 25 mM solution of a specific metal ion was added to a CD suspension at concentration of 20 mg mL À1 . The PL intensity was collected 2 min after the addition of the metal ions to allow the system to reach equilibrium.
In order to investigate the quenching mechanism, we conducted quenching experiments using increasing concentrations of ferric ions at different temperatures (20, 30 and 40 C).

Characterization methods
Fourier transform infrared (FTIR) spectroscopy of the CDs was recorded using a Perkin Elmer Spectrum RXI spectrometer with the KBr pellet technique. The UV-Vis spectroscopy was performed using a Hitachi U-2010 Spectrophotometer. Measurements of the CDs aqueous suspensions were performed in 1.0 cm quartz cells to obtain the spectra. Transmission electron microscopy (TEM) images of the carbon nanoparticles were taken using a Tecnai G2-20 SuperTwin FEI (200 kV). The dilute aqueous suspensions of the CDs were deposited onto an ultrathin carbon film on a lacey carbon support film. The PL spectra were recorded using a Varian Cary Eclipse fluorescence spectrometer in different excitation wavelength.
To calculate the quantum yield of fluorescence (QY), quinine sulfate was used as reference standard. [33] The spectra of the different samples and the quinine sulfate (in 0.1 M H 2 SO 4 ) were acquired at different concentrations and at the k exc ¼340 nm. The QY was calculated by comparing the PL intensities and the absorbance values of the sample with the reference quinine sulfate using the following equation: In this equation, Ø is the quantum yield, ST is the standard, CD is the carbon dot sample, S is the slope (of the PL intensity against absorbance), and g is the refractive index.
Potentiometric titration curves were performed using an automatic titrator (Titroline 7000, SI Analytics). The experimental procedure describing the electrode calibration and the algorithm used to fit the experimental data were described earlier. [34,35] Around 5 mg of the CDs were dispersed in an HCl solution (3.4 Â 10 À3 mol L À1 ) in an electrochemical cell and then titrated with a CO 2 -free NaOH solution (0.0152 mol L À1 ). Afterward, the curves were fitted using an appropriate non-linear fitting method to determine the amount of each functional group on the surface of the CDs from their acid strength (pKa). [36] 3. Results and discussion

Characterization of the CDs
In this work, besides the different precursors used in the preparation of different CDs, distinct bottom-up methods were also performed in order to compare the physicochemical and optical properties of each nanoparticle obtained, which can allow to establish advantages and disadvantages of each experimental method used.
The experimental approach used for the preparation of carbon nanoparticles can result in significant physical-chemical differences and, consequently, differences in the optical properties of the CDs, such as the maximum fluorescence intensity (emission color) and quantum yield of fluorescence (QY). These two parameters are important for the characterization of carbon nanoparticles, especially the QY values, since for most of the CDs applications, including their use as sensors or as markers in the biomedical area, a minimum value of quantum yield is required.
It is important to mention that the sample obtained using suberic acid and L-alanine resulted in a sample with a considerable amount of precipitate, possibly due to the formation of macroscopic carbons and thus, a lower reaction yield related to the formation of carbon nanoparticles. Therefore, the preparation of CDs using these specific precursors was not studied in this work. Table 1 shows the QY values obtained for the different samples prepared.
The PL quantum yields of the obtained CDs were calculated using quinine sulfate as a standard. Figure S1 shows the PL intensity against the absorption obtained from different concentrations of quinine sulfate and the samples prepared by the CHC method.
The heating reflux reaction approach, which is little described in the literature as an alternative method for obtaining carbon dots, [20] normally presents the specific characteristic of using low temperatures to obtain the CDs when water is used as solvent. This could make it possible to prepare these materials with lower energy consumption. However, our results indicate that the use of this method (compared to other methods using the same precursors), resulted in nanoparticles with lower QY values.
On the other hand, the CHC and MHC methods showed considerably higher QY values compared to the reflux method. Furthermore, for the majority of the reactions, the conventional hydrothermal carbonization method showed higher values of QY when compared to the microwaveassisted method, indicating the advantages of using conventional hydrothermal carbonization over microwave-assisted carbonization.
Regarding the influence of the precursors using the same experimental method, it is known that the N element (dopant) contributes to enhance the QY of CDs. Furthermore, it has been reported that, in general, primary amines (when used as precursors) increase the QY more than secondary and tertiary amines, i.e., -NH 2 > -NH>-N. [37] Since both sources of N (EDA and L-alanine) used in this study were primary amines, the influence of each nitrogen element on the final properties of CDs can be comparable. Therefore, it is expected that the samples prepared from EDA as the Nprecursor will present higher QY values than those prepared using L-alanine as the nitrogen source, due to the higher amount of nitrogen in the EDA precursor (which contains two -NH 2 groups). This behavior was observed when hydrothermal carbonization was used as the experimental method. The samples with the highest QY values obtained by the microwave and reflux methods also had EDA as the nitrogen source in their formulations. For these latter methods, the results show that the use of suberic acid as the carbon source also played an important role in increasing the QY.
The quantum yield values can be studied mainly in terms of the surface groups on the carbon nanoparticles. In this work, the characterization of the surface functional groups was carried out using FTIR and potentiometric titration techniques. Figure 1 shows the FTIR spectra of the CDs obtained through the CHC method.
The FTIR spectrum of the SuEtA sample shows a signal at 3298 cm À1 , which can be attributed to N-H stretching vibrations due to formation of amide functional groups. The presence of these groups is confirmed by the typical bands of this functional group at 1642 (amide I) and 1550 cm À1 (amide II). Aromatic structures also absorb in these regions. [38] The presence of a shoulder around 3100 cm À1 suggests the presence of ¼ C-H stretch of sp 2 carbon. The bands at 2930 and 2856 cm À1 are assigned to C-H stretching. [39] The presence of these bands together with the signal at 1398 cm À1 suggests the presence of aliphatic carbon chains, whereas the C ¼ O stretching of carboxylic acid appears at 1709 cm À1 . [38] Reactions between alcohols and primary amines can lead to the formation of secondary amines (N-alkylation) and/or amides (through oxidative amidation of alcohols). [40] The sample BuEtA presented several of the functional groups found on the SuEtA carbon nanoparticles. The large band around 3100-3500 (centered at 3390 cm À1 ) is assigned to the O-H and N-H stretching. Similar to the SuEtA sample, the presence of typical C-H stretching of sp 3 carbon is observed at 2930 and 2857 cm À1 , as well as a shoulder around 3100 cm À1 assigned to ¼ C-H stretching of sp 2 carbon. The bands at 1658 and 1550 cm À1 are assigned to C ¼ O stretching of amides and -NH bending vibrations (amides and amines), respectively. The band occurring at 1050 cm À1 is characteristic of the C-O stretching of alcohols. [38] For the sample CaLaA, the wide band around 3430 cm À1 is related to O-H and N-H stretching. The significant widening in this region of the spectrum suggests the presence of hydrogen bonds in the structures, due to the presence of high concentrations of carboxylic functional groups. In fact, the results obtained from the potentiometric titration technique confirmed this high concentration, since 88.1% of the surface groups of this sample were identified as carboxylic acid groups ( Table 2). The signal at 2992 cm À1 is attributed to C-H stretching bond. [39] The band at 1722 cm À1 is assigned to C ¼ O from carboxylic acid functional groups, whereas the signal at 1642 cm À1 is due to C ¼ O stretching of amides. [38] Also, it can be mention that aromatic C ¼ C ring absorptions also appear in this region. The signal at 1208 cm À1 can be attributed to C-O-C stretching. [41] The spectrum of the BuLaA sample shows a broad band around 3400 cm À1 assigned to O-H and N-H stretching, whereas the signal around 3090 cm À1 represents ¼ C-H stretching of sp 2 carbon. [38] Typical bands of amides are observed in the region 1650-1500 cm À1 . In the fingerprint region of the spectrum (from about 1500 to 500 cm À1 ), several characteristic peaks of the precursors used (mainly from L-alanine) appeared, which may be due to incomplete polymerization and carbonization, and can explain the low QY value of fluorescence found for this sample.
The samples obtained by microwave-assisted reactions showed similar infrared spectra, with only small shifts observed for the signal related to -NH bending vibration  Figure S2a).
The same incomplete reaction (as observed for the CHC method) was observed between 1,4-butanediol and L-alanine precursors, resulting in the appearance of multiple signals in the fingerprint region. These signals showed different characteristic features of the L-alanine precursor. Thus, in the same way as the CDs obtained by the conventional hydrothermal carbonization, the nanoparticles obtained by the microwave method with these precursors also presented low QY of fluorescence.
The samples obtained by the reflux reaction method showed significantly different FTIR spectra (Supplementary Material, Figure S2b) compared to the samples obtained from CHC or MHC methods, mainly when comparing the spectra in the fingerprint region. The FTIR spectra of the samples obtained from the reflux method showed a considerable number of characteristics signals from the precursors, indicating incomplete reactions of polymerization and/or carbonization of the material. For example, the CDs BuLaR and CaLaR show several characteristic features of the precursor L-alanine in the FTIR spectra (in the region between 500-1500 cm À1 ). This suggests that the initial steps in the carbon nanoparticle formation mechanism were not completed. This can explain the lower quantum yield values for the samples obtained by the reflux method, which used significantly lower temperatures to obtain the nanoparticles.
The CD formation mechanism is characterized by a multi-step process. [21] The first steps in the "bottom-up" approach starts with polymerization reactions of the precursors, initially forming oligomers and then crosslinked polymers. With an increasing extent of these reactions, rigid entanglements are formed inside of the nanoparticles (core), while the formation of a shell also takes place, which is  formed by different functional groups on the outside of the CDs. Finally, the carbonization process takes place (from the core of the nanoparticles), decreasing the number of polymeric structures (with reactions such as deamination and dehydration), and increasing the relative amount of carbon in the CDs. It is expected that the process of nanoparticle formation and carbonization will only be effective if the system reaches a minimum value of temperature. This process includes the initial steps of oligomers formation, polymerization and the formation of highly crosslinked polymer nanoclusters, which are essential for the creation of surface defects (due to the presence of functional groups) on the surface of the CDs. The surface states (surface defects) due to the presence of different functional groups play a pivotal role in the fluorescence mechanism of CDs, and therefore, on the QY values. Thus, our results suggest that the temperature reached in the reflux system, using water as a solvent, was not sufficient for the effective formation of carbon nanoparticles with appreciable QY values. In this way, higher temperature values than those reached in the reflux method should be achieved to increase the extent of the main reactions that occur in the first steps of the CDs formation.
It is important to mention that the reflux method may only be suitable for obtaining CDs if the precursors used undergo a rapid and facilitated reaction in the early stages of nanoparticle formation, even at relatively low temperatures. This may depend on the specific precursors used. In such cases, the use of the reflux method may offer advantages as it requires lower energy consumption in the CD preparation process. Liu et al. [20] prepared CDs by the reflux method using only PEG as precursor, and obtained a material that was demonstrated to be useful for detecting low concentrations of Hg 2þ . The authors, however, did not report the value of QY of fluorescence obtained.
Additionally, when comparing the CHC and MHC methods, it is known that convection and conduction-based heating techniques require longer reaction times compared to the microwave-assisted method to ensure that the reaction medium fully reaches the desired temperature. [42] Furthermore, in microwave-assisted synthesis, microwave irradiation heats directly the precursor molecules, and excessive irradiation can cause partial decomposition of the precursors and oligomers formed in the early stages of the CD formation, leading to CDs with lower QY values of fluorescence. [42] Thus, even considering that we have used shorter reaction times in the case of microwave reactions, the "optimum time" in the case of microwave irradiation can be crucial for CDs formation (which includes polymerization, dehydration, carbonization, etc). On the other hand, the convection heating in the autoclave, which can be considered a preparation method under mild heating conditions (compared to the use of microwave irradiation), promoted the formation of CDs with higher QY values.
Potentiometric titration technique was also used to characterize the surface of the carbon dots as a complementary technique to FTIR spectroscopy.
CDs are usually rich in oxygen-and nitrogen-containing functional groups on their surface. The presence of these functional groups interferes directly with the physico-chemical and optical properties of the CDs. Then, Potentiometric titration can represent a useful technique to quantify the different acidic groups on these nanocarbons by the determination of the pKa values and quantification of the different surface functional groups. [36] Figure S3 shows the potentiometric titration curves of the samples obtained by CHC, MHC and reflux methods. From the potentiometric titration curves, the pKa's values were obtained. Table 2 shows the pKa's values of the surface functional groups for the samples obtained using CHC experimental method whereas Tables S1 and S2 (Supplementary Material) show the pKa's values of the surface groups of the samples obtained through MHC and reflux methods, respectively.
Comparing the CDs obtained from the same precursors and using different techniques (Tables 2, S1 and S2), the samples showed similar relative amounts of functional groups, indicating that the precursor plays a more important role than the experimental method (concerning the relative amount of functional groups), provided that the reactions have occurred to a minimal extent for the formation of the core/shell structure of the CDs. An exception is the sample obtained from 1,4 butanediol and EDA, which showed a higher number of phenolic and amines groups when obtained by CHC than the sample obtained by microwaveassisted reaction. The higher number of amino groups in the BuEtR sample was possibly due to the amount of residual ethylenediamine. The samples obtained from 1,4butanediol and L-alanine showed higher amounts of phenolic/amines groups.
It is difficult to establish a direct relationship between the relative amounts of functional groups and optical properties of carbon dots. Despite various efforts in the literature to establish a direct relationship between the optical properties of carbon dots (for example, QY values) and the presence and the number of nitrogen-or oxygen-containing groups, the examples shown in this work suggest that different factors can influence the optical properties of these nanoparticles. These factors may include the extent to which the initial reactions occur during the formation of carbon nanoparticles, such as those that take place between acids or alcohols and amines. This first step, as well as the subsequent ones (oligomer formation and polymerization) play an important role in the formation of the surface states, which are important to define the QY of fluorescence.
Since the chemical characteristics of the samples obtained from CHC or MHC methods are similar and the CDs obtained from CHC method presented higher values of QY (which is an important characteristic for different applications), the subsequent characterizations will be focused on these samples. Figure 2 shows transmission electron microscopy (TEM) images of the CDs samples obtained using the conventional hydrothermal carbonization method. The images were obtained from diluted suspensions of the nanoparticles that were dried at room temperature on an ultrathin (3 nm thickness) carbon film.
Some of the images obtained by the TEM technique showed low contrast of the carbon nanoparticles against the carbon film used as substrate. To aid visualization, a few of the CDs and also some agglomerates were marked with red circles.
The CDs were partially dispersed, showing the formation of some nanoparticle agglomerates in some images ( Figure  2a and c). The appearance of aggregates is expected, mainly when the dispersing medium is removed during the TEM sample preparation, due to the strong interaction established between the surface polar groups of the CDs. Using several TEM images for each sample, we were able to determine the size distribution of isolated CDs and the average size of the nanoparticles for the different samples. The histograms of the samples with Gaussian fitting curves are shown in Figure S4 (Supplementary Material), and reveal the corresponding particle size distribution of the CDs. The average particle size of the samples BuEtA and CaLaA were determine to be 1.2 nm, whereas for the SuEtA and BuLaA samples, these values were 1.7 and 2.2 nm, respectively.
It may be difficult to explain the subtle differences in size of different particles obtained from different carbon and nitrogen sources. However, in general, CDs with sizes smaller than 5 nm are expected to form, and they often have diameter values $ 2 nm, when nanoparticles are obtained from low molar precursors and using a bottom-up approach. [43][44][45] The presence of strong hydrogen-bonding intermolecular interactions between the precursors favors the formation of high-density cross-linked polymer nanocluster in the first stage of the nanocarbons formation. [21] Then, reactions such as dehydration and deamination take place, while the number of polymer structures decreases, and thereby increasing the relative amount of carbons in the cross-linked polymer chains. Thus, the denser the crosslinked polymer nanoclusters formed in the first stage, the smaller the nanoparticles formed in the successive stages of carbonization. Figure 3 shows UV-VIS absorption and PL spectra recorded at different excitation wavelengths in the UV region for the different carbon nanoparticles obtained using conventional hydrothermal carbonization. The samples were dispersed in water and the inset in each figure shows the nanoparticles samples illuminated under visible (left) and  The UV-VIS absorption spectra for the majority of the samples exhibited a band around 330-350 nm due to n!p Ã transitions of C ¼ O from different functional groups. [12] In addition to this absorption band, the CaLaA sample also exhibited an aromatic p-p Ã transition associated with the conjugated C ¼ C. [12] The sample BuLaA presented only the p-p Ã transition at $270 nm. The presence of a clear absorption band at 270 nm for these samples (CaLaA and BuLaA), indicates the presence of conjugated system and/or auxochrome.
The fluorescence curves of the different CDs showed strong emissions in the blue/green region of the spectrum when excited with UV light (from 300 to 460 nm wavelength). Moreover, the maximum emission peaks of the different CDs showed a shift depending on the excitation wavelength used. Additionally, a decrease in the intensity of the emitted light was observed as the excitation wavelength increased. This excitation-dependent PL behavior is explained by the distribution of different surface functional groups on CDs that create a broader distribution of surface defects with different energy levels. [46] 3.

CD as selective probe for different metal ions
In general, CDs exhibit distinct physico-chemical and optical properties depending on the different functional groups that may be present on the nanoparticles surface. In this work, since the CDs were prepared from distinct precursors, each nanoparticle may have different characteristics, so each obtained CD interacts differently with different metal ions. Thus, the different CDs prepared by CHC method (which showed the higher values of QY) were studied as probes for various metal ions.
Firstly, a selectivity study was carried out using the photoluminescence quenching effect. Solutions of different metal ions (Fe 2þ , Co 2þ , Cu 2þ , Ni 2þ , Pb 2þ , Cr 3þ , Mn 2þ , Hg 2þ , Zn 2þ and Fe 3þ ) at a concentration of 0.01 mol L À1 , were added to 2.5 mL of a CD suspension at a concentration of 20 mg L À1 . After 2 min of stabilization, the PL intensity was recorded. In order to show that the effect of decreasing intensity with the dilution of the CD sample in deionized water is negligible, the same amount of water (comparing with the ion solution), was also added to record the PL intensity.
In the presence of Fe 3þ ions, the samples SuEtA and BuLaA showed a photoluminescence quenching of around 40% (results not shown here), whereas for the other metal ions, the PL intensity remained almost unchanged. Even though this result can be considered as a selective behavior for Fe 3þ ions, the quenching was not significant enough to consider these CDs as probes with considerable sensitivity. On the other hand, the sample BuEtA exhibited a considerable quenching effect for various metal ions, which suggests a lack of specificity when considering the different metal ions studied.
In contrast, the sample CaLaA exhibited considerable specificity for Fe 3þ (Figure 4). It showed a PL quenching of around 83% in the presence of Fe 3þ ions, while the PL intensity with the addition of other ions remained almost unchanged compared to the addition of the same amount of DI water. Thus, it clearly shows the selectivity of the sample CaLaA for Fe 3þ ions.
The CaLaA sample has a higher amount of carboxylic acid groups on its surface (FTIR results, Figure 1 and Potentiometric Titration, Table 2). Fluorescence quenching by metal ions in carbon dots has been described in the literature as a result of strong interaction and coordination of metal ions with oxygenated and nitrogenated functional groups on CDs surface. [47] In this way, the results obtained in this work suggest a strong interaction between ferric ions and the main functional groups found on the CaLaA sample, i.e., carboxylic groups, which represent strong sites for metal ions attachment. [48,49] This great interaction leads to a coordination between the nanoparticles and the metal ion, which is associated with the "excellent" coupling of the electronic levels of the CDs dots with the half-filled 3d orbital of the ferric ions, after the breakdown of degeneracy. [47,49] Once the metal is coordinated on the CDs surface, the excited electrons of the nanoparticles can be effectively transferred to the electronic level of the Fe 3þ , which leads to a non-radiative electron-hole annihilation, inducing strong PL quenching. [47] To investigate the sensitivity of this sample in the presence of ferric ions, we also studied the photoluminescence behavior as a function of the Fe 3þ ion concentration. After adding different amounts of 25 mM Fe 3þ solution to 2.5 mL of the CD suspension at a concentration of 0.02 mg mL À1 , a gradual decrease in PL intensity was observed (Figure 5a). The relation between the reduction in PL intensity of the CaLaA sample and the concentration of ferric ions is linear (R 2 ¼ 0.99) in the concentration range studied (from 0 to 22 mM).
The Stern-Volmer equation describes fluorescence behavior by the following equation: In this equation, K SV is the Stern-Volmer quenching constant, F o is the photoluminescence intensity of the neat CD and F is the PL intensity of the carbon nanoparticles in the presence of the metal ion. The Stern-Volmer plot for the sample CaLaA in the presence of the ferric ions is shown in Figure 5b, showing good linearity from 0 to 22 mM. The limit of detection (LOD) can be calculated using the relation 3d/S, where d is the standard deviation and S is the slope. Using this relation, the LOD was calculated as being 0.89 mM.
Although there are some reported works where the CDbased photoluminescence method performed better than the one described in this work [25] , the LOD value obtained in this study is similar to most values reported in the literature for CDs used as nanoprobes for Fe 3þ ions. [30,31,50] Additionally, this value is lower than the maximum permissible concentration of ferric ions in drinking water by the World Health Organization (WHO) which is 5.36 mM. [51] In order to elucidate whether the quenching mechanism that occurs with the CaLaA sample in the presence of Fe 3þ ions was static or dynamic, a study with incremental concentrations of ferric ions at different temperatures was carried out. In a dynamic mechanism, an increase of the quenching effect with temperature is expected because the collisions are stronger between the quencher and the fluorophore. This leads to an increase in the Stern-Volmer quenching constant, K SV , with temperature. On the other hand, in the static mechanism, a decrease in the quenching effect with temperature is observed, since the thermal agitation at higher temperatures decreases the stability of the molecular complex formed between the CD and the quencher. This leads to a partial recovery of PL emission, and therefore, a decrease of K SV at higher temperatures is observed. [52,53] Figure 6 shows the Stern-Volmer plot for CaLaA sample in the presence of incremental concentrations of Fe 3þ ions at different temperatures. The calculated values of the Stern-Volmer constant were 8.20 Â 10 3 , 7.52 Â 10 3 and 7.10 Â 10 3 L mol À1 at 20, 30 and 40 C, respectively. This decrease in K SV values with temperature indicates a static quenching mechanism.
The results obtained for the CaLaA sample show that the obtained nanoparticles can be used as a selective and sensitive probe to Fe 3þ ions in an aqueous environment. In addition, the preparation of the nanoprobes can be achieved by a simple and one step method, and the CD-based photoluminescent sensor method is cheap, simple and environmentally friendly.

Conclusion
In this work, different carbon dots were obtained using distinct bottom-up methods and precursors to investigate the physicochemical and optical properties of the obtained nanoparticles and to determine the advantages and drawbacks of each experimental approach.
Compared with the MHC method, the CHC method showed higher QY values, demonstrating the advantage of using conventional hydrothermal carbonization over microwave-assisted carbonization. On the other hand, the reflux method showed the lowest QY values due to incomplete reactions that are important for the formation of carbon nanoparticles. Regarding the precursors, the samples prepared from EDA as N-precursor showed higher QY values than those that used L-alanine as N-source, due to the higher amount of N in the EDA precursor. The average particle size of the different samples obtained by the CHC method was determined to be between 1.2 to 2.2 nm, and these low diameter values were attributed to the dense crosslinked polymer nanoclusters formed in the early stages of the nanoparticles formation. The fluorescence spectra of the different CDs showed strong emissions in the blue/green region of the spectrum when excited with UV light, and the peak position of the emission spectra shifted depending on the excitation wavelength due to the distribution of different surface functional groups on the CDs, resulting in a broader distribution of surface defects with different energy levels.  Furthermore, the obtained CDs were studied as metal ion sensors, and the sample that presented greater number of carboxylic acid groups on the surface (obtained from citric acid and L-alanine as precursors) showed high selectivity and sensitivity for Fe 3þ ion detection with LOD ¼ 0.89 mM.
Finally, the approach used here can be extended to develop different selective and sensitive nanoprobes for different meal ions such as Hg and Pb, varying the precursors and thus the functional groups on the nanoparticles surface.