Multichannel Sensor for Detection of Molybdenum Ions Based on Nitrogen-Doped Carbon Quantum Dot Ensembles

: Trace elements such as cobalt (Co), molybdenum (Mo), and zinc (Zn) play necessary roles in different biological functions. Co is a microelement that influences the vascular system. Mo works as an enzymatic cofactor of three enzymes (aldehyde oxidase, sulfite oxidase, and xanthine oxidase dehydrogenase). However, these elements are difficult to detect, since the analytical methods developed have a high cost, which restrict their applicability. In this sense, fluorescent sensors are an alternative for detecting trace elements, such as Mo 4+ ions. Herein, a new multichannel trace elements sensor has been proposed to detect Mo entities. In this sense, two different N-CQDs were synthesized and fully characterized. The N-CQDs presented quantum yield values of 25.93% and 6.02% and excellent solubility in water. Also, a mixture of these two carbon-based nanoparticles was used to identify and to quantify Mo in water between seven different trace elements. The method was found to reach 1.28 and 3.88 ppm for limit of detection (LOD) and quantification (LOQ), respectively. To further verify the potential of the detection platform, the multichannel sensor was applied to identify the different concentrations of metal ions (Fe 2+ , Co 2+ , Mn 2+ , Cu 2+ , Zn 2+ , Mg 2+ , and Mo 4+ ) in water. The data matrix was treated using different algorithms, such as K-Means and Discriminant Analysis (DA). The detection strategy has successfully identified the molybdenum ions at 5 ppm. This result shows the potential application of a multichannel sensor toward the detection of Mo entities, since it is comparable with the molybdenum test already available on the market.


Introduction
Trace elements are essential for different physiological functions.Some examples include copper (Cu), molybdenum (Mo), and cobalt (Co).For instance, Cu is vital for many living organisms and plays an important role in physiological and pathological processes, such as bone formation and cellular respiration [1].Mo acts as an enzymatic cofactor of several enzymes (such as aldehyde oxidase, sulfite oxidase, and xanthine oxidase dehydrogenase), which catalyzes several substrates' hydroxylation [2].However, abnormal levels of these elements in the human body pose a severe threat, since it may cause distinct diseases.An excessive intake of Co 2+ can lead to serious health issues, including blindness and deafness [3].So, it is clear the demand for the detection of trace elements.Molybdenum ions are easily detected due to their interactions with carbonaceous C 2024, 10, 57 2 of 16 matrices.For example, Amin and collaborators developed a polymeric membrane to incorporate the metal.The authors attribute the interaction mechanism, which resulted in changes from the optical signal to electrostatic attraction [4].Consequently, many methods have been developed to detect these kinds of chemical entities.In this sense, inductively coupled plasma mass spectroscopy (ICP-MS), atomic absorption spectroscopy/emission spectroscopy, potentiometry, and voltammetry may be mentioned.These techniques are found to be costly, technically complicated, time-consuming, and require expensive instruments.Therefore, the development of an optional method is imperative.In this context, fluorescent sensors emerged as an alternative strategy for metal detection owing to their simplicity, high sensitivity, good selectivity, and rapid response [5,6].
Carbon quantum dots (CQDs) are a new class of fluorescent nanoparticles (NPs) that have attracted the interest of many researchers worldwide.These carbon-based nanomaterials display biocompatibility, chemical stability, green synthesis, as well as surface modifications with different functional groups [7][8][9].Furthermore, it is possible to adjust their structural composition to improve spectral properties [10].Currently, numerous reports have been focused on nitrogen, boron, sulfur, phosphorus, or silicon doped-CQDs [11,12].By introducing atomic impurities into CQDs, their electronic structure can be adjusted, generating n-type or p-type carriers [13].Therefore, due to the versatility of the CQDs, their applications have broadened in the sensing field, where it was already used to detect organic molecules (e.g., quinolinic acid in human serum) [14], and inorganics (boric acid and hydroxylamine hydrochloride) [15].Also, cationic (e.g., Fe 3+ and Cu 2+ ) [16,17] and anionic species (NO 2 − in meat) [18] were detected.However, identifying the target analyte is highly desired, since it would allow for a fast decision toward the best medical treatment.Therefore, in recent years, the CQDs were considered to be excellent candidates to develop an array-based strategy to discriminate distinct analytes like proteins [19,20], natural amino acids [21], pesticides [22], food additives [23], antibiotics [24], phosphate anions [25], metallic ions [26,27], and others [28].
Sensor arrays are based on the "chemical nose-tongue" system, defined as an ensemble of receptors (for instance, synthetic molecules and/or nanomaterials), to mimic mammalian olfactory systems [29].A "nose"-based strategy developed using CQDs provides differential events to create a unique fluorescent pattern, which is further analyzed using computational tools [30,31] to discriminate/classify the target analytes.This type of strategy was developed based on electronic nose and colorimetric sensors, which presented impressive detection results [32,33].Thereby, this kind of methodology may differentiate a broad range of analytes in a complex matrix using relatively few single receptors to generate fingerprint-like signals [34].In addition to these advantages, the strategy may be time-consuming, which is undesired.So, a multichannel sensing [35][36][37] strategy came out to surpass this difficulty.The combination of different receptors creates a cross-reactive array able to provide multiple events by performing one single measurement.Therefore, it is possible to reduce the time of analysis, as well as the volume used.Also, it allows us to increase the sample throughput.Herein, we developed a multichannel sensing strategy using two different types of N-based CQDs to identify and quantify trace elements, such Mo 4+ ions.By using our methodology, the two "nose"-based receptors were able to identify the molybdenum ions with 100% of accuracy.

Apparatus
Regarding CQDs characterization, the optical properties were investigated by acquiring the UV-Vis spectrum in the spectrophotometer Uv-Vis Shimadzu UV-2600, Kyoto, Japan.The fluorescence spectra were obtained in the Shimadzu RF-6000 fluorophotometer, Kyoto, Japan.The structural-functional groups on the surface were studied through X-ray photoelectron spectroscopy (XPS) (Physical Electronics Quantum 2000 spectrometer) using a monochromatic Al Kα excitation at a spot size of 10 mm with a pass energy of 46.95 eV, Turku-Finland and FT-IR VERTEX 70 V, Columbus, OH, USA, and Raman Spectroscopy investigated the carbon structure of the nanomaterials.The Raman Spectrum was recorded in a range of 1000-3000 cm −1 using a Raman HORIBA spectrometer, Columbus, OH, USA.The spectral excitation was performed in a laser using a 785 nm line.The Atomic Force Microscopy (AFM) image was obtained using an Asphalt Research microscope of the MFP-3D AFM type, Zurich, Switzerland.

Synthesis of N-CQDs
Initially, the first N-doped-CQD (labeled as CQD-NH 2 ) was synthesized according to the well-described methodology of Wang et al. [38].Briefly, 500 mg of citric acid (CA) with 500 µL of ethylenediamine were added into a 10 mL beaker.Then, it was led to the muffle, previously heated at 180 • C, where it remained for 30 min.After this step, the recipient was allowed to cool down naturally, and the product was solubilized in Milli-Q water.The second N doped-CQD (labeled as CQD-CONH 2 ) was produced following the protocol previously described by Chen et al. [39].In this regard, 1.0 g of CA and 2.0 g of urea were dissolved in 20 mL of DMF and stirred to form a transparent solution.Subsequently, the solution was transferred into a 25 mL Teflon-lined stainless autoclave.The sealed autoclave vessel was heated at 200 • C and held for 6 h.Afterward, the mixture was held at rest until it cooled to room temperature.Once both CQDs dispersions were cooled down, their purification was carried out through the dialysis membrane (cut-off tubing in the range of 1 kDa) for 48 h.Finally, each colloidal dispersion was lyophilized to obtain the powder, and was stored in the fridge at 4 • C for further characterization and application.

Quantum Yield
Quinine sulphate was used as a standard for the determination of quantum yield (QY) of the prepared CQDs.The fluorescence spectra of quinine sulphate (in 0.1 mol.L −1 H 2 SO 4 as a solvent, QY = 0.54) and CQDs solution, as well as their absorbances, were recorded at 350 nm.The following Equation was used to calculate the QY (ϕ): where ϕ re f is the standard of quinine sulfate, I is the integrated emission intensity, A is optical absorbance at 350 nm, and n is the refractive index of the solvent.Subscript ref denotes the values obtained for the standard, representing the known quantum yield [40].

Metallic Ions Sensing Study
Firstly, stock solutions of different metallic cations (Co 2+ , Cu 2+ , Mg 2+ , Mn 2+ , Fe 2+ , Mo 4+ , and Zn 2+ ) were previously prepared in Milli-Q water.In addition, to accurately calculate the concentration of the CQDs, a calibration curve (concentration x absorbance) was previously constructed for each sample based on the mass/volume ratio.Then, for the sensing studies, the concentration of CQDs was measured at 350 nm and 400 nm for CQD-NH 2 and CQD-CONH 2 , respectively.The selected wavelengths correspond to the absorption band maximum of the respective CQDs.This procedure was carried out separately, due to the distinct chemical structure of the two nanoparticles.So, by knowing the precise concentration of both CQDs, a dispersion containing CQD-NH 2 and CQD-CONH 2 was prepared.Subsequently, two emission fluorescence spectra were recorded, taking into consideration the conditions for each CQD.In this sense, the first one was taken in the range of 380-650 nm using 360 nm as the excitation wavelength, while the second one was measured from 440 to 650 nm, considering 420 nm as the excitation wavelength.Each spectrum was used as the blank.Afterward, we carried out a procedure where eight aliquots of 10 µL were removed from a metal ion stock solution and added into the N-CQDs dispersion previously prepared and measured.For each addition, two spectrum were recorded, regarding the previously reported measurement conditions.This protocol was repeated six times for each metal ion investigated (Co 2+ , Cu 2+ , Mg 2+ , Mn 2+ , Fe 2+ , Mo 2+ , and Zn 2+ ).

Data Treatment
Taking into consideration the protocol described in Section 2.5, a raw data matrix of 2 nose-based receptors × 7 metal ions × 8 concentrations × 6 replicates were generated for each metal cation analyzed.The ratio I 0 I (where I and I 0 denote the final and initial intensity of the fluorescence spectrum at 450 nm and 515 nm, respectively) was the signal investigated as a function of ion concentration.The LOD and LOQ parameters were estimated according to the International Conference on Harmonization's (ICH) guidelines, which indicates the following formulas: LOD = 3.3 × SD S and LOQ = 10 × SD S .SD denotes the standard deviation of y-intercept, while S represents the slope of the regression line.The curves were further fitted using the Stern-Volmer equation, and the n parameter was evaluated.In addition, the whole raw data matrix was evaluated using the K-Means, PCA, and LDA methods to identify/classify each target analyte [22].

CQDs Characterization
UV-Visible and fluorescence spectroscopies were carried out to investigate the optical properties of the CQD-NH 2 and CQD-CONH 2 , both being prepared in an aqueous solution.Figure 1a,b displays the presence of two strong absorption bands around 350 and 335 nm, respectively.These transitions can be assigned to the n-π* transition of C-O, C=O, and N-H functional groups [41,42], which evidences the doping of N in both samples.Moreover, the absorption spectrum for CQD-CONH 2 (Figure 1b) displayed additional bands centered around 430 nm, 510, and 561 nm.These transitions may be attributed to the presence of abundant N functional groups, as the N doping process is capable of inducing N-related defect states between the HOMO and LUMO [39].In Figure 1c, the intensity of the photoluminescence emission of the CQD-NH 2 at 450 nm remained unchanged when the excitation wavelength was increased from 300 to 400 nm.This clearly indicates that the CQD-NH 2 maximum fluorescence emission is independent of the excitation wavelength.This behavior can be explained by surface defects, but also by the uniform size distribution of the nanoparticles.This distribution can be confirmed using AFM imaging [43][44][45].Still, the emission of CQD-CONH 2 (Figure 1d) is clearly excitation-dependent, since the fluorescence could be observed in the range 448-620 nm when excited at different wavelengths (360-520 nm). Figure S1 shows this dependence in point graph style.In this case, the electron-rich nitrogen atoms could offer more active sites, and the excitation-dependent fluorescence behavior could be related to the different surface states of the CDQ-CONH 2 [46].The quantum yield (QY) of CQD-NH 2 and CQD-CONH 2 in the water was measured to be 25.93% and 6.02%, respectively, using quinine sulphate as a reference.X-ray photoelectron spectroscopy (XPS) analyses were performed to characterize functional groups into CQDs.The CQD-NH2 and CQD-CONH2 showed compositions C1s, N1s, and O1s (Figure 2a,b).In CQD-NH2, the high resolution of C1s (Figure 2c) a the binding energy peak at 284.6 eV confirms the graphitic structure sp 2 (C-C and C= The peak at 286.3 eV suggests the presence of C-N, while the peak around 288.0 eV cou be assigned to C=O bonds [47,48].The O1s spectrum exhibited two peaks at 530.8 a 532.0 eV (Figure S2), which can be assigned to the C=O and C-O-C/C-OH groups, resp tively [49].The high-resolution N1s (Figure 2e) show two peaks, 399.8 and 401.6eV, signed to the absorption of N-H and C-N-C groups, respectively [50][51][52][53].In CQD-CON the C1s band can be mainly deconvoluted into two peaks (Figure 2d), which repres graphitic sp 2 carbons (C=C/C-C, 284.6 eV) and carbonyl groups (C=O, 286.5 eV), and peak around 288.0 eV could be attributed to C=O bonds.The O1s band can be assigned two peaks (Figure S1b), C=O (532.3 eV) and C-O (533.4 eV).The N1s band can be fitt into one peak (Figure 2f), which corresponds to pyrrolic N (399.5 eV) [39].However, attribution of this peak to the N-H groups can not be discarded, due to the proximity the bound energy values.X-ray photoelectron spectroscopy (XPS) analyses were performed to characterize the functional groups into CQDs.The CQD-NH 2 and CQD-CONH 2 showed compositions of C1s, N1s, and O1s (Figure 2a,b).In CQD-NH 2 , the high resolution of C1s (Figure 2c) and the binding energy peak at 284.6 eV confirms the graphitic structure sp 2 (C-C and C=C).The peak at 286.3 eV suggests the presence of C-N, while the peak around 288.0 eV could be assigned to C=O bonds [47,48].The O1s spectrum exhibited two peaks at 530.8 and 532.0 eV (Figure S2), which can be assigned to the C=O and C-O-C/C-OH groups, respectively [49].The high-resolution N1s (Figure 2e) show two peaks, 399.8 and 401.6eV, assigned to the absorption of N-H and C-N-C groups, respectively [50][51][52][53].In CQD-CONH 2 , the C1s band can be mainly deconvoluted into two peaks (Figure 2d), which represent graphitic sp 2 carbons (C=C/C-C, 284.6 eV) and carbonyl groups (C=O, 286.5 eV), and the peak around 288.0 eV could be attributed to C=O bonds.The O1s band can be assigned to two peaks (Figure S1b), C=O (532.3 eV) and C-O (533.4 eV).The N1s band can be fitted into one peak (Figure 2f), which corresponds to pyrrolic N (399.5 eV) [39].However, the attribution of this peak to the N-H groups can not be discarded, due to the proximity of the bound energy values.Figure 3 shows the Raman spectra for N-CQDs samples, which was used to characterize the intrinsic structure of the particles.From the Raman spectra, we performed a deconvolution into four other spectra, using Gaussian curves, whose statistical parameters are presented in Table S1.Among these four curves, the two with the highest intensities correspond to bands D (A1g mode) and G (E2g mode) for both CQDs.In this context, the G bands were found to be at 1548 cm −1 and 1505 cm −1 for CQD-NH2 (Figure 3a) and CQD-CONH2 (Figure 3b), respectively.According to the literature [54], this band can be attributed to the vibration of sp 2 carbon atoms in a two-dimensional hexagonal (2D) graphite grid.The D Band is centered at 1386 cm −1 for CQD-NH2 (Figure 3a) and at 1393 cm −1 for CQD-CONH2 (Figure 3b).This band corresponded to the vibration of the sp 3 hybridized carbon atoms in the termination plane of the disordered graphite [55,56].Furthermore, it is well-known that the ID/IG ratio can provide relevant information about the defects of the structure.So, the ratio values were calculated to be 1.12 and 0.82 for CQD-NH2 and CQD-CONH2, respectively.However, our results evidence the presence of a greater amount of amorphous nature in CQD-CONH2.From XPS results, this sample Figure 3 shows the Raman spectra for N-CQDs samples, which was used to characterize the intrinsic structure of the particles.From the Raman spectra, we performed a deconvolution into four other spectra, using Gaussian curves, whose statistical parameters are presented in Table S1.Among these four curves, the two with the highest intensities correspond to bands D (A 1g mode) and G (E 2g mode) for both CQDs.In this context, the G bands were found to be at 1548 cm −1 and 1505 cm −1 for CQD-NH 2 (Figure 3a) and CQD-CONH 2 (Figure 3b), respectively.According to the literature [54], this band can be attributed to the vibration of sp 2 carbon atoms in a two-dimensional hexagonal (2D) graphite grid.The D Band is centered at 1386 cm −1 for CQD-NH 2 (Figure 3a) and at 1393 cm −1 for CQD-CONH 2 (Figure 3b).This band corresponded to the vibration of the sp 3 hybridized carbon atoms in the termination plane of the disordered graphite [55,56].Furthermore, it is well-known that the I D /I G ratio can provide relevant information about the defects of the structure.So, the ratio values were calculated to be 1.12 and 0.82 for CQD-NH 2 and CQD-CONH 2 , respectively.However, our results evidence the presence of a greater amount of amorphous nature in CQD-CONH 2 .From XPS results, this sample C 2024, 10, 57 7 of 16 presented a greater amount of C-N and C=O bonds in the structure in comparison to the CQD-NH 2 .Thus, it is fair to assume that the presence of these kinds of chemical bonds in the termination plane of the carbon structure may contribute to increasing the disorder.
CONH2 displayed bands at 1665-1594 cm −1 , which can be assigned to the stretching vibration of the amide linkage, indicating the incorporation of many desired amino groups in the N-doped CQD's structure.These revealed C=O stretching, O-H bending, and N-H bending bands indicate the presence of carboxylic and amide functional groups on CQDs surface.The bands C-N 1368, 1188, and 1117 cm −1 are assigned C=C, C-N, and C-O groups [57][58][59].The morphology of CQDs was characterized using AFM. Figure 3c,d shows the topographic images of CQD-NH2 and CQD-CONH2, respectively.It can be seen that the particle size distribution is relatively uniform for both N-CQD samples, and their height profiles are in the range of 1-4 nm (inset Figure 3c,d), which can be understood to be nanoparticles with quasi-spherical morphology [60,61].

N Doped-CQDs as "Nose"-Receptors
Once the CQDs were fully characterized and their properties were revealed, the next step was to find the "nose"-based receptors to build up a multichannel sensing methodology.Our hypothesis is that the presence of CQD-NH2 and CQD-CONH2 in a mixture can help to obtain an unique fluorescence pattern, which will further allow us to perform the analyte discrimination.In this sense, both CQDs were firstly mixed, requiring an appropriate concentration to acquire a balance between the fluorescence intensity.As a result, a unique dispersion of concentrations of 1.0 and 1.8 mg.mL −1 for CQD-NH2 and CQD- The functional groups of CQD-NH 2 and CQD-CONH 2 were investigated through FTIR spectra (Figure S3).The typical N-H bands at 3268 for CQD-N and 3325 cm −1 for CQD-CONH 2 were observed to originate from conjugated amine groups (stretching vibrations of N-H) connected to the carbon core.Furthermore, the FTIR spectrum for CQD-CONH 2 displayed bands at 1665-1594 cm −1 , which can be assigned to the stretching vibration of the amide linkage, indicating the incorporation of many desired amino groups in the N-doped CQD's structure.These revealed C=O stretching, O-H bending, and N-H bending bands indicate the presence of carboxylic and amide functional groups on CQDs surface.The bands C-N 1368, 1188, and 1117 cm −1 are assigned C=C, C-N, and C-O groups [57][58][59].The morphology of CQDs was characterized using AFM. Figure 3c,d shows the topographic images of CQD-NH 2 and CQD-CONH 2 , respectively.It can be seen that the particle size distribution is relatively uniform for both N-CQD samples, and their height profiles are in the range of 1-4 nm (inset Figure 3c,d), which can be understood to be nanoparticles with quasi-spherical morphology [60,61].

N Doped-CQDs as "Nose"-Receptors
Once the CQDs were fully characterized and their properties were revealed, the next step was to find the "nose"-based receptors to build up a multichannel sensing methodology.Our hypothesis is that the presence of CQD-NH 2 and CQD-CONH 2 in a mixture can help to obtain an unique fluorescence pattern, which will further allow us to perform the analyte discrimination.In this sense, both CQDs were firstly mixed, requiring an appropriate concentration to acquire a balance between the fluorescence intensity.As a result, a unique dispersion of concentrations of 1.0 and 1.8 mg mL −1 for CQD-NH 2 and CQD-CONH 2 , respectively, was prepared.Figure 4a shows the emission behavior of the mixture when excited at different wavelengths (360-420 nm).Clearly, there is an emission band for each CQD, depending on the excitation wavelength used.For instance, a strong emission from CQD-NH 2 , centered at 450 nm, was obtained by using 360 nm as excitation wavelength.On the other hand, when excited at 420 nm, the emission from CQD-CONH 2 was found to be centered at 515 nm.In addition, no significant overlap was observed, which helps it to obtain a clean emission with no interference from other CQDs.Therefore, after this optimization, the dispersion containing both CQDs was tested against seven different metal ions (Co 2+ , Fe 2+ , Mo 4+ , Cu 2+ , Mg 2+ , Mn 2+ , and Zn 2+ ). Figure 4b-h displays the fluorescence responses of CQD mixtures against these seven metal ions at different concentrations in the range of 100-700 µmol L −1 .
C 2024, 10, x FOR PEER REVIEW 8 of 17 CONH2, respectively, was prepared.Figure 4a shows the emission behavior of the mixture when excited at different wavelengths (360-420 nm).Clearly, there is an emission band for each CQD, depending on the excitation wavelength used.For instance, a strong emission from CQD-NH2, centered at 450 nm, was obtained by using 360 nm as excitation wavelength.On the other hand, when excited at 420 nm, the emission from CQD-CONH2 was found to be centered at 515 nm.In addition, no significant overlap was observed, which helps to obtain a clean emission with no interference from other CQDs.Therefore, after this optimization, the dispersion containing both CQDs was tested against seven different metal ions (Co 2+ , Fe 2+ , Mo 4+ , Cu 2+ , Mg 2+ , Mn 2+ , and Zn 2+ ).As can be seen, the titration experiments pointed out that each metallic cation was able to quench both CQDs in the mixture.To properly explain the interaction, as well as the quenching mechanism, it is important to understand the structure of the carbon-based NPs synthesized in this work.In this regard, according to our XPS results, the CQDs have different functional groups on their surfaces.CQD-NH 2 was found to be rich in amine groups, while CQD-CONH 2 contained amide groups.Therefore, the presence or absence of these groups on the surface of N doped-CQDs can strongly affect their interactions.In order to obtain a better insight, the titration curves were described using the Stern-Volmer equation [62]: where I and I 0 denote the fluorescence intensity recorded with the presence absence of metallic cations, respectively, K SV is the Stern-Volmer quenching constant, and [C] is the molar concentration of quencher.Herein, by assuming a linear relationship between I 0 I and [C] with an intercept of 1, K SV can be easily obtained through graphical determination.The magnitude of this constant may help to understand why some metallic cations are more effective to quench the fluorescence intensity than others.Also, to go further, the number n of binding sites on the surface of the CQDs were calculated by utilizing a modified version of the Stern-Volmer equation [63].The data are summarized in Table 1.The R 2 values for all fits were found to be in the range of 0.9129-0.9935.This evidences that more than 90% of the data variance in the y-axis (normalized fluorescence intensity) can be explained by the x-axis (quencher concentration) using a linear model.So, it is fair to assume that there is a strong linear relationship between I 0 I and [C].However, the deviation of linearity may signal the presence of a dynamic (collisional) or static (bindingrelated) quenching mechanism [64].Once it is not possible to determine which one has the quenching mechanism from a single linear Stern-Volmer plot, we carried out titrations at different temperatures (298.15 and 310.15 K) for Co 2+ , Cu 2+ , Fe 2+ , and Mo 4+ ions.In this sense, to identify the mechanism that better defines the interaction between CQDs and metallic cations, it is fundamental to know that if the quenching is dynamic, the value of K SV should increase.On the other hand, if the mechanism is static, K SV should decrease by increasing the temperature.In this sense, the predominant quenching mechanism for Co 2+ C 2024, 10, 57 10 of 16 and Fe 2+ ions was found to be dynamic, while for Cu 2+ and Mo 4+ ions, the static one was the observed one.Additionally, n values for Fe 2+ , Zn 2+ , Mg 2+ , Mn 2+ , and Co 2+ are near to 1.0 for both CQDs, indicating that one empty site on the CQDs surface interacts with the mentioned metals.
However, the situation differs when Cu 2+ ions are used as a quencher.In this case, the presence of amide groups on the surface of the CQDs seems to increase the number of binding sites for Cu 2+ ions, since the n values were found to be 0.89 ± 0.04 and 1.51 ± 0.06 for CQD-NH 2 and CQD-CONH 2 , respectively.Additionally, n ≈ 2 was observed for Mo 4+ ions when used as quencher for both CQDs.Compared to the other metallic cations, Mo 4+ ions are the ones with highest K SV values.Taking into consideration that this ion is the quencher, the values found were 61.67 × 10 2 ± 214.67 and 17.20 × 10 2 ± 115.89 M −1 for CQD-NH 2 and CQD-CONH 2 , respectively, while the K SV values for the other ions were found to be in the range of 1.80 × 10 2 -12.55 × 10 2 M −1 .Once this constant signals for the interaction between the CQD's surface and the metallic cations, it is fair to assume that the Mo 4+ ions are the ones presenting the strongest interaction.This may be explained based on the electronic configuration of the cation, since the Mo 4+ is the ion with the highest amount of empty d-orbitals, which may favor the interaction with an electron-rich surface.
Overall, the results demonstrated the great potential of the multichannel sensor developed for detecting Mo 4+ ions.Therefore, the LOD and LOQ parameters were calculated according to the ICH guidelines to further deepen the analytical parameters regarding the sense of this cation.However, to take advantage of the multichannel sensor, the interaction of the Mo 4+ ions with the two N-CQDs was used.Herein, the difference between K SV and n values calculated through the titration studies demonstrates the diversity of interactions on the CQDs surface.So, this that the sensibility of the methodology may increase, taking into consideration the interaction of the target ions with CQD-NH 2 and CQD-CONH 2 .In this sense, Figure 5 displays a schematic illustration of the dual-response sensor array strategy.By using this strategy to obtain the data, the LOD and LOQ parameters were estimated based on the standard deviation (SD) of the response, as well as the slope of the calibration curve of the target analyte (Mo 4+ ions).The response (R) was calculated using the formula: where I 0 represents the initial intensity (from each N-CQD), while I 1 and I 2 denote the fluorescence intensity of CQD-NH 2 and CQD-CONH 2 , respectively.R was plotted as a function of the concentration of the Mo 4+ ions, and the data were fitted using a regression line method to obtain the SD of the y-intercept and S values.The LOD and LOQ parameters were found to be 1.28 and 3.88 ppm, respectively, which are smaller than the ones calculated using each N-CQD separately.For instance, the LOD and LOQ were estimated by utilizing the intensity from the CQD-NH 2 , and were calculated to be 3.25 and 9.86 ppm, respectively.This confirms our initial hypothesis that the use of both N-CQDs may increase the sensitivity of the multichannel sensor, compared to the same strategy using a single channel.Furthermore, based on the values of LOD and LOQ and the detection range used, the multichannel sensor strategy developed in this work can be easily used to detect Mo entities, since some molybdenum detection kits are already on the market with similar analytical sensibility.For instance, the MQuant ® , a molybdenum test from Merck (Darmstadt, Germany), is used to sense Mo in the range of 5-250 ppm.
Another example is the Macherey-Nagel™ Quantofix™ from Thermo Fischer Scientific (Waltham, MA, USA), which estimates the Mo concentration in a similar range.
As can be seen, CQD-NH 2 and CQD-CONH 2 were able to produce differential binding events regarding the detection of Mo 4+ ions in comparison to the other metallic cations.Therefore, in order to go further on challenging the multichannel sensor strategy, the differentiation was also tested.In this sense, the dual-response sensor array yielded a training matrix of 2 "nose"-based receptors × 7 metals ions × 8 concentrations × 6 replicates, which was firstly analyzed through K-Means and PCA.Both algorithms are unsupervised, and the whole matrix is treated to find patterns with no groups identified.Therefore, the analysis of the data matrix can confirm if the multichannel sensor strategy is able to discriminate Mo 4+ ions from the other metallic cations.
sense of this cation.However, to take advantage of the multichannel sensor, the interaction of the Mo 4+ ions with the two N-CQDs was used.Herein, the difference between KSV and n values calculated through the titration studies demonstrates the diversity of interactions on the CQDs surface.So, this that the sensibility of the methodology may increase, taking into consideration the interaction of the target ions with CQD-NH2 and CQD-CONH2.In this sense, Figure 5 displays a schematic illustration of the dual-response sensor array strategy.By using this strategy to obtain the data, the LOD and LOQ parameters were estimated based on the standard deviation () of the response, as well as the slope of the calibration curve of the target analyte (Mo 4+ ions).The response () was calculated using the formula:

𝑅 = ( 𝐼
where  0 represents the initial intensity (from each N-CQD), while  1 and  2 denote the fluorescence intensity of CQD-NH2 and CQD-CONH2, respectively. was plotted as a function of the concentration of the Mo 4+ ions, and the data were fitted using a regression Following the preprocessing and normalization of the dataset, the elbow method was applied to determine the most efficient number of clusters for dividing the dataset.The results, as plotted in Figure 6a, indicate a significant change in the slope when the number of clusters reaches three.This suggests that the algorithm can differentiate between three distinct types of metals out of the seven analyzed in the study.Based on the previous analysis, the decision was made to employ the K-means algorithm with three clusters, running the method 100 times with randomly selected centers.The distribution of cluster sizes was 6-6-30.As summarized in Figure 6b, the analysis reveals a 100% accuracy rate in detecting Mo and an 83% success rate for identifying Zn.The procedure proved to be incapable of distinguishing between the other metals.The differentiation of the metal ions was further investigated through a discriminant analysis.The results are displayed in Figure 7.
The canonical score plots shown in Figure 7 evidence the ability of our multichannel sensor approach to easily discriminate metallic Mo 4+ ions out of seven different metal ions.However, each ion has some specific characteristics.So, the behavior of each one on the N-CQDs surface is expected to be different, and this could influence the position of the clusters in the canonical score plots.In this regard, the position of the clusters related to the Mo 4+ ions can be readily identified in Figure 7b-d.This reveals the potential of CQD-NH 2 and CQD-CONH 2 to create a unique fluorescent response to perform the identification of Mo species.Interestingly, Zn 2+ ions were also identified.Therefore, the DA analysis corroborated the K-Means analysis, which reveals the optimization through three clusters of classification.This demonstrates the robustness of the statistical treatment used in this work and, once again, confirms that our multichannel sensor is able to differentiate Mo 4+ and Zn 2+ ions.This may be explained based on the titration studies, since they demonstrate that the presence of a distinct number of surface sites was found.Mo 4+ ions were estimated to have 1.80 ± 0.17 and 1.96 ± 0.15 surface sites on CQD-NH 2 and CQD-CONH 2 , respectively.For Zn 2+ ions, CQD-NH 2 had an 0.66 ± 0.02 surface site, while CQD-CONH 2 was found to have an 0.97 ± 0.03 surface site.Both CQDs presented n values near to 1.0, which may explain the poor discrimination.According to these results, the interactions between the CQD's surface and Fe 2+ , Zn 2+ , Mg 2+ , Mn 2 , and Co 2+ ions are similar, which did not allow our multichannel sensor to differentiate between them.The CQD-NH 2 and CQD-CONH 2 were able to produce three different behaviors on their surface, regarding the seven target analytes investigated.In addition, the concentration level did not increase the identification ability towards Fe 2+ , Zn 2+ , Mg 2+ , Mn 2+ , and Co 2+ ions.However, the Mo 4+ ions are easily identified at 9.90, 14.78, and 19.6 ppm, since the cluster from this metallic cation is separated from the other ones, as can be seen in Figure 7b-d, respectively.These results pointed out the great potential of the methodology to discriminate Mo 4+ ions.The discrimination power of our "nose"-based receptors may rely on the interaction between the functional groups on the N-CQD's surface and each ion and its ability to produce differential binding events.So, each metallic cation interacts with both CQD-NH 2 and CQD-CONH 2 in order to generate a fingerprint-like pattern.Once the receptors are the same, the characteristics of the ions are also important to obtain a unique interaction.In this regard, we believe that the discrimination of our methodology towards Mo 4+ ions may be explained based on the electronic properties of the ion, since the electronic configuration of this cation presents a greater number of empty d-orbitals, which favors the interaction with the electron-rich surface of the N-CQDs.It is important to highlight that the proposed sensing strategy has limitations.In the future, other optimization studies should be carried out, as suggested in Figure 5.Samples containing humic, fulvic acids, or proteins require prior treatment, as such substances can interfere with the analysis process.

Conclusions
In summary, we present a simple and easy strategy for building up a multichannel sensor approach.In this regard, the nose-based multichannel sensor allowed for the development of a methodology to detect Mo 4+ ions.Interestingly, the titration curves at different temperatures revealed a sensing mechanism through static quenching Mo 4+ ions, which indicates the formation of a non-fluorescent complex on the surface of the CQD-NH 2 and CQD-CONH 2 .Furthermore, the dataset obtained from the titration studies were used to build up a training matrix to carry out a statistical treatment through K-Means and Discriminant analysis to check the identification of the Mo species among the other ions.Mo 4+ ions were easily identified at 5 ppm with the "nose"-based probes.The presence of two empty sites on the surface of the N-CQDs was a key factor to explain this result.Along with this, the presence of a greater number of empty d-orbitals in the electronic structure of the Mo 4+ ions is another important feature, since it allows for a strong interaction of this cation with the N-CQDs surface.Overall, the methodology to detect Mo 4+ ions based on our multichannel sensor was found to have 1.28 and 3.88 ppm for LOD and LOQ parameters, respectively.These values are comparable to the molybdenum test already available on the market, such as MQuant ® (Merck) and Macherey-Nagel™ Quantofix™ (Thermo Fischer Scientific).This demonstrates the potential of the multichannel sensor strategy used in this work to carry out the detection of Mo species.
Figure 4b-h displays the fluorescence responses of CQD mixtures against these seven metal ions at different concentrations in the range of 100-700 μmol.L −1 .

Figure 6 .
Figure 6.Training matrix statistical treatment: (a) Elbow Method for the complete dataset, and (b) Facet wrap of the clustering process for metal detection, using the multichannel sensor strategy.The variables 1 and 2 are arbitrary and both collected from the complete dataset.Also, groups 1, 2, and 3 are associated with Mo, Zn, and remaining metals, respectively.

Figure 7 .
Figure 7. Canonical score plot of CQDs-N and CQ-CONH2 for the discrimination of seven metal ions at about (a) 4.98, (b) 9.90, (c) 14.78, and (d) 19.6 ppm.Each metal group is composed of six replicates.However, a boxplot methodology was applied to exclude outliers.

Figure 6 . 17 Figure 6 .
Figure 6.Training matrix statistical treatment: (a) Elbow Method for the complete dataset, and (b) Facet wrap of the clustering process for metal detection, using the multichannel sensor strategy.The variables 1 and 2 are arbitrary and both collected from the complete dataset.Also, groups 1, 2, and 3 are associated with Mo, Zn, and remaining metals, respectively.

Figure 7 .
Figure 7. Canonical score plot of CQDs-N and CQ-CONH2 for the discrimination of seven metal ions at about (a) 4.98, (b) 9.90, (c) 14.78, and (d) 19.6 ppm.Each metal group is composed of six replicates.However, a boxplot methodology was applied to exclude outliers.

Figure 7 .
Figure 7. Canonical score plot of CQDs-N and CQ-CONH 2 for the discrimination of seven metal ions at about (a) 4.98, (b) 9.90, (c) 14.78, and (d) 19.6 ppm.Each metal group is composed of six replicates.However, a boxplot methodology was applied to exclude outliers.

Table 1 .
Values of λ exc , R 2 , K SV , and n for each titration carried out at 298.15 K.