Amorphous Carbon Dots and their Remarkable Ability to Detect 2,4,6-Trinitrophenol

Apparently mundane, amorphous nanostructures of carbon have optical properties which are as exotic as their crystalline counterparts. In this work we demonstrate a simple and inexpensive mechano-chemical method to prepare bulk quantities of self-passivated, amorphous carbon dots. Like the graphene quantum dots, the water soluble, amorphous carbon dots too, exhibit excitation-dependent photoluminescence with very high quantum yield (~40%). The origin and nature of luminescence in these high entropy nanostructures are well understood in terms of the abundant surface traps. The photoluminescence property of these carbon dots is exploited to detect trace amounts of the nitro-aromatic explosive — 2,4,6-trinitrophenol (TNP). The benign nanostructures can selectively detect TNP over a wide range of concentrations (0.5 to 200 µM) simply by visual inspection, with a detection limit of 0.2 µM, and consequently outperform nearly all reported TNP sensor materials.

human health problems 20,21 . The availability and detrimental potential of TNP, makes it an important candidate to be detected. The existing methods for detection of such compounds are spectrophotometry 22 , gaschromatography 23 , capillary electrophoresis 24 , and high performance liquid chromatography 25 . Most of these techniques are expensive and require sophisticated instruments which cannot be easily employed. Consequently, there is an ongoing effort to develop materials and techniques for detecting TNP in a simple and cost effective manner.
In 2013, Nagarkar et al. 26 developed a cadmium based, 3D fluorescent metal-organic framework (MOF) for sensitive (up to 4 µM) and selective detection of TNP. In the following year Dinda et al. 20 reported even better performance (selective detection of TNP up to 0.5 µM) by 2,6-diamino pyridine functionalized graphene oxide. Subsequently, Li et al. 27 used blue fluorescent graphene quantum dots to develop a sensor for the analysis of TNP in water samples in the concentration range of 0.1-15 μM. Na et al. 28 demonstrated that lysozyme-capped CdS quantum dots are able to detect TNP in the range of 0.5-15 μM with a detection limit of 0.1 μM. More recently, Wang et al. 29 developed a Cd based MOF and utilized it as a chemosensor to detect various nitro-compounds including TNP. A larger detection range of TNP (10-600 µM, with a detection limit of 2 µM) using amine functionalized CDs has also been very recently reported by Campos et al. 30 . In addition to the fluorescence based methods other techniques have also been employed for sensitive detection of picric acid. Pal et al. 31 used crystalline carbon dots and polypyrrole based composite films for electrical detection of TNP and achieved a detection limit of 0.14 µM.
It follows from the above discussion that some sensitive and selective techniques for detection of TNP are available. However, a major problem with some of the best techniques is the toxicity of nanomaterials. Health hazards of the Cd and other heavy metals present in MOFs are well documented 26,32 . A benign alternative based on CDs is therefore much preferred over the MOFs. In this regard, crystalline CDs and functionalized quantum dots of carbon have already shown substantial promise. Here, we demonstrate that apparently mundane, amorphous CDs having very high PL yield are excellent candidates since their synthetic strategy can result in low-cost, high output material suitable for industrial production. Additionally, we show that the amorphous CDs can detect over a much wider range (up to 200 µM) of TNP concentrations with a detection limit of 0.2 µM, which perhaps is the best performance reported so far (Supplementary Information).

Results and Discussion
Ultrasonication of a mixture of dextrose and HCL followed by centrifugation and drying produces well dispersed CDs as evidenced from the high resolution transmission electron microscopy (HR-TEM) image shown in Fig. 1a. Due to the low contrast between CDs and carbon coated copper grids, bright field images of amorphous carbon based samples are seldom distinct. Nevertheless, the clear dark spots seen in the image are indicative of the formation of localized clusters of carbon atoms. The selected area electron diffraction pattern (SAEDP), shown as top inset, Fig. 1a, consists of a diffused halo devoid of any spots or rings, which is typical of an amorphous sample. These apparently amorphous clusters are nearly spherical with diameters varying from ~4-18 nm as shown in the histogram (bottom inset, Fig. 1a). No signature of fringes is present in the high magnification image shown in Fig. 1b, which captures one isolated CD. The dominant amorphous character of the CDs is also reflected in the broad hump centered around ~2 26 θ =° in the x-ray diffraction (XRD) profile shown in Fig. 1c. Such a hump has been widely observed in XRD patterns of amorphous carbon 35 . Elemental maps of C, H and O obtained by energy filtered TEM (EF-TEM) measurements are shown in Fig. 1d. The EF-TEM images clearly reveal the clustering of carbon atoms which is manifested by a brighter contrast in the elemental C-map. The maps of the other two elements -H and O, show their distribution, which appears to be concentrated on the surface surrounding the carbon cluster. A cluster of carbon atoms surrounded by H and O is clear from the composite mapping of all three elements, shown in the extreme right panel of Fig. 1d. Therefore, the micro-structural features of the synthesized CDs suggest that the one-step method leads to the formation of amorphous clusters of well-dispersed carbon nanostructures surrounded by O and H.
In the last few years a significant amount of work has been dedicated to understand the mechanism of formation of CDs 16,36,37 . Nearly all the proposed mechanisms of bottom-up synthesis hinge of the following steps: (i) pyrolysis of carbon rich precursors at elevated temperatures; which leads to (ii) carbonization and nucleation; followed or accompanied by (iii) surface passivation with stabilizing agents. Tang et al. 36 suggested that any carbohydrate containing C, H, and O in the ratio of 1:2:1, where H and O exist in a form that allows dehydration under hydrothermal conditions, can be used for preparation of CDs. We show that carbonization of carbohydrates can be achieved through simple ultrasonic agitation at room temperature. As shown in the Fig. 2, simple mechanical agitation leads to carbonization followed by nucleation of carbon. The growth of amorphous clusters of carbon is assisted by diffusion of other molecules towards the particle surface resulting in the formation of CDs which are self-passivated with H and O related functional groups.
Surface passivation and functionalization of CDs are of paramount importance since fluorescence from these nanostructures is believed to be largely linked to the surface character 38 . Effective surface passivation is considered to be an essential step in order to produce CDs with high fluorescence yields 39 . We have nonetheless obtained highly luminescent CDs without any additional step involving external surface passivation.
In an attempt to form an idea of the bonds present in the synthesized CDs, Fourier transform Infra-red (FTIR) spectroscopy was performed and is shown in Fig. 3. The FTIR spectrum displays several peaks related to the surface bonds of CDs. The absorption band spanning from 3640 to 3160 cm −1 correspond to stretching vibrations of the OH bond 40 . The presence of such bonds is possibly responsible for imparting hydrophilicity and consequent water dispersibility of the CDs. The bands peaking at 1760 cm −1 , 1680 cm −1 and 1150 cm −1 may be assigned to the stretching vibrations of C=O, C=C and C-O, respectively 41 , suggesting once again that the surface of carbon clusters are passivated by surface groups that are spontaneously derived out of the carbonization process of dextrose.
X-ray photoelectron spectroscopy (XPS) measurements were carried out to further investigate the composition and surface groups of the CDs. Two strong peaks at 285.5 eV and 532.5 eV are detected in the wide-scan XPS spectrum as shown in Fig. 4a, which are attributed to oxygen and carbon, respectively 42 . The de-convoluted C1s spectrum (Fig. 4b) has four components, corresponding to C=C (sp 2 carbon) at 283.2 eV, C-C (sp 3 carbon) at 284.7 eV, C-OR and COOR at 285.4 eV and 289.5 eV, respectively [42][43][44] . The measured O1s spectrum (Fig. 4c) can be de-convoluted to three components peaking at 529.6, 531.1 and 532 eV, which are due to the C=O, C-OH and C-O-C groups, respectively. Hence, the XPS data allows us to find out the surface composition, although the relative composition of the core of the CDs, which are evidently made of C=C, cannot be obtained. Importantly, the results of XPS and FTIR reinforce each other and affirm the proposition of spontaneous self-passivation of the CDs.
In an attempt to strengthen our understanding of the structure of the CDs, electron-energy loss spectroscopy (EELS) was performed. Figure 5a,b, show the carbon K-edge and the low-energy-loss EELS spectra of the CDs, respectively. Sufficient care was taken and the samples were tested before and after the EELS measurements to ensure protection from electron-beam induced damage and possible contamination. The carbon-K energy-loss near edge structure (ELNES) spectrum in Fig. 5a, exhibits a peak at 292 eV along with a higher energy shoulder like feature appearing at ~311 eV. The band peaking at 292 eV corresponds to 1 s → σ * transition and is consistent with ELNES spectrum of amorphous carbon 45 . A typical and distinctive feature of graphitic carbon is a sharp peak at the onset (~285 eV) of the σ * band 36,46,47 . Muller et al. 48 have shown that the absence of this pre-peak (at ~285 eV) is a feature of fully amorphous carbon or graphitic structures with sizes less than 1.5 nm. Furthermore, the shoulder appearing after σ * band is reported to be sensitive to the long range order 49 . A broad, plateau-like feature observed in the profile of our sample is due to the occurrence of damaged ring structures and absence of any long-range order 50 . Figure 5b is the low-loss EELS spectrum of the CDs. The peak at 0 eV is due to the elastically scattered (or un-scattered) electrons. The peak appearing at ~5 eV is the π plasmon peak, while the π + σ bulk plasmon peak manifests as a broad hump peaking at 21.6 eV. These results are in excellent agreement with both theoretical and experimental findings reported for amorphous carbon [51][52][53] . The approximate density of the CDs estimated from the bulk plasmon energy peak position (21.6 eV) is found to be 1.5 g/cc, which is consistent with the reported density of graphite-like amorphous carbon ( Supplementary Information) 52,54 .
Nearly all reports published so far suggest that PL from unpassivated or self-passivated CDs is usually low with a QY less than 10% 55,56 . Perhaps, the only exception is a work reported by Shen et al. 57 where they showed that hydrothermally prepared crystalline CDs may have QYs as high as 40.5%, if they are treated with NaBH 4 , so as to dramatically increase surface defects. However, such studies on correlation of surface treatments and luminescence are primarily based on crystalline CDs 58,59 . In amorphous nanostructures empty or densely packed local atomic defects are abundant. Additionally, the high-entropy disordered nature of the entire structure produces multiple luminescence centers which contribute to intense PL. Despite enormous efforts in understanding the mechanism of luminescence from carbon nanostructures, a consensus about the exact mechanism is still awaited. Earlier reports in this regard focused primarily on quantum size effects and provided evidences in support of quantum confinement 60,61 . Later studies have revealed that emissive traps and surface groups are primarily responsible for bright PL from carbon nanosystems. The fact that both crystalline and amorphous CDs exhibit very similar PL characteristics indicates that defects/traps related emissive centers and surface states play dominant roles in light emission from CDs.
The PL and the UV-visible absorption profiles of the synthesized CDs are shown in Fig. 6a. The shoulders appearing at 275 nm and 335 nm in the absorption spectrum are typical of colloidal CDs absorption characteristics 39,62 . The origin of these features are related to π electron transitions in C=C and oxygen-containing functional groups, respectively. The shoulder at ~275 nm is probably due to π-π* transition of C=C and the absorption hump at ~335 nm possibly corresponds to n-π* transition of the C=O bond. The PL spectra on the other hand show that the dominant emissions from the CDs are in the blue-green region, which can be tuned by changing the excitation energy as shown in Fig. 6b. The photographs of the luminescent colloids exhibiting blue-green-red emissions under different excitation energies are shown as inset, Fig. 6b. It is evident that the yield is maximum in the blue-green region corresponding to UV-blue excitation and decreases with decreasing energies. The QYs for PL emission corresponding to 330 nm excitation was calculated by method proposed by Pålsson and Monkman 33 , and de Mello et al. 34 and was found to be 39.5% (Supporting Information). The intense blue-green emission is also evident from the photograph of the glowing colloids. To the best of our knowledge this is the first report of such high yield self-passivated, amorphous CDs prepared at room temperature.
Nearly all variants of carbon nanostructures exhibit excitation dependent multicolour luminescence -a feature that has been widely discussed in literature for crystalline as well as amorphous samples 12,63 . Size-controlled synthesis of CDs has also been reported in literature, but, unlike semiconductor dots, tuning of PL colour by controlling the size of CDs has been challenging 3,4 . As observed by us, the as-synthesized CDs have strong PL from blue to green. Optimal emission in longer wavelength regions, though observed, is very rare for colloidal CDs 60,64,65 . Additionally, the PL bands of CDs, as seen from Fig. 6b and as reported by several other groups [66][67][68] , is usually pretty broad, indicating inhomogeneous chemical structure and diverse PL centers. Based on the above observations we are strongly inclined to believe that PL from CDs are essentially for emissive traps and surface functional groups. The excitation dependence of PL emission arises from the fact that at a given wavelength only certain emissive sites can be excited. Increase in excitation energy correspondingly increases the number of sites since higher energy traps can now be excited. This affects a consistent blue shift of the PL maxima accompanied by an increase in intensity with excitation energy. Thus, PL spectra of amorphous CDs also reflect the distribution of emissive centers present in the sample. In addition, we believe, discretization of phonon density of states in nanostructures also contributes towards the observed excitation dependence of emission energy. It is well known   that in colloidal nanostructures the solid-liquid boundaries act as potential barriers for phonons, resulting in phonon confinement within the nanostructures 69 . The photo-excited carriers in such systems relax by emitting phonons corresponding to some discrete states. Consequently, the emission energy becomes dependent on the excitation energy. However, a detailed investigation on this particular aspect is a separate study and is being presently pursued. At this stage we think distribution of emissive trap states and discretization of phonon energies play a combined role for excitation dependence of emission from CDs.
As a potential application of these CDs, we investigated the efficacy of these nanostructures in sensing the explosive nitro-compound -2,4,6-TNP. A remarkable quenching of luminescence detectable with the unaided eye, as seen in the photographs (Fig. 7), was observed when trace amounts of TNP was added to the CD colloids. The visible bright blue luminescence vanishes immediately upon addition of TNP solution. Detection of the presence of TNP in solution, simply by visual inspection, was possible up to a concentration of 0.5 µM.
To explore the quantitative sensitivity of CDs in detecting TNP, fluorescence-quenching titrations were performed with incremental addition of TNP over a wide range of concentrations between 0.5 to 200 µM in 15 mL aqueous solution of CDs. A plot of the PL intensity profiles are shown in Fig. 8 and variation of PL maxima with different concentrations of TNP additions are shown in Figure S2 in the supplementary material. These plots, demonstrate that the variation is linear over a wide range (0.5 to 50 µM), while at higher concentrations ( > 50 µM) the intensity decreases exponentially with TNP addition (Supplementary Information).
In order to be an effective sensor, in addition to being sensitive, the material must be selective. To assay the selectivity of CDs towards TNP sensing, fluorescence quenching titrations were also performed with other nitro-compounds -4-nitrophenol (4NP), 2-nitrophenol (2NP), dinitro-benzene (DNB), nitroaniline (NAni), phenol (PH) and nitrobenzene (NB).We can see from Fig. 9a, that all the nitro-compounds affect a PL quenching, but the rate of quenching with TNP is by far the maximum. While, 80% PL quenching is affected at 60 µM of TNP addition, the closest competitors like 4NP, 4NAni and NB can cause ~30% or less quenching for the same concentration. The colloidal CDs therefore, exhibit higher selectivity towards TNP compared to other nitro-compounds investigated here.
Quantitative assessment of the quenching efficiencies of different quenchers was performed by standard Stern-Volmer (S-V) method (Supporting Information). The S-V quenching coefficient (K SV ), which is a measure  of sensitivity towards an analyte, is found to be significantly larger for TNP suggesting dominant selectivity of TNP over others as shown in Fig. 9b. The K SV values of different analytes provided in Table S2 (Supplementary Information) demonstrate that K SV of TNP is at least three times or more compared to the other nitro-aromatics.
In recent times, different materials, including many carbon based nano-systems have been investigated for detection of TNP by PL quenching [70][71][72] . Interestingly, in almost all cases, it has been reported that TNP affects a preferential quenching of luminescence. Mechanisms such as Fröster (fluorescence) resonance energy transfer (FRET), charge transfer, inner filter effect, etc. have been suggested to be responsible for the observed selectivity 73 . In this communication we will refrain from dealing with the mechanism of selective detection. We will simply note that at lower concentrations, the S-V plots of TNP are close to linear ( Figure S3a-c) but for higher concentrations, it deviates from linearity and increases almost exponentially ( Figure S3d). The linear S-V plots can be attributed to the ground state interactions between the TNP and the CDs, whereas the non-linear S-V plots of TNP suggest that quenching is mainly due to energy and/or electron transfer processes between the analyte and the CDs 74 . It is also observed ( Figure S4) that the emission band of the CDs overlaps with the absorption band of the TNP, although the overlap integral is not sufficiently large to unambiguously confirm FRET 75 . At this stage we believe, multiple mechanisms involving the surface groups are involved in selective detection of TNP by the amorphous CDs. A detailed analysis of this phenomenon warrants a separate investigation.
In summary, a simple and inexpensive method to prepare scalable quantities of amorphous CDs and their application in sensing TNP are demonstrated. Carbonization of glucose by ultrasonic waves leads to the formation self-passivated clusters of carbon atoms -the amorphous CDs. The CDs exhibit intense room temperature PL, which closely resemble the luminescent character of their crystalline counterparts. Abundant surface traps and dominant π-π* and n-π* transitions are responsible for excitation dependent PL from the CD colloids. These CDs are found to be excellent candidates for sensitive and selective detection of 2,4,6-TNP. Rapid and intense quenching of PL intensity of the CD colloid was observed due to the addition of TNP, which allowed detection of TNP simply by visual inspection, up to a concentration of 0.5 µM. In general, the CDs can selectively detect TNP over a wide range of concentrations (up to 200 µM) with a detection limit of 0.2 µM, which is one of the best performances reported so far.
Preparation of carbon dots. The preparation of amorphous CDs involved simple ultrasonication (Piezo U-Sonic) of a 1:1 (volume ratio) mixture of 1 M dextrose solution and HCl. With increasing time of ultrasonication the solution changed from colorless to brown. After 12 h of sonication the brown colored solution was oven-dried at 80 °C under ambient pressure for 24 hours to obtain a dark brown colored dry powder. This powder was dissolved in DI water and centrifuged at 12000 rpm for 15 min. The light yellowish supernatant was separated and found to contain fine dispersion of water-soluble, amorphous CDs. It was found that on an average 9 gms of dextrose produces 6.5 gms of dry CD powder implying a production yield of 72.2% (Supplementary Information).
Characterization. Investigation of crystalline property of the synthesized CDs was carried out by XRD using a Brucker D8 advanced diffractometer operating at 40 kV, using Cu Kα 1 (λ = 1.54056 Å) radiation. Structure of the samples was investigated by HR-TEM using a JEOL JEM-2200FS electron microscope equipped with a 200 kV field emission gun and in-column energy filter (Omega) along with Gatan software. The samples were five times diluted prior to deposition and 2 μl of samples were deposited on carbon coated cupper grids for HR-TEM measurements. EELS spectra were collected around the carbon K-edge to understand the chemical bonding and low-energy-loss EELS was performed to extract additional information including mass density. FTIR spectra were recorded using a Bruker, Tensor 27 FTIR spectrometer. XPS measurements were conducted on an Omicron Multiprobe (Omicron NanoTechnology Gmbh., UK) spectrometer fitted with an EA125 (Omicron) hemispherical analyzer. Monochromatic Al-K α source operated at 150 W was used and the pass energy of the analyzer was kept at 40 eV. A low-energy electron gun (SL1000, Omicron) with a large spot size was used for sample neutralization. The voltage of the electron gun was fixed at −3 V. UV-visible spectroscopy was performed using a JASCO V-750, UV-VIS spectrophotometer. Steady state PL property of different samples was investigated using a Horiba JobinYvon, Fluorolog-3 (Nanolog) spectrofluorometer (model FL3-11) fitted with a 450 W xenon lamp source, photomultiplier tube detector and single grating monochromator. During observation, entry and exit slit widths were kept at 1 nm and the integrating time was 0.5 s. PL yields of the samples were estimated by recording the PL spectra using an integrating sphere following the method proposed by Pålsson and Monkman 33 , and de Mello et al. 34 . The integrating sphere was mounted inside the Fluorolog-3 spectrofluorometer and the samples were mounted into the holder inside the sphere. The measured spectra were background corrected by recording a blank data.