Supporting Information Synthesis and characterization of Mono-disperse Carbon Quantum Dots from Fennel Seeds: Photoluminescence analysis using Machine Learning

Herein, we present the synthesis of mono-dispersed C-QDs via single-step thermal decomposition process using the fennel seeds (Foeniculum vulgare). As synthesized C-QDs have excellent colloidal, photo-stability, environmental stability (pH) and do not require any additional surface passivation step to improve the fluorescence. The C-QDs show excellent PL activity and excitation-independent emission. Synthesis of excitation-independent C-QDs, to the best of our knowledge, using natural carbon source via pyrolysis process has never been achieved before. The effect of reaction time and temperature on pyrolysis provides insight into the synthesis of C-QDs. We used Machine-learning techniques (ML) such as PCA, MCR-ALS, and NMF-ARD-SO in order to provide a plausible explanation for the origin of the PL mechanism of as-synthesized C-QDs. ML techniques are capable of handling and analyzing the large PL data-set, and institutively recommend the best excitation wavelength for PL analysis. Mono-disperse C-QDs are highly desirable and have a range of potential applications in bio-sensing, cellular imaging, LED, solar cell, supercapacitor, printing, and sensors.


Figure S4
7. Optimization of the pyrolysis process parameters for the synthesis of C-QDs.

Figure S5
8. XPS analysis of as-synthesized C-QDs Table S1-Wide scan of as-synthesized C-QDs Table S2-Peak fitting results obtained after deconvolution of carbon peak Table S3-Peak fitting results obtained after deconvolution of oxygen peak 9. Normalized PL emission spectra of C-QDs.  11. Effect of pH on PL of as synthesized C-QDs (Environmental stability).

Figure S8
12. PCA analysis of as synthesized C-QDs.  Photoluminescence Data structure: PL data was acquired using the excitation at 200, 220, 240, 260, 280, 300, 320 and 340 nm, respectively for the following pH 3, 5, 7, 9, 11, 13. One spectrum, i.e., intensity vs. wavelength is stored in one-dimensional vector of length (n, the number of data point acquired in spectral range 1: n). Consequently, all the spectra was put together to form the two dimensional (2D) matrix; i.e., D (48 x n) of size (samples x n) containing the entire PL data set. Each row in the 2D data matrix represents one spectrum. Since, the data have good signal to noise ratio, no additional data processing was done. analysis of D matrix resulted to two-sub matrix of S (score) and L (loading) using NIPAL algorithm [1].

D = CS T
where D, S T , C and E are the data set (PL spectra), loading (spectral profile), and score matrix, respectively.

(ii) Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS): MCR is useful for
resolving the spectroscopic data featuring peaks that contain relevant spectral profile (S T ), even estimating the concentration (C) from mixture spectra [2]. Since, the PL spectrum of as synthesized C-QDs is quite broad, MCR-ALS was performed to extract the possibility of the other peaks. The MCR-ALS with non-negative constrained was performed using MCR-ALS tool [2]. The major hindrance for the MCR-ALS is that ones must know the expected number of components in advance to optimize the MCR-ALS regression. An incorrect choice can lead to overestimation (inclusion of noise) and underestimation (loss of information). PCA was used to choose the number of components with objectivity. The number of components using PCA was found to be two.   Figure S2. Higher resolution TEM image of C-QDs shows that C-QDs have crystalline structure and the distance between the lattices fringes is 0.21 nm assigned to (100) plane.  Figure S5. Symmetric and Asymmetric peaks of CH2 as an indicator to optimize the pyrolysis process parameters (sensor to distinguish the insufficient carbonization). FTIR spectra taken from a typical C-QDs synthesis trial at 300 ºC for 3 hours. The presence of CH2 peaks in FTIR at 2921 and 2851 cm -1 shows incomplete pyrolysis of ground fennel powder.  Figure S6. Normalized PL emission spectra of C-QDs excited at various wavelength (240 -340 nm).

XPS analysis of as-synthesized C-QDs
PL emission spectrum of C-QDs was independent to the excitation wavelength (no redshift was observed).

Figure S8.
Effect of pH on PL of as synthesized C-QDs (acidic to basic pH). PL of carbon quantum dots (Normalized) in strong acidic (pH 3) and basic media (pH 13) shifts towards the shorter and longer wavelength, respectively. However, the shift is not very significant as can be seen in the zoomed image of PL (excited wavelength 260 nm).