Carbon dots (CDs) were first discovered in 2004 as fluorescent fragments during electrophoretic purification of single walled carbon nanotubes. Since then, a very active research community has formed around the topic of CDs. Often, CDs are compared to semiconductor quantum dots (QDs), as they can have similar dimensions, in the range of 2 – 10 nm diameter. QDs have gained tremendous interest due to their beneficial optical properties, such as very bright emission with a high quantum yield and narrow linewidth. This has led to wide-scale commercial use of QDs in various fields, such as QLED TVs. In contrast, real-world applications-related aspects of CD research are still in their infancy. CDs are however very promising nanomaterials due to their facile and low-cost synthesis and especially their negligible cytotoxicity. Unlike many QDs that often employ toxic elements as Cd, Pb and Hg, CDs are generally free from any heavy metals and mostly just comprise of carbon atoms as well as a small number of dopants such as nitrogen or sulfur. Recent years saw the advent of various companies that started to develop and market CD applications, which include CDs as in-product markers to increase traceability of chemicals or fibers, as tags to verify the authenticity of genuine products, and as security inks. Also, there are companies that aim to market CDs for agricultural use, for example to allow temporary gene expression.

Most of these prospective commercial applications use CDs as marker/sensor or as a drug carrier, exemplifying the advantages of their high biocompatibility and fluorescent stability under challenging conditions. In this regard, designing and engineering the surface of CDs to the requirements of certain applications is crucial to advance their applications. Likewise, inherent drawbacks of CDs such as lower color purity and brightness require smart approaches to enhance those important parameters.

The surface design of CDs has witnessed a noteworthy advancement through the introduction of chiral CDs, making a significant contribution to the field. In this thesis, we thoroughly examine the emerging field of chiral CDs, which represents a relatively new avenue of research dedicated to imparting CDs with chiral characteristics. Although we do not provide experimental findings regarding their production within the scope of this thesis (our work addressing this aspect is predominantly ongoing), our comprehensive literature review on this subject offers significant insights into the realm of chiral CDs.

This thesis introduces several approaches to modify the surface chemistry of CDs. A new strategy to incorporate CDs into a silsesquioxane matrix is presented, with the aim to make CDs dispersible in a wide range of solvents. Furthermore, two sensing applications with CDs are developed, where ethanol sensing based on a deep neural network that uses optical spectra of CDs is shown. Several different temperature sensing strategies based on luminescence of CDs are compared, and a multiple linear regression model is developed to significantly increase the temperature sensing accuracy using CDs as probes. CDs employed for studies in this thesis are synthesized within the solvothermal and hydrothermal approaches using precursors such as citric acid, polyhedral oligomeric silsesquioxanes and phenylenediamines.

At first, a method of incorporating CDs in a matrix material to allow the formation of stable dispersions in almost any solvent is presented. Composite nanospheres, which consist of luminescent CDs in a dielectric matrix of polyhedral oligomeric silsesquioxane (POSS) are synthesized. The POSS−CD composite structure is built from N-doped sp²-hybridized carbon domains formed by solvothermal treatment in inverse micelles of citric acid surrounded by POSS molecules. The size of the nanospheres can be tuned across a range of 20 − 60 nm by varying the precursor ratio of citric acid and POSS; their emission is centred at 460 − 480 nm with a high photoluminescence quantum yield (PLQY) of up to 50%. Furthermore, the POSS−CD nanospheres which are initially soluble in nonpolar solvents can be easily transferred into a broad range of polar and/or aprotic solvents (such as water, methanol, ethanol, dimethylsulfoxide, dimethylformamide, ethylene glycol, etc.) by partial chemical etching of the POSS matrix with tetramethylammonium hydroxide. At the final stage of the etching procedure, the initially large (several tens of nanometers) composite nanospheres transformed into small (∼5 nm) carbon nanoparticles resembling “classical” chemically synthesized CDs, which confirmed their “raisin bun”-like architecture.

With the demonstrated potential of designing CDs to the need of various applications based on their environment, this thesis establishes new sensing methods to employ CDs as sensors for a variety of environmental parameters, such as solvent composition and temperature. One disadvantage of CDs used for optical sensing is their broad emission profile, leading to less specific sensing signals and a potential overlap of their luminescence with the autofluorescence of the samples. Machine learning (ML) approaches can address these short-comings and greatly enhance sensing accuracy with CDs. To showcase the potential of this method, CDs are utilized as optical probes, while ML algorithms are applied to optimize ethanol content determination in ethanol/water mixtures as well as in alcohol-containing beverages. A simple neural network is used to understand the importance of different optical parameters on sensing, while a modular deep learning model is developed for increased generalizability towards samples with strong autofluorescence. Drawing from multiple input channels, the deep learning model is able to predict ethanol concentrations with a mean absolute error of 0.4 vol % in pure solvents. Moreover, the deep learning model has the ability to handle samples with strong autofluorescence, in turn a mean absolute error of 6.7 vol % is reached for alcoholic beverages like beers, wines, and spirits.

As a next step, temperature sensing is also demonstrated as an application for CDs. CDs with cyan, green, and orange emission are synthesized, and tested for their applicability in temperature sensing. Various luminescent sensing strategies that are commonly reported in literature for all kinds of materials are compared, such as PL intensity, PL peak broadening, peak shifting, ratiometric as well PL-lifetime based sensing. Green and cyan CDs show PL quenching with increased temperature, which is fitted with a linear approximation for temperature sensing with an accuracy of up to 2.43 K. At the same tie, PL lifetime variations of CDs are proven to be a far more accurate sensing method, reaching accuracies of up to 0.70 K. Using the amplitudes of Gaussian fits achieves a noticeable accuracy improvement for PL-intensity-based sensing, reaching 1.63 K.

Orange CDs display dual emission, with the longer wavelength emission peak undergoing a PL intensity increase with rising temperature, while the shorter wavelength emission peak experiences PL quenching. As result, a ratiometric temperature sensing approach based on red CDs is demonstrated, reaching an accuracy of 1.67 K. This method benefits from “self-calibration” of the PL intensities, which significantly reduces the impact of excitation fluctuations. Finally, a multiple linear regression sensing strategy is developed, which utilizes 10 spectral parameters that have been extracted from experimental data using exponential and Gaussian fits. With the multiple linear regression model, the accuracy is improved to 0.55 K.

Overall, this thesis offers several important contributions to the field of CDs, by showcasing the potential of designing the surface of CDs towards the requirements of specific applications. In addition, the use of ML methods to compensate for some inherent CD drawbacks is introduced, leading to significant improvements in their sensing ability.