Conclusions
This dissertation features an investigation of inorganic luminescent materials, activated by transition metal or lanthanide impurities. These materials, which are in this context referred to as phosphors, form a key building block of white LEDs, a technology that has already revolutionized, and will continue to revolutionize, electric lighting in terms of functionality, design and consumption. LED technology is not limited to lighting, it improves display technology in terms of color gamut, contrast
and user experience. In this work, phosphors are developed, investigated and optimized for either lighting or display applications, the former requiring broadband emission, guaranteeing a pleasant light source providing a good color rendering,
the latter requiring narrow emission, allowing displays with a superior color range.
Next to this spectral condition, five other requirements have to be fulfilled simultaneously before a material can be considered as a good candidate for applications.
These are described in chapter 1.
Given the different requirements that have to be validated for every candidate-material and the huge number of possible combinations between the available activator ions and thinkable host materials, an urge has originated for a thoughtful engineering of materials. Two different strategies are explored in this dissertation, i.e. computational methods, calculating properties of luminescent materials in different mathematical formalisms and rules, and the mining of scientific literature in the hope to
excavate the luminescent material one was searching for. Both strategies require few or no experimental input.
Computational methods can be rooted in a strong theoretical framework, being constructed by the careful inspection of empirically found trends and correlations or find itself into the spacious gray zone in between both extremes. No matter which computational technique is applied, it is of great importance to know exactly which assumptions are made, implicit or explicit, their impact on the predicted physical properties and the uncertainties or systematic errors that are to be expected. Energy level schemes are the computational tools that stand out to describe luminescent
properties and happen to be intrinsic quantum mechanical concepts. For this reason, it is endeavored in the first chapters of this work to give a detailed description of how different luminescence phenomena, be them desirable or undesirable for applications, can be understood in terms of quantum mechanical theory. Chapter 2 offers the theoretical basis by illuminating those aspects of the theory of light, matter and their interaction to understand spectroscopic experiments with phosphors.
Chapter 3 builds on this by focusing on the matter part. Very useful approximations are introduced, allowing to consider the nuclear and electronic motion separately to a certain extent. The nuclear motion governs the microscopic dynamics that are associated with luminescent properties such as spectral shapes, Stokes shifts, temperature-dependence and non-radiative decay of excited states, while the electronic motion is responsible for the luminescent transitions themselves.
To describe the electronic motion which is associated with luminescent transitions,
two different types of energy level schemes exist, i.e. single-particle and manyparticle schemes. Whatever theoretical framework or computational technique one uses, the resulting energy level scheme will always be one of these two. Chapters 4 and 5 discuss respectively the relevance of many-particle and single-particle energy level schemes in the study of luminescent materials.
Crystal field theory is used to explain the use of many-particle or multiplet schemes in chapter 4. In simplified terms, this theory starts from the atomic states of the activator ion and regards the host crystal as a perturbation. A Python program was developed, allowing crystal field calculations for the electron configuration that are relevant in the study of phosphors. This chapter is rather extensive, allowing to cast a look under the hood of the Python code and appreciating the assumptions and limitations of crystal field theory. The theory is illustrated and the developed program validated by applying it to the emblematic Dieke diagram and the Tanabe-Sugano diagrams for lanthanide and transition metal ions respectively. Both diagrams can be considered as historical landmarks. Of more recent relevance are the excited 4fN−15d1 configurations of lanthanide ions which are essential in many LED phosphors. Many-particle schemes are constructed for the currently relevant LED phosphor K2SiF6:Mn4+, explaining its peculiar decay behavior and to the well-known
afterglow material SrAl2O4:Eu2+, explaining the origin of a somewhat mysterious blue emission which is only present at low temperatures.
In chapter 5, band theory is explained as an example of a single-particle theory. It is addressed how defects alter the properties in a perfect crystal structure by introducing discrete levels into the band gap of the host material. For certain types of impurities, these impurity single-particle levels give a good idea of the luminescence properties, while for the activators that are of interest in this work, the singleparticle description intrinsically fails due to electron correlation. In order to stick to the single-electron picture, a generalization of impurity levels, called charge-state transition levels, is introduced, attaching meaning to impurity levels of highly correlated activators. Density functional theory is discussed with the goal to calculate charge-state transition levels, while illuminating the limitations of the technique and focusing on the meaning of the obtained results. Finally, luminescent transitions involving single-particle states of both the host and activator ion are discussed. During the last 40 years, several empirical methods and relationships were devised for constructing single-particle schemes, containing the charge-state transition levels of lanthanide defects in wide band gap solids. Chapter 6 reviews these empirical rules and explains the notable systematic in lanthanide spectroscopy. Up to now, an unbiased assessment of the accuracy of the obtained values of the calculated parameters is still lacking to a large extent. To address this issue, error margins for calculated electronic and optical properties are deduced. It is found that optical transitions can be predicted within an acceptable error margin, while the description of phenomena involving conduction band states is limited to qualitative interpretation. This is due to the large error margins for physical observables, such as thermal quenching temperature, corresponding to standard deviations in
the range 0.3-0.5 eV for the relevant energy differences. As an example, the electronic structure of CaGa2S4:LnQ+ is determined, taking the experimental spectra of LnQ+ = Ce3+, Eu2+ and Tm3+ as input. Two different approaches to obtain the shape of the zig-zag curves connecting the 4f levels of the different lanthanides are explored and compared. When these empirical rules are applied, it is implicitly assumed that all lanthanide ions form isostructural defects. However, in practice,
multiple nonequivalent defects related to the same lanthanide can occur or different lanthanides can even incorporate in different ways. The consequences of these complications on the impurity energy levels are discussed. It seems that small structural differences around the lanthanide dopant can give rise to important spectral differences in its emission. These are not always clearly reproduced by the charge-state transition level schemes. Improvements to the existing procedure are suggested, potentially decreasing the uncertainties, which are then applied to the lanthanide ions
in the host crystals SrAl2O4, Sr2Si5N8 and SrGa2S4. In the second part of this dissertation, multiple phosphors are selected based on reports in scientific literature describing promising luminescent properties. In these chapters, it is described how the phosphors are prepared and it is validated to which extent the technological requirements are fulfilled. To investigate the quantum efficiency of the phosphors in a quantitative way, next to the more traditional experimental techniques, a setup with an integrating sphere was designed, purchased and characterized (see chapter 7).
The luminescence properties of the blue emitting phosphor Sr0.25Ba0.75Si2O2N2:Eu2+
are extensively investigated and compared to other members of the europium doped
MSi2O2N2 oxonitridosilicates in chapter 8. This phosphor features strong 4f65d1 ↔4f7
luminescence originating from the Eu2+ ion, with a narrow emission band peaking at 467 nm and a full width at half maximum (FWHM) of only 41 nm. Thermal quenching of the blue luminescence only sets in above 450 K, making this material an interesting candidate as LED conversion phosphor. The fast decay of the luminescence prevents the phosphor to be susceptible to saturation effects at high
excitation fluxes. Furthermore it is proven to be chemically stable against moisture.
The only drawback is the relatively low quantum efficiency of the synthesized powder, provisionally preventing this material to be used in applications. In addition, the phosphor features a weak yellow emission band, originating from small domains featuring a different crystal structure. It is shown that the majority of the powder grains only exhibit blue emission. Finally, the spectrum of a white LED, based on a UV pumping LED and three (oxo)nitridosilicate phosphors is simulated in order to assess the usefulness of blue phosphors in LEDs for lighting. Only a marginal improvement in terms of color quality can be achieved with a narrow banded phosphor, at the expense of a decrease in luminous efficacy and overall electrical to optical power efficiency.
Subsequently, in chapter 9, the interesting class of thiogallate and thioaluminate host materials is considered. From the general overview of their properties upon doping with divalent europium, two thiogallates, SrGa2S4 and ZnGa2S4 are selected for further investigation. The luminescent properties of Sr1−xEuxGa2S4 phosphors are studied over a wide dopant concentration range (x = 0.01-0.3) as function of temperature. The phosphors show a saturated green emission over the entire studied
range, with a typical peak wavelength around 536 nm and a FWHM of 50 nm. The internal quantum efficiency is 71% for x = 0.04. For this concentration, the emission intensity at 400 K is still 90% of the intensity at room temperature. By measuring both decay and thermal quenching profiles as a function of europium concentration, the emission properties can be explained on the basis of the local environment of the europium ions in the lattice. As SrGa2S4:Eu2+ achieves a very good score with respect to the technological requirements, a fully optimized powder is used to deliver a proof-of-concept remote phosphor white LED, suitable for display applications.
The red component is provided by CdSe/CdS quantum dots and different remote hybrid phosphor layers are prepared, featuring different stacking geometries of the green and red components. The optimized white LEDs show favorable properties such as internal quantum efficiencies in the 75-80% range, high luminous efficacies and saturated primary colors. The different stacking geometries provide a means to select the most cost-efficient layer design given the relative cost of the green powder and red quantum dot components.
The second europium doped thiogallate which is studied is ZnGa2S4:Eu2+. This material has been reported as a saturated green emitting phosphor, suitable as conversion phosphor in white LEDs for lighting or displays. No direct proof for the incorporation of Eu2+ in ZnGa2S4 has however been given. Here, X-ray diffraction (XRD), cathodoluminescence in electron microscopy (SEM-CL) and X-ray absorption spectroscopy (XAS) are combined to study the incorporation of the europium
ions in the host material. The previously reported green luminescence was found to originate from small amounts of unintentionally formed EuGa2S4, and not from europium ions incorporated into ZnGa2S4. EuGa2S4 has a low quantum efficiency (<20%) and shows strong thermal quenching, already below room temperature. The XAS data analysis suggests that a certain amount of europium might occupy octahedral voids inside the zinc thiogallate lattice in a divalent state. The zinc ion next to these interstitial dopants is then removed for charge compensation. Notwithstanding the possible, but limited, incorporation of Eu2+ in ZnGa2S4, these ions do not activate any luminescence as was shown with SEM-CL.The final chapter, chapter 10, features a combined experimental-theoretical study of the luminescent material CaZnOS:Mn2+. This compound features orange broadband luminescence, peaking at 612 nm, originating from intraconfigurational 3d5 transitions within the Mn2+ ion. DFT calculations at PBE+U level and X-ray absorption spectroscopy indicate that the Mn impurity is incorporated on the Zn site in a divalent charge state. The electronic structure of the MnZn defects is obtained by two complementary techniques. On the one hand, impurity levels in the band diagram, i.e. the single-particle energy level scheme, are obtained from defect formation energies calculated with PBE+U. On the other hand, the excited state landscape of the Mn2+ 3d5
electron configuration is assessed through the spin-correlated crystal field, yielding the multiplets, i.e. the many-particle energy level scheme, of the optical dopant. Experimental photoluminescence spectra at room and low temperature are analyzed in detail and a good correspondence is found between the calculated energy levels and the experimental transition energies. The electron-phonon interaction is investigated from the luminescence spectra showing that at least three different vibrational modes are active in the transition. These are also found in the Mn-projected phonon density of states. This case study demonstrates how physical information can be extracted from the two complementary, but different types of energy level schemes. The CaZnOS:Mn2+ phosphor is finally evaluated for the use as red phosphor in white LEDs. Despite its favorable emission spectrum, the forbidden nature of the intraconfigurational 3d5 transitions yield a much too low external
quantum efficiency.
Perspectives
It is clear that white LEDs, and the conversion phosphors being an essential part of it, will remain relevant technologies in the future and will continue to consolidate their increasing market share in lighting and display markets. Although an incredible advantage has already been achieved on the phosphor point-of-view, going from cold white YAG:Ce3+-based LEDs towards two- or three-phosphor LEDs with a wellbalanced color, incremental improvements can still be expected in the future.
In this work, phosphors based on different types of activator ions were investigated,
largely inspired by the unstable prices of lanthanides precursors. Straightforward alternatives for the dipole allowed 4fN ↔ 4fN−15d1 transitions of Ce3+ and Eu2+ are the dipole forbidden intraconfigurational 4fN or 3dN transitions of respectively lanthanides and transition metals. A huge drawback of both transitions is their forbidden nature, prohibiting an efficient excitation with blue pump light, making it difficult to achieve the required values for the absorption around 80%. The external quantum efficiency is hence limited by the selection rules. Sensitizing these ions by
another absorbing ion is a straightforward solution, potentially showing high internal and external quantum efficiency, at least in theory. Up to now, no convincing lanthanide-free system exploiting this strategy can be found. An alternative from this series which is possibly viable is the tetravalent Mn ion, Mn4+. Recent reports indicate that this ion seems to suffer less from a low external quantum efficiency than its fellow 3d ions. Future research, and specifically critical feasibility studies, have to be performed to unambiguously assess whether this small ion can dethrone the dominant lanthanide ions. Until then, the ultimately color-tunable divalent europium ion, Eu2+ is still the king.
Regarding energy level modeling, novel empirical rules can be expected, complementing the existing empirical rules for lanthanide, transition metal or s2 ions.
These rules might relate properties of ions of different groups or limit the required experimental input even further. Despite that, a rational footnote should however be made that the most obvious empirical rules have probably already been found.
Critically scrutinizing the data used to construct the existing rules might decrease the uncertainty of predictions to some extent, although no wonders can be expected from this angle.
From the quantum mechanical side, progress is to be expected especially from the high-level-of-theory side. The limitations pertaining to computational resources become less problematic due to rapidly increasing computer power. A large-scale computational screening of dopant-host combinations can hence be expected in the near to mid-long term. A fundamental restriction of density functional theory remains however the inability to describe excited states. Possibly more important, the most complete computational techniques, such as multireference calculations, will become more standard to interpret spectroscopic experiments, potentially leading to new insights into functional materials.
In general, despite what is often claimed, designing novel functional materials with well-defined properties, solely from the computational drawing board is currently still far away. The reason for this is not only the barrier of computational cost, but also the error margins of computational results which are still largely exceeding the stringent restrictions by technological requirements. This pertains to essentially all computational techniques with predictive potential. The distance in energy between blue and red light is after all only 850 meV, a tiny energy interval which contains all
colors of the rainbow!