In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science.

Exceptional molecules and materials with one or more extraordinary properties are both technologically valuable and fundamentally interesting, because they often involve new physical phenomena or new compositions that defy expectations. Historically, exceptionality has been achieved through serendipity, but recently, machine learning (ML) and automated experimentation have been widely proposed to accelerate target identification and synthesis planning. In this Perspective, we argue that the data-driven methods commonly used today are well-suited for optimization but not for the realization of new exceptional materials or molecules. Finding such outliers should be possible using ML, but only by shifting away from using traditional ML approaches that tweak the composition, crystal structure, or reaction pathway. We highlight case studies of high-Tc oxide superconductors and superhard materials to demonstrate the challenges of ML-guided discovery and discuss the limitations of automation for this task. We then provide six recommendations for the development of ML methods capable of exceptional materials discovery: (i) Avoid the tyranny of the middle and focus on extrema; (ii) When data are limited, qualitative predictions that provide direction are more valuable than interpolative accuracy; (iii) Sample what can be made and how to make it and defer optimization; (iv) Create room (and look) for the unexpected while pursuing your goal; (v) Try to fill-in-the-blanks of input and output space; (vi) Do not confuse human understanding with model interpretability. We conclude with a description of how these recommendations can be integrated into automated discovery workflows, which should enable the discovery of exceptional molecules and materials.

Twists and Puckers: Tuning Crystal Chemistry in the La(AuxGe1-x)2 Compositional Series.

The physical properties of solid-state materials are closely tied to their crystal structure, yet our understanding of how competing structural arrangements energetically compare is limited. In this work, we explore how small differences in composition affect the structure in the La(AuxGe1-x)2 series of compounds, which comprises four unique structure types between LaGe2 and LaAu2. This family includes the previously unknown AlB2-type compound with stoichiometry La(Au0.375Ge0.625)2 as well as La(Au0.25Ge0.75)2, an intergrowth of the AlB2 and ThSi2 structure types. We then study the chemical forces driving the structure changes and use phonon band structure calculations and DFT-Chemical Pressure to evaluate atomic-size effects. These calculations show that the parent AlB2 structure type is disfavored in Au-rich compounds due to soft atomic motions along the c axis. The instability of AlB2-type LaAuGe is confirmed by the presence of imaginary modes in the phonon band structure that correspond to a "puckering" of the hexagonal AlB2-type lattice, resulting in the experimentally observed LiGaGe structure type. The impact of size effects is less clear for Au-poor compositions; instead, twisting the AlB2 structure type to form the ThSi2 type opens a pseudogap at the Fermi level in the electronic density of states. This investigation demonstrates how crystal structure in solid-state materials can be compositionally tuned based on balancing size and electronics when multiple structure types are in close thermodynamic competition.

Non-destructive food-quality analysis using near-infrared luminescence from Mg3Gd2Ge3O12:Cr3.

Rapid, non-destructive food-quality analysis using near-infrared (NIR) photoluminescence spectroscopy produced by phosphor-converted light-emitting diodes (pc-LEDs) has fascinating prospects for future food-safety monitoring. However, covering the energy window for organic molecular vibrations of interest in these applications requires NIR-emitting phosphors that are highly energy-efficient with ultra-broadband photoluminescence. This remains a materials design challenge. Here, a Cr3+-substituted garnet phosphor, Mg3Gd2Ge3O12, is found to possess a desired broadband NIR emission (λem = 815 nm, fwhm = 172 nm; 2513 cm-1) covering from 700 nm to 1200 nm with a photoluminescence quantum yield of 60.8% and absorption efficiency of 44.1% (λex = 450 nm). Fabricating a prototype NIR pc-LED device using the title material combined with a 455 nm emitting InGaN LED chip produces a NIR output power of 23.2 mW with photoelectric efficiency of 8.45% under a 100 mA driving current. This NIR light source is then used to demonstrate the quantitative detection of ethanol in solution. These results highlight the feasibility of this material for NIR spectroscopy and validate the prospects of using NIR pc-LEDs in food-quality analysis.

Smartphone-readable RPA-LFA for the high-sensitivity detection of Leishmania kDNA using nanophosphor reporters.

Early diagnosis of infectious diseases improves outcomes by enabling earlier delivery of effective treatment, and helps prevent further transmission by undiagnosed persons. We demonstrated a proof-of-concept assay combining isothermal amplification and lateral flow assay (LFA) for early diagnosis of cutaneous leishmaniasis, a vector-borne infectious disease that affects ca. 700,000 to 1.2 million people annually. Conventional molecular diagnostic techniques based on polymerase chain reaction (PCR) require complex apparatus for temperature cycling. Recombinase polymerase amplification (RPA) is an isothermal DNA amplification method that has shown promise for use in low-resource settings. Combined with lateral flow assay as the readout, RPA-LFA can be used as a point-of-care diagnostic tool with high sensitivity and specificity, but reagent costs can be problematic. In this work, we developed a highly-sensitive smartphone-based RPA-LFA for the detection of Leishmania panamensis DNA using blue-emitting [(Sr0.625Ba0.375)1.96Eu0.01Dy0.03]MgSi2O7 (SBMSO) persistent luminescent nanophosphors as LFA reporters. The greater detectability of nanophosphors allows the use of a reduced volume of RPA reagents, potentially reducing the cost of RPA-LFA. The limit of detection (LOD) of RPA with gold nanoparticle-based LFA readout is estimated at 1 parasite per reaction, but LOD can be 100-fold better, 0.01 parasites per reaction, for LFA based on SBMSO. This approach may be useful for sensitive and cost-effective point-of-care diagnosis and contribute to improved clinical and economic outcomes, especially in resource-limited settings.

From Laves Phases to Quasicrystal Approximants in the Na-Au-Cd System.

The exploratory synthesis of gold-based polar intermetallic phases has revealed many new compounds with unprecedented crystal structures, unique bonding arrangements, and interesting electronic features. Here, we further understand the complexity of gold's crystal chemistry by studying the Na-Au-Cd ternary composition space. A nearly continuous structure transformation is observed between the seemingly simple binary NaAu2-NaCd2 phases, yielding three new intermetallic compounds with the compositions Na(Au0.89(5)Cd0.11(5))2, Na(Au0.51(4)Cd0.49(4))2, and Na8Au3.53(1)Cd13.47(1). Two compounds adopt different Laves phases, while the third crystallizes in a complex decagonal quasicrystal approximant. All three compounds are related through Friauf-Laves polyhedral building units with the gold/cadmium ratio found to control the transition among the unique crystal structures. Electronic structure calculations subsequently revealed the metallic nature of all three compounds with a combination of polar covalent Na-(Au/Cd) interactions and covalent (Au/Cd)-(Au/Cd) bonding interactions stabilizing each structure. These results highlight the crystal and electronic structure relationship among Laves phases and quasicrystal approximants enabled by the unique chemistry of gold.

A novel technology for home monitoring of lupus nephritis that tracks the pathogenic urine biomarker ALCAM.

Introduction: The gold standard for diagnosis of active lupus nephritis (ALN), a kidney biopsy, is invasive with attendant morbidity and cannot be serially repeated. Urinary ALCAM (uALCAM) has shown high diagnostic accuracy for renal pathology activity in ALN patients. Methods: Lateral flow assays (LFA) for assaying uALCAM were engineered using persistent luminescent nanoparticles, read by a smartphone. The stability and reproducibility of the assembled LFA strips and freeze-dried conjugated nanoparticles were verified, as was analyte specificity. Results: The LFA tests for both un-normalized uALCAM (AUC=0.93) and urine normalizer (HVEM)-normalized uALCAM (AUC=0.91) exhibited excellent accuracies in distinguishing ALN from healthy controls. The accuracies for distinguishing ALN from all other lupus patients were 0.86 and 0.74, respectively. Conclusion: Periodic monitoring of uALCAM using this easy-to-use LFA test by the patient at home could potentially accelerate early detection of renal involvement or disease flares in lupus patients, and hence reduce morbidity and mortality.

Producing Tunable Broadband Near-Infrared Emission through Co-Substitution in (Ga1-xMgx)(Ga1-xGex)O3:Cr3.

Broadband near-infrared (NIR) phosphors are in high demand for creating "smart" NIR phosphor-converted light-emitting diode (pc-LED) sources. In this work, a series of Cr3+-substituted NIR-emitting materials with highly efficient, broad, tunable emission spectra are achieved by modifying the simple oxide Ga2O3 using [Mg2+-Ge4+] and [Ga3+-Ga3+] co-unit substitution. The results show that the emission peak can be shifted from 726 to 830 nm while maintaining a constant excitation peak in the blue light region, enabling extensive application. The optical properties stem from changes in the Cr3+ crystal field environment upon substitution. Intriguingly, the temperature-dependent photoluminescence emission peak position shows virtually no change in the [Mg2+-Ge4+] co-substituted materials. This abnormal phenomenon is found to be a comprehensive embodiment of a weakening crystal field environment (red-shift) as the temperature increases and reduced local structure distortion (blue-shift) with increasing temperature. The high quantum yield, NIR emission, and net-zero emission shift as a function of temperature make this phosphor class optimal for device incorporation. As a result, their performance was studied by coating the phosphor on a 450 nm emitting LED chip. The fabricated device demonstrates an excellent NIR output power and NIR photoelectric conversion efficiency. This study provides a series of efficient, tunable, broadband NIR materials for spectroscopy applications and contributes to the basic foundation of Cr3+-activated NIR phosphors.

Treating Superhard Materials as Anomalies.

Superhard materials are among the most scarce functional inorganic solids in existence. Indeed, recent research suggested that less than 0.1% of all known materials are likely to have a Vickers hardness ≥40 GPa. Here, an anomaly detection framework is created to treat these materials as rare occurrences by encoding and reconstructing the input composition and crystal structure information without supervision. The resulting model can quantitatively identify outliers from "normal" behaving materials, leading to the discovery of materials with exceptional properties such as a superhard response. Moreover, examining the difference between the encoded and decoded crystal structure provides fundamental insights into the crystal-chemical origin of hardness. The presented methodology is ultimately generalizable, enabling the design of other outlier materials with rare and unexpected physical properties.

Phosphorescence in Mn4+-Doped R+/R2+ Germanates (R+ = Na+ or K+, R2+ = Sr2+).

Single crystals of three new compounds, Na0.36Sr0.82Ge4O9 (1, proposed composition), Na2SrGe6O14 (2), and K2SrGe8O18 (3), were obtained and characterized using single-crystal X-ray diffraction. Their structures contain three-dimensional (3D) anionic frameworks built from GeO4 and GeO6 polyhedra. The presence of octahedral Ge4+ sites makes the new phases suitable for Mn4+ substitution to obtain red-emitting phosphors with a potential application for light conversion. Photoluminescence properties of Mn4+-substituted Na2SrGe6O14 (2) and K2SrGe8O18 (3) samples were studied over a range of temperatures, and red light photoluminescence associated with the electronic transitions of tetravalent manganese was observed. The Na2SrGe6O14 (2) phase was also substituted with Pr3+ on the mixed Na-Sr site similar to the previously studied Na2CaGe6O14:Pr3+. The red emission peak of the Pr3+ activator occurs at a shorter wavelength (610 nm) compared to that of Mn4+ (662-663 nm). Additionally, second harmonic generation (SHG) data were collected for the noncentrosymmetric Na2SrGe6O14 (2) phase, indicating weak SHG activity. Diffuse reflectance spectroscopy and density of states calculations were performed to estimate the band gap values for pristine Na2SrGe6O14 (2) and K2SrGe8O18 (3) phases.

Sensitive and Reliable Fluorescent Thermometer Based on a Red-Emitting Li2MgHfO4:Mn4+ Phosphor.

Contactless fluorescent thermometers are rapidly gaining popularity due to their sensitivity and flexibility. However, the development of sensitive and reliable non-rare-earth-containing fluorescent thermometers remains a significant challenge. Here, a new rare-earth-free, red-emitting phosphor, Li2MgHfO4:Mn4+, was developed for temperature sensing. An experimental analysis combined with density functional theory and crystal field calculations reveals that the sensitive temperature-dependent luminescence arises from nonradiative transitions induced by lattice vibration. Li2MgHfO4:Mn4+ also exhibits reliable recovery performance after 100 heating-cooling cycles due to the elimination of surface defects, which is rare but vital for practical application. This study puts forward a new design strategy for fluorescent thermometers and sheds light on the fundamental structure-property relationships that guide sensitive temperature-dependent luminescence. These considerations are crucial for developing next-generation fluorescence-based thermometers.

The Role of the Bi3+ Lone Pair Effect in Bi(H3O)(SO4)2, Bi(HSO4)3, and Bi2(SO4)3.

Three new members in the Bi2O3-SO3-H2O system are identified by single crystal X-ray diffraction and Rietveld refinement after a fundamental examination of this phase space. Bi(H3O)(SO4)2 crystallizes in space group P21/c (no. 14, a = 1203.5(4), b = 682.9(2), c = 821.2(2) pm, ß = 102.99(1)°, 861 independent reflections, 88 refined parameters, wR2 = 0.14) homeotypic with Nd(H3O)(SO4)2 featuring edge-sharing BiO9 polyhedra. Bi(HSO4)3 crystallizes in a new structure type in space group P1 (no. 2, a = 492.04(7), b = 910.8(1), c = 1040.8(2) pm, α = 85.443(5)°, ß = 86.897(5)°, γ = 74.542(4)°, 3227 independent reflections, 154 refined parameters, wR2 = 0.05) comprising dimers of edge-sharing BiO8 polyhedra. For Bi2(SO4)3, a new modification crystallizing in space group P21/n (no. 14, a = 1308.03(7), b = 473.25(3), c = 1452.61(8) pm, ß = 100.886(2)°, 3189 independent reflections, 155 refined parameters, wR2 = 0.03) isotypic to Sb2(SO4)3 with noncondensed BiO7 polyhedra is presented. The role of the Bi3+ lone pair effect as elucidated by density functional theory (DFT) calculations is discussed for all three compounds with respect to their structural and optical properties. Additionally, the Bi3+ lone pair activity is compared to the recently reported borosulfates Bi(H3O)[B(SO4)2]4 and Bi2[B2(SO4)6]. Geometrical calculations based on structural data are correlated with electron localization function (ELF) calculations to establish the origin of the direction and strength of the lone pair stereoactivity of Bi3+ in oxidic compounds. Finally, the thermal properties of the three compounds are reported.

Photoluminescent Molecules and Materials as Diagnostic Reporters in Lateral Flow Assays.

The lateral flow assay (LFA) is a point-of-care diagnostic test commonly available in an over-the-counter format because of its simplicity, speed, low cost, and portability. The reporter particles in these assays are among their most significant components because they perform the diagnostic readout and dictate the test's sensitivity. Today, gold nanoparticles are frequently used as reporters, but recent work focusing on photoluminescent-based reporter technologies has pushed LFAs to better performance. These efforts have focused specifically on reporters made of organic fluorophores, quantum dots, lanthanide chelates, persistent luminescent phosphors, and upconversion phosphors. In most cases, photoluminescent reporters show enhanced sensitivity compared to conventional gold nanoparticle-based assays. Here, we examine the advantages and disadvantages of these different reporters and highlight their potential benefits in LFAs. Our assessment shows that photoluminescent-based LFAs can not only reach lower detection limits than LFAs with traditional reporters, but they also can be capable of quantitative and multiplex analyte detection. As a result, the photoluminescent reporters make LFAs well-suited for medical diagnostics, the food and agricultural industry, and environmental testing.

The Deep Eutectic Solvent Precipitation Synthesis of Metastable Zn4V2O9.

A precipitation method involving a deep eutectic solvent (DES)─a mixture of hydrogen bond donor and acceptor─is used to synthesize a ternary metal oxide. Without toxic reagents, precipitates consisting of Zn3(OH)2V2O7·nH2O and Zn5(OH)6(CO3)2 are obtained by simply introducing deionized H2O to the DES solution containing dissolved ZnO and V2O5. Manipulation of the synthetic conditions demonstrates high tunability in the size/morphology of the two-dimensional nanosheets precipitated during the dynamic equilibrium process. According to differential scanning calorimetry and high-temperature powder X-ray diffraction, Zn3V2O8 and ZnO obtained by the annealing of the precipitate are intermediates in the reaction pathway toward metastable Zn4V2O9. Intimate mixing of the metal precursors achieved by the precipitation method allows access to the metastable zinc-rich vanadate with unusually rapid heat treatment. The UV-vis and surface photovoltage spectra reveal the presence of sub-band gap states, stemming from the reduced vanadium (V4+) center. Photoelectrochemical measurements confirm weak photoanodic currents for water and methanol oxidation. For the first time, this work shows the synthesis of a metastable oxide with the DES-precipitation route and provides insight into the structure-property relationship of the zinc-rich vanadate.

Spectral Design of Phosphor-Converted LED Lighting Guided by Color Theory.

In the race to develop new luminescent materials for the next generation of light-emitting-diode (LED)-based solid-state lighting and display applications, it is often forgotten that color theory and human perception should be some of the principal factors guiding materials design. In this Viewpoint, we explore some of the antiquated colorimetrics established originally for incandescent and fluorescent lighting and discuss how they are still widely applied in the literature today to interpret the color quality of luminescent materials, like inorganic phosphors and quantum dots, and to analyze prototype devices, despite their shortcomings. We then shift our analysis toward contemporary ideas in color theory that more accurately describe the color quality of modern LED light bulbs and flat-panel displays. Finally, the perspective examines the opportunities and challenges of applying these new concepts to guide the design of luminescent materials used in LED-based applications.

From simple to complex crystal chemistry in the RE-Au-Tt systems (RE = La, Ce, Pr, Nd; Tt = Ge, Pb).

Polar intermetallics are an intriguing class of compounds with complex relationships between composition and structure that are not fully understood. This work reports a systematic study of the underexplored ternary composition space RE-Au-Tt (RE = La, Ce, Pr, Nd; Tt = Ge, Pb) to expand our knowledge of the intriguing chemistry and diversity achievable with these metallic constituents. These composition spaces are particularly interesting because of the potential to find Au-bearing, highly polar intermetallic compounds. The elements were first reacted through arc welding under an inert atmosphere, followed by annealing at 850 °C. X-ray diffraction of the products identified seven unreported compounds ranging from the simple NaTl-type compounds La1.5Au2Pb0.5 and Nd2-x Au2Pb x to the more structurally complex La5AuPb3 in the Hf5CuSn3-type structure and Pu3Pd4-type RE3Au3Ge (RE = La, Ce, Pr, Nd). First-principles electronic structure calculations investigate the combination of Fermi surface-Brillouin zone interactions, electrostatic interactions, and delocalized metallic bonding that contributes to the formation of these phases. These calculations show that a mixture of electrostatic and metallic bonding plays a dominant role in these phases. The RE-Au-Tt composition space remains full of potential for discovering materials with relevant magnetic and quantum properties, provided the crystal chemistry can be comprehended.

Determining Temperature-Dependent Vickers Hardness with Machine Learning.

Assessing the hardness of structural materials at elevated temperatures is experimentally and computationally challenging, yet crucial for their success. In this work, a machine-learning method was developed to determine a material's temperature-dependent hardness based on its chemical composition and crystal structure. A total of 593 Vickers hardness data collected at various temperatures were extracted from the literature and used to train an extreme gradient boosting (XGBoost) machine-learning model. Applying a combination of composition descriptors and smooth overlap of atomic positions (SOAP) structural descriptors to represent these materials resulted in outstanding accuracy (R2 = 0.91; MAE = 2.52 GPa). The model's intrinsic variance was also measured by using a bootstrap aggregating (bagging) method, and the subsequent predictions showed strong agreement with the experimental data. The capability of the trained model was finally verified by demonstrating the model's ability to discriminate polymorphs, separate the properties of similar compositions, and reproduce the high-temperature hardness of several classic structural materials.

Efficient and Tunable Luminescence in Ga2-xInxO3:Cr3+ for Near-Infrared Imaging.

Broadband near-infrared (NIR) emitting materials are in great demand as next-generation smart NIR light sources. In this work, a Cr3+-substituted phosphor capable of efficiently converting visible to NIR light is developed through the solid solution, Ga2-xInxO3:Cr3+ (0 ≤ x ≤ 0.5). The compounds were prepared using high-temperature solid-state synthesis, and the crystal and electronic structure, morphology, site preference, and photoluminescence properties are studied. The photoluminescence results demonstrate a high quantum yield (88%) and impressive absorption efficiency (50%) when x = 0.4. The NIR emission is tunable across a wide range (713-820 nm) depending on the value of x. Moreover, fabricating a prototype of a phosphor-converted NIR light-emitting diode (LED) device using 450 nm LED and the [(Ga1.57Cr0.03)In0.4]O3 phosphor showed an output power that reached 40.4 mW with a photoelectric conversion efficiency of 25% driven by a current of 60 mA, while the resulting device was able to identify damaged produce that was undetectable using visible light. These results demonstrate the outstanding potential of this phosphor for NIR LED imaging applications.

Advancing Human-Centric LED Lighting Using Na2MgPO4F:Eu2.

The proliferation of energy-efficient light-emitting diode (LED) lighting has resulted in continued exposure to blue light, which has been linked to cataract formation, circadian disruption, and mood disorders. Blue light can be readily minimized in pursuit of "human-centric" lighting using a violet LED chip (λem ≈ 405 nm) downconverted by red, green, and blue-emitting phosphors. However, few phosphors efficiently convert violet light to blue light. This work reports a new phosphor that meets this demand. Na2MgPO4F:Eu2+ can be excited by a violet LED yielding an efficient, bright blue emission. The material also shows zero thermal quenching and has outstanding chromatic stability. The chemical robustness of the phosphor was also confirmed through prolonged exposure to water and high temperatures. A prototype device using a 405 nm LED, Na2MgPO4F:Eu2+, and a green and red-emitting phosphor produces a warm white light with a higher color rendering index than a commercially purchased LED light bulb while significantly reducing the blue component. These results demonstrate the capability of Na2MgPO4F:Eu2+ as a next-generation phosphor capable of advancing human-centric lighting.

Opportunities for Next-Generation Luminescent Materials through Artificial Intelligence.

Luminescent materials are continually sought for application in solid-state LED-based lighting and display applications. This has traditionally required extensive experimental effort. More recently, the employment of data-driven approaches in materials science has provided an alternative avenue to accelerate the discovery and development of luminescent materials. In this Perspective, we give an overview of luminescent materials used for lighting and display applications with a specific focus on inorganic phosphors, quantum dots, and organic light-emitting diodes. We discuss recent progress using data-driven approaches to discover new compounds, predict optical properties, and optimize synthesis, among other topics for each type of material. We then highlight future research directions focusing on using artificial intelligence (AI) to advance these fields and address some cross-cutting challenges limiting the current application of AI techniques in luminescence-related research.

Finding the Next Superhard Material through Ensemble Learning.

An ensemble machine-learning method is demonstrated to be capable of finding superhard materials by directly predicting the load-dependent Vickers hardness based only on the chemical composition. A total of 1062 experimentally measured load-dependent Vickers hardness data are extracted from the literature and used to train a supervised machine-learning algorithm utilizing boosting, achieving excellent accuracy (R2  = 0.97). This new model is then tested by synthesizing and measuring the load-dependent hardness of several unreported disilicides and analyzing the predicted hardness of several classic superhard materials. The trained ensemble method is then employed to screen for superhard materials by examining more than 66 000 compounds in crystal structure databases, which show that 68 known materials have a Vickers hardness ≥40 GPa at 0.5 N (applied force) and only 10 exceed this mark at 5 N. The hardness model is then combined with the data-driven phase diagram generation tool to expand the limited number of reported high hardness compounds. Eleven ternary borocarbide phase spaces are studied, and more than ten thermodynamically favorable compositions with a hardness above 40 GPa (at 0.5 N) are identified, proving this ensemble model's ability to find previously unknown materials with outstanding mechanical properties.