Perovskite Materials: Crystal Chemistry, Synthesis, and Selected Functional Applications

Abstract

Perovskites provide a flexible structural platform for tailoring electronic, optical, ferroic, catalytic, and transport properties. Their performance depends on composition, octahedral connectivity, defects, microstructure, and interfaces, but the effects of these variables differ between oxide and halide systems. This review examines selected advances in these material families, with emphasis on the links between crystal chemistry and device behavior. It first introduces the ideal perovskite structure, the scope and limitations of geometric stability descriptors, and the distinction between compositional and structural classifications. It then discusses synthesis routes and representative applications in photovoltaics, light emission, radiation detection, gas sensing, electrocatalysis, and electrochemical energy storage. The review also discusses the trade-offs associated with compositional substitution, dimensional reduction, and defect control. Computational methods and machine learning are considered as complementary tools for exploring these relationships. Across the reviewed applications, improved performance in a laboratory test does not necessarily establish operational stability or scalability; meaningful comparisons require consistent measurement conditions and a clear distinction between materials, electrodes, and complete devices. The review identifies priorities for more reliable reporting and for integrating structural characterization, computational predictions, and experimental validation.

1. Introduction

Perovskites are a widely studied class of functional materials mainly due to their structures having a high capacity to accommodate diverse chemical compositions and physical properties, suitable for a wide range of technological applications [1,2]. The family takes its name from the mineral calcium titanate, CaTiO₃, as described in 1839 [3]. Research now extends from oxide ceramics to halide semiconductors and layered derivatives, with applications ranging from capacitors and catalysts to solar cells, electroluminescence devices and radiation detectors. This breadth makes it useful to compare how composition and structure influence performance across different perovskite systems.

The properties associated with a perovskite chemical family are not shared by every member of the family. Depending on composition and structure, they include ferroelectricity, piezoelectricity, magnetic ordering, ionic conductivity, luminescence, and photovoltaic or catalytic activity [4,5,6]. Their common structural framework provides a basis for studying how chemical substitution and lattice distortions affect these responses [7]. BaTiO₃ is an established dielectric and ferroelectric material used in capacitors and electromechanical devices [8,9]. SrTiO₃ is an important model oxide and is also investigated in photocatalysis [10]. BiFeO₃ combines ferroelectric and antiferromagnetic order, although exploiting its coupled responses requires control of defects, leakage, and microstructure [11]. Manganites such as La_{1 − x}Sr_xMnO₃ illustrate the sensitivity of electronic and magnetic behavior to composition, while LaCoO₃, LaNiO₃, and related oxides are studied in catalysis and solid oxide cells [12,13,14]. Metal halide perovskites, including MAPbI₃, FAPbI₃, and CsPbBr₃, have attracted attention for their optical absorption, emission, and accessible processing routes [15,16,17,18]. These properties support applications in photovoltaics and optoelectronics, but do not remove the need to control recombination, ion migration, and degradation [18]. Lead-free compositions and double perovskites offer alternative chemistries; their environmental impact and stability must nevertheless be assessed individually rather than inferred from the absence of Pb [19,20].

Progress has been supported by advances in both synthesis and characterization [21,22]. Solid-state processing remains important for oxide ceramics, while solution and vapor deposition enable thin films, nanocrystals, and heterostructures [23,24,25]. Diffraction, microscopy, and spectroscopy can connect processing conditions to phase formation, local structure, defects, and interfacial behavior. Computational screening provides a complementary way to explore candidate compositions before synthesis [26,27]. Machine learning can extend this screening to larger datasets, provided that its predictions remain within a validated domain of applicability [28,29]. High-throughput calculations have also been used to examine thermal transport in oxide and fluoride perovskites [30].

The challenges differ substantially among material families. Many halide perovskites are sensitive to moisture, heat, illumination, or electrical bias, and their degradation may involve ion migration, phase segregation, and interfacial reactions. Oxide perovskites are often more robust thermally, but their performance can depend strongly on oxygen stoichiometry, grain boundaries, and processing route. Manufacturing and lifetime requirements also vary among ceramic capacitors, photovoltaic modules, and catalytic electrodes. Existing reviews cover broad applications as well as specific material classes and techniques [1,3,5]. The purpose of the present review is to focus on selected examples suitable for technological applications; to compare how structural and chemical modifications affect functional behavior and the evidence needed to substantiate those relationships.

The review addresses three questions: which structural or chemical variable is modified; what evidence connects that modification to the measured response; and what are the trade-off limits for practical and valuable applications. Oxide and halide perovskites are considered alongside layered and ordered derivatives, while anti-perovskites are included as a related platform for ionic transport. The discussion focuses on selected applications and computational approaches rather than attempting a complete account of every perovskite property. It closes by identifying common limitations in performance comparison, stability assessment, and the translation of predictions into experiments.

Unlike existing reviews that primarily focus on specific perovskite families, individual applications, or synthesis strategies, this review adopts a comparative, structure–property perspective to examine how compositional changes, structural dimensionality, and defect engineering influence functional performance across selected oxide and halide perovskites and related derivatives. Its objective is to critically evaluate the evidence linking these modifications to material and device behavior, identify the associated trade-offs and limitations, and highlight opportunities for integrating experimental characterization, computational methods, and machine learning to guide the development of more reliable functional materials.

2. Crystal Structure and Chemistry of Perovskite Materials

Perovskite chemistry allows substitution on several crystallographic sites, but chemical formula alone does not establish a perovskite structure. The ideal ABX₃ framework contains corner-sharing BX₆ octahedra, with A-site cations occupying the cavities between them. In oxide perovskites, X is O²⁻; in common metal halide perovskites, it is a halide anion. Substitution, ordering, and changes in connectivity generate a broader set of perovskite-derived materials, which should be distinguished from the ideal three-dimensional framework [31,32]. Dimensional engineering changes both the connectivity and the electronic environment [33]. Organic–inorganic hybrids add molecular degrees of freedom [34], whereas defects and substitution alter local bonding and charge compensation [35]. These features link crystal chemistry to optical anisotropy, chiral responses, and other functional behavior in selected systems [36,37].

2.1. Crystal Structure of the Ideal Perovskite

The ideal perovskite crystal adopts a cubic structure belonging to the Pm$\overline{3}$m space group. In this arrangement, the B-site cation occupies the center of an octahedron coordinated by six equivalent anions, while the larger A-site cation is surrounded by twelve anions within a cuboctahedral coordination environment [35]. Each BX₆ octahedron shares its corners with neighboring octahedra, producing a continuous three-dimensional framework that extends throughout the crystal [36,37]. This interconnected octahedral network plays a central role in determining the physical properties of perovskites. Since adjacent octahedra are directly connected through shared anions, any variation in bond lengths or bond angles propagates throughout the crystal lattice [38]. Consequently, relatively small structural distortions often produce significant modifications in electronic conductivity, dielectric behavior, magnetic ordering, optical absorption, and ionic transport. In the ideal cubic perovskite structure, all B–X bond lengths are equivalent, and the B–X–B bond angles are 180°, reflecting the high symmetry of the crystal lattice. As illustrated in Figure 1, the structure consists of A-site cations occupying the cavities within a three-dimensional framework of corner-sharing BX₆ octahedra, with B-site cations at the octahedral centres and X-site anions at the vertices [39].

Figure 1. Schematic representation of the ideal cubic ABX₃ perovskite structure, showing the three-dimensional framework of corner-sharing BX₆ octahedra and the positions of the A-, B-, and X-site ions [39].

The ideal cubic structure is a reference rather than a universal description of real perovskites. Table 1 summarizes its main geometric features. Cation ordering, as demonstrated in NaLaZr₂O₆, introduces additional structural variables [40]. Octahedral distortions must also be distinguished from effects arising from interfaces or composite architectures, including plasmonic coupling [41,42]. Entropy engineering and coupled lattice–electronic responses illustrate why a geometric description alone cannot predict functionality [43,44].

Table 1. Geometric characteristics of the ideal ABX₃ perovskite framework [42,45].

Structural Parameter	Ideal Geometric Feature	Interpretation and Limitation
General formula	ABX3	Nominal stoichiometry; the formula alone does not establish connectivity.
Crystal symmetry	Cubic, $\mathrm{Pm}\overline{3}$m	Ideal reference structure; real materials may have lower symmetry.
A-site coordination	12-fold cuboctahedral	Describes the ideal cavity; coordination changes with distortion.
B-site coordination	6-fold octahedral	Local geometry relevant to bonding and electronic structure.
Octahedral framework	Corner-sharing BX6 network	Defines the ideal 3D topology; it does not guarantee efficient transport.
B–X bond lengths	Equivalent in the ideal structure	Distortions may introduce inequivalent bonds.
B–X–B angle	180° in the ideal structure	Electronic effects also depend on orbitals, chemistry, and defects.

2.2. Goldschmidt Tolerance Factor

The Goldschmidt tolerance factor, t, estimates the geometric compatibility of the constituent ionic radii in an ABX₃ framework. It is commonly written as follows [46]:

$$\mathrm{t}={\displaystyle \frac{{\mathrm{r}}_{\mathrm{A}}+{\mathrm{r}}_{\mathrm{X}}}{\sqrt{2}\left({\mathrm{r}}_{\mathrm{B}}+{\mathrm{r}}_{\mathrm{X}}\right)}}$$

(1)

where r_A, r_B, and r_X represent the ionic radii of the A-site cation, B-site cation, and anion, respectively.

A value of t close to unity indicates approximate size matching for the ideal cubic geometry. Values below or above unity can indicate a tendency toward distortions or competing structures, but no universal interval uniquely determines a crystal system or space group. The choice of ionic radii, coordination number, and oxidation state must be stated; effective radii for molecular cations introduce additional uncertainty. Structural and functional changes also depend on electronic configuration and defect chemistry, as illustrated by substituted oxides and double perovskites [47,48]. Geometric screening should therefore be combined with crystallographic evidence and, where appropriate, energetic, mechanical, or dynamical stability calculations [49,50]. Table 2 summarizes these limitations qualitatively. The distinction is relevant when comparing inorganic solar absorbers, layered double perovskites, and other energy materials.

Table 2. Qualitative use and limitations of geometric tolerance-factor analysis [45,51,52,53].

Geometric Descriptor	What It Indicates	What It Cannot Establish	Additional Evidence Needed
t near unity	Approximate ionic-size matching to the ideal cubic geometry	A unique space group or operational stability	Diffraction and assessment of competing phases
t below unity	Increasing size mismatch in the ideal construction	A universal tetragonal, orthorhombic, or rhombohedral interval	Measured symmetry, temperature, and tilt pattern
t above unity	A-site/B-site size mismatch; alternative structures may compete	A universal boundary for a hexagonal or non-perovskite phase	Composition-specific structural and energetic analysis
Choice of radii	Depends on valence, coordination, and radius convention	Direct comparability if different conventions are mixed	Explicit radius source and site assumptions
Stability prediction	A screening input combined with other descriptors	Thermodynamic, mechanical, and dynamical stability from t alone	Energy comparisons, elastic response, phonons, and experiment

In addition to the Goldschmidt tolerance factor, the Bartel tolerance factor, τ, provides an alternative descriptor for evaluating the likelihood of perovskite formation. Bartel et al. [54] developed this descriptor using a data-driven analysis of chemical and structural parameters and reported improved classification of experimentally reported ABX₃ compounds compared with an optimized Goldschmidt tolerance-factor criterion. The descriptor can therefore complement conventional geometric approaches when screening candidate compositions. Nevertheless, tolerance factors should not be interpreted as universal measures of thermodynamic stability, because phase competition, bonding, electronic effects, temperature, pressure, and synthesis conditions can also influence whether a perovskite structure forms.

2.3. Classification of Perovskite Materials

Perovskites can be classified along complementary axes: anion chemistry, the nature of the A-site species, cation ordering, and structural dimensionality. These categories overlap. A material may be a metal halide, an organic–inorganic hybrid, a layered structure, and a double perovskite at the same time. The following subsections use these labels to organize the discussion rather than to define mutually exclusive families. Perovskite-derived structures, including vacancy-ordered and layered compounds, should be described by their actual octahedral connectivity. Figure 2 illustrates representative three-dimensional and layered arrangements.

Figure 2. Representative perovskite and perovskite-derived structures. (a) Ordered halide double perovskite with monovalent and trivalent B-site cations [55]. (b) Three-dimensional ABO₃ perovskite framework [56]. (c) Perovskite, Dion–Jacobson, Ruddlesden–Popper, and Aurivillius structures, illustrating different arrangements of perovskite-like slabs and interlayers [57].

2.3.1. Oxide Perovskites

Oxide perovskites commonly have the nominal formula ABO₃, although oxygen nonstoichiometry and substitution are often central to their behavior. Their relatively robust oxide frameworks support applications at temperatures that are inaccessible to many halide systems, including solid oxide cells [58]. A- and B-site chemistry can modify electronic and ionic transport, dielectric response, ferroelectricity, magnetism, and catalysis [59]. These properties remain composition- and condition-dependent: an oxide perovskite is not necessarily an insulator, a wide-band-gap semiconductor, or chemically stable in every environment, opening several application frameworks. Representative materials include BaTiO₃, SrTiO₃, BiFeO₃, LaMnO₃, LaCoO₃, and LaNiO₃ [5,60].

2.3.2. Metal Halide Perovskites

In metal halide perovskites, halide anions coordinate the B-site cations, while the A site may contain inorganic or organic species. Many lead-halide compositions combine strong optical absorption, tunable emission, and processing at relatively low temperatures. Their soft lattices also permit substantial structural fluctuations and ionic motion, which can influence both performance and degradation. MAPbI₃, FAPbI₃, CsPbBr₃, CsPbI₃, and their alloys are representative systems. Stability and defect tolerance must be evaluated for the specific composition, phase, and operating conditions rather than assumed for the whole family [61,62].

2.3.3. Hybrid Organic–Inorganic Perovskites

Hybrid organic–inorganic perovskites combine inorganic frameworks with organic molecular components, introducing structural and dynamic degrees of freedom that are generally absent from conventional oxide perovskites [63]. The organic cation can influence lattice dimensions, octahedral tilting, hydrogen bonding, dielectric response, and phase stability. These effects may modify the optical band gap, charge-carrier behavior, and response to temperature or illumination [64]. Three-dimensional hybrid halide perovskites are particularly important for optoelectronic applications, whereas layered structures provide additional opportunities to control dimensionality and environmental stability [65]. However, the influence of the organic component is composition-dependent, and improvements in one property may be accompanied by disadvantages in charge transport, moisture resistance, or operational stability.

2.3.4. Double Perovskites

B-site-ordered double perovskites are commonly written as A₂BB′O₆ for oxides or A₂BB′X₆ for halides. Ordering, antisite defects, and the valence states of B and B′ influence their properties [66]. The designation does not, by itself, imply very different properties like, for instance, low toxicity or a particular band-gap character. Cs₂AgBiBr₆ is a well-studied lead-free halide example, but its optical transitions and transport differ from those of conventional lead halide absorbers [67]. Oxide double perovskites span different electronic and magnetic behaviors. La₂NiMnO₆-based compositions have been investigated for dielectric tunability [68], while SrAgNiMnO₆ provides an example of impedance and electrical-transport analysis [69].

Lead-free double perovskites provide an alternative approach to modifying the composition and electronic structure of halide perovskite materials. In a typical A₂B(I)B(III)X₆ structure, two different metal cations occupy an ordered arrangement of octahedral sites, replacing the single B-site cation found in the conventional ABX₃ structure [70,71]. This ordering changes the electronic states and optical transitions of the material. Candidate compositions have been investigated for photodetection, light emission, and other optoelectronic applications. Nevertheless, their performance is not necessarily equivalent to that of lead-based halide perovskites: some compositions exhibit indirect band gaps, weak optical transitions, or limitations in charge transport [72]. Their suitability should therefore be assessed using composition-specific structural, optical, and device measurements rather than lead content alone.

2.3.5. Layered Perovskites

Layered perovskites consist of inorganic slabs separated by interlayers, and their properties depend on slab thickness, spacer chemistry, layer orientation, and interlayer bonding. Reduced dimensionality does not necessarily lead to improved performance [73,74]. In layered halide perovskites, two important structural families are Ruddlesden–Popper (RP) and Dion–Jacobson (DJ) phases, which differ in spacer charge and interlayer arrangement. Their general formulas can be expressed as S₂A_{n − 1}B_nX_{3n + 1} for RP phases and DA_n−1B_nX_{3n + 1} for DJ phases, where S is a monovalent organic spacer, D is a divalent organic spacer, A is a monovalent cation, B is a divalent cation, X is a halide anion, and n represents the number of inorganic octahedral layers within each perovskite slab. Although organic spacers can improve moisture resistance, they may also hinder charge transport perpendicular to the inorganic layers. Quantum confinement and exciton binding influence the optical properties of these materials, while film orientation and phase distribution are important for charge extraction and device performance [75,76]. These structural considerations complement compositional strategies for optimizing luminescence and solar-fuel applications [77,78]. The formulas presented here describe representative layered halide perovskites and do not encompass all layered oxides or other perovskite-derived structures. The broader scope of perovskite-related chemistry is further illustrated by ferroic oxides, narrow-band-gap halides, hybrid materials, and mixed-dimensional films [79,80,81,82,83,84].

Oxide, halide, and lead-free double perovskites offer different combinations of structural stability, electronic properties, and processing requirements [1,3]. Oxide perovskites generally exhibit good thermal and chemical stability and are attractive for applications such as dielectric components, catalysis, and gas sensing [3,6]. However, many oxide compositions have relatively wide band gaps, which can limit their use as the sole light-absorbing layer in photovoltaic devices [3]. Halide perovskites offer strong optical absorption, composition-dependent band gaps, and promising charge-carrier properties, making them attractive for photovoltaics, light emission, and photodetection [61,62]. Their limitations include sensitivity to environmental conditions and operational stress, while ion migration and phase segregation can affect the stability of compositions [20,85]. Lead-free double perovskites provide alternatives to lead-containing compositions, but their optical and electronic properties depend on their chemical composition, crystal structure, and defect chemistry [55,67,68]. However, the absence of lead alone is insufficient to establish that a material has low toxicity or a lower overall environmental impact; this requires assessment of its constituent elements and, ideally, its life cycle. Consequently, the suitability of each material family should be assessed in relation to its intended function, processing requirements, stability, and environmental considerations [1,3,20]. Table 3 summarizes the main classifications, their structural characteristics, potential advantages, key limitations, and representative examples.

Table 3. Classification of perovskite materials by composition and structural features [86,87,88,89,90,91,92,93,94].

Classification	Formula or Structural Criterion	Useful Features	Limitations to Assess	Examples
Oxides	ABO3; nonstoichiometric variants possible	Broad ferroic, electronic, ionic, and catalytic responses	Processing temperature, oxygen stoichiometry, leakage, surface chemistry	SrTiO3; BiFeO3; LaCoO3
Metal halides	ABX3 with a halide X	Composition-dependent absorption and emission	Phase stability, ion migration, interfaces, element-specific toxicity	MAPbI3; FAPbI3; CsPbBr3
Organic–inorganic hybrids	Organic species within an inorganic framework	Molecular and lattice degrees of freedom	Orientation, molecular stability, and environmental response	MAPbI3; FAPbI3; layered hybrids
B-site double perovskites	A2BB′O6 or A2BB′X6	Cation ordering and diverse valence combinations	Antisite disorder; optical and transport properties vary	La2NiMnO6; Cs2AgBiBr6; LaBa0.5Ag0.5FeMnO6; BiBaFeZnO6
Layered halides	RP: S2An − 1BnX3n + 1; DJ: DAn − 1BnX3n + 1	Control through slab thickness and spacer chemistry	Anisotropic transport, phase distribution, and orientation	RP and DJ structures

3. Synthesis Strategies of Perovskite Materials

Synthesis determines phase purity, stoichiometry, crystallinity, particle morphology, microstructure, and interfacial quality, all of which influence the functional properties of perovskite materials [22,31]. The appropriate route depends on the chemical composition and intended form, including powders, bulk ceramics, single crystals, or thin films. Conventional oxide perovskites are commonly prepared by solid-state reaction, sol–gel processing, co-precipitation, or hydrothermal and solvothermal methods. These routes may involve drying, precursor decomposition, calcination, and, for bulk ceramics, sintering at elevated temperatures to obtain the desired phase and microstructure.

Halide perovskites require a different approach because their precursor chemistry, crystallization behavior, and thermal stability differ from those of many oxide ceramics [59,80]. Their preparation commonly involves solution processing, antisolvent-assisted crystallization, sequential deposition, or vapor-based methods [95]. The solvent, precursor coordination, deposition rate, annealing conditions, and atmosphere can influence nucleation, film coverage, defect formation, and phase composition. Moisture and oxygen sensitivity, as well as possible thermal degradation, must be considered for susceptible compositions. Although some halide perovskites can be prepared as powders or nanocrystals, the processing conditions must be adapted to the composition. Therefore, oxide ceramic routes and halide processing methods should be compared according to their chemical compatibility, temperature requirements, product form, reproducibility, and environmental constraints [24,25,96,97].

3.1. Solid-State Method

Solid-state synthesis usually starts from mixed oxides, carbonates, or other suitable precursors [98]. Milling reduces diffusion distances, and calcination drives reaction and phase formation; reported processing temperatures depend on composition and commonly lie within a broad high-temperature range [99]. Repeated grinding and heating may be needed to reduce residual phases. The method is well suited to many bulk oxide ceramics, but phase purity and reproducibility require control of precursor mixing, atmosphere, temperature, and volatile components. Grain coarsening, energy consumption, and incomplete reaction are practical limitations.

Controlled heat treatment of SiO₂–Na₂O–Nb₂O₅ glass generated NaNbO₃ surface crystals, with corresponding changes in dielectric behavior [100]. In B₂O₃–Na₂O–Nb₂O₅ glass, crystallization of NaNbO₃ altered the dielectric response [101], while LiNbO₃ formation in SiO₂–Li₂O–Nb₂O₅ glass–ceramics affected conductivity and dielectric relaxation [102].

3.2. Sol–Gel Method

Sol–gel routes distribute precursor species within a gel or polymeric network before thermal conversion to the oxide [103]. Depending on the chemistry, network formation may involve alkoxide hydrolysis and condensation or complexation of metal salts. These routes should not be treated as chemically identical. Shorter diffusion distances can lower the temperature needed for phase formation and improve compositional mixing, but drying, organic removal, and calcination still affect porosity, agglomeration, and residual phases [104]. LiNbO₃ powders prepared by the Pechini method crystallized at 450 °C, and their dielectric relaxation varied with the duration of the heat treatment [105]. In sol–gel, Sr substitution modified the structure and transport of La_0.75Ba_0.25−xSr_xFeO₃ [106].

3.3. Hydrothermal and Solvothermal Synthesis

Hydrothermal and solvothermal methods promote reaction or crystal growth in a sealed vessel using aqueous or nonaqueous media, respectively [107]. Temperature, pressure, mineralizers, solvent chemistry, and precursor ratios influence phase selection and morphology. These routes can produce controlled particles or crystals, but compatibility with the target composition must be established; conditions suitable for an oxide cannot be assumed suitable for a moisture-sensitive halide [108].

3.4. Co-Precipitation Method

Co-precipitation brings several precursor species into a solid phase through controlled solution chemistry [109]. The precipitate is then separated, washed, dried, and, where necessary, calcined. The main advantage is close mixing of the constituent species, but different solubilities and precipitation kinetics can produce compositional deviations. pH, addition order, aging, and washing therefore require control [110]. Figure 3 summarizes representative routes and highlights precursor formation, and final crystallization may occur in separate steps [111,112,113,114].

Figure 3. Schematic representations of representative synthesis strategies for perovskite and related oxide materials: (a) solid-state reaction method for Ba₂NiWO₆ double perovskite [111], (b) step-by-step sol–gel synthesis route [112], (c) schematic illustration of the MgZrO₃ synthesis route [113], and (d) schematic of the co-precipitation method [114].

3.5. Mechanochemical Synthesis and Ball Milling

Mechanochemical synthesis uses mechanical energy to promote chemical reactions between solid precursors [115]. In ball milling, repeated collisions between the milling media and the precursor powders improve mixing, reduce particle size, introduce defects, and may facilitate phase formation without requiring a solvent or prolonged high-temperature treatment. Depending on the composition and milling conditions, the desired phase may form during milling or require a subsequent annealing step [116]. Mechanochemical processing can therefore provide an alternative to conventional ceramic synthesis when reducing solvent use, processing time, or reaction temperature is important.

The outcome depends on the milling equipment, milling frequency or rotational speed, duration, ball-to-powder ratio, milling-media composition, atmosphere, and precursor properties. Vibrational ball milling is one configuration in which oscillatory motion produces repeated impacts and mixing. However, room-temperature or solvent-free processing should not be assumed to be universally applicable: some perovskite compositions require controlled atmospheres, additives, or post-milling heat treatment to achieve the desired phase purity and crystallinity. Potential limitations include contamination from the milling media, local heating, agglomeration, incomplete reaction, and energy consumption. To provide an overview of the synthesis strategies discussed above, Table 4 compares the fundamental principles, potential advantages, and main limitations of representative methods used to prepare perovskite materials.

Table 4. Comparison of the principles, potential advantages, and limitations of representative synthesis and deposition methods for perovskite materials [95,98,99,103,104,107,108,109,110,115,116].

Method	Principle	Potential Advantages	Main Limitations
Solid-state reaction	Reaction between mixed solid precursors, often followed by calcination or sintering	Suitable for many bulk oxide ceramics; relatively straightforward equipment	May require high temperatures and repeated grinding; possible incomplete reaction and grain coarsening
Sol–gel	Molecular-scale mixing through solution chemistry followed by gel formation and thermal conversion	Good compositional mixing; potential control over powder homogeneity and particle size	Drying and calcination affect porosity and agglomeration; precursor chemistry must be controlled
Hydrothermal/solvothermal	Reaction or crystal growth in a sealed vessel using aqueous or nonaqueous media	Can support morphology and crystal-growth control	Requires pressure-rated equipment; solvent and temperature conditions must suit the target composition
Co-precipitation	Precipitation of multiple components from solution under controlled chemical conditions	Potentially homogeneous mixing and relatively simple processing	Sensitive to pH, precipitation rates, precursor solubility, and washing conditions
Mechanochemical/ball milling	Mechanical activation and reaction of solid precursors	Can improve mixing and may reduce solvent use; potentially promotes solid-state reactions	Possible milling-media contamination, agglomeration, amorphization, and need for post-annealing
Vapor-based deposition	Delivery of precursors through vapor-phase processes to form films	Can offer control over film thickness and coverage; useful for selected device architectures	Equipment and process complexity; precursor delivery and composition control may be challenging

4. Selected Functional Applications of Perovskite Materials

Different applications impose different requirements on perovskite chemistry. Photovoltaics requires efficient charge generation and extraction, whereas light emission requires radiative recombination. Sensors exploit a measurable interaction with the environment, while stable electronic devices often need to suppress that same interaction. Catalytic and energy-storage electrodes introduce further requirements involving redox chemistry and mass transport. The examples below examine these differences and identify where structural modifications improve performance or create a competing limitation. Figure 4 summarizes representative applications [117].

Figure 4. Representative applications of perovskite materials in photovoltaics, light-emitting and laser devices, photodetectors, sensors, memory devices, field-effect transistors, and other emerging technologies [117].

4.1. Perovskite Photovoltaic Devices

Research on perovskite photovoltaics has developed rapidly since early demonstrations in 2009 [118]. Many lead halide absorbers combine strong absorption with band gaps suitable for solar conversion and charge transport compatible with thin-film devices. These properties make them attractive, but efficiency depends on the complete device, including selective contacts, interfaces, thickness, and recombination losses. Absorber properties alone do not establish device performance.

In a perovskite solar cell, incident photons with sufficient energy are absorbed by the photoactive layer, generating electronic excitations that contribute to charge-carrier generation [119]. Electrons and holes must then be transported and extracted through appropriate contacts to produce an external current. Electron- and hole-selective transport layers are designed to facilitate extraction of the desired carrier while limiting interfacial recombination. Consequently, photovoltaic performance depends not only on the band gap and absorption coefficient of the absorber, but also on carrier mobility and lifetime, defect density, film morphology, contact energetics, interfacial chemistry, and optical losses. Recombination at defects or interfaces can reduce the open-circuit voltage and fill factor, while incomplete extraction can limit the short-circuit current [120]. Optimization therefore requires coordinated control of the absorber, transport layers, interfaces, and device architecture.

MAPbI₃ helped establish the field through its accessible thin-film processing and band gap near 1.55 eV [121]. Formamidinium-rich absorbers and mixed-cation or mixed-halide compositions offer additional control over optical absorption and phase behavior [122]. Alloying can improve a selected property, but its effects depend on composition and conditions: mixed halides may also undergo light-induced segregation. Claims of improved stability should therefore identify the stress applied, test duration, and comparison material. An appropriate composition must balance band-gap requirements, phase stability, and interface compatibility. Focused on stability, de Morais et al. [17] show a remarkable result using MAPbI3-based perovskite solar cells, with a modification/passivation of the perovskite layer with salophen, with a final improvement in film stability operating under normal environment conditions (25 °C, ∼50% rh) for at least 30 days with efficiencies over 18%.

Interfacial and compositional engineering can help mitigate performance losses in wide-band-gap halide perovskites. For example, Park et al. [123] investigated trivalent metal chloride doping as a strategy for interfacial passivation and improved charge transfer in wide-bandgap perovskite solar cells. Their results illustrate how targeted chemical modification can influence device performance and the stability of mixed-halide compositions. Such strategies should nevertheless be assessed using both photovoltaic metrics and appropriate stability tests, because improvements in initial efficiency do not alone establish long-term operational durability.

All-inorganic compositions such as CsPbI₃ and CsPb(I,Br)₃ remove organic-cation decomposition pathways, but are not automatically stable against moisture, light, or phase transformation [124]. A central challenge is retaining the photoactive phase of CsPbI₃ under the intended operating conditions [125]. Alloying, additives, dimensional control, and strain can alter phase stability, although the resulting optical and transport trade-offs must also be considered.

Lead-free double perovskites, including Cs₂AgBiBr₆, replace Pb with alternative combinations of cations [126]. Their different band structures and optical transitions can limit absorption and carrier collection relative to lead halide absorbers. Substitution and heterostructure design may address individual limitations, but predicted improvements should be distinguished from measured device efficiencies. The absence of Pb is a compositional advantage, not a complete assessment of environmental safety.

Oxide perovskites can also serve as transport or interfacial layers. Lu et al. [96] compared inkjet-printed TiO₂, SrTiO₃, and SnO₂ electron-transport layers. Stanić et al. [127] combined experimental work on a BaTiO3-containing FAPbI₃ device with numerical optimization; the reported experimental efficiency was 11%, while further improvements were modeled rather than experimentally demonstrated. Figure 5 shows the corresponding architectures and energy-level diagrams. Ahmad et al. [97] investigated sequential deposition of Cs_yFA_1-yPbI_3-xBr_x, illustrating how processing and composition can be adjusted for wide-band-gap absorbers.

Figure 5. Representative energy-level alignment and device architectures of oxide-perovskite-based perovskite solar cells: (a) mp-TiO₂ and (b) mp-SrTiO₃-based PSCs [96]; and (c) device layer structure and (d) energy-band diagram of a BaTiO₃-containing PSC [127].

Alanazi. [128] used SCAPS-1D to examine Cs₂AgBiBr₆ solar-cell architectures and their response to changes in layer properties and illumination. Figure 6 summarizes these modeled results. Their value is in testing parameter sensitivity and identifying candidate designs; experimental validation is required before the calculated efficiencies can be interpreted as achievable device performance.

Figure 6. Numerical analysis of a lead-free Cs₂AgBiBr₆ double-perovskite solar cell: (a) device architecture of the ITO/SnO₂/Cs₂AgBiBr₆/Spiro-OMeTAD/Au structure, (b) corresponding energy-band alignment, and the simulated dependence of photovoltaic parameters on light intensity for (c) the initial single-electron-transport-layer (ETL) device and (d) the optimized double-ETL device [128].

Dong et al. [129] investigated electron-transport-layer-free RP devices based on (BA)₂(MA)₃Pb₄I₁₃. Modification of FTO with PEIE improved the contact energetics, yielding an average efficiency of 12.7% and a best-device value of 13.0% in the reported study. Figure 7 presents the architecture, morphology, and energy-level alignment. The diagram places the FTO Fermi level at approximately −4.72 eV before and −4.08 eV after modification relative to vacuum; these correspond to work-function magnitudes of approximately 4.72 and 4.08 eV. The example illustrates interface engineering, but comparisons of lifetime still require the actual storage or operating conditions.

Figure 7. Characteristics of the electron-transport-layer-free Ruddlesden–Popper two-dimensional perovskite solar cells: (a) schematic device architecture, (b) cross-sectional scanning electron microscopy image, (c) schematic energy-level alignment of the functional layers, and (d) chemical structure of PEIE [129].

4.2. Perovskite Light-Emitting Diodes

Perovskite light-emitting diodes use electrical injection and radiative recombination within a perovskite-containing emissive layer. Their performance depends on charge balance, non-radiative losses, and optical outcoupling, as well as the intrinsic emission spectrum [130]. High photoluminescence quantum yield is useful but does not alone determine electroluminescent efficiency. Electrically driven PeLEDs should also be distinguished from conventional LEDs that use perovskite phosphors for color conversion.

Hybrid halide materials support several optical functions, including polarization-sensitive photodetection and electroluminescence [131]. For lighting, morphology and interface quality influence charge injection and radiative losses. Compositional mixing and passivation can improve individual device characteristics, but operational stability must be assessed under specified current or luminance. Figure 8 summarizes the roles of device architecture, optical management, and degradation pathways in PeLED operation [132].

Figure 8. Fundamental operating principles, device architectures, and stability considerations of perovskite light-emitting diodes: (a) major factors affecting PeLED operational lifetime and corresponding strategies for improving device stability, (b) schematic illustration of the charge-injection and radiative-recombination mechanism, and device architectures of (c) bottom-emitting, (d) top-emitting, and (e) transparent PeLEDs [132].

CsPbX₃ nanocrystals offer composition-dependent emission across much of the visible spectrum [133]. Green-emitting CsPbBr₃ can provide narrow emission and is also used in color-conversion structures, including glass composites [134]. Blue emission remains challenging because wide-band-gap compositions and confined structures can introduce additional transport and stability constraints. Reduced dimensionality changes exciton confinement and energy transfer, but does not automatically suppress every non-radiative pathway [135]. In RP and DJ films, the distribution of slab thicknesses and the meaning of the reported average n-value require care [136].

Li et al. [137] introduced pentafluorophenyl diphenylphosphinate (FDPP) during film formation to modify crystallization and passivate undercoordinated sites in quasi-2D PeLEDs. The devices used an ITO/PEDOT:PSS/perovskite/TmPyPB/LiF/Al architecture (Figure 9a,b). Emission remained near 524 nm, with a full width at half maximum of approximately 25 nm. The optimized device reached a reported maximum luminance of about 24,230 cd m⁻² and an external quantum efficiency of 21.06%. Figure 9f shows improved retention of emission relative to the control at an initial luminance of 500 cd m⁻² over the measured time window. This establishes a relative improvement in that experiment, not a technology-level lifetime. The result illustrates the coupled effects of film formation and interfacial or defect passivation.

Figure 9. Device structure, electroluminescence characteristics, efficiency, and operational stability of FDPP-treated quasi-two-dimensional perovskite light-emitting diodes: (a) schematic device architecture, (b) cross-sectional scanning electron microscopy image, (c) electroluminescence spectra, (d) current density–voltage–luminance characteristics, (e) external quantum efficiency as a function of luminance, and (f) operational lifetime at an initial luminance of 500 cd m⁻² [137].

Padhiar et al. [138] used ZrBr₄-mediated phase engineering to modify CsPbBr₃-based nanocrystals for white-light emission. The perovskite component was combined with complementary emitters in a white-light device, so its role should be distinguished from direct electrical injection into a perovskite emissive layer. Figure 10 summarizes spectra, temperature-dependent behavior, and a 50 h operating test. The reported device combined a luminous efficacy of approximately 158 lm W⁻¹ with a color-rendering index near 98 and chromaticity coordinates around (0.35, 0.32). These measurements illustrate the performance of the assembled light source; they should not be compared directly with PeLED external quantum efficiency or treated as evidence of long-term service life.

Figure 10. Characteristics of the ZrBr₄-engineered CsPbBr₃-based white-light device: (a) emission spectra at different driving currents; (b) temperature-dependent spectra; (c) thermal images; (d) spectra during a 50 h operating test at 100 mA; (e) luminous-output and color-rendering metrics over time; and (f) chromaticity diagram and device photograph [138].

4.3. Photodetectors and X-Ray Detectors

Halide perovskites have been investigated as photoconductors, photodiodes, phototransistors, and radiation detectors [139] for many years. Relevant properties include absorption, carrier mobility and lifetime, resistivity, and ionic transport. Their relative importance depends on the detector architecture and the radiation spectrum. Layered structures can increase resistivity or limit ion migration, but anisotropic transport may restrict charge collection. Single crystals reduce grain-boundary contributions, whereas polycrystalline films can offer more accessible large-area fabrication [140].

Bismuth-containing materials offer lead-free routes to radiation detection. Their stoichiometry and connectivity should be specified: compounds such as Cs₃Bi₂I₉ are perovskite-related structures and should not be assumed to possess the three-dimensional ABX₃ framework. Bismuth halides also exhibit resistive switching in memristors, illustrating their varied electrical response under bias [141]. In a related radiation application, Zhao et al. [142] investigated mixed FA/Cs lead-bromide single crystals for gamma-ray spectroscopy, linking crystal-growth control to reduced defect density and improved charge collection. Across these systems, sensitivity must be interpreted alongside field, thickness, noise, response time, and measurement conditions.

Li et al. [143] reported a layered (F-PEA)₂PbI₄ single-crystal X-ray detector stabilized through interactions involving the fluorinated organic spacers. The material exhibited a reported bulk resistivity of 1.36 × 10¹² Ω cm, and the device used Au/perovskite/C₆₀/BCP/Cr contacts (Figure 11). Under the reported 120 keV_p X-ray conditions, the sensitivity was 3402 μC Gy⁻¹ cm⁻², and the lowest detectable dose rate was 23 nGy s⁻¹, with both dose quantities expressed for air. The peak-energy notation describes the source conditions and should not be read as monochromatic 120 keV irradiation. The study connects spacer chemistry to resistivity and ionic stability, but comparison with other detectors also requires the applied field, spectral conditions, and noise criterion.

Figure 11. Characteristics of the layered (F-PEA)₂PbI₄ single-crystal detector: (a) architecture; (b) calculated attenuation coefficients; (c) current-density response versus dose rate; (d) response after different storage periods; and (e) transient photocurrent following pulsed optical excitation [143].

Lédée et al. [144] deposited PEA₂PbBr₄ microcrystalline films directly on patterned flexible substrates. The reported detector response extended to irradiation with a tube operated at 150 kVp. The study reported a sensitivity of approximately 806 μC Gy⁻¹ cm⁻² and a detection limit near 42 nGy s⁻¹ under its measurement conditions. Figure 12 includes dose–response, bias, aging, and bending measurements. The results illustrate the feasibility of flexible layered-perovskite detectors. They do not establish a direct performance ranking against the single-crystal device above because geometry, field, irradiation, and noise evaluation may differ.

Kyesmen et al. [145] prepared colloidal Cs₂AgBiBr₆ nanosheets for photodetection. Their quasi-2D morphology should be distinguished from the spacer-separated layered structures discussed above. The device showed a reported on/off ratio up to 2.39 × 10⁴, a 3 dB bandwidth of 363 Hz, and rise and fall times of 857 and 829 μs. Reported responsivity and detectivity reached approximately 0.12 A W⁻¹ and 1.15 × 10¹² Jones under the study conditions (Figure 13). Photocurrent was retained after 80 days of storage, while structural characteristics were reported after more than 200 days. These are storage-stability observations rather than equivalent periods of continuous operation. Detectivity comparisons also require noise measurement or estimation methods.

Figure 12. Flexible PEA₂PbBr₄ detector: (a) attenuation comparison; (b) response under 150 kVp irradiation; (c,d) dose-dependent response and sensitivity; (e) signal-to-noise ratio; (f) bias dependence; (g) response after aging; and (h) response under bending [144].

Figure 12. Flexible PEA₂PbBr₄ detector: (a) attenuation comparison; (b) response under 150 kVp irradiation; (c,d) dose-dependent response and sensitivity; (e) signal-to-noise ratio; (f) bias dependence; (g) response after aging; and (h) response under bending [144].

Figure 13. Cs₂AgBiBr₆ nanosheet photodetector: (a) frequency response; (b) rise and fall times; (c) responsivity and detectivity versus incident power density; and (d) external quantum efficiency [145].

Figure 13. Cs₂AgBiBr₆ nanosheet photodetector: (a) frequency response; (b) rise and fall times; (c) responsivity and detectivity versus incident power density; and (d) external quantum efficiency [145].

4.4. Sensors

Perovskite sensors exploit changes in electrical or optical response caused by adsorption or other external stimuli. Composition, defects, morphology, and interfaces influence the magnitude and reversibility of these changes. Gas-sensing studies cover both oxides and halides, but reported responses are not directly comparable unless the response definition, gas concentration, temperature, humidity, and exposure protocol are specified [146]. Selectivity and long-term drift are as important as the largest response measured under optimized conditions. Simultaneously, the chemical affinity with some analytes of interest opens infinite possibilities, as the perovskite-based active material can be tailored for applications.

Oxide gas sensors commonly link surface adsorption and redox chemistry to changes in electrical resistance [147]. The details depend on the oxide, its carrier type, the gas, and the operating temperature. Substitution and nanostructuring can modify both electronic transport and accessible surface sites. Humidity and interfering gases must be included in the assessment; studies of humidity sensors, for example, may aim to suppress CO₂ interference rather than detect CO₂ itself [148].

Duan et al. [149] examined Co substitution in LaFeO₃ for CO₂ sensing. This illustrates how B-site modification can change the response of a given oxide, but the mechanism should be supported by surface and electronic evidence rather than attributed to doping alone. Wider comparisons of perovskite VOC sensors must also account for temperature, selectivity, and response definitions.

Metal halide sensors use adsorption-induced changes in conductivity, luminescence, or related signals [150]. Their environmental sensitivity can be useful for detection, but problematic for baseline stability. Dimensional and interface engineering must therefore preserve access to the target gas while limiting irreversible reactions or unwanted humidity response. Neuromorphic sensing introduces a further design requirement: persistent electrical changes may be useful for memory but undesirable for a reversible concentration sensor.

Li et al. [151] studied Cs₂AgBiCl₆ for room-temperature H₂S sensing. Figure 14 shows responses at low-ppb concentrations, a wider concentration series, selectivity tests, and repeated exposure cycles. Measurements down to 5 ppb demonstrate response at that tested concentration; a formal detection limit additionally requires stated noise or a statistical criterion. It supports further investigation of lead-free double perovskites for gas sensing, while humidity dependence, response definition, and long-term reversibility remain essential to comparison with other sensors.

Figure 14. H₂S sensing characteristics of the lead-free Cs₂AgBiCl₆ double-perovskite sensor: (a) low-concentration sensing response, (b) real-time response under progressively increasing H₂S concentrations, (c) response and recovery characteristics, (d) current–voltage behavior at different H₂S concentrations, (e) comparative gas selectivity, and (f) repeatability of the sensor response during cyclic exposure to H₂S [151].

Ye et al. [152] investigated dimensional reduction from Cs₂AgBiBr₆ to hybrid double perovskites containing alkylammonium spacers, including (BA)₄AgBiBr₈, (HA) ₄AgBiBr₈, and (OA)₄AgBiBr₈. The organic interlayers were associated with improved resistance to ambient exposure. The work reported a 20-ppb CO detection limit for (BA)₄AgBiBr₈ and retention of sensing capability after more than 270 days in ambient air. This links spacer chemistry to environmental resistance but does not imply continuous operation for 270 days. Figure 15 summarizes the strategy. The key trade-off is maintaining a reversible gas interaction while reducing unwanted environmental degradation.

Figure 15. Conceptual representation of the dimensional reduction of three-dimensional Cs₂AgBiBr₆ into alkylamine-based two-dimensional hybrid double perovskites and their application in long-term stable CO sensing [152].

Shen et al. [153] prepared Mo-doped LaFeO₃ by a sol–gel route for triethylamine (TEA) sensing. At 140 °C, the optimized 4% Mo composition gave a reported response near 70 to 100 ppm TEA, approximately eleven times the pristine-material response. This numerical response must be read using the definition adopted in the original study. Figure 16 shows temperature dependence, repeated cycles, and concentration-dependent behavior from 2 to 100 ppm; the reported fit had a coefficient of determination R² = 0.99736. The example demonstrates compositional optimization under defined test conditions. A high fitted R² alone does not establish a detection limit, selectivity in mixtures, or accuracy outside the calibration range.

Figure 16. Gas-sensing characteristics of Mo-modified LaFeO₃ toward triethylamine: (a) dependence of sensor response on operating temperature, (b) repeatability during successive TEA exposure cycles, (c) dynamic response and recovery at different TEA concentrations, and (d) concentration-dependent response with corresponding linear fitting [153].

Despite these exceptional outputs in perovskite-based sensors, there are still some drawbacks to overcome: the sensors drift with time, the possibility of building useful capacitance sensors and, naturally, both recovery and lifetimes. Moreover, the issues related to absorption/adsorption of analytes in sensor active material need to be fully understood before moving to technological applications. Nevertheless, the examples mentioned above (and in their references) point to an interesting future in this field.

4.5. Electrocatalysis

Perovskite electrocatalysts allow composition and defect chemistry to be varied over a broad range. Changes in cation valence, surface composition, oxygen stoichiometry, and morphology can affect activity, but the active surface under reaction conditions may differ from the initial bulk structure [154]. Xu et al. [155] studied La_xCo_0.4Fe_0.6O_{3 − δ} supported on nickel–iron foam, with x = 1.0, 0.9, and 0.8. The x = 0.9 sample required a reported HER (Hydrogen Evolution Reaction) overpotential of 160.5 mV at 10 mA cm⁻², compared with 190.5 and 174.9 mV for the other two compositions. These are comparisons within the reported electrode preparation and measurement protocol.

The same study reported a Tafel slope of 80.8 mV dec⁻¹ and a charge-transfer resistance of 2.81 Ω for the optimized electrode. Its double-layer capacitance was 78.1 mF cm⁻². This capacitance is an indirect indicator of an electrochemically accessible interface; conversion to ECSA requires an appropriate specific-capacitance assumption. La deficiency was associated with changes in defect chemistry and surface response, but charge compensation need not occur solely through oxygen vacancies. Figure 17 also includes cycling and time-dependent tests. These support stability over the reported window, while practical durability requires sustained operation, gas-efficiency measurements, and separation of catalyst and support contributions.

Figure 17. HER electrocatalytic characteristics of La-deficient La_xCoFe/NFF perovskite catalysts: (a) polarization curves, (b) Tafel plots, (c) comparison with previously reported perovskite electrocatalysts, (d) electrochemical impedance spectra, (e) double-layer capacitance, (f) long-term HER stability, and (g–i) comparative catalytic-performance parameters of the investigated LaCoFe/NFF catalysts [155].

Wu et al. [156] used electrospinning and calcination to prepare La_0.2Sr_0.8Cu_0.4Co_0.6O_3-δ nanofibers. Within the reported series, the material calcined at 650 °C showed the lowest OER (Oxygen Evolution Reaction) overpotential, 209 mV at 10 mA cm⁻². Calcination can affect phase formation, surface area, conductivity, and residual species simultaneously, so the improvement should not be assigned to morphology alone.

The optimized nanofibers had a reported Tafel slope of 48.7 mV dec⁻¹ and double-layer capacitance of 1.48 mF cm⁻². Figure 18 compares the polarization curves, capacitance measurements, cycling response, and impedance. The limited change after 5000 voltammetry cycles indicates resistance to that cycling protocol; it is not equivalent to a long-duration industrial operating test. Meaningful comparison with other catalysts requires matching electrolyte concentration, loading, area normalization, reference scale, and iR correction [156].

Figure 18. OER characterization of LSCC nanofibers prepared at different calcination temperatures: (a) polarization curves; (b) overpotential at 10 mA cm⁻²; (c) Tafel plots; (d) double-layer capacitance analysis; (e) polarization before and after 5000 cycles; and (f) impedance spectra [156].

Zhong et al. [157] examined La_1.6Sr_0.4MnCoO₆ as an ORR (Oxygen Reduction Reaction)/OER catalyst for zinc–air batteries. Substitution and B-site chemistry can alter oxygen stoichiometry, cation valence, and the surface under polarization. Figure 19 compares this composition with related oxides and benchmark catalysts. Such within-study comparisons are useful, if differences in loading, accessible area, and electrode preparation are considered.

Figure 19. ORR/OER characterization of La_1.6Sr_0.4MnCoO₆ and comparison materials: (a) ORR curves; (b) electron-transfer number and peroxide yield; (c) iR-corrected OER curves; (d) Tafel plots; (e) combined ORR/OER response; and (f) O 1s XPS spectra of surface oxygen-containing species [157].

The reported ORR behavior was consistent with a pathway close to four-electron reduction, while OER analysis gave a Tafel slope near 72 mV dec⁻¹. Using an ORR half-wave potential near 0.63 V and an OER potential near 1.73 V at 10 mA cm⁻² gives a bifunctional gap of approximately 1.10 V under the stated protocol. That value should be assessed against the benchmarks shown in Figure 19 rather than described as intrinsically small. The O 1s spectra indicate differences in surface oxygen-containing species, but peak fitting alone does not uniquely identify or quantify oxygen vacancies. Defect assignments require complementary evidence and consideration of adsorbates, hydroxyls, and surface reconstruction.

4.6. Energy Storage

Perovskite-based electrodes are studied for capacitors and batteries primarily since substitution, defects, and morphology can modify transport and redox behavior [158]. Comparisons must distinguish single-electrode tests from two-electrode cells and distinguish capacitance from capacity. Ghosh et al. [159] investigated nominal Bi_0.05Ca_0.95SnO₃ (BCSO) and reported a symmetric cell with a specific capacity near 73.3 mAh g⁻¹ at 0.5 A g⁻¹ and about 94% retention after 2000 cycles. The nominal composition describes the substituted material; its charge-compensation mechanism requires experimental evidence. The authors attributed improved behavior to changes in charge transfer and electronic response.

The study also reported a maximum energy density near 88 Wh kg⁻¹ at a power density of 2.10 kW kg⁻¹ [159]. These values must be interpreted using the original voltage window, calculation method, and mass basis; they should not be treated as packaged-device metrics. Figure 20 presents electrode-level measurements, including CV, discharge profiles, and capacitance analysis. Reported electrode values were approximately 193 F g⁻¹ at 5 mV s⁻¹ and 162.8 F g⁻¹ at 1 A g⁻¹, with about 95% retention after 2000 cycles. These differ in metric and test configuration from the symmetric-cell capacity. Nonlinear discharge and redox features indicate Faradaic contributions, but do not by themselves establish pseudocapacitive kinetics or exclude battery-type storage.

Figure 20. BCSO electrode measurements: (A) cyclic voltammetry; (B) galvanostatic charge–discharge; (C) specific capacitance derived from CV and discharge measurements; and (D) capacitance retention and coulombic efficiency during cycling [159].

Chatterjee et al. [160] studied Cs₃Bi₂Cl₉ and Cs₃Bi₂Br₉, which belong to a perovskite-related halide family rather than the ideal ABX₃ network. The chloride composition gave a reported specific capacitance near 1617 F g⁻¹ at 2 mV s⁻¹, compared with approximately 613 F g⁻¹ for the bromide under the same protocol. The proposed storage mechanism involved Bi redox chemistry and interaction with hydroxide species. Establishing whether the initial halide phase remains the active phase requires structural and chemical evidence after or during cycling. An asymmetric Cs₃Bi₂Cl₉/activated-carbon cell with PVA–KOH gel electrolyte was also demonstrated.

Figure 21 summarizes the assembled-cell measurements. The study reported an energy density near 27 Wh kg⁻¹ at a power density of 689 W kg⁻¹ and approximately 80% capacitance retention after 5000 cycles [160]. The electrode value of 1617 F g⁻¹ and the cell energy density should not be treated as directly interchangeable measures. Interpretation requires the voltage window, active-mass convention, charge balance between electrodes, and treatment of Faradaic discharge. A laboratory two-electrode demonstration is an important step beyond a single-electrode test, but it does not establish commercial viability.

Figure 21. (a) Photograph of the fabricated asymmetric supercapacitor; (b) cyclic voltammetry curves at different scan rates; (c) galvanostatic charge–discharge profiles at different current densities; (d) specific capacitance at different current densities; (e) Ragone plot showing the energy- and power-density characteristics; and (f) cycling stability and coulombic efficiency over 5000 charge–discharge cycles [160].

Baskoro et al. [161] investigated three-dimensional and layered hybrid halides as lithium-ion battery anodes. The study reported improved cycling after chlorine incorporation in a three-dimensional composition, including testing over 1000 cycles at 500 mA g⁻¹. Layered lead-chloride compositions with different organic ammonium spacers were also examined, with reported capacities exceeding 1000 mAh g⁻¹ at 100 mA g⁻¹. Figure 22 summarizes cyclic voltammetry, capacity, rate, and cycling measurements. These values should be evaluated against the initial coulombic efficiency, reversible capacity, active-mass basis, and possible conversion or interfacial storage processes. High capacity alone does not demonstrate preservation of the initial perovskite framework.

Figure 22. (a) cyclic voltammetry curves, (b) capacity profiles, (c) rate performance, and (d) long-term cycling stability [161].

In addition, in supercapacitors, perovskite-related oxides may contribute to charge storage through surface or near-surface redox reactions, while their electrical conductivity and accessible surface area influence rate performance [162]. In batteries, selected compositions are investigated as electrode materials or functional components where ion insertion, conversion, or redox reactions can contribute to capacity [163]. These applications impose different requirements: supercapacitors typically emphasize rapid charge–discharge, power capability, and cycle life, whereas batteries require a favorable balance of capacity, energy density, rate capability, and long-term cycling stability. Important limitations for perovskite-based electrodes can include low intrinsic conductivity in some compositions, limited accessible active sites, volume or structural changes during cycling, dissolution or degradation in the electrolyte, and synthesis costs. Comparisons should therefore report the electrode composition and loading, electrolyte, current or scan-rate conditions, capacity or capacitance calculation method, and cycling protocol. Measurements from a three-electrode configuration should also be distinguished from full-device results, because the latter more directly reflect practical energy and power performance. Taken together, these application-specific examples demonstrate that perovskite performance depends on composition, processing, structure, interfaces, and measurement conditions. Table 5 summarizes representative studies across the six application areas, distinguishing experimental results from simulations and highlighting the principal performance metrics and limitations.

Table 5. Representative perovskite and perovskite-related materials for selected functional applications: composition, modification strategy, reported performance, and principal limitations.

Application	Materials	Main Strategy	Key Reported Results	Main Limitations
Photovoltaics	Salophen-modified MAPbI3 [17]	Molecular modification/passivation	Efficiency above 18%; stability reported for at least 30 days at 25 °C and approximately 50% RH	The 30-day result should not be interpreted as continuous operational lifetime unless confirmed by the original study
Photovoltaics	TiO2, SrTiO3 and SnO2 transport layers [96]	Inkjet-printed electron-transport layers	Optimum efficiency of 17.37% for the SnO2-based device	Performance depends on printing conditions, layer morphology and device architecture
Photovoltaics	CsyFA1-yPbI3-xBrx [97]	Sequential deposition; mixed-cation, mixed-halide absorber	Processing and composition investigated for wide-band-gap solar cells	Quantitative efficiency and stability values are not provided in the supplied discussion
Photovoltaics	Wide-band-gap halide perovskites [123]	Trivalent-metal-chloride modification	Charge transfer and device performance investigated	Quantitative performance and stability values are not provided in the supplied discussion
Photovoltaics	BaTiO3-containing FAPbI3 device [127]	Oxide incorporation and numerical optimization	Experimental efficiency of 11%	Modelled improvements must be distinguished from experimentally measured results
Photovoltaics	Cs2AgBiBr6 [128]	SCAPS-1D solar-cell simulation	Device architecture and transport-layer properties optimized numerically	Experimental validation is required
Photovoltaics	(BA)2(MA)3Pb4I13 [129]	Layered RP perovskite; PEIE-modified FTO contact	Average efficiency of 12.7%; best-device efficiency of 13.0%	Device lifetime requires evaluation under defined storage or operating conditions
Light-emitting devices	FDPP-treated quasi-2D perovskite [137]	Crystallization control and defect passivation	Emission peak of 524 nm; FWHM of approximately 25 nm; maximum luminance of 24,230 cd m−2; EQE of 21.06%	The reported retention test does not establish technology-level lifetime
Light-emitting devices	ZrBr4 -engineered CsPbBr3-based white-light device [138]	Phase engineering and combination with complementary emitters	Luminous efficacy of approximately 158 lm W−1; CRI near 98; 50 h operating test at 100 mA	The 50 h test does not establish long-term service life
Photodetection and X-ray detection	Mixed FA/Cs lead-bromide single crystals [142]	Single-crystal growth and defect control	Investigated for gamma-ray spectroscopy and charge collection	Numerical detector performance and irradiation conditions are not provided in the supplied discussion
X-ray detection	(F-PEA)2PbI4 single crystal [143]	Layered structure with fluorinated organic spacers	Sensitivity of 3402 μC Gy−1 cm−2; detection limit of 23 nGy s−1	Comparisons require consistent irradiation, bias, geometry and noise criteria
X-ray detection	PEA2PbBr4 microcrystalline films [144]	Flexible layered-perovskite detector films	Sensitivity of approximately 806 μC Gy−1 cm−2; detection limit near 42 nGy s−1	Results depend on irradiation, bias and detector geometry
Photodetection	Cs2AgBiBr6 nanosheets [145]	Colloidal quasi-2D nanosheets	On/off ratio of 2.39 × 104; responsivity of approximately 0.12 A W−1; detectivity of approximately 1.15 × 1012 Jones	Storage stability does not establish equivalent continuous operational stability
Gas sensing	Co-substituted LaFeO3 [149]	B-site Co substitution	LaCo0.1Fe0.9O3: CO2 detection range of 25–10,000 ppm; response/recovery times of 6/42 s at 5000 ppm	Response depends on concentration, temperature, humidity and measurement protocol
Gas sensing	Cs2AgBiCl6 [151]	Lead-free double perovskite	H2S response reported at concentrations down to 5 ppb	The lowest tested concentration is not automatically a formally determined detection limit
Gas sensing	(BA)4AgBiBr8, (HA) 4AgBiBr8 and (OA)4AgBiBr8 [152]	Dimensional reduction and organic-spacer engineering	CO detection limit of 20 ppb for (BA)4AgBiBr8; sensing performance retained after 270 days of continuous ambient-air exposure	Ambient-air exposure is not equivalent to continuous sensing under a controlled gas stream
Gas sensing	4% Mo-doped LaFeO3 [153]	Sol–gel synthesis and Mo substitution	TEA response of approximately 70 at 100 ppm and 140 °C	Response definition, selectivity and detection-limit criteria must be considered
Electrocatalysis	LaxCo0.4Fe0.6O3 − δ [155]	Composition tuning on nickel–iron foam	At $\backslash (\mathrm{x}=0.9\backslash )$: HER overpotential of 160.5 mV at 10 mA cm−2; Tafel slope of 80.8 mV dec−1	Durability and catalyst-versus-support contributions require further evaluation
Electrocatalysis	La0.2Sr0.8Cu0.4Co0.6O3 − δ nanofibers [156]	Electrospinning and calcination	OER overpotential of 209 mV at 10 mA cm−2; Tafel slope of 48.7 mV dec−1	Stability over 5000 cycles does not establish long-term industrial operating life
Electrocatalysis	La1.6Sr0.4MnCoO6 [157]	Composition and surface-chemistry engineering	ORR half-wave potential near 0.63 V; OER potential near 1.73 V at 10 mA cm−2	Reference electrode, electrolyte and correction protocol must be consistent for comparisons
Energy storage	Bi0.05Ca0.95SnO3 [159]	Bi-substituted oxide	Symmetric-cell capacity near 73.3 mAh g−1; energy density near 88 Wh kg−1; electrode capacitance near 193 F/g at 5 mV/s	Electrode and cell metrics must be distinguished; the mass basis and voltage window should be checked
Energy storage	Cs3Bi2Cl9 [160]	Bismuth-halide electrode material	Electrode capacitance near 1617 F g−1; asymmetric-cell energy density near 27 Wh kg−1	Electrode capacitance cannot be directly equated with full-cell energy performance
Energy storage	Cs3Bi2Br9 [160]	Comparative halide composition	Electrode capacitance near 613 F g−1	Full-cell performance should not be inferred from electrode-level capacitance
Energy storage	Three-dimensional and layered hybrid halides [161]	Chlorine incorporation and organic-spacer variation	Up to 1000 cycles reported for a chlorine-incorporated 3D composition; some layered compositions exceeded 1000 mAh g−1 at 100 mA g−1	Reversible capacity, initial coulombic efficiency, cycling retention and structural evolution must be considered

The comparison highlights the importance of evaluating performance alongside stability, testing conditions, and device-level validation. Photovoltaic efficiency should be considered alongside operational stability and scalability; detector sensitivity alongside noise and response speed; and sensor response alongside selectivity, reversibility, and baseline drift. Similarly, electrocatalytic activity and energy-storage performance require clearly defined testing protocols and, where possible, validation at the complete-device level.

5. Computational Materials Design and Machine Learning

Computational methods help relate perovskite chemistry to structure, electronic behavior, defects, and transport. Machine learning can extend this analysis to larger compositional spaces by learning from calculated or measured data [164]. The approaches are complementary: calculations provide model-dependent physical information, while statistical models enable rapid predictions within their validated domain [28,29]. Figure 23 summarizes an iterative workflow linking candidate selection, prediction, synthesis, and measurement.

Figure 23. Computational design and machine-learning workflow for the discovery and optimization of perovskite materials.

5.1. Computational Methods for Perovskite Materials

First-principles and atomistic methods can examine structural and transport questions that are difficult to isolate experimentally [55]. DFT is used to compare candidate structures, electronic states, and defect configurations, with examples ranging from oxide work functions to mixed-halide electronic properties and lithium-containing compositions [165]. Studies of multiferroics and substituted oxides further illustrate links between lattice distortions and electronic or ferroic response [166]. The results depend on the chosen functional, structural model, magnetic configuration, and treatment of defects; they should not be described as uniformly accurate across all perovskites. Besides the traditional use of DFT to get suitable information on electronic structure, the extension to sensors becomes of particular interest, as it would be possible to study about change/affinity among active film and different molecular analyses. Therefore, sensor improvement becomes achievable in specific applications, such as in the biofields.

Huang et al. [167] combined DFT and ab initio molecular dynamics (AIMD) to examine Li-ion transport in the anti-perovskite Li₃OBr_0.5Cl_0.5 under strain and doping. Antiperovskites are structurally related materials with a different arrangement of ionic species and should be distinguished from the ABX₃ cation framework introduced earlier. Figure 24 presents temperature-dependent diffusion and Arrhenius analysis. Extrapolated room-temperature transport is a prediction that depends on the simulated temperature range, sampling, and assumed transport regime; it is not a direct room-temperature measurement.

Figure 24. Temperature-dependent Li-ion diffusion in Li₃OBr_0.5Cl_0.5 anti-perovskite obtained from AIMD simulations and Arrhenius analysis [167].

Calculations on hybrid and inorganic halides address phase energetics, interfaces, electronic structure, and the effects of external conditions [168]. Spin–orbit coupling can be important for heavy-element systems, and conventional exchange-correlation approximations may give inaccurate band gaps. Hybrid functionals and quasiparticle methods can improve selected electronic predictions, while treatment of optical excitations requires additional assumptions or methods. Mechanical, dynamical, and thermodynamic stability are distinct criteria: a favorable formation energy alone does not establish stability against all competing phases or under-development operation.

Molecular dynamics follows atomic motion within an assumed interaction model. It can probe diffusion, structural fluctuations, and degradation pathways, but its conclusions depend on the force field or electronic-structure method and on accessible length and time scales. Reactive simulations of CsPbI₃ have examined the effects of field and temperature on surface behavior [169]. Simulations of multicomponent oxides have explored how disorder and vacancies affect oxygen transport. Such trends should be evaluated within the modeled compositions and conditions rather than generalized to every high-entropy perovskite.

Monte Carlo methods sample configurations within a specified statistical model. They can be used for magnetic ordering, cation arrangements, and disorder, as illustrated by studies of PrFeO₃ [170]. Their physical interpretation depends on the Hamiltonian, parameters, and sampling method. DFT, molecular dynamics, and Monte Carlo therefore address different levels of the problem; combining them is useful when the transfer of assumptions and parameters is made explicit.

5.2. Machine Learning for Perovskite Materials

Machine learning relates descriptors to target properties and can reduce the number of expensive calculations needed to explore a composition space. Shibata et al. [171] used tailored descriptors for predicting band gaps in perovskite-type oxides, illustrating the value of chemically meaningful inputs. Prediction accuracy and interpretability should nevertheless be assessed separately: a descriptor associated with a prediction is not automatically a causal mechanism. Validation on compositions distinct from the training set is needed to assess transferability.

Fatima et al. [172] developed models using 1185 computationally characterized halide compositions to predict energetic and electronic properties and used SHAP analysis to examine feature contributions. Ni et al. [173] combined DFT and ML for A₂BX₆-type inorganic materials, using 546 calculated data points and reporting R² values of 0.92 and 0.93 for band gap and formation energy. The models were then applied to 3276 candidate compositions and used to identify a smaller set for further assessment. Figure 25 shows prediction plots for four models. The scores describe performance under the original validation procedure; they do not independently demonstrate experimental synthesizability or out-of-domain accuracy.

Figure 25. Comparative performance of four machine-learning models for band-gap prediction: (a) XGBoost, (b) Random Forest, (c) Support Vector Machine, and (d) AdaBoost [173].

DFT-derived labels inherit the limitations of the underlying calculations, and ML inherits both those limitations and dataset bias. Reliable assessment therefore requires the train/test strategy, absolute error metrics, uncertainty estimates, and an independent validation route.

Further studies have used ML to screen lead-free double perovskites and all-inorganic candidates for photovoltaic applications [174,175]. Screening should report on the candidate space, rejection criteria, and validation of selected predictions. Many screened formulas are not themselves evidence of discovery; the relevant outcome is a validated improvement in a material or device.

Defect prediction offers another use of ML, but defect charge state, chemical potentials, and structural environment must be represented consistently. Models that omit these variables may confuse composition trends with changes in calculation conditions. Active learning can prioritize new calculations or experiments where uncertainty is high, if uncertainty is calibrated and the selection process remains physically constrained.

5.3. Critical Assessment and Future Opportunities

The main limitation of data-driven discovery is often the dataset rather than the algorithm. Experimental records may omit unsuccessful syntheses, while computational datasets can mix different approximations or structural assumptions. Models trained on these records can learn biases that are difficult to detect from a single score. More reliable studies should document data provenance, avoid near-duplicate compositions across training and test sets, and assess performance using both absolute errors and chemically separated tests.

Interpretability is useful when it generates a hypothesis that can be tested independently. Feature-importance or SHAP plots describe how a model uses its inputs; they do not establish a microscopic mechanism. Transfer to new chemical families is a separate challenge, particularly for disordered, layered, or multicomponent systems. External tests and uncertainty estimates should therefore accompany claims of generalization, and predicted candidates should be evaluated against phase stability and feasible synthesis conditions.

Physics-informed models and closed-loop experimentation can reduce some of these limitations, but they require explicit constraints and well-defined validation targets. A practical loop should connect a prediction to a synthesis attempt, structural characterization, and a measurement under specified conditions, then use both successful and unsuccessful outcomes to update the model. The goal is not only faster screening but a more reliable link between predicted chemistry and reproducible function.

As a point of interest, different ML cores are nowadays available, and the choice of the best possible one is always a trade-off between the final purpose and the dataset characteristics (and data extension/features). With the huge increase in perovskite-based research and publications, creating a reasonable dataset that can be used in any of the available ML methods will not be an issue. Therefore, it is expected that these methods will gain extreme relevance in the future.

6. Environmental Sustainability and End-of-Life Considerations

Environmental sustainability is an important consideration in the development of perovskite materials, particularly when laboratory-scale performance is used to motivate future technological applications. For lead-containing halide perovskites, the potential release of lead during material degradation, device damage, or disposal requires appropriate encapsulation, collection, recycling, and end-of-life management [176]. Encapsulation can reduce exposure under certain conditions, but its effectiveness must be evaluated under realistic ageing, mechanical damage, and environmental scenarios. Recycling strategies should aim to recover valuable components while limiting the release of hazardous substances.

Lead-free compositions may reduce one important environmental concern, but they should not automatically be considered environmentally benign. Their sustainability also depends on the toxicity and availability of constituent elements, precursor synthesis, solvent use, processing temperatures, manufacturing yield, device lifetime, and the energy required for material recovery. Some oxide perovskites offer high thermal and chemical stability, but high-temperature synthesis or sintering may increase processing-energy requirements. Conversely, lower-temperature solution processing can reduce some manufacturing demands but may involve hazardous solvents or require careful control of waste streams [177,178].

A meaningful sustainability assessment should therefore consider the entire material and device life cycle, including raw-material extraction, precursor preparation, manufacturing, operational durability, repair or replacement, and end-of-life treatment. Comparisons between candidate materials should use clearly stated system boundaries and should distinguish experimentally measured environmental performance from qualitative expectations. Improving efficiency is important, but it should be considered alongside lifetime, resource use, process safety, recyclability, and the prevention of harmful releases.

7. Challenges and Future Perspectives

The reviewed studies show that performance gains arise from different combinations of composition, dimensionality, defects, and interfaces. Their transfer between applications is limited by competing requirements. Halide optoelectronic devices generally need to suppress non-radiative recombination and unwanted ionic motion, whereas sensors rely on an accessible and reversible interaction with the environment. Catalysts and storage electrodes require redox activity while retaining a useful structure or forming a stable active phase. Oxides are already used in established technologies, so commercialization should not be presented as an unresolved challenge for the entire perovskite family. The more specific question is which emerging compositions or architectures can meet their intended lifetime and manufacturing requirements.

Stability reporting is a common weakness. Storage in air, cycling, and continuous biased operation probe different processes and should not be combined into a single lifetime claim. Future comparisons should report temperature, atmosphere, illumination or current, geometry, encapsulation, and the criterion used for failure or retention. Scalable processing also requires more than reproducing a small-area result: yield, uniformity, precursor control, and batch variability matter. Environmental assessment should address the full composition, solvent and energy inputs, exposure pathways, and end-of-life handling. Lead-free chemistry alone is not sufficient evidence of sustainability.

Three priorities follow from these limitations. First, compare structural strategies using measurements that distinguish bulk, surface, and interfacial effects. Second, connect performance metrics to the relevant operating conditions and device configuration. Third, validate computationally selected materials experimentally and report negative outcomes as well as successful ones. These steps would make it easier to identify which improvements are transferable, which depend on a narrow processing window, and which require a different device design.

In addition, the challenges facing perovskite materials differ according to their composition, structure, and intended application. For oxide perovskites, important considerations include the temperature and energy required for synthesis, control of oxygen stoichiometry, phase purity, grain-boundary effects, and the relationship between defect chemistry and long-term performance. Some oxide systems require high-temperature processing, and their functional properties may be sensitive to atmosphere and thermal history. For metal halide perovskites, major concerns include sensitivity to moisture, heat, illumination, and electrical bias; ion migration and phase segregation in some compositions; interface degradation; and the environmental implications of lead-containing materials. Solution processing can enable relatively low-temperature film fabrication, but reproducible crystallization, uniform coverage, and scalable manufacturing remain important challenges.

For lead-free double perovskites, the avoidance of lead does not automatically guarantee high performance or environmental sustainability. Some compositions exhibit indirect band gaps, limited absorption in the visible region, or less favorable charge-carrier transport than leading lead-halide absorbers. Cation ordering, defects, and phase stability can also influence their functional properties. Further optimization is needed to establish which compositions can deliver a useful balance of stability, electronic performance, processability, and cost. Across all families, comparisons should be based on well-defined measurement conditions and should distinguish intrinsic material properties from electrode or device performance. Future progress will depend on reproducible synthesis, clear reporting of uncertainties, long-term testing under relevant operating conditions, and careful assessment of manufacturing and end-of-life requirements.

8. Conclusions

Perovskite materials provide a versatile platform for developing functional materials because their composition, crystal structure, defects, and interfaces can be modified to tailor their properties. This review has examined selected oxide and halide perovskites and related derivatives, highlighting the relationships between crystal chemistry, synthesis, and applications in photovoltaics, light emission, radiation detection, sensing, electrocatalysis, and electrochemical energy storage. The comparison shows that the properties required for a particular application depend strongly on the material family and device configuration. Halide perovskites are attractive for optoelectronic applications because of their optical absorption and tunable electronic properties, whereas oxide perovskites offer useful combinations of thermal stability, dielectric behavior, catalytic activity, and adjustable defect chemistry. Hybrid organic–inorganic and lead-free double perovskites introduce further opportunities for compositional and structural design, although their performance and stability must be evaluated individually.

Synthesis conditions, phase purity, microstructure, defects, and interfaces play important roles in determining functional behavior. Computational approaches can support the interpretation and screening of candidate materials, but their predictions require careful assessment of methodological limitations and, where possible, experimental validation. Similarly, performance improvements observed in individual laboratory tests should not automatically be interpreted as evidence of long-term stability, scalability, or commercial readiness. Future progress will require reproducible synthesis, comprehensive structural and functional characterization, transparent reporting of measurement conditions, and more systematic evaluation of operational lifetime and environmental impact. Comparisons should distinguish intrinsic material properties from the behavior of complete devices and should consider processing requirements, resource use, and end-of-life management. Integrating crystal-chemical understanding with experimental validation and application-specific engineering will be essential for translating promising perovskite materials into reliable technologies.