Particle-resolved optical diagnostics of solid fuel combustion for clean power generation: a review

Chemical energy carriers are crucial for addressing challenges that arise from time lag, large distances, and temporal fluctuations in renewable energy production, which lead to unbalanced energy production and demand. The thermochemical utilization of chemical energy carriers such as solid fuels must be decarbonized to achieve a climate-neutral circular economy as soon as possible, while remaining important for reliable electricity generation and stable economics. To accomplish this, extensive fundamental research is required to understand the underlying chemical and physical processes that can potentially be realized at an industrial scale. This paper reviews optical diagnostics used for particle-level combustion studies for clean power generation applications. The focus is on particle-resolved optical experiments for oxy-fuel coal combustion, biomass combustion, and utilization of iron in regenerative oxidation–reduction schemes. Previous studies are categorized by fuel and reactor types, investigated parameters, and experimental methodology. Phenomenological aspects of the multi-stage combustion process at the particle level are outlined using examples of bituminous coal and iron particle burning in hot gas. A selection of experimental studies is highlighted, with a particular methodological focus on measuring quantities at the particle level. These representative examples address relevant parameters, including particle number density, particle size and shape, surface temperature, ignition and combustion time, gas flame structure, gas temperature and species, nanoparticle formation, gas velocity, and particle dynamics. Finally, issues and problems that require further effort to improve diagnostics for solid fuel combustion studies are discussed.


Solid fuel for low-carbon electricity generation
In the first two decades of the 21st century, worldwide annual primary energy consumption has consecutively increased and reached a historical peak in 2021 [1]. Despite an increasing share of energy from renewable sources in recent years, the overall consumption of fossil fuels including oil, coal and natural gas has not changed and accounted for approximately 82% of the primary energy supply in 2021 [1]. Fossil fuel use inevitably leads to increasing atmospheric greenhouse gas concentrations and intensification of global warming, supported by solid scientific evidence [2]. Growing primary energy consumption and long-term global warming result in substantial climate change, which deeply impacts life on earth and the sustainable development of society and economics. 195 countries signed the Paris Agreement at the twenty-first session of the Conference of the Parties (COP 21) in 2015, agreeing to increase efforts to address climate change by limiting global temperature rise this century well below 2 • C above preindustrial levels and pursuing to limit the temperature increase even further to 1.5 • C.
Among the wide variety of fossil fuels, coal remains the major source (44.8% in 2021) [3] of worldwide energy-related carbon emissions. A dominant share of 36% in electricity generation relied on coal-fired power plants [3]. To reduce greenhouse gas emissions, various technologies have been proposed and developed to retrofit existing coal-fired power plants. This review focuses on fundamental research in two technical pathways: oxy-fuel combustion technology and recyclable metal fuel utilization technology.
Oxy-fuel combustion is a carbon capture and storage (CCS) technology that encompasses pre-combustion capture, post-combustion capture, and chemical looping [4]. Previous studies have been substantially reviewed in the literature [4][5][6][7][8][9][10], exploring concepts, challenges and urgent demands for research, spanning from fundamental studies to industrial applications. Among multiple proposals for reducing CO 2 emission, oxy-fuel coal combustion is a promising technology for coal-fired power plants [6]. Figure 1 is a schematic layout of a power plant using oxyfuel combustion of solid carbonaceous fuels [4,6,7,11]. To enable CO 2 capture and storage of the flue gas, the solid fuel is oxidized in an atmosphere of recirculated flue gas and pure oxygen (O 2 ) obtained from an air separation unit. The main combustion products are carbon dioxide (CO 2 ) and water (H 2 O). After cleaning the flue gas (removing H 2 O, SO x , NO x and ash), CO 2 can be compressed, transported, and stored. Due to high flame temperatures in pure O 2 , flue gas is recirculated to the boiler, regulating the temperature of the combustion chamber within a range comparable to conventional air-fired conditions, avoiding damage to the boiler.
Air separation and CO 2 compression account for 7%-10% efficiency penalty during the power generation process [4]. Nevertheless, oxy-fuel combustion is the most energy and cost-efficient approach among the available CCS technologies [6]. However, the combustion processes are significantly impacted by introducing CO 2 as the primary inert species due to its chemical and thermal properties, such as the molar heat capacity, diffusivity, emissivity, and density of CO 2 are different from N 2 . These properties impact combustion behavior, such as ignition, flame propagation, and flame stabilization.
Pulverized coal combustion is a multi-phase and multiparameter physicochemical process that adds further complexity to the practical implementation of oxy-fuel technologies. Comprehensive understanding of oxy-fuel combustion requires detailed experimental and numerical investigations spanning multiple orders of magnitude of scales in time and space, from complex real systems down to generic configurations. Experimental studies under wellcontrolled conditions resolving the behavior of individual particles are essential to understand the individual influence of the underlying sub-processes occurring during oxy-fuel combustion.
Another technology proposed to reduce the carbon footprint of electricity production is the power cycle using renewable metal fuels [12][13][14]. Metals, as an alternative carbonfree chemical carriers to hydrogen and ammonia, have great potential to store and transport renewable energy produced from wind and photovoltaics [14]. Previous studies on metal combustion have been comprehensively reviewed in [12,13,[15][16][17][18][19]. Among numerous metals, iron is a promising recyclable e-fuel [13]. The essential processes are schematically presented in figure 2. Energy can be stored by reducing iron oxides in either electrolysis using electricity from renewable sources or thermochemical reduction processes using green hydrogen produced from renewable energy. The heat can be released by direct combustion of iron particles with air in dustfiring or fluidized-bed processes [14], independent of time and space. Carbon-free, efficient combustion and abundant reserves of iron make it feasible to retrofit existing power plants with adaptations for flame stabilization and oxides collection [20]. The circular reduction-oxidation concept, using the clean recyclable iron fuel, has great potential for future energy systems that are carbon neutral but raises a lot of scientific questions concerning combustion characteristics and thermodynamic efficiency that request extensive fundamental research [19]. Compared to oxy-fuel combustion, iron combustion in a reduction-oxidation cycle is not well understood, and thorough systematic investigation is urgently needed.  [4,6], Copyright (2021), with permission from Elsevier. Sustainable and carbon-free energy supply of iron as renewable energy carrier in a reduction-oxidation cycle. Reproduced with permission from [13]. CC BY-NC-ND 4.0.

Scope of this review
This review is motivated by the global challenge of climate change and urgency to reduce CO 2 emissions in the electricity sector. As discussed in section 1.1, oxy-fuel and recyclable iron combustion have great potential to decarbonize electricity generation but raise plenty of scientific questions. Nonintrusive optical diagnostics are powerful tools that aid understanding and help answer these questions [21,22]. This review summarizes optical diagnostics for solid fuel combustion at the particle level, which requires that the particle's location and/or size is spatially resolved. This is important because physio-chemical interactions between particle and gas phase can be better understood if the particle is optically resolved. At the particle level, experimental studies are usually conducted in generic systems on a laboratory scale, where boundary conditions can be clearly defined. In this review, only in-situ non-intrusive methods for solid fuel particle combustion experiments are considered. Micrometer-sized particles with realistic heating rates in the order of 10 4 K s −1 are focused on as they are close to the realistic combustion environments of power plants.
Previous reviews on solid fuel combustion towards electricity generation have focused on different scientific aspects, providing essential insights into oxy-fuel coal combustion [4][5][6][7][8][9][10]23] and metal combustion [12,13,15,16,[18][19][20]24]. Supplementary to the previous work, this review focus on the particular aspect of particle-resolved optical experiments, which requires more investigation due to the lack of knowledge. For the sake of completeness, it should be mentioned that single particles are also relevant in other areas such as rocket propulsion or plasma sprays. However, a discussion of all these research areas is beyond the scope of this paper.
This review is structured as follows. Section 2 provides an comprehensive overview of particle-scale or particle-resolved experiments in the literature. The differences between single particle combustion and particle group combustion are discussed (section 2.1) and previous investigations are summarized in various tables classified by the fuel type (section 2.2) and the reactor used (section 2.3). Physio-chemical subprocesses of a single solid fuel particle in high-temperature oxidizing environments are described in section 3 to provide a basic understanding of combustion stages at realistic particle heating rates as well as relevant parameters of interest. This includes schematic introductions of combustion stages for a single coal (section 3.1) and metal particle (section 3.2) burning in hot oxidizing environments. Optical combustion diagnostics that address these parameters and processes are presented in section 4 using examples from the literature. Relevant experimental studies are summarized in tables 1-4 classified by type of fuels, reactors, parameters, and diagnostics. Section 5 summarizes scientific issues and indicates the research needed. The main results of this review are concluded in section 6.

Single and group particle combustion
Fundamental studies on solid fuel combustion can be categorized into single particle combustion (SPC) and particle group combustion (PGC). The key difference between these two modes is the influence of particle-particle interaction on transport processes and chemical reactions. Figure 3 provides a schematic representation of these two modes, depicting ideally spherical high-volatile particles burning in hot oxidizing gas atmospheres. In SPC, particles are isolated from other particles, and reactions depend only on the mass and heat transfer within the particle and at the particle-solid interface. In contrast, individual particles are influenced by neighboring particles in PGC, and the physical and chemical processes of each particle in the particle cloud are closely coupled. The intensity of this interaction mainly depends on the volume fraction of particles in the gas. Important measures of particle interactions include particle number density (PND) and inter-particle distances L pp (discussed in section 4.1). As particle interaction significantly alters flame behavior, previous experiments are classified based on SPC/PGC categories, which will be used consistently throughout this review.
Regardless of whether particles interact or not, the combustion processes at the particle level exhibit similarities between single and group combustion. Figure 3(c) illustrates some important sub-processes and parameters. Most of these are common subjects in fundamental research on solid fuel combustion. The thermochemical states in the solid, vapor, and gas Schematic illustration of (left) single-particle and (right) particle group combustion of high-volatile solid fuels in hot oxidizing gas atmospheres. Fundamental aspects and processes of solid fuel combustion at the particle level (middle).
phases and their temporal development are determined by various factors, such as solid fuel types, gas conditions (including temperature and species), flow conditions, and burner geometry. Typically, particles are non-spherical and have sizes in the order of 10-100 µm. For two-way coupling, particles subjected to laminar or turbulent flows will change their dynamics, which, at higher particle number densities, will impact gas dynamics in return. Combustion-relevant processes, such as heating, ignition, and volatile and solid oxidation, are closely correlated to particle and gas phase temperatures. Further detailed discussion on these topics is in section 3.

Solid fuel types
Solid fossil fuels can be characterized based on their degree of carbonization or coalification, the natural evolution process of buried plant matter into dense, dry, carbon-rich, hard material due to heat and pressure acting over time [25]. The degree of coalification determines the coal rank, of which there are four major types: lignite, sub-bituminous coal, bituminous coal, and anthracite, while peat is considered the precursor to coal. These types differ in color, hardness, and, fundamentally, elemental composition. The Van Krevelen diagram shown in figure 4 illustrates the change in composition from biomass through peat to coal and its various stages. In this diagram, the atomic H/C and O/C ratios have high values for solid fuel types with a low degree of carbonization (biomass and peat) and decrease with an increase in coal rank.
Solid fuels can be characterized not only by their elemental composition (ultimate analysis), but also by their proximate analysis which includes fixed carbon, volatile material, ash, and moisture content. These parameters are used to determine the lower heating value (LHV) of the fuel, which varies with coal rank. Anthracite and bituminous coals generally have  [26,27]. Reproduced with permission from [28]. CC BY-SA 4.0 higher LHVs than other fuels, while lignite typically has lower LHVs with comparatively higher volatile fractions. Biomass, on the other hand, typically contains a larger fraction of volatile material, often above 60% [29].
As discussed in section 1.1, metals are prospective chemical energy carriers for renewable and carbon-free electricity generation [13]. Utilization of metal fuels can be realized by direct combustion [12] or a wet cycle through metal-water reactions [18]. According to a systematic assessment [30], metal fuels, including Fe, Al, Si, Mg, and Ti are promising candidates, as they have higher energy per volume than many other energy carriers such as NH 3 , coal, diesel, and gasoline [13]. However, only Fe and Si have the potential to be recyclable fuels for direct combustion as their oxidation with air is mainly heterogeneous [12]. In particular, iron is the most promising choice due to its large reserves and low market price [20]. This work includes a review of particle-level experiments for iron combustion. Table 1 provides a summary of optical experiments conducted at the particle level using various solid fuel particles. Bituminous coal is the most studied fuel, followed by lignite and biomass. While single particle experiments dominate the literature, there has been some progress in particle group investigation. In recent years there have been significant advancements in iron combustion experiments by research groups such as Aldén [31][32][33], de Goey [34,35], and Dreizler [36,37].

Flow reactors
The most common laboratory experiments consider furnace conditions that lead to particle heating rates of around 10 4 K s −1 or higher. In order to achieve such high heating rates, particles are subjected to high temperatures, which can be achieved through electrical heating or flame exhaust gases. Due to laboratory restrictions, many experimental studies focus on laminar flow conditions. Several commonly used devices for such studies are shown in figure 5, including (a) drop tube furnace (DTF), (b) flat flame burner (FFB), (c) jet-incross-flow reactor, and (d) counterflow burner. DTFs are typically electrically heated and can provide wall temperatures up to 1600 K. The advantage of these furnaces is thee flexibility of being able to select gas compositions in the furnace. In contrast, other reactors use premixed or diffusion flames, which means that the atmosphere depends on flame products and adiabatic flame temperature. Compared to DTF, flame-assisted reactors can usually achieve higher gas temperatures above 2000 K. Table 2 provides a summary of previous optical experiments on solid fuel combustion using different reactor configurations. The DTF was first used for optical experiments on solid fuel combustion in the 1980s by Timothy [47], Levendis [103], and Wall [51] and their coworkers. The furnace was electrically heated, and particles were carried by gas and dropped at the top, falling along the furnace centerline. Optical access was usually available via a glass window on the wall. This reactor configuration has been widely used by different research groups internationally. In particular, Levendis' research group has made insightful contributions to the understanding of single particle oxidation processes in DTF through optical pyrometry and imaging measurements over the last two decades [39,55,57,97,129]. Reactors for solid fuel single particle and particle group combustion with high particle heating rates.
FFBs can be designed and operated in various ways, with the three major configurations being the Hencken burner, McKenna burner, and laminar flow reactor (LFR). The Hencken burner is a rectangular entrained flow reactor that utilizes a matrix of diffusion flamelets stabilized on the burner surface [46,130]. Particles are seeded through a central tube using carrier gas. One significant advantage of operating the Hencken burner is the wide range of gas temperatures and compositions that can be achieved by burning different fuel mixtures with the optional addition of inert gas. Flame flashback is not an issue since fuel and oxidizer are fed through separate tubes. This reactor concept originated in the Sandia National Laboratories and has been further developed by groups such as Schiemann [66,88] and Li [127,131]. Pioneering experiments by Molina and Shaddix [40,52] on this burner provided novel insights into combustion physics and motivated the years-long development of optical diagnostics in the field of solid fuel combustion.
The second type of FFB is the McKenna burner, which utilizes a porous sintered plate to stabilize a premixed flat flame  [31,[115][116][117]120] above the surface [132]. Particles are seeded through a central tube embedded in the porous plate, and the flat flame is shielded by an annular co-flow of ambient air. By adjusting the inlet gas velocity based on the laminar burner velocity S L , the flat flame can be either attached or lifted from the surface. The temperature of the plate can be controlled using an internal cooling circuit. Flame flashback is mitigated by the heat lost when the flame propagates into the porous sintered plate. However, careful consideration must be given to the radial homogeneity of species concentration and temperature [132]. The McKenna burner has been adapted by many laboratories for solid fuel experiments, including the groups of Costa [73,87] and Aldén [32,33,99]. The third type of FFB is the LFR, which employs a rectangular ceramic honeycomb structure to homogenize the inlet gas mixtures. An array of premixed flamelets forms an enclosed thin flat flame, which is stabilized on the honeycomb surface. Particles are introduced through a central capillary tube using a particle dispersion unit. The flame is shielded from the environment by a glass chimney, providing welldefined boundary conditions for gas temperature, composition, and velocity in the post-flame zone. The fused quartz glass ensures optical access in all directions, and the excellent flatness of the flame allows for precise determination of particle heating points. While the Hencken and McKenna burners are typically used as calibration burners, the LFR was originally designed for solid fuel combustion experiments by Dreizler and Böhm's group [70], and has been applied in single particle [36,81,83] and particle group [122][123][124] combustion experiments.
Jet-in-cross-flow reactors utilize the hot products of a premixed flat flame, sustained through a ceramic honeycomb, to create a gas environment with a uniform temperature and velocity field. Unlike the LFR, particles are horizontally injected into the atmosphere through a tube. This configuration has been used for single-particle visualization by Choi [43,67] and Mock [44,95,133]. Another jet-in-cross-flow configuration was used by Adamczyk [104,105].
In addition, Sawada et al [78,82] used a vertical counterflow burner, which forms a H 2 /O 2 diffusion flame from the bottom port that solid fuel particles from the top port pass through. The counter-flow burner is a less used configuration for solid fuel combustion, as indicated by table 2, which also considers experiments performed in other burners and reactor systems. For instance, Wu et al [74,117,120] employed an annular burner using two co-axial tubes, where the inner particle jet was surrounded by a pilot methane flame. Ning et al [34,35,109] also used two coaxial tubes but supplied them with non-reacting mixtures and ignited particles from the central tube with laser pulses. Other reactors, such as Bunsentype burners [31,134] and swirl burners [115,116], which usually focus on flames with higher particle seeding rates, are also employed.

Coal and biomass particle combustion
This section provides a basic outline of the physico-chemical processes involved in solid fuel combustion, using the example of a single micrometric high-volatile bituminous (hvb) coal particle subjected to a hot gaseous environment. Figure 6 illustrates this scenario by showing the particle and gas temperatures over the particle's residence time t. Note that the actual temperature and residence time depend on various boundary conditions, such as coal rank, particle size, and oxygen concentrations. The temporal profiles in figure 6 are for illustration purposes only.
Consider an individual hvb coal particle with a diameter of d p and an initial temperature T p of 300 K suddenly exposed to a high-temperature oxidizing environment. The particle undergoes rapid heating due to convective and radiative heat transfer, followed by devolatilization and various combustion stages. The duration of each phase and the accompanying particle temperature rise varies depending on the particle size, flow conditions, and fuel types. Here, we assume a homogeneous ambient gas temperature T g,amb of 1800 K and constant gas velocities. In most laminar flow studies, uniform temperature and velocity fields are desired. Figure 6 schematically presents the temporal behavior of the particle and gas temperature rise. Several essential stages of the combustion process are described below for high-volatile fuels such as hvb coal and biomass.
3.1.1. Particle heating Particle heating is an important phenomenon throughout the entire combustion process, and its rate is crucial to understanding the combustion of pulverized coal at the single-particle level. The heating rate is strongly influenced by particle size, thermal convection, and radiation near the solid-gas interface. During the pre-ignition stage, which includes evaporation and devolatilization, the current gas temperature T g is much higher than the particle temperature T p . Heat transfer from the gas to the solid phase first leads to moisture evaporation, which slightly cools down the surrounding hot gases. As a result, the local gas temperature slightly decreases from T g,amb , defined by the given boundary condition. Subsequently, devolatilization occurs as the particle temperature increases, and fuel gases, such as CH 4 , H 2 , and tars, are released into the gas phase. The particle heating rate significantly affects the duration of moisture evaporation, the rate of volatile release, and the composition of the released fuel gases. Temporal variations of the particle temperature can be described by the following expression, assuming a spherical particle, which considers the balance of the internal energy and heat transfer (adapted from [137]) Here, the convective and radiative energy exchange are represented byQc andQr, respectively. The variables d p , T p , ρ p , c p , and ε p correspond to the diameter, temperature, density, heat capacity, and emissivity of particles, respectively, while h represents the convection coefficient of heat transfer. The variables T g and ε g denote the temperature and emissivity of gases, respectively, and σ is the Stefan-Boltzmann constant. The heat required for evaporation and released from devolatilization reactions are denoted by Q evap and Q devol , respectively, with the latter often considered negligible prior to ignition. The heating rate is commonly determined through the temporally resolved measurements of the particle temperature, which often relies on two-or three-color pyrometry techniques [103]. However, obtaining temperatures lower than 1200 K can be challenging due to the insufficient quantum efficiency of conventional detectors in the infrared range, which will be discussed further in section 4.3. Therefore, experimental measurements of the particle heating rate are difficult in pre-ignition stages. To inspect the heating process, parameters such as particle diameter d p , velocity V p , and slip velocity V slip are often targeted in experiments. These parameters are extensively investigated using non-intrusive optical diagnostics (section 4). The burner wall temperature T w , required for estimating thermal radiation, can also be quantified experimentally using phosphor thermometry [136,138] or thermocouples.
The process of particle heating directly affects particle ignition and combustion. For example, coal particles with higher moisture content require more energy for evaporation, resulting in delayed homogeneous ignition time and extended volatile combustion time [86,126]. Larger particle diameters lead to lower heating rates [83], delayed ignition [41,81], and longer volatile combustion [75,82]. Additionally, a higher wall temperature has been found to reduce char burnout time and increase char temperature [54,55]. An increased slip velocity promotes ignition by enhancing convective heat transfer [81,139].

Ignition
For a bituminous coal or biomass particle with high volatile content, two combustion modes can be characterized: homogeneous and heterogeneous combustion. Homogeneous combustion, also known as gas phase combustion, occurs when a mixture of volatile and oxidizer is ignited. Heterogeneous combustion occurs through surface reactions at elevated temperatures. During devolatilization, a flammable gas mixture is formed by the molecular diffusion of volatiles and oxidizers into each other, which can be facilitated by flow convection. Under laminar conditions, an annular area with flammable mixture fractions can be expected in the vicinity of the particle, as illustrated in figure 6. Auto-ignition occurs when the gas mixture reaches a proper temperature and states within flammability limits. The ignition delay time τ i of a reactant gas mixture for auto-ignition, following the classical thermal explosion problem can be expressed as [140]: where c v is the specific heat capacity of the mixture, q c is the reaction heat release rate, T 0 is the initial temperature, T a is the activation temperature, Y F,0 is the fuel mass fraction, and A is the pre-exponential factor for reaction kinetics. The ignition time increases with specific heat capacity and is highly dependent on T a and T 0 . T a is defined as T a = E a /R, where E a is the activation energy for the reaction. The ignition mode of coal particles depends on various factors, including the combustion atmosphere, coal rank, and particle heating rates [45,89,135]. For hvb and biomass particles, homogenous ignition occurs first, releasing volatile matter, such as fuel gases and water, at increasing particle temperatures. However, as the particle heating rates decrease, transition from homogeneous to heterogeneous ignition mode is observed [89]. The homogeneous ignition time is influenced by several parameters, including oxygen mole fractions, particle size, PND, particle slip velocity, and the presence of other gases, such as water and CO 2 .
For instance, it was reported that the ignition delay time (τ i ) increased when N 2 was replaced by CO 2 due to the higher volumetric heat capacity (cv) of CO 2 (approximately 1.7 times higher than that of N 2 ) [40]. Similarly, in O 2 /H 2 O conditions, ignition is delayed due to the higher heat capacity of H 2 O, which is about 1.3 times that of N 2 [90]. However, if the initial mixture temperature (T 0 ) in a hot ambient is reduced by the particle-gas heat transfer, a different scenario may occur, wherein faster ignition occurs in CO 2 than in N 2 . In this case, the presence of CO 2 inhibits temperature reduction due to its higher heat capacity, leading to faster volatile release and formation of flammable mixtures. Consequently, the mixture could ignite even faster in CO 2 , as observed in experiments and numerical simulations [81].

Homogeneous and heterogeneous combustion
Homogeneous combustion of volatile matter from coal and biomass particles involves complex chemical reactions. Pyrolysis is the thermal decomposition of organic matter in the absence of oxygen, which primarily depends on solid fuel types, volatility, and heating rates. The main products of pyrolysis are hydrocarbons, including methane, ethane, propane, and butane, as well as small amounts of carbon monoxide and hydrogen. The released volatile matter reacts with oxygen to form carbon dioxide, water, and pollutants such as SO x and NO x . Gas phase reactions can lead to the formation of intermediate species such as formaldehyde, acetaldehyde, and acetone. At single particle level, an enveloping diffusion flame usually exists with released fuel gases and surrounding oxidizers mixing together. The flame structures of homogeneous combustion can be visualized by thermal radiation of soot particles, chemiluminescence (CL) or laser-induced fluorescence (LIF) of relevant gas phase radicals, such as OH and C 2 . To study SPC, flame luminosity or intensities of radicals, stand-off distances d sod , and burn time t vol are important quantities to characterize homogeneous combustion, whereas quantitative gas temperature and species measurements are challenging.
Heterogeneous combustion is a complex process that involves various reactions for char oxidation, which take place at elevated temperatures subsequent to or simultaneously with the gas phase reaction. The initial step is the adsorption of oxygen molecules onto the char surface, followed by a series of reactions. The most important reactions during heterogeneous combustion are: (1) the oxidation of char by oxygen to form CO 2 and CO, releasing heat at the same time, (2) exothermic gasification reactions where CO 2 and H 2 O react with the char surface to form CO and H 2 , and (3) a secondary reaction that occurs in the absence of oxygen, where carbon reacts with CO 2 to form CO. Several factors influence the rate of these char reactions, including particle temperature, oxygen mole fraction, and gas diffusivity. Experimental studies typically focus on the burn time t char and surface temperature T p as the most important parameters.
An increase in the oxygen mole fraction in the atmosphere can have various effects on volatile and char combustion. It has been reported that increasing the oxygen mole fraction decreases volatile burnout time [53,81] and char burnout time [54], while increasing the volatile and char temperature [55,135]. The volatile burnout time increases with the mass ratio of volatile matter, and the char burnout time increases with carbon content [39]. In addition, compared to an N 2 atmosphere, CO 2 extends the volatile combustion time due to its lower oxygen diffusivity [40].

Single iron particle combustion
Unlike coal and biomass particles, non-volatile iron particles primarily undergo heterogeneous combustion. The adiabatic flame temperature for iron oxidation is lower than its boiling temperature of around 3140 K, which means iron oxidation occurs mainly in the solid phase [13]. High-speed luminosity measurements from [33,37,109] indicate the surface temperature of a single iron particle as a function of residence time, which is schematically depicted in figure 7. Note that the measured temperature and residence time depend on various parameters such as particle size and oxygen concentrations, and the temporal profiles in figure 7 are for illustration purposes only.
In a scenario similar to that described for bituminous coal particles in figure 6, we consider an iron particle with a diameter of about 100 µm entering an oxidizing environment with a gas temperature of 1800 K. In the first stage, the particle is heated by convection at the solid-gas interface, resulting in an increase in surface temperature. Meanwhile, oxidizers such as vapor-phase water, oxygen, and radicals diffuse onto the particle surface and react with iron. At a relatively low temperature of about 800-1100 K, the particle ignites, leading to a rapid temperature increase due to exothermic oxidation reactions. Particle ignition represents a transition between kineticlimited and diffusion-limited combustion regimes [141]. As the temperature reaches the melting point of iron at 1811 K, the iron particle melts, starting from the surface. If the initial particle is non-spherical, a clear shape transition to a sphere can be observed with high magnification imaging [36]. During melting, the luminosity intensity of the particle shows a clear plateau. The correlation of this intensity plateau to melting has been experimentally confirmed by measuring the temperature [109] and the particle shape transition [37] as a function of time.
After melting, the temperature of the iron particle continues to rise until it reaches a peak temperature near 3000 K, which depends on the particle size and gas phase compositions. The high temperature approaches or may even exceed the iron boiling point, resulting in the formation of vapor phase iron and iron oxides in the vicinity of the iron core, which can lead to the possible formation of nanoparticles [36]. Subsequently, the particle cools down due to heat loss to the environment, although full oxidation might not be achieved during this stage, which is referred to as reactive cooling. During cooling, when the liquid particle solidifies, there is a second peak observed as an intensity jump [33,109]. This jump indicates a sudden temperature increase attributed to latent heat release during Fe 3 O 4 solidification [109]. Finally, the particle cools down further to the ambient gas temperature.
The surface temperature is a crucial parameter for understanding the multi-stage combustion of single iron particles. Ning et al conducted detailed experiments using pyrometry, that provided the temperature history of a burning iron particle in a cold flow [35,109]. Experiments conducted in hot conditions, supported by a gas flame, focus on different aspects, such as particle melting and nanoparticle formation [36], combustion stages [33], and the micro-explosion phenomenon [32]. Particle diameter, particle dynamics, ignition delay time, ignition temperature, and burn time are all critical parameters studied in these experiments. However, determining the ignition temperature can be challenging because it falls outside the measurement limits of conventional pyrometry, which will be discussed in section 4.3.

Other combustion phenomena
Several unique combustion behaviors have been observed during the SPC process, apart from those shown in figures 6 and 7. These behaviors include particle swelling, fragmentation, micro-explosion, and nanoparticle release. These phenomena are highly dependent on the specific conditions and fuel used and therefore require careful analysis and interpretation of experimental observations. This section aims to provide a brief introduction to these behaviors to help gain a basic understanding of their characteristics.
The rate of char oxidation is influenced by various factors, such as the size of the char particle, its surface area, and porosity. The swelling ratio is a commonly used parameter in devolatilization models. In the case of bituminous coal, particle swelling occurs during thermal pyrolysis, leading to an increase in particle size. This phenomenon is attributed to the growth and expansion of volatiles and gas bubbles, resulting in the formation of thin-walled cenosphere structures. However, at high heating rates, the volatile release time might be shorter than the time required for particle expansion, resulting in insignificant particle swelling [142]. Therefore, particle swelling is highly dependent on the particle heating rate and exhibits a two-regime characteristic, with the swelling ratio increasing first with heating rates and then declining. The turn-over heating rate for this behavior is typically around 0.5-1 × 10 4 K s −1 [142][143][144].
Zygourakis [145] reported that the swelling ratio and char porosity of bituminous coal increased with the particle heating rate in a wide range from 1 to 10 3 K s −1 . Fletcher [146] conducted sampling measurements at high particle heating rates above 10 4 K s −1 and suggested that mitigation of swelling and porosity occurs with increasing heating rates. They reported that the swelling ratio was also influenced by gas species, other than oxygen, present in the devolatilization or combustion environment. A further study on bituminous coal particles by Gale et al [142]. reported a rapid decrease of the particle swelling ratio as heating rates increased from 2 × 10 4 to 7 × 10 4 K s −1 . Based on their experimental observations, they proposed that swelling exhibits a two-regime behavior with increasing heating rates, and the turnover point was around 5 × 10 3 K s −1 . This two-regime behavior was also reported by Yu et al [147] for bituminous coal particles burning in a drop tube reactor. Recently, the swelling phenomenon was determined in-situ using the temporally and spatially resolved diffuse backlight-illumination (DBI) imaging technique at high heating rates [83]. The swelling ratio of bituminous coal particles was found to depend on particle size and revealed a temporal evolution during the volatile combustion process. The dynamic behavior of the swelling phenomenon was interpreted and combined with insightful information from numerical simulations.
Similar to swelling, primary fragmentation is caused by the release of volatiles after the particle is heated. As the volatile gases and bubbles form inside the particle, the pressure inside the particle increases. If this pressure becomes too high, the particle breaks up instead of undergoing elastic deformation which causes swelling to accommodate further volatile release [148]. Generally, fragmentation refers to particles breaking up into relatively large particles comparable in size to the parent particles. Primary fragmentation occurs during devolatilization, while secondary fragmentation occurs due to the weakening (by combustion) and breaking up (by collisions) of bridges connecting the elements of the char particle [149]. An earlier review of fragmentation was carried out by Chirone et al [149].
Primary fragmentation has been observed in drop tube reactor experiments [56,103,135]. These studies reported that lignite particles underwent frequent fragmentation, while subbituminous coal showed limited fragmentation. Fragmentation frequency for bituminous and anthracite coal was influenced by gas temperature, particle size, and residence time, as measured by Friedemann et al [150] using backlight imaging in a DTF reactor. However, no fragmentation was observed for brown coal (lignite) in their experiments. Wu et al [91] employed high-speed digital in-line holography to measure particle fragments in a McKenna burner, characterizing the velocity, size, and fragmentation modes for lignite coal particles. Their results showed that temperature and residence time influenced the probability of fragmentation occurring.
Micro-explosions have been observed for various metal powders, such as aluminum [151], iron [107], and composite metal [152]. This occurs due to gas dissolution or vaporization of the unreacted metal core, leading to bubble formation and pressure growth inside the particle. Wainwright et al [152] used x-ray imaging to visualize bubble growth inside Al:Zr micro-particles, and explained that micro-explosions were caused by the rapid growth of N 2 bubbles, which were dissolved in the liquid metal during combustion. Tang et al [151] measured particle temperatures to investigate the microexplosion of Al particles burning in a Hencken burner. The authors explained that micro-explosion was mainly driven by the vaporization of the unreacted aluminum core at high temperatures. Recently, micro-explosions of iron particles were visualized in different reactors [32,107,108,153,154]. Huang et al [32] utilized pyrometry and high-speed imaging to determine the particle temperature before and after microexplosion, as well as the velocity, respectively. However, the detailed mechanism of iron particle explosion is largely unexplored.
Nanoparticle formation in iron powder flames is a challenging issue in electricity generation applications. The classification of metal particle combustion as homogeneous or heterogeneous largely depends on the flame temperature T f and the boiling temperature T b of the metal and metal oxides involved. Vapor-phase homogeneous combustion has been observed for metals like magnesium and aluminum, which have a boiling point lower than the flame temperature. In contrast, iron powder is categorized as undergoing heterogeneous combustion since its predicted flame temperature of 2228 K is significantly higher than its boiling point of around 1800 K. However, recent in-situ [36,106,107,109] and ex-situ [35,154] experiments have reported the formation of nanoparticles during iron powder combustion. This is attributed to the actual surface temperature of the iron particles probably being higher than the boiling temperature of Fe and even close to the boiling temperature of FeO [35,155]. The heat generated from the exothermic heterogeneous iron oxidation reactions cannot be dissipated quickly enough to the gas through convention and diffusion, leading to a higher temperature in the iron core than in the gas phase. This enhances the evaporation of Fe and FeO, increasing their vapor pressure in the gas phase [35], which promotes the formation of nanoparticles.
Li et al [107] used time-resolved shadowgraphy and thermal imaging to observe the nanoparticle release from 70 µm iron particles burning in the exhaust gas of a McKenna burner flame. They also detected spectrally resolved vaporphase FeO * emission lines, which were not observed in similar experiments [35,36]. Ning et al [109] employed highspeed shadow imaging and pyrometry to study the correlation between nanoparticle formation and surface temperature for a laser-ignited iron particle burning in cold flows. Recently, Li et al [36] visualized the formation of nanoparticle clouds of iron particles burning in hot gases in detail using the spatiotemporally resolved DBI technique. Different topologies of nanoparticle clouds were observed, depending on the slip velocity, particle surface temperature (indicated by the simultaneously performed luminosity measurements), and particle diameter. The authors also hypothesized that a vapor-phase layer of Fe and/or FeO was formed near the surface of the iron core, which further oxidized, liquified and solidified to form nanoparticles.

Optical combustion diagnostics
The primary focus of this section is on the application of optical combustion diagnostics to study solid fuel combustion processes and their most relevant parameters. Given the complex and multi-phase nature of solid fuel combustion (discussed in section 3), a variety of parameters must be considered in experiments. This review only considers optical diagnostics at the particle level and focuses on spatially resolved particle locations. Table 3 summarizes the experimental studies that have been reviewed and categorized based on important parameters, such as particle size, number density, temperature, ignition/burn time, flame structures, particle and gas dynamics, and pollutants. Table 4 provides an overview of studies that used common optical imaging techniques in solid fuel combustion experiments.
The following sections describe a selection of experimental studies, with a particular focus on the methodology's ability to address parameters at the particle level. Due to space limitations, only representative examples selected from tables 3 and 4 are discussed in this review.

Particle number density
Distinguishing between SPC and PGC depends on whether the combustion processes of individual particles are affected by particle-particle interactions. Measures and PNDs.
Digital in-line holography (DIH) is a commonly used diagnostic approach for determining PND in aluminum [159,160], iron [31,161], and coal combustion [76,117,162]. Recent developments and applications of DIH in multiphase flows were reviewed by Huang et al [163]. In [31], DIH was applied to investigate the burning of iron particles within a methane-air Bunsen flame. A continuous wave laser was used that was attenuated, spatially filtered, and collimated to illuminate the particles, as depicted in figure 8(a). The particle holograms were recorded using a complementary metaloxide-semiconductor (CMOS) camera at a rate of 5 kHz and a magnification of approximately 1.5. Iron particles were seeded using a high-voltage particle dispersion unit, as shown in figure 8, and were also used in several iron and aluminum experiments [34,164,165]. The methane Bunsen burner was operated at various equivalence ratios to increase the oxygen volume fraction in the exhaust gas, which allowed for the individual iron particles to oxidize. Using a clusteringbased particle detection method developed by the same authors [166], the particle edge was reconstructed to derive the particle size and location. The resulting PND of iron powder suspended in CH 4 -air flames decreased with excess oxygen, which can be explained by the stronger thermal expansion effect due to the intensive iron oxidation reactions.
Li et al [122][123][124] utilized DBI to quantify the number density of bituminous coal particles in an LFR, as described in section 2.3. Figure 9 is a detailed schematic of the DBI imaging setup that has been used in other studies [81,83,122] as well. The setup involved a high-power pulsed LED (peak emission at 518 nm) equipped with an optical diffuser to illuminate particles from one side of the burner. The particle shadow was captured at 10 kHz by a CMOS camera at the opposite side, aided by a long-distance microscope that enabled high optical magnification to resolve particle size and shape. To suppress thermal radiation from the particle surface, a band-pass filter (centered at 525 nm with a full width half maximum FWHM of 25 nm) was employed to match the emission peak of the green LED. The LED pulse width was restricted to 1-2 µs to avoid motion blurring. Particle number N prt was detected using an adaptive threshold method (developed by the same authors [167]), which determined particle edges at the maximum gradient locations. PND and particle-particle distances were then computed after defining a 3D particle residence volume V, denoted by PND = N prt /V. An increase in PND reduced mean particle velocities. This can be attributed to a decrease in kinetic energy due to particle collisions and weakened thermal expansion, as the particles acted as heat sinks, reducing the gas temperature before ignition.
In line-of-sight methods like DBI, uncertainties in PND arise primarily from two sources: (1) estimation of the particle distribution volume V, and (2) determination of particle number N prt , which can be biased when particles overlap in a projection view. These two sources of errors were investigated to estimate PND uncertainties in coal PGC [122]. Figure 10(a) shows the increase in PND with particle number for different particle volumes V. The steeper increase in PND is due to the Figure 10. (a) PND (in mm −3 ) with increasing particle numbers Nprt for different particle residence volume V (different particle jet diameter d jet ). (b) Correlation of real particle number N prt,3D and projected particle number N prt,proj from a 3D particle simulation. Reprinted from [122], Copyright (2020), with permission from Elsevier. Reprinted from [124], Copyright (2022), with permission from Elsevier. use of a smaller injection tube (0.8 mm in diameter) in [122] compared to that (2.9 mm in diameter) used in [124], which reduced the particle residence volume V. As PND increases, V increases since particles expand outwards and occupy a large residence volume. Using a single average V obtained from different particle loadings underestimates the real PND at lower N prt and overestimates it at higher N prt . However, this effect is negligible compared to the influence of the injection tube geometry. Figure 10(b) shows the estimation of the error in N prt by performing 3D particle simulations. In these simulations, particles were generated using diameters and aspect ratios obtained from experiments and randomly distributed in a space V. These synthetic particles were then projected onto a 2D view resembling a DBI camera with the real pixel resolution used in the experiment. By applying the same particle detection method, the number of projected particles N prt,proj was compared to the number of 3D-dispersed particles N prt,3D . For N prt over 100 (PND ≈ 0.7 mm −3 ), the particle number was gradually underestimated in a 2D projection, and the discrepancy increased with an increasing number of particles: up to 20% for 250 particles. These uncertainties should be carefully considered when interpreting experimental observations in PGC.
The transition from single particle to PGC occurs as the inter-particle distance decreases which can be evaluated based on the PND [122]. Despite a large number of investigations on single particle experiments, further research is needed on PGC, as indicated by the limited number of publications in tables 3 and 4. One reason for this is the practical difficulties involved in conducting controlled and wellcharacterized PGC experiments. Efforts have been made by Levendis and coworkers [168,169] to visualize the transition from single particle to PGC using polymer and pulverized coal particles. Recent studies utilizing an in-situ determined PND have enabled a more detailed evaluation and a deeper understanding of the transition process. Non-flammable Figure 11. Determination of the optical resolution for particle sizing by the optical transfer function from a Siemens star pattern. Reprinted from [83], Copyright (2022), with permission from Elsevier. regions appear inside the enveloping flame as the PND increases [122]. Non-flammability becomes more pronounced when the PND exceeds a limit of approximately 0.37 mm −3 under the investigated conditions [122], which corresponds to an inter-particle distance of 4d prt . This critical distance is consistent with the stand-off distance of a single-particle volatile flame [123], indicating the minimum spacing required for marginally interacting particles.

Particle size and shape
The size and shape of micro-sized particles can be spatially resolved in single-shot or high-speed imaging using techniques such as DBI or shadowgraphy, provided that the camera magnification is sufficiently high. For example, Ning et al [109] used high temporal and spatial resolutions to study the size evolution of laser-ignited iron particles and identified different behaviors in particle size transitions. Li et al [83] investigated the swelling behavior of bituminous coal particles by evaluating their size growth and characterizing the swelling regime depending on the particle heating rates. They also studied the rotational motion of non-spherical particles by approximating the particle shape with ellipse fitting and tracking the orientation over time.
In experiments that involve particle size and shape measurements, optical resolution is a crucial aspect. Li et al [83] evaluated the resolution capability of their DBI setup using a Siemens star target. The Siemens star is composed of a pattern of bright and dark spokes emanating from a single point. The gaps between spokes decrease towards the center of the star, making it useful for determining the resolution of a detection system. In [83], a pattern with 36 spokes was imaged at various depth positions, and the optical transfer function was evaluated to quantify the imaging resolution. Figure 11 shows that the spatial resolution is a function of imaging sharpness. At the focus plane, the highest resolution value of 51 lp mm −1 was achieved based on the Rayleigh criterion. Displaced away from the focus plane, the spatial resolution decreases in both forward and backward directions. With the well-characterized imaging system, the size and shape of bituminous coal particles were determined in DBI measurements, which were further used to quantify particle swelling effects. All bituminous coal particles swelled during volatile combustion and then shrunk. The swelling ratio was found to be a function of time, and the peak swelling ratio decreased for larger diameters with lower particle heating rates.
The accuracy of particle sizing depends significantly on the imaging processing algorithm employed. Conventional edge detection algorithms, such as the Canny method, detect the particle boundary based on local intensity maxima [170]. More effective algorithms are required when micro-sized particles are closely associated with interfering signals from soot or nanoparticle clouds. For instance, Li et al [123] used an adaptive threshold method, first proposed for flame front detection in turbulent gas flames [167,171], to define the boundaries of particle clouds. In this method, particles were imaged using a high-speed DBI system with a pixel resolution of 9.2 µm, corresponding to a magnification of 2.17. The intensity gradient maps were computed from the 2D particle shadow images, sorted in ascending order, and an iso-contour was drawn on the gradient map by setting a percentile value of 10%, which represented the top 10% gradient magnitudes. An average intensity within the gradient iso-contour was then determined and used to localize the particle boundary. The apparent particle edge represented the significant gradients and performed robustly even for 100 µm coal particles surrounded by heavy soot formation.
More recently, Ning et al [108] applied the light attenuation (LA) method to determine the size of iron particles sieved in the range of 45-55 µm. They employed a DBI system that consisted of an LED illumination at 395 nm and a high-speed camera for particle sizing. The particle shadow image was normalized by the background intensity and then intensity-inverted, revealing particle intensities close to unity, while the background noise was around zero. The laser attenuation was computed as an integration of pixel intensity over the particle, and the particle diameter was obtained by setting a threshold value for the intensity integration. This method delivered robust results, even when nanoparticle clouds were present, and was not sensitive to the background intensities. However, the laser attenuation method can only obtain a circle-equivalent diameter, whereas edge detection allows for particle shape and sphericity characterization.
One way to validate the accuracy of particle sizing is to use a precise shadowgraphy target, such as the LaVision one for shadow imaging techniques [123]. This target has pre-defined circular non-transparent dots in a large diameter range printed on transparent glass, making it suitable for back-illumination within the particle region of interest. The dot diameters can be evaluated using the same detection system and algorithm and compared to the defined values. Figure 12 shows an example of such an assessment with a group of circular dots measured using laser attenuation [108] and the adaptive edge detection  [108] and adaptive edge detection approaches [83]. The red numbers on the x-axis denotes the ground truths of the target diameters. (c) The relative measurement error as a function of measured particle diameter. Reproduced with permission from [37]. CC BY-NC 4.0.
method [123]. The dot pattern detected by the LA method is highlighted by red circles. It shows that both methods can measure all dots that are larger than 10 µm. By summarizing the diameter distribution from 100 independent measurements, the LA approach has a higher measurement certainty than the edge detection method, as indicated by the more narrowly distributed results. However, the adaptive edge detection method delivers a more accurate mean diameter than the LA method. The relative measurement error of the LA method increases exponentially with decreasing dot diameters, as shown in figure 12(c). This error behavior needs to be considered for in-situ particle sizing and can be corrected by introducing a correction factor formulated by fitting the measurement error with an exponential function, as performed in [37].

Particle surface temperature
Non-intrusive methods for measuring surface temperature typically rely on spectral characteristics of thermal radiation, which are described by Planck's law, and the emission is often assumed to follow the grey body assumption. One widely used technique for such measurements is optical pyrometry, which is based on detecting intensity ratios from two or more wavelengths. The popularity of this technique is reflected by the number of publications listed in table 4. To provide a better understanding of the technical limitations and measurement uncertainties, it is necessary to briefly discuss how temperature is evaluated using pyrometry before diving into some of the experimental studies in the literature.
Pyrometry measurements are based on the thermal radiation theory, described by Planck's law, which state that the spectral radiation density B(λ, T) is a function of wavelength λ and temperature T: where h is the Planck constant, c is the speed of light, and k B is the Boltzmann constant. Two constants, c 1 = 2hc 2 and c 2 = hc/k B can be applied to simply equation (3).
Taking two-color pyrometry as an example, the method involves using two detectors (such as a photodiode or camera) to capture the thermal radiation intensity of a surface at two different wavelengths, λ 1 and λ 2 . The intensity S recorded by the detector is an integration over a narrow bandpass filter with bandwidths of ∆λ 1 and ∆λ 2 . For example, S λ1 represents the intensity recorded by the camera for the wavelength λ 1 : where ϵ λ1 is the emissivity and η λ1 is the overall efficiency for converting photons into camera intensities. However, the emissivity is temperature and wavelength dependent for a real body, and for practical purposes it is usually assumed that emissivities at the same temperature and different wavelengths remain approximately constant, ϵ λ1 (T) ≈ ϵ λ2 (T). This gray body assumption is often the main source of uncertainty in most pyrometry measurements. The efficiency η λ1 includes the transmittance of optics, geometry of the object, and quantum efficiency at the detector [103]. Therefore, the signal ratio at a temperature T can be expressed as: The properties of the optical elements used and the quantum efficiency of the sensors can be summarized into an instrument constant C λ1,λ2 . To obtain C λ1,λ2 , calibration with a light source with known emissivity and temperature is required beforehand. From equation (6), the surface temperature can be derived using two approaches. The first approach uses the Wien approximation: exp( c2 In this case, the temperature can be computed as: The second approach deduces temperature directly from equation (6) [103]: Since T appears on both sides of equation (8), a solution can only be determined through iterative methods, which start with an initial value obtained from equation (7) under the Wien assumption. This value is then used as an input to the right-hand side of equation (8) to update the temperature until convergence is achieved after a few iterations. These two approaches usually show good agreement for temperature calculation. Nevertheless, the Wien approximation is more often adopted due to its simplicity. For temperature calculation using equation (7) or equation (8), the gray body assumption is commonly used, namely, ϵ λ1 = ϵ λ2 . In the following sections, several representative pyrometry setups and experiments are discussed, including those related to coal and metal particle combustion at the particle level.
Murphy and Shaddix [111] conducted experiments using two-color pyrometry to measure the temperature of coal chars in oxygen-enriched atmospheres generated by a Hencken burner. The burner was operated with hydrogen and ethylene as fuel and provided different furnace oxygen concentrations (i.e. 6-36 mol%) and temperature levels (i.e. 1600-1800 K). Their setup, as shown in figure 13, combined simultaneous particle sizing, collection, and two-color pyrometry techniques. A HeNe laser induced light scattering of the particle in focus, triggering data acquisition. The pyrometer consisted of two photomultipliers equipped with bandpass filters (FWHM = 40 nm) with central wavelengths of 550 nm and 700 nm. The selection of wavelengths and filters avoided interfering signals from sodium and potassium emission lines. Additionally, particle sizing was performed by recording the 700 nm emission signal using the coded aperture geometry used in [156]. The char particles were collected using a helium-quench, water-cooled sampling probe, and were subjected to chemical and physical characterization. The combination of these diagnostics enabled the study of high-temperature char kinetics, originated from one sub-bituminous and one bituminous coal type in the range of 106-125 µm.
Levendis and coworkers [54] conducted research on coal char combustion using a two-color pyrometer in a laminar electrically-heated DTF, as shown in figure 14. The furnace was operated under various conditions with wall temperatures ranging from 1300 to 1500 K and oxygen concentrations from 21vol% to 100 vol%. The pyrometer used a bifurcated optical fiber bundle with a collimating lens and an aperture on the top to collect signals. The signals were then split and focused on two silicon-based, amplified photo diodes. Two filters centered at 800 nm and 1000 nm were used before the detector with a bandwidth of 70 nm (FWHM). This setup was initially developed by Levendis et al [172] and was derived from the system by Timothy et al [47]. Over the decades, the setup has been continuously improved, for example, by extending it to a three-color pyrometer using wavelengths of 640 nm, 810 nm, and 998 nm, allowing measurement of a wide temperature range [55,103]. Recently, high-speed imaging was incorporated to enable direct visualization of particle flames. Bejarano and Levendis performed various pyrometry measurements for coal [39,57,129,173], biomass [42, 93, 97, 100,135], and iron particles [174] under different N 2 /O 2 and CO 2 /O 2 atmospheres.
Schiemann et al [63,65,79] developed a stereoscopic camera system for thermography (SCOT) to enable multiparameter combustion analyses in a Hencken burner. The SCOT system was an extension of their earlier experiments on a DTF, which relied solely on imaging [63,94]. As depicted in figure 15, the SCOT system is composed of four chargecoupled device (CCD) cameras that are coupled with longdistance microscopes (Questar QM1). Camera 1 and 2 form an optical two-color pyrometer, in which incoming thermal radiation is first split by a long-pass dichroic mirror at 700 nm and then imaged onto two cameras equipped with bandpass filters centered at 785 nm (FWHM = 62 nm) and 650 nm (FWHM = 100 nm). The wavelengths were selected to avoid signal contamination by Na and K emission lines. In the other branch, camera 4 forms a back-illumination system along with a bright lamp on the opposite side of the burner, enabling particle size and shape measurements by recording shadowgraphic images. Camera 3 captures the luminosity of burning particles in a double-frame mode to estimate particle velocity via particle tracking velocimetry (PTV) and 3D particle shape approximation supported by camera 1 (two-view 3D reconstruction). This multi-parameter detection system enabled the construction of a comprehensive experimental dataset for the SPC of bituminous coal and torrefied biomass under various conditions [80]. Recently, the same approach was used to investigate the PGC of bituminous coal particles, supported by additional particle sampling measurements [85].
In iron combustion research, Ning et al [35] utilized time-resolved two-color pyrometry to measure the temperature of single iron particles at a kHz temporal resolution. Simultaneous high-speed temperature and luminosity measurements were supported by time-integrated spectral measurements for particle peak temperatures, as illustrated in figure 16. The particles were ignited using a diode laser and burned in various oxygen concentrations at room temperature. The thermal radiation from a burning particle was split equally between two SA3 CMOS cameras using a 50/50 beam Figure 16. Schematic of simultaneous high-speed two-color pyrometry, luminosity imaging and spectrum measurements for laser-ignited iron particles. Reproduced from [35]. CC BY 4.0. splitter cube. Two bandpass filters, centered at 850 nm and 950 nm (FWHM = 10 nm), were chosen to increase sensitivity at the lower temperature range of the particle. The pyrometric ratio and particle temperature were obtained by integrating signals over the bright particle in each camera. Combustion stages, such as melting, oxidation, reactive cooling, and phase transition, were identified and correlated with particle temperature using synchronized luminosity imaging. For example, melting was recognized by the luminosity plateau, which revealed an average temperature that stayed relatively constant at different oxygen mole fractions and quite close to the iron melting point (figure 10 in [35]). In their subsequent studies, time-resolved pyrometry was combined with shadow imaging to investigate particle size evolution [109] and nanoparticle formation [108].
In pyrometry experiments, calibration is a crucial step that involves determining the instrument constant or calibration coefficient, C λ1λ2 , in equation (6). This is typically achieved by measuring the signal ratio of an object at a known temperature T and emissivities ε 1 and ε 2 . For example, Bejarano and Levendis [54] calibrated their pyrometer by melting wires of various metals with known melting temperatures, including copper (1358 K), platinum (2043 K), and tungsten (3683 K). They also used a tungsten lamp calibrated by National Institute of Standards and Technology (NIST) which allowed calibration for a wide temperature range with adjustable set points. The work in [55,103,129] shows that calibration with melting wires was more accurate than with a tungsten lamp. Murphy and Shaddix [111] and Schiemann et al [63] calibrated their pyrometer using a black-body source with emissivity near to unity at different temperatures.
Spectral emissivity is a critical issue in pyrometry measurements since it varies with material, phase, and wavelength. Levendis et al [103] summarized the emissivity at the melting point temperature of the metals they used to calibrate their three-color pyrometer. Schiemann et al [63] Figure 17. Spectral emissivity of tungsten at temperatures from 1400 K to 2800 K and wavelengths from 380 nm to 2600 nm using analytic expressions [176] fitted to experimental data in [177]. calibrated their high-resolution two-color pyrometer using a black-body source at Pyhsikalisch Technische Bundesanhalt (PTB) with an emissivity near 1. However, for the commonly used tungsten lamp, numerous emissivity data have been reported in figure 234 of [175], indicating that wavelength and temperature dependency may be problematic for accurate calibration. Pon and Hessler [176] provided an analytic expression of the spectral emissivity of tungsten based on experiments performed by de Vos et al [177], covering temperatures from 1400 K to 2800 K and wavelengths from 380 nm to 2600 nm. As shown in figure 17, the emissivity of tungsten decreases with wavelength and shifts with temperature. For example, in a two-color pyrometer with channels centered at 850 nm and 950 nm, the ratio ϵ 850 /ϵ 950 is approximately 1.05 in the temperature range between 1400 K and 2800 K. Neglecting the wavelength dependence of emissivity in a limited spectral range in the calibration with a tungsten lamp can lead to significant experimental errors.
It is useful to discuss whether the gray body assumption is valid for real objects in experiments. Mitchell and McLean [178] examined the gray body assumption of sub-bituminous coal char by comparing measured emission signals as a function of wavelength to that from a gray body. Their measured signal ratio from 550 nm to 850 nm to the signal at 850 nm showed good agreement with the gray body assumption at a mean temperature of 1700 K. However, slight discrepancies were still observed in their logarithmic plot (refer to figure 2 of [178]). Schiemann and his coworkers also conducted a series of experiments to measure coal char emissivity in the infrared range. They observed that spectral emissivity depends on particle size, rank, temperature, and combustion atmospheres [66,69,179,180]. The emissivity of char has also been experimentally studied by Wall and Becker [181] through spectral band emissivity measurements. These findings improve our understanding of radiative heat transfer in the infrared range >1 µm. Nevertheless, the significant variation in emissivity still challenges the gray body assumption and the accuracy of pyrometric temperature measurements. For iron and aluminum, emissivity varies largely with wavelength, as documented in [175]. Regarding iron oxides, an obvious variation in emissivity in the infrared spectrum was observed as oxidation time (layer thickness) or temperature changes [182]. Unfortunately, emissivity data on char and iron oxides are limited in the literature, making it difficult to reduce uncertainties for pyrometers using conventional cameras and detectors. This issue becomes more critical for temperature measurements when soot particles or nano-sized metal oxides form and evolve, as quantification of the spectral emissivity during such a complex transient combustion process is very challenging.
The design of an accurate pyrometer requires careful selection of wavelengths and filter bandwidths. Two or three wavelengths should be selected to ensure that the signal ratio S λ1 /S λ2 properly reflects temperature changes with high sensitivity. These wavelengths should be spectrally close enough to satisfy the gray body assumption and minimize measurement uncertainty. The selection of wavelengths should avoid any elemental emission lines, such as Na (589.0 and 589.6 nm) and K (766.5 nm and 769.9 nm) in coal and biomass combustion, and CL of species such as OH * (306 nm), CH * (314 nm, 390 nm, and 431 nm), and C 2 * (470-550 nm) in hydrocarbon flames. In iron combustion, FeO emission peaks at around 600 nm [183] are possible interference signals. Additionally, larger wavelengths towards the infrared spectrum are preferred since the wavelength for the thermal radiation peak λ peak increases at lower temperatures. For example, according to Wien's displacement law, λ peak = 1.16 µm can be calculated at a given temperature T = 2500 K, where λ peak · T = 2898 µm·K. However, CCD and CMOS sensors are only sensitive in the visible and near-infrared range, typically from 400 nm to 1000 nm with decreasing quantum efficiency towards the infrared spectrum, making it practically impossible to measure very low temperatures via pyrometry, with the usual lower limit being around 1200 K. Finally, it has been reported that bandwidths are not critical [184], and they should be defined to optimize temperature sensitivity and the signal-to-noise ratio, with typical FWHM values ranging from 10 nm to 150 nm.
Other parameters, such as exposure time, aperture, and focus quality of pyrometer detectors, also influence measurement accuracy and should be optimized according to the requirements in the experiment. Additional sources of uncertainties, particularly when measuring coal particle temperatures, are discussed in [129].
Molina and Shaddix investigated the ignition and volatile combustion of hvb coal particles in the exhaust gas of a Hencken burner [53]. The experimental facility was described in more detail in [111], and the setup is shown in figure 13. The reactor was operated under N 2 /O 2 and CO 2 /O 2 atmospheres with oxygen mole fractions of 21 vol% and 30 vol% in the post-flame region at a gas temperature of approximately 1250 K. The CH * CL was recorded using a CCD camera equipped with a bandpass filter centered at 431 nm. The camera was operated with a long exposure time of 1 s to capture time-averaged CH * profiles of individual particles. An example of the CL signal from a burning hvb particle is shown in figure 18. The ignition time (τ i ) and volatile combustion time (τ vol ) were determined by fitting the CH * intensity profiles with a 2nd-order polynomial and identifying the intersection point with the background signal (baseline). The results indicated that CO 2 increased the ignition delay time, which was attributed to the difference in heat capacity of N 2 and CO 2 . While a lower fuel diffusivity could lead to reduced consumption rates, the volatile combustion time was not affected by CO 2 . A higher oxygen concentration shortened τ i , but oxygen enrichment did not measurably impact τ vol . The particle ignition delay was significantly longer than the volatile combustion time, indicating that the particle heating rates were relatively low at the reported gas temperature of 1250 K.
Köser et al pioneered the use of OH-LIF imaging to visualize ignition and volatile combustion of single coal particles with high temporal resolution [61,70]. They extended this technique in a subsequent study [75] to simultaneously determine particle ignition time τ i , volatile combustion time τ vol , and particle sizes d p of bituminous coal particles in the exhaust gas of a pre-mixed flat flame burner using spatiallyand temporally-resolved measurements. The experimental setup included 10 kHz planar OH-LIF, luminescence imaging, and single-shot DBI measurements from two projected views. Particle shadow images from two 90 • arranged views provided information on aspect ratio and circle-equivalent diameters, which were statistically similar. By applying a band-pass filter of 380-492 nm, the luminescence signals were mainly attributed to the thermal radiation of soot, tar, and coal particles themselves. Figure 19 compares OH-LIF and luminescence signals on and around a single coal particle at different stages of volatile combustion: ignition, middle stage, and end of the volatile flame. The results show that OH-LIF is a more accurate method for ignition time measurements. For example, using luminescence signals for ignition, the ignition was observed to occur 1.2 ms later than that from OH-LIF measurements for the particle shown in figure 19. However, luminescence imaging was an accurate technique for determining the end of volatile combustion, whereas OH-LIF signals remained visible in high-temperature products from the volatile combustion, as indicated by event C in figure 19. The authors concluded that OH-LIF and luminescence should be used to detect the beginning and end of volatile combustion, respectively, and that combining both techniques can reliably determine volatile burn time.
Li et al [81] followed the methodology proposed by [75] and extended the experimental setup to include simultaneous OH-PLIF, luminosity, and DBI measurements at 10 kHz using the same LFR. They studied bituminous coal particles with average sizes of 120 µm and 205 µm in N 2 /O 2 and CO 2 /O 2 atmospheres with increasing levels of oxygen enrichment (10-40 vol%). Sequential combustion stages, such as the pre-ignition phase, volatile combustion, and char combustion, were identified using OH-LIF and luminescence signals, as shown in figure 20. The termination time t vol,end of the volatile flame can be clearly identified at the minimum of the luminescence signal surface area A LU , which is visualized in figures 20(b)-(e).
Similar to [75], it was found that the luminosity measurement largely overestimated gas phase ignition times by several milliseconds since the thermal radiation dominated the signal intensity [81]. Gas phase ignition time t ign can be accurately captured by following the intensity and structural changes of OH-LIF signals in the vicinity of coal particles. Ignition is defined as the time when a volatile flame revealed higher OH-LIF intensity against the background. Due to the weak and noisy OH-LIF signals at the onset of ignition, an image analysis algorithm is crucial for accurate ignition detection. It is challenging to algorithmically determine when evident ignition happens based on the initial size and intensity growth of OH-LIF signals. In a subsequent investigation [84], advanced object recognition algorithms, as well as deep learning approaches, were applied to improve the accuracy of ignition detection. It was found that the blob feature detection algorithm and ResNet object detection models significantly reduced errors in the ignition delay time. They concluded that OH-LIF is a suitable technique for accurate detection of homogeneous ignition, which is enhanced by using sophisticated feature detection algorithms [84].
One of the objectives of the experiments conducted in [81] was to provide validation data for numerical simulations. It was important to characterize the experimental boundary conditions carefully. OH-LIF combined with laser absorption spectroscopy and PIV were used to measure gas temperature and velocity, respectively. Particle velocity was determined by analyzing DBI data using PTV. The collected data facilitated a comprehensive comparison between the experiments and detailed simulations conducted within a Eulerian-Lagrangian framework with finite rate chemistry. The experiments showed that the homogeneous ignition process was slightly accelerated in a CO 2 atmosphere, which was explained numerically by the difference in the local gas temperature before ignition. During heating, the particle acted as a heat sink due to solid-gas heat transfer, leading to a reduction in temperature in the gas surrounding the particle. Hence, the temperature in a CO 2 atmosphere fell less due to the higher heat capacity of CO 2 , promoting the ignition process. Both ignition delay and volatile combustion time decreased as the oxygen mole fraction increased. Additionally, increasing particle diameters prolonged ignition delay t ign and volatile combustion duration t vol , which was attributed to the slower particle heating and lower volatile release rates.
Khatami et al [135] performed a study on SPC behavior in an electrically heated drop tube reactor with a wall temperature of 1400 K. Three-color pyrometry was used to track the temporal evolution of single particles (and associated volatile shooting flames) from a top projection view, similar to [55,172]. In [129], temperature-time profiles for five fuels were obtained, including bituminous coal, sub-bituminous coal, lignite, and biomass particles, in different oxygen mole fractions (20%-100%) in N 2 /O 2 and CO 2 /O 2 atmospheres (refer to figure 3 in [135]). SPC was categorized into one-mode combustion, in which homogeneous and heterogeneous combustion could not be clearly distinguished, and two-mode combustion events, in which homogeneous combustion occurred first, and heterogeneous ignition followed. The researchers reported that combustion modes depended on many factors, such as coal rank, oxygen mole fractions, and N 2 or CO 2 atmosphere. Additionally, the particle burnout time was derived, indicating a significant reduction with increasing oxygen and a slight reduction in N 2 compared to CO 2 .
Temperature-time profiles obtained from various studies, such as those conducted in the Levendis laboratory [39,55,93,135], have provided valuable insights into the different stages of SPC. For example, in [55], distinct pyrometric profiles of single bituminous, lignite, and char particle combustion were observed (see figure 21). The volatile and char combustion stages can be differentiated for bituminous coal particles, but are temporally superposed for lignite particle combustion. The two-stage behavior of bituminous coal was mainly attributed to soot formation in the volatile enveloping flames. High-speed luminosity photographs showed sooting flames forming for bituminous coal in different oxygen mole fractions, but not for lignite coal [56,135]. Chemical composition analyses indicated higher oxygen and lower carbon content in lignite coal [55,135,136], leading to non-sooting volatile flames despite its higher volatile content than bituminous coal. High-speed planar OH-LIF measurements [136] showed similar ignition behavior for bituminous and lignite coal particles. However, the volatile combustion stage might not be distinguishable in pyrometrically determined temperature profiles because thermal radiation is weak in non-sooting volatile flames. Therefore, LIF-based techniques such as OH-LIF and PAH-LIF, are necessary to better identify the volatile combustion stage and measure ignition delay in lignite and biomass combustion. For non-volatile particles, including char and metals such as silicon and iron, temperature profiles provide reliable burn times.
Ning et al [34] conducted experiments on the burn time and combustion stages of single laser-ignited iron particles in gas flows at room temperature and various oxygen mole fractions. They used a high-voltage particle generator to seed single particles into cold flows and ignite them with a diode laser. The high-voltage generator was developed for lifted aluminum jet flames [164], and a similar concept was employed to study single aluminum particle combustion in hot gas flows [165]. The authors dispersed metal powder between two horizontal electrodes applied with a high DC voltage of several kV. Particles were transported by a carrier gas flow along a capillary tube to the reactor exit surrounded by a laminar coflow. After laser ignition, particle luminosity was recorded using a high-speed camera and a color camera to visualize  the entire oxidation process (see figure 22). Based on the luminosity-time profiles, oxidation time until the luminosity peak t max and the entire oxidation time t tot were determined for different particle sizes and oxygen mole fractions. It was reported that the first combustion stage time t max was sensitive to oxygen concentrations, indicating that iron combustion was mainly limited by oxygen diffusion. After the luminosity peak, the combustion time was not significantly affected by the oxygen concentration. The entire combustion time t tot was well described by a md n power law, with the exponent decreasing at higher oxygen concentrations.

Volatile flame structures
The visualization of single-particle gas phase flames dates back to the 1980s, when luminescence imaging [48,185] or the Schlieren method [49] were used. In recent years, techniques such as luminescence [43, 44, 95-99, 131, 133, 186] and LIF techniques [28,61,70,75,116,[121][122][123][124]136] have been used to measure the structures of particle-loaded flames, similar to the approaches used for ignition and burn time measurements. Laser-induced incandescence (LII) [116,131] and Mie-scattering [40,116,124,131] of soot particles have been used to study high-volatile hydrocarbon fuels. Select examples from the literature are discussed below, including recent developments in volumetric imaging approaches that aim to resolve the intrinsic 3D nature of gas phase flames.
High-speed luminescence imaging has been extensively utilized to provide visual and direct impressions of singleparticle flames, complementing temperature-time profiles in pyrometry measurements [39,56,57,64,89,93,97,100,135]. In some cases, particles are back-illuminated, enabling their shadows to be observed with the luminous flames in the same image. For instance, Panahi et al [96] employed a highspeed and high-resolution camera to study single miscanthus and beechwood particle combustion in a DTF heated to a wall temperature of 1400 K. To enhance the optical magnification, a long-distance microscope lens was attached to the camera, which was operated at 2 kHz. Figure 23 shows a time sequence of an initially non-spherical miscanthus particle during volatile combustion. A luminous envelope flame, burning around the particle, was formed by gas phase volatile oxidation. The initial particle was about 1.7 mm × 0.16 mm in size and highly non-spherical (needle-like in shape) which shrank, melted, and spherodized as the particle devolatilized. This kind of shape transition has been consistently observed for raw and torrefied biomass particles. The flame size was characterized and appeared to be typically 3-5 times larger than the char size.
Lee and Choi [43,67] investigated the volatile flame structures of individual bituminous coal particles in a jetin-cross-flow reactor. The reactor was supported by a laminar premixed propane flame to generate an initial hot gas flow with varying oxygen mole fractions (16.7-40.2 vol%) and temperatures up to 1024 K. Bituminous coal particles of different sizes were introduced horizontally using a cold carrier flow. High-magnification luminescence imaging was employed to observe the flame structures, facilitated by extension rings, and with continuous backlighting provided by a halogen lamp. Figure 24, illustrates a luminous particle envelop flame and the methodology employed for analyzing the flame structure. The luminescence intensities were first normalized to the maximum and then separated between coal particle and volatile flames using a threshold. The volatile flame area A f , effective flame radius r eff , and elongated distance r c (a distance from flame center to particle center) were derived to describe the volatile flame structures. The flame size increased, and the flame shape became more elongated as the oxygen mole fractions decreased. This phenomenon was likely due to lower reaction rates in the locally fuel-rich diffusion flame, which elongated downstream of the particle due to particle slip velocity. In addition, hot spots were observed on the particle surface that initiated char oxidation, which then rapidly spread to the entire surface. This analysis of the flame structure provided valuable insights into the evolution of volatile flames and was extended to the study of various biomass particles in the same reactor, as performed by Mock et al [44,95,133].
Köser et al [61,70] achieved the experimental visualization of single-particle volatile flame structures for the first time by using high-speed planar OH-LIF imaging. Figure 25(a) shows a sequence of OH-PLIF images captured at 10 kHz, illustrating a burning bituminous coal particle from the onset of gas phase ignition. Upon ignition at t = 0 ms, OH-LIF signal intensity and size increased continuously, forming a nearly spherical reaction zone surrounding the coal particle. The particle location can be identified at the center of the flame, leaving a dark streak to the right-hand side, as the particle attenuated the laser sheet with around 100 µm in thickness. The averaged radial intensity profiles in figure 25(b) display a rapid intensity increase and spatial expansion. A sharp gradient existed between the particle center and peak OH-PLIF signals, marking the reaction zone of the diffusion flame, while the more gradual decrease after the OH-peak resulted from the hot combustion products. The location of the reaction zone remained Figure 24. The luminous volatile flame of a single bituminous coal particle and image processing procedure to derive flame area A f , effective flame radius r eff , and elongated distance rc. Reprinted from [43], Copyright (2015), with permission from Elsevier. almost constant over time, indicating a balance between volatile release and consumption rates. In a subsequent study [123], 3D OH-LIF visualization and volumetric reconstruction of volatile flames were applied to calculate the flame stand-off distance r flame as the distance between the particle and the peak OH-LIF intensity location. It was found that r flame increased with increasing particle diameter. However, the effective flame size r flame /r p normalized by the particle size decreased with increasing r p , experimentally demonstrating a slower volatile release rate resulting from lower particle heating rates as particle size increased.
Similar to other gaseous flames, volatile flames surrounding fuel particles possess complex and inherently threedimensional structures that are crucial to understanding the physics of combustion, particularly for PGC. In recent years,  3D Mie-scattering and LIF-based imaging techniques, facilitated by high-frequency laser scanning technology [167,[187][188][189][190] or tomographic imaging [191][192][193], have been developed and implemented for solid fuel combustion experiments [122][123][124]. Specifically, conventional planar LIF measurements have been extended to 3D imaging using acoustooptical defectors (AOD) for visible or UV pulsed lasers. The AOD technology involves rapidly scanning a thin laser sheet across the probe volume, allowing quasi-3D reconstruction by combining multi-plane information at different depth positions. The working principle of AOD for laser scanning was described in [167,187]

in detail.
Illustrative examples of the 3D reconstructed volatile flame structures are presented in figure 26 for increasing PNDs from case A to D. PND was determined in-situ from projected DBI images for each flame, and it increased by over ten times from 0.18 mm −3 to 1.9 mm −3 for the examples presented. The enclosed non-flammable region observed within the enveloping flame grew extensively with increasing PND. Local flame extinction was caused by high heat loss due to particlegas heat transfer and the formation of zones exceeding flammability limits within the enveloping flame. This conclusion was based on combining experimental evidence of reducing gas temperature with results of a detailed numerical study for the same cases [194]. The study classified three zones of particle heating, volatile flame, and nonflammable region, which allowed the quantitative determination of flame extinction. The phenomenon of flame extinction became more pronounced, and the non-flammable region grew larger with increasing PNDs. Additionally, non-flammability was strongly inhibited by increasing oxygen mole fraction in the hot atmospheres but behaved similarly in N 2 and CO 2 conditions [124].

Gas velocity
The aerodynamic effects of the gas phase surrounding solid fuel particles can have a significant impact on the combustion behavior of single particles and particle groups as they affect the advective transport of heat, volatiles and oxidizer. This section deals with the fundamentals of the experimental methods used to determine the gas velocity independent of the presence of solid fuel particles while section 4.7 discusses the determination of particle dynamics and the simultaneous measurement of gas and particle velocities.
Before the advent of optical techniques in flow measurements, hot wires were the key experimental method for measuring gas phase velocities [195]. Due to their intrusive nature, point-wise measurement imitations and limited use in hightemperature combustion environments due to the thermal stability of the wire probe, optical techniques have replaced hot wires in most flow measurement applications, especially when combustion processes and high temperatures are involved.
Tracer-based optical techniques rely on the seeding of chemically inert particles into the flow, whose properties should be chosen such that these tracer particles follow the fluid motion as closely as possible and withstand the high temperatures present in the hot environments of the burners presented in section 2.3 [195][196][197]. The first criterion means that particles should exhibit a small Stokes number St = τ p /τ f , where τ denotes the relaxation times of particle (τ p ) and fluid phase (τ f ), as well as a uniform size distribution, such that only minor differences in scattering intensity and slip velocity between gas and differently sized particles should occur. However, if the particle size is too small, the scattering intensity creating the particle images on the camera sensor will be too low resulting in images with a low signal-to-noise ratio. For these reasons, micron-sized particles made from ceramic materials, such as titanium dioxide or aluminum oxide, are commonly used in such applications [195], as demonstrated by the flow field measurements within the LFR reported in [81]. The motion of these tracer particles and hence the fluid velocity is determined using minimally invasive laser-based methods.
Laser-Doppler velocimetry (LDV) was the first widespread laser-based particle velocimetry method. It was established in the 1960s and provides point-wise measurements [198]. The basic principle of LDV relies on the crossing of two slightly frequency-shifted laser beams, such that an interference pattern is formed in the probe volume. As tracer particles pass through this pattern, they will scatter the laser light onto a detector that records a time-resolved intensity trace. The higher the velocity of the tracer particle, the higher the detected oscillation frequency of the time-resolved recording of the scattered light from which the velocity of the tracer particle is computed. This technique can be expanded to measure multiple velocity components by using beams crossed perpendicularly. However, LDV is limited to single-point measurements, providing no instantaneous information about the spatial structure of the velocity field. Li et al reported a recent example of an LDV-based characterization of the gas velocity in a McKenna burner for the study of iron particle combustion [33].
With the increasing usage of digital cameras as scientific instruments in the 1980s and the development of flexible pulsed lasers, particle image velocimetry (PIV) became defacto standard velocity measurement technique in experimental fluid mechanics, including applications in combustion systems. Figure 1 of [199] shows that the number of publications in the literature on PIV largely surpasses the number of publications on hot wire probes or LDV for more than 20 years already. A comprehensive introduction to PIV and its state of the art is given by [196,197,199]. The basic principle of PIV is the recording of two subsequent images of the scattered light of laser-illuminated tracer particles and the algorithmic evaluation of their displacement inside interrogation windows. Within these windows, cross-correlation of the two recorded images is computed and the position of the peak of the correlation function determines the magnitude and direction of the velocity vector. PIV requires a certain number of tracer particles to be present within each interrogation window to enable a robust calculation of the local fluid velocity. If the seeding density is too sparse, PTV is often employed. The movement of individual particles is tracked, which results in a velocity vector for each detected particle [196]. In its most simple form, PIV and PTV are conducted in a planar arrangement, where one camera perpendicularly faces a laser light sheet that illuminates tracer particles, resulting in the measurement of two velocity components within a plane. Both PIV and PTV have been extended to time-resolved volumetric approaches, as shown by the recent development of novel Langrangian particle tracking approaches [200] and the use of tomographic PIV, including turbulent combustion systems [188].
Notable examples of using PIV in burner systems to study solid combustion are the studies by Li et al [81], García Llamas et al [102] and Balusamy et al [115] (within the scope of this review). The latter study compared the performance of PIV and LDV to determine gas velocity using alumina tracer particles and the velocity of bituminous coal particles, whereas the measurements with tracers and coal particles were conducted separately.
As indicated by table 3, gas velocity is often not specifically measured in generic burners even though it represents an important boundary condition for the interaction of the oxidizing gas phase with solid fuel particles. For low particle concentrations in laminar flows, which represent the majority of the studies discussed in this review, flow velocity is only minimally affected by the presence of solid fuel particles. Still, knowledge of the local gas velocity is highly important when analyzing the effect of the surrounding gas flow on combustion and particle movement dynamics, making measurement of it highly desirable for comprehensive characterization of solid fuel combustion at the particle level. Characterization of the instantaneous fluid velocity, ideally simultaneously with the solid fuel particle velocity is necessary, especially considering solid fuel combustion in unsteady turbulent flows and at higher PNDs. The following section discusses both of these aspects, based on the methods presented for gas phase velocimetry.

Particle dynamics and particle-fluid interaction
Reliable experimental determination of solid fuel particle dynamics in burner systems is highly important in the study of phenomenological aspects of solid fuel combustion. In this review, we use the term dynamics to refer to both the translational and rotational movement of the solid fuel particle. The Figure 27. Different stages of particle motion of a rocketing and a non-rocketing Norwegian spruce particle. The particle velocity is positive for motion in the direction of gravity. Reproduced from [102]. CC BY 4.0.
detection of rotation speeds of solid fuel particles with a diameter of only a few 100 µm is especially challenging, as it requires high optical magnification and spatial resolution of the imaging system to resolve the shape of the particle over time as discussed in section 4.2.
Similar to the experimental determination of gas phase velocities discussed in the previous section, the translational velocities of solid fuel particles can be measured using LDV or, more commonly in recent experiments, PIV/PTV based on the time-resolved or dual-frame imaging of solid fuel particles. In the literature, the temporally resolved solid fuel particle position was detected using the luminosity of burning particles, illuminated by a laser sheet, as in gas phase PIV/PTV or holographic imaging. Tables 3 and 4 provide a comprehensive overview of recent studies.
To determine the particle rotation speed, the particle shape needs to be spatially resolved and tracked over time. For this reason, the rotation speed within the imaging plane ω p,z is usually measured using highly resolved shadowgraphy or DBI measurements, which are necessary to determine the spatial orientation of the particle as demonstrated in [83] for single solid fuel particles. Most commonly, an ellipse is fitted to the shape of the particle and the major or minor axis orientation is tracked over time to compute the planar rotation speed. Combining multiple DBI systems in a stereoscopic fashion enables the particles' 3D orientation to be visualized (as demonstrated in [79]). There is, however, no study in the literature that has presented this approach in a time-resolved manner, which limits available measurements of particle rotation to a single rotational velocity component. In contrast, 3D translational velocities have been measured using stereoscopic imaging to study micro-explosions during iron particle combustion [32].
For SPC, a notable example of particle velocity measurement is García Llamas et al [102], in which Norwegian spruce particles were recorded inside a drop tube reactor equipped with an upside-down McKenna burner. Using PTV, they tracked individual biomass particles over time to investigate their motion within the reactor. Due to their high volatile content, the elongated spruce particles release a relatively high amount of volatile matter during pyrolysis, which creates a Stefan flow around the particle. In some cases, sudden acceleration during devolatilization could be observed, often occurring with a change of direction. This rocketing behavior was attributed to the sudden release of volatiles, as shown in figure 27. During the temporal evolution of the effective volatile gas velocity, the rocketing particle releases a large amount of volatiles in the opposite direction to particle movement during the sudden acceleration event.
In all of the particle-resolved studies in laminar burners discussed in this review, the relative velocity between the particle and gas phase (also called slip velocity) is not measured instantaneously but inferred from separate velocity measurements of both phases. This assumption and approach is only valid as long as the particle velocity does not influence the gas phase, which cannot be assumed for high PNDs and turbulent flows: instantaneous two-phase velocity measurements in such complex systems are needed. However, simultaneous measurement of particle and gas phase is very challenging, even in non-reactive dispersed multi-phase flows [201]. The current methods for two-phase velocimetry are outlined below, emphasizing that they have not yet been employed in reactive conditions with sieved solid fuel particles in particleresolved studies.
Both LDV and PIV/PTV can be used for two-phase velocimetry, with the latter currently being more popular [201]. Two-phase PIV/PTV can be realized using a single camera, and phase separation using image processing algorithms [202] or through the emitted wavelength of scattered light by flow tracers and larger particles including fluorescence [203], phosphorescence [204], incandescence [205], or a combination of them [206]. Wavelength-dependent separation methods require at least two monochrome cameras to capture both phases separately, using color filters, with fluorescence and phosphorescence demanding multiple excitation wavelengths. Single camera two-phase PIV/PTV using ceramic flow tracers have the highest potential to measure reactive flows, including reacting solid fuel particles, similar to PIV experiments in combustion systems using gaseous fuels.
Geschwindner et al [207] is a recent example of multiphase velocimetry of single miscanthus biomass particles in a non-reactive turbulent jet at a bulk Reynolds number of 7500. Oil droplets were added as flow tracers and simultaneously seeded into the flow with the biomass particles. Both particles were illuminated by an ultra-high repetition rate fiber laser, enabling the movement of the biomass particle and its surrounding gas phase to be tracked at a repetition rate of 200 kHz. Figure 28 shows the two-phase PIV measurement of a representative particle moving through the centerline region of the turbulent jet. The ultra-high recording rates were necessary as the particles had to be tracked while moving at more than 30 m s −1 at a pixel resolution of just 10 µm to resolve the particle shape and hence determine rotation speed. As the instantaneous slip velocity field and the particle size from a simultaneously running DBI system were known, the time-resolved Reynolds particle number could be computed, providing information on the local flow field around the biomass particle.

Gas temperature, species, and nanoparticles
Non-intrusive probing of gas temperature and species in multiphase reactive flows poses significant challenges due to the presence of solid particles. Raman/Rayleigh spectroscopy is a powerful tool for probing the thermochemical state of the gas phase, but the Rayleigh signals can be easily distorted by the Mie-scattering of micron-sized solid particles. The high laser energy required for Raman measurements would induce particle breakdown with mission intensities of several orders of magnitude above that of Raman scattering. Coherent anti-Stokes Raman spectroscopy (CARS) can provide the gas phase temperature in near-surface processes and solid fuel flames, but its capability is commonly limited to point-wise measurements. Line-of-sight laser absorption thermometry techniques, such as tunable diode laser absorption spectroscopy (TDLAS) [208] and OH-LIF combined with absorption thermometry [81], cannot provide particle-level resolved measurements. Consequently, Raman/Rayleigh spectroscopy, CARS, TDLAS, and LIF have been used to characterize boundary conditions or macroscopic temperature distribution in multi-phase combustion. In general, the particle-level resolved measurement of temperature in the vicinity of a burning particle has been challenging, and only a few studies have been conducted.
One such study was performed by Bucher et al [209], who applied LIF of AlO to investigate temperature and species in the gas phase around a single burning aluminum particle. AlO-LIF was used to determine both the AlO distribution and the gas temperature around ∼220 µm aluminum particles. They used two LIF systems to excite the (0,1) and (0,2) band heads of the B 2 Σ + ← X 2 Σ + electronic system of AlO at 507.935 µm and 533.659 µm. The systems were operated successively, with a time separation of 50 ns. The laser fluence used was well above the saturation threshold, which made the assumption of vibrational equilibrium in the electronic B-state unnecessary. When the fluorescence ratio was formed to determine the temperature, effects such as LIF reabsorption or radiative trapping were canceled out. Broadband emission was recorded perpendicular to the two superimposed light sheets using two cooled fiber-coupled intensified CCD cameras. The signal was collected with a single f /4.5 lens (105 mm focal length) and split between the two cameras using a coated, non-polarizing beam splitter cube. A 10 nm interference filter, centered at 488 nm, was used to block the elastically scattered laser light. The field of view of both cameras was 4 × 2.7 mm and the pixel resolution was 12 µm/pixel. The measurements were limited to a region of the laser sheet where the fluctuations of the local fluence were within 25%. The resulting variation of the PLIF signal was within 2% since it was recorded in the saturation regime. The presence of a condensed phase in the particle-enveloping flame creates measurement challenges. In the high-temperature regions, the thermal radiation from the hot oxides requires short exposure times, which was accounted for with an intensifier gate of 25 ns. Mie scattering from the condensed phase broadens the spatial distribution of the laser light. So, rather than using a thin light sheet, a wide one that included the whole flame around the particle was used. With the help of another CCD camera, aligned almost coaxially to the laser sheet, it was ensured that only particles were evaluated whose flames were within the light sheet. Assuming rotational symmetry, an Abel transformation was used to determine the AlO and temperature profiles around the particles during the first phase of combustion, characterized by a spherical flame that burns with a distance to the particle's surface. As shown in figure 29(a), the results indicated a finite-thickness flame structure in which AlO is an intermediate. The temperature increased from 2350 K at the particle surface to a nearly constant value of 3800 K at a distance of ∼4 particle radii from the particle surface. AlO is formed by the outward transport of Al and the reaction with molecular oxygen diffusing inward, as shown in figure 29(b).
Regarding intermediate gas species, the OH radical is a well-accepted flame marker that has been commonly used for hydrocarbon fuels to determine ignition [28] and visualize flame structures using OH-LIF techniques [124]. The excitation wavelength for OH-LIF is typically around 283 nm, depending on the transition line chosen, and the emission is red-shifted to approximately 310 nm. Polycyclic aromatic hydrocarbons (PAHs), which can be produced or released during coal particle combustion, are considered a crucial precursor to soot particle formation. Measuring the spatial distribution of PAHs can aid better comprehension of the pathways of soot formation. PAHs can be visualized by LIF techniques using excitation wavelengths in the UV and visible range [210]. The emitted light is broadband, and the wavelength is dependent on the molecular structure and the number of aromatic rings [211]. While OH-LIF has been Comparison of local equilibrium model prediction of radial temperature and normalized species mole fractions with experimental data for 220 µm aluminum particle combustion in a gas mixture of 21% O 2 . Reprinted from [209], Copyright (1998), with permission from Elsevier. applied at the single-particle level in previous studies [61,70,75,81,122,124,131,157], PAH-LIF has only been utilized to visualize turbulent jet pulverized coal flames, such as in [212]. Currently, there is a significant lack of experimental data and understanding of PAH formation during coal and biomass combustion at the particle level, which is highly desirable for future research.
Species such as SO x and NO x are of great interest in pollutant formation. LIF can be used to quantitatively measure NO in gaseous flames with laser excitation at 226 nm [213]. SO 2 can be visualized using LIF at 266 nm as demonstrated for turbulent flames [214] and internal combustion engines [215]. However, applying these diagnostics to coal and biomass flames may result in interference from PAH-LIF signals, as PAHs can be excited in a broad wavelength range and their emission spectra overlap with SO 2 -LIF and NO-LIF, making these measurements challenging in practice. This may explain why LIF of SO x and NO x has been rarely used in the past. Nonetheless, these diagnostics could be very useful for studying the combustion of carbon-free metal fuels such as iron and aluminum.
Nanoparticle formation is another critical issue in the gas phase reaction of solid fuels. In coal combustion, soot particles are a common type of nanoparticle, while metal combustion can produce nanometric oxides. LII is a widely used technique to measure soot particles [216]. LII requires a sufficiently high laser fluence to heat soot particles instantaneously. The thermal radiation can be captured to visualize the spatial distribution of soot and quantify the volume fraction and size distribution. However, LII has only been applied to large-scale flames so far [116,131,212,217,218], and few particle-level measurements have been reported in the literature (table 4). To image soot particles, Mie-scattering has been applied to single-particle volatile flames [40]. The disadvantage of Mie-scattering is that it superimposes signals from both nano-sized and micro-sized particles which cannot be distinguished from each other [124].
Recent studies have investigated the visualization of nanoparticle clouds during single iron particle combustion [36,107,109]. These studies all utilize backlight illumination to image the nanoparticles formed around burning iron particles. Figure 30 shows temporal image sequences of the typical topology of nanoparticle clouds formed near three single burning particles [36]. The sizes of the molten parent iron cores were 31 µm, 54 µm, and 73 µm in events EV-C, EV-D, and EV-E, respectively, as determined in-situ. The topology of nanoparticle clouds clearly correlated with the size of the parent particles. In event EV-C, a spherical cloud was formed within the 2.5 ms shown, whereas EV-D and EV-E showed a parachute-like shape and a contrail-like shape, respectively. This was due to the increased slip velocity between the iron core and nanoparticles, which led to lower iron particle velocities with increasing particle diameter. Nanoparticles were smaller and could better follow the gas velocity. It was also observed that the formation of nanoparticles was delayed with increasing diameter due to longer ignition delay and combustion duration, due to the difference in particle heating rates. Nanoparticle formation terminated at the peak luminosity point of the iron core, showing a clear correlation between temperature and nanoparticle formation. Similar phenomena have also been investigated by Ning et al [109] by combining pyrometry and shadowgraphy. Experimental evidence of nanoparticles challenges the assumption of the heterogeneous combustion regime of iron, where reactions remain in the solid phase. Further experimental studies are urgently needed to understand the mechanism behind nanoparticle formation, via measurements of (1) detailed composition and concentration of nano oxides, (2) time-resolved nanoparticle topology and particle temperature history, and (3) vapor-phase components near the particle surface. Optical diagnostics, including laser-induced breakdown spectroscopy (LIBS), LII, DBI, Mie-scattering, and LIF of FeO, could be useful tools for future studies.

Unresolved issues and implications
A lot of effort have been dedicated to the development and application of diagnostics to gain insights into the detailed chemical and physical processes involved in solid fuel combustion at the particle level. Previous experimental studies are summarized and categorized in tables 1-4. Despite studies over many years, the combustion process of many solid fuel particles, including carbonaceous and noncarbonaceous fuels, is not fully understood. In particular, combustion characteristics of carbon-free metal fuels are largely unexplored. For carbon-neutrality, fundamental studies of both carbon-neutral and carbon-free fuels at the particle level and corresponding diagnostics for generic and close-to-reality applications are highly desired. Several open issues and challenges that need to be addressed in future studies are listed below.
(1) Measuring the surface temperature of particles below 1200 K using pyrometry or thermal radiation is challenging due to the limitations of conventional CCD and CMOS sensors for infrared wavelengths. In the visible range, the grey-body radiation intensity is too low to provide a reasonable signal-to-noise ratio. However, it is crucial to measure the particle temperature in the lower temperature range of 300-1200 K to gain insights into the particle heating and heat release during the initial oxidation stages. For instance, determining the ignition temperature of iron particles, which is estimated to be between 800 K and 1300 K, is essential for developing accurate kinetic models for iron combustion. One possible solution for measuring the particle surface temperature in the difficult-to-measure range of 300-1200 K is to use thermal imaging cameras that operate in the infrared range. Usually, these cameras require a onepoint calibration on the objective material at a known temperature, hence, the accuracy may be affected by changes in emissivity due to material oxidation and temperature changes. As a result, thermal cameras may not be suitable for quantitative measurements but could be useful for qualitative imaging purposes. Alternatively, pyrometry with two or three IR cameras would extend the measurable limits towards lower temperatures. Another potential solution is to apply a thin phosphor film on the particle surface, which can be excited by a laser source [138]. The resulting emittance can be used to derive temperatures using a pre-calibration procedure. This method could be used for non-volatile solid fuels in which phosphors are not broken down by outgassing or evaporation. However, the influence of the phosphor on ignition and chemical reactions must be considered. (2) Obtaining intermediate species, pollutants such as NO, and temperature in single coal and biomass particle combustion through experiments has been a challenging task. These parameters are crucial for understanding the diffusion flame of volatile matter release and improving combustion processes. Furthermore, these data are necessary for developing kinetic models and conducting numerical simulations that involve realistic heating rates and interacting gas flows, which are not available in thermal gravimetric analysis measurements. Because many gas phase flame techniques cannot be readily applied to multi-phase combustion, more effort needs to be directed towards developing suitable diagnostics. For example, PAH-LIF on the particle level could provide insights that would be helpful to understanding soot formation. (3) Further efforts are needed to develop diagnostics for gas phase species in single-particle iron combustion. LIF of AlO has been utilized to measure the flame structure and gas temperature of an Al particle [209,219]. The gas phase processes in single iron particle combustion have been poorly studied due to the assumption that micrometric iron powder burns heterogeneously without any gas phase reactions. Recent experimental observations of nanoparticle formation suggest that iron oxidation partially originates from the gas phase reactions, for which there is a significant lack of knowledge. One possible diagnostic for understanding gas phase processes in single iron particle combustion is LIF of Fe atoms. This technique has been applied successfully in nanoparticle synthesis via spray combustion [220]. (4) Since the formation of nanoparticles during iron combustion is a serious issue, there is a pressing need for diagnostics that can shed light on the mechanisms behind this phenomenon and help reduce its occurrence. In particular, in-situ and quantitative measurements of nanoparticles are essential, including determination of oxide composition, nanoparticle size distribution, and volume fraction. To this end, LIBS is a promising technique for probing the iron oxidation stage in real-time [221]. It can also be applied to probe oxidation stages of micrometric iron and aluminum oxides, which helps to identify the reaction progress. Additionally, LII, which is established for measuring soot, could be adapted to measure nanoparticles of iron oxides. (5) Simultaneous measurement of the particle and gas phase velocities is essential for determining particle-fluid interaction effects. For two-phase PIV, simultaneous seeding of thermally stable small tracer particles and solid fuel particles is necessary to achieve this. For single solid fuel particles in laminar flows, the Stefan flow resulting from the out-gassing of pyrolysis gases might push away tracer particles, limiting the measurement of this phenomenon in classic PIV/PTV approaches. García Llamas et al [102] have tracked the luminosity of the incandescent clouds of volatiles to determine the Stefan flow around biomass particles, which however assumes that the spread of this luminosity signal is equal to the fluid motion. For increasing PNDs, discrimination between solid fuel particles and tracer particles will become a major challenge, motivating detailed investigations using two-phase velocimetry in such environments. For high-speed turbulent flows, ultrahigh repetition rates are necessary to track particles at high optical magnifications. Geschwindner et al [207] showed that two-phase velocimetry of shape-resolved biomass particles in such flows is possible for non-reactive conditions using a fiber laser system. An extension of this approach to reactive flows using ceramic tracer particles or alternative approaches will have to be explored. (6) Closer collaboration between experiments and numerical simulations is essential for gaining a better understanding of complex solid fuel combustion processes. To achieve this, more effort is required in developing multi-parameter optical measurements that can simultaneously acquire data needed to understand various sub-processes. Recent studies have demonstrated the use of multi-parameter diagnostics to address gas phase reactions, surface temperature, and particle dynamics at the same time [79,81,85]. However, current optical diagnostics still have limitations and cannot provide a complete understanding of the combustion process. Combining experimental data with detailed numerical simulations can provide more in-depth analysis and interpretation of observed phenomena, including hard-to-measure parameters from simulations. Recent progress have been demonstrated in [36,81,122,139,194,222], but more collaboration and effort are needed from both experimentalists and numerical modelers.

Conclusion
This paper reviewed particle-resolved optical diagnostics used in solid fuel combustion experiments for clean power generation applications. The key outcomes from this review can be summarized in the following. (1) Oxy-fuel coal combustion, biomass combustion, and iron fuels in an oxidation-reduction cycle have been identified as the most promising technical pathways for carbon-neutral power generation and require extensive experimental investigation. (2) Laser and optical combustion diagnostics are powerful tools for probing various relevant parameters in-situ and improving fundamental understanding of combustion processes at the particle level.
(3) Solid fuel combustion is either single particle or PGC, depending on whether particle-particle interaction plays a role. (4) Previous experiments were summarized in tables and classified by practical aspects, including fuel and reactor type, investigated parameters and processes, and methodology. (5) Several important processes and parameters were highlighted and representative examples were discussed in detail. Although solid fuel combustion is crucial for decarbonizing power generation systems, it has been studied using optical combustion diagnostics less than other gas phase combustion processes. While coal combustion is fairly well understood, biomass and iron combustion require further experimental efforts, especially at the particle level. Adapting state-of-theart laser diagnostics to solid fuel combustion presents numerous challenges, highlighting the urgent need for fundamental studies to improve experimental technology and deepen understanding of underlying thermochemical processes. This review paper demonstrated these efforts by providing in-depth discussion of a selection of recent experimental developments, covering different combustion processes (section 4) and a collection of unresolved technical issues (section 5). This review aims to promote and inspire further technical advancements in optical diagnostics for different types of solid fuel combustion and utilization, facilitating their practical applications in the carbon-neutral society of the future.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).