Contemporary tools for experimentation and computational modeling of unsteady reacting flow open new opportunities for engineering insight into dynamic phenomena. In the work presented here, a novel use of proper orthogonal decomposition (POD) is described to validate the structure of dominant heat release and flow features in the flame, shear-layer, and wake of a bluff-body-stabilized flame. A general validation process is presented which involves a comparison of experimental and computational results, beginning with single-point mean statistics and then extending to the dynamic modes of the data using POD to reduce the ensemble of instantaneous flow field snapshots. The results demonstrate the use of this technique by applying it to large eddy simulations of the bluff body stabilized premixed combustion experiment. Large-eddy simulations (LES) using both Fluent and OpenFOAM were conducted to reproduce experiments conducted in an experimental test rig which was built as part of this work to study the behavior of turbulent premixed flames stabilized by bluff bodies. Planar Particle-Image Velocimetry (PIV) and filtered chemiluminescence were used to characterize the flow in the experiment's reacting and non-reacting regimes respectively. While PIV measurements could be compared directly to the velocity field in the simulations, the chemiluminescence measurements represented a line-of sight signal which was not directly comparable to the LES model. To account for this, the heat release in the LES models was integrated along simulated lines of sight by solving an additional discretized differential equation with heat release as the source term. The results show generally good agreement between the dominant modes of the experiment with those of the numerical simulations. By isolating the dynamic modes from each other via the proper orthogonal decomposition, it was shown the models were able to accurately reproduce the size, shape, amplitude, and timescale of various dynamic modes which exist the experiment, some of which are dwarfed by the other flow features and are not apparent using time-averaging approaches or by inspection of instantaneous snapshots of the flow. / Ph. D.

Yes / Dual-fuel HCCI engines allow a relatively small quantity of diesel fuel to be used to ignite a variety of fuels such as natural gas or methane in HCCI mode. The gaseous fuel is mixed with the incoming air, and diesel fuel is sprayed into the cylinder by direct injection. Mathematical modelling is used to investigate the effects of parameters such as premixed ratio (fuel ratio) and pilot fuel injection timing on combustion of a dual-fuel HCCI engines. A CFD package is used with AVL FIRE software to simulate dual-fuel HCCI combustion in detail. The results establish a suitable range of premixed ratio and liquid fuel injection timing for low levels of NOx, CO and HC emissions along with a reliable and efficient combustion. Dual-fuel HCCI mode can increase NOx emission with lower premixed ratios in comparison to normal HCCI engines, but it is shown that the NOx emission reduces above a certain level of the premixed ratio. Due to the requirement of homogenous mixing of liquid fuel with air, the liquid fuel injection is earlier than for diesel engines. It is shown that, with careful control of parameters, dual-fuel HCCI engines have lower emissions in comparison with conventional engines.

Ce travail s'intéresse à la compréhension du comportement des flammes non-prémélangées issues d'un jet de méthane assisté par un coflow d'air dilué avec du CO2, ou d'autres gaz chimiquement inertes pour discriminer les différents phénomènes impliqués dans la dilution. Les phénomènes transitionnels, décrochage et extinction, quantifiés par des limites de stabilité, sont analysés à l'aide de grandeurs physiques représentatives. Le domaine de stabilité de flamme est limité par des surfaces 3D dans le domaine physique ( Qdiluant/Qair (taux de dilution), Uair (vitesse d'air), UCH4 (vitesse de méthane)), révélant un effet compétitif entre l'aérodynamique et la dilution. Des cartographies génériques de décrochage et d'extinction communes à tous ces diluants sont proposées. Des grandeurs liées à la stabilisation sont toutes soumises à des lois d'évolution auto-simlilaires. Il en ressort que la vitesse de propagation de flamme est l'élément clé du mécanisme de stabilisation lors de la dilution. / This work focuses on the understanding of the behaviours of non-premixed methane flame inside an air coflow diluted by carbon dyoxide (CO2) or by other chemically inert diluents in order to discriminate different phenomena involved in dilution. Transitional phenomena (liftoff and extinction) quantified trough the stability limits, are analyzed trough representative physical quantities. The flame stability domain is limited by 3D-surfaces (liftoff and extinction) in the physical domain (Qdiluant/Qair (dilution level), Uair (air velocity), UCH4 (methane velocity)) revealing a competitive effect between aerodynamics and dilution. Generic diagrams of flame liftoff and extinction are proposed for all the diluents. Physical quantities related to flame stabilization process are all submitted to, regardless of diluent, self-similar laws. This is explained by flame burning velocity which is considered as the key element in the flame stabilization mechanism with air-side dilution.

La Simulation aux Grandes échelles (LES) est de plus en plus présentée comme un outil à part entière dans le développement des chambres de combustion des turbomachines. Dans ce contexte, les écoulements réactifs considérés sont complexes et, dans un souci de validation, la LES doit montrer ses capacités sur des configurations modèles. Le but de cette thèse est de démontrer le potentiel de la LES pour la prédiction des écoulements vrillés réactifs non prémélangés de chambres de combustion modèles. - La LES est tout d'abord appliquée sur une configuration turbulente avec une hypothèse de prémélange parfait, afin d'étudier l'influence de la modélisation de la cinétique chimique, des modèles de combustion turbulente et de leur paramètres internes. Dans ces conditions, chacun de ces modèles montre ses avantages et désavantages. - L'hypothèse de prémélange parfait est ensuite retirée et l'étude réalisée permet d'évaluer l'influence de la prise en compte du mélange air/carburant dans un injecteur vrillé, des pertes thermiques et des conditions limites acoustiques. - Enfin, une chambre de combustion non prémélangée est simulée afin de démontrer les capacités du modèle de flamme épaissie sur ce type de flamme, pour lequel il n'a pas été initialement développé. Les résultats obtenus sont encourageants et démontrent, entre autres, la bonne représentation du positionnement de la flamme. / Large Eddy Simulation (LES) is considered as the next generation tool for the development of turbomachinery combustion chambers. In this specific context, reactive flows are of very complex nature and, as a validation goal, LES needs to prove its capabilities on academic configurations. This dissertation aims at demonstrating LES capabilities for the simulation of non-premixed reactive flows that can be found in swirled academic combustion chambers. - LES is first applied to a turbulent reacting configuration with a perfect premixing assumption. Chemical kinetics, turbulent combustion models and their internal parameters are studied. For this flow condition, each model shows his advantages and disadvantages. - Then, the perfect premixed hypothesis is removed, allowing the evaluation of mixing, thermal losses and acoustic boundary conditions for this swirled injector. - Finally, a non premixed combustion chamber is simulated with the dynamically thickened flame model, which was not developped for this kind of reactive flow. However, results are encouraging and demonstrate that the flame localisation is well represented by LES.

Turbulent premixed propane/air and methane/air flames were studied using planar Rayleigh scattering and particle image velocimetry on a stabilized Bunsen type burner. The fuel-air equivalence ratio was varied from Φ=0.7 to 1.0 for propane flames, and from Φ=0.6 to 1.0 for methane flames. The non-dimensional turbulence intensity, u'/SL (ratio of fluctuation velocity to laminar burning velocity), covered the range from 3 to 24, equivalent to conditions of corrugated flamelets and thin reaction zones regimes. Temperature gradients decreased with the increasing u'/SL and levelled off beyond u'/SL > 10 for both propane and methane flames. Flame front thickness increased slightly as u'/SL increased for both mixtures, although the thickness increase was more noticeable for propane flames, which meant the thermal flame front structure was being thickened. A zone of higher temperature was observed on the average temperature profile in the preheat zone of the flame front as well as some instantaneous temperature profiles at the highest u'/SL. Curvature probability density functions were similar to the Gaussian distribution at all u'/SL for both mixtures and for all the flame sections. The mean curvature values decreased as a function of u'/SL and approached zero. Flame front thickness was smaller when evaluated at flame front locations with zero curvature than that with curvature. Temperature gradients and FSD were larger when the flame curvature was zero. The combined thickness and FSD data suggest that the curvature effect is more dominant than that of the stretch by turbulent eddies during flame propagation. Integrated flame surface density for both propane and methane flames exhibited no dependance on u'/SL regardless of the FSD method used for evaluation. This observation implies that flame surface area may not be the dominant factor in increasing the turbulent burning velocity and the flamelet assumption may not be valid under the conditions studied. Dκ term, the product of diffusivity evaluated at conditions studied and the flame front curvature, was a magnitude smaller than or the same magnitude as the laminar burning velocity.

The laminar burning velocity is an important fundamental property of a fuel-air mixture at given conditions of temperature and pressure. Knowledge of burning velocities is required as an input for combustion models, including engine simulations, and the validation of chemical kinetic mechanisms. It is also important to understand the effect of stretch upon laminar flames, to correct for stretch and determine true (unstretched) laminar burning velocities, but also for modelling combustion where stretch rates are high, such as turbulent combustion models. A constant volume combustion vessel has been used in this work to determine burning velocities using two methods: a) flame speed measurements during the constant pressure period, and b) analysis of the pressure rise data. Consistency between these two techniques has been demonstrated for the first time. Flame front imaging and linear extrapolation of flame speed has been used to determine unstretched flame speeds at constant pressure and burned gas Markstein lengths. Measurement of the pressure rise during constant volume combustion has been used along with a numerical multi-zone combustion model to determine burning velocities for elevated temperatures and pressures as the unburned gas ahead of the spherically expanding flame front is compressed isentropically. This burning velocity data is correlated using a 14 term correlation to account for the effects of equivalence ratio, temperature, pressure and fraction of diluents. This correlation has been modified from an existing 12 term correlation to more accurately represent the dependence of burning velocity upon temperature and pressure. A number of fuels have been tested in the combustion vessel. Biogas (mixtures of CH₄ and CO₂) has been tested for a range of equivalence ratios (0.7–1.4), with initial temperatures of 298, 380 and 450 K, initial pressures of 1, 2 and 4 bar and CO₂ fractions of up to 40% by mole. Hydrous ethanol has been tested at the same conditions (apart from 298 K due to the need to vaporise the ethanol), and for fractions of water up to 40% by volume. Binary, ternary and quaternary blends of toluene, n-heptane, ethanol and iso-octane (THEO) have been tested for stoichiometric mixtures only, at 380 and 450 K, and 1, 2 and 4 bar, to represent surrogate gasoline blended with ethanol. For all fuels, correlation coefficients have been obtained to represent the burning velocities over wide ranging conditions. Common trends are seen, such as the reduction in burning velocity with pressure and increase with temperature. In the case of biogas, increasing CO₂ results in a decrease in burning velocity, a shift in peak burning velocity towards stoichiometric, a decrease in burned gas Markstein length and a delayed onset of cellularity. For hydrous ethanol the reduction in burning velocity as H₂O content is increased is more noticeably non-linear, and whilst the onset of cellularity is delayed, the effect on Markstein length is minor. Chemical kinetic simulations are performed to replicate the conditions for biogas mixtures using the GRI 3.0 mechanism and the FlameMaster package. For hydrous ethanol, simulations were performed by Carsten Olm at Eötvös Loránd University, using the OpenSMOKE 1D premixed flame solver. In both cases, good agreement with experimental results is seen. Tests have also been performed using a single cylinder optical engine to compare the results of the hydrous ethanol tests with early burn combustion, and a good comparison is seen. Results from tests on THEO fuels are compared with mixing rules developed in the literature to enable burning velocities of blends to be determined from knowledge of that of the pure components alone. A variety of rules are compared, and it is found that in most cases, the best approximation is found by using the rule in which the burning velocity of the blend is represented by weighting by the energy fraction of the individual components.

<p> Lifted non-premixed turbulent jet flames in the Transverse Instability Combustor (TIC) have been analyzed using qualitative and quantitative methods. Lifted flames in the TIC have been observed to stabilize about zero to five injector exit diameters downstream of the dump plane into the chamber and exhibit pulsating, unsteady burning. Anchored flames immediately begin reacting in the injector recess and burn evenly in a uniform jet from the injector exit through the entire optically accessible region. Statistically significant, repeatable behavior lifted flames are observed. It is shown that the occurrence of lifted flames is most likely for an injector configuration with close wall-spacing, second greatest for a configuration with close middle-element spacing, and lowest for a configuration with even element-spacing. For all configurations, of those elements that have been observed to lift, the center element is most likely to lift while the second element from the wall was likely. Flames at the wall elements were never observed to lift. Evidence is shown to support that close injector element spacing and stronger transverse pressure waves aid lateral heat transfer which supports flame stability in the lifted position. It is hypothesized that the stability of lifted flames is influenced by neighboring ignition sources, often a neighboring anchored flame. It is also shown that instances of lifted flames increase with the root-mean-squared magnitude of pressure fluctuation about its mean (P’ RMS) up to a threshold, after which flames stabilize in the anchored recess position.</p> <p>Dynamic mode decomposition (DMD) and proper orthogonal decomposition (POD) analyses of CH* chemiluminescence data is performed. It is found that lateral ignition of the most upstream portion of lifted flames is dominated by the 1W mode. Furthermore, it is shown that low-frequency high energy modes with spatial layers resemble intensity-pulses, possibly attributable to ignition. These modes are trademarks of CH* chemiluminescent intensity data of lifted flames. It was also shown that the residence time in the chamber may be closely associated with those low-frequency modes around 200 Hz. DMD and POD were repeated for a downstream region on the center element, as well as a near-wall element, highlighting differences between the lifted flame dynamics in all three regions. </p> <p>It is shown that lifted flames are best characterized by their burning behavior and in rare cases may stabilize in the recess, while still being “lifted”. Furthermore, it is shown that flame position differentiation can extend into an initial period of highly stable combustor operation. Dynamic mode decomposition is explored as potential method to understand physical building blocks of proper orthogonal spatial layers. Non-visual indicators of lifted flames within the high-frequency (HF) pressure signal are sought to seek a method that allows for observation of lifted flames in optically inaccessible combustors, such as those in industry. Some attributes of power-spectral diagrams and cross-correlations of pressure signals are provided as potential indicators. </p>

Influence of dust particles on the characteristics of premixed methane-air flames has been studied in this dissertation. Experiments are performed in a Bunsen burner type experimental set-up called Hybrid Flame Analyzer (HFA), which can be used to measure the burning velocity of gas, dust, and hybrid (gas and dust) premixed flames at constant pressure operating conditions. In the current study, analysis of particle-gas-air system of different types of dust particles (at particle size, dp = 75-90 µm) in premixed methane-air (ϕg = 0.8, 1.0 and 1.2) flames. Coal, sand, and sodium bicarbonate particles are fed along with a premixed methane-air mixture at different concentrations (λp = 0-75 g/m3) in both laminar and turbulent conditions. First, the variation of laminar burning velocity with respect to the concentration of dust particles, and type of dusts are investigated for different equivalence ratios. Second, the laminar premixed flame extinction with inert and chemical suppressant particles are studied. Third, the variation of turbulent burning velocity of these hybrid mixtures are investigated against different turbulent intensities apart from the different concentrations and types of dusts. Fourth, the radiative fraction of heat released from turbulent gas-dust premixed flames are also presented against the operating parameters considered. Combustible dust deflagration hazard is normally quantified using the deflagration index (Kst) measured using a constant volume explosion sphere, which typically is a sealed 20-liter metal sphere where a premixed mixture is ignited at the center and the progression of the resulting deflagration wave is recorded using the pressure measured at the vessel wall. It has been verified from prior studies that the quantification of the turbulence by this method is questionable and there is a need to analyze the controlling parameters of particle-gas-air premixed system accurately through a near constant pressure operated experimental platform. Thus, the main objective of this study is to analyze the influence of dust particles on premixed methane-air flames at near constant pressure conditions. The turbulent burning velocity is calculated by averaging the measured flame heights and the laminar burning velocity is calculated through the premixed cone angle measurements from several high-speed shadowgraph images obtained from the experiments. The turbulent intensity and length scale of turbulence generated by a perforated plate in the burner is quantified from the hot-wire anemometer measurements. Radiative heat flux is also measured for each of the turbulent test conditions. The outcomes from these experiments are: 1. An understanding of the variation of turbulent burning velocity of gas-dust premixed flames as a function of dust type, turbulent intensity, integral length scale, dust concentration and gas phase mixture ratio. 2. An understanding of the flame extinction characteristics and variation of laminar burning velocity of gas-dust premixed flames as a function of dust concentration and gas phase mixture ratio. 3. Quantify the radiative heat flux and radiative fraction of heat released from gas-dust turbulent premixed flames as a function of dust type, turbulent intensity, dust concentration and gas phase mixture ratio. Dust type and concentration play an important role in deciding the trend in the variation of both laminar (SL) and turbulent burning velocity (ST). Coal particles, with the release of volatile (methane), tend to increase burning velocities except for fuel rich conditions and at higher coal concentrations at larger turbulent intensities. At a higher turbulent intensity and larger concentrations, higher ST values are observed with the addition of sand. Sodium bicarbonate addition, with the release of CO2 and H2O, decreased the burning velocity at all the concentrations, turbulent intensities and equivalence ratios. Laminar flame extinction was observed with the addition of sand and sodium bicarbonate particles at conditions exceeding certain critical dust concentrations. These critical concentrations varied with the equivalence ratios of gaseous premixed flames. The turbulence modulation exhibited by particles and particle concentration is evident in these observations. The independent characteristic time scale analysis performed using the experimental data provided further insights to the results. The chemical and convective times in gas phase confirm the broadened preheat thin reaction zone regime in the current test cases, which has an effect of attenuating turbulence and thereby the resulting turbulent burning velocity. The particle time scale analysis (Stokes number) show that the effect of particles and particle concentration is to slightly enhance the turbulence and increase the turbulent burning velocity at lower concentrations. However, the time scale analysis of particle vaporization (vaporization Damköhler number) indicate an increase in the vaporization rate for particles (coal and sodium bicarbonate) resulting in a decrease in their turbulent burning velocities at higher concentrations and turbulent intensities. Sodium bicarbonate has higher evaporation rate than coal at same level of turbulence and the absence of this effect for inert (sand) results in higher turbulent burning velocities at higher concentrations. An increase in the turbulent intensity increases the vaporization rate of particles. The investigation on radiative fraction of heat released by methane-air-dust turbulent premixed flames identified that, the addition of dust particles increases the radiative fraction irrespective of the dust type due to the radial and axial extension of flame. A unified approach to couple this multiple complex phenomenon of turbulence, particle interaction, particle vaporization and combustion in particle laden premixed gaseous flames is the direction for future research.

A simulação numérica de chamas turbulentas é ainda hoje um desafio para as práticas de mecânica dos fluidos computacional. Compreendendo que as abordagens numéricas mais completas e realísticas atualmente disponíveis podem ser computacionalmente proibitivas, diversos modelos vêm sendo desenvolvidos com o objetivo de reproduzir os fenômenos envolvidos na combustão de uma forma simplificada, mas ainda fisicamente consistente. Este trabalho é, portanto, dedicado à comparação de diferentes modelos de fechamento para a turbulência baseados nas equações de Navier-Stokes em médias de Reynolds e de modelos para simplificação da cinética química baseados no conceito de flamelets, com e sem a modelagem da radiação térmica, esta última através do modelo de soma-ponderada-de-gasescinzas. Para tanto, na primeira parte do presente trabalho são comparados seis modelos de turbulência na solução de um jato turbulento de propano, não reativo e isotérmico, circundado por uma corrente paralela de ar, quanto a sua eficiência na predição dos valores médios da velocidade longitudinal e transversal, fração mássica de propano e massa específica da mistura. Os modelos são o k- Padrão (empregado na sua versão original e com mais duas modificações nas suas constantes conforme propostas encontradas na literatura), o k- Realizable, o k- Padrão e o k- Shear-Stress Transport. Um dos modelos de melhor desempenho é então usado na simulação de uma chama turbulenta não pré-misturada de metano/hidrogênio/nitrogênio circundada por um escoamento coaxial de ar de baixa velocidade, no qual são então comparados os modelos para redução da cinética química baseados no conceito de flamelets, o Steady Laminar Diffusion Flamelet (SLDF) e o Flamelet-Generated Manifold (FGM), tendo os seus resultados comparados aos dados experimentais para os valores médios da velocidade longitudinal, fração de mistura, temperatura e frações mássicas das espécies químicas. Dentre os modelos de turbulência avaliados, é observado que as duas versões ajustadas do k- Padrão e o k- Padrão se mostraram com melhor concordância em relação às medições experimentais do que os demais. No presente estudo é também avaliada a consistência dos dados experimentais reportados e uma discrepância é identificada neste jato, mas que, conforme verificado, não compromete a comparação dos modelos aqui proposta. Na solução do escoamento reativo, o modelo SLDF se mostrou com resultados bastante próximos aos resultados experimentais (exceto para o NO), sendo aprimorados ainda mais com a inclusão da modelagem da radiação térmica, sobretudo para regiões mais distantes do bico injetor do combustível, após o pico de temperatura da chama. O modelo FGM, contudo, apresentou resultados muito aquém dos esperados, sobretudo para as frações mássicas das espécies químicas, mesmo utilizando malhas com nível de refinamento muito maior e com o teste de diversas combinações de espécies para a variável de progresso da reação, e no qual a inclusão da radiação na modelagem também não trouxe benefícios perceptíveis. Todas as simulações numéricas foram realizadas empregando o código comercial ANSYS Fluent, versão 15.0.0. / The numerical simulation of turbulent flames is still a challenge for today's computational fluid dynamics practices. Understanding that the most complete and realistic numerical approaches available today may be computationally prohibitive, several models have been developed in order to reproduce the phenomena involved in combustion in a simplified, but still physically consistent, way. Therefore, this work is dedicated to compare different models for turbulence closure based on the Reynolds-averaged Navier-Stokes equations and models for simplification of the chemical kinetics based on the flamelet concept, with and without thermal radiation modeling through the weighted-sum-of-gray-gases model. Thus, in the first part of the current work six turbulence models are employed to solve a turbulent nonreactive isothermal flow, a propane jet surrounded by a parallel stream of air. The models are compared through their effectiveness in predicting the mean values of longitudinal and transversal velocities, propane mass fraction and mixture density. The models are the Standard k- (employed in its original version and with two modifications according to proposals found in the literature), the Realizable k- , the Standard k- and the Shear-Stress Transport k- . One of the best performing models is then used to simulate a turbulent nonpremixed flame of methane/hydrogen/nitrogen surrounded by a low-velocity air coflow, in which are compared the models to reduce the chemical kinetics based on the flamelets concept, the Steady Laminar Diffusion Flamelet (SLDF) and the Flamelet-Generated Manifold (FGM), being the numerical results compared to the experimental data for the mean values of longitudinal velocity, mixture fraction, temperature and species mass fractions. Among the six turbulence models evaluated, it is observed that the two adjusted versions of the Standard k- and the Standard k- showed better agreement with the experimental measurements than the other models. In the current study it is also evaluated the consistency of the reported experimental data and a discrepancy is identified, which, as verified, does not compromise the models comparison here proposed. In the solution of the reactive flow, the SLDF model showed results very close to the experimental results (except for NO), being further enhanced with the inclusion of the thermal radiation modeling, especially for regions far from fuel nozzle, after the peak of temperature of the flame. The FGM model, however, showed results far below the expected, especially for the mass fractions of chemical species, even using meshes with much higher refinement level and testing of various species combinations for the reaction progress variable. The inclusion of the radiation modeling did not brought noticeable benefits. All the numerical simulations were performed employing the ANSYS Fluent version 15.0.0 commercial code.

Reduction of the pollutant emissions is a challenge for the gas turbine industry. A solution to this problem is to employ the low swirl burner which can operate at lower equivalence ratios than a conventional swirl burner. However, flames in the lean regime of combustion are susceptible to flow perturbations and combustion instability. Combustion instability is the coupling between unsteady heat release and combustor acoustic modes where one amplifies the other in a feedback loop. The other method for significantly reducing NOx and CO2 is increasing fuel reactivity, typically done through the addition of hydrogen. This helps to improve the flammability limit and also reduces the pollutants in products by decreasing thermal NOx and reducing CO2 by displacing carbon. In this work, the flammability limits of a low swirl burner at various operating conditions, is studied and the effect of pressure, bulk velocity, burner shape and percent of hydrogen (added to the fuel) is investigated. Also, the flame structure for these test conditions is measured using OH planar laser induced fluorescence and assessed. Also, the OH PLIF data is used to calculate Rayleigh index maps and to construct averaged OH PLIF intensity fields at different acoustic excitation frequencies (45-155, and 195Hz). Based on the Rayleigh index maps, two different modes of coupling between the heat release and the pressure fluctuation were observed: the first mode, which occurs at 44Hz and 55Hz, shows coupling to the flame base (due to the bulk velocity) while the second mode shows coupling to the sides of the flame. In the first mode, the flame becomes wider and the flame base moves with the acoustic frequency. In the second mode, imposed pressure oscillations induce vortex shedding in the flame shear layer. These vortices distort the flame front and generate locally compact and sparse flame areas. The local flame structure resulting from these two distinct modes was markedly different.