Journal Publications

Optical resonances in multilayer cavities arise from constructive interference of electromagnetic waves and govern field enhancement, energy localization, and photothermal energy conversion. Although resonance phenomena are commonly characterized through real-frequency spectral responses, their underlying modal structure remains insufficiently understood. Here, a complex-frequency framework is developed to investigate optical resonance modes in metal–polymer–metal (M–P–M) cavities. Analytical solutions of the frequency-domain Maxwell–Helmholtz equation are assembled through interfacial continuity conditions to construct a global fieldresponse function, whose poles define the intrinsic resonance modes of the cavity. The framework is applied to a Cr–PDMS–Cr cavity to investigate how complex-frequency resonance poles underlie resonance matching, field localization, and energy redistribution as functions of polymer-layer thickness, refractive index, and metal-layer thickness. Resonance matching to a 532 nm excitation wavelength is achieved through cavity-thickness tuning, and the corresponding complex-frequency resonance poles are systematically tracked as the structural and material parameters are varied. The pole-based predictions are validated by real-frequency electric-field calculations, confirming the correspondence between the resonance pole associated with the excitation wavelength and the resonance-matching condition. The results further show that refractive-index contrast and metal-layer attenuation govern resonance-pole migration and the resulting resonance characteristics of the cavity. Resonance matching produces strong electric-field localization within the polymer cavity while substantially reducing metallic heat generation through resonance-induced energy redistribution. These findings establish a consistent physical relationship between complex-frequency resonance poles and real-frequency optical and thermal responses, providing a unified modal framework for interpreting resonance matching, field localization, and resonance-mediated energy redistribution and a basis for the analysis and design of photothermal multilayer structures.

This study presents a comprehensive three-dimensional unsteady computational fluid dynamics (CFD) investigation into the transient slurry transport mechanisms within a 300 mm silicon dioxide (SiO2) chemical mechanical polishing (CMP) system. The numerical model incorporates realistic pad groove geometries and detailed retaining ring slots to accurately capture slurry flow. Six simulation cases were analyzed under varying platen/head rotational speeds and slurry flow rates. Results indicate that accelerated pad rotation induces significant centrifugal thinning of the slurry film, while increased flow rates facilitate a broader radial footprint of slurry near the intake. Slurry recirculation zones were observed within the wafer-pad gap, decreasing in intensity at higher velocities. Furthermore, the simulations established a robust spatial correlation between the predicted slurry volume fraction and experimental material removal rates (MRR). These insights provide a physical basis for optimizing slurry delivery efficiency and mitigating within-wafer non-uniformity (WIWNU) in advanced semiconductor manufacturing processes.

Direct numerical simulations are performed to explore drag reduction mechanisms in turbulent Couette flow with a transverse (spanwise) alternating slip and no-slip (SNS) boundary condition on the bottom wall, compared to the case with a no-slip (NS) condition on the bottom wall. The streamwise length ratio of the slip and no-slip phases is set to 2.5:1. The SNS condition allows for a reduction in skin-friction coefficient of approximately 31%, with over 95% of the drag reduction attributed to the turbulent term, as demonstrated by the Fukagata– Iwamoto–Kasagi identity analysis. The results of the quadrant analysis reveal that Q2 ejection and Q4 sweep motions play a dominant role, contributing 75.1% and 57.8% to drag reduction, respectively. Energy spectra analysis shows that the SNS condition redistributes energy across turbulent scales, not only reducing the wavelength of the outer peak and its energy, but extending the energy closer to the wall, compared to the NS case. Further analysis of scale-separated Reynolds shear stress demonstrates the distinct contributions from different turbulent scales, where a cutoff wavelength is set to kx =h 1⁄4 10. The smaller-scale structures of kx =h < 10 contribute to drag reduction mainly near the bottom wall, accounting for 55.4% of the total drag reduction, while the larger-scale structures of kx =h > 10 contribute the reduction of 39.6%, mainly near the channel center.

Physics-regulated dynamic mode decomposition (PRDMD) is proposed as an improved framework over conventional dynamic mode decomposition (DMD), which allows robust and accurate predictions for two-dimensional laminar flow around a square cylinder. PRDMD addresses the limitations of the conventional DMD by incorporating physical laws, such as mass and momentum conservation, as regularizers, which struggle in noisy environments and with limited training data. To handle non-linear terms, such as convection terms as regularizers, this study treats these terms as independent features to be considered. The proposed method demonstrates significant improvements in noise robustness and physical consistency, reconstructing the flow field with reduced errors even in high-noise scenarios. Comparative analyses demonstrate the ability of PRDMD to reconstruct the vorticity field while adhering to the governing equations. It outperforms DMD in both the reconstruction and prediction phases. In addition, PRDMD maintains the stability of the eigenvalues, ensuring consistent oscillatory dynamics over extended time horizons. These findings highlight its potential as a reliable reduced-order modeling technique for complex fluid dynamics applications. It is particularly effective in scenarios with data scarcity and measurement uncertainty.

This study investigates the unsteady dispersion characteristics of hydrogen jets under varying flow velocities and inert gas purging conditions using computational fluid dynamics (CFD). Simulations were performed for six cases with identical total hydrogen mass release but different jet velocities and nitrogen purging conditions. A two-dimensional axisymmetric domain and the SST k-ω turbulence model were employed to predict transient concentration fields and jet behavior. Results show that higher jet velocities extend jet penetration and enhance initial mixing, leading to a wider flammable region near the nozzle, but also accelerate subsequent dilution. After hydrogen release stops, buoyancy dominates jet dispersion, causing residual hydrogen to accumulate in upper regions. Nitrogen purging effectively reduces hydrogen concentrations below the lower flammability limit more rapidly by displacing hydrogen and enhancing entrainment. These findings provide quantitative insights into hydrogen jet dispersion and offer practical implications for safety assessment and mitigation in high-pressure storage systems.

To analyze the flow and noise characteristics of a positive displacement hydrogen compressor, a computational fluid dynamics (CFD) simulation based on the Lattice-Boltzmann Method (LBM) was performed. For detailed analysis the discharge flow of the compressor was divided into three regions: the area around the valve, the discharge jacket, and the outlet pipe. Hydrogen passing through the valve formed turbulent jets, which induced high-frequency noise. Inside the discharge jacket, changes in the cross-sectional area caused continuous variations in the average velocity and turbulence characteristics. In the outlet pipe with an elbow, high flow velocity corresponding to the relatively narrow cross-section was observed. Additionally, flow separation occurring on the inner corner of the elbow enhanced turbulence, contributing to the amplification of high-frequency noise. This study provides a physical understanding of the flow and noise issues in positive displacement hydrogen compressors and presents a fundamental analysis to support the design of efficient and low-noise compressors.

The relation between flow rate and pressure drop was investigated numerically for a flame arrester with a crimped ribbon element. Permeability and inertial/viscous resistance factor of the micro-scaled element were obtained from a laminar simulation. Then, the porous zone model was adopted to simulate the flow in a flame arrester. The simulation results of the pressure drop were in good agreement with the experimental in the low flow-rate range. The analysis process developed in this study can be extended to different configurations/shapes of flame arresters and working fluids such as hydrogen.

Generally, the homogeneity of grease increases as the stirring time becomes longer. However, excessive stirring leads to a decrease in performance and production quantity due to the formation of bubbles. The performance of the product is determined by the homogeneity and the amount of foam generated when additives are added to grease. It is difficult to predict the appropriate stirring time, which depends on the viscosity of the grease and the type of additives, and it is measured through experiments. However, there are very limited ways to determine how much stirring has occurred depending on the viscosity and the type of additive. To solve this problem, we propose a method to measure the degree of stirring of grease using an image sensor. The proposed method involves floating a colored ball on the grease and measuring its movement. With the proposed method, the degree of movement of the ball can be used to represent the degree of stirring of the grease. Real-time performance was implemented to increase the method's usability. To verify the proposed method, we increased the speed of the stirrer and confirmed that the result indicates the degree of stirring of the grease increased when applying the proposed formula. This method could be useful for measuring the appropriate stirring time according to various environmental variables in grease production and other stirring processes.

The effect of the open-centered swirl generator (SG) on the flow in a pipe elbow was investigated. Steady RANS analysis with the shear stress transport (SST) model was performed to obtain flow fields using commercial CFD software. The working fluid and the bulk velocity were set to air and 1 m/s, respectively. When the SG of this study was installed downstream of the elbow, the maximum velocity region after the elbow was quickly guided to the center of the pipe and generated an axial swirl. Instead, the pressure drop increased due to the flow resistance compared to the case without the SG. Depending on the location of the SG, the difference in the flow field was shown as a result of the interaction with the flow passing through the elbow. Among the three locations considered in this study, the flow resistance was minimized when the SG was installed upstream. When the SG was installed downstream, both the vortex and the TKE were predicted to be the maximum.

Ultrasonic non-destructive testing (NDT) consists of three stages involving the following: an ultrasonic transducer, an impedance matching medium, and a test object. Generally, water, oil, or grease is used as a matching medium and increases the energy transmission rate between the transducer and test object. However, during non-destructive testing of large structures, the media can cause inconvenience before and after testing, and conventional three-phase tests cannot be performed on test subjects such as batteries that cannot use matching media. In this case, ultrasonic NDT consisting of two stages (transducer ↔ inspector) can be considered, but research on such NDT is scarce. In this study, finite element analysis was performed considering the contact condition between the probe and specimen without a matching medium as a dynamic contact condition involving the relative speed between the two contact objects; the resulting waveforms were analyzed. In addition, based on the linear elasticity theory, the energy transmittance between the ultrasonic transducer and specimen was analytically defined. This index can also be used as an indicator to evaluate the level of two-phase NDT technology and confirm the possibility of commercialization in the future. Furthermore, we propose a technique that can increase the ratio of energy transferred from the excitation transducer to the test object through simple modification of the experimental configuration.

The present study performs direct numerical simulations of turbulent channel flows using a spectral method in a large computational domain. Because of applying Fourier discretisation in the spectral method, parallelisation of the method may incur heavy communication overhead, thereby resulting in poor scalability. We design and improve the spectral code by exploring parallel techniques, including domain decomposition and data transposition algorithms. We focus particularly on the 2D domain decomposition and data transpose algorithm with the non-blocking collective operations improves parallel performance, thereby enabling latency mitigation by overlapping the computation and communication. Finally, we evaluate the code on the Nurion supercomputer at KISTI supercomputing centre. The transpose algorithm based on the non-blocking collective operations shows the best performance, which enables 3.55 times faster computing on 256 nodes using 16,384 MPI ranks for the L550 case of 1536 × 257 × 1536 grid points than the non-optimised 2D decomposition case.

Since the early work of Johnston [Johnston, Halleent, and Lezius, J. Fluid Mech. 56, 533 (1972)], the mean flow scaling in a spanwise rotating channel has received much attention. While it is known that the mean velocity near the pressure, turbulent side follows a linear scaling U = 2y + C at high rotation speeds, the functional dependence of C on the Reynolds number and the rotation number has been an open question. Here, U is the mean velocity, is the constant rotating speed in the spanwise direction, and C is a constant. In this work, we show that C + = log(l+ )/K, where the superscript + denotes normalization using wall units at the pressure side; l = uτ,p /2 is a rotation-induced length scale; K is a constant and K ≈ κ, where κ is the von Kármán constant; and uτ,p is the wall friction velocity at the pressure side.

Existing methods to identify the interfaces separating different regions in turbulent flows, such as turbulent/nonturbulent interfaces, typically rely on subjectively chosen thresholds, often including visual verification that the resulting surface meaningfully separates the different regions. Since machine learning tools are known to help automate such classification tasks, we here propose to use an unsupervised self-organizing map (SOM) machine learning algorithm as an automatic classifier. We use it to separate a boundary layer undergoing bypass transition into two distinct spatial regions, the turbulent boundary layer (TBL) and non-TBL regions, the latter including the laminar portion prior to transition and the outer flow which possibly contains weak free-stream turbulence. Both regions are separated by the turbulent boundary layer interface (TBLI). The data used in this study are from a direct numerical simulation and are available on an open database system. In our analysis of one snapshot in time, every spatial point is characterized by a 16-dimensional vector containing the magnitudes of the components of total and fluctuating velocity, magnitudes of the velocity gradient tensor elements, and the streamwise and wall-normal coordinates, all normalized by their global standard deviation. In an unsupervised fashion, the SOM classifier separates the points into TBL and non-TBL regions, thus identifying the TBLI without the need for user-specified thresholds. Remarkably, it avoids including vortical streaky structures that exist in the laminar portion prior to transition as well as the weak free-stream turbulence in the turbulent boundary layer region. The approach is compared quantitatively with existing methods to determine the TBLI (vorticity magnitude, cross-stream velocity fluctuation). Also, the SOM classifier is cast as a linear hyperplane that separates the two clusters of data points, and the method is tested by finding the TBLI of other snapshots in the transitional boundary layer data set, as well as in a fully turbulent boundary layer with similar levels of free-stream turbulence. Variants in which the approach failed are also summarized.

A robust algorithm is introduced for detection of large-scale coherent structures in transitional and intermittent flows that feature turbulent/non-turbulent (T/NT) interfaces. The algorithm is applicable the instantaneous flow fields of wall-bounded and free shear flows, and can effectively identify coherent events in the velocity or vorticity fields, or sweep/ejection motions. A database from direct numerical simulation (DNS) of transitional boundary layer is used to develop and demonstrate the capabilities of the algorithm which consists of three steps. The first is identification of the T/NT interface by comparing the normalized vorticity magnitude to a threshold value that is independent of the Reynolds number. The vorticity normalization is specifically designed to be applicable in transitional flows, where regions of the flow can host juxtaposed regions of laminar and turbulent flow. With the definition of the T/NT interface, conditional statistics are computed and perturbation quantities are defined relative to their respective conditional means. Second, the influence of the small-scale turbulence is excluded by applying an anisotropic Gaussian filter. The filter size is determined from the spatial characteristics of the smallscale vortical motions. In the third step, one-dimensional cores and two-dimensional surfaces within the flow structures of interest are identified from local extrema in the fields, and are tracked as Lagrangian objects. Using the algorithm, the population trends and advection speeds of large-scale sweep/ejection events are computed in the transitional boundary layer. Two additional flow configurations are also considered: turbulent jet flow emerging from a circular nozzle and the turbulent flow in a channel with a wavy surface.

The effect of large-scale motions (LSMs) on the turbulent/non-turbulent (T/NT) interface is examined in a turbulent boundary layer. Using flow fields from direct numerical simulation, the shape of the interface and near-interface statistics are evaluated conditional on the position of the LSM. The T/NT interface is identified using the vorticity magnitude and a streak detection algorithm is adopted to identify and track the LSMs. Two-point correlation and spectral analysis of variations in the interface height show that the spatial undulation of the interface is longer in streamwise wavelength than the boundary-layer thickness, and grows with the Reynolds number in a similar manner to the LSMs. The average variation in the interface height was evaluated conditional on the position of the LSMs. The result provides statistical evidence that the interface is locally modulated by the LSMs in both the streamwise and spanwise directions. The modulation is different when the coherent structure is high- versus low-speed motion: high-speed structures lead to a wedge-shaped deformation of the T/NT interface, which causes an anti-correlation between the angles of the interface and the internal shear layer. On the other hand, low-speed structures are correlated with crests in the interface. Finally, the sudden changes in turbulence statistics across the interface are in line with the changes in the population of low-speed structures, which consist of slower mean streamwise velocity and stronger turbulence than the high-speed counterparts.

Direct numerical simulation (DNS) of a turbulent channel flow with the slip boundary condition was performed to investigate the spatial characteristics of large-scale motions (LSMs) in a drag-reducing flow and the influence of LSMs on drag reduction. The slip boundary condition was applied in the streamwise direction. DNS of a channel flow with the no-slip boundary condition was also conducted for comparison. The overall mean skin-friction coefficient of the flow with the slip boundary condition was reduced by 35% relative to the flow with the no-slip condition at the same bulk Reynolds number of 10,333. The streamwise and spanwise length scales of the LSMs in the outer region are larger with the slip boundary condition than with the no-slip condition due to a large population of long LSMs. These LSMs influence the near-wall region via roll-cell motions and footprints. Although the number of total streaks decreases, a portion of the footprints increases due to the influence of the outer LSMs. The streamwise vorticities around the low-speed footprints are attenuated more than those around the high-speed footprints. The LSMs and the near-wall footprints contribute 45% of the total reduction in the skin friction coefficient, with 25% of this contribution due to low-speed LSMs.

Direct numerical simulation data from turbulent pipe and channel flows at Reτ ≈ 930 are used to investigate their statistical difference by focusing on large-scale motions (LSMs). The contribution to the bulk production of turbulent kinetic energy shows marked differences in the overlap and core regions. These discrepancies arise from the dominant contributions of the LSMs (λx > 3δ) to the Reynolds shear stress in the channel flow. The spectrum of the net Reynolds shear force reveals that the LSMs accelerate the mean flow in the overlap region. The net force spectrum is further decomposed into the spectra of velocity–vorticity correlations, hvωz i and h−wωy i, which are related to the advective vorticity transport and the change-of-scale effect, respectively. The dominance of large-scale accelerating motions (LSAMs) in the overlap region of the channel flow is due to the contribution of h−wωy i at longer wavelengths (λx > 3δ), The LSAMs are related to the long low-speed regions, and these regions are longer and wider in the channel flow than in the pipe flow. Due to the pipe curvature, the spanwise size of the LSMs is restricted by neighbouring LSMs and the spanwise velocity fluctuations are reduced. The contribution of h−wωy i to the acceleration is prominent in the channel flow, leading to the dominance of the LSAMs associated with the change-of-scale effect.

Direct numerical simulation of a turbulent boundary layer over a rod-roughened wall is performed with a long streamwise domain to examine the streamwise-scale growth mechanism of streamwise velocity fluctuating structures in the presence of two-dimensional (2-D) surface roughness. An instantaneous analysis shows that there is a slightly larger population of long structures with a small helix angle (spanwise inclinations relative to streamwise) and a large spanwise width over the rough-wall compared to that over a smooth-wall. Further inspection of time-evolving instantaneous fields clearly exhibits that adjacent long structures combine to form a longer structure through a spanwise merging process over the rough-wall; moreover, spanwise merging for streamwise scale growth is expected to occur frequently over the rough-wall due to the large spanwise scales generated by the 2-D roughness. Finally, we examine the influence of a large width and a small helix angle of the structures over the rough-wall with regard to spatial two-point correlation. The results show that these factors can increase the streamwise coherence of the structures in a statistical sense.

Direct numerical simulation data of turbulent channel flow (Reτ = 930) are used to investigate the statistics of long motions of streamwise velocity fluctuations (u), and the interaction of these structures with the near-wall disturbances, which is facilitated by their associated large-scale circulations. In the log layer, the negative-u structures are organized into longer streamwise extent (>3δ) in comparison to the positive-u counterparts. Near the wall, the footprint of negative-u structures is relatively narrow in comparison to the footprint of positive-u structures. This difference is due to the opposite spanwise motions in the vicinity of the footprints, which are either congregative or dispersive depending on the circulation of the outer roll cells. Conditional sampling of the footprints shows that the spanwise velocity fluctuations (w) are significantly enhanced by the dispersive motions of high-speed structures. On the other hand, the near-wall congregative motions of negative-u structures generate relatively weak w but intense negative-u regions due, in part, to the spanwise collective migration of near-wall streaks. The concentrated near-wall regions of negative-u upwell during the merging of the outer long scales – an effect that is demonstrated using statistical analysis of the merging process. This leads to a reduction of the convection speed of downstream negative-u structures and thus promotes the merging with upstream ones. These top-down and bottom-up interactions enhance the spatial coherence of long negative-u structures in the log region.

In turbulent flows over streamwise-aligned superhydrophobic surface (SHS) textures, the percent drag reduction is dependent on Reynolds number. This dependence is examined using direct numerical simulations of channel flow over SHS texture at three bulk Reynolds numbers, Reb = 2800, 6785 and 10975. Simulations of regular no-slip channel flows at the same bulk velocities are also performed for comparison. Changes in the flow due to the SHS texture are examined with particular focus on phase averaged statistics and coherent turbulent motions. As the Reynolds number is increased, the texture performance, or the percent drag reduction, is enhanced. In terms of turbulent motions, the weakened shear stresses are manifest as a decrease in the population of the small-scale vortical structures and a weakening of large-scale motions. The influence of stochastic fluctuations near the large-scale motions diminishes at higher Reynolds number. The enhanced performance is therefore due to a significant drop in the strength of small- and large-scale vortical structures. The weaker vortical motions yield a considerable reduction in the Reynolds shear stress and the transport of mean momentum. The SHS texture introduces a performance penalty in the form of a local increase in drag at the no-slip–free-slip edge. This adverse effect is caused by the inhomogeneity of the near-wall flow field. However, the affected region narrows in width at higher Reynolds numbers. As a result, the percent drag reduction approaches the gas fraction of the texture.

A direct numerical simulation of a turbulent pipe flow at a high Reynolds number of Reτ = 3008 over a long axial domain length (30R) was performed. The streamwise mean velocity followed the power law in the overlap region ( y + = 90–300; y/R = 0.03–0.1) based on the power law indicator function. The scale separation of the Reynolds shear stresses into two components of small- and large-scale motions (LSMs) revealed that the LSMs in the outer region played an important role in constructing the constant-stress layer and the mean velocity. In the pre-multiplied energy spectra of the streamwise velocity fluctuations, the bimodal distribution was observed at both short and long wavelengths. The k x −1 region associated with the attached eddies appeared in λ x /R = 2–5 and λ x /y = 18–160 at y + = 90–300, where the power law was established in the same region. The k z −1 region also appeared in λ z /R = 0.3–0.6 at y + = 3 and 150. Linear growth of small-scale energy to large-scale energy induced the k x −1 region at high Reynolds numbers, resulting in a large population of the LSMs. This result supported the origin of very-large-scale motions in the pseudo-streamwise alignment of the LSMs. In the pre-multiplied energy spectra of the Reynolds shear stress, the bimodal distribution was observed without the k x −1 region.

Statistical measures of turbulence intensities in turbulent pipe and channel flows at a friction Reynolds number of Reτ ≈ 930 were explored by a population of large-scale motions (LSMs) and very-large-scale motions (VLSMs). Although the statistical measures characterizing these internal turbulent flows were similar in the near-wall region, the extents of the mean streamwise velocities and cross-stream components of the turbulence intensities differed in the core region. The population density of VLSMs/LSMs decreased/increased significantly in the core region of the pipe flow. The survival time of VLSMs of the pipe flow was shorter than that of the channel flow. The area fractions of the VLSMs displayed similar trends to the population density. The wall-normal and spanwise turbulence intensities in the pipe flow increased in the core regions due to the high-speed large-scale structures and associated motions above the structures. The large-scale structures increased the streamwise intensity and the Reynolds shear stress in the pipe and channel flows, whereas the effective streamwise intensities and the Reynolds shear stress were equivalent in both flows.

Swimming jellyfish deliver momentum to the surrounding fluid in the form of vortices. A three-dimensional computational model was adopted to investigate the characteristic flow patterns produced by jellyfish with a jet-based locomotion and the process of vortex generation. The interaction between jellyfish and the surrounding fluid may be simulated using the immersed boundary method. The vortex structures generated in the wake were elucidated in detail. The vortices were formed due to the contraction and expansion of the elastic bell. A dimensionless temporal parameter was employed to analyze the vortex formation process. During the early stage of contraction, the vortices were dominantly generated by the stroke. The ejected fluid from the inside of the bell was then entrained into the vortices, thereby decreasing the vorticity at the core and increasing the total circulation within the vortex ring. The Froude propulsion efficiency increased as the vortex formation number increased, implying that the propulsion in the way of growing the vortex structures was favorable in terms of the efficiency.

Direct numerical simulations (DNSs) of spatially developing turbulent boundary layers (TBLs) over sparsely-spaced two-dimensional (2D) rod-roughened walls were performed. The rod elements were periodically arranged along the streamwise direction with pitches of px/k = 8, 16, 32, 64 and 128, where px is the streamwise spacing of the rods, and k is the roughness height. The Reynolds number based on the momentum thickness was varied from Reh = 300–1400, and the height of the roughness element was k = 1.5hin, where hin is the momentum thickness at the inlet. The characteristics of the TBLs, such as the friction velocity, mean velocity, and Reynolds stresses over the rod-roughened walls, were examined by varying the spacing of the roughness features (8 6 px/k 6 128). The outer-layer similarity between the rough and smooth walls was established for the sparsely-distributed rough walls (px/k P 32) based on the profiles of the Reynolds stresses, whereas those are not for px/k = 8 and 16. Inspection of the interaction between outer-layer large-scale motions and near-wall small-scale motions using two-point amplitude modulation (AM) covariance showed that modulation effect of large-scale motions on near-wall small-scale motions was strongly disturbed over the rough wall for px/k = 8 and 16. For px/k P 32, the flow that passed through the upstream roughness element transitioned to a smooth wall flow between the consecutive rods. The strong influence of the surface roughness in the outer layer for px/k = 8 and 16 was attributed to large-scale erupting motions by the surface roughness, creating both upward shift of the near-wall turbulent energy and active energy production in the outer layer with little influence on the near-wall region.

Direct numerical simulations (DNS) of turbulent boundary layers (TBLs) over isothermally heated walls were performed, and the influence of the wall-heating on the thermal boundary layers was investigated. The DNS adopt an empirical relation for the temperature-dependent viscosity of water. The Prandtl number therefore changes with temperature, while the Péclet number is constant. Two wall temperatures (Tw = 70 °C and 99 °C) were considered relative to T1 = 30 °C, and a reference simulation of TBL with constant viscosity was also performed for comparison. In the variable viscosity flow, the mean and variance of the scalar, when normalized by the friction temperature deficit, decrease relative to the constant viscosity flow. A relation for the mean scalar which takes into account the variable viscosity is proposed. Appropriate scalings for the scalar fluctuations and the scalar flux are also introduced, and are shown to be applicable for both variable and constant viscosity flows. Due to the modification of the near-wall turbulence, the Stanton number and the Reynolds analogy factor are augmented by 10% and 44%, respectively, in the variable viscosity flow. An identity for the Stanton number is derived and shows that the mean wall-normal velocity and wall-normal scalar flux cause the increase in the heat transfer coefficient. Finally, the augmented near-wall velocity fluctuations lead to an increase of the wall-normal scalar flux, which contributes favorably to the enhanced heat transfer at the wall.

Direct numerical simulations were carried out to investigate the spatial features of large- and very-large-scale motions (LSMs and VLSMs) in a turbulent channel flow (Reτ = 930). A streak detection method based on the streamwise velocity fluctuations was used to individually trace the cores of LSMs and VLSMs. We found that both the LSM and VLSM populations were large. Several of the wall-attached LSMs stretched toward the outer regions of the channel. The VLSMs consisted of inclined outer LSMs and near-wall streaks. The number of outer LSMs increased linearly with the streamwise length of the VLSMs. The temporal features of the low-speed streaks in the outer region revealed that growing and merging events dominated the large-scale (1–3δ) structures. The VLSMs (>3δ) were primarily created by merging events, and the statistical analysis of these events supported that the merging of large-scale upstream structures contributed to the formation of VLSMs. Because the local convection velocity is proportional to the streamwise velocity fluctuations, the streamwise-aligned structures of the positive- and negative-u patches suggested a primary mechanism underlying the merging events. The alignment of the positive- and negative-u structures may be an essential prerequisite for the formation of VLSMs.

Direct numerical simulations (DNS) of turbulent boundary layers over isothermally heated walls were performed, and the effect of viscosity stratification on the turbulence statistics and skin friction were investigated. An empirical relation for temperaturedependent viscosity for water was adopted. Based on the free-stream temperature (30◦ C), two wall temperatures (70◦ C and 99◦ C) were selected. In the heated flows, the turbulence energy diminishes in the buffer layer, but increases near the wall. The reduction in turbulence kinetic energy in the buffer layer is accompanied by smaller levels of Reynolds shear stresses and, hence, weaker turbulence production. The enhanced turbulence energy near the wall is attributed to enhanced transfer of energy via additional diffusion-like terms due to the viscosity stratification. Despite the lower fluid viscosity near the wall, dissipation is also increased owing to the augmented nearwall fine-scale motion. Wall heating results in reduction in the skin-friction coefficient by up to 26 %. An evaluation of the different contributions to the skin friction demonstrates that drag reduction is primarily due to the changes in the Reynolds shear stresses across the boundary layer. Quadrant and octant analyses showed that ejections (Q2) and sweeps (Q4) are significantly reduced, a result further supported by an examination of outer vortical structures from linear stochastic estimation of the ejection events and spanwise vortices.

Direct numerical simulations (DNS) were carried out to investigate the turbulence statistics and coherent structures in a fluid flow through a conical diffuser, called an Azad diffuser, for three opening angles of 2◦ , 4◦ , and 8◦ . The adverse pressure gradient affected the axisymmetric flow such that the streamwise component of the turbulence intensity was significantly increased in the outer region of the diffuser. The maximum Reynolds shear stress increased as the opening angle increased. The premultiplied energy spectra showed that the most energetic wall-normal height and its wavelength increased when the opening angle increased. Time-resolved instantaneous flow fields and twopoint correlation functions provided evidence for the streamwise merging of low-speed streaking structures. The outer vortical structures shifted toward the diffuser axis as the flow passed the diffuser throat. The swirling motions of the individual hairpins in the outer region were stronger in the diffuser flow than in the straight pipe flow.

Direct numerical simulations (DNSs) of open-channel flows in the presence of an air–water interface were performed to examine the effects of interface deformation on the turbulence structures. In the water-driven turbulence, flows characterized by either of the two Froude numbers (Fr = 0.2 and 0.8) were examined and compared. A coupled level-set and volume-of-fluid (CLSVOF) method was employed to track the interface. The mean and root-mean-square of the air–water interface elevation varied rapidly with the spanwise distance as Fr increased. The air that contacted the water was entrained into the turbulent flow. At high Fr, all turbulent normal stresses on the air side of the interface were high near the sidewall. Two-point correlations between the streamwise vortex and the velocity fluctuations provided structural information about the near-wall streaky structures and the inner secondary flows in the cross-stream plane. Linear stochastic estimates of the conditionally averaged flow field showed that the inner secondary flow consisted of not only in-plane velocity components but also streamwise velocity components.

Coherent wall-normal structures in a turbulent boundary layer (TBL) have been investigated by scrutinizing a direct numerical simulation (DNS) dataset with Reθ = 2500. It was found that groups of hairpin legs induce wall-normal-aligned structures and these large-scale regions are mostly induced by the combination of the attached and detached vortical structures. By using the feature extraction algorithm, we found that more than approximately 60% of the attached momentum regions extend beyond the log region, while most of the detached regions formed locally away from the wall have relatively short wall-normal extent. The two-point correlation of the streamwise velocity fluctuations showed that the wall-normal length scales vary linearly with the distance from the wall, and, finally, the two-point correlations between the signed swirling motions showed that interactions between counterrotating vortices are predominant throughout the boundary layer and especially frequent in the near-wall region.

The coherent structures in turbulent boundary layers (TBLs) subjected to adverse pressure gradients (APGs) were investigated by analyzing a database of direct numerical simulations. The equilibrium adverse pressure gradient flows were established by using a power law free-stream distribution. The population trends of the spanwise vortices show that the outer regions of the APG TBLs are densely populated with hairpin-like vortices. These vortical structures induce low-momentum regions in the middle of the boundary layers, which result in an outer peak in the Reynolds shear stress. The 3-D features of the hairpin packets were deduced from their spatial characteristics in the spanwise-wall-normal plane. The conditionally averaged velocity fields show that there are counter-rotating v − w swirling motions that represent cross-sectional evidences of the packets. Moreover, two-point correlations and linear stochastic estimations were used to provide statistical information about the hairpin packet motions in the crossstream planes of the APG TBLs.