3.1. From Classical Tesla Geometry to Modern Valve Architectures
The evolution of Tesla valves reflects a progressively deeper understanding of directional flow-control mechanisms. Since Tesla introduced the concept of the valvular conduit in 1920, a wide range of modified configurations has emerged for applications in microfluidics, thermal management, and intelligent water systems [
41,
42,
43,
44,
45,
46]. Despite their structural diversity, these developments have been driven by a common objective: maximizing the dissipation asymmetry between forward and reverse flows through channel-topology design, thereby achieving efficient passive rectification and flow regulation without moving parts [
47,
48,
49].
Based on current studies, the geometric evolution of Tesla valves can be broadly summarized into three main pathways (
Table 4).
Early studies primarily focused on evaluating the hydrodynamic validity of the classical configuration. With advances in experimental diagnostics and computational fluid dynamics, researchers gradually recognized that rectification does not arise from simply increasing flow-path length, but from the generation of highly dissipative flow phenomena during reverse flow, including flow diversion, redirection, boundary-layer separation, recirculation, and jet interaction [
41]. This mechanistic understanding provided the foundation for subsequent structural innovations.
As illustrated in
Figure 7, typical configurations include the T45-R valve, D-valve, and GMF valve. The T45-R valve is characterized by an approximately 45-degree deflection structure, offering compact geometry and improved rectification performance compared with conventional diffuser-type check valves in piezoelectric micropumps. The D-valve enhances reverse flow resistance by aligning inlet and outlet while introducing strongly curved side branches and has been applied in oscillating heat pipe systems. The GMF valve further extends the T45-R configuration by lengthening and strengthening branch channels to intensify disturbance and dissipation during reverse flow. These designs demonstrate that Tesla valves have evolved from a single prototype into a family of structures with distinct flow logic [
42].
Importantly, the development of modern configurations does not follow the simplistic assumption that greater geometric complexity always leads to better performance. Recent studies on bifurcated structures indicate that the branching section in classical Tesla valves does not universally enhance rectification [
42,
43,
44,
45,
46,
47]. As illustrated in
Figure 8, removing the bifurcation in the T45-R valve can improve diodicity at a high Reynolds numbers, whereas in the GMF valve, such structures are essential for maintaining performance. The D-valve exhibits a transitional behavior dependent on the Reynolds number [
42]. These findings suggest that geometric evolution is fundamentally driven by the identification and reinforcement of dominant dissipation mechanisms rather than the preservation of traditional structural elements [
47,
48,
49].
As shown in
Table 5, the three valves respond very differently to the removal of the bifurcated section. The bifurcated D-valve offers the most robust performance across the entire
Re range (
Di = 1.94 at
Re = 2000), while the non-bifurcated T45-R actually outperforms its bifurcated counterpart at high
Re (1.66 vs. 1.49). In contrast, the GMF valve suffers a significant drop without bifurcation and its bifurcated version peaks early then declines. These results demonstrate that the optimal choice depends strongly on the target Reynolds number and the specific role of the bifurcated geometry. Therefore, the evolution from classical two-dimensional configurations to modern variants can be interpreted as a process of decomposition of geometric functions, selection of key structural elements, and reconstruction of flow mechanisms [
50]. This perspective reinforces that no single design universally excels; instead, the geometry should be tailored to the intended operating conditions.
Building upon modern variants, Tesla valve development has further progressed toward performance-driven and functionally integrated designs. Research emphasis has shifted from configuration comparison to objective-oriented design, focusing on rectification enhancement, operating-condition expansion, and multifunctional integration [
66].
3.2. Geometry-Induced Regulation of Dissipation Mechanisms
Geometric parameter optimization represents a critical link between structural design and hydraulic performance in Tesla valves. Rather than identifying a universal set of optimal dimensions, the primary objective of parameter studies is to establish how geometric features influence local flow structures and, consequently, rectification performance. From this perspective, apparently inconsistent conclusions reported in the literature can often be interpreted within a common framework that relates geometry, flow mechanisms, and performance outcomes [
42,
44].
Existing studies generally distinguish between two categories of geometric parameters. The first comprises local parameters, including width ratio, curvature radius, side-branch dimensions, and transition-region geometry, which primarily control local separation, recirculation, jet interaction, and remerging processes [
42]. The second includes global parameters, such as branching angle, merging angle, and main-channel dimensions, which determine the overall organization of flow paths and momentum redistribution [
46]. Despite their different roles, the common objective of geometric optimization is to strengthen reverse-flow dissipation while maintaining acceptable forward-flow resistance, thereby maximizing directional asymmetry in pressure loss [
47,
48,
49].
Among the global design parameters, branching and merging angles are widely recognized as major determinants of rectification performance. Small branching angles allow reverse flow to pass through relatively direct pathways, limiting flow diversion and momentum loss, whereas excessively large angles may increase dissipation in both directions and reduce overall hydraulic efficiency. Consequently, most studies report an optimal intermediate range that reflects a balance between low-resistance forward transport and enhanced reverse-flow dissipation. Numerical investigations further indicate that angle variations alter vortex formation, jet interaction, and velocity-field asymmetry, leading to different rectification behaviors across Reynolds-number regimes [
48]. Similar observations have been reported in Tesla-inspired micromixers, where geometric angles govern flow splitting, redirection, and remerging, all of which are closely associated with energy-dissipation mechanisms [
54].
The dimensions of the main channel also play an important role because they influence both the relative importance of inertial and viscous effects and the flow distribution between the main passage and side branches [
46]. Liu et al. demonstrated that geometric scaling significantly affects the response of the flow field to channel asymmetry, implying that optimal parameters are not universally transferable across different size scales [
27]. At small scales, viscous effects suppress flow separation and recirculation, whereas at larger scales or higher Reynolds numbers, inertial effects promote the formation of high-dissipation structures in the reverse flow [
41]. As a result, geometric optimization must always be considered within the context of the intended operating conditions.
Local geometric features exert a more direct influence on the development of dissipation structures. Width ratio, for example, affects local velocity distribution and pressure-loss allocation. Increasing the contraction ratio generally accelerates the flow and promotes separation and vortex formation in expansion or merging regions, thereby enhancing reverse-flow dissipation. However, excessive contraction may simultaneously increase forward-flow losses, indicating a trade-off between stronger reverse-flow resistance and acceptable hydraulic efficiency. Similar conclusions have been reported by Gamboa et al., who showed that local width distribution strongly affects flow partitioning and pressure-drop characteristics in Tesla-type micropump structures [
49].
Curvature radius influences rectification performance through its effect on centrifugal forces and secondary-flow development. Smaller curvature radii tend to intensify flow separation and recirculation under reverse-flow conditions, increasing energy dissipation. However, excessive curvature may also introduce significant losses in the forward direction. Differences in Reynolds-number sensitivity among configurations such as the D-valve and GMF valve have been partially attributed to variations in branch curvature and the resulting dissipation mechanisms [
42].
The dimensions and placement of side branches likewise play a crucial role in determining flow behavior. Branches that are too small may fail to induce sufficient flow diversion, whereas excessively large branches can create extensive stagnant regions and unnecessary hydraulic volume. Studies by Wiley and Huang further demonstrated that the influence of bifurcation structures is strongly configuration-dependent: removing or weakening a bifurcated section may either improve or degrade rectification performance depending on the valve architecture [
42]. These findings suggest that local geometric complexity alone does not guarantee superior performance; rather, effective dissipation depends on the coordinated interaction between branches, junctions, and the main flow path.
Although broad consensus exists regarding the mechanisms responsible for rectification—namely flow diversion, jet interaction, separation, recirculation, and remerging losses—substantial variation remains in the reported optimal geometric parameters [
41]. Much of this variation originates from differences in operating conditions and evaluation criteria rather than genuine contradictions among studies. The Reynolds number is particularly influential, as geometric features that have little effect under low-Reynolds-number conditions may become dominant once flow separation and vortex formation emerge at higher Reynolds numbers [
41]. For example, Wiley and Huang reported that removing bifurcation structures from a D-valve reduced diodicity at low Reynolds numbers but produced progressively less detrimental effects as Reynolds number increased [
42].
Performance metrics also influence optimization outcomes. Fundamental studies often seek to maximize diodicity, whereas practical applications typically require simultaneous consideration of pressure drop, pumping power, mixing efficiency, heat-transfer enhancement, or other system-level objectives [
61]. Consequently, the geometry that maximizes diodicity is not necessarily the geometry that delivers the best overall performance. Similar observations have been reported in Tesla-inspired micromixer optimization, where power-dissipation-based objectives produced different optimal designs from those obtained using mixing efficiency alone [
52].
Additional discrepancies arise from differences in modeling assumptions and flow conditions, including two-dimensional versus three-dimensional geometries, steady versus unsteady flow, and single-phase versus multiphase systems [
56]. Therefore, geometric optimization should not be viewed as the search for universally optimal parameter values. Instead, it should be understood as the process of identifying how geometric features modify dominant flow mechanisms under specific operating conditions and how those mechanisms ultimately determine rectification performance [
66].
3.3. Multi-Stage and Three-Dimensional Strategies for Dissipation Amplification
The limited rectification capability of a single Tesla valve has driven the development of advanced dissipation-enhancement strategies. Beyond conventional geometric optimization, recent studies increasingly focus on multi-stage cascading, three-dimensional architectures, and function-oriented designs to amplify asymmetric energy dissipation and extend operating ranges [
43].
Multi-stage cascading is the most widely adopted approach for rectification enhancement. By connecting multiple Tesla units in series, reverse-flow dissipation can be accumulated through repeated processes of flow splitting, separation, recirculation, and recombination, thereby increasing the overall pressure-drop asymmetry [
45]. However, performance improvement is generally non-linear because stage-to-stage flow interactions, pressure recovery, and additional forward-flow losses gradually reduce the marginal benefit of adding more stages [
47].
Three-dimensional and helical configurations provide another important route for dissipation amplification. By introducing spatial curvature, twisting, and cross-sectional flow circulation, these structures generate secondary flows and Dean vortices that are absent in conventional planar designs [
50]. The resulting three-dimensional momentum exchange enhances reverse-flow energy dissipation while improving spatial compactness. Similar principles have also been exploited in Tesla-inspired micromixers, where chaotic advection significantly improves mixing performance [
53].
To broaden application capabilities, Tesla valves have increasingly been integrated with nozzle, throttling, contraction–expansion, and obstacle-assisted structures [
58]. These modifications strengthen jet impingement, flow separation, and shear-layer instability under reverse flow, making the valves more suitable for high-pressure-drop, pulsating, and compressible-flow conditions [
59]. In parallel, irregular, obstacle-enhanced, and adaptive configurations have been proposed to improve operational flexibility and multifunctionality, extending Tesla valves from passive fluidic diodes to integrated platforms for flow regulation, thermal management, mixing enhancement, and multiphase-flow control [
62].
Overall, the evolution from multi-stage cascading to three-dimensional and functionally integrated architectures reflects a common objective: maximizing asymmetric dissipation while adapting Tesla valves to increasingly complex engineering applications [
44].
3.4. Design Methodologies for Next-Generation Tesla Valves
The evolution of the Tesla valve architecture has been accompanied by a fundamental shift in design philosophy, from intuition-driven geometry modification to data-driven and objective-oriented optimization. Early Tesla valve development largely relied on empirical design and iterative trial-and-error approaches, where a candidate geometry was proposed based on fluid–mechanical intuition and subsequently evaluated through experiments or computational fluid dynamics (CFD) simulations [
44]. While effective for revealing physical mechanisms, this forward-design paradigm becomes increasingly inefficient when confronted with high-dimensional design spaces and multiple competing performance objectives.
As application requirements expand beyond rectification alone to include pressure loss, compactness, mixing efficiency, thermal performance, and structural constraints, Tesla valve design has progressively evolved toward intelligent inverse-design frameworks [
51]. Rather than evaluating predefined geometries, modern approaches formulate optimization objectives first and then search the design space systematically to identify optimal flow configurations.
The earliest stage of this transition was based on parametric studies and sensitivity analysis. Key geometric variables, including branch angle, channel-width ratio, and curvature radius, were varied within established valve configurations such as T45-R and D-valve designs to assess their influence on performance [
42,
55]. These studies provided valuable physical insight into the relationship between geometry and flow behavior, revealing how individual parameters affect vortex formation, flow separation, and momentum dissipation. However, because geometric variables often interact nonlinearly, conventional parameter scanning is prone to local optima and becomes computationally expensive as the number of design variables increases [
57].
To address these limitations, surrogate-model-based optimization has emerged as a powerful alternative. As illustrated in
Figure 9, by combining design-of-experiment techniques with response surface models, Kriging methods, or machine-learning-based surrogate models, researchers can establish efficient mappings between geometric parameters and performance metrics [
60,
63]. These reduced-order models significantly decrease computational cost while enabling global optimization through evolutionary algorithms such as NSGA-II and particle swarm optimization. The resulting Pareto-optimal solutions allow designers to balance competing objectives, including rectification performance, pressure loss, mixing efficiency, and energy consumption.
As illustrated in
Figure 10 and
Figure 11, more recently, topology optimization has extended Tesla valve design beyond predefined geometric templates. Instead of modifying existing flow passages, topology optimization treats the entire design domain as a continuous medium and determines the optimal material distribution directly from governing equations and objective functions [
64]. Density-based and level-set methods have been widely applied to fluidic diode design, generating highly non-intuitive channel topologies that differ substantially from traditional Tesla valve configurations. Studies have shown that such generated structures can achieve superior reverse-flow dissipation and broader operating ranges compared with conventional designs [
65].
Looking forward, topology optimization is increasingly being integrated with machine learning, physics-informed neural networks, and generative artificial intelligence. This convergence is expected to enable automated design workflows that combine physical constraints, performance prediction, and advanced manufacturing, providing a promising pathway toward the next generation of high-performance passive flow-control devices [
41,
66] (
Figure 12).
As shown in
Table 6, surrogate-based approaches (RSM, Kriging, machine learning) require significant upfront investment in training data and model training but offer fast online predictions, enabling global optimization that is infeasible with direct CFD alone. The key trade-off is that while direct methods (parametric studies, topology optimization) avoid offline training, their per-evaluation cost is substantially higher, making them impractical for high-dimensional design spaces. For Tesla-valve optimization involving more than 5–10 geometric variables, machine-learning-based surrogates or hybrid approaches generally offer the most favourable balance between computational investment and optimisation efficacy.
3.5. Summary and Design Guidelines for Structural Evolution
As shown in
Table 7, the evolution of Tesla-valve designs from classical single-stage configurations to modern planar variants, multistage cascades, three-dimensional architectures, and topology-optimized geometries reflects a progressive enhancement in rectification capability, but each design category entails distinct trade-offs. Modern planar variants such as the D-valve offer robust performance (
Di = 1.94 at
Re = 2000) with moderate forward losses, making them suitable for a wide range of applications [
42]. Multistage cascading significantly increases diodicity (up to 3.58 for
N = 10 at
Re = 450) but suffers from diminishing returns and elevated forward pressure drops as stages are added [
43,
45,
47]. Three-dimensional and helical configurations introduce additional dissipation mechanisms (e.g., Dean vortices) that are absent in planar designs, though experimental validation remains limited [
50,
53]. Topology-optimized and bio-inspired designs represent the frontier of Tesla-valve development, offering the potential for superior performance and broader operating ranges, but they face substantial challenges in computational cost, manufacturing complexity, and experimental verification [
60,
63,
64,
65,
66]. Overall, the optimal design choice depends strongly on the target Reynolds number, acceptable forward losses, fabrication constraints, and specific application requirements.