Abstract
Transient growth due to non-normality is investigated for the Taylor–Couette problem with counter-rotating cylinders as a function of aspect ratio $\eta$ and Reynolds number $Re$. For all $Re \le 500$, transient growth is enhanced by curvature, i.e., is greater for $\eta < 1$ than for $\eta = 1$, the plane Couette limit. For fixed $Re < 130$ it is found that the greatest transient growth is achieved for $\eta$ between the Taylor–Couette linear stability boundary and one, while for $Re > 130$ the greatest transient growth is achieved for $\eta$ on the linear stability boundary. Transient growth is shown to be approximately 20% higher near the linear stability boundary at $Re = 310, \eta = 0.986$ than at $Re = 310, \eta = 1$. The energy in the optimal inputs is primarily meridional; that in the optimal outputs is primarily azimuthal.
Introduction
In fluid dynamics, two "Couette" flows serve as fundamental benchmarks: Taylor-Couette flow (TCF), the fluid between two rotating cylinders, and Plane Couette flow (PCF), the fluid between two sliding flat plates. While TCF is well-understood through linear stability analysis, PCF remains a "paradigm of mystery." Mathematically, PCF is linearly stable at all speeds, yet it abruptly transitions to turbulence in practice. This study explores the hypothesis that "non-normal" operators allow for transient growth—a temporary spike in energy that triggers nonlinear effects—and investigates how the curvature of TCF influences this phenomenon.
Mapping the Geometry
To investigate the transition from flat plates to curved cylinders, the researchers treated TCF as a tunable "embedding" for PCF. By adjusting the radius ratio ($\eta$), they could move from a highly curved system to a nearly flat one. This requires a precise mapping of coordinates and wavenumbers between the two systems to ensure consistent comparison.
Methodology
The study analyzed the linearized Navier-Stokes equations for an incompressible fluid, focusing on counter-rotating cylinders ($\mu = -1$) and axisymmetric perturbations ($m=0$). Key technical components included:
- Pseudospectral Methods: Chebyshev polynomial expansions were used to discretize the equations.
- Energy Norm & SVD: Singular Value Decomposition (SVD) identified the "optimal growth" ($G_{max}$), finding the initial conditions (optimal inputs) that yield the maximum possible energy gain at a later time (optimal outputs).
- Pseudospectra: The researchers calculated $\epsilon$-pseudospectra to quantify the sensitivity of the system's eigenvalues to small perturbations, a hallmark of non-normal operators.
Key Insights: Curvature as an Enhancer
Contrary to intuition, adding curvature ($\eta < 1$) actually increases the maximum transient growth compared to the flat-plate limit ($\eta = 1$). This enhancement is driven by the "Lift-Up" mechanism: optimal inputs consist of meridional vortices (circular rolls) that "lift up" fluid to create optimal outputs in the form of high-energy azimuthal streaks.
For low speeds ($Re < 130$), the highest growth occurs at an intermediate curvature. However, for higher speeds ($Re > 130$), the maximum growth is consistently found at the edge of the linear stability boundary. Near the transition threshold observed in flat plates ($Re \approx 310$), the transient growth in a narrow-gap TCF ($\eta = 0.986$) is approximately 20% higher than in the flat PCF limit.
Pseudospectral Signatures
In narrow gaps ($\eta = 0.99$), the pseudospectra protrude significantly into the unstable half-plane. This signals that the system is highly sensitive and prone to non-normal growth, even when technically stable. For large curvature ($\eta = 0.5$), the pseudospectra adhere more closely to the spectrum, indicating less sensitivity compared to the narrow-gap case.
Conclusion
This research demonstrates that curvature is a significant driver of transient growth, potentially lowering the threshold for turbulence transition. While the study focuses on linear mechanisms, the 20% growth boost observed near the PCF transition threshold suggests that curvature plays a vital role in the nonlinear "spark" that leads to turbulence. Future work will need to address non-axisymmetric spiral modes and experimental validation to see if turbulence appears in TCF at $Re \approx 300$ outside of mathematically defined instability zones.
Summary updated