Longitudinal vortices in unsteady Taylor-Couette flow: solution to a 60-year-old mystery

This paper numerically resolves a 60-year-old mystery regarding Coles' 1965 observation of transient longitudinal vortices in unsteady Taylor-Couette flow by demonstrating that these structures arise from an inflection in the azimuthal velocity profile during deceleration, linking the instability to the Stokes oscillating boundary layer problem while explaining its historical elusiveness through radius ratio dependencies and competition with Görtler rolls.

The Gist

Imagine two giant, hollow pipes, one sitting inside the other. If you spin the inner pipe, the fluid in between usually forms neat, donut-shaped rings that circle around the pipes. This is a classic physics puzzle known as Taylor–Couette flow, and scientists have been studying it for over a century.

However, back in 1965, a scientist named Coles noticed something weird. When he suddenly stopped the outer pipe after spinning it, the fluid didn't just slow down smoothly. Instead, it briefly formed strange, long, straight lines running up and down the pipes, like the stripes on a candy cane. These "longitudinal vortices" were a mystery for 60 years. Why did they appear? Why didn't they show up more often?

This paper solves that 60-year-old mystery using powerful computer simulations guided by a recent, lucky experiment. Here is the story of what they found, explained simply.

The Mystery of the Candy Cane Stripes

For decades, scientists thought these strange stripes might be caused by a specific type of friction instability (called Tollmien instability) that happens when a fluid speeds up against a wall. It's like the ripples you see when wind blows over a calm lake.

But the authors of this paper discovered that's not the whole story. They found that these stripes actually appear during the deceleration phase—the moment the outer pipe is slowing down to a stop.

The "Speed Bump" in the Fluid

To understand the cause, imagine the fluid's speed as a hill.

• Normal flow: The speed changes smoothly from the spinning wall to the stationary wall, like a gentle, curved ramp.
• The Mystery Moment: When the outer wall suddenly slows down, the fluid near the wall slows quickly, but the fluid in the middle is still moving fast. This creates a weird "speed bump" or a sharp kink in the flow profile.

The authors found that this sharp kink (which they call an inflection point) is the trigger. It's like a speed bump on a highway that causes cars to swerve. In the fluid, this kink causes the smooth flow to break apart and snap into those straight, vertical stripes.

The Connection to a Classic Wave Problem

The paper links this phenomenon to a very old physics problem solved by George Stokes in the 1800s regarding waves in a fluid caused by a moving plate. The authors show that the Taylor–Couette system, when it speeds up and slows down, behaves mathematically like Stokes' oscillating wave problem.

Think of it like this: The fluid in the gap is acting like a drum skin. When you hit it (start) and let it go (stop), it doesn't just vibrate randomly; it creates a specific, predictable pattern of ripples. The authors proved that the "candy cane stripes" are essentially the fluid's version of these Stokes waves, triggered specifically when the outer wall is braking.

Why Was This So Hard to Find?

You might wonder, "If this happens, why didn't everyone see it before?" The paper explains three main reasons:

1. The "Goldilocks" Gap: The size of the gap between the pipes matters immensely.

   • If the gap is too wide, the fluid gets confused by the curvature of the pipes, and the stripes get swallowed up by a different, more chaotic type of swirl (called Görtler rolls).
   • If the gap is too narrow, the effect is too small to see.
   • Coles happened to use a gap size that was just right to see the stripes, but he didn't realize how sensitive the effect was to that specific size.

2. The Timing is Fleeting: These stripes are incredibly short-lived. They only exist for a split second while the outer pipe is slowing down. If you look too early (while it's speeding up) or too late (after it stops), they are gone. It's like trying to photograph a hummingbird's wings; if your camera shutter is even a fraction of a second off, you miss it.

3. The Need for a Nudge: The fluid is very stable. To get these stripes to form, you need a tiny bit of "noise" or disturbance to kick things off. In a perfectly smooth, idealized lab, the stripes might never start. In the real world, vibrations or the ends of the pipes provide that tiny nudge.

The Bottom Line

The paper concludes that the "candy cane stripes" Coles saw were not a fluke, but a specific, predictable instability caused by the fluid's speed profile getting "kinked" during deceleration. It is a beautiful example of how a simple action—stopping a spinning cylinder—can reveal a hidden, complex dance in the fluid that had been hiding in plain sight for 60 years.

The authors suggest that with modern laser cameras (which can see these tiny, fast movements much better than old-school photography), we might start seeing these stripes in many more experiments, provided we get the gap size and the stopping speed just right.

Technical Summary

Technical Summary: Longitudinal Vortices in Unsteady Taylor–Couette Flow

Problem Statement
This study addresses a 60-year-old mystery regarding a specific flow instability observed in the classic work of Coles (1965). In experiments involving a "start-stop" motion of the outer cylinder in a Taylor–Couette system, Coles recorded the formation of longitudinal vortices (rolls aligned with the cylinder axis) rather than the standard azimuthally aligned Taylor-vortex rolls. While Coles hypothesized these might result from Tollmien–Schlichting instability in the unsteady viscous layer, he provided no parameters for the specific start-stop event, and the phenomenon has remained elusive in subsequent systematic studies. Recent, limited experimental observations by Burin & Czarnocki (2012) confirmed the existence of these structures during the deceleration phase of the outer cylinder but only for a very narrow gap width. The primary objective of this work is to numerically simulate this transitory flow state to identify the precise instability mechanism, determine the conditions under which longitudinal vortices emerge, and explain why they have been difficult to observe.

Methodology
The authors employ direct numerical simulations (DNS) using a modified variant of the Openpipeflow solver. The governing equations are the incompressible Navier–Stokes equations, non-dimensionalized using the gap width $d$ and the viscous diffusion timescale $d^2/\nu$.

• Flow Configuration: The study models the flow between concentric cylinders with a fixed inner cylinder ($Re_i = 0$) and an outer cylinder subjected to a time-dependent rotation rate. The outer cylinder's Reynolds number, $Re_o(t)$, follows a linear ramp-up and ramp-down profile (start-stop motion) to approximate the experimental conditions.
• Numerical Approach: The velocity field is decomposed into a time-dependent base flow (the instantaneous Couette profile) and a perturbation field. The perturbation is discretized using a double-Fourier decomposition in the axial and azimuthal directions and finite differences (Chebyshev collocation) in the radial direction.
• Analysis: The authors analyze the linear stability of the "frozen" mean flow profiles at various time steps during the acceleration and deceleration phases. They calculate eigenvalues (growth rates) and eigenfunctions to identify the dominant instability modes. The study also draws a theoretical parallel to the Stokes oscillating boundary layer problem to contextualize the instability mechanism.

Key Results

1. Mechanism of Instability: The simulations confirm that the longitudinal vortices arise specifically during the deceleration phase of the outer cylinder. The instability is linked to the formation of an inflection point in the azimuthal velocity profile, which occurs as the outer wall slows down. This finding contradicts the initial hypothesis that the instability is a Tollmien–Schlichting phenomenon (which would be associated with the acceleration/boundary layer formation phase).
2. Connection to Stokes' Problem: Despite the transient nature of the start-stop flow, the authors demonstrate a strong link to the linear stability of Stokes' oscillating boundary layer problem. By mapping the start-stop ramp-down to half a period of a sinusoidal oscillation, the numerical results align with theoretical predictions for the most unstable wavenumber and growth rates derived from the Stokes problem.
3. Dependence on Radius Ratio ($\eta$): The study investigates the sensitivity of the instability to the radius ratio $\eta = R_i/R_o$.
   • For narrow gaps (e.g., $\eta = 0.97$), the longitudinal vortices are clearly observable.
   • As the gap widens (lower $\eta$), the instability is rapidly overwhelmed by axisymmetric Görtler rolls (centrifugal instability).
   • Coles' original choice of $\eta = 0.874$ is identified as a serendipitous "borderline" case where both the longitudinal instability and Görtler rolls might coexist, explaining why the phenomenon was visible in his specific setup but likely suppressed in wider gaps where Görtler rolls dominate.
4. Why the Mystery Persisted: The paper outlines several reasons for the elusiveness of this instability:
   • Parameter Sensitivity: It requires a specific combination of start-stop motion and Reynolds number that is not typical for standard stability studies.
   • Competition with Other Modes: In wider gaps, Görtler rolls grow much faster, masking the longitudinal vortices.
   • Visualization Challenges: The instability occupies a small fraction of the gap depth (scaling with the viscous penetration depth $\delta$), making it difficult to detect with traditional visualization techniques (e.g., crystalline platelets) unless the contrast is high.
   • Transient Nature: The instability exists only briefly during deceleration, requiring precise timing for observation.

Significance and Claims
The authors claim to have provided the first systematic numerical observation and explanation of the longitudinal vortices described by Coles. The primary significance of the work lies in:

• Resolving the Mechanism: Correctly identifying the instability as an inflection-point instability occurring during deceleration, distinct from Tollmien–Schlichting instability.
• Unifying Framework: Linking a complex, unsteady Taylor–Couette phenomenon back to the fundamental Stokes oscillating boundary layer problem, thereby placing the observation within a broader context of fluid mechanics history.
• Explaining Observational Gaps: Providing a physical rationale for why this instability has remained elusive, specifically highlighting the competition with Görtler rolls at lower radius ratios and the practical difficulties of visualizing thin, transient structures in wider gaps.

The paper concludes modestly, noting that while the mechanism is now understood, the instability remains difficult to observe experimentally without specific conditions (narrow gaps, high Reynolds numbers, and precise start-stop timing) or advanced diagnostic tools like Particle Image Velocimetry (PIV).