A turbulent velocity correlation can redistribute momentum into a structured mean flow without adding net momentum to the whole periodic domain. The local acceleration depends on the spatial gradient of that correlation, not simply on whether the correlation is positive.
The short answer is to write the averaged momentum balance, retain its boundaries and check the resulting energy transfer separately. We will solve an original local model, reconstruct a periodic flow, add damping conditionally and test what a measured covariance can identify. Every velocity, length, time and coefficient is fictional teaching data. The model is a deliberately restricted bridge to the plasma Zonal Flow root, not a prediction or control prescription for a fusion device.
eduKateSG · Zonal Flow
Find your next learning step
Choose the question closest to your task, or follow the original worked model in order. All teaching chapters remain open.
ROUTE 1 · CHAPTERS 1–6
Define the local momentum model
Name coordinates, covariance, acceleration and boundary transport.
ROUTE 2 · CHAPTERS 7–13
Work the periodic example
Check velocity, net momentum and the separate energy account.
ROUTE 3 · CHAPTERS 14–19
Add damping and measurement structure
Examine damping, viscosity, fluctuation phase and finite-volume conservation.
ROUTE 4 · CHAPTERS 20–25
Check evidence and transfer
Test resolution, geometry and an independent changed-scale calculation.
ROUTE 5 · CHAPTERS 26–32
Challenge the interpretation
Reverse the initial flow, compare explanations and close boundary and sign checks.
Chapter contents
Define the local momentum model · Chapters 1–6
1. Define the local coordinates and mean
2. Name the covariance before calling it a drive
3. Derive the restricted averaged momentum equation
4. Distinguish acceleration from shear
Work the periodic example · Chapters 7–13
7. Evolve an initially stationary mean flow
8. Read the spatial ledger with its signs
9. Check a constant covariance as a counterexample
10. Reconstruct the covariance from an acceleration profile
11. Calculate mean-flow kinetic energy separately
12. Derive the covariance work term
13. Recognise the role of the velocity gradient in energy transfer
Add damping and measurement structure · Chapters 14–19
14. Add linear damping as an explicit new term
15. Check the damped energy ledger
16. Include viscosity without hiding its spatial operator
17. Relate covariance to fluctuation phase
18. Use a realizability bound without treating it as a drive law
Check evidence and transfer · Chapters 20–25
20. Separate a cell average from an unresolved local peak
21. Check the noise cost of a small spatial interval
22. Keep density and geometry outside the constant-coefficient shortcut
23. Use the primary plasma study for its interpretive lesson
Challenge the interpretation · Chapters 26–32
26. Test energy transfer when the initial flow opposes the drive
27. Compare damping explanations using the same observed velocity
28. Scale the response without claiming universal plasma behaviour
29. Make a boundary experiment close its momentum ledger
30. Check the signs with a small displacement
CHAPTER 1 OF 32 · Define the local momentum model
1. Define the local coordinates and mean
Let x represent a local cross-surface direction and y a local along-surface direction. The root’s plasma setting is important: these are not automatically geographical east–west and north–south coordinates. A full toroidal plasma description requires its actual magnetic geometry and averaging conventions.
For the teaching model, assume a local Cartesian, constant-density, incompressible velocity field with zero mean radial velocity. Let U(x,t) be the averaged along-surface velocity. Write the radial fluctuation as u′ and the along-surface fluctuation as v′, each with zero mean under the specified average.
The average may idealise a suitable periodic direction or ensemble. It must be defined consistently when splitting a signal into mean and fluctuation. A correlation computed after one background subtraction need not equal a correlation computed after another. We will keep the averaging operator fixed so the algebra addresses one clearly defined mean-flow quantity.
CHAPTER 2 OF 32 · Define the local momentum model
2. Name the covariance before calling it a drive
Define S(x,t) = ⟨u′v′⟩. It has units of squared velocity, metres squared per second squared. In a constant-density momentum-flux account, multiplying it by density supplies the corresponding dimensional momentum-flux term.
Some sign conventions define a turbulent stress with an additional minus sign. Our S is explicitly the velocity covariance, so its sign and the sign in the momentum equation must be read together. A positive plotted number does not establish a positive acceleration without that definition.
The covariance differs from ⟨u′⟩⟨v′⟩, which is zero by the fluctuation convention. It records a joint relationship between the fluctuations. Separate average velocities do not supply it. The relation between this covariance and a mean flow is dynamical: the spatial organisation of momentum transport enters the averaged balance, rather than the covariance acting as a uniform source term at every point.
CHAPTER 3 OF 32 · Define the local momentum model
3. Derive the restricted averaged momentum equation
For the stated local incompressible model, the fluctuating contribution to the along-surface momentum flux is S. Averaging the conservation equation and suppressing the other terms in this first exercise gives ∂U/∂t = −∂S/∂x.
This is a model contract, not a complete plasma momentum equation. We have excluded mean radial advection, pressure and geometry contributions, magnetic stresses, damping, viscosity and external forcing unless added explicitly later. Real zonal-flow interpretation must account for the relevant omitted terms.
The minus sign follows the divergence form of the flux. If more along-surface momentum leaves a small radial interval than enters it, that interval’s mean momentum decreases. If transport converges, it increases. This is the physical meaning of the derivative. Before calculating, label the direction of positive x, the positive along-surface velocity and the definition of S so that the sign of a result can be interpreted consistently.
CHAPTER 4 OF 32 · Define the local momentum model
4. Distinguish acceleration from shear
The stress-gradient contribution aR = −∂S/∂x has units of metres per second squared. It is an acceleration. Flow shear ∂U/∂x has units of inverse seconds. The velocity U itself has units of metres per second.
A large covariance can have a small gradient if it varies slowly. A large velocity can have little shear if it is nearly uniform. A large instantaneous acceleration does not establish how much velocity has already accumulated, because the initial state and time history matter.
These distinctions are useful when comparing plots. A curve labelled Reynolds stress, another labelled shear and another labelled zonal velocity are not interchangeable views of one numerical quantity. The model connects them through derivatives and time evolution, with their own units and boundary assumptions. Preserving those operations prevents a physically interesting correlation from being promoted directly into a confinement result that the plotted variable does not establish.
CHAPTER 5 OF 32 · Define the local momentum model
5. Integrate the momentum balance over a domain
Integrate from x = zero to x = D. The mean-flow momentum per unit density changes according to d∫U dx/dt = −[S(D,t) − S(0,t)]. The boundary flux difference controls the integrated effect.
For a periodic domain, the boundary values agree, so the Reynolds-stress term does not change the domain-integrated mean momentum. Local accelerations can still be nonzero and alternate in sign. Momentum is redistributed within the domain rather than supplied as a net input through its boundaries.
This conclusion concerns the isolated flux term under the stated boundaries. If other forces or unequal boundary fluxes are present, they contribute to the integrated balance. A zero integral also does not mean every local flow vanishes. The next example constructs a nonuniform velocity field with zero spatial mean, making the difference between local structure and net momentum explicit.
CHAPTER 6 OF 32 · Define the local momentum model
6. Prescribe an original periodic stress profile
Choose a domain of length D = two πL and define S(x) = S0 sin(x/L), constant in time over the illustrative interval. Let S0 = twelve metres squared per second squared and L = 0.2 metre.
Differentiation gives aR(x) = −(S0/L)cos(x/L) = −sixty cos(x/L) metres per second squared. The spatial average is zero over the complete period, while the acceleration is locally positive or negative.
The stress amplitude S0 is not itself an acceleration. Dividing by the stated length scale produces the acceleration scale. This dimensional step also explains why the same covariance amplitude distributed over a shorter spatial scale can produce a larger gradient in the mathematical model. It does not imply that a real plasma can freely change that scale without altering its other dynamics or the validity of the local approximation.
CHAPTER 7 OF 32 · Work the periodic example
7. Evolve an initially stationary mean flow
Take U(x,zero) = zero and retain the prescribed stress with no other terms. Integrating in time gives U(x,t) = −(S0/L)t cos(x/L). At t = 0.05 second, the velocity is minus three cos(x/L) metres per second.
The flow has positive and negative bands, with zero spatial mean. Its shear is (S0/L²)t sin(x/L), which at that time is fifteen sin(x/L) inverse seconds. The strongest acceleration and strongest shear occur at different positions in this example.
This solution is conditional on holding the stress profile prescribed over the interval. It does not solve the turbulence that generated S or the feedback of U on that turbulence. A growing mean-flow response in this forced exercise should therefore not be described as a complete self-consistent saturation theory. The calculation isolates the consequences of one supplied momentum-flux profile so those consequences can be checked cleanly.
CHAPTER 8 OF 32 · Work the periodic example
8. Read the spatial ledger with its signs
At t = 0.05 second, the original model gives the following values at four phase positions. The phase coordinate ξ means x/L.
| Phase ξ | Covariance S, square metres per square second | Stress-gradient acceleration, metres per square second | Mean velocity, metres per second |
|---|---|---|---|
| zero | zero | minus sixty | minus three |
| π/two | twelve | zero | zero |
| π | zero | sixty | three |
| three π/two | minus twelve | zero | zero |
A positive covariance at ξ = π/two coincides with zero acceleration there. At ξ = zero, the covariance is zero but its gradient produces a negative acceleration. The table directly challenges the shortcut that a positive covariance must be a positive local drive. Sign interpretation requires the derivative, coordinate convention and momentum equation together.
CHAPTER 9 OF 32 · Work the periodic example
9. Check a constant covariance as a counterexample
If S is the same nonzero constant at every x, then ∂S/∂x is zero and the restricted Reynolds-stress contribution produces no local acceleration. The covariance can be large without generating a gradient-driven mean-flow change in this model.
On a finite domain, equal constant flux at both ends also gives no integrated momentum change. This does not say that no momentum crosses either boundary; it says that the entering and leaving contributions cancel in the net balance.
The example distinguishes a transported quantity from its convergence. It also limits an interpretation based only on a spatially averaged stress magnitude. Averaging can remove the derivative structure that determines the local acceleration. An informative evidence record therefore preserves position and boundary information, rather than report one covariance amplitude as if it uniquely described the mean-flow forcing everywhere.
CHAPTER 10 OF 32 · Work the periodic example
10. Reconstruct the covariance from an acceleration profile
If the restricted model’s acceleration aR(x) is known, integration gives S(x) = S(x0) − ∫ from x0 to x of aR(s) ds. The covariance is identified only up to an additive constant unless one boundary value or equivalent constraint is supplied.
For a periodic solution, the acceleration must have zero domain integral. That compatibility condition is necessary for S to return to the same value after one period. A proposed periodic acceleration with a nonzero mean cannot be generated solely by a periodic covariance divergence in this model.
This is an inverse-information limit. Measuring the acceleration more precisely does not identify the missing constant flux level. Conversely, a constant offset in S leaves the predicted local acceleration unchanged. State whether the task concerns the gradient-driven response or the absolute momentum-flux value before deciding which additional boundary information is needed.
CHAPTER 11 OF 32 · Work the periodic example
11. Calculate mean-flow kinetic energy separately
Define Emean = one-half ⟨U²⟩x, where the brackets now denote the spatial average over the periodic domain. It is kinetic energy per unit mass in the local model, with units of metres squared per second squared.
For U = −At cos(x/L), with A = S0/L, the average of cos² is one-half. Therefore Emean = A²t²/four. At A = sixty and t = 0.05, the result is 2.25 in those units.
The energy is positive even though the domain-mean velocity and integrated momentum are zero. Oppositely directed flow bands can carry kinetic energy while their momenta cancel. This is why the momentum and energy accounts must both be checked. A zero net momentum input does not imply that generating a structured mean flow requires no energy or that the turbulence-to-mean exchange has vanished.
Multiply ∂U/∂t = −∂S/∂x by U and average over a periodic domain. Integration by parts removes the boundary term and gives dEmean/dt = ⟨S ∂U/∂x⟩x.
In the original solution, this is S0²t/(two L²), equal to A²t/two. At t = 0.05 the transfer rate is ninety metres squared per second cubed. Differentiating the energy expression from the previous chapter returns the same result.
This is the work contribution of the prescribed covariance to the mean-flow energy. It does not create energy from a zero-momentum condition. A self-consistent turbulence model would include the corresponding exchange and the sources and losses of the fluctuating field. Our supplied S does not track its own depletion, so its continued forcing cannot be interpreted as a closed energy-conserving system by itself.
CHAPTER 13 OF 32 · Work the periodic example
13. Recognise the role of the velocity gradient in energy transfer
The energy-transfer integrand is S times the mean shear. A stress profile’s sign alone does not determine the net work unless its relationship to the shear is known. Different relative spatial patterns can yield positive, negative or zero averaged transfer.
In the worked example, S and ∂U/∂x are proportional to the same sine, so their product has a positive average after the flow begins. If the supplied mean shear had the opposite sign, the same covariance could remove mean-flow energy through that term.
This does not contradict the acceleration equation. Momentum redistribution and energy transfer are linked but different summaries. The useful interpretation preserves the full profiles and their correlation. A statement that turbulence drives the mean flow should specify the transfer definition and the relevant averaged balance rather than rely only on a coincident rise in two plotted amplitudes.
CHAPTER 14 OF 32 · Add damping and measurement structure
14. Add linear damping as an explicit new term
Now stipulate ∂U/∂t = aR(x) − rU with positive damping rate r and the same time-independent prescribed covariance. Starting from zero gives U = aR(x)[one − exp(−rt)]/r.
For r = two inverse seconds, the original cosine velocity amplitude at t = 0.05 is thirty times [one − exp(−0.1)], approximately 2.85488 metres per second. It is smaller than the undamped amplitude three. The eventual amplitude under this restricted forcing is thirty metres per second.
The long-time value is a model equilibrium between a prescribed drive and linear damping. It is not a predicted zonal-flow saturation level of a real device. In a plasma, the covariance itself can respond to the growing shear, and several damping or transfer mechanisms may contribute. The added term clarifies one balance while leaving those omitted relationships explicit.
CHAPTER 15 OF 32 · Add damping and measurement structure
15. Check the damped energy ledger
With the damping term, dEmean/dt = ⟨S Ux⟩x − r⟨U²⟩x. If U = −B(t)cos(x/L), the stress contribution is A B/two and the damping contribution is r B²/two.
At the steady amplitude B = A/r, these terms agree, so the mean-flow energy stops changing. For A = sixty and r = two, each rate is nine hundred in metres squared per second cubed. The finite energy is maintained through a balance, not through a disappearance of the energy transfer.
The domain mean remains zero for this symmetric initial state and forcing pattern. That fact does not make local damping irrelevant or turn the model into an isolated total-momentum system. The damping represents an exchange with omitted degrees of freedom. Its physical meaning would need to be specified in a real interpretation instead of treating r as a universal empirical constant for every zonal flow.
CHAPTER 16 OF 32 · Add damping and measurement structure
16. Include viscosity without hiding its spatial operator
Suppose another restricted model includes ν Uxx with positive kinematic viscosity ν. For the single cosine shape, Uxx = −U/L², so viscosity acts like an additional damping rate ν/L² on that particular mode.
With ν = 0.01 square metre per second and L = 0.2 metre, this contribution is 0.25 inverse seconds. Together with r = two, the effective rate for the prescribed shape is 2.25. The steady cosine amplitude would be sixty/2.25, approximately 26.6667 metres per second.
This replacement works because the chosen spatial shape is an eigenfunction of the second derivative under the periodic boundary condition. It is not a reason to replace viscosity by the same scalar rate for every possible profile. Different spatial scales have different curvature and viscous damping. Retain the original operator when moving beyond the single-mode example.
CHAPTER 17 OF 32 · Add damping and measurement structure
17. Relate covariance to fluctuation phase
For a simple local time-series example, stipulate u′ = a cos(ωt) and v′ = b cos(ωt + φ). Averaging over complete periods gives S = ab cosφ/two. Their individual variances are a²/two and b²/two.
Changing φ changes the covariance without changing either separate variance. At φ = zero the covariance is positive; at φ = π/two it is zero; at φ = π it is negative. Separate fluctuation-power summaries therefore do not determine the momentum-flux correlation.
This is a measurement lesson rather than a claim that plasma turbulence consists of one sinusoid. A real covariance depends on the joint signals, sampling, filtering and coordinate definitions. The example shows precisely which information is lost when only separate amplitudes are reported. That missing relationship can matter before one even begins to take the spatial gradient needed for an acceleration estimate.
CHAPTER 18 OF 32 · Add damping and measurement structure
18. Use a realizability bound without treating it as a drive law
Cauchy–Schwarz gives |S| ≤ urms vrms for zero-mean fluctuations with finite second moments. In the original profile, an amplitude twelve is compatible with rms values four and three, since their product is twelve.
The inequality constrains possible covariances. It does not establish the actual value, sign or spatial gradient. Equal rms products can accompany many correlation profiles, including constant stress with zero gradient or alternating stress with local accelerations.
A fitted stress profile that violates the supplied rms bound would be inconsistent with those signal definitions or moments. Passing the bound is only a necessary consistency check, not proof of the model’s causal account. This distinction is useful in simulation audits: a quantity can satisfy a general mathematical constraint while its interpretation still depends on numerical resolution, averaging and omitted physical terms.
CHAPTER 19 OF 32 · Add damping and measurement structure
19. Preserve a finite-volume momentum balance
For a radial interval of width Δx, the cell-average stress contribution is minus [Sright − Sleft]/Δx. This follows by integrating the derivative over the cell rather than guessing an acceleration from its midpoint stress.
For two equal-width cells of one metre with interface covariance values zero, two and zero in squared-velocity units, the cell accelerations are minus two and plus two metres per second squared. Their integrated contributions cancel because the shared interior flux leaves one cell and enters the other.
This discrete account makes conservation visible. The cancellation depends on using the same interior-face flux in both cells. Inconsistent face values can create a spurious net momentum source. A fine-looking local plot does not substitute for the summed balance. Check the boundary fluxes and the integrated change when evaluating a discretised model of gradient-driven flow.
CHAPTER 20 OF 32 · Check evidence and transfer
20. Separate a cell average from an unresolved local peak
A cell whose two boundary stresses are equal has zero cell-average stress-gradient acceleration. It can still contain positive and negative local accelerations that cancel within the interval. The finite-volume result does not resolve every interior feature.
For example, a complete sine period has equal boundary stress values and zero average acceleration, while the original pointwise profile ranges from minus sixty to plus sixty. Increasing covered length and resolving smaller structures are different observation tasks.
A model or diagnostic should therefore state its spatial response. An averaged covariance trace may not preserve the local gradient needed for a peak acceleration estimate. Recovering an unresolved structure requires additional information rather than simply dividing a coarse difference by a smaller number. The conservation result remains valid for its defined average even when it does not identify the detailed interior profile.
CHAPTER 21 OF 32 · Check evidence and transfer
21. Check the noise cost of a small spatial interval
If each of two boundary-stress readings has an independent absolute error bound ε, the finite-difference gradient error is bounded conservatively by two ε/Δx. A smaller separation increases this error bound when the reading errors are held fixed.
Higher spatial resolution can therefore reduce averaging error while increasing sensitivity to measurement uncertainty in a numerical derivative. The best interpretation needs both parts of the observation model. Smoothing may reduce some noise but also change the resolved stress structure and its gradient.
The formula is a deterministic input-error bound, not a statistical confidence level or a universal instrument specification. Shared errors can cancel partly in a difference, while different response functions can introduce other effects. Record the actual error relationship and spatial averaging before attaching precision to a derived acceleration. More printed digits in a stress measurement do not by themselves establish the derivative’s physical accuracy.
CHAPTER 22 OF 32 · Check evidence and transfer
22. Keep density and geometry outside the constant-coefficient shortcut
The simple derivative used constant density and a Cartesian coordinate. If density varies, the relevant momentum-flux divergence can contain density within the derivative. A geometry-weighted surface average can also introduce volume or metric factors.
Consequently a term such as minus one/ρ times ∂(ρS)/∂x is not generally equal to minus ∂S/∂x. Nor can a local radial derivative be promoted to a full toroidal balance without the appropriate geometry and definitions.
This chapter marks the intended boundary of the worked model. Plasma zonal flows can involve Reynolds and magnetic stresses, neoclassical effects, geodesic transfers and other mechanisms depending on regime. The article does not decide which terms dominate a device. It teaches how to read one conservative transport contribution so that a more complete balance can be assembled without confusing covariance, gradient, shear and integrated momentum.
CHAPTER 23 OF 32 · Check evidence and transfer
23. Use the primary plasma study for its interpretive lesson
Alonso and colleagues’ 2012 study of zonal-flow-like structures in the TJ-II stellarator examines possible Reynolds-stress drive and damping mechanisms. Its abstract states that the observed turbulence-driven forces could provide the necessary acceleration, while a firm causal relation was not established, potentially because the stress measurements were local.
That is relevant methodological context for retaining geometry, locality and competing terms. Our periodic stress, damping rates, velocities and energy ledgers are original fictional calculations, not the study’s measurements or a reconstruction of that device.
Read the primary study of zonal-flow-like dynamics and possible drive mechanisms, checked on 25 October 2026. The distinction between a compatible acceleration scale and an established causal balance should remain visible when interpreting another experiment or simulation.
CHAPTER 24 OF 32 · Check evidence and transfer
24. Try an independent changed-scale example
Take S(x) = eight sin(x/0.4) in square metres per square second on one complete spatial period. Begin with zero mean flow and retain only the restricted stress-gradient term for 0.1 second.
Calculate the acceleration amplitude, velocity amplitude, shear amplitude and mean-flow kinetic energy per unit mass. Verify that the domain-integrated momentum change is zero. Then add a damping rate of four inverse seconds and calculate the velocity amplitude at 0.1 second and its eventual steady value.
As an inverse check, ask whether adding a spatially constant value to S changes any of those gradient-driven quantities. Finally state which additional evidence would be required before assigning the same calculation to a real toroidal plasma. Work from the definitions and units rather than reusing the original amplitudes, because both the stress amplitude and length scale have changed.
CHAPTER 25 OF 32 · Check evidence and transfer
25. Check the changed-scale solution
The acceleration amplitude is eight/0.4 = twenty metres per second squared. After 0.1 second without other terms, the velocity amplitude is two metres per second. The shear amplitude is two/0.4 = five inverse seconds, and the spatially averaged kinetic energy is two²/four = one square metre per square second.
The cosine flow still has zero spatial mean, so the integrated momentum change vanishes over its full period. With damping r = four, the velocity amplitude is five times [one − exp(−0.4)], approximately 1.64840 metres per second. The steady amplitude in that prescribed-drive model is five metres per second.
A constant offset in S changes none of these derivative-driven results. A real plasma interpretation still needs the actual averaging, density and magnetic geometry, measurement response, other momentum terms and evidence concerning the stress-flow relationship. The arithmetic does not supply those missing physical inputs.
CHAPTER 26 OF 32 · Challenge the interpretation
26. Test energy transfer when the initial flow opposes the drive
The covariance work depends on the existing gradient. Keep S = S0 sin ξ and write the mean velocity as U = −B(t) cos ξ. The undamped equation gives dB/dt = A, where A = S0/L. Our original flow started with B = 0. Instead let B(0) = −2 m/s, so the initial velocity has the opposite sign from the later driven pattern. With A = 60 m/s², B(t) = −2 + 60t.
The mean energy is B²/4 and its rate is AB/2. At the initial instant the rate is −60 m²/s³. The prescribed covariance initially removes mean-flow energy, even though its spatial profile is identical to the one that later supplies energy. At t = 1/30 s the velocity passes through zero; thereafter its energy grows. This calculation distinguishes the signed work term from a blanket description of Reynolds stress as an energy source. Neither the sign of S at one location nor the fluctuation amplitudes alone determine the domain-averaged work.
A useful learner check is to sketch B as a straight line and B²/4 as a parabola. The minimum of the energy occurs at zero amplitude, while the acceleration remains nonzero. There is no contradiction: an acceleration can slow a moving fluid before reversing it. The same separation between force, velocity and work appears in elementary mechanics.
CHAPTER 27 OF 32 · Challenge the interpretation
27. Compare damping explanations using the same observed velocity
Suppose a stationary velocity profile U = −B cos ξ has been measured, with B positive. In the restricted damped model, stationarity requires −S_x = rU. A compatible covariance is S = rBL sin ξ + C, where C is any spatial constant. Differentiating gives −S_x = −rB cos ξ, which exactly balances rU.
For B = 3 m/s and L = 0.2 m, a damping rate r = 2/s requires a sinusoidal covariance amplitude 1.2 m²/s². A damping rate r = 4/s requires 2.4 m²/s². Both explanations reproduce the same stationary velocity profile. A velocity measurement therefore does not identify the covariance drive and damping separately. If the covariance is independently measured with adequate spatial resolution, its gradient adds a distinct constraint; if the damping is independently established, that also changes the inference.
This example does not declare either damping rate physically correct. It constructs two conditional explanations and specifies the additional evidence that could distinguish them. Adding viscosity introduces another term that must be resolved rather than folded into an unnamed force. In the single cosine example its contribution has the same shape, which makes separating the coefficients from that one stationary shape especially difficult.
CHAPTER 28 OF 32 · Challenge the interpretation
28. Scale the response without claiming universal plasma behaviour
With r positive, choose the velocity scale Ustar = S0/(rL), the coordinate ξ = x/L and dimensionless time τ = rt. For zero initial flow, the local damped solution becomes U/Ustar = −(1 − exp(−τ)) cos ξ. The fraction of the eventual amplitude reached depends only on τ within this restricted equation. At τ = 1 it is approximately 0.632; at τ = 2 it is approximately 0.865.
The scaling separates two questions that are often mixed. S0/(rL) sets the stationary velocity amplitude, while 1/r sets the response time. Doubling S0 changes the amplitude without changing the fractional response time. Doubling r halves that eventual amplitude and also halves the time needed to reach a specified fraction of it. Comparing two traces in seconds without stating their damping rates can conceal this distinction.
These are properties of a prescribed, constant-coefficient forcing model. They do not imply that measured stellarator or tokamak flows collapse onto this curve. In a self-consistent plasma, the covariance, damping and geometry can change as the flow evolves. A failed collapse may indicate a failed assumption rather than an arithmetic error. The responsible use of scaling is to make the assumptions testable, including the quantities held fixed.
CHAPTER 29 OF 32 · Challenge the interpretation
29. Make a boundary experiment close its momentum ledger
A periodic domain was chosen to make the net covariance contribution vanish. Consider instead a finite interval of width D with S at its left boundary equal to 1 m²/s² and at its right boundary equal to 5 m²/s². Integrating the restricted equation gives the mean acceleration −4/D. If D = 2 m, that is −2 m/s². This domain can gain negative mean momentum through the covariance flux imbalance even when there is no separate body-force term in the interior equation.
The result does not contradict the periodic example. The boundary conditions have changed. Calling the interior process a redistribution is insufficient unless the boundary transport is included in the chosen system. Increasing both boundary values by the same constant leaves the imbalance unchanged, while increasing only one boundary value changes it.
Before comparing an experiment with a momentum budget, state the spatial domain, its boundary fluxes, the averaging interval and any terms omitted from the model. A measured local acceleration cannot by itself establish a net momentum input to a larger system. Conversely, a zero integral over a carefully closed domain does not imply that every local acceleration vanishes. The boundary ledger is what connects those statements.
CHAPTER 30 OF 32 · Challenge the interpretation
30. Check the signs with a small displacement
Choose a location where S rises toward the right. A narrow cell there receives a smaller positive covariance flux through its left face than it loses through its right face. Its along-flow momentum therefore decreases, giving a negative acceleration. Drawing those two boundary arrows before differentiating is a useful check on the minus sign. Repeat the argument where S falls: the imbalance reverses. This spatial ledger is independent of whether the covariance itself is positive or negative at the cell centre.
Can a positive covariance give a negative acceleration?
Yes. In our convention the contribution is minus its spatial derivative. A positive covariance that increases with x gives a negative local contribution. Its value and gradient are different quantities.
Can mean-flow energy grow while net momentum stays zero?
Yes. Oppositely directed flow bands can have positive kinetic energy while their momenta cancel. The energy transfer still requires an account of the fluctuating field, sources and losses in a complete model.
Does this equation prove that Reynolds stress causes every zonal flow?
No. It isolates a restricted local transport contribution. A complete plasma balance can contain other stresses, geometry, damping and transfers. Compatible magnitudes and correlations do not automatically establish causation.
Does finer spatial sampling always give a more accurate gradient?
Not automatically. It can resolve structure better while amplifying uncertainty in finite differences. The spatial response, reading errors and their covariance must be considered together.
CHAPTER 32 OF 32 · Challenge the interpretation
32. Continue through the relevant physical routes
Return to Zonal Flow for plasma geometry, E-cross-B motion, shear, diagnostics and confinement limits. Keep its plasma setting distinct from the atmospheric meaning of zonal wind.
Continue through Learning That Transfers into Life and How X Works. A useful explanation names the covariance convention, differentiates it with respect to the correct coordinate, checks the boundary momentum balance and accounts for mean-flow energy separately.