Well-posed Self-Similarity in Incompressible Standard Flows

Rescaling of the vorticity depends on the derivatives

rescaling of the energy is given by

and rescaling of the velocity is given by

All three exponents will be positive, if \beta_x/\beta_t>3/2. For numercial examples, see the Appendix.

Any non-self-similar function can be written as f(x,y,z) = C_4f(x/C_1,y/C_2,z/C_3), where the fixed constants C_1=C_2=C_3=C_4=1. For arbitrary values of the fixed constants one can always write C_4f(x/C_1,y/C_2,z/C_3) = p_4 f(x/p_1,y/p_2,z/p_3), where the scaling parameters p_1=C_1,p_2=C_2,p_3 = C_3, p_4 = C_4. Rescaling of the parameters yileds p_i^{\prime} = k^a p_i, k>0, and since we seek p_i^{\prime} = C_i, it follows that k=1, or a = 0, which is the trivial, or identity scaling. Thus, any non-self-similar function f is contained into a self-similar function g having the same functional form as f, where f is contained into g at the identity scale.

• Embedding of the initial data. By Lemma 1.2 and Corollary 1.3, the given smooth, divergence-free initial field u_{0}(x,y,z) is realized as the k=1 member of a one-parameter self-similar family

  u_{\mathrm{SS}}(x,y,z,t) \;=\; T^{(\beta_x-\beta_t)/\beta_t}\, F\!\bigl(x/L_1,y/L_2, z/L_3,t/T\bigr).

  This provides a candidate solution profile matching u_{0} exactly at t=0.

• Existence and uniqueness. Invoking the Kato–Fujita theory in the periodic setting (or Leray–Hopf theory in the Schwartz-class setting) [Cla06], [Wik25] - [Ber02], one obtains a unique, smooth NSE solution (u,p) on an interval [0,\tau) with \;u(x,y,z,0)=u_{0}(x,y,z). By uniqueness, this solution must coincide with the self-similar field u_{\mathrm{SS}}(x,y,z,t) for all t\in[0,\tau).

• A priori norm estimates. Lemma 1.1 provides exact scaling exponents for the velocity, energy and vorticity norms under the standard flow scaling (\alpha_x=1,\alpha_t=2). It shows that for \beta_x/\beta_t>3/2, none of these norms grows with the scaling parameter, yielding uniform bounds for all t.

• Blow-up exclusion. The classical Beale–Kato–Majda criterion [BKM] is applied: since

  \displaystyle\int_{0}^{T}\|\omega(\cdot,\tau)\|\,d\tau<\infty

  by the uniform bound on \|\omega\|, no finite-time singularity can occur. This establishes smoothness of u_{\mathrm{SS}} on [0,\infty).

• Conclusion of Theorem 3.1 Together, these steps show that the k=1 self-similar member is the unique, globally smooth, space-periodic solution of the incompressible Navier–Stokes equations evolving from u_{0}, with all relevant norms remaining bounded for all t\ge0.

All self-similar solutions (?) exist as mathematically valid solutions of the NSE. This includes the standard flow cases \alpha_x=1, \alpha_t=2, where the viscosity is the same at all scales: such solutions are given in (?) and (?). Here, we study only (?), which can be summarized as

where the isobaric weight of \vec{u} is equal to \beta_x-\beta_t and comes from the multiplying parameter while \mathbf{F} has zero isobarity. Here L_1,L_2, L_3, T, U are nonzero parameters that scale the same way as x,y,z,t,u accordingly. NSE solutions of this form are not only known to exist (some examples given above), but are very common in physics; and one can then conclude that the NSE solutions (Proof) always exist.

We require the self-similar solutions (Proof) to have non-standard isobarity \beta_x-\beta_t that does not match the standard flow \alpha_x=1, \alpha_t=2; we also require \beta_x/\beta_t>3/2, as per Lemma 1.1.

The space-periodic solutions

are subset of (Proof); however, space-periodic NSE solutions of the form (Proof) are known to exist (e.g. the solutions of the Taylor-Green vortex above), and one can then conclude that the NSE solutions (Proof) always exist. Note, that we have chosen (Proof) to be periodic in space only, and not in time.

At t=0

and all space-periodic self-similar solutions with isobaric weight equal to \beta_x-\beta_t (Proof) reduce to space-periodic self-similar functions of x,y,z and isobaric weight of \beta_x-\beta_t, where \boldsymbol{\psi} need not be steady-state solutions of the NSE.

Suppose we are given a smooth, real-valued, periodic and divergence-free vector field \vec{u}_0, defined on a rectangular domain [0, L_1] \times [0, L_2] \times [0, L_3]. It can always be expanded in a real Fourier series as:

where:

• U is a constant,

• \mathbf{C}_{n_1, n_2, n_3}, \mathbf{S}_{n_1, n_2, n_3} \in \mathbb{R}^3 are real-valued vector coefficients,

• The sum is taken over all non-negative integers n_1, n_2, n_3 \in \mathbb{Z}_{+},

• The terms \frac{2\pi n_1}{L_1}, \frac{2\pi n_2}{L_2}, and \frac{2\pi n_3}{L_3} are the spatial frequencies in the x, y, and z directions, respectively.

Now we use Lemma 1.2- we construct a self-similar function \boldsymbol{\psi} with mathematical form identical to \vec{u}_0, with the exception that L_1, L_2, L_3, U are no longer constants, but parameters that rescale as x,y,z,u accordingly:

It is seen, that as long as \vec{u}_0 is divergence-free, so is U\boldsymbol{\psi}. It is also seen, that \vec{u}_0 \in\boldsymbol{\psi}, that is \vec{u}_0 is embedded in the self-similar function \boldsymbol{\psi} at the identity scale, so that

The original initial condition \vec{u}_0 is reproduced at scale k=1 by rescaling the parameter U. Notice that U\boldsymbol{\psi} is a self-similar space-periodic function, or a family, a set of functions corresponding to various rescalings of U; while \vec{u}_0 is only one space-periodic function of the aforementioned family, achieved at scale k=1.

It is well known that, given a smooth, space-periodic, divergence free, self-similar initial condition U\boldsymbol{\psi}, one can always construct an unique, infinitely differentiable, space-periodic, self-similar NSE solution

such that

which exists in the time interval [0,\tau) [Cla06], [Wik25] - [Ber02]. We see that, the space-periodic solution U\boldsymbol{\varphi} of Eq.(Proof), being a self-similar solution with isobarity equal to \beta_x-\beta_t, belongs to the set of self-similar solutions of Eq. (Proof).

Analogously, it is well known that, given a smooth, space-periodic, divergence free, non-self-similar initial condition \vec{u}_0, one can always construct an unique, infinitely differentiable, space-periodic, non-self-similar NSE solution

such that

which exists in the time interval [0,\tau) [Cla06], [Wik25] - [Ber02]. In this case, L_1,L_2,L_3,T,U are fixed, non-rescaleable constants. Analogously with (Proof), it follows that U\vec{u}^* \in U\boldsymbol{\varphi}, that is, U\vec{u}^* is embedded in the self-similar solution U\boldsymbol{\varphi} so that

The original solution U\vec{u}^* corresponds to the original initial condition \vec{u}_0, such that U\vec{u}^*(\cdot,0)=\vec{u}_0 and is reproduced at scale k=1 by rescaling of the parameter U of the self-similar function U\boldsymbol{\varphi}. Notice that U\boldsymbol{\varphi} is a self-similar solution, or a family, a set of solutions corresponding to different rescalings of U; while U\vec{u}^* is only one solution of the aforementioned family, achieved at scale k=1.

The self-similar solutions (Proof) have isobaric weight \beta_x-\beta_t with \beta_x/\beta_t>3/2; but rescaled under standard conditions \alpha_x=1, \alpha_t=2 remain smooth for all time according to the Beale-Kato-Majda criterion [BKM], as shown in Lemma 1.1. The self-similar solutions (Proof) do not exhibit scaling-induced blow-up of the vorticity, the energy or the velocity. It follows, that the energy of such solutions will remain bound for all time.

Because U\vec{u}^* (corresponding to the original initial condition \vec{u}_0) is embedded in the self-similar solution U\boldsymbol{\varphi} at scale k = 1, and both are solutions to the NSE with the same initial data, uniqueness implies that U\vec{u}^* = U\boldsymbol{\varphi}|_{k=1} for all time. Since U\boldsymbol{\varphi} is globally smooth, it follows that U\vec{u}^* remains smooth throughout \mathbb{R}^3 \times [0, \infty).

Regarding the physics of the self-similar solutions (Proof): smoothness, or infinite differentiability is easily shown to be a property of the self-similar solutions (Proof). Except for the multiplying constant, they are isobaric polynomials of zero weight, or can be Taylor expansions of closed-form functions which depend on zero-weight arguments, e.g. x/L, t/T; or are the ratio of such functions. Since the energy of the system comes solely from the smooth, space-periodic initial velocity \vec{u}_0 (and pressure), the initial energy is bound; (Proof) cannot grow at infinity. This is always possible in a zero-weight function (note that the isobarity comes solely through U), since decay can always be ensured through the presence of exponential functions, e.g. \mbox{exp}[-(x^2+y^2+z^2)/L^2]. Smoothness and decay in t is also required, since the system energy is conserved. This also is always possible in a zero-weight function, since decay can always be ensured through the presence of exponential functions, e.g. \mbox{exp}[-t/T].

We can therefore conclude, that given any arbitrary, smooth, divergence free, space-periodic function (Proof) in the initial moment t=0, the incompressible NSE always has a self-similar solution with isobarity of \beta_x-\beta_t with \beta_x/\beta_t>3/2 (Proof), which is infinitely differentiable, being a constant U multiplying a zero-weight isobaric function; non-increasing in space due to the requirement of bound initial energy at t=0 and non-increasing in time within the interval [0,\tau) due to the energy conservation requirement. The latter guarantees that system energy will be bound in [0,\tau). To maintain these properties for all time under infinite rescaling while we are modeling the fluid as a continuum, rescaling must be performed according to standard flow conditions \alpha_x=1, \alpha_t = 2. For all time, the original solution U\vec{u}^* corresponding to the original initial condition \vec{u}_0, will remain embedded into the self-similar solution U\boldsymbol{\varphi} at scale k=1; this embedding prevents any rescalings of U\vec{u}^*, while the rescalings of the self-similar solution U\boldsymbol{\varphi} always keep the vorticity norm bound by virtue of the mismatch between the isobaric weight \beta_x-\beta_t of the self-similar solution and the standard flow environment \alpha_x=1, \alpha_t = 2, as shown in Lemma 1.1.

In this section, we follow closely the proof given in the previous section. Here, we study only Bouton’s self-similar solutions with isobaric weight \beta_x-\beta_t (Proof) under standard flow conditions \alpha_x=1, \alpha_t=2. They are continuous at t=0; their isobaric weight comes from the multiplying parameter while \mathbf{F} has zero isobarity. L_1, L_2, L_3, T, U are nonzero parameters that scale the same way as x,y,z,t,u accordingly.

The solutions

are a subset of (Proof); here we have a self-similar function \mathbf{S}, which has zero isobaric weight by virtue of the scaling parameters L_1,L_2, L_3, T.

At t=0

and all such self-similar solutions with isobaric weight equal to \beta_x-\beta_t (Proof) reduce to self-similar functions of x,y,z and isobaric weight of \beta_x-\beta_t, where \mathbf{s} need not be steady-state solutions of the NSE.

Schwartz-class functions are generally non-self-similar and have Gaussian peak form exp(-r), a polynomial multiplying a Gaussian P(x,y,z)exp(-r), a bump function form or a convolution of the above functions. Suppose we are given a smooth, real-valued and divergence-free vector field \vec{u}_0 of Schwartz-class, defined with the fixed constant scale factors L_1, L_2, L_3

where U is a fixed constant.

Now we use Lemma 1.2- we construct a self-similar function \mathbf{s} with mathematical form identical to \vec{u}_0, with the exception that L_1, L_2, L_3, U are no longer constants, but parameters that rescale as x,y,z,u accordingly:

It is seen, that as long as \vec{u}_0 is divergence-free, so is U\mathbf{s}. It is also seen, that \vec{u}_0 \in\mathbf{s}, that is \vec{u}_0 is embedded in the self-similar function \mathbf{s}, so that

The original initial condition \vec{u}_0 is reproduced at scale k=1 by rescaling the parameter U. Notice that U\mathbf{s} is a self-similar function, or a family, a set of functions corresponding to different rescalings of U; while \vec{u}_0 is only one Schwartz-class function of the aforementioned family, achieved at scale k=1.

It is well known that, given a smooth, divergence free, self-similar initial condition U\mathbf{s}, one can always construct an unique, infinitely differentiable, self-similar NSE solution

such that

which exists in the time interval [0,\tau) [Cla06], [Wik25] - [Ber02]. We see that, the solution U\mathbf{S} of Eq.(Proof), being a self-similar solution with isobarity equal to \beta_x-\beta_t, belongs to the set of self-similar solutions Eq. (Proof).

Analogously, it is well known that, given a smooth, Schwartz-class, divergence free, non-self-similar initial condition \vec{u}_0, one can always construct an unique, infinitely differentiable, Schwartz-class, non-self-similar NSE solution

such that

which exists in the time interval [0,\tau) [Cla06], [Wik25] - [Ber02]. In this case, L_1,L_2,L_3,T,U are fixed, non-rescaleable constants. Analogously with (Proof), it follows that U\vec{u}^* \in U\mathbf{S}, that is, U\vec{u}^* is embedded in the self-similar solution U\mathbf{S} so that

The original solution U\vec{u}^* corresponds to the original initial condition \vec{u}_0, such that U\vec{u}^*(\cdot,0)=\vec{u}_0 and is reproduced at scale k=1 by rescaling of the parameter U of the self-similar function U\mathbf{S}. Notice that U\mathbf{S} is a self-similar solution, or a family, a set of solutions corresponding to different rescalings of U; while U\vec{u}^* is only one solution of the aforementioned family, achieved at scale k=1.

Analogously with the reasoning in Theorem A.1, we limit our study only to self-similar solutions with isobarity \beta_x-\beta_t with \beta_x/\beta_t>3/2 and we argue that by virtue of Lemma 1.1, the self-similar solution remains smooth for all time when rescaled under standard conditions \alpha_x=1, \alpha_t=2 according to the Beale-Kato-Majda criterion [BKM], since scaling-indiced blowup is impossible in this case.

Because U\vec{u}^* (corresponding to the original Schwartz-class initial condition \vec{u}_0) is embedded in the self-similar solution U\mathbf{S} at scale k=1, it will possess the characteristics of U\mathbf{S} and will thus remain smooth throughout ({\rm I\!R}^3 \times [0,\infty)), because U\mathbf{S} remains smooth.

We can be confident in the smoothness of U\mathbf{S}, as well as in its bound energy for all time through arguments, analogous to those given at the end of the proof of Theorem A.1.

Let u_{0}(x,y,z) be any smooth, divergence‐free vector field on \mathbb{R}^{3} which also lies in L^{2}(\mathbb{R}^{3}) (hence in H^{s}(\mathbb{R}^{3}) for all s>0). By Lemma 1.2 and Corollary 1.3, u_{0} embeds as the k=1 member of a one‐parameter self‐similar family

which formally satisfies the incompressible NSE for all t\ge0. Classical existence and uniqueness theorems (e.g. Leray–Hopf for L^2–data [Tem77, Maj02]) then guarantee that there is a unique, smooth solution (u,p) evolving from u_{0} on an interval [0,\tau). Lemma 1.1 provides uniform bounds on the vorticity and energy norms under the standard flow scaling (\alpha_x=1,\alpha_t=2), and the Beale–Kato–Majda criterion [BKM] ensures no finite‐time singularity can occur.