Doubly Degenerate Nutrient Taxis Model

• The doubly degenerate nutrient taxis model is defined by equations where both diffusion and taxis flux vanish when either bacterial density or nutrient concentration is low.
• It elucidates bacterial aggregation and pattern formation through rigorous analysis of global weak and classical solvability across one, two, and three dimensions.
• Methodologies include kinetic derivations, cross-diffusion techniques, and entropy-based compactness arguments to determine critical thresholds and long-term behavior.

The doubly degenerate nutrient taxis model denotes a class of chemotaxis-consumption systems in which the transport of a motile population $u$ depends simultaneously on population density and nutrient or signal concentration $v$, so that diffusion and, in many formulations, the taxis flux itself vanish when either $u$ or $v$ becomes small. Representative equations include the one-dimensional system

$u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$

its multidimensional variants

$$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$

and parabolic-elliptic consumption systems of the form

$$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$

The class arose in the modeling of bacterial aggregation, especially for \emph{Bacillus subtilis} in nutrient-poor environments, and has developed into a distinct branch of degenerate cross-diffusion theory with results on derivation, weak and classical solvability, uniqueness, boundedness, sharp thresholds, and long-time stabilization (Leyva et al., 2013, Plaza, 2017, Tran et al., 5 Aug 2025).

1. Defining structure and principal model families

The defining feature is the simultaneous degeneration of bacterial transport with respect to more than one state variable. In the prototypical cross-diffusion form,

$$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$

the diffusion coefficient $uv$ vanishes if either $u=0$ or $v$0, while the taxis coefficient $v$1 vanishes even more strongly in low-density or low-nutrient regions. This is the sense in which the model is “doubly degenerate” in the literature on bacterial nutrient taxis (Chen et al., 12 Jan 2025, Herrero-Hervás, 24 Feb 2026).

Several closely related formulations appear in current research. One family replaces $v$2 by $v$3 and $v$4 by $v$5, leading to

$v$6

with growth conditions such as $v$7 (Wu, 2024). Another family uses the exponent $v$8 in

$v$9

or its logistic counterpart with $u$0 במקום $u$1 (Zhang et al., 2024, Zhang et al., 2024, Zhang et al., 8 Jan 2026). A further variant places the degeneracy in the signal dependence of the diffusion,

$u$2

with $u$3 and $u$4, so that the $u$5-equation loses ellipticity as nutrient approaches vacuum (Winkler, 2022). The parabolic-elliptic consumption system

$u$6

is not doubly degenerate in the same explicit $u$7-sense, but it belongs to the same nutrient-taxis/consumption line and exhibits a sharp diffusion threshold under radial symmetry (Winkler, 8 Jan 2026).

Across these formulations, $u$8 is interpreted as bacterial or organism density and $u$9 as nutrient or signal concentration. The term $v$0 in the $v$1-equation models consumption, while $v$2 or $v$3 in the $v$4-equation models nutrient-supported proliferation. No-flux boundary conditions are standard on bounded domains; the whole-line Cauchy problem and parabolic-elliptic problems with prescribed nutrient boundary data also occur (Herrero-Hervás, 10 Aug 2025, Winkler, 8 Jan 2026).

2. Biological motivation and kinetic derivation

The biological origin of the class lies in models for bacterial colony formation on thin agar plates. A foundational formulation extends the Kawasaki et al. aggregation model by taking the bacterial diffusivity to be nonlinear and degenerate,

$v$5

and by introducing a nutrient-chemotactic flux

$v$6

which is explicitly constructed to be compatible with the same nonlinear mobility law (Leyva et al., 2013). After nondimensionalization, this yields

$v$7

Here the transport vanishes when either bacteria or nutrient are scarce, reflecting reduced motility in nutrient-poor or low-density regions.

A formal kinetic justification was later given through the parabolic limit of a velocity-jump process of Othmer-Dunbar-Alt type (Plaza, 2017). In that derivation, individual-cell motion is described by a Boltzmann-type transport equation with an isotropic leading turning rate

$v$8

and a Schnitzer-type first-order perturbation

$v$9

Under parabolic scaling and a Hilbert expansion, the effective diffusion tensor becomes proportional to $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$0, and the chemotactic drift becomes proportional to $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$1. The resulting macroscopic bacterial equation has the form

$u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$2

which recovers the doubly degenerate diffusion and nutrient-taxis flux of the phenomenological model (Plaza, 2017).

This derivation is significant because it identifies the degeneracy as a consequence of microscopic movement rules rather than a purely ad hoc constitutive choice. It also explains why diffusion and taxis carry the same nutrient-density prefactors: both emerge from the same turning-rate mechanism.

3. Solvability theory across dimensions and geometries

Existence theory depends strongly on dimension, whether the nutrient equation is parabolic or elliptic, and whether damping terms are present. The current state of the theory covers whole-space, bounded-domain, parabolic-parabolic, and parabolic-elliptic settings.

Setting	Representative feature	Main conclusion
$u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$3, Cauchy problem (Herrero-Hervás, 10 Aug 2025)	doubly degenerate diffusion and taxis	global weak solution under $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$4 and extra entropy/gradient conditions
1D bounded interval (Herrero-Hervás, 24 Feb 2026)	parabolic-elliptic model with external supply $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$5	global weak solution for any nonnegative $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$6
2D bounded convex domain (Zhang et al., 2024)	$u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$7	global bounded weak solution, continuous in $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$8, essentially smooth in $u_t=(uvu_x)_x-(u^2vv_x)_x+uv,\qquad v_t=v_{xx}-uv,$9
2D smooth domain (Zhang et al., 8 Jan 2026)	quadratic logistic source $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$0	global weak solution uniformly bounded in time
3D bounded convex domain (Tran et al., 5 Aug 2025)	bacterial response parameter $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$1	global weak solvability for all $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$2
Ball in arbitrary dimension (Winkler, 8 Jan 2026)	parabolic-elliptic consumption system with radial data	global bounded classical solution for $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$3

In one space dimension, the bounded-interval system

$$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$4

already has a global weak-solution theory, and uniqueness is available under the additional condition $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$5 (Chen et al., 12 Jan 2025). On the whole line, global weak solutions were obtained by regularizing on expanding intervals $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$6, deriving $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$7-independent entropy bounds, and passing to the limit via Aubin-Lions (Herrero-Hervás, 10 Aug 2025). For the parabolic-elliptic problem

$$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$8

global weak solvability holds for nonnegative $$u_t=\nabla\cdot(uv\nabla u)-\chi\nabla\cdot(u^\alpha v\nabla v)+\ell uv,\qquad v_t=\Delta v-uv,$$9 and $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$0, $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$1 (Herrero-Hervás, 24 Feb 2026).

In two dimensions, the model

$$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$2

admits global bounded weak solutions in smooth bounded convex domains for reasonably regular initial data (Zhang et al., 2024). The logistic variant

$$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$3

admits global weak solutions uniformly bounded in time on general smooth planar domains; the paper attributes the strengthening primarily to the quadratic degradation term $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$4 (Zhang et al., 8 Jan 2026). More general two-dimensional large-data results are available for

$$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$5

with weak solvability for $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$6 and classical solvability for $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$7 under explicit parameter ranges (Wu, 2024). For the $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$8 family, global weak solutions exist in 2D for $$u_t=\nabla\cdot(D(u)\nabla u)-\nabla\cdot(u\nabla v),\qquad 0=\Delta v-uv.$$9, and uniform boundedness is established for $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$0 (Zhang et al., 2024).

In three dimensions, a major recent result gives global weak solvability of

$$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$1

for the full range $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$2, after splitting the analysis into weak, moderate, and strong chemotactic regimes (Tran et al., 5 Aug 2025). A separate three-dimensional theory proves global continuous weak solutions and stabilization for $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$3 (De-Ji-Xiang-Mao et al., 18 Jan 2026).

4. Critical exponents, thresholds, and dimension effects

A central theme is the identification of parameter thresholds at which degenerate diffusion ceases to dominate nutrient-driven aggregation. These thresholds are highly model-dependent.

The sharpest such result is the parabolic-elliptic consumption theorem in a ball,

$$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$4

with radial data and

$$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$5

For $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$6, every nonnegative radially symmetric $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$7, $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$8, generates a unique global bounded classical solution in arbitrary dimension $$u_t=\nabla\cdot(uv\nabla u)-\nabla\cdot(u^2v\nabla v)+\text{reaction terms},$$9. Earlier work gives nonexistence of global solutions for some radial data when $uv$0 with $uv$1. The critical exponent

$uv$2

is therefore sharp and dimension-independent for this nutrient-taxis/consumption system (Winkler, 8 Jan 2026).

For explicitly doubly degenerate parabolic-parabolic systems, the thresholds are more geometry- and structure-sensitive. In the family

$uv$3

the one-dimensional theory yields global boundedness for all $uv$4, whereas the two-dimensional theory yields global weak solvability for $uv$5 and uniform boundedness for $uv$6; at $uv$7 in 2D, only local-in-time boundedness is asserted (Zhang et al., 2024). For the generalized large-data system with $uv$8 and $uv$9, the admissible ranges of the taxis growth exponent $u=0$0 depend explicitly on $u=0$1 and the spatial dimension $u=0$2; in particular, for $u=0$3 in two dimensions the admissible large-data range is extended from $u=0$4 to $u=0$5 (Wu, 2024).

The three-dimensional model

$u=0$6

requires a finer subdivision. The analysis is separated into $u=0$7 as weak chemotaxis, $u=0$8 as moderate chemotaxis, and $u=0$9 as strong chemotaxis, with different bootstrap mechanisms in each regime (Tran et al., 5 Aug 2025). This suggests that, unlike the radial parabolic-elliptic consumption problem, no single dimension-independent threshold governs the entire doubly degenerate parabolic-parabolic class.

5. Analytical mechanisms, positivity theory, and uniqueness

The standard construction begins with regularized classical problems. Typical devices are the replacement $v$00, regularization on expanding intervals $v$01, and positivity-preserving approximate systems with homogeneous Neumann data (Herrero-Hervás, 10 Aug 2025, Herrero-Hervás, 24 Feb 2026, Zhang et al., 2024). Because the original equations are not uniformly parabolic, compactness is usually sought not for $v$02 directly but for nonlinear composites such as $v$03 or $v$04 together with time-derivative bounds in dual spaces.

A recurring structural tool is an entropy or quasi-energy coupling the two components. Examples include

$v$05

for the whole-line Cauchy problem (Herrero-Hervás, 10 Aug 2025), the functional

$v$06

in the two-dimensional $v$07 system (Zhang et al., 2024), and weighted logarithmic functionals such as

$v$08

or

$v$09

in the three-dimensional theory (De-Ji-Xiang-Mao et al., 18 Jan 2026). In the $v$10 three-dimensional analysis, the identity

$v$11

encodes the competition between diffusion and taxis in a single dissipation structure (Tran et al., 5 Aug 2025).

Positivity of the nutrient is decisive because it restores coercivity in degenerate bacterial diffusion. For the signal-dependent diffusion model $v$12, a quantitative strong parabolic maximum principle yields a uniform positive lower bound for $v$13 after any positive waiting time, allowing the $v$14-equation to become uniformly parabolic on $v$15 (Winkler, 2022). For the parabolic-elliptic one-dimensional system with external supply, a Harnack-type inequality for the elliptic nutrient equation

$v$16

provides pointwise comparability of $v$17 and $v$18, compensating for the lack of uniform coercivity (Herrero-Hervás, 24 Feb 2026).

Uniqueness has historically lagged behind existence. In one dimension, uniqueness of weak solutions for the bounded-interval parabolic-parabolic system is proved by introducing

$v$19

and estimating the difference $v$20 via a Grönwall argument, using in particular the reciprocal estimate $v$21 when $v$22 (Chen et al., 12 Jan 2025). A more general 2025 draft establishes continuous dependence and uniqueness of bounded weak solutions for triangular degenerate cross-diffusion systems by an $v$23-method based on the Neumann inverse Laplacian, with doubly degenerate nutrient taxis models as a principal application (Chen et al., 24 Aug 2025).

6. Long-time behavior, persistent heterogeneity, and pattern formation

The long-time dynamics of doubly degenerate nutrient taxis systems differ markedly from those of uniformly parabolic chemotaxis models. In the signal-dependent diffusion system

$v$24

solutions stabilize to a limit $v$25 in $v$26 for all $v$27, while $v$28 in $v$29. If $v$30 is nonconstant and $v$31 is sufficiently small in $v$32, then the limiting profile $v$33 is also nonconstant (Winkler, 2022). The paper identifies this explicitly as a qualitative effect due to diffusion degeneracy.

A closely related phenomenon appears in three dimensions for

$v$34

For $v$35, global continuous weak solutions converge to an equilibrium $v$36, and $v$37 is nonconstant when $v$38 and the initial nutrient mass is sufficiently small (De-Ji-Xiang-Mao et al., 18 Jan 2026). In both cases, nutrient depletion eventually freezes a spatially heterogeneous bacterial profile rather than enforcing homogenization. A plausible implication is that degeneracy changes not only well-posedness thresholds but also the asymptotic selection mechanism.

Numerical work motivated much of this analytical program. High-resolution GPU simulations of the chemotaxis-extended Kawasaki model reported that chemotaxis increases the propagation speed of the colony envelope and modulates morphology in a regime-dependent way (Leyva et al., 2013). In semi-solid agar with dense branching morphologies, adding chemotaxis leaves the branching pattern largely intact while accelerating the front. In soft-agar, low-nutrient regimes, increasing chemotactic strength was reported to suppress branching, close fjords, and promote more homogeneous disk-like growth at sufficiently large $v$39 (Leyva et al., 2013). The same paper also derived an effective scalar front equation with

$v$40

and proved that the threshold front speed increases with chemotactic sensitivity.

Taken together, these results place the doubly degenerate nutrient taxis model at the intersection of cross-diffusion theory, chemotaxis-consumption dynamics, and pattern formation. Its distinctive mathematical signature is the collapse of transport in low-density or low-nutrient regions; its distinctive analytical consequence is the need for weighted entropy, positivity, and compactness mechanisms adapted to loss of ellipticity; and its distinctive dynamical effect is that nutrient exhaustion can stabilize nonhomogeneous structures rather than erase them.