Devil's Staircase: Locking and Hierarchies

• Devil's Staircase is a continuous, non-decreasing function characterized by dense plateaus and an exceptional set where abrupt state transitions occur.
• Mathematical formulations reveal DS behavior in non-integer expansions, outer billiards, and matrix-product problems, linking symbolic dynamics with experimental observations.
• In physics, DS organizes discrete states in magnetic, dynamical, and topological systems, offering practical insights into mode locking and emergent order.

A devil’s staircase (DS) denotes a singular staircase-like dependence of an observable on a control parameter, classically exemplified by continuous monotone functions with dense plateaus, and in physics by sequences of distinct ordered, locked, or quantized states selected as temperature, field, interaction strength, or drive is varied. Across current literature, the term covers both canonical commensurate–incommensurate hierarchies and several noncanonical extensions: magnetic and magnetostructural plateaus, frequency-locking staircases in driven systems, topological sector staircases, and singular functions in symbolic dynamics and fractal geometry (Komornik et al., 2015, Matsuda et al., 2014, Isohätälä et al., 2012).

1. Core definition and scope

In the classical mathematical sense, a devil’s staircase is a continuous non-decreasing function that is constant on a dense family of intervals and changes only on a thin exceptional set. That formulation appears explicitly in the outer-billiard literature, where the rotation number as a function of an area parameter is proved to be a devil’s staircase (Yao, 2014). In closely related matrix-product problems, the same structure appears as a continuous monotone non-decreasing $1$-ratio function $\mathfrak r(\alpha)$, with every rational value attained on an interval and irrational values attained only on a Hausdorff-dimension-zero set (Morris et al., 2011).

In physical usage, the term usually refers to a stepwise response with many plateaus, each plateau corresponding to a distinct stable or metastable ordered state. In magnetic systems, this often means a stepwise magnetization curve whose plateaus encode different commensurate spin configurations or stackings (Lopez-Bezanilla, 24 Jul 2025). In layered or modulated solids, it can also mean a temperature-driven succession of magnetostructures rather than a field-driven magnetization staircase, as in CeSb below $T_{\rm N}\sim 17$ K (Kuroda et al., 2020).

The literature now uses several related labels. “Complete” DS denotes a hierarchy in which all rational plateaus occur in the relevant effective description; “incomplete” DS denotes a truncated or partially replaced hierarchy, often because an incommensurate phase intervenes; “hidden” DS denotes an underlying zero-temperature hierarchy of ordered states of which only a finite subset is experimentally visible at finite temperature; and “DS-like” is used when finite size or aperiodicity prevents a strict infinite rational hierarchy while preserving the hallmark stepwise organization (Burnell, 2010, Trzop et al., 2016, Ruzzi et al., 2020, Lopez-Bezanilla, 24 Jul 2025).

Setting	Staircase variable	Plateau meaning
Non-integer expansions	$D(q)=\dim_H U_q$	constant symbolic entropy sector
Outer billiards	rotation number $\tau(a)$	rational rotation interval
Matrix pairs	$1$-ratio $\mathfrak r(\alpha)$	Sturmian slope plateau
Magnetic materials	magnetization or $n_{HS}$	commensurate ordered phase
Driven systems	$\omega_2/\omega$	rational frequency locking

This breadth does not erase the common structure. In all cases, a DS organizes a large family of nearby states by replacing smooth variation with a hierarchy of discrete selections.

2. Mathematical and symbolic formulations

One major modern mathematical realization arises in non-integer base expansions. For alphabet $\{0,1,\dots,M\}$ and base $\mathfrak r(\alpha)$0, the univoque set $\mathfrak r(\alpha)$1 consists of numbers with a unique base-$\mathfrak r(\alpha)$2 expansion, and its Hausdorff dimension satisfies

$\mathfrak r(\alpha)$3

for every $\mathfrak r(\alpha)$4, where $\mathfrak r(\alpha)$5 is the topological entropy of the symbolic univoque set (Komornik et al., 2015). The function $\mathfrak r(\alpha)$6 is continuous and has bounded variation, but its DS behavior is nonclassical: on $\mathfrak r(\alpha)$7, where $\mathfrak r(\alpha)$8 is the Komornik–Loreti constant, one has $\mathfrak r(\alpha)$9 for all $T_{\rm N}\sim 17$0 while $T_{\rm N}\sim 17$1 almost everywhere (Komornik et al., 2015). This is a reversed or dual staircase rather than the standard Cantor–Lebesgue prototype.

A second formulation concerns generalized devil’s staircases as distribution functions of Gibbs measures on self-conformal sets. For a Gibbs measure $T_{\rm N}\sim 17$2 on a self-conformal invariant set $T_{\rm N}\sim 17$3, the staircase is

$T_{\rm N}\sim 17$4

The fine regularity of $T_{\rm N}\sim 17$5 is described through the Hausdorff dimensions of the sets $T_{\rm N}\sim 17$6, $T_{\rm N}\sim 17$7, and $T_{\rm N}\sim 17$8, where the $T_{\rm N}\sim 17$9-Hölder derivative is respectively $D(q)=\dim_H U_q$0, $D(q)=\dim_H U_q$1, or nonexistent in the general sense (Troscheit, 2013). The analysis uses thermodynamic formalism, with pressure equation $D(q)=\dim_H U_q$2, multifractal spectrum $D(q)=\dim_H U_q$3, and an auxiliary function $D(q)=\dim_H U_q$4 defined by $D(q)=\dim_H U_q$5 (Troscheit, 2013). In this setting, DS regularity is controlled by the local-dimension spectrum of the underlying Gibbs measure.

A third formulation appears in dynamical systems. For a convex polygon $D(q)=\dim_H U_q$6, an outer billiard map obtained by cutting off a fixed area from the polygon interior induces a circle homeomorphism on $D(q)=\dim_H U_q$7. If $D(q)=\dim_H U_q$8 is rational, the paper proves that the iterate $D(q)=\dim_H U_q$9 is not the local identity at a periodic point, and this implies that the rotation number $\tau(a)$0 as a function of the area parameter is a devil’s staircase (Yao, 2014).

A fourth formulation comes from the joint spectral radius of matrix pairs. For the one-parameter family $\tau(a)$1, the slope of the Sturmian weakly extremal products defines a continuous monotone non-decreasing function $\tau(a)$2. Every rational $\tau(a)$3 has a plateau $\tau(a)$4 that is a closed interval with nonempty interior, while irrational slopes occur only on a Hausdorff-dimension-zero set and have singleton preimages (Morris et al., 2011). Here the staircase is not a response function of matter, but of extremal symbolic growth.

3. Competing interactions and magnetic staircases

In condensed-matter physics, the canonical DS mechanism is a hierarchy of nearly degenerate commensurate structures selected by competing interactions. SrCo$\tau(a)$5O$\tau(a)$6 provides a direct magnetic example: resonant soft x-ray scattering reveals coexistence of many magnetic periodicities with almost the same energies, including reflections at

$\tau(a)$7

with additional fitted components at $\tau(a)$8 and $\tau(a)$9, and small magnetic fields reshuffle which periodicity dominates (Matsuda et al., 2014). The authors explicitly connect this to ANNNI-like DS physics, while arguing that longer-period structures likely require interactions beyond nearest and next-nearest layers, plausibly of RKKY type (Matsuda et al., 2014). In that material, the staircase is not incidental to transport: field selection among nearly degenerate layered spin configurations underlies its spin-valve-like magnetoresistance (Matsuda et al., 2014).

SrCu$1$0(BO$1$1)$1$2 realizes a different magnetic DS context. Its high-field magnetization curve contains plateaux such as $1$3, $1$4, $1$5, and $1$6, but thermodynamic and ultrasonic measurements show that the staircase does not simply start at the first visible $1$7 plateau above $1$8 T (Imajo et al., 2022). A previously unresolved anomaly near $1$9 T carries large low-temperature entropy and lacks the Schottky-type gapped behavior expected for a simple precursor state, leading to the proposal of a magnetically hidden spin-nematic phase built from $\mathfrak r(\alpha)$0 bound triplet pairs beneath the first visible step (Imajo et al., 2022). This shifts the onset of the staircase from a purely dipolar magnetic crystal to a broader sequence in which a hidden quadrupolar phase precedes crystallization.

CeSb is a canonical devil’s-staircase solid in zero field. Below $\mathfrak r(\alpha)$1 K it passes through AFP1–AFP6 before reaching an AF ground state below $\mathfrak r(\alpha)$2 K, with ordered states built from Ising-like ferromagnetic (001) Ce layers and paramagnetic layers distinguished by the modulation $\mathfrak r(\alpha)$3 (Kuroda et al., 2020). Bulk-sensitive laser ARPES shows that each staircase step is accompanied by a distinct reconstruction of the low-energy electronic structure: folded bands appear, hybridization gaps open, quasiparticle peaks shift or disappear at phase boundaries, and in one momentum sector the coherent band picture near $\mathfrak r(\alpha)$4 collapses in AFP6 and recovers in the AF ground state (Kuroda et al., 2020). CeSb therefore demonstrates that a DS can reorganize not only order parameters but the full spectral function.

The same material also exhibits a step-like change in electron-boson coupling across the staircase. Combining laser ARPES, Raman, and neutron scattering, another study identified a $\mathfrak r(\alpha)$5 meV kink in the Sb $\mathfrak r(\alpha)$6 band and attributed it to coupling with quadrupolar crystal-electric-field excitations $\mathfrak r(\alpha)$7, introducing the “multipole polaron” as a mobile electron dressed by a cloud of quadrupole CEF polarization (Arai et al., 2021). The coupling constant $\mathfrak r(\alpha)$8 changes anomalously through the staircase, reaching $\mathfrak r(\alpha)$9 in AFP6/AFF and $n_{HS}$0 in the AF phase (Arai et al., 2021). A plausible implication is that DS order in correlated solids can couple directly to low-energy quasiparticle dressing rather than merely to static folding.

4. Topological, elastic, and aperiodic extensions

Recent work has generalized DS physics beyond periodic commensurate lock-in. In a one-dimensional XXZ–Bose–Hubbard model with bond spins modulating boson hopping, antiferromagnetic XXZ interactions generate a DS of commensurate Peierls bond-ordered insulators satisfying

$n_{HS}$1

with ordering wavevector

$n_{HS}$2

For negative magnetization, each staircase step is simultaneously a symmetry-protected topological Peierls insulator with inversion protection, fractionalized $n_{HS}$3 bosonic edge states, and quantized local many-body Berry phases (Chanda et al., 2020). Around that commensurate hierarchy lies an incommensurate ordered region of Peierls supersolids, so the staircase organizes both insulating and supersolid sectors (Chanda et al., 2020).

A distinct “topological Devil’s staircase” appears in a constrained kagome Ising antiferromagnet. In the limit $n_{HS}$4 with finite $n_{HS}$5, the low-temperature phase already contains a finite density of zero-energy system-spanning domain walls. On heating, additional system-spanning $n_{HS}$6-lines condense, but noncrossing constraints quantize the number of $n_{HS}$7-lines inserted between consecutive zero-energy $n_{HS}$8-lines. The integer

$n_{HS}$9

therefore labels the sequence $\omega_2/\omega$0, giving an infinite series of thermal first-order transitions (Rufino et al., 9 May 2025). The crucial distinction from ANNNI-type staircases is explicit: inside each phase the wavevector is not fixed to commensurate values, because $\omega_2/\omega$1 and $\omega_2/\omega$2 vary continuously while their ratio remains quantized (Rufino et al., 9 May 2025). Here the staircase variable is a topological sector label rather than a locked modulation vector.

Spin-crossover materials add another family of noncanonical staircases. A microscopic 2D elastic lattice-mismatch model maps exactly to a long-range Ising model with couplings $\omega_2/\omega$3 generated by integrating out harmonic elastic modes; on the square lattice these couplings decay asymptotically as $\omega_2/\omega$4 (Ruzzi et al., 2020). At $\omega_2/\omega$5, sweeping $\omega_2/\omega$6 yields many commensurate HS/LS ordered phases and corresponding steps in $\omega_2/\omega$7, consistent with a hidden DS; at finite temperature, only some of those steps survive as visible plateaus (Ruzzi et al., 2020). In the molecular coordination polymer $\omega_2/\omega$8, thermal spin conversion proceeds through a commensurate spin-state concentration wave at $\omega_2/\omega$9 and an incommensurate one at $\{0,1,\dots,M\}$0, which the authors interpret as an incomplete DS rather than a complete rational staircase (Trzop et al., 2016).

Aperiodicity and finite size also do not preclude DS phenomenology. In quasiperiodic qubit lattices realized on a quantum annealer, nearest-neighbor purely antiferromagnetic couplings on Ammann–Beenker and Pentaplexity graphs produce stepwise magnetization curves and sharp susceptibility peaks as field is varied (Lopez-Bezanilla, 24 Jul 2025). Because the lattices are finite and the plateaus do not align with a simple rational hierarchy of saturation fractions, the authors use “DS-like” rather than a strict DS label; nevertheless, increasing system size produces additional steps, and local commensurability rather than translational periodicity stabilizes the spin sectors (Lopez-Bezanilla, 24 Jul 2025).

5. Nonequilibrium and dynamical staircases

A DS need not encode equilibrium phases. In semiconductor superlattices driven by a monochromatic AC electric field, a Hopf bifurcation of the symmetric periodic state produces a self-generated second frequency $\{0,1,\dots,M\}$1. Depending on parameters, the system is quasiperiodic, frequency-locked with

$\{0,1,\dots,M\}$2

or chaotic; plotting $\{0,1,\dots,M\}$3 versus drive amplitude $\{0,1,\dots,M\}$4 yields flat rational plateaus, Arnol'd tongues in $\{0,1,\dots,M\}$5 space, and a DS of frequency locking (Isohätälä et al., 2012). The same work shows that if either $\{0,1,\dots,M\}$6 or $\{0,1,\dots,M\}$7 is even, the locked orbit breaks the half-period symmetry and generates a spontaneous DC bias, while overlapping tongues produce chaos (Isohätälä et al., 2012). The staircase variable is thus a frequency ratio rather than a density or magnetization.

Josephson junctions under external radiation provide a second dynamical realization. Numerical simulations of the RCSJ model reveal a DS of subharmonic Shapiro steps between $\{0,1,\dots,M\}$8 and $\{0,1,\dots,M\}$9, organized by continued fractions such as

$\mathfrak r(\alpha)$00

and more generally by

$\mathfrak r(\alpha)$01

What is distinctive is the presence of structured chaotic windows between plateaus: Lyapunov exponents and Poincaré sections show regular phase-locked steps interleaved with chaotic intervals that inherit the staircase’s arithmetic ordering and scaling (Shukrinov et al., 2014). The box-counting dimension of the resulting object was reported as

$\mathfrak r(\alpha)$02

close to the value expected for a complete DS, and the onset of chaos near a subharmonic step follows Feigenbaum period doubling (Shukrinov et al., 2014). In this context, the staircase acts as an organizing backbone for synchronized and chaotic dynamics simultaneously.

These nonequilibrium examples show that DS physics is not confined to ground-state selection. The common ingredient is mode locking or state locking on rational manifolds, with chaos or quasiperiodicity appearing between plateaus rather than thermal disorder.

6. Mechanisms, variants, and recurring misconceptions

A common misconception is that a DS requires a periodic lattice with competing long-range interactions and explicit particle-hole symmetry. That is correct for several canonical examples, but modern work has broadened the mechanism set considerably. One-dimensional lattice gases with short-range attraction and long-range repulsion can produce hybrid staircases whose low-filling and high-filling sectors are not related by particle-hole symmetry; in those models the staircase is controlled by a two-body interaction rather than a chemical potential, and the relevant effective particles can be dimers, trimers, or other $\mathfrak r(\alpha)$03-mers (Lan et al., 2017). Closely related Rydberg-dressed models generate an emergent complete DS without microscopic particle-hole symmetry because the low-energy objects are dimer particles below half filling and dimer holes above half filling, both with effective convex interactions (Lan et al., 2015).

Another misconception is that a DS is necessarily exact and complete whenever the phrase is used. Dipolar bosons in a one-dimensional optical lattice do realize a complete DS in the strict classical convex limit $\mathfrak r(\alpha)$04 with sufficiently large onsite repulsion: every rational filling $\mathfrak r(\alpha)$05 is stable on a finite chemical-potential interval, and every chemical potential lies in a gapped commensurate phase (Burnell, 2010). But the same paper shows that finite hopping turns the staircase into an infinite hierarchy of Mott lobes competing with a superfluid, while reducing the onsite interaction creates alternative commensurate states with double occupancies and one-dimensional supersolids (Burnell, 2010). In other words, DS structure can be exact in one limit and only organizing in another.

A further misconception is that the plateau variable must be a commensurate wavevector. The literature now contains counterexamples in which the staircase variable is the high-spin fraction $\mathfrak r(\alpha)$06, a rotation number, a frequency ratio, a topological sector label $\mathfrak r(\alpha)$07, a Hausdorff dimension function $\mathfrak r(\alpha)$08, or a Sturmian slope $\mathfrak r(\alpha)$09 (Ruzzi et al., 2020, Yao, 2014, Isohätälä et al., 2012, Rufino et al., 9 May 2025, Komornik et al., 2015, Morris et al., 2011). This suggests that “Devil’s staircase” is best understood as a structural description of singular state selection, not as a single microscopic mechanism.

The most stable unifying statement is therefore narrow and technical. A DS appears when a control parameter scans a landscape of many nearby candidate states, and the system resolves that competition by producing a hierarchy of discrete plateaus separated by sharp reorganizations or thin exceptional sets. What changes from one domain to another is the object being locked—commensurate modulation, symbolic entropy, rotation number, topological line count, or dynamical frequency ratio—and the sense in which the staircase is complete, hidden, incomplete, or only DS-like.