MT-breaker: Disambiguation & Applications

• MT-breaker is a nonstandard term with context-specific meanings in combinatorial games, power-system security, and nanoelectronic reliability.
• In Maker–Breaker graph games, it encompasses outcome problems and quantitative invariants, highlighting strategic play in domination variants.
• In power-grid attacks and MTJ reliability, MT-breaker denotes breaker status manipulations and dielectric breakdown, underscoring challenges in system security and spin transport.

MT-breaker is a nonstandard, context-dependent label rather than a stable technical term. In current arXiv usage, it denotes markedly different objects across distinct research literatures: a family of Maker–Breaker graph games, especially domination and total-domination variants; breaker-centered hidden topology attacks on power grids; and, in nanoelectronic reliability, the breakdown of a magnetic tunnel junction (MTJ) tunnel barrier under time-dependent dielectric breakdown (TDDB). Its meaning is therefore determined by the surrounding formalism rather than by a field-independent definition (Duchêne et al., 2018, Deka et al., 2015, Reza et al., 2016).

1. Terminological status and principal senses

The term is best understood as a disambiguation problem. In combinatorial game theory, “breaker” is the adversarial player in a Maker–Breaker positional game. In power-systems security, “breaker” refers literally to transmission-line breaker statuses manipulated in a hidden topology attack. In MTJ reliability modeling, the expression refers to an MTJ whose dielectric barrier has “broken” electrically under TDDB, and one paper explicitly states that it is not a separate device class or a standard acronym (Reza et al., 2016).

Context	Meaning	Representative source
Combinatorial game theory	Maker–Breaker game framework; in one paper the relevant object is the Maker–Breaker domination game	(Duchêne et al., 2018)
Power-systems security	Breaker-jammer hidden topology attack using breaker-status changes and flow-measurement jamming	(Deka et al., 2015)
Spintronics / reliability	MTJ barrier breakdown under TDDB; not a separate device class or standard acronym	(Reza et al., 2016)

A plausible implication is that lexical similarity alone is not sufficient for retrieval or citation. Nearby terms can be unrelated: for example, “MultiBreak” denotes a multi-turn jailbreak benchmark for LLM safety rather than a breaker-centered graph game or device-failure model (Song et al., 3 May 2026).

2. MT-breaker in Maker–Breaker domination and total-domination games

In graph-theoretic usage, the most direct interpretation is the Maker–Breaker domination game. The game is played on a graph $G=(V,E)$, with a position modeled as $G=(V,E,c)$, where $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$. Dominator and Staller alternately choose unplayed vertices. Dominator wins when the set of vertices he has chosen becomes a dominating set; Staller wins when she occupies the closed neighborhood $N[u]$ of some vertex $u$, thereby isolating that vertex from Dominator (Duchêne et al., 2018). The outcome classes are $D$, $S$, $N$, and $P$, but the paper proves that $P$ never occurs: if a player can win as second player, that player can also win as first player by an imagination strategy (Duchêne et al., 2018). The same paper establishes that deciding the outcome is PSPACE-complete on bipartite graphs and on split graphs, while admitting polynomial-time algorithms on cographs and trees; its central structural certificate is the pairing dominating set, a family of pairs $G=(V,E,c)$0 such that

$G=(V,E,c)$1

If such a set exists, then Dominator has a winning strategy (Duchêne et al., 2018).

The Maker–Breaker total domination game is the total-domination analogue of this framework. Dominator now wins only when his chosen vertices form a total dominating set, so every vertex must have a neighbor in Dominator’s set; Staller wins by selecting all vertices in the open neighborhood $G=(V,E,c)$2 of some vertex $G=(V,E,c)$3 (Gledel et al., 2019). This variant is strictly harder for Dominator. The paper proves that no minimum-degree condition is sufficient to guarantee that Dominator wins when Staller starts, and that there are infinitely many connected cubic graphs in which Staller wins (Gledel et al., 2019). It also classifies several graph classes: for grids,

$G=(V,E,c)$4

is $G=(V,E,c)$5 if both $G=(V,E,c)$6 and $G=(V,E,c)$7 are even and $G=(V,E,c)$8 otherwise; for cacti, the outcome is determined by whether the graph decomposes into induced $G=(V,E,c)$9-blocks or reduces to an $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$0-star cactus under iterative deletion of end-blocks $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$1 (Gledel et al., 2019). The same paper places the game in a hypergraph setting by taking hyperedges to be open neighborhoods $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$2, so that the MBTD game becomes a Maker–Breaker game on a neighborhood hypergraph (Gledel et al., 2019).

A later extension introduces quantitative game-length invariants. The Maker–Breaker total domination number $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$3 is the minimum number of Dominator moves needed to win when Dominator starts, and $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$4 is defined analogously when Staller starts. The paper proves the sharp bounds

$$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$5

together with

$$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$6

and strong realizability results: for any $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$7, there exist connected graphs realizing $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$8, $$c:V\to \{\text{Dominator},\text{Staller},\text{Unplayed}\}$$9, and $N[u]$0 (Divakaran et al., 23 Jul 2025). This shows that, within the Maker–Breaker interpretation, “MT-breaker” may refer either to an outcome problem or to a quantitative invariant of optimal play.

3. Threshold phenomena in Maker–Breaker graph games

A broader combinatorial interpretation places Breaker in probabilistic and extremal graph games. In the Maker–Breaker $N[u]$1-game on the random graph $N[u]$2, Maker and Breaker alternately claim unclaimed edges of $N[u]$3, and Maker wins if his claimed edges contain all edges of some copy of a fixed graph $N[u]$4. The threshold function $N[u]$5 is defined by

$N[u]$6

The key density parameter is

$N[u]$7

If $N[u]$8 contains a subgraph $N[u]$9 such that $u$0, $u$1 is strictly 2-balanced, and $u$2 is not a tree or a triangle, then there exist constants $u$3 such that the threshold is of order $u$4: $u$5 This matches the corresponding random Ramsey threshold. Triangles are exceptional: for $u$6, although $u$7, the Maker–Breaker triangle game has threshold $u$8, not $u$9 (Nenadov et al., 2014).

Random-player Maker–Breaker games shift the asymmetry from the board to the players. In the $D$0 random-Breaker game on $D$1, Breaker chooses $D$2 unclaimed edges uniformly at random while Maker plays strategically. For the Hamilton cycle game, the perfect-matching game, and the $D$3-vertex-connectivity game, the critical bias satisfies

$D$4

and the paper gives explicit winning strategies for Maker in the corresponding range $D$5 (Krivelevich et al., 2015). In the dual $D$6 random-Maker model, Maker is random and Breaker is strategic. Here the critical scale is logarithmic: $D$7 Breaker can isolate a vertex when $D$8, while Maker can typically force Hamiltonicity and $D$9-connectivity once $S$0 for a suitable constant $S$1 (Krivelevich et al., 2015). The contrast shows that “Breaker” may encode either a sparsification mechanism or a local isolation strategy, depending on which side plays randomly.

On complete graphs with no randomization, exact small-board thresholds are also known. In the unbiased Hamiltonicity game on $S$2, Maker wins if and only if $S$3, independent of who starts. For the Hamiltonian Path game, Maker wins if and only if $S$4. For the Fixed Hamiltonian Path game, Maker wins if and only if $S$5 when Maker starts and if and only if $S$6 when Breaker starts (Stojaković et al., 2017). These results were obtained by a computer-assisted minimax search with canonical labeling and isomorphism pruning, and they also yielded the extremal bound

$S$7

for the minimum number of edges in an $S$8-vertex graph on which Maker can still win Hamiltonicity (Stojaković et al., 2017).

4. Breaker as separator on random trees, lattices, and random boards

A distinct but related line of work studies Breaker as a separator rather than as an occupation-based blocker. On a Galton–Watson tree $S$9 with root $N$0, Breaker deletes one available edge per round, thereby removing the child endpoint and all descendants, while Maker fixates one available edge and secures it permanently. Breaker wins when the root lies in a finite component; Maker wins by sustaining an infinite path from the root (Vilkas, 2024). If $N$1 is the offspring distribution with probability generating function $N$2, then the winning probability of Breaker depends sharply on the information regime. With full information, the probabilities $N$3 and $N$4 that Breaker wins when Breaker or Maker starts satisfy

$N$5

With no extra information and no leaves, the game reduces to a left-continuous random walk, and $N$6 satisfies

$N$7

while in the subtree-size-information regime the fixed-point equation becomes

$N$8

where $N$9 is the extinction probability of the original Galton–Watson process (Vilkas, 2024). The paper emphasizes that more information tends to favor Breaker and that the phase transition may be discontinuous in the full-information regime.

On planar lattices, Breaker often appears as a dual obstruction. In the $P$0-Maker–Breaker crossing game on the triangular grid graph $P$1, Maker tries to claim a left-to-right path, while Breaker tries to prevent it. The paper proves that Maker has a winning strategy for sufficiently tall grids when $P$2, specifically when $P$3, and that Breaker has a winning strategy for sufficiently wide grids when $P$4, specifically when $P$5 (Wallwork, 2022). The proof uses “secure” components, blue paths, and brackets in the primal and dual lattices; the non-self-duality of the triangular grid forces the Breaker threshold to be expressed through the dual hexagonal geometry rather than by a symmetry argument (Wallwork, 2022).

On random boards, the Maker–Breaker percolation game on $P$6 sharpens the separator viewpoint. Each edge of $P$7 is kept independently with probability $P$8, Maker chooses a protected vertex $P$9, and then Maker and Breaker alternately claim $P$0 and $P$1 unclaimed edges. Breaker wins if the component of $P$2 becomes finite (Dvořák et al., 2024). For the $P$3 game, Breaker almost surely has a winning strategy for every $P$4, answering a question posed for the full lattice $P$5, where Maker wins at $P$6 (Dvořák et al., 2024). For the $P$7 game, the threshold is nontrivial: $P$8 (Dvořák et al., 2024). The paper further introduces the threshold

$P$9

and derives general sufficient conditions for Breaker in terms of the connective constant of a vertex-transitive graph (Dvořák et al., 2024).

5. Breaker-jammer attacks in power-grid state estimation

In power-systems security, MT-breaker refers to a hidden topology attack centered on circuit breakers rather than on a game-theoretic adversary. The grid is modeled as a graph $G=(V,E,c)$00, and breaker statuses are encoded by a diagonal matrix $G=(V,E,c)$01, initially $G=(V,E,c)$02 for all-operational lines. The estimator uses the DC model with state vector $G=(V,E,c)$03 and measurement vector

$G=(V,E,c)$04

where $G=(V,E,c)$05 are line-flow measurements and $G=(V,E,c)$06 are bus injection measurements (Deka et al., 2015). The adversary is agnostic: it does not know the true system state $G=(V,E,c)$07 or the line susceptance matrix $G=(V,E,c)$08, cannot alter meter values arbitrarily, but can change breaker statuses on selected transmission lines and jam some flow measurements so that those measurements do not reach the control center (Deka et al., 2015).

The postattack estimated state is written as $G=(V,E,c)$09. For the unjammed flow measurements, consistency requires

$G=(V,E,c)$10

and if a breaker status is changed on a line carrying a flow meter, that flow measurement must also be jammed: $G=(V,E,c)$11 For injection meters, the paper derives

$G=(V,E,c)$12

which, using the flow constraint and the breaker-jamming condition, is rewritten as

$G=(V,E,c)$13

Uniqueness of the postattack state estimate requires

$G=(V,E,c)$14

(Deka et al., 2015).

The paper’s main analytical device is a graph-coloring formulation. Buses with equal values of $G=(V,E,c)$15 are assigned the same color. If a flow measurement on line $G=(V,E,c)$16 is not jammed, then the flow constraint forces $G=(V,E,c)$17, so only jammed-flow lines can connect different colors. Collapsing each color class yields a reduced graph $G=(V,E,c)$18, and boundary injection constraints take the reduced Laplacian form

$G=(V,E,c)$19

The key structural theorem states that following a breaker-jammer attack, the number of injection measurements at the boundary buses should be one less than the number of distinct colors in the grid buses; under the stated normal-operating assumptions, this condition is both necessary and sufficient for feasibility (Deka et al., 2015).

The strongest result concerns sparsity of the physical action. If a feasible attack can be designed with $G=(V,E,c)$20 breaker status changes, then there exists a feasible attack in which all but one of those breakers are restored to status $G=(V,E,c)$21, while keeping the corresponding flow measurements jammed. Hence an optimal breaker-jammer attack needs to modify only one transmission-line breaker (Deka et al., 2015). The attack design problem first minimizes the number of breaker-status changes and then, among those, the number of jammed flow measurements. After the one-breaker reduction, it is simplified to

$G=(V,E,c)$22

subject to

$G=(V,E,c)$23

The paper notes that an $G=(V,E,c)$24-to-$G=(V,E,c)$25 relaxation can approximate this combinatorial optimization (Deka et al., 2015). Simulations on IEEE 14-, 30-, and 57-bus systems, with flow measurements on all lines and injection measurements on a random fraction of buses, show that the optimal attack is often achievable with a single breaker change and that the number of flow measurements that must be jammed generally increases as the number of injection measurements increases (Deka et al., 2015).

6. MTJ barrier breakdown, Büttiker probes, and spin-transport degradation

In nanoelectronics and spintronics, MT-breaker refers not to a game or a power-grid attack but to the electrical breakdown of the dielectric barrier in an MTJ. One paper states explicitly that the term is not a separate device class or a standard acronym; in context it refers to the breakdown of the magnetic tunnel junction’s MgO tunnel barrier, with traps and possibly a percolation path forming under TDDB stress (Reza et al., 2016). The motivation is shared across MOS devices and MTJs: ultrathin dielectrics accumulate traps under high fields, producing stress-induced leakage current (SILC), soft breakdown (SBD), and hard breakdown (HBD) (Reza et al., 2016).

The transport model embeds traps as virtual Büttiker probes within the non-equilibrium Green’s function (NEGF) formalism. The retarded Green’s function without probes is

$G=(V,E,c)$26

and the transmission from left to right is

$G=(V,E,c)$27

A trap is represented by a Büttiker self-energy, and the paper emphasizes that this trap self-energy is spin independent: unlike ferromagnetic contacts, it does not include exchange splitting $G=(V,E,c)$28, because traps randomize spin (Reza et al., 2016). Multiple trap-assisted routes are treated as parallel conduction channels. For example, if traps are inserted at positions 2 and 5, the total transmission is written as

$G=(V,E,c)$29

with the sequential pathway decomposed through series-like transmission rules such as

$G=(V,E,c)$30

Hard breakdown is modeled by inserting the probe self-energies directly into the device Green’s function,

$G=(V,E,c)$31

reflecting the emergence of a strongly conducting percolation path (Reza et al., 2016).

The model is applied to SILC and post-breakdown current in conventional dielectrics and to MTJ-specific spin transport. The paper states that an interface trap density around $G=(V,E,c)$32 yields simulated SILC that matches experiment closely in an HfSiON/SiO$G=(V,E,c)$33 gate stack (Reza et al., 2016). In MTJs, the central physical claim is that trap-assisted tunneling degrades spin filtering because traps are spin independent and captured electrons emerge with effectively randomized spin. The spin-current expression is written as

$G=(V,E,c)$34

and the tunnel magnetoresistance is

$G=(V,E,c)$35

The paper reports that $G=(V,E,c)$36 decreases faster than $G=(V,E,c)$37, so TMR degrades significantly after soft breakdown; one example states that at around $G=(V,E,c)$38 and $G=(V,E,c)$39, TMR can drop by about $G=(V,E,c)$40 after one soft breakdown (Reza et al., 2016). In this literature, MT-breaker therefore denotes a reliability and transport-failure regime rather than an adversarial action.

The disparate meanings of MT-breaker share only the surface vocabulary of “breaking.” In combinatorial game theory, Breaker blocks Maker’s target structure; in power systems, breakers are switched to falsify topology while measurements are jammed; in MTJs, the barrier itself breaks electrically. The term is thus encyclopedically best treated as a field-dependent homonym rather than as a single unified concept.