---
title: Iyama–Wemyss Mutations
url: https://www.emergentmind.com/topics/iyama-wemyss-mutations
type: topic
---

# Iyama–Wemyss Mutations

Iyama–Wemyss mutations are operations on modifying or maximal modifying modules over Gorenstein rings that replace a chosen summand by an exchange kernel obtained from approximation theory. If \(M\) gives a non-commutative crepant resolution (NCCR), then the mutated module again gives an NCCR, and the corresponding endomorphism rings are derived equivalent. In recent work, these mutations appear as the module-theoretic mechanism behind connectivity of toric NCCRs, wall-crossing for magic windows in geometric invariant theory, symmetries of BPS and Gopakumar–Vafa invariants for cDV singularities, wall-and-chamber structures in \(K\)-theory and Bridgeland stability, and mutation networks for NCCRs of anticanonical cones over del Pezzo surfaces [2510.26252] [2310.11057] [2207.13540] [2603.04858] [2604.11319].

## 1. Modifying modules and the formal mutation operation

Let \(R\) be a normal Gorenstein \(\Bbbk\)-algebra, or more generally a Gorenstein normal domain of dimension at least \(2\), and let \(M\in \mathrm{refl}\,R\). A reflexive \(R\)-module \(M\) is called modifying if \(\End_R(M)\in \mathrm{CM}\,R\). It gives an NCCR if, in addition, \(\End_R(M)\) has finite global dimension. In the toric setting, the summands of \(M\) may be rank-one divisorial modules \(S_g\) indexed by the divisor class group [2310.11057] [2510.26252].

Write
\[
M=\bigoplus_i M_i
\]
and fix a nonzero summand \(N\in \add M\). With \((-)^*=\Hom_R(-,R)\), one forms two approximation sequences. For a right \((\add N)\)-approximation of \(M\),
\[
0\to K_0\to N_0\to M\to 0,
\qquad N_0\in \add N,
\]
the right mutation is
\[
\mu_N^+(M):=N\oplus K_0.
\]
Dually, for
\[
0\to K_1\to N_1^*\to M^*\to 0,
\qquad N_1^*\in \add N^*,
\]
the left mutation is
\[
\mu_N^-(M):=N\oplus K_1^*.
\]
Iyama–Wemyss show that if \(M\) is modifying, then so are \(\mu_N^\pm(M)\), and if \(M\) gives an NCCR, then so do \(\mu_N^\pm(M)\); moreover \(\End_R(M)\) and \(\End_R(\mu_N^\pm(M))\) are derived equivalent [2510.26252].

## 2. Exchange sequences, tilting realizations, and the three-dimensional case

For a \(3\)-dimensional basic NCCR
\[
\Lambda=\End_R(M),
\qquad
M=M_k\oplus N
\]
with \(M_k\) indecomposable and non-projective, one chooses a right \(\add(N)\)-approximation
\[
f\colon N'\longrightarrow M_k,
\qquad N'\in \add(N),
\]
fitting into
\[
0\longrightarrow U_k\longrightarrow N'\xrightarrow{\,f\,} M_k\longrightarrow 0.
\]
The right mutation at \(M_k\) is then
\[
\mu_k^+(M)=N\oplus U_k.
\]
Dually, one takes a right \(\add(N^*)\)-approximation of \(M_k^*\) and dualizes to obtain \(\mu_k^-(M)\). In dimension \(3\), these two operations agree up to grading shift, so one writes simply
\[
\mu_k(M)=\mu_k^+(M)=\mu_k^-(M).
\]
In the \(3\)-dimensional isolated Gorenstein case, if
\[
M=N\oplus N^c
\]
with \(N=M_i\), there is a unique minimal right \((\add N^c)\)-approximation
\[
a^+\colon N^+\to N
\]
and an exchange sequence
\[
0\to K_i\to N^+\to N\to 0;
\]
dually one obtains \(K_i^*\), and in this isolated case one shows \(K_i^*\cong K_i\), so the right and left mutations coincide [2604.11319] [2603.04858].

The categorical realization is by tilting. If \(M=N\oplus N^c\) gives an NCCR, then there are canonical derived equivalences
\[
\Phi_N^\pm\colon
D^b(\mod \End_R(M))
\xrightarrow{\sim}
D^b(\mod \End_R(\mu_N^\pm(M))).
\]
Concretely, one constructs a tilting bimodule \(T_N\) with
\[
\Phi_N^+=R\Hom(T_N,-),
\qquad
T_N\cong \Hom_R(M,\mu_N^+(M)).
\]
In the cDV setting, the mutation can also be represented by a \(2\)-term tilting complex
\[
T=\bigoplus_{k\neq i}P_k\oplus T_i
\]
with \(\End_{D(\Lambda)}(T)\cong \Lambda_i\), producing inverse equivalences by \(R\Hom\) [2310.11057] [2207.13540].

A recurrent compression in the literature is to suppress the sign and write \(\mu_i(M)\). That shorthand is justified only in the \(3\)-dimensional isolated or graded setting described above; in general, right and left Iyama–Wemyss mutations are defined separately.

## 3. Rank-one toric singularities and upper-set combinatorics

For a Gorenstein toric singularity
\[
R=k[\sigma^\vee\cap M]
\]
with class group \(\operatorname{Cl}R\cong G\) of rank one, Tomonaga gives a classification of toric NCCRs in terms of upper sets. After choosing generators \(x_1,\dots,x_\ell\) of positive class and \(x_{\ell+1},\dots,x_{\ell+\ell'}\) of negative class, one sets
\[
H=G/(\text{torsion part}),
\qquad
p=\sum_{i=1}^{\ell}\pi(x_i)= -\sum_{j=1}^{\ell'}\pi(x_{\ell+j})>0,
\]
and defines a partial order on \(H\) by
\[
h'\ge h
\iff
h'-h\in
\sum_{i=1}^{\ell}\mathbb Z_{\ge0}\cdot \pi(x_i)
+
\sum_{j=1}^{\ell'}\mathbb Z_{\ge0}\cdot (-\pi(x_{\ell+j})).
\]
An upper set \(I\subset H\) is a subset such that \(h\in I\) and \(h'\ge h\Rightarrow h'\in I\); it is non-trivial if \(I\neq \varnothing,H\). To such an \(I\) one associates
\[
J(I):=\pi^{-1}(I\cap (I^c+p))\subset G,
\qquad
M_I:=\bigoplus_{g\in J(I)}S_g.
\]
The key theorem is that
\[
I\longmapsto M_I
\]
gives a bijection between non-trivial upper sets \(I\subset H\) and splitting modules \(M\subset \mathrm{Ref}\,R\) which are NCCRs [2510.26252].

Because the set of non-trivial upper sets is a finite lattice under inclusion, any two such upper sets are connected by elementary moves
\[
I\to I\setminus\{m\},
\]
where \(m\in I\) is minimal. In the toric module picture this operation corresponds exactly to right or left Iyama–Wemyss mutation at the summand \(S_m\). More precisely, if \(I'=I\setminus\{m\}\) and
\[
N=\bigoplus_{g\in J(I)\setminus J(I')}S_g=S_m,
\]
then
\[
(\mu_N^+)^{\ell'-1}(M_I)\cong M_{I'},
\qquad
(\mu_N^-)^{\ell-1}(M_I)\cong M_{I'}.
\]
By induction, any two splitting NCCRs are related by iterated Iyama–Wemyss mutations and hence are derived equivalent. In rank one, the only criterion for mutation is that \(m\in I\) be minimal; geometrically one may think of \(H\) as the lattice points in the line \(\operatorname{Cl}R/\mathrm{tors}\cong \mathbb Z\), and upper sets are rays of the form \(\{m,m+1,m+2,\dots\}\) [2510.26252].

The basic example is the \(3\)-dimensional \(A_1\) toric singularity, where \(G=\mathbb Z\), the positive generators \(x,y\) have weight \(+1\), the negative generators \(z,w\) have weight \(-1\), and \(p=2\). Up to translation, the only non-trivial upper set is
\[
I=\{0,1,2,3,\dots\},
\]
so
\[
J(I)=\{0,1\},
\qquad
M_I=S_0\oplus S_1.
\]
Its unique minimal element is \(m=0\). Removing \(m\) gives \(I'=\{1,2,3,\dots\}\), hence
\[
J(I')=\{1,2\},
\qquad
M_{I'}=S_1\oplus S_2.
\]
On the module side, taking \(N=S_0\), the left mutation produces \(S_1\oplus S_2\). Concretely, the exact sequence
\[
0\to S(-2)\to S(-1)^2\to S(0)\to 0
\]
resolves \(S_0\), and dualizing shows that the new summand is \(S(1)\). Thus a single Iyama–Wemyss mutation replaces \(S_0\oplus S_1\) by \(S_1\oplus S_2\), exactly matching the combinatorics of \(I\to I'\) [2510.26252].

## 4. Wall-crossing in geometric invariant theory

In the GIT framework of Hara–Hirano, let \(X\) be a generic quasi-symmetric representation of a connected reductive group \(G\), with character lattice \(M=\Hom(T,\mathbb G_m)\). For a generic parameter \(\delta\in M_\mathbb R\) avoiding a finite \(W\)-periodic hyperplane arrangement \(\mathscr H\), one defines the magic window subcategory
\[
\mathcal M(\delta+V)\subset D^b(\coh[X/G])
\]
generated by the vector bundles
\[
\{\,V(\lambda)\otimes \mathcal O_X\mid \lambda\in (\delta+V)\cap M^+\,\}.
\]
Halpern–Leistner and Sam’s theorem identifies \(\mathcal M(\delta+V)\) with the derived category of the GIT quotient stack and, under the induced algebra description, with
\[
D^b(\mod A_\delta),
\qquad
A_\delta=\End_R(M_\delta),
\qquad
M_\delta=\bigoplus_{\lambda\in (\delta+V)\cap M^+}M(\lambda).
\]
For adjacent generic parameters \(\delta,\delta'\), the window-shift functor
\[
\Phi_{\delta\to\delta'}=
\mathrm{res}(\delta')\circ \mathrm{res}(\delta)^{-1}
\]
is induced by the tilting module
\[
T(\delta,\delta'):=\Hom_R(M_\delta,M_{\delta'}),
\]
and under the identifications above one has
\[
\Phi_{\delta\to\delta'}\cong R\Hom(T(\delta,\delta'),-).
\]
Thus wall-crossing equivalences between magic windows coincide with derived equivalences between NCCRs induced by tilting modules [2310.11057].

The same paper identifies these wall-crossings with Iyama–Wemyss mutation. Writing
\[
M_\delta=N\oplus N',
\qquad
N=\bigoplus_{\lambda\in C\setminus C'}M(\lambda),
\qquad
N'=\bigoplus_{\lambda\in C\cap C'}M(\lambda),
\]
the matching \(C\setminus(C\cap C')\simeq C'\setminus(C\cap C')\) reproduces the exchange kernel, and one gets
\[
M_{\delta'}\cong \mu_N^+(M_\delta).
\]
When \(G=T\) is a torus, exactly one summand crosses the wall, say \(N=M(\lambda)\). In that case, if
\[
d=\#\{\beta_i\mid (\beta_i,\ell)>0\},
\qquad
d'=\#\{\beta_i\mid (\beta_i,\ell)<0\},
\]
then
\[
\mu_N^{d-1}(M_\delta)\cong M_{\delta'},
\qquad
\mu_N^{d'-1}(M_{\delta'})\cong M_\delta,
\]
so the mutation is periodic of period \(d+d'-2\). In the one-parameter Calabi–Yau complete-intersection setting, the corresponding window-shifts generate an action of
\[
\pi_1(\mathbb P^1\setminus\{0,1,\infty\})
\]
on \(D^b(\coh Y)\), with generators identified with appropriate compositions of Iyama–Wemyss mutations [2310.11057].

A concrete torus example is
\[
G=\mathbb G_m,
\qquad
X=\Sym^3(\Bbbk^2)\oplus \Sym^3(\Bbbk^2)^*,
\]
for which the affine quotient is the cone over the Segre cubic threefold. For any chamber between \(m-\tfrac12\) and \(m+\tfrac12\),
\[
M_\delta=\bigoplus_{i=0}^3 R(i).
\]
Crossing the wall at \(m+\tfrac12\) exchanges \(R(m+3)\) with \(R(m-3)\), and iterating the corresponding mutation six times returns to the original module, matching the general period-\((d+d'-2)=6\) statement [2310.11057].

## 5. cDV singularities, restricted roots, and BPS/GV wall-crossing

For a cDV singularity admitting an NCCR
\[
\Lambda=\End_R(M)
\]
of affine Dynkin type \((\Delta_{\mathrm{aff}},J)\), write
\[
M=\bigoplus_{k\in J^c} M_k
\]
and fix a vertex \(i\in J^c\). Iyama–Wemyss define exchange sequences
\[
0\to K\to Z^*\to M_i^*\to 0
\]
in \(\mathrm{ref}\,R\)-mod, where \(Z\in \add(M/M_i)\), and after dualizing,
\[
0\to M_i\to Z\to K^*\to 0.
\]
One then sets
\[
\mu_i^-(M):=(M/M_i)\oplus K^*,
\qquad
\Lambda_i:=\End_R(\mu_i^-(M)).
\]
At the derived level, if \(P_k=e_k\Lambda\), the object
\[
T=\bigoplus_{k\neq i}P_k\oplus T_i
\]
is a \(2\)-term tilting complex with \(\End_{D(\Lambda)}(T)\cong \Lambda_i\), giving derived equivalences
\[
\mu_i^+=R\Hom_\Lambda(T,-)\colon D^b(\mod \Lambda)\to D^b(\mod \Lambda_i),
\]
\[
\mu_i^-=R\Hom_{\Lambda_i}(T,-)\colon D^b(\mod \Lambda_i)\to D^b(\mod \Lambda).
\]
At the level of modules, the short exact sequences
\[
0\to P_i\to \bigoplus_{j\to i}P_j\to T_i^+\to 0,
\qquad
0\to T_i^-\to \bigoplus_{i\to k}P_k\to P_i\to 0
\]
realize the two one-step mutations [2207.13540].

The same affine Dynkin data governs a restricted-root combinatorics. If
\[
\pi_J\colon \mathbb Z\Delta_{\mathrm{aff}}\to \mathbb ZJ^c
\]
is the restriction map, then
\[
R(\Delta_{\mathrm{aff}},J)
=
\{\pi_J(r)\neq 0\mid r\in \mathrm{Rts}\,\Delta_{\mathrm{aff}}\}
\]
is the set of restricted roots. A key vanishing theorem states that if \(\delta\in \mathbb N J^c\) has \(\gcd(\delta)=d\) and
\[
\delta/d\notin R(\Delta_{\mathrm{aff}},J),
\]
then the noncommutative BPS invariant \(\Omega(\delta)\) vanishes. The associated hyperplane arrangement has walls
\[
\{\theta\mid \theta(\delta)=0\}
\quad \text{for } \delta\in R(\Delta_{\mathrm{aff}},J),
\]
and the chambers are labelled by the \(2\)-term tilting complexes of \(\Lambda\) [2207.13540].

Mutation at \(i\) defines an element \(\omega_i\) in the wall-crossing groupoid. For adjacent \(2\)-term tilting complexes, there is a derived equivalence between the corresponding stability-theoretic hearts,
\[
\mu_i^+=R\Hom_\Lambda(T,-)\colon S_\theta(\Lambda)\simeq S_{\omega_i\cdot \theta}(\Lambda_i).
\]
Passing to BPS invariants, if \(\mu_i\Lambda\cong \Lambda_i\) is again given by a quiver with potential and \(\delta\in R(\Delta_{\mathrm{aff}},J)\) is indivisible and not colinear with \(\pi_J(\alpha_i)\) or \(\pi_J(r_{\mathrm{imag}})\), then
\[
\Omega_\Lambda^{\mathrm{num}}(\delta)=\Omega_{\Lambda_i}^{\mathrm{num}}(\omega_i\cdot \delta).
\]
In the geometric crepant-resolution picture, with \(n_\beta\) the genus-zero Gopakumar–Vafa numbers of a crepant resolution \(Y\), one obtains
\[
n_\beta=n^+_{\omega_i\cdot \beta}
\quad \text{if } C_i \text{ flops},
\]
and
\[
n_\beta=n_{\omega_i\cdot \beta}
\quad \text{otherwise},
\]
for every effective class \(\beta\notin \mathbb Z\cdot [C_i]\). The example \(R=\mathbb A^3/D_{2n}\) realizes mutation at a vertex \(i\in D_{n+1}\) as the corresponding finite-Weyl reflection on restricted roots, yielding explicit identifications among nonzero invariants [2207.13540].

## 6. Mutation cones and Bridgeland stability conditions

For a \(3\)-dimensional complete local Gorenstein isolated singularity \((R,\mathfrak m)\) and a basic maximal modifying \(R\)-module
\[
M=M_1\oplus \cdots \oplus M_n,
\qquad
\Lambda=\End_R(M),
\]
Hara–Hirano define a wall-and-chamber structure in real \(K\)-theory called the mutation cone. For any basic modifying module \(N\in (M)\), write \(\Lambda_N=\End_R(N)\) and
\[
K^N:=K_0(\mathrm{Perf}\,\Lambda_N),
\qquad
K^N_\mathbb R=K^N\otimes_\mathbb Z \mathbb R.
\]
If \(P_1,\dots,P_n\) are the indecomposable projectives of \(\Lambda_N\), the open positive cone is
\[
C_+^N=\sum_{i=1}^n \mathbb R_{>0}[P_i]\subset K^N_\mathbb R.
\]
For a basic tilting \(\Lambda_N\)-module \(T=\bigoplus T_i\),
\[
C_T^N=\sum_{i=1}^n \mathbb R_{>0}[T_i]
\]
is an open simplicial cone whose closure is polyhedral. Mutation functors induce isomorphisms on \(K\)-groups; composing along a path \(p\) of simple mutations gives \(\phi_p\), and one shows that \(\phi_p\) depends only on the endpoints. Fixing \(M\), the mutation cone is
\[
\mathrm{Cone}(M):=\bigcup_{N\in (M)} \phi_N^M(C_+^N)\subset K^M_\mathbb R.
\]
The open chambers \(\phi_N^M(C_+^N)\) are disjoint, their closures meet along codimension-one faces exactly when the corresponding modules differ by a single simple mutation, and there is a bijection
\[
\{\text{chambers of }\mathrm{Cone}(M)\}
\longleftrightarrow
\{\text{elements of }(M)\}.
\]
Thus crossing a single wall is exactly applying one simple mutation [2603.04858].

The same paper introduces the tilting-noetherian property for \(\Lambda\), meaning that there are no infinite ascending chains in the poset of basic tilting \(\Lambda\)-modules. It proves that, for maximal modifying modules, \(\Lambda\) is tilting-noetherian if and only if all maximal modifying \(R\)-modules are connected by iterated mutations. This makes mutation connectivity an intrinsic finiteness property of the tilting theory [2603.04858].

On the triangulated side, let \(\mathscr D_M\subset D^b(\mod \Lambda)\) be the full subcategory whose cohomology modules have support \(\mathfrak m\), and let \(A_M=\mathrm{fl}\,\Lambda\) be its standard finite-length heart. A heart \(A\subset \mathscr D_M\) is called modifying if it is obtained from a heart \(A_N\) by a composition of mutation functors. The resulting space \(\mathrm{Stab}^{\mathrm{mdf}}\mathscr D_M\) of modifying Bridgeland stability conditions is a disjoint union of chambers \(U_\alpha\simeq H^n\), and the forgetful map gives a regular covering
\[
\pi\colon \mathrm{Stab}^{\mathrm{mdf}}\mathscr D_M\to \mathrm{Cone}(M)_\mathbb C,
\]
where
\[
\mathrm{Cone}(M)_\mathbb C=\bigcup_N \phi_N^M(H_+^n).
\]
The subgroup \(PBr\,\mathscr D_M\subset \Aut \mathscr D_M\) generated by mutation functors acts freely and transitively on the set of chambers and is precisely the Galois group of this covering. The same framework also describes the subgroup of autoequivalences preserving \(\mathrm{Stab}^{\mathrm{mdf}}\mathscr D_M\) in terms of mutation functors and class-group twists [2603.04858].

## 7. Anticanonical cones over del Pezzo surfaces and polygonal models

For a del Pezzo surface \(X\), let
\[
R_X=\bigoplus_{k\ge 0}\Gamma(X,\omega_X^{-k}),
\qquad
Y=\Tot(\omega_X)\xrightarrow{\;\pi\;} \Spec R_X.
\]
A very strong exceptional collection \((E_0,\dots,E_{n-1})\) of vector bundles on \(X\) is a full exceptional collection satisfying
\[
\Ext_X^\ell(E_i,E_j\otimes \omega_X^{-k})=0
\quad
(\ell\neq 0,\; k\ge 0),
\]
equivalently, with slopes increasing in the interval \([\mu(E_0),\mu(E_0)+K_X^2)\). Such a collection generates a geometric helix
\[
\mathcal H=\{E_j\}_{j\in \mathbb Z},
\qquad
E_{j-n}=E_j\otimes \omega_X.
\]
Any thread of \(\mathcal H\) is very strong, and the rolled-up helix algebra
\[
B(\mathcal H)=\End_Y\!\Bigl(\pi^*\!\Bigl(\bigoplus_{i=1}^n E_i\Bigr)\Bigr)
\]
is a \(3\)-Calabi–Yau NCCR of \(R_X\). Conversely, every graded NCCR of \(R_X\) is Morita equivalent to such a rolled-up helix algebra [2604.11319].

In this setting, Iyama–Wemyss mutation can be read directly on the helix. A mutation of \(B(\mathcal H)\) at the summand corresponding to \(E_i\) is realized by a sequence of left or right braid mutations on a thread until the result is again a very strong thread, yielding a mutated helix \(\mathcal H'\). On the toric-system–polygon side of Hille–Perling, the helix is encoded by line segments
\[
\ell_{i,i+1}\in \mathbb R^2
\]
whose midpoints trace a convex polygon \(P\) containing the origin. Mutation at index \(i\) is then a simple affine shear in the plane replacing \(\ell_{i,i+1}\) by a new segment parallel to \(\ell_{i-1,i}\), with the rest of the polygon shifted accordingly. Algebraically, this shear is the DWZ quiver-with-potential mutation, and it agrees with the Iyama–Wemyss mutation on the NCCR [2604.11319].

The polygon also encodes the quiver. If
\[
P=\bigl|\ell_{0,1},\ell_{1,2},\dots,\ell_{n-1,0}\bigr|\subset \mathbb R^2,
\]
then convexity of \(P\) is equivalent to the very-strong condition, and the number of arrows \(i\to j\) is
\[
c_{ij}
=
\omega\bigl(\ell_{i,i+1}-\ell_{i-1,i},\;\ell_{j,j+1}-\ell_{j-1,j}\bigr),
\]
with respect to the normalized volume form \(\omega\). A quiver mutation at vertex \(i\) is precisely the shear
\[
\ell_{i,i+1}\mapsto
\ell'_{i,i+1}
=
\ell_{i-1,i}
+
\frac{\omega(\ell_{i-1,i},\,\ell_{i+1,i+2})}
{\omega(\ell_{i,i+1},\,\ell_{i+1,i+2})}
\,m_{i+1}.
\]
Up to tensoring the helix by a line bundle, shifting all objects in degree, permuting orthogonal blocks, and rotating the labeling, every helix can be reached from any other by a finite sequence of such quiver mutations. Accordingly, all NCCRs of anticanonical cones over del Pezzo surfaces are connected by a finite chain of Iyama–Wemyss mutations. This is presented as the non-commutative analogue of the statement that crepant resolutions are connected by flops [2604.11319].

Source: https://www.emergentmind.com/topics/iyama-wemyss-mutations