---
title: Weighted State t-Designs
url: https://www.emergentmind.com/topics/weighted-state-t-designs
type: topic
---

# Weighted State t-Designs

Weighted state \(t\)-designs are finite ensembles of pure quantum states whose discrete \(t\)-th moments reproduce the Haar moment on complex projective space. In dimension \(d\), a weighted ensemble \(\{(|\psi_i\rangle,w_i)\}\), with \(w_i\ge 0\) and \(\sum_i w_i=1\), is a weighted state \(t\)-design when
\[
\sum_i w_i \,\big(|\psi_i\rangle\!\langle\psi_i|\big)^{\otimes t}
\;=\;
\int_{\mathbb{CP}^{d-1}} \big(|\psi\rangle\!\langle\psi|\big)^{\otimes t}\, d\mu_{\rm Haar}(\psi)
\;=\;
\frac{P_{\mathrm{sym}^{(t)}}}{\binom{d+t-1}{t}},
\]
where \(P_{\mathrm{sym}^{(t)}}\) is the projector onto the totally symmetric subspace of \((\mathbb{C}^d)^{\otimes t}\). This notion organizes exact Haar-moment quadrature on \(\mathbb{CP}^{d-1}\), links directly to SIC-POVMs and unions of MUBs in the case \(t=2\), and has recently been expanded by projective toric constructions, weighted minimality results, and explicit nearly tight weighted 2-designs in every dimension [2311.13479][2101.11756][2501.14938].

## 1. Moment identities and equivalent characterizations

In the complex projective setting, weighted state \(t\)-designs and weighted projective \(t\)-designs are the same object described in two equivalent languages. On the unit sphere \(S(\mathbb{C}^d)\), a finite family of unit vectors \(\{x_k\}_{k\in[n]}\subset \mathbb{C}^d\) with weights \(\{w_k\}\) is a weighted projective \(t\)-design if
\[
\sum_{k\in[n]} w_k (x_k^{\otimes t})(x_k^{\otimes t})^*
=
\binom{d+t-1}{t}^{-1}\Pi^{(t)},
\]
where \(\Pi^{(t)}\) is the orthogonal projection onto \((\mathbb{C}^d)^{\otimes t}_{\mathrm{sym}}\). Equivalently, the design reproduces the average of every polynomial in \(\mathrm{Hom}_a(t)\), and in integral form one has
\[
\int d\psi \, (|\psi\rangle\langle\psi|)^{\otimes t}
=
\binom{d+t-1}{t}^{-1}\Pi_{\mathrm{sym}}^{(t)}.
\]
The unweighted case is recovered by setting \(w_k=1/n\) [2101.11756].

A second characterization is through the weighted \(t\)-frame potential
\[
\mathrm{FP}_t(\{\psi_i\},\{w_i\})
=
\sum_{i,j} w_i w_j \, |\langle \psi_i|\psi_j\rangle|^{2t}.
\]
A weighted ensemble is a \(t\)-design if and only if this equals the Haar value
\[
\int |\langle \psi|\phi\rangle|^{2t}\,d\mu(\psi)\,d\mu(\phi)
=
\frac{t!\,(d-1)!}{(d+t-1)!}.
\]
For \(t=2\),
\[
\sum_i w_i (|\psi_i\rangle\!\langle\psi_i|)^{\otimes 2}
=
\frac{2}{d(d+1)}\Pi_{\mathrm{sym}}^{(2)}
=
\frac{I+F}{d(d+1)},
\qquad
\mathrm{FP}_2=\frac{2}{d(d+1)},
\]
with \(F\) the swap operator on \(\mathbb{C}^d\otimes \mathbb{C}^d\) [2311.13479][2101.11756][2501.14938].

Recent work also uses an equivalent unnormalized convention for \(t=2\), requiring
\[
\sum_{i=1}^N w_i (\Pi(\psi_i))^{\otimes 2}
=
P_{\mathrm{sym}}^{(2)}
=
\frac{I+F}{2},
\]
so that \(\sum_i w_i=d(d+1)/2\). This is simply a rescaling of the normalized convention by the factor \(\binom{d+1}{2}\) [2501.14938].

## 2. Tightness, weighted minimality, and cardinality bounds

For complex projective 2-designs, the sharp cardinality threshold is \(d^2\). If \(X=\{x_k\}_{k\in[n]}\subset S(\mathbb{C}^d)\) is an unweighted projective 2-design, then \(n\ge d^2\), with equality precisely when \(X\) is equiangular. In that equality case, \(X\) is a tight projective 2-design, or SIC-POVM, characterized by
\[
|\langle \psi_i|\psi_j\rangle|^2=\frac{1}{d+1}\qquad (i\neq j).
\]
The weighted analogue is equally rigid: every weighted projective 2-design in \(\mathbb{C}^d\) has size at least \(d^2\), and if equality holds then the weights are all equal. Thus weighted tightness coincides with the unweighted tight case [2101.11756].

For general dimensions, existence is known without achieving tightness. A weighted projective 2-design of size \((d+1)^2\) exists for each \(d\), via Carathéodory’s theorem applied to the convex cone spanned by rank-1 symmetric projectors in \((\mathbb{C}^d)^{\otimes 2}_{\mathrm{sym}}\) [2101.11756]. More recent constructions improve this substantially: if \(n(d)\) denotes the minimal cardinality of a weighted complex projective 2-design in \(\mathbb{CP}^{d-1}\), then
\[
n(d)\le d^2+O(d^{1.525}),
\]
so \(n(d)=(1+o(1))d^2\). This establishes explicit nearly tight weighted 2-designs in every dimension [2501.14938].

Projective toric designs introduce a separate family of lower bounds. If \(X\subset P(T^n)\) is a finite projective toric \(t\)-design, then
\[
|X|\ge G_{n-1}\!\Big(\big\lfloor\tfrac{t}{2}\big\rfloor\Big),
\]
where \(G_{n-1}(s)\) is the crystal ball number in the root lattice \(A_{n-1}\),
\[
G_{n-1}(s)=\sum_{i=0}^s \binom{n-1}{i}^2 \binom{n-i+s-1}{s-i}
=
{}_3F_2(1-n,-s,n;1,1;1).
\]
For even \(t\), minimal projective toric designs are forced to be uniformly weighted [2311.13479].

## 3. Projective toric designs and the lift to quantum state designs

A central structural development is the use of the diagonal subgroup of the unitary group. The maximal torus of \(U(n)\) is \(T^n=(\mathbb{R}/2\pi\mathbb{Z})^n\), and the projective torus is the quotient
\[
P(T^n)\coloneqq T^n/T,
\]
where the embedded diagonal copy of \(T\) accounts for global phase. Projective toric monomials have equal degree and conjugate degree and lift to exponentials of the form
\[
\exp\!\Big(i\sum_{k=1}^t (\phi_{a_k}-\phi_{b_k})\Big).
\]
A measure space \((X\subset P(T^n),\Sigma,\nu)\) is a \(P(T^n)\) \(t\)-design if, for all \(a,b\in I_n^t\),
\[
\int_X \exp\!\Big(i\sum_{j=1}^t(\phi_{a_j}-\phi_{b_j})\Big)d\nu(\phi)
=
\int_{P(T^n)} \exp\!\Big(i\sum_{j=1}^t(\phi_{a_j}-\phi_{b_j})\Big)d\mu_{n-1}(\phi),
\]
and the Haar integral equals \(1\) precisely when the multisets \(\{\!\{a_1,\dots,a_t\}\!\}\) and \(\{\!\{b_1,\dots,b_t\}\!\}\) coincide, and \(0\) otherwise [2311.13479].

The connection to weighted state \(t\)-designs is through a factorization of the Fubini–Study measure. With \(\Delta^{d-1}\) the probability simplex, define
\[
\pi:\Delta^{d-1}\times P(T^d)\longrightarrow \mathbb{CP}^{d-1},
\qquad
(p,[\phi])\mapsto [\,|p,\phi\rangle\,],
\]
where
\[
|p,\phi\rangle=\sum_{n=1}^d \sqrt{p_n}\,e^{i\phi_n}|n\rangle.
\]
The pullback of the Fubini–Study measure factorizes as
\[
\pi^*(d\mu_{\mathrm{FS}})=dp\,d\phi,
\]
that is, Lebesgue measure on \(\Delta^{d-1}\) times Haar measure on \(P(T^d)\). Consequently, if \(D\subset \Delta^{d-1}\) is a \(\Delta^{d-1}\) \(t\)-design and \(X\subset P(T^d)\) is a \(P(T^d)\) \(t\)-design, then the set of states \(\{[|p,\phi\rangle]:p\in D,\phi\in X\}\) with product weights is a weighted complex-projective \(t\)-design in dimension \(d\) [2311.13479].

This factorized construction is important because it separates amplitudes and phases. A plausible implication is that exact quantum-state cubature can be built from simpler exact cubature rules on the simplex and on the projective torus, rather than directly on \(\mathbb{CP}^{d-1}\).

## 4. Explicit constructions from Sidon sets, Singer sets, and finite abelian groups

Projective toric designs admit several explicit families. For any \(n,t\in\mathbb{N}\),
\[
X=\Big\{\big(0,\tfrac{2\pi d_1}{t+1},\dots,\tfrac{2\pi d_{n-1}}{t+1}\big): d\in\mathbb{Z}_{t+1}^{\,n-1}\Big\},
\qquad
w=(t+1)^{-(n-1)},
\]
is a \(P(T^n)\) \(t\)-design. For \(n\in\mathbb{N}\) and a prime \(p>\max(2,n)\), the set
\[
X=\Big\{\big(0,\tfrac{2\pi}{p}(q_1+q_2), \tfrac{2\pi}{p}(2q_1+4q_2),\dots,\tfrac{2\pi}{p}((n-1)q_1+(n-1)^2q_2)\big): q_1,q_2\in \mathbb{Z}_p\Big\}
\]
with uniform weights is a \(P(T^n)\) 2-design. Singer sets and more general \(B_t\bmod m\) sets provide denser constructions: if \(n-1\) is a prime power, there exist \(P(T^n)\) 2-designs of size \(n(n-1)+1\), and more generally \(P(T^n)\) \(t\)-designs of size
\[
\frac{(n-1)^{t+1}-1}{n-2}.
\]
These constructions also imply the additive-combinatorial bound
\[
m\ge G_{n-1}\!\Big(\big\lfloor\tfrac{t}{2}\big\rfloor\Big)
\]
for a \(B_t \bmod m\) set of size \(n\); for \(t=2\), this becomes \(m\ge n(n-1)+1\), with equality for Singer’s construction when \(n-1\) is a prime power [2311.13479].

A different but closely related construction yields weighted projective 2-designs directly from finite abelian groups. Let \(G\) be a finite abelian group and \(S\subset G\) a Sidon set. For each character \(\alpha\in \widehat{G}\), define \(x_\alpha\in \mathbb{C}^{|S|}\) by
\[
x_\alpha(s)=\frac{\alpha(s)}{\sqrt{|S|}},
\]
and for each \(r\in S\), define the standard basis vector \(e_r\) by \(e_r(s)=\delta_{r,s}\). Then, with the unnormalized convention,
\[
\sum_{\alpha\in\widehat{G}} w_\alpha (\Pi(x_\alpha))^{\otimes 2}
+
\sum_{r\in S} w_r (\Pi(e_r))^{\otimes 2}
=
P_{\mathrm{sym}}^{(2)}
\]
if one chooses
\[
w_\alpha=\frac{|S|^2}{2|G|}\quad \text{for all }\alpha\in \widehat{G},
\qquad
w_r=\frac12\quad \text{for all }r\in S.
\]
Under the normalized convention, all weights are scaled by \(2/(|S|(|S|+1))\). The resulting weighted projective 2-design in \(\mathbb{CP}^{|S|-1}\) has size
\[
N=|G|+|S|.
\]
Its structure has two layers: a harmonic layer of \(|G|\) vectors \(x_\alpha\) and a coordinate layer of \(|S|\) basis vectors \(e_r\); weights are uniform within each layer but nonuniform across layers [2501.14938].

Dense Sidon families produce explicit asymptotic bounds. The paper lists Erdős–Turán, Singer, Bose, Spence, and Hughes families, all with \(|S|\) near \(\sqrt{|G|}\). Since subsets of Sidon sets remain Sidon, one may tune \(|S|\) to a target dimension. Combining Erdős–Turán with the Baker–Harman–Pintz estimate \(p(d)\le d+O(d^{0.525})\) yields
\[
n(d)\le d^2+O(d^{1.525}),
\]
while Singer and Bose recover the sharp families \(N=d^2+1\) and \(N=d^2+d-1\) under the stated prime-power hypotheses [2501.14938].

## 5. Uniform designs, MUBs, SICs, and counterexamples

The best-known unweighted projective 2-designs arise from SICs and complete sets of MUBs, but weighted and toric constructions show that these do not exhaust the landscape. A complete set of MUBs, when it exists, consists of \(N=d(d+1)\) uniformly weighted states with overlap structure
\[
|\langle \psi|\phi\rangle|^2=
\begin{cases}
0, & \text{if in same basis and orthogonal},\\[3pt]
1/d, & \text{if in different bases}.
\end{cases}
\]
Projective toric geometry makes this equivalent to the existence of a uniformly weighted \(P(T^d)\) 2-design \(X\) of size \(d^2\) such that
\[
\forall\,\phi\neq \theta\in X:\quad
\Big|\sum_{j=1}^{d} e^{\,i(\phi_j-\theta_j)}\Big|^2\in \{0,\,d\}.
\tag{4.1}
\]
When such an \(X\) exists, concatenation with the simplex 2-design produces the standard uniform \(d(d+1)\)-state 2-design with MUB overlap structure [2311.13479].

The simplex ingredient is explicit. Proposition 4.1 uses the \(d\) extremal points \(e^{(j)}\) and the centroid \(c\), with weights
\[
w(e^{(j)})=\frac{1}{d(d+1)},
\qquad
w(c)=\frac{d}{d+1}.
\]
If \(X\subset P(T^d)\) is a uniform \(P(T^d)\) 2-design of size \(d^2\) arising from a \(B_2\bmod d^2\) set, and one concatenates only the centroid part with \(X\), then every resulting state has weight \(1/[d(d+1)]\), and the total number of states is \(d+d^2=d(d+1)\). This gives a uniformly weighted \(\mathbb{CP}^{d-1}\) 2-design that need not satisfy the MUB overlap condition [2311.13479].

That mechanism disproves Zhu’s 2015 conjecture asserting that any uniformly weighted quantum state 2-design in dimension \(d\) of size \(\le d(d+1)\) is either a complete set of MUBs or a SIC. The counterexamples are explicit: choosing a Sidon set modulo \(d^2\) that yields a uniform \(P(T^d)\) 2-design violating (4.1) produces a uniform \(\mathbb{CP}^{d-1}\) 2-design of size \(d(d+1)\) that is not a complete set of MUBs. In dimension \(3\), the example
\[
z=(0,1,3)\ \text{mod}\ 9,
\qquad
X=\Big\{\big(0,\tfrac{2\pi k}{9},\tfrac{6\pi k}{9}\big): k=0,\dots,8\Big\},
\]
yields the 12 states
\[
\{|1\rangle,|2\rangle,|3\rangle\}
\ \cup\
\Big\{\tfrac{1}{\sqrt 3}\big(|1\rangle + e^{2\pi ik/9}|2\rangle + e^{6\pi ik/9}|3\rangle\big): k=0,\dots,8\Big\},
\]
each with weight \(1/12\); this is a uniformly weighted \(\mathbb{CP}^2\) 2-design but not a complete set of MUBs [2311.13479].

SICs remain the minimal case. A SIC-POVM is a minimal \(\mathbb{CP}^{d-1}\) 2-design with \(N=d^2\), uniform weights, and
\[
|\langle \psi_i|\psi_j\rangle|^2=\frac{1}{d+1}\qquad (i\ne j).
\]
Projective toric methods produce almost-minimal designs of size \(d^2+1\), but do not resolve SIC existence in arbitrary dimension [2311.13479].

## 6. Channel-theoretic reformulation, extensions, and open directions

Weighted projective 2-designs admit an exact reformulation in terms of quantum channels. Consider the depolarizing channel
\[
\Phi_d(X):=\frac{1}{d+1}\big(X+\operatorname{tr}(X)I_d\big).
\]
If \(T(X)=X^T\) is transpose, then the scaled Choi matrix of \(T\circ \Phi_d\) is
\[
C_{T\circ\Phi_d}=\frac{2}{d+1}\Pi^{(2)}.
\]
The entanglement breaking rank \(\operatorname{ebr}(\Psi)\) of a channel \(\Psi\) is the smallest \(n\) for which \(\Psi\) has a rank-1 Kraus decomposition
\[
\Psi(X)=\sum_{k\in[n]} R_k X R_k^*,
\qquad
\sum_{k\in[n]} R_k^*R_k=I_d,
\qquad
\operatorname{rank}(R_k)=1.
\]
The main equivalence is exact: the smallest weighted projective 2-design for \(\mathbb{C}^d\) has size \(\operatorname{ebr}(\Phi_d)\). Moreover,
\[
\operatorname{ebr}(\Phi_d)\ge d^2,
\]
with equality if and only if there exists a tight projective 2-design, and
\[
\operatorname{ebr}(\Phi_d)\le d^2+d
\]
whenever \(d\) is a prime power. Additional bounds include
\[
\operatorname{ebr}(\Phi_d)\le d^2+1 \quad \text{when } d-1 \text{ is a prime power},
\]
and
\[
\operatorname{ebr}(\Phi_d)\le (d+1)^2 \quad \text{for all } d.
\]
This makes minimal weighted 2-design size a channel-theoretic invariant [2101.11756].

The theory also extends beyond complex projective space. Over finite fields, projective 2-designs are defined through tight-frame conditions on \(F_{q^2}^d\) and on the symmetric square \((F_{q^2}^d)^{\otimes 2}_{\mathrm{sym}}\); the analogue of the lower bound remains \(n\ge d^2\), and explicit constructions achieving \(d^2\) points are obtained from Singer difference sets under stated congruence conditions [2101.11756]. In the quaternionic setting, every tight projective 2-design for \(\mathbb{H}^d\) determines an equi-isoclinic tight fusion frame of
\[
d(2d-1)
\]
subspaces of \(\mathbb{R}^{d(2d+1)}\), each of dimension \(3\). Such tight quaternionic 2-designs are known only for \(d\in\{2,3\}\) [2101.11756].

Several open problems remain central. Zauner’s conjecture is equivalent to the statement that the minimal weighted size is always \(d^2\) [2101.11756]. For projective toric designs, the bound
\[
|X|\ge G_{n-1}\!\Big(\big\lfloor\tfrac{t}{2}\big\rfloor\Big)
\]
is conjectured tight for even \(t\) [2311.13479]. In dimension \(6\), exhaustive search shows that no group \(P(T^6)\) 2-design of size \(36\) satisfies the MUB phase condition (4.1), so any hypothetical complete set of MUBs in \(d=6\) must arise from a non-group projective toric design [2311.13479]. The 2025 Sidon-set construction, while nearly tight, cannot reach \(N=d^2\): any group containing a \(d\)-point Sidon set must satisfy \(|G|\ge d^2-d+1\), forcing \(|G|+|S|\ge d^2+1\) [2501.14938].

Weighted state \(t\)-designs are therefore both a rigidity notion and a construction framework. They encode exact Haar-moment identities, interpolate between harmonic analysis and quantum information, support explicit algebraic constructions, and furnish optimal or near-optimal ensembles for tasks such as state tomography, average fidelity estimation, randomized benchmarking, and POVM design [2101.11756][2501.14938].

Source: https://www.emergentmind.com/topics/weighted-state-t-designs