Second-Chern Bounds in Non-Abelian Quantum Geometry

Abstract

We study the quantum geometry of doubly degenerate energy levels in a four-dimensional parameter space. For degenerate pairs with $SU(2)$ gauge structures, the quantum geometry obeys $\big(\textrm{tr }g\big)^{2}/16\geq\sqrt{\det g}\geq|\textrm{Tr}(F\wedge F)|/12$. The first inequality characterizes the anisotropy in the metric. The second determinant inequality measures the self-duality of the curvature under Hodge star operation and the inter-level processes that do not close under the three $SU(2)$ rotations of the doubly degenerate levels. The saturation of the determinant bound induces a quaternion Kähler structure on the four-dimensional parameter space, analogous to the complex structure induced by the ideal-band condition in two-dimensional Chern insulators. As examples, four-band Dirac Hamiltonians automatically saturate the determinant bound and possess a topological zero in $\textrm{Tr}(F\wedge F)$. We discuss the comparison to degenerate pairs with $U(2)$ gauge structures.

Quantum metric and Berry curvature describe how quantum states vary with respect to parameters. They correspond to the symmetric and antisymmetric parts of the quantum geometric tensor (QGT) 31; 5; 2; 44. In two dimensions, the integral of Berry curvature gives the first Chern number, which determines the quantized Hall response and obstructs exponentially localized Wannier functions 38; 11; 7. Meanwhile, quantum metric quantitatively characterizes the localization length of Wannier functions and optical conductivity 23; 33; 35. It can contribute to the superfluid weight and support superconductivity even in a flat band 30; 13; 39. The quantum metric is known to be bounded by the Berry curvature 34; 25. This imposes fundamental relations among the aforementioned physical phenomena 18; 28. Landau levels provide basic ideal examples where the metric and Berry curvature are exactly locked to each other 30; 25. This structure has guided the construction of ideal Chern bands and fractional Chern insulators 8; 15; 42.

The geometric bound between the quantum metric and Berry curvature originates from the semi-positivity of the Abelian QGT 22; 34; 25. Saturation of the bound implies a reciprocal-space Cauchy-Riemann equation for the band structure 8; 25. This exactly corresponds to the descriptions of Landau levels and moiré graphene in the chiral limit 37; 15; 42. Intuitively, the complex structure imposed by the Cauchy-Riemann equation fixes the Abelian Berry curvature to the volume determined by the metric.

The geometry changes qualitatively for degenerate multi-band structures. Adiabatic transport in the parameter space rotates the states within the degenerate occupied subspace, and the QGT becomes matrix-valued 43; 22. This leads to intrinsic non-Abelian quantum geometry whose components cannot be fixed by complex analytic equations. An example is the doubly-degenerate energy level with a $SU(2)$ gauge structures. In four-dimensional (4D) parameter space, the $SU(2)$ gauge structure produces the topological second Chern number that probes matrix rotations in the degenerate doublet. Its basic realization is the Yang monopole, and it underlies the four-dimensional quantum Hall effect and higher-dimensional $SU(2)$ Landau levels 46; 47; 16; 17; 14; 36; 48. The non-Abelian QGTs and their relation to topological invariants have attracted interests recently 9; 26.

In this Letter, we consider a gapped and doubly degenerate band structure and derive geometric bounds for the quantum metric. From Hodge decomposition, we obtain the bounds relating the quantum metric to the second Chern class. Their saturation induces three Cauchy-Riemann equations forming a quaternion algebra. Minimal four-band Dirac projectors saturate the determinant bound while exhibit singular points in the geometry. We also discuss the case of $U(2)$ and clarify the essentiality of $SU(2)$ for the second-Chern bounds.

$SU(2)$ geometry of a degenerate doublet.— Consider a smooth $N$-band Hamiltonian $H(\bm{\lambda})$ in a 4D $\bm{\lambda}$-space $M$. Two degenerate occupied bands $|u_{n}\rangle$ are gapped from other empty bands. We collect these two states as a local frame as $\Psi=(|u_{1}\rangle,|u_{2}\rangle)$ and introduce the projector $P$ and the matrix $X_{\mu}$,

$$P=\Psi\Psi^{\dagger},\quad X_{\mu}\equiv(1-P)\partial_{\mu}\Psi\equiv D_{\mu}\Psi.$$

Here we abbreviate $\partial_{\mu}\equiv\partial_{\lambda^{\mu}}$ and introduce the derivative $D_{\mu}$ tangent to $\Psi$. The matrix $X_{\mu}$ describes the virtual transitions to empty eigenstates $|u_{\ell}\rangle$:

$$(X_{\mu})_{\ell n}\equiv\langle u_{\ell}|\partial_{\mu}u_{n}\rangle=\frac{\langle u_{\ell}|\partial_{\mu}H|u_{n}\rangle}{\varepsilon_{n}-\varepsilon_{\ell}}.$$ (1)

This is the standard interband amplitude used in band geometry and response theory 44; 45; 24; 20. The non-Abelian Berry connection $[A_{\mu}]_{mn}=A_{\mu}^{mn}\equiv\mathrm{i}\langle u_{m}|\partial_{\mu}u_{n}\rangle$ describes the variation of the local basis.

The non-Abelian QGT is given by the Gram matrix:

$$Q_{\mu\nu}=X_{\mu}^{\dagger}X_{\nu}=G_{\mu\nu}-\frac{\mathrm{i}}{2}F_{\mu\nu},$$ (2)

whose symmetric part gives the quantum metric $G_{\mu\nu}=\frac{1}{2}\big(Q_{\mu\nu}+Q_{\nu\mu}\big)$ and the antisymmetric part gives the non-Abelian Berry curvature $F_{\mu\nu}=\mathrm{i}\big(Q_{\mu\nu}-Q_{\nu\mu}\big)$. The component-wise expressions are explicitly given by $[Q_{\mu\nu}]_{nm}=\langle\partial_{\mu}u_{n}|(1-P)|\partial_{\nu}u_{m}\rangle$. Within the Berry connection $A_{\mu}$, the curvature $F_{\mu\nu}$ can be formulated in the conventional non-Abelian curvature form,

$$F_{\mu\nu}=\partial_{\mu}A_{\nu}-\partial_{\nu}A_{\mu}-\mathrm{i}[A_{\mu},A_{\nu}]$$ (3)

From hereby, we introduce $F=\sum_{\mu,\nu}F_{\mu\nu}\,\mathrm{d}\lambda^{\mu}\wedge\mathrm{d}\lambda^{\nu}/2$. We denote the scalar quantum metric as

$$g_{\mu\nu}=\operatorname{Tr}G_{\mu\nu}=\operatorname{Re}\operatorname{Tr}(X_{\mu}^{\dagger}X_{\nu}).$$ (4)

Here $\operatorname{Tr}$ traces over the occupied-band indices. It sums the transition probability from both occupied bands to all empty bands. This measures the total probability transferred out of the occupied doublet by an infinitesimal change of parameters.

As $\bm{\lambda}$ varies, the frame $\Psi$ rotates. When $\bm{\lambda}$ is scanned over a loop in the parameter space $M$, the parallel-transported frame changes by a rotation $\Psi^{\prime}=\Psi U$. We focus on situations where $U\in SU(2)$, i.e., no net phase is produced in parallel transports of $\Psi$ along any loops in $M$. Under this condition, the Abelian part of the Berry curvature vanishes $\textrm{Tr }F_{\mu\nu}=0$. Such situations occur when $|u_{1}\rangle$ and $|u_{2}\rangle$ rotate in opposite phases, for example, when they form a Kramers pair corresponding to an antiunitary symmetry $\Theta^{2}=-1$. As $SU(2)\simeq Sp(1)$, a quaternion structure appears in the Hilbert spaces spanned by $\Psi$. Under this identification, the negative-energy doublet of four-band Dirac Hamiltonians resembles the canonical $\mathbb{H}P^{1}$ Yang monopole 46.

Bounds of the second Chern class.— For the $SU(2)$ case, the curvature can be decomposed into

$$F_{\mu\nu}=\sum_{a}F_{\mu\nu}^{a}\sigma_{a},\quad F^{a}=\frac{1}{2}\sum_{\mu,\nu}F_{\mu\nu}^{a}\,\mathrm{d}\lambda^{\mu}\wedge\mathrm{d}\lambda^{\nu}$$ (5)

where the Pauli matrices $\sigma_{a}$ ($a=1,2,3$) act on the two-dimensional band index. Changing the local basis of the two occupied bands rotates the three $\sigma_{a}$ components among themselves. For traceless curvature $F$, the second Chern class (density) $c_{2}$ is reduced to

$$c_{2} =\frac{1}{8\pi^{2}}\operatorname{Tr}(F\wedge F)=\frac{1}{4\pi^{2}}\sum_{a}F^{a}\wedge F^{a}.$$ (6)

Its integrals give the second Chern number $C_{2}=\int c_{2}\in\mathbb{Z}$. Since $c_{2}$ is a four-form proportional to the volume form, we parameter $c_{2}(\bm{\lambda})=s(\bm{\lambda})|c_{2}(\bm{\lambda})|\mathrm{d}^{4}\lambda$, where $s(\bm{\lambda})=\pm 1$ is the sign of $c_{2}$ relative to the volume form.

The main result of this work is the following inequalities,

$$\frac{1}{16}\big(\textrm{tr }g\big)^{2}\geq\sqrt{\det g}\geq\frac{2\pi^{2}}{3}|c_{2}|.$$ (7)

Here “tr” is for spatial indices $\textrm{tr }g=\sum_{\mu}g_{\mu\mu}$. The first inequality refers to a chosen Cartesian frame of the physical parameters; the volume inequality is invariant under reparametrization. These inequalities compare three levels of information carried by the interband response. The trace of the quantum metric $g_{\mu\nu}$ measures the total strength of the deformation of the occupied space, whereas $\sqrt{\det g}$ measures the four-dimensional volume spanned by four independent deformations. The second Chern class retains only the $SU(2)$ geometric part of this volume.

For a closed parameter space $M$, the integral of Eq. (7) leads to a topological bound

$$\int\frac{d^{4}\lambda}{16}\big(\textrm{tr }g\big)^{2}\geq\int\sqrt{\det g}\ d^{4}\lambda\geq\frac{2\pi^{2}}{3}|C_{2}|.$$ (8)

The bound may thus be viewed as the minimal quantum-geometric cost of carrying a quantized $SU(2)$ instanton charge. In the following, we give a geometric proof for Eq. (7) and explain the saturation conditions.

Hodge decomposition of the curvature.— At every point where $g$ is nondegenerate ($\det g>0$), the metric defines a Hodge star operation $*_{g}$, maping two-forms to two-forms. The norm convention for a given two-form $K$ is defined as $\left\lVert K\right\rVert_{g}^{2}=\sum_{\mu,\nu,\rho,\sigma}K_{\mu\nu}K_{\rho\sigma}g^{\mu\rho}g^{\nu\sigma}/2\geq 0$, where $g^{\mu\nu}\equiv[g^{-1}]^{\mu\nu}$. The two-form $F^{a}$ can be decomposed into self-dual and anti-self-dual sectors,

$$F_{\pm}^{a}=\frac{1}{2}(F^{a}\pm*_{g}F^{a}),\qquad*_{g}F_{\pm}^{a}=\pm F_{\pm}^{a}.$$ (9)

This decomposition reflects the structure $SO(4)\simeq{SU(2)_{+}\times SU(2)_{-}}/{\mathbb{Z}_{2}}$. Geometrically, Hodge star simply pairs complementary 2D planes, $(12)\leftrightarrow(34)$, $(13)\leftrightarrow(42)$, and $(14)\leftrightarrow(23)$. The parity under Hodge star operation generalizes the two-dimensional handedness/chirality to four dimensions.

The Hodge decomposition reorganizes the Chern class into

$$c_{2}=\frac{1}{4\pi^{2}}\left(\sum_{a}\left\lVert F_{+}^{a}\right\rVert_{g}^{2}-\sum_{a}\left\lVert F_{-}^{a}\right\rVert_{g}^{2}\right)\sqrt{\det g}\,\mathrm{d}^{4}\lambda.$$ (10)

This expression gives $c_{2}$ an interpretation as the difference between the two handed Hodge sectors. The value of $|c_{2}|$ is maximized when all $F^{a}$ belong to the same Hodge sector corresponding to the sign of $c_{2}$. With the decomposition Eq. (10), we write the difference between the determinant bound and the Chern class as

$$\Delta=\sqrt{\det g}-\frac{2\pi^{2}}{3}|c_{2}|=\frac{1}{6}(\Delta_{Q}+\Delta_{H})\sqrt{\det g},$$ (11)

where the two contribution are given by

$$\Delta_{H}=2\sum_{a}\left\lVert F_{-s}^{a}\right\rVert_{g}^{2},\quad\Delta_{Q}=6-\sum_{a}\left\lVert F^{a}\right\rVert_{g}^{2},$$ (12)

We will show that both contributions are non-negative $\Delta_{H}\geq 0,\Delta_{Q}\geq 0$. The determinant inequality in Eq. (7) follows. The saturation is reached when $\Delta_{H}=\Delta_{Q}=0$.

The Hodge contribution $\Delta_{H}$ provides a local measure of the deviation from the Bogomolny self-duality condition in four-dimensional Yang-Mills theory 4; 6, which is explicitly non-negative. It is the Berry-curvature weight with the handedness opposite to the local sign of $c_{2}$. It is saturated when

$$*_{g}F^{a}=sF^{a}\ (a=1,2,3)\Leftrightarrow\Delta_{H}=0.$$ (13)

The $\Delta_{Q}$ part measures the failure of the four-dimensional tangent image to preserve the Pauli rotations of its tangent vectors. To demonstrate this, we introduce the $(1,1)$-form $[J^{a}]^{\mu}_{\nu}=\sum_{\rho}g^{\mu\rho}F^{a}_{\rho\nu}$ and its conjugate $[(J^{a})^{\dagger}]^{\mu}_{\nu}=\sum_{\rho,\sigma}g^{\mu\rho}[J^{a}]^{\sigma}_{\rho}g_{\sigma\nu}$. We have the relation $\left\lVert F^{a}\right\rVert_{g}^{2}=\textrm{tr}[(J^{a})^{\dagger}J^{a}]/2$. We treat the QGT as a quaternion-valued bilinear form in the tangent space of M, where $g$ is the scalar component and $F^{a}$ are the coefficients of the three quaternions. The action of $Q$ on four dimensional tangent vectors $v,v^{\prime}\in TM$ is done by contracting the $\mu,\nu$ indices of $v,v^{\prime}$. We have $|Q(v,v^{\prime})|_{\mathbb{H}}\leq\sqrt{g(v,v)g(v^{\prime},v^{\prime})}$ from the quaternion Cauchy-Schwarz inequality. On the other hand, by definition, we have $g(v,J^{a}v^{\prime})=F^{a}(v,v^{\prime})$. As $F^{a}(v,v^{\prime})$ is just a summand in the quaternion $Q(v,v^{\prime})$, we have $|g(v,J^{a}v^{\prime})|\leq\sqrt{g(v,v)g(v^{\prime},v^{\prime})}$. Taking $v=J^{a}v‘$, this shows that $g(J^{a}v^{\prime},J^{a}v^{\prime})\leq g(v^{\prime},v^{\prime})$. The $J^{a}$ makes the length of vectors smaller. This demonstrates $||(J^{a})^{\dagger}J^{a}||\leq 1$ (operator norm). Tracing over all spatial indices, we have $\left\lVert F^{a}\right\rVert_{g}^{2}=\textrm{tr}[(J^{a})^{\dagger}J^{a}]/2\leq 2$. Summing over all $a=1,2,3$, this yields $\Delta_{Q}\geq 0$.

Equality and its physical meaning.— Now we explore when the determinant inequality is saturated, i.e, $\Delta=0$. The $\Delta_{H}=0$ part has been shown before. We focus on when $\Delta_{Q}=0$. This requires $\textrm{tr}[1-(J^{a})^{\dagger}J^{a}]=0$. As $1-(J^{a})^{\dagger}J^{a}$ is semi-positive definite, this in fact requires $(J^{a})^{\dagger}J^{a}=1$. Since $(J^{a})^{\dagger}=-J^{a}$, this requires $[J^{a}]^{2}=-1$, a complex structure on the tangent space of $M$.

On the other hand, the condition $\Delta_{Q}=0$ requires $F^{a}(J^{a}v,v)=g(J^{a}v,J^{a}v)=g(v,v)$. The geometric structure for $v$ originates from the tangent derivatives $D_{v}\Psi$. By further analyzing the quaternion Cauchy-Schwarz inequality, we find the following conditions at $\det g(\bm{\lambda)}>0$ points 1:

$$\sum_{\nu}[J^{a}]^{\nu}_{\mu}D_{\nu}\Psi+D_{\mu}\Psi(-i\sigma_{a})=0,$$ (14)

where $J^{a}$ satisfies the quaternion algebra: $J^{1}J^{2}=J^{3},(J^{a})^{2}=-1_{4\times 4}$. This equation means that a quaternion action $(-i\sigma_{a})$ on the doublet isospin space commutes with an quaternion action $J^{a}$ in the tangent space on the manifold $M$. This condition is in fact a triple Cauchy-Riemann equation. In two dimensions, if we set $J_{a}^{b}=-\epsilon^{ab}$ where $a,b=1,2$, the Cauchy-Riemann equation takes the form:

$$\sum_{b}J^{b}_{a}\partial_{b}|\psi\rangle+\partial_{a}|\psi\rangle i=0,\ \textrm{2D Cauchy Riemann}.$$ (15)

Comparing the two equations, Eq. (14) corresponds exactly to three Cauchy-Riemann equations related by $SO(3)$ rotations in the quaternion space. It is the four-dimensional analogue of the complex closure $J^{2}=-1$ at the ideal point of an Abelian Chern band 12.

The preceding demonstration proves $\Delta(\bm{\lambda})\geq 0$ for $\det g(\bm{\lambda})>0$. When $\det g(\bm{\lambda})=0$, the amplitudes $X_{\mu}$ are linear-dependent. The vector space spanned by them has dimensions lower than four. All four forms on this space have to vanish and so does $c_{2}$, which is an exterior power based on $X_{\mu}$. The determinant inequality simply becomes 0 on both sides.

Four-band Hamiltonians.— We demonstrate two facts in four-band Dirac Hamiltonians: (1) All such four-band models saturate the determinant bound; (2) For $M$ being a 4D torus, there exists $\bm{\lambda}$ such that $\det g(\bm{\lambda})=\textrm{Tr}[F(\bm{\lambda})\wedge F(\bm{\lambda})]=0$. It is impossibible build a topological four-band model with constant second Chern density.

After neglecting an overall constant, four-band Dirac Hamiltonians are parametrized by five real coefficients $d_{A}\in\mathbb{R}$ in front of $4\times 4$ Dirac Gamma matrices $\Gamma_{A}$ 46; 47; 32

$$H(\bm{\lambda})=\sum^{5}_{A=1}d_{A}(\bm{\lambda})\Gamma_{A},$$ (16)

The quantum geometric tensor and the projector of the two negative-energy bands are uniquely expressed by the normalized vector $\hat{\bm{d}}$:

$$g_{\mu\nu}=\frac{1}{2}\partial_{\mu}\hat{\bm{d}}\cdot\partial_{\nu}\hat{\bm{d}},\qquad P=\frac{1}{2}\left(I_{4}-\sum_{A}\hat{d}_{A}\Gamma_{A}\right).$$ (17)

while the second-Chern density is given by 32; 26

$$\operatorname{Tr}(F\wedge F)=\pm\frac{1}{8}\epsilon_{ABCDE}\hat{d}_{A}d\hat{d}_{B}\wedge d\hat{d}_{C}\wedge d\hat{d}_{D}\wedge d\hat{d}_{E}.$$ (18)

The orthogonal projector uniquely corresponds to the sub-Hilbert space spanned by the states $|u_{1}\rangle,|u_{2}\rangle$. As a result, the sub-Hilbert space spanned by $|u_{1}\rangle,|u_{2}\rangle$ is characterized by the four dimensional sphere $\hat{\bm{d}}\in S^{4}$. This is the four-band analog of the two-band Bloch sphere $\mathbb{C}P^{1}\simeq S^{2}$, by replacing $\mathbb{H}P^{1}\simeq S^{4}$.

Now we prove the two statements. At points $\bm{\lambda}$ where $\det g(\bm{\lambda})>0$, the tangent space $TM$ is isomorphically mapped to the tangent space at $P(\bm{\lambda})$ on $S^{4}$. So the image $dP\ TM_{\bm{\lambda}}$ is invariant under all quaternion actions $J_{a}$ in the tangent space of $S^{4}$, $\sqrt{\det g}=|\textrm{Tr }F\wedge F|/12$. At $\bm{\lambda}$ where $\det g(\bm{\lambda})=0$, the previous results show that both sides are equal to zero. The determinant bound is thus always saturated. Assume that $\det g>0$ for all $\bm{\lambda}$, this tells the tangent map at every $\bm{\lambda}$ is an isomorphism, leading to a local diffeomorphism. A local diffeomorphism between two compact spaces is a covering map. This would map the fundamental group $\pi_{1}(T^{4})=(\mathbb{Z})^{4}$ injectively into $\pi_{1}(S^{4})=0$, which is impossible. Thus, there must be $\bm{\lambda}$ where $\det g(\bm{\lambda})=\textrm{Tr }F(\bm{\lambda})\wedge F(\bm{\lambda})=0$.

These two properties are reminiscent of the previous results on 2D two-band models. In 2D two-band Hamiltonians, it is known that the Abelian Berry curvature is always saturates the determinant bound $\sqrt{\det g}=|F|/2$, and there exist points in the Brillouin zone where the Berry curvature vanishes 25; 41. This obstructs the possibility of realizing constant Berry curvature in two-band topological models, a condition suggested to favor fractional Chern insulators 10; 29. The statements presented here are the 4D counterparts.

Situations for $U(2)$ doublet.— We comment that the $SU(2)$ structure is essential for the topological bound from second Chern number. A generic degenerate doublet $\Psi$ transforms by a $U(2)$ rotation when parallel transported along a parameter loop. The Abelian Berry curvature $\textrm{Tr }F$ is non-vanishing. The second Chern class receives one more contribution 40:

$$c_{2} =\frac{1}{8\pi^{2}}\left[\operatorname{Tr}(F\wedge F)-(\operatorname{Tr}F)\wedge(\operatorname{Tr}F)\right].$$ (19)

The second term may change its sign relative to the first term at different $\bm{\lambda}$. Meanwhile, the metric also receives modification in the band-index parts. We cannot directly compare $\det g$ with the second Chern number.

On the other hand, the Wirtinger inequality yields a bound on Abelian curvature for $U(2)$ doublets. Note that the first Chern classes $c_{1}=\textrm{Tr }F/(2\pi)$ may be non-vanishing. Besides the second Chern number, there is another 4D Chern number $C_{1,1}=\int c_{1}\wedge c_{1}\in\mathbb{Z}$ 27. The Wirtinger inequality asserts 3:

$$\int\sqrt{\det g}\ d^{4}\lambda\geq\Big|\int\frac{1}{8}(\operatorname{Tr}F)\wedge(\operatorname{Tr}F)\Big|=\frac{\pi^{2}}{2}|C_{1,1}|.$$ (20)

Thus, we see that for $U(2)$ the geometric bound in 4D is on the first Chern class instead of the second Chern class 26. An intuitive way to interpret this difference is as follows. With $U(2)$, we can only employ the complex structure to fix the Abelain quantum geometry, while $SU(2)\simeq Sp(1)$ supplements a quaternion structure that fixes the non-Abelian components in the curvature.

A minimal saturation example is a three-level Hamiltonian on $M=\mathbb{C}P^{2}$, locally parameterized by  21,

$$H=2|v\rangle\langle v|-I_{3},\quad|v\rangle =\frac{(1,w_{1},w_{2})^{T}}{\sqrt{1+\left\lvert w_{1}\right\rvert^{2}+\left\lvert w_{2}\right\rvert^{2}}},$$ (21)

The two complex numbers $w_{1},w_{2}$ give four real parameters. The state $|v\rangle$ is the empty level at energy $+1$, while the orthogonal two-dimensional subspace consists of two degenerate occupied levels at energy $-1$. The global parameter space of the normalized three-component vector modulo its phase is denoted $\mathbb{C}P^{2}$. All $(X_{\mu})_{\ell n}$ amplitudes pass to the single empty level, and Eq. (20) is saturated locally 21:

$$c_{2}=-c_{1}\wedge c_{1},\qquad\sqrt{\det g}=\frac{\pi^{2}}{2}\left\lvert c_{1}^{2}\right\rvert=\frac{\pi^{2}}{2}\left\lvert c_{2}\right\rvert.$$ (22)

Discussion.— The geometric bounds place constraints on non-Abelian quantum geometry, which characterizes interband polarization amplitudes. It does not assume flat bands, weak coupling, or a particular lattice model. Unlike the two dimensional geometric bound whose saturation selects a chirality in the wavefunctions, saturation of the four dimensional bounds selects a Hodge sector. The metric must be fully converted into a single-handed set of relative $SU(2)$ rotations without spectator transition channel.

This aspect is useful for experiments and numerics. Quantum metrics can be extracted from excitation rates under weak parameter modulation, while non-Abelian curvature can be reconstructed from state tomography or Wilson-loop protocols 14; 9; 36. The two contributions in the determinant inequality correspond to different physical processes. The contribution $\Delta_{H}$ detects cancellation between complementary parameter loops. In comparison, the contribution $\Delta_{Q}$ detects leakage into non-adiabatic interband channels. The latter should be especially sensitive to remote-band mixing.

The determinant bound can be saturated by an anisotropic QGT, while the trace bound requires an isotropy. This might lead to differences when the geometry is defined with respect to pumping parameters $\bm{\lambda}=(k_{x},k_{y},k_{z},\phi)$. For example. the relative weight assigned to the phase $\phi$ must be fixed before the metric trace has a physical meaning. Note that saturation of the bounds at every parameter point does not immediately yield the topological bounds. Local geometric bounds test the closure of virtual interband transitions at each parameter. Saturation of the topological bound further requires a fixed sign of the second-Chern density. Neither condition by itself implies a flat dispersion or a correlated many-body phase. They instead identify the band geometry that minimizes the wave-function volume needed to carry a given $SU(2)$ topological density. This provides a starting point for extending ideal-band ideas from Chern bands to four-parameter non-Abelian systems.

Note added.— While this manuscript was being prepared, Ref. 19 appeared, developing a quaternion-Kähler formulation of quantum geometry for Kramers pairs. That work derives an equivalent second-Chern quantum-volume bound and characterizes its saturation through quaternionic tangent-space closure, including the automatic saturation of minimal quaternionic four-band models. The present work derives the gap between the quantum volume and second Chern class from the Hodge decomposition and considers the difference to $U(2)$ cases.

Acknowledgments.— C.W. is supported by the National Natural Science Foundation of China under Grants No. 12234016 and No. 12174317. Z.P. is supported by the National Natural Science Foundation of China under Grant No. 12504219. KY thanks Bruno Mera for pointing out the Wirtinger inequality and is supported by a start-up grant from Westlake University. This work has been supported by the New Cornerstone Science Foundation.