Symplectic Bootstrap Analysis

• Symplectic bootstrap is a technique that adapts conformal bootstrap methods to Anderson localization in the symplectic symmetry class by enforcing Δϕ=0 and replica degeneracy.
• It employs truncated determinant methods to solve crossing-symmetry constraints and extract scaling dimensions, providing estimates for the localization exponent ν.
• Quantitative analyses using various truncations yield ν values that align well with numerical data, notably reproducing accurate estimates in two dimensions.

Symplectic bootstrap designates the application of the conformal bootstrap to the symplectic Anderson localization problem by treating localization as a non-unitary critical phenomenon in a conformal-field-theory-like setting and extracting scaling dimensions from truncated crossing-symmetry constraints via the determinant, or minor, method. In "Conformal Bootstrap Analysis for Localization: Symplectic Case" (Hikami, 2018), the central objective is to determine whether this construction can reproduce or approximate the critical exponents of symplectic localization in arbitrary dimension $D$, especially the localization exponent $\nu$, by encoding the localization critical point as a constrained bootstrap problem, imposing localization-specific scaling conditions, truncating the operator spectrum to a small set of low-lying operators, and solving for scaling dimensions from vanishing minors.

1. Symplectic localization as the target critical phenomenon

The physical target is the symplectic symmetry class of Anderson localization. In the formulation used here, time-reversal symmetry is preserved, spatial inversion symmetry is broken, and strong spin-orbit coupling is the characteristic physical realization. This class is distinguished from the orthogonal and unitary classes by the existence of a metal-insulator transition even in two dimensions (Hikami, 2018).

The field-theoretic background is the nonlinear $\sigma$ model on a symmetric space. In replica language, the symplectic case is described through a symplectic Grassmannian structure with the correspondence

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$

In the localization limit, the parameter choice emphasized is

$p=N=0,$

together with a sign change in the coupling,

$t\to -t,$

associated with non-compactness. The same analysis also allows an alternative reading in terms of the $Sp(N)/U(N)$ nonlinear $\sigma$ model in the $N\to 0$ limit, so the symmetry identification is explicitly not claimed to be absolutely unique.

Within this framework, the bootstrap is not derived from an explicit microscopic Lagrangian. Instead, it is used as a nonperturbative consistency method for a non-unitary critical point. The problem is therefore posed in a form analogous to earlier non-unitary bootstrap applications to the Yang-Lee edge singularity, polymers, branched polymers, and random-field Ising models, but with constraints specific to localization.

2. Localization-specific bootstrap constraints

The defining step in the symplectic bootstrap is the imposition of two scaling conditions that encode localization physics. The starting scaling law is

$\beta = \nu (D - 2 + \eta),$

with field dimension

$\nu$0

For the localization problem, the density of states has no singularity, implying

$\nu$1

The bootstrap analysis therefore imposes

$\nu$2

in any dimension $\nu$3 (Hikami, 2018).

A second condition comes from replica-limit degeneracy:

$\nu$4

where $\nu$5 is the energy operator scaling dimension and $\nu$6 is the crossover scaling dimension. The bootstrap is thus performed under the combined constraints

$\nu$7

These two conditions are the defining localization assumptions of the construction. They reduce the space of allowed scaling dimensions and make it possible to search for isolated intersections of determinant zero-loci in the $\nu$8 plane. A plausible implication is that the method is less a generic conformal bootstrap than a localization-constrained conformal bootstrap: crossing symmetry is not used in isolation, but is supplemented by phenomenological input tied to the density of states and replica structure.

3. Crossing symmetry, conformal blocks, and Casimir structure

The bootstrap problem begins from the four-point function of a scalar field $\nu$9,

$\sigma$0

The cross ratios are introduced as

$\sigma$1

and the reduced amplitude is expanded in conformal blocks,

$\sigma$2

Crossing symmetry under $\sigma$3 gives the constraint

$\sigma$4

This is the fundamental consistency equation from which the spectrum is constrained (Hikami, 2018).

The conformal blocks are eigenfunctions of the quadratic Casimir differential operator,

$\sigma$5

with eigenvalue

$\sigma$6

At the symmetric point $\sigma$7, the spin-zero block is written explicitly as

$\sigma$8

The analysis is performed around

$\sigma$9

using variables $$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$0 defined by

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$1

These choices are standard for the determinant approach: derivative data at the symmetric point furnish a finite-dimensional surrogate for the full infinite bootstrap system. In the symplectic case, that surrogate is then combined with the localization constraints $$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$2 and $$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$3.

4. Truncated determinants and the minor method

Because the full bootstrap contains infinitely many exchanged operators, the spectrum is truncated and the small-determinant method is applied. The derivative data are

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$4

These satisfy one inhomogeneous equation,

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$5

and homogeneous equations,

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$6

A nontrivial solution exists only when the relevant determinants, or minors, vanish (Hikami, 2018).

The analysis studies minors $$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$7 defined schematically by

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$8

The derivative-label dictionary is

$$Sp(N)/Sp(p)\times Sp(N-p)= O(-2N)/O(-2p)\times O(-2N+2p).$$9

The zero-loci of these minors are not independent. For a $p=N=0,$0 matrix, Plücker relations are given by

$p=N=0,$1

and

$p=N=0,$2

These relations constrain how multiple determinant conditions can intersect.

The truncation itself is organized by retained operator content. In the $p=N=0,$3 truncation, the operators are the scalar energy-like operator $p=N=0,$4, the crossover operator $p=N=0,$5, and a spin-2 stress-tensor-like operator with dimension $p=N=0,$6. In the $p=N=0,$7 truncation, a spin-4 operator $p=N=0,$8 is added. The representative $p=N=0,$9 determinant is

$t\to -t,$0

Including higher-spin operators increases the number of unknown scaling dimensions, so higher truncations are not parameter-free.

5. Quantitative estimates for scaling dimensions and $t\to -t,$1

With $t\to -t,$2 and $t\to -t,$3, the $t\to -t,$4 analysis identifies intersection points of minor zero-loci in the $t\to -t,$5 plane. The relation used for the localization exponent is

$t\to -t,$6

For the simplest truncation, the reported values include $t\to -t,$7, where $t\to -t,$8 gives $t\to -t,$9; $Sp(N)/U(N)$0, where $Sp(N)/U(N)$1 gives $Sp(N)/U(N)$2; and $Sp(N)/U(N)$3, where $Sp(N)/U(N)$4. Near $Sp(N)/U(N)$5, no solution appears. The disappearance of solutions near $Sp(N)/U(N)$6 is interpreted as indicating a lower critical dimension around that value in the simplest approximation (Hikami, 2018).

The $Sp(N)/U(N)$7 truncation improves agreement with numerics, especially in $Sp(N)/U(N)$8. For $Sp(N)/U(N)$9, an appropriate choice of the spin-4 parameter $\sigma$0 yields either $\sigma$1, with $\sigma$2 and hence $\sigma$3, or $\sigma$4, with $\sigma$5 and hence $\sigma$6. The value $\sigma$7 is highlighted as very close to the numerical finite-size scaling estimate

$\sigma$8

for symplectic localization in two dimensions. For $\sigma$9, one finds approximately $N\to 0$0 and $N\to 0$1. The table of $N\to 0$2 estimates also shows $N\to 0$3 at $N\to 0$4 and $N\to 0$5 at $N\to 0$6.

Dimension and truncation	Scaling data	Exponent estimate
$N\to 0$7, $N\to 0$8	$N\to 0$9	$\beta = \nu (D - 2 + \eta),$0
$\beta = \nu (D - 2 + \eta),$1, $\beta = \nu (D - 2 + \eta),$2	$\beta = \nu (D - 2 + \eta),$3	$\beta = \nu (D - 2 + \eta),$4
$\beta = \nu (D - 2 + \eta),$5, $\beta = \nu (D - 2 + \eta),$6	$\beta = \nu (D - 2 + \eta),$7	not stated
$\beta = \nu (D - 2 + \eta),$8, $\beta = \nu (D - 2 + \eta),$9, $\nu$00	$\nu$01	$\nu$02
$\nu$03, $\nu$04, $\nu$05	$\nu$06	$\nu$07
$\nu$08, $\nu$09	$\nu$10	$\nu$11
$\nu$12, $\nu$13	not stated	$\nu$14
$\nu$15, $\nu$16	not stated	$\nu$17

These estimates are described as somewhat rough in the $\nu$18 case and improved, but still approximate, in the $\nu$19 case. The strongest success is therefore the two-dimensional symplectic transition, where the higher truncation reproduces the numerical scale of $\nu$20 with notable accuracy.

6. Interpretation, ambiguities, and relation to the broader bootstrap program

The principal conclusion is that conformal bootstrap can be adapted to localization by imposing $\nu$21 and $\nu$22, and that the symplectic Anderson localization class is the natural target because it has a genuine transition in $\nu$23 (Hikami, 2018). Small determinant truncations capture the correct order of magnitude of the localization exponent $\nu$24, while the $\nu$25 truncation provides a notably good estimate in two dimensions. At the same time, the method is explicitly exploratory rather than exact: fusion rules are not known for the localization problem, the operator spectrum is truncated aggressively, higher-spin dimensions introduce extra unknown parameters, and higher-truncation results depend on parameter choices. For that reason, the solutions are best viewed as semi-quantitative approximations.

Two unresolved structural issues are emphasized. First, the lower critical dimension suggested by the simplest truncation, around $\nu$26, is not claimed as definitive. Second, the upper critical dimension remains unresolved. Around $\nu$27, the bootstrap loci do not collapse in the way one might expect at a free-field upper critical dimension, so the results do not support $\nu$28 as the upper critical dimension; the possibility $\nu$29 remains open and is not excluded.

The universality-class identification is also presented cautiously. The same bootstrap structure may describe a different universality class, such as quantum Hall criticality or the $\nu$30 sigma model in the replica limit. This suggests that the determinant solution space may encode broader non-unitary localization-like critical behavior than a single unequivocal fixed point.

In the wider bootstrap landscape, the symplectic construction belongs to a family of approaches that strengthen crossing by adding extra structural information beyond generic bosonic consistency. A distinct example is the long-multiplet superconformal bootstrap, where one bootstraps the full superfield and uses superdescendant relations to obtain stronger constraints than those available from superprimaries alone (Cornagliotto et al., 2017). The comparison is methodological rather than substantive: the symplectic case imposes localization-specific conditions such as $\nu$31 and replica degeneracy, whereas the long-multiplet program exploits superspace nilpotent invariants and Casimir equations for full supermultiplets. A plausible implication is that "symplectic bootstrap" is best understood not as a separate formalism with autonomous axioms, but as a problem-specific instantiation of the broader bootstrap program in which the decisive input comes from the physics of localization.