Local-Variance Asymmetry: Definitions and Applications

• Local-variance asymmetry is defined as the unequal behavior of locally measured variance-like quantities, reflecting directional or structural imbalances.
• It manifests in multiple domains, from fractional uncertainty and deformed interference (showing skewed local moments) to variance dipoles in CMB and imbalanced hydrogen bonds in water.
• Empirical and theoretical studies leverage these asymmetries to probe underlying physics, revealing anisotropies and nonlocal effects across various scientific disciplines.

Local-variance asymmetry denotes an asymmetry in a locally defined variance-like quantity, but the term is not tied to a single universal formalism. In the recent literature it appears in at least four technically distinct settings. In fractional uncertainty, it names the structural asymmetry between a nonlocal “position” form and a fractional energy that has a clear local limit (1803.02384). In interference theory, it denotes a left-right asymmetry of local statistical moments around a fringe maximum, produced by a deformation of the Born rule that preserves fringe positions and quadratic curvature while generating a cubic skewness (Zhang, 19 Mar 2026). In cosmology, it usually denotes directional variation of locally estimated variance or power on the sky, encoded in variance maps and their dipoles in CMB temperature or polarization data (Akrami et al., 2014). In molecular physics, it describes the instantaneous imbalance between the two donor and/or the two acceptor hydrogen bonds of a given water molecule, quantified through asymmetry parameters built from charge-transfer energies (Elgabarty et al., 2020).

1. Principal meanings and formal scope

Across the cited literature, the phrase refers either to a variance map on a domain, to an asymmetry of local moments around a point, or to a structural nonequivalence between two variance-like functionals. The shared feature is not a single estimator, but the appearance of locally defined second-moment or variance-analog objects whose behavior is directionally, spatially, or structurally unequal.

Domain	Local quantity	Asymmetry
Fractional uncertainty	$I_1(\varphi,s)$, $I_2(\varphi,s)$	nonlocal position form vs local-limit energy
Quantum interference	local moments around a fringe maximum	left-right cubic skewness
CMB analysis	patchwise variance map $V(\hat n)$	dipole or higher multipoles on the sky
Low-$z$ cosmography	hemispherical $H_0$ fits	directional $\delta H_0$
Liquid water	donor/acceptor HB strengths	one strong and one weak “leg”

This suggests a family of related constructions rather than a single invariant concept. In some cases the asymmetry is between opposite hemispheres or opposite sides of a local maximum; in others it is between two formally different “variance-like” quantities, or between two bonds attached to the same microscopic unit.

2. Fractional uncertainty and structural asymmetry

In the fractional uncertainty inequality, the relevant pair of quadratic forms is

$$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$

$$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$

for $0<s<\tfrac12$, with $\|\varphi\|_{L^2(\mathbb R)}=1$, and the main result is

$I_2(\varphi,s)$0

Here $I_2(\varphi,s)$1 is the fractional energy quadratic form of order $I_2(\varphi,s)$2, while $I_2(\varphi,s)$3 is introduced as a nonlocal substitute for the position quadratic form (1803.02384).

The asymmetry arises because the two factors are not of the same structural type. The “frequency-side” object $I_2(\varphi,s)$4 has the standard interpretation

$I_2(\varphi,s)$5

and, in the limit $I_2(\varphi,s)$6, recovers the local energy $I_2(\varphi,s)$7, which the paper identifies with the local form of $I_2(\varphi,s)$8. By contrast, the “position-side” object $I_2(\varphi,s)$9 is not a second moment; it is a double integral involving $V(\hat n)$0 and has no simple local representation like $V(\hat n)$1 (1803.02384).

The paper makes this distinction precise through a dyadic model on $V(\hat n)$2 with dyadic metric $V(\hat n)$3, dyadic energy $V(\hat n)$4, and dyadic position form $V(\hat n)$5. On Haar functions $V(\hat n)$6,

$V(\hat n)$7

so their product is independent of $V(\hat n)$8, and the dyadic inequality

$V(\hat n)$9

follows from orthogonality and Cauchy–Schwarz. In the Euclidean setting this yields

$z$0

The key interpretive point is that the frequency side has a clear local limit, whereas the position side remains intrinsically nonlocal; this is precisely the “fractional local-variance asymmetry” identified in the paper (1803.02384).

The limiting behavior reinforces that asymmetry. As $z$1, $z$2; as $z$3, $z$4. Formally, as $z$5, $z$6, while the continuum analogue of $z$7 is only motivated through the scaling $z$8, which matches the $z$9-scaling of classical variance only at $H_0$0 (1803.02384).

3. Local moments and interference asymmetry

In the interference setting, local-variance asymmetry is tied to a deformation of the Born rule rather than to a variance map. The model starts from

$H_0$1

with $H_0$2 and

$H_0$3

The probability density is defined as

$H_0$4

The Schrödinger equation remains linear in $H_0$5, probability conservation follows from the continuity equation, and the deformation is carried entirely by the map from $H_0$6 to $H_0$7 (Zhang, 19 Mar 2026).

For a two-path superposition

$H_0$8

the undeformed interference profile near a bright fringe maximum is locally even in the phase deviation $H_0$9: $\delta H_0$0 With the deformed rule, the first-order correction near a bright fringe contains a cubic term,

$\delta H_0$1

Accordingly,

$\delta H_0$2

There is no linear term, so the fringe position is unchanged; there is no quadratic $\delta H_0$3-correction at this order, so the curvature is unchanged; the leading shape change is odd and cubic, so the local profile becomes left-right asymmetric (Zhang, 19 Mar 2026).

The paper formalizes the effect through the normalized third central moment

$\delta H_0$4

computed in a narrow region around a single bright fringe. In standard quantum mechanics, $\delta H_0$5. In the deformed theory,

$\delta H_0$6

The denominator is the variance to the power $\delta H_0$7, so the formal observable is skewness rather than variance itself. The paper nevertheless states that once the cubic term is present, local variances restricted to $\delta H_0$8 and $\delta H_0$9 become different, $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$0. In that sense, local-variance asymmetry is not an independent primitive but a side-dependent consequence of the same cubic distortion that generates nonzero $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$1 (Zhang, 19 Mar 2026).

A recurrent misconception is that such an effect should shift fringes or broaden them. The paper argues the opposite: the positions of maxima and the quadratic curvature are protected, while the leading modification is a cubic skewness. It also argues that conventional phase noise, path-length fluctuations, and finite detector resolution generically produce symmetric broadening or random shifts, not a systematic cubic term that survives averaging (Zhang, 19 Mar 2026).

4. CMB variance maps, hemispherical asymmetry, and morphological extensions

In CMB analysis, local-variance asymmetry usually means that the temperature or polarization variance estimated on sky patches is not statistically uniform over the sphere. A local-variance map is built by assigning to each patch center the sample variance of the field inside a hemisphere or disc, and the asymmetry is then quantified by the dipole or low-$$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$2 multipoles of that map (Bernui et al., 2014).

A widely used construction places discs on an $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$3 HEALPix grid, computes the local variance in each disc, subtracts the simulation mean field, and fits a dipole to the resulting variance map. Applied to Planck 2013 SMICA, none of the 1000 isotropic FFP6 simulations had a larger variance dipole amplitude than the data for $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$4, implying a significance of at least $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$5, with a preferred direction $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$6 for $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$7 (Akrami et al., 2014). A related hemisphere-based estimator on Planck foreground-cleaned maps found a variance-map dipole pointing to $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$8 and a maximal statistical significance of $$I_1(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{2s-1}|\varphi(x)|\,|\varphi(y)|\,dx\,dy,$$9 CL in the scales ranging from $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$0 to $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$1; notably, removing the quadrupole and octopole made the asymmetry stronger rather than weaker (Bernui et al., 2014).

The temperature variance asymmetry is scale dependent. Extending the disc analysis to smaller scales and including the CMB Doppler dipole showed that, after removing large-scale features up to $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$2, local variance maps can measure the Doppler dipole in Planck 143 and 217 GHz channel maps at about $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$3. At these small scales, there is no power asymmetry in the direction of the anomalous large-scale power asymmetry beyond that expected from cosmic variance, while at large scales the hemispherical power asymmetry remains at at least $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$4 (Adhikari, 2014).

The same local-variance methodology has been applied to Planck 2015 polarization. On filtered E-mode maps, the fitted local-variance dipole amplitudes are in the range $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$5, depending on the galactic mask and disc radius, and the preferred direction lies broadly toward the CMB kinetic dipole direction. The signal is significant on large and intermediate scales but becomes consistent with isotropy for $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$6; the paper therefore treats it as tentative evidence, explicitly noting the possibility of systematics in Planck polarization data (Aluri et al., 2017).

Recent work has broadened the object of interest beyond variance alone. A stripe-based analysis of weighted local variance maps aligned to the most prominent dipole direction found that some higher multipoles of the variance field can have significance comparable to the dipole, depending on smoothing scale: for $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$7 caps the dipole and the $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$8 mode both have $$I_2(\varphi,s)=\iint_{\mathbb R^2}|x-y|^{-2s-1}|\varphi(x)-\varphi(y)|^2\,dx\,dy,$$9, while for $0<s<\tfrac12$0 caps the $0<s<\tfrac12$1 mode has $0<s<\tfrac12$2, more significant than the dipole. The same paper emphasized that the interpretation depends on the look-elsewhere effect: the dipole remains unusual if one privileges dipolar asymmetry, but the total number of outlying multipoles up to $0<s<\tfrac12$3 is not anomalous when all are treated symmetrically (Jamshidi et al., 2024).

A still broader generalization replaces patch variance by local morphological descriptors. Using the three Minkowski Functionals on $0<s<\tfrac12$4 discs in the Planck SMICA map, one can fit the local MF curves to the Gaussian-isotropic expectations and extract not only a local estimate of the field variance $0<s<\tfrac12$5 but also a local estimate of the gradient variance $0<s<\tfrac12$6 and a local goodness-of-fit statistic. In that framework the variance dipole remains highly significant, with p-value $0<s<\tfrac12$7, but there is also a moderately significant gradient-variance dipole, with p-value $0<s<\tfrac12$8, and a mild dipole in the goodness-of-fit to Gaussian-isotropic predictions, with p-value $0<s<\tfrac12$9. The Planck dipoles for all three quantities point toward the same region of the sky, and the paper concludes that the hemispherical asymmetry extends beyond local variance into local morphology (Duque et al., 23 Mar 2026).

5. Cosmological extensions: parity, inflationary modeling, and local $\|\varphi\|_{L^2(\mathbb R)}=1$0

The CMB literature contains several nearby constructions that do not use the standard variance-map estimator but address related asymmetries in local power. A pixel-domain parity analysis decomposes the temperature field into antipodally symmetric and antisymmetric parts,

$\|\varphi\|_{L^2(\mathbb R)}=1$1

and defines the local statistic

$\|\varphi\|_{L^2(\mathbb R)}=1$2

Large negative $\|\varphi\|_{L^2(\mathbb R)}=1$3 identifies odd-parity-dominated local regions, while large positive $\|\varphi\|_{L^2(\mathbb R)}=1$4 identifies even-parity-dominated ones. The paper finds that the global parity asymmetry is mainly associated with a deficit of symmetric peaks rather than an excess of asymmetric ones, and that one of the strongest local peaks, at $\|\varphi\|_{L^2(\mathbb R)}=1$5, lies close to the dipole-modulation direction of the power spectrum (Creswell et al., 2021). This suggests a parity-resolved analogue of local-variance asymmetry.

Inflationary model building has also been tied directly to local-variance observables. In a model where translational invariance is broken during inflation by a primordial domain wall, the wall induces an anisotropic correction to the curvature perturbation power spectrum and therefore to the variance on the CMB sphere. If the CMB sphere is centered at distance $\|\varphi\|_{L^2(\mathbb R)}=1$6 from the wall, with radius $\|\varphi\|_{L^2(\mathbb R)}=1$7 and $\|\varphi\|_{L^2(\mathbb R)}=1$8, the variance correction takes the form

$\|\varphi\|_{L^2(\mathbb R)}=1$9

which can be expanded in Legendre multipoles $I_2(\varphi,s)$00. For $I_2(\varphi,s)$01,

$I_2(\varphi,s)$02

so the dipole dominates over higher multipoles. Using approximate observational values of local-variance dipole, quadrupole, and octopole amplitudes, the paper finds a best fit near $I_2(\varphi,s)$03 and concludes that a configuration in which the CMB sphere does not intersect the wall provides a good fit to the data (Jazayeri et al., 2014).

Directional local-variance ideas also appear in low-$I_2(\varphi,s)$04 cosmography. A hemispherical comparison of Type Ia supernovae at $I_2(\varphi,s)$05 constructs a Hubble-map by fitting $I_2(\varphi,s)$06 separately on each hemisphere of an $I_2(\varphi,s)$07 HEALPix tessellation. The resulting maximal variance is

$I_2(\varphi,s)$08

toward $I_2(\varphi,s)$09. The direction agrees with bulk-flow estimates in the literature, and Monte Carlo tests give a moderate statistical significance, $I_2(\varphi,s)$10 CL. The paper therefore argues that the local variance of $I_2(\varphi,s)$11 can plausibly be caused by the bulk flow motion of the local Universe, while also stressing that current supernova uncertainties and sky coverage do not allow a definitive conclusion (Jr, 2015).

6. Microscopic local asymmetry in liquid water

In liquid water, local-variance asymmetry refers to a microscopic imbalance between the strengths of the two donor and/or the two acceptor hydrogen bonds of a single molecule. The relevant strength measure is the charge-transfer energy $I_2(\varphi,s)$12 from ALMO-EDA, and the asymmetry parameters are defined as

$I_2(\varphi,s)$13

Here $I_2(\varphi,s)$14 corresponds to a perfectly symmetric pair and $I_2(\varphi,s)$15 to an extreme asymmetry where the second interaction vanishes in strength (Elgabarty et al., 2020).

The quantitative asymmetry is substantial. The review reports that 75% of molecules have $I_2(\varphi,s)$16 or $I_2(\varphi,s)$17, meaning that the strongest donor or acceptor interaction is at least twice as strong as the second. About 25% have $I_2(\varphi,s)$18, so the strongest interaction is more than six times stronger than the second (Elgabarty et al., 2020). At the same time, the time-averaged geometry remains consistent with a nearly tetrahedral four-fold coordination. The asymmetry is therefore instantaneous and electronic rather than a permanent departure from tetrahedral coordination.

The dynamics are fast. Tracking the same first- and second-strongest donor partners shows that the initially strongest hydrogen bond becomes slightly weaker than the one that was second strongest at $I_2(\varphi,s)$19 after about $I_2(\varphi,s)$20, and that both converge toward a common average over hundreds of femtoseconds. The main relaxation is associated with one cycle of the intermolecular hydrogen-bond stretch, $I_2(\varphi,s)$21, corresponding to roughly $I_2(\varphi,s)$22 (Elgabarty et al., 2020).

The asymmetry has direct spectroscopic consequences. In ultrafast X-ray absorption, the pre-edge feature is dominated by the 25% most asymmetric molecules. In infrared spectroscopy, local asymmetry decouples the two OH stretch modes and broadens the OH stretch band. In terahertz response, molecules with $I_2(\varphi,s)$23 are more orientationally labile under the applied field (Elgabarty et al., 2020). A common misconception is that such observations require a majority of molecules with broken hydrogen bonds. The review instead argues that a relatively small but significant fraction of strongly asymmetric configurations is sufficient.

7. Comparative interpretation

The modern uses of local-variance asymmetry fall into three broad classes. First, there are directional estimators on a space of patches or hemispheres, where the primary object is a variance map or a closely related local-power field; this is the dominant CMB and low-$I_2(\varphi,s)$24 cosmography usage. Second, there are local-moment asymmetries around a point, where the key observable is a skewness or a side-dependent variance generated by an odd local term, as in deformed interference. Third, there are structural asymmetries between formally distinct variance-like quantities, as in fractional uncertainty, or between two local interactions attached to the same microscopic unit, as in liquid water.

This suggests that the phrase is best read as a methodological label rather than as the name of a single invariant object. In one literature, “local” means a sky patch; in another, a neighborhood of a fringe maximum; in another, a single molecule and its nearest hydrogen-bond partners; in another, the local limit of a quadratic form. The asymmetry can be dipolar, cubic, or bondwise; it can be static or transient; and it can reflect either estimator geometry, physical anisotropy, or the absence of a symmetric local representation.

The main controversies are correspondingly domain specific. In CMB work, significance depends on masking, smoothing scale, simulation fidelity, and the look-elsewhere effect (Jamshidi et al., 2024). In polarization, the inferred axis lies broadly toward the CMB kinetic dipole direction, and the authors explicitly warn that future data will decide whether the signal is cosmological or systematic (Aluri et al., 2017). In fractional uncertainty, the issue is not observational significance but the fact that only one side of the inequality has a clean local limit (1803.02384). In quantum interference, the central claim is precisely that the effect cannot be mimicked by conventional symmetric noise, because it changes neither fringe positions nor quadratic curvature (Zhang, 19 Mar 2026). In water, the central clarification is that strong instantaneous asymmetry coexists with nearly tetrahedral average structure (Elgabarty et al., 2020).

Taken together, these works establish local-variance asymmetry as a cross-disciplinary pattern of analysis: one studies a local variance-like quantity, asks how it fails to be symmetric under an expected operation—hemispherical exchange, left-right reflection, donor-acceptor equivalence, or local/nonlocal correspondence—and then uses that failure as a probe of underlying structure.