Zero-Deviation: Theory and Applications

• Zero-deviation is a multifaceted concept describing conditions under which deviation quantities exactly vanish—be it via Sobolev norm minimization, uniform quantum fidelity, precise resistance standards, or unbiased federated updates.
• In the Sobolev setting and quantum teleportation, specific algebraic and variational conditions (such as monic extremality or identical eigenvalues) guarantee the exact nullification of deviation metrics.
• Experimental and algorithmic implementations in quantum Hall metrology and federated LoRA exemplify how zero-deviation ensures reproducible measurements and structurally unbiased model aggregations.

In the cited literature, “zero-deviation” does not denote a single invariant notion. It instead labels several exact conditions in which a deviation quantity vanishes, or a polynomial minimizes deviation from the zero function. Díaz-González, Pijeira-Cabrera and Quintero-Roba study monic polynomials of least deviation from zero in Sobolev $p$-norms (Díaz-González et al., 2021). Ghosal et al. define zero fidelity deviation in two-qubit teleportation as the dispersion-free, or universal, regime in which fidelity is independent of the input state (Ghosal et al., 2019). Schopfer and Poirier use “zero-deviation” for the zero-dissipation extrapolation of the discrepancy between nominally identical quantum Hall resistance standards (Schopfer et al., 2013). DEeR defines zero aggregation deviation for federated low-rank adaptation as the exact equality between naïve LoRA averaging and the true average update (Zhu et al., 2024).

1. Cross-domain meaning and formal structure

Across these works, the central object is always a deviation quantity with a precise operational interpretation. In approximation theory, the deviation is from the zero function under a Sobolev norm. In teleportation, it is the standard deviation of fidelity over all input states. In quantum Hall metrology, it is the relative discrepancy $\Delta R/R$ between Hall resistors on the $\nu=2$ plateau. In federated LoRA, it is the matrix difference between two aggregation rules.

Domain	Quantity	Defining condition
Sobolev extremal polynomials	$\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$	$P_n$ is monic of exact degree $n$
Two-qubit teleportation	$$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$	$\sigma=0$
Quantum Hall metrology	$\Delta R/R=(R_H-R_K/2)/(R_K/2)$	zero-dissipation extrapolation
Federated LoRA	$\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$	$\Delta R/R$0

A common technical feature is that zero-deviation is stronger than small average error. In the teleportation setting, $\Delta R/R$1 is stronger than merely having $\Delta R/R$2. In quantum Hall metrology, a flat plateau is not sufficient if $\Delta R/R$3 remains finite. In federated LoRA, unbiased-looking factor averaging is not sufficient unless the factorization constraint eliminates $\Delta R/R$4. In the Sobolev setting, the issue is different: the relevant object is not a variance-like deviation but extremality with respect to zero.

2. Least deviation from zero in Sobolev $\Delta R/R$5-norm

For real-coefficient polynomials, let $\Delta R/R$6 be the polynomial space and let $\Delta R/R$7 denote the monic polynomials of exact degree $\Delta R/R$8. An $\Delta R/R$9th minimal or extremal polynomial is a monic $\nu=2$0 satisfying

$\nu=2$1

In this framework, “least deviation from zero” means minimization of the Sobolev $\nu=2$2-norm of a monic polynomial (Díaz-González et al., 2021).

For a vector of finite positive Borel measures $\nu=2$3 and $\nu=2$4, the continuous Sobolev $\nu=2$5-norm is

$\nu=2$6

when each $\nu=2$7 is supported on a compact subset of $\nu=2$8 and has no atoms. If some $\nu=2$9 has positive mass at finitely many points, the same formula defines a discrete Sobolev $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$0-norm, often written

$\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$1

For $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$2,

$\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$3

and a point-mass contribution reduces to $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$4.

For $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$5, a monic polynomial $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$6 is extremal if and only if

$\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$7

For $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$8, if a monic $\|P_n\|=\inf_{Q\in\mathbb P_n^1}\|Q\|$9 satisfies

$P_n$0

then $P_n$1 is extremal; and when $P_n$2 is continuous, this condition is also necessary. The paper emphasizes that $P_n$3 loses uniqueness in general. With $P_n$4, the family

$P_n$5

is extremal of degree $P_n$6. For the discrete norm

$P_n$7

the monic family

$P_n$8

is again extremal.

The discrete case also supports asymptotic zero-distribution results. If $P_n$9 is regular in the logarithmic potential-theory sense and $n$0 is in the Reg class, then for the monic Sobolev extremal polynomials $n$1,

$n$2

Under the additional sequentially ordered condition on the discrete part, with

$n$3

the $n$4th extremal polynomial has at least $n$5 simple real zeros in the convex hull of $n$6. In this context, the zero-related statement concerns approximation to the zero function rather than the vanishing of a fluctuation metric.

3. Zero fidelity deviation and universality in two-qubit teleportation

For a shared two-qubit resource $n$7 and an arbitrary pure input qubit $n$8, the teleportation fidelity is

$n$9

The average fidelity is

$$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$0

where $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$1 is the normalized Haar measure on the Bloch sphere, and the fidelity deviation is

$$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$2

Ghosal et al. identify $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$3 as the measure of fidelity fluctuations over all input states (Ghosal et al., 2019).

In the Hilbert-Schmidt decomposition,

$$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$4

In an optimal local-unitary-preprocessed protocol one may take $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$5 and $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$6 diagonal. If the real eigenvalues of $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$7 are $$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$8, then

$$\sigma=\sqrt{\langle f^2\rangle-\langle f\rangle^2}$$9

or, since $\sigma=0$0 is diagonal,

$\sigma=0$1

The universality, or dispersion-free, condition is $\sigma=0$2. Requiring $\sigma=0$3 gives

$\sigma=0$4

which for three real numbers holds if and only if $\sigma=0$5. Hence the necessary and sufficient condition for zero fidelity deviation is

$\sigma=0$6

In that case the teleportation fidelity is independent of the input $\sigma=0$7.

The relation to quantum advantage is explicit. For useful teleportation, the relevant case is $\sigma=0$8, and the maximal average fidelity is

$\sigma=0$9

If $\Delta R/R=(R_H-R_K/2)/(R_K/2)$0, then

$\Delta R/R=(R_H-R_K/2)/(R_K/2)$1

Since the classical teleportation bound is $\Delta R/R=(R_H-R_K/2)/(R_K/2)$2, universal states beating the classical bound satisfy

$\Delta R/R=(R_H-R_K/2)/(R_K/2)$3

These are precisely the canonical Werner-like states with $\Delta R/R=(R_H-R_K/2)/(R_K/2)$4, $\Delta R/R=(R_H-R_K/2)/(R_K/2)$5.

The examples sharpen the classification. For pure entangled states in Schmidt form,

$\Delta R/R=(R_H-R_K/2)/(R_K/2)$6

one has

$\Delta R/R=(R_H-R_K/2)/(R_K/2)$7

$\Delta R/R=(R_H-R_K/2)/(R_K/2)$8

In terms of the concurrence $\Delta R/R=(R_H-R_K/2)/(R_K/2)$9,

$\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$0

Thus $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$1 only if $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$2, namely the maximally entangled case. For Bell-diagonal states $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$3, with $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$4, one has $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$5 as soon as $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$6, and

$\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$7

Zero deviation implies $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$8, hence a Werner state. Ghosal et al. also exhibit two one-parameter families of $\mathcal O=\bar B\bar A-\frac1K\sum_k B_kA_k$9-states for which $\Delta R/R$00, hence $\Delta R/R$01, while the marginals are not maximally mixed; in each case $\Delta R/R$02, so these states are universal for $\Delta R/R$03 and useless for $\Delta R/R$04.

4. Zero-deviation near the zero-dissipation state in quantum Hall metrology

In the quantum Hall effect, “zero-deviation” refers to the experimental demonstration that four nominally identical GaAs/AlGaAs quantum-Hall-resistance standards, each realizing $\Delta R/R$05 at filling factor $\Delta R/R$06, agree within a few $\Delta R/R$07 of the nominal quantized value (Schopfer et al., 2013). Schopfer and Poirier obtain this result with a modified Wheatstone-bridge technique.

The four Hall bars have width $\Delta R/R$08, identical GaAs/AlGaAs two-dimensional electron gases, $\Delta R/R$09, and $\Delta R/R$10. The “triple-connection” of Delahaye makes contact and lead resistances enter only in third order,

$\Delta R/R$11

so wire or contact errors are negligible. An AC excitation $\Delta R/R$12 at low frequency ($\Delta R/R$13–$\Delta R/R$14) is fed into one diagonal of the bridge, and the bridge unbalance current $\Delta R/R$15 is detected in the other diagonal by a cryogenic current comparator and SQUID with ultimate resolution $\Delta R/R$16. In the DC limit,

$\Delta R/R$17

and

$\Delta R/R$18

By measuring $\Delta R/R$19 versus $\Delta R/R$20 and extrapolating the in-phase bridge unbalance to $\Delta R/R$21, the method removes reactive or frequency-dependent error.

At $\Delta R/R$22, the ideal quantized value is

$\Delta R/R$23

At the central magnetic field $\Delta R/R$24, where $\Delta R/R$25 for all four bars and longitudinal dissipation is minimal, linear extrapolation to zero mean longitudinal resistance yields

$\Delta R/R$26

This demonstrates reproducibility between the four standards with a relative uncertainty of $\Delta R/R$27.

The paper further shows that the mean macroscopic longitudinal resistance

$\Delta R/R$28

follows the phenomenological resistivity rule

$\Delta R/R$29

From the slope of $\Delta R/R$30 versus $\Delta R/R$31, one extracts

$\Delta R/R$32

and this $\Delta R/R$33 remains constant for currents $\Delta R/R$34–$\Delta R/R$35. Physically, $\Delta R/R$36 with $\Delta R/R$37, because spatial carrier-density fluctuations $\Delta R/R$38 generate microscopic $\Delta R/R$39-variations and macroscopic dissipation $\Delta R/R$40 is dominated by these $\Delta R/R$41 fluctuations rather than the intrinsic $\Delta R/R$42. The vanishing of the plateau slope therefore requires $\Delta R/R$43.

$\Delta R/R$44 ($\Delta R/R$45A)	$\Delta R/R$46 ($\Delta R/R$47)	$\Delta R/R$48 ($\Delta R/R$49, combined)
40	$\Delta R/R$50	$\Delta R/R$51
80	$\Delta R/R$52	$\Delta R/R$53
120	$\Delta R/R$54	$\Delta R/R$55

Even a few-$\Delta R/R$56 longitudinal resistance produces nonzero $\Delta R/R$57 on the plateau. Plotting $\Delta R/R$58 versus $\Delta R/R$59 yields a straight line of slope $\Delta R/R$60, and extrapolation to $\Delta R/R$61 recovers the zero-deviation intercept. Below about $\Delta R/R$62, the longitudinal resistance is purely temperature-limited and independent of $\Delta R/R$63. Sweeping $\Delta R/R$64 through the $\Delta R/R$65 plateau to identify the pivot field $\Delta R/R$66, and ensuring density homogeneity $\Delta R/R$67, are the stated conditions for driving all four bars simultaneously to the zero-dissipation state.

5. Zero aggregation deviation in privacy-preserving federated LoRA

DEeR studies the combination of low-rank adaptation and federated learning, focusing on aggregation deviation and differential-privacy noise amplification (Zhu et al., 2024). With $\Delta R/R$68 clients, each client $\Delta R/R$69 holds LoRA factors $\Delta R/R$70 and $\Delta R/R$71, and the true model update is

$\Delta R/R$72

A naïve FedAvg on LoRA separately averages the factors,

$\Delta R/R$73

and uses the global update

$\Delta R/R$74

The correct average of true updates is

$\Delta R/R$75

The deviation is therefore

$\Delta R/R$76

An algebraic rearrangement yields

$\Delta R/R$77

The paper states that

$\Delta R/R$78

Hence the necessary and sufficient condition for zero-deviation is that, at aggregation time, one of the two factor sets must be identical across all clients.

DEeR enforces this condition every round by alternating minimization. In the $\Delta R/R$79-step, the server broadcasts $\Delta R/R$80, each client solves

$\Delta R/R$81

and the server aggregates

$\Delta R/R$82

In the $\Delta R/R$83-step, the server broadcasts $\Delta R/R$84, each client solves

$\Delta R/R$85

and the server aggregates

$\Delta R/R$86

During the $\Delta R/R$87-step, all clients use the same $\Delta R/R$88; during the $\Delta R/R$89-step, all clients use the same $\Delta R/R$90. The paper therefore concludes that every mini-aggregation satisfies $\Delta R/R$91.

The interaction with differential privacy is handled separately. If Gaussian noise is injected independently into $\Delta R/R$92 and $\Delta R/R$93, then

$\Delta R/R$94

The linear terms scale with $\Delta R/R$95 and $\Delta R/R$96, so the noise is amplified over rounds. DEeR’s Noise Regulator instead samples a single matrix noise $\Delta R/R$97 and solves

$\Delta R/R$98

or

$\Delta R/R$99

This yields

$\nu=2$00

The entire mechanism therefore adds exactly one Gaussian noise matrix to each update, with no dependence on the growing norms of $\nu=2$01 or $\nu=2$02.

6. Recurrent distinctions and common misconceptions

Several distinctions recur across these four literatures. In the Sobolev setting, extremality does not imply uniqueness when $\nu=2$03; the examples $\nu=2$04 for $\nu=2$05 and $\nu=2$06 for $\nu=2$07 show explicit non-uniqueness (Díaz-González et al., 2021). In two-qubit teleportation, usefulness and universality are not the same: $\nu=2$08 does not by itself imply $\nu=2$09, and the pure-state formula

$\nu=2$10

shows that zero fidelity deviation occurs only at maximal entanglement for that family (Ghosal et al., 2019).

In quantum Hall metrology, the presence of a Hall plateau does not by itself establish zero-deviation. Schopfer and Poirier show that even a few-$\nu=2$11 mean longitudinal resistance can produce finite $\nu=2$12, and that the correct zero-deviation statement is the extrapolated intercept

$\nu=2$13

rather than a visual judgment about plateau flatness (Schopfer et al., 2013). In federated LoRA, averaging factor matrices is not automatically equivalent to averaging their products; DEeR isolates the exact obstruction in $\nu=2$14 and removes it by forcing one factor set to be identical across clients at aggregation time (Zhu et al., 2024).

This suggests a common interpretive pattern: zero-deviation is typically a structural guarantee, not merely a small-error regime. In the teleportation and federated-learning cases, the zero condition is an exact algebraic constraint on the correlation matrix or on the LoRA factors. In the quantum Hall case, it is an experimentally defined zero-dissipation limit. In the Sobolev case, the relevant phrase identifies optimal approximation to the zero function and is therefore conceptually adjacent, but not identical, to a vanishing fluctuation or discrepancy metric.