A counterexample to the Kato conjecture for positive commutators

Abstract: We disprove the conjectural converse to Kato's positivity criterion for commutators of functions of the canonical position and momentum operators $Q$ and $P$ by showing that the operator [ i\,[\,\arctan(P),\,\arctan(Q)\,] ] is nonnegative and nonzero.

• The paper proves that the commutator i[arctan(P), arctan(Q)] is nonnegative, nonzero, trace class, and has the exact trace π/2, providing a counterexample to the converse Kato conjecture.
• The authors establish positivity by expressing the commutator kernel through a conditionally negative definite exponent, using parity properties of an auxiliary operator and a Fock-space Gram factorization.
• The result shows that Kato analytic-strip conditions are sufficient but not necessary, with a scalable family of positive commutators existing beyond the conjectured Kato framework.

A counterexample to the Kato conjecture for positive commutators

Problem setting and principal result

The paper studies positivity of commutators generated by the canonical position and momentum operators

$P=-i\frac{d}{dx}, \qquad Q=x$

on $L^2(\mathbb R)$. For bounded real-valued functions $f$ and $g$, the operator

$i[f(P),g(Q)]$

is bounded and self-adjoint. The central question is to characterize those pairs $(f,g)$ for which this commutator is nonnegative and nonzero.

Kato established a sufficient analytic condition. If $f$ and $g$ belong to Kato classes $K_a$ and $K_b$, respectively, defined through bounded analytic continuation to lower half-strips together with the sign constraint

$L^2(\mathbb R)$0

then

$L^2(\mathbb R)$1

implies

$L^2(\mathbb R)$2

The conjectural converse, formulated explicitly by Herbst and Kriete, asserted that every nonzero positive commutator of this form should arise, up to almost-everywhere equivalence and simultaneous sign reversal, from a pair of Kato-class functions with strip widths satisfying $L^2(\mathbb R)$3.

Frank and Ivanisvili disprove this converse by proving that

$L^2(\mathbb R)$4

is nonnegative, nonzero, and trace class, with the exact trace

$L^2(\mathbb R)$5

This provides a direct counterexample to the Kato conjecture (2608.07805).

The contradiction with the conjecture follows from the analytic structure of the arctangent. The function $L^2(\mathbb R)$6 belongs to $L^2(\mathbb R)$7, but to no $L^2(\mathbb R)$8 with $L^2(\mathbb R)$9, because its analytic continuation has singularities at $f$0. Moreover, $f$1 belongs to no positive-width Kato class. Consequently, no choice of signs and strip widths can place both functions in Kato classes whose product equals $f$2.

Relation to the Howland–Kato problem

The result addresses a longstanding operator-theoretic problem originating in the study of dense point spectrum. Earlier work had identified several positive commutators, including examples involving $f$3 and $f$4, and Kato developed a systematic sufficient construction based on analytic continuation and positive-measure representations by translates of the hyperbolic tangent.

For $f$5, Kato’s representation expresses $f$6, modulo an additive real constant, as

$f$7

where $f$8 is a finite nonnegative measure. Positivity of commutators for such functions can therefore be deduced from positivity properties of the elementary $f$9 commutators.

The conjecture proposed that this representation captured all nonzero positive commutators. The counterexample demonstrates that positivity can arise from a substantially different mechanism. In particular, the derivatives

$g$0

have no nontrivial exponential moments, placing the example outside the regimes in which previous results had established the conjectured Kato-strip conclusion.

This feature is important: the failure is not caused by an irregular or discontinuous symbol. The functions are smooth, bounded, strictly increasing, and absolutely continuous. The obstruction is instead their borderline complex-analytic behavior and the absence of exponential decay in their derivatives.

Integral-kernel representation

The proof begins with an explicit kernel formula for the scaled family

$g$1

Writing

$g$2

the authors obtain

$g$3

where

$g$4

and

$g$5

The kernel is not manifestly positive semidefinite. Indeed, the divided-difference kernel associated with $g$6 is not positive semidefinite on the whole real line; such positivity would imply global operator monotonicity of the bounded function $g$7, which is impossible. Thus, positivity does not follow merely from the monotonicity of the symbols or from a positive divided difference. It emerges from the interaction between the divided difference and the Fourier multiplier generated by the derivative of $g$8.

The crucial structural identity rewrites $g$9 as a quadratic form. Define $i[f(P),g(Q)]$0 by the signed indicator of the interval between $i[f(P),g(Q)]$1 and $i[f(P),g(Q)]$2, and introduce an operator $i[f(P),g(Q)]$3 with kernel

$i[f(P),g(Q)]$4

Then

$i[f(P),g(Q)]$5

Therefore, positivity of the operator $i[f(P),g(Q)]$6 implies that $i[f(P),g(Q)]$7 is conditionally negative definite. Schoenberg’s theorem then yields positive semidefiniteness of $i[f(P),g(Q)]$8. The paper implements this implication explicitly through a symmetric Fock-space factorization.

The auxiliary operator and its parity structure

The operator $i[f(P),g(Q)]$9 is unitarily equivalent to $(f,g)$0 on $(f,g)$1, where $(f,g)$2 has kernel

$(f,g)$3

The function $(f,g)$4 satisfies

$(f,g)$5

and admits the positive integral representation

$(f,g)$6

This representation reveals a decisive parity decomposition. The even and odd restrictions of $(f,g)$7 satisfy

$(f,g)$8

The signs follow from the decomposition

$(f,g)$9

The first term generates a positive quadratic form on even functions, while the second generates a negative quadratic form on odd functions.

Consequently,

$f$0

is nonnegative provided

$f$1

Equivalently, for positive parameters $f$2 satisfying

$f$3

one obtains

$f$4

The symmetric scaling reduction shows that the two-parameter assertion reduces to the one-parameter family $f$5.

Fock-space factorization and exact trace

Once $f$6 is known to be nonnegative, set

$f$7

The quadratic-form identity becomes

$f$8

The kernel therefore has the form of a Gaussian-type Gram kernel. Introducing exponential vectors in the symmetric Fock space over $f$9, the authors construct vectors $g$0 such that

$g$1

If $g$2 denotes the corresponding operator from $g$3 into Fock space, then

$g$4

This proves nonnegativity directly and simultaneously establishes trace class.

The diagonal norm is particularly simple: $g$5 Hence

$g$6

The trace is independent of the scale parameter $g$7. In particular, the counterexample is not merely a positivity statement: it produces a positive trace-class commutator with an exact, scale-invariant trace.

Quantitative control of the positivity range

To verify that the relevant parameter range includes $g$8, the paper proves

$g$9

A simpler Schur-test estimate gives the weaker bound

$K_a$0

which is already sufficient for the principal theorem. The sharper estimate is obtained through a trace–residual argument.

The positive operator $K_a$1 has trace

$K_a$2

For the normalized trial vector

$K_a$3

the Rayleigh quotient is

$K_a$4

with numerical value in the interval

$K_a$5

The residual satisfies

$K_a$6

An abstract two-dimensional compression estimate then yields

$K_a$7

The numerical evidence reported in the paper suggests

$K_a$8

indicating that the analytic bound is close to optimal. The estimate also shows that the chosen cosine vector is an accurate approximate top eigenvector, providing information about the spectral geometry of the auxiliary kernel beyond what is required for the counterexample.

Analytic obstruction for the arctangent

The final section determines the exact Kato-strip behavior of the arctangent. For $K_a$9, $K_b$0 belongs to $K_b$1. Its analytic continuation satisfies

$K_b$2

after the corresponding scaling. This has the same sign as $K_b$3 within the strip $K_b$4.

The width $K_b$5 is maximal. If an analytic continuation existed to a larger lower half-strip, differentiating along the real axis would force

$K_b$6

throughout that strip by the identity theorem. Evaluating at $K_b$7 produces a contradiction because $K_b$8 vanishes there. The negative arctangent fails the Kato sign condition in every nontrivial strip, since its imaginary part has the opposite sign from the imaginary coordinate.

Thus, the positivity of

$K_b$9

cannot be reconciled with the Kato conjecture by selecting alternative analytic representatives or by changing the signs of the functions.

Theoretical and practical implications

The principal theoretical implication is a strict separation between positivity of canonical commutators and the Kato analytic-strip mechanism. Kato’s condition remains a robust sufficient criterion, but it is not a necessary characterization. The positive-commutator cone is therefore larger than the cone generated by Kato-class symbols and their positive-measure mixtures.

The proof suggests a broader principle. Positivity may be established by showing that a kernel exponent is conditionally negative definite, even when the underlying divided-difference kernel lacks positive definiteness. The relevant object is not the divided difference in isolation, but its modification by a Fourier-analytic factor. This places the problem at the intersection of commutator theory, Schoenberg kernels, conditional negative definiteness, and Fock-space realizations.

The construction also gives a parameterized family rather than a single isolated example. All positive $L^2(\mathbb R)$00 in the explicit range

$L^2(\mathbb R)$01

produce nonnegative trace-class commutators with trace $L^2(\mathbb R)$02. Using the bound $L^2(\mathbb R)$03, the admissible product includes values at least as large as $L^2(\mathbb R)$04. The conjectured Kato threshold $L^2(\mathbb R)$05 is therefore not an upper boundary for positivity in this class.

Future work will likely concern classification rather than verification of positivity alone. The parity decomposition and conditional-negative-definiteness representation may provide a framework for identifying further symbol pairs outside Kato classes. A complete theory would need to characterize the admissible structure of the kernels

$L^2(\mathbb R)$06

under positivity, perhaps through spectral factorizations, operator-valued measures, or intrinsic negative-type metrics. The arctangent example indicates that such a theory cannot be based solely on scalar analytic continuation widths or exponential moment conditions.

Conclusion

Frank and Ivanisvili prove that

$L^2(\mathbb R)$07

is a nonzero, nonnegative trace-class operator with exact trace $L^2(\mathbb R)$08, thereby disproving the converse Kato conjecture (2608.07805). The proof combines explicit commutator kernels, a parity-sensitive auxiliary operator, conditional negative definiteness, and a Fock-space Gram factorization. The result shows that Kato classes furnish an important sufficient mechanism for positive commutators but do not exhaust the positive-commutator phenomenon.