Some Gruss type inequalities for Frechet differentiable mappings

Amir Ghasem Ghazanfari; Teimouri Azadbakht

REVIEW 3 major objections 5 minor 9 references

Some Gruss type inequalities for Frechet differentiable mappings

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fréchet derivatives in Hilbert C*-modules obey Grüss-type covariance bounds with constant one quarter.

desk verdict Novel Grüss-type inequalities for differentiable maps into C*-modules that are mostly sound, but the companion Banach algebra theorem is false as stated. read the letter →

arxiv 1908.05326 v1 pith:QPFANVCX submitted 2019-08-14 math.FA

Teimouri Azadbakht , Amir Ghasem Ghazanfari This is my paper · ORCID

classification math.FA MSC 26D1046C0546L08

keywords GrüssinequalityFréchetdifferentiablemappingsHilbertC*-modulessemi-innerproductBanach*-algebraKorkineidentityoperator-valuedinner

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that the classical Grüss inequality—which bounds the gap between the average of a product and the product of averages by one quarter of the product of the ranges—has a genuine analogue for Fréchet derivatives of functions that take values in a Hilbert C*-module. The authors build an operator-valued semi-inner product on the space of differentiable mappings and use it to control weighted sums of derivatives at a point. If the main inequality is right, the usual discrete and integral Grüss estimates are not special to scalar functions: they hold when the inner product is valued in a C*-algebra and the maps are operator-valued. The paper also proves that the space of such differentiable maps between Banach *-algebras is itself a Banach *-algebra.

What carries the argument

The load-bearing object is the modified $A$-valued inner product $\langle x,y\rangle_1 = \langle x,y\rangle - \langle x,e\rangle\langle e,y\rangle$, with $e$ chosen so that $\langle e,e\rangle$ is an idempotent in the C*-algebra; subtracting $e\langle e,\cdot\rangle$ is what turns the derivative space into a semi-inner product module. On $D_p(A,X)$ this yields the bracket $[f,g]_a$ above, whose positivity and Schwarz inequality drive every bound. The other main mechanism is the Korkine identity, which rewrites the weighted covariance as $\frac{1}{2}\sum_{i,j}r_ir_j\langle Df_i(p)(a)-Df_j(p)(a), Dg_i(p)(a)-Dg_j(p)(a)\rangle$; this identity makes the covariance a semi-inner product and reduces the Grüss bound to two applications of Schwarz plus the diameter hypotheses.

What would settle it

Work in the finite-dimensional case $A = X = M_2(\mathbb{C})$ with the standard module inner product $\langle a,b\rangle = a^*b$, choose $e$ with $\langle e,e\rangle$ idempotent, and take two low-degree matrix polynomials $f,g$ whose derivatives satisfy the diameter hypotheses in Lemma 2. A direct numerical evaluation of the left side of (2.2) for a nontrivial probability vector and a matrix $a$ either produces a value exceeding $\frac{1}{4}\|x_0-y_0\|\|x_1-y_1\|\|a\|^2$, which would refute the claim, or confirms the bound in that test case; the same calculation can be run for many random choices of $f,g$.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that for differentiable maps $f,g$ from a C*-algebra $A$ into a Hilbert C*-module $X$, the operator-valued expression $$(f,g)_a = \sum_{i=1}^n r_i \langle Df_i(p)(a), Dg_i(p)(a)\rangle - \left\langle \sum_{i=1}^n r_i Df_i(p)(a), \sum_{i=1}^n r_i Dg_i(p)(a)\right\rangle$$ is bounded by $\frac{1}{4}\|x_0-y_0\|\|x_1-y_1\|\|a\|^2$ whenever the derivative vectors stay within the indicated diameters of the midpoint maps $\frac{x_0+y_0}{2}$ and $\frac{x_1+y_1}{2}$. The machinery is a generalized semi-inner product on $D_p(A,X)$, namely $[f,g]_a = \langle Df(p)(a),Dg(p)(a)\rangle_1 + \langle f(p),g(p)\rangle_1 - D\langle f(\cdot),g(\cdot)\rangle_1(p)(a)$, where $\langle x,y\rangle_1 = \langle x,y\rangle - \langle x,e\rangle\langle e,y\rangle$ for an element $e$ with $\langle e,e\rangle$ idempotent. From the main bound the paper derives weighted variance inequalities for derivatives and explicit bounds for linearly and quadratically indexed families.

Load-bearing premise

The whole argument depends on the modified inner product $\langle x,x\rangle - \langle x,e\rangle\langle e,x\rangle$ being a nonnegative element of $A$ for every $x$ whenever $\langle e,e\rangle$ is idempotent; if that positivity is false, the bracket is not a semi-inner product and the Grüss bounds do not follow.

Editorial extensions

If this is right

If inequality (2.2) holds, the classical constant $\frac{1}{4}$ from Grüss's 1934 theorem survives verbatim in the C*-module setting, with the extra factor $\|a\|^2$ measuring the direction in which the derivatives are evaluated.
Corollary 1 makes the bound a variance inequality: the weighted deviation $\sum r_i\alpha_i Df_i(p)(a) - (\sum r_i\alpha_i)(\sum r_i Df_i(p)(a))$ is controlled by the weighted variance of the coefficients $\alpha_i$ and the diameter of the derivatives.
The explicit formulas in Corollary 2 show that linearly weighted sums of $n$ derivatives grow at most like $n^{3/2}$ and quadratically weighted sums like $n^{5/2}$, up to constants and the diameter factor.
Theorem 4 gives $D_p(A,B)$ a Banach *-algebra structure, so differentiable maps between Banach *-algebras form a complete algebra with an involution compatible with the Fréchet derivative.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The weighted-covariance form of (2.2) reads like an empirical covariance bound for operator-valued random variables: if the $r_i$ are probabilities and the $Df_i(p)(a)$ are samples, the inequality controls the covariance of two dependent samples by the product of their ranges, so it could feed concentration or bootstrap arguments for matrix-valued derivatives.
Because $e$ is arbitrary subject to $\langle e,e\rangle$ being idempotent, the bounds carry a free parameter; choosing $e$ to minimize the right-hand side for a specific module is an optimization problem the paper leaves open.
The Banach *-algebra theorem suggests that repeated differentiation and functional calculus on $D_p(A,B)$ are available, so one could define higher-order Grüss inequalities by iterating the bracket construction; the paper does not pursue this.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Thanks for the report. The core of this paper is a genuinely new object: a bracket [f,g]_a = _1 on a space of Fréchet differentiable maps into a Hilbert C*-module. Using it, the authors prove Grüss-type bounds (Lemmas 1 and 2) that are legitimate extensions of earlier work on inner product modules. The Korkine identity for D_p(A,X^n) is a nice touch. After filling small gaps — the positivity of [·,·]_a follows from the C*-Schwarz inequality, not from the Gram-Schmidt-like equality written in Theorem 3 — Lemmas 1 and 2 reconstruct cleanly. Corollary 1 is also fine once you replace the garbled equality with the correct triangle-plus-Cauchy-Schwarz step.

The soft spots are real but localized. Theorem 4, the Banach *-algebra claim, is false as stated. The norm in (3.1) is infinite on D_p(A,B) because that space includes unbounded functions — f(z)=z^2 sin(1/z^2) on C is a concrete example. Even on bounded functions, the max of sup|Df| and sup|f| is not submultiplicative; think high-frequency small-amplitude oscillations. The proof also assumes a common domain U for a Cauchy sequence, which isn't part of the definition. This section needs a rewrite or removal. Corollary 2's second constant is off by a factor of 5: the substitution into (2.3) gives n/(12√5) times the stated square-root term, not n√5/12. The first half of Corollary 2 is correct. Nothing in the flaw in Theorem 4 contaminates the Grüss results.

The citation practice is honest; the main external tool, Theorem 2, comes from the authors' earlier paper and is used legitimately. No circularity or fitting here.

Who should read it: people working on operator inequalities in C*-modules. The inequality work is worth knowing about, but the Banach algebra section should be treated with caution. I'd send it to a referee — the core results are novel and repairable, and a referee can help sort the correct parts from the broken ones. But it needs major revision before it's publishable, not just copy-editing.

Referee Report

3 major / 5 minor

Summary. The paper studies Grüss-type inequalities for Fréchet differentiable maps from a C*-algebra A into a Hilbert C*-module X. It defines the space Dp(A,X), introduces several A-valued semi-inner products on this function space (one using an idempotent element e to define a Gram–Schmidt-type bracket, and one weighted-covariance bracket on X^n), and proves two main inequalities (Lemmas 1 and 2) bounding the deviation of a weighted sum of inner products from the inner product of weighted sums in terms of diameter conditions on f(p), Df(p)(a), and analogous quantities for g. Corollaries give variance-type bounds for weighted sums of derivatives. A separate section claims that Dp(A,B) is a Banach *-algebra with a sup-norm.

Significance. If fully justified, Lemmas 1 and 2 provide a genuine operator-valued extension of the classical Grüss inequality to differentiable maps into C*-modules, with explicit constants and a natural Korkine-type identity. Lemma 2 in particular gives a clean bound depending only on the diameters of the derivative maps, and Corollaries 1 and 2 recover discrete variance inequalities whose constants check out. The paper also advertises a Banach *-algebra completeness theorem, but that theorem is false as stated; this substantially weakens the secondary claim. The Grüss-inequality core appears correct and repairable, while the Banach-algebra claim needs a genuine restriction and a different norm.

major comments (3)

[Section 3, Theorem 4, Eq. (3.1)] The theorem is false as stated. The set Dp(A,B) was defined in Section 1 with no boundedness condition, so the quantity sup_{a∈A}‖f(a)‖ can be infinite even for very smooth f. For A=B=C and p=0, the function f(0)=0, f(z)=z^2 sin(1/z^2) for z≠0 is continuous and Fréchet differentiable on a neighborhood of 0, but sup_{z∈C}|f(z)|=∞ and sup_{x∈U}|Df(x)|=∞ for every neighborhood U of 0. Thus (3.1) is not a norm on Dp(A,B), and the completeness proof applies only to a proper subspace. Moreover, the proposed max norm is not submultiplicative: with A=R and B=C, the functions f(x)=g(x)=e^{ix} satisfy ‖f‖=‖g‖=1, while ‖fg‖=2, so the pointwise product inequality ‖fg‖≤‖f‖‖g‖ fails. The theorem requires a restriction to functions with finite sup and finite derivative sup on a fixed neighborhood, and either a different norm (for example, the sum of the two suprema) or an explicit proof of submultiplicativity.
[Section 2, proof of Corollary 1] The displayed equality at the beginning of the proof of (2.3) is not valid: it writes the norm of the weighted sum as |Σ r_i(α_i−Σ r_j α_j)| times a single norm ‖Df_i(p)(a)−(x0+y0)/2·a‖, but the index i is free and the expression is not an equality. The statement of Corollary 1 is nevertheless correct, and a standard proof is available: for any c∈X, Σ r_i α_i u_i − (Σ r_i α_i)(Σ r_i u_i)=Σ r_i(α_i−ar α)(u_i−c), so Cauchy–Schwarz together with the diameter assumption yields (2.3). The proof should be rewritten along these lines.
[Section 2, Lemma 1 and Theorem 2] Lemma 1 applies Theorem 2, which is quoted only for Hilbert C*-modules, to a semi-inner product C*-module X. As written this is a gap: the quoted theorem requires a complete, nondegenerate inner product module, while X need be neither. The inequality presumably extends by quotienting by the null space of ⟨·,·⟩ and completing, but this step should be stated explicitly. A brief justification would make the proof of Lemma 1 complete.

minor comments (5)

[Section 1, Eq. (1.2)] The displayed mean value formula has a missing closing norm: it should read ‖f(x)−f(y)‖ ≤ ‖x−y‖ sup_{0<θ<1}‖Df((1−θ)x+θy)‖.
[Section 2, Theorem 3] The proof that [·,·]_a is a generalized semi-inner product is incomplete: additivity and A-linearity in the second argument are asserted but not shown. These are straightforward verifications and should be included.
[Section 2, Lemma 1] The last step of the proof is omitted: after the estimates for ‖[f,f]_a‖ and ‖[g,g]_a‖, the Schwarz inequality should be applied to obtain the stated bound. Note that the displayed estimates actually give the stronger constant 1/4, so the stated 1/2 is valid but the final line should be written out.
[Section 2, Corollary 2] The substitutions r_i=1/n and α_i=k or k^2 introduce an extra factor n on the left-hand side of (2.3); the inequalities (2.4) and (2.5) are correct, but the proof should explicitly multiply the right-hand side of (2.3) by n before passing to the displayed forms.
[Section 3, Eq. (3.1)] The notation sup_{x∈U} is ambiguous because the neighborhood U is not part of the data determining a function f∈Dp(A,B); the theorem must either fix a common neighborhood for the entire space or quantify over the admissible neighborhoods.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the Grüss-type bounds are derived from explicit semi-inner products and the cited Theorem 2 is independent published support; the Section 3 completeness gap is a correctness issue, not circularity.

full rationale

The derivation chain in Sections 2 and 3 is not circular. Lemma 1 and Lemma 2 construct explicit A-valued semi-inner products on D_p(A,X) and D_p(A,X^n), verify positivity using the idempotence of <e,e> and the standard Korkine identity, and then obtain the bounds (2.2)-(2.5) by applying the Cauchy-Schwarz inequality to those semi-inner products plus the stated diameter hypotheses. No quantity is fitted to a subset of data and then presented as a prediction, and no claimed inequality is just the definition of an object introduced for the purpose. The only external result used in the main chain is Theorem 2 from [3], a published parameter-free inequality for arbitrary elements of a Hilbert C*-module; although [3] shares an author, its assumptions do not include the target function-space inequalities and its content is not equivalent to them, so it is independent mathematical support rather than a self-referential loading. The skeptical concern about Theorem 4 is that the norm in (3.1) need not be finite on all of D_p(A,B) and the completeness proof therefore has a gap; that is a correctness defect, not a circular reduction of the paper's claims to its inputs.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central inequalities rest on standard module theory and the prior theorem [3]. There are no fitted numerical parameters and no new postulated entities. The only ad hoc element is the finiteness assumption needed to make the Theorem 4 norm well-defined.

assumptions (5)

standard math Generalized Schwarz inequality for semi-inner product C*-modules, <x,y><y,x> <= ||<x,x>|| <y,y>, as in (2.1).
Invoked for the new brackets [·,·]_a and (·,·)_a in Lemma 1 and Lemma 2; cited to Lance [7, Proposition 1.1].
standard math Theorem 2 of [3], the prior Grüss-type inequality for inner product modules over Banach *-algebras.
Used as a black box in Lemma 1 to bound individual inner product terms. One author of [3] is a co-author here, but the result is a published theorem and is not the target claim.
domain assumption A is a C*-algebra and X is a semi-inner product A-module; Dp(A,X) is a right A-module under (f a)(t)=f(t)a.
Needed for the module actions and A-valued inner products used throughout Section 2; stated in Section 2.
domain assumption There exists e in X with <e,e> idempotent, and the boundedness hypotheses place f(p), Df(p)(a), g(p), Dg(p)(a) in strips around e.
Explicit assumptions in Theorem 3 and Lemma 1; the positivity of the projected bracket and the bounds depend on them.
ad hoc to paper For Theorem 4, all functions f in Dp(A,B) have finite sup over A and over U of f and Df respectively.
The norm (3.1) is otherwise infinite for many differentiable maps; the paper does not restrict Dp to such functions, so the finiteness is glossed over.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Some Gruss type inequalities for Frechet differentiable mappings." pith.science (2026). https://pith.science/paper/QPFANVCX

@misc{pith2026190805326,
  author       = {Pith},
  title        = {Pith review of: Some Gruss type inequalities for Frechet differentiable mappings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QPFANVCX}},
  note         = {Machine review of arXiv:1908.05326}
}

read the original abstract

Let X be a Hilbert C^*-module on C^*-algebra A and p in A. We denote by Dp(A;X) the set of all continuous functions f on A, which are Frechet differentiable on a open neighborhood U of p. Then, we introduce some generalized semi-inner products on Dp(A;X), and using them some Gruss type inequalities in semi-inner product C^*-module Dp(A;X) and Dp(A;X^n) are established.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages

[1]

S. S. Dragomir, Advances in Inequalities of the Schwarz, Gr¨ uss and Bessel T ype in Inner Product Spaces , Nova Science puplishers Inc., New York, 2005

work page 2005
[2]

S. S. Dragomir, A Gr¨ uss type discrete inequality in inner product spaces an d applica- tions, J. Math. Anal. Appl., 250 (2000), 494-511

work page 2000
[3]

Ghazanfari, S.S

A.G. Ghazanfari, S.S. Dragomir, Bessel and Gr¨ uss type inequalities in inner product modules over Banach ∗-algebra, Linear Algebra Appl. 434 (2011), 944-956

work page 2011
[4]

Gr¨ uss, ¨Uber das Maximum des absoluten Betrages von 1 b− a ∫ b a f (x)g(x)dx − 1 (b− a)2 ∫ b a f (x)dx ∫ b a g(x)dx, Math

G. Gr¨ uss, ¨Uber das Maximum des absoluten Betrages von 1 b− a ∫ b a f (x)g(x)dx − 1 (b− a)2 ∫ b a f (x)dx ∫ b a g(x)dx, Math. Z. 39(1934), 215-226

work page 1934
[5]

Iliˇ sevi´ c and S

D. Iliˇ sevi´ c and S. Varoˇ sanec,Gr¨ uss type inequalities in inner product modules , Proc. Amer. Math. Soc. 133 (2005), 3271-3280

work page 2005
[6]

A. I. Kechriniotis and K. K. Delibasis, On generalizations of Gr¨ uss inequality in inner Product Spaces and applications , J. Inequal. Appl. Vol(2010), Article ID 167091

work page 2010
[7]

Lance, Hilbert C∗ -Modules, London Math

E.C. Lance, Hilbert C∗ -Modules, London Math. Soc. Lecture Note Series 210, Cam- bridge Univ. Press, 1995

work page 1995
[8]

X. Li, R. N. Mohapatra and R. S. Rodriguez, Gr¨ uss-type inequalities, J. Math. Anal. Appl. 267 (2002), no. 2, 434-443

work page 2002

Show all 9 references

[9]

D. S. Mitrinovi´ c, J. E. Peˇ cari´ c, and A. M. Fink, Classical and New Inequalities in Analysis, Kluwer Academic, Dordrecht, 1993. 1,2Department of Mathematics, Lorestan University, P.O. Box 4 65, Kho- ramabad, Iran. E-mail address : 1t.azadbakhat88@gmail.com, 2ghazanfari.a@lu.ac.ir

1993

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

S. S. Dragomir, Advances in Inequalities of the Schwarz, Gr¨ uss and Bessel T ype in Inner Product Spaces , Nova Science puplishers Inc., New York, 2005

work page 2005

[2] [2]

S. S. Dragomir, A Gr¨ uss type discrete inequality in inner product spaces an d applica- tions, J. Math. Anal. Appl., 250 (2000), 494-511

work page 2000

[3] [3]

Ghazanfari, S.S

A.G. Ghazanfari, S.S. Dragomir, Bessel and Gr¨ uss type inequalities in inner product modules over Banach ∗-algebra, Linear Algebra Appl. 434 (2011), 944-956

work page 2011

[4] [4]

Gr¨ uss, ¨Uber das Maximum des absoluten Betrages von 1 b− a ∫ b a f (x)g(x)dx − 1 (b− a)2 ∫ b a f (x)dx ∫ b a g(x)dx, Math

G. Gr¨ uss, ¨Uber das Maximum des absoluten Betrages von 1 b− a ∫ b a f (x)g(x)dx − 1 (b− a)2 ∫ b a f (x)dx ∫ b a g(x)dx, Math. Z. 39(1934), 215-226

work page 1934

[5] [5]

Iliˇ sevi´ c and S

D. Iliˇ sevi´ c and S. Varoˇ sanec,Gr¨ uss type inequalities in inner product modules , Proc. Amer. Math. Soc. 133 (2005), 3271-3280

work page 2005

[6] [6]

A. I. Kechriniotis and K. K. Delibasis, On generalizations of Gr¨ uss inequality in inner Product Spaces and applications , J. Inequal. Appl. Vol(2010), Article ID 167091

work page 2010

[7] [7]

Lance, Hilbert C∗ -Modules, London Math

E.C. Lance, Hilbert C∗ -Modules, London Math. Soc. Lecture Note Series 210, Cam- bridge Univ. Press, 1995

work page 1995

[8] [8]

X. Li, R. N. Mohapatra and R. S. Rodriguez, Gr¨ uss-type inequalities, J. Math. Anal. Appl. 267 (2002), no. 2, 434-443

work page 2002