Applications of microlocal analysis to inverse problems

Mikko Salo

REVIEW 2 major objections 3 minor 21 references

Applications of microlocal analysis to inverse problems

T0 review · 2 major / 3 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 These lecture notes argue that the classical theorems of three inverse problems—Radon inversion, wave-equation coefficient recovery, and boundary conductivity determination—can be proved by one elementary quasimode construction.

desk verdict Largely sound lecture notes that hit the microlocal-inverse-problems highlights, but the proof of Proposition 4.4(4.4) has a phase-cancellation error that invalidates the claimed elementary proof of the Rakesh–Symes theorem. read the letter →

arxiv 1908.03041 v1 pith:M5CGXXVD submitted 2019-08-08 math.AP

Mikko Salo This is my paper · ORCID

classification math.AP MSC 35R3035S3044A1235A27

keywords microlocalanalysisinverseproblemsDirichlet-to-NeumannmapRadontransformpseudodifferentialoperatorsFourierintegralquasimodeconstructionboundarydetermination

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

These lecture notes aim to show that the core theorems of three inverse problems can be proved with a single elementary tool: testing boundary measurements against highly oscillatory trial solutions. The first result is the planar Radon transform identity $R^*R = 4\pi|D|^{-1}$, which makes the normal operator an elliptic pseudodifferential operator and explains which singularities are stably recoverable from limited data. The second is that equality of hyperbolic Dirichlet-to-Neumann maps forces equality of line integrals of the unknown potential over every segment of length less than the measurement time $T$. The third is that equality of elliptic Dirichlet-to-Neumann maps forces the boundary Taylor series of two conductivities to match. The payoff is a concrete picture of how microlocal ideas connect visible singularities, null bicharacteristics, and boundary symbol extraction.

What carries the argument

The central object is the quasimode, an approximate solution of the form $v_\lambda = \lambda^{-1/2} e^{i\lambda \Phi} a$, with phase $\Phi$ chosen so that the principal symbol of the operator vanishes (the eikonal equation), and amplitude $a$ built recursively by transport equations so that the leftover error is $O(\lambda^{-N})$ in $L^2$. For the wave equation the phase is $t - x_n$, concentrating the solution along a line; for the conductivity equation the phase has imaginary part equal to distance to the boundary, so $e^{i\lambda\Phi}$ decays exponentially inward. The amplitude is assembled from Taylor data at the boundary using Borel summation, which lets one prescribe the normal derivatives of $a$ so that all lower-order terms vanish to infinite order at the boundary. These quasimodes are then paired with the Alessandrini-style integral identity for the difference of Dirichlet-to-Neumann maps; taking $\lambda\to\infty$ extracts precisely the line integral or boundary Taylor coefficient sought.

What would settle it

Evaluate the limit formula (5.2) numerically for a specific curved domain, e.g. the unit ball in $\mathbb{R}^2$ with constant conductivity and the boundary point $x_0=(1,0)$, by constructing the quasimode phase and amplitude explicitly. If the limit $\lambda^k \int_{\Omega} \operatorname{dist}(x,\partial\Omega)^k f\,\nabla v\cdot\nabla\tilde v\,dx$ does not converge to $c_k \int_{\partial\Omega} f|\chi|^2\,dS$ for some smooth $f$, the curved-boundary step of Proposition 5.5 is false and the proof of Theorem 5.1 fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is expository: it claims that the microlocal paradigm for inverse problems can be substantiated by direct oscillatory testing, without developing a full pseudodifferential or Fourier integral operator calculus. Concretely, it proves three classical statements: the identity $R^*R = 4\pi|D|^{-1}$ for the Radon transform in the plane; the theorem that equal hyperbolic Dirichlet-to-Neumann maps imply $\int_\gamma q_1\,ds = \int_\gamma q_2\,ds$ for every maximal line segment $\gamma$ of length $<T$; and the theorem that equal elliptic Dirichlet-to-Neumann maps imply that $\gamma_1$ and $\gamma_2$ have identical Taylor series at every boundary point. The proofs work by inserting highly oscillatory approximate solutions into an integral identity and letting the frequency $\lambda$ tend to infinity, so that the interior terms either vanish or concentrate on a line or boundary point. In the boundary-determination argument, the exponential decay $e^{-\lambda x_n}$ of the quasimode isolates the boundary, and the limit formula (5.2) converts the inner product of gradients into a boundary integral of $f|\chi|^2$.

Load-bearing premise

The load-bearing premise is that the quasimode constructions deliver the stated small remainders, including the curved-boundary case of Proposition 5.5 where the phase is solved only to infinite order on the boundary and the transport estimates are merely sketched.

Editorial extensions

If this is right

The identity $R^*R = 4\pi|D|^{-1}$ implies the normal operator is an elliptic pseudodifferential operator of order $-1$, so the simple backprojection $R^*Rf$ recovers the singular support of $f$ and, with the filtered variant, inverts the transform exactly.
If the wave-equation boundary measurements agree for time $T > \operatorname{diam}(\Omega)$, Theorem 4.1 plus injectivity of the X-ray transform forces $q_1 = q_2$; for shorter times only line integrals over segments of length $<T$ are determined.
Equality of conductivity boundary measurements forces the two conductivities to agree to all orders at the boundary, so a real-analytic conductivity is uniquely determined by the Dirichlet-to-Neumann map.
The same oscillatory-testing scheme recovers two different data types from two different boundary maps: integrals of a potential over interior lines, and boundary Taylor coefficients of a conductivity.

Reading between the lines

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

The $\lambda^{-1/2}$ and $\lambda^{-N}$ remainder rates visible in the quasimode proofs suggest quantitative versions: how many boundary measurements at which frequencies are needed to recover a prescribed number of Taylor coefficients of a conductivity. The notes do not address this.
The same construction should apply to other elliptic boundary value problems whose Dirichlet-to-Neumann map has a scalar principal symbol, such as magnetic Schrödinger equations, yielding boundary Taylor determination of additional coefficients. This is an extrapolation from the method, not a claim of the notes.
For the wave-equation problem, replacing straight line segments by geodesics of a Riemannian metric should convert Theorem 4.1 into a geodesic X-ray transform recovery statement, matching the paper's remark that microlocal analysis is used in seismic imaging.
The visible-singularity dichotomy for limited-angle Radon data is an instance of a more general principle: whenever the forward map is a Fourier integral operator, the stable information is exactly the intersection of the data with the canonical relation. Extending the dichotomy to nonlinear inverse problems is a natural test.

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

The notes are a pleasant, honest overview of the microlocal paradigm in inverse problems, and they do the Radon and Calderón sections well. The exposition of visible singularities, the FIO viewpoint, and the boundary-determination proof via complex-phase quasimodes are all clear and correctly attributed to the classical literature. If you want a compact entry point into these ideas, this is a good place to look.

But there's a real problem in Section 4. The proof of Proposition 4.4(4.4) is wrong as printed. The forward and terminal solutions are both constructed with the same phase φ = t − x_n, so their product contains e^{2iλφ} = e^{4iλw} after the change of variables. When you scale w = εσ, the phase becomes λ^{1−1/(n+8)}σ, which oscillates and forces the integral to zero by Riemann–Lebesgue, not the claimed line integral. The sentence "since φ and a0 are independent of q, the same argument as above proves (4.4)" is exactly where the argument breaks: the same phase is what produces the oscillation, and the O(λ^{-1/2}) remainders are too small to rescue the limit. The theorem itself is classical, so the conclusion is true, but this particular elementary proof does not establish it. The fix is straightforward—use opposite phases or conjugate one solution—but as written it's a gap in the main proof of Theorem 4.1.

That weakness is contained to one section. The Radon normal-operator computation is clean, and the Calderón boundary-determination proof is solid, modulo the expected handwave for the curved-boundary case. I also noticed a harmless factor-of-two issue in the change of variables for (4.3), but it's absorbed into the limit and doesn't affect correctness.

Where does that leave the paper? It's genuinely useful as a survey, and the three classical theorems are correctly stated. But the notes currently deliver an invalid proof for one of the three advertised results. A serious referee would send it back for revision, not desk-reject it outright. The error is fixable, the rest is good, and the intended audience (graduate students, researchers new to microlocal inverse problems) would still benefit after a correction. If this were submitted to a journal, I'd recommend peer review with the request that the author fix (4.4).

Referee Report

2 major / 3 minor

Summary. The manuscript is a set of lecture notes for a minicourse on applications of microlocal analysis to inverse problems. It gives informal introductions to pseudodifferential operators, wave front sets, and Fourier integral operators, and then presents three classical results: Theorem 3.3 (the normal operator of the planar Radon transform satisfies R*R = 4π|D|^{-1}), Theorem 4.1 (equality of hyperbolic Dirichlet-to-Neumann maps implies equality of line integrals of the potential over maximal segments of length < T, due to Rakesh-Symes), and Theorem 5.1 (equality of elliptic Dirichlet-to-Neumann maps implies equality of Taylor series of conductivities at the boundary, due to Kohn-Vogelius and Sylvester-Uhlmann). The exposition is deliberately elementary: the theorems are proved by quasimode/WKB constructions rather than by the full microlocal calculus, and all results are attributed to their classical sources.

Significance. If the elementary proofs are valid, the notes fulfill a useful pedagogical role: they connect the microlocal point of view to concrete inverse problems while keeping the technical machinery relatively light. Theorem 3.3 is proved by a clean Fourier-slice and polar-coordinates computation, and the integral identities in Lemmas 4.3 and 5.4 are standard and correctly derived. Theorems 4.1 and 5.1 are classical, so the value of the manuscript is primarily expository: it demonstrates how highly oscillatory test functions can replace full FIO calculus in these examples. However, the exposition currently contains a load-bearing gap in Section 4 and an under-supported step in Section 5, so the advertised elementary route is not yet fully established as written.

major comments (2)

[Section 4, Proposition 4.4, Eq. (4.4)] The claimed limit in (4.4) is not proved by the preceding construction and, as printed, is false. Both u and \tilde u are constructed with the same phase e^{iλ(t-x_n)} and with amplitudes supported where w=(t-x_n)/2=O(ε), ε=λ^{-1/(n+8)}. After the change of variables (4.8) and the scaling w=εσ, the product u\tilde u contains the oscillatory factor e^{4iλεσ} (up to a constant in the exponent), and since λε=λ^{1-1/(n+8)}→∞, the Riemann-Lebesgue lemma forces the integral to tend to 0 for generic test functions ψ, not to ∫ ψ(γ(s),s)ds. The sentence 'Since φ and a0 are independent of q, the same argument as above proves (4.4)' does not repair the problem: using the same phase is exactly what creates the oscillation, and the correction term r is only O(λ^{-1/2}), too small to produce a nonzero limit. The proof of Theorem 4.1 uses (4.4) at the limiting step after Eq. (4.5), so the proof of Theorem 4.1 as written has a load-bearing gap. This is an internal gap in the exposition rather than a false theorem, since Theorem 4.1 is classical and cited, and the gap is repairable (for example by complex-conjugating one factor or using the opposite phase -t+x_n for the terminal solution), but the manuscript must be corrected here. A related inconsistency is that Eq. (4.5) writes u1 u2 instead of u1 \bar u2 as in Lemma 4.3.
[Section 5, Proposition 5.5, final paragraph] The curved-boundary case is the part of the proof of Theorem 5.1 that applies to the actual geometry of the theorem, and it is not proved in the manuscript. In boundary normal coordinates the conductivity equation becomes ∇·(γ A∇u)=0 with a nonconstant positive matrix A, and the eikonal equation p_2(x,∇Φ)=0 cannot be solved globally by the affine phase used in the flat case. The text states that the equation is solved only to infinite order on {x_n=0} and that the transport and error machinery proceeds 'in a similar way as above' without details. Since formula (5.2) and the boundary limit in the proof of Theorem 5.1 depend on the exponential decay and the structure of the approximate solution, this omission is load-bearing. The theorem itself is classical ([KV84], [SU88]) and the unpublished draft [FSU] is cited, but the paper's advertised elementary proof needs either a complete argument for the curved-boundary construction or an explicit reference for this step.

minor comments (3)

[Section 4, proof of Theorem 4.1] In the sentence introducing u2, the manuscript writes 'with u1 = ∂tu1 = 0 on {t=T}'; this should be u2 = ∂tu2 = 0 on {t=T}.
[Section 5, Eq. (5.2) and proof of Theorem 5.1] Lemma 5.4 and the proof of Theorem 5.1 use the conjugate ∇\bar u2 and ∇\bar v2, but formula (5.2) in Proposition 5.5 is stated with ∇v · ∇\tilde v without a conjugate. The notation should be made consistent, since the quasimodes are complex-valued.
[References] The proof of Theorem 5.1 relies on the unpublished draft [FSU] for the quasimode construction; for a self-contained published account, the manuscript should cite [KV84] and [SU88] explicitly at the point of Proposition 5.5 as well as in the statement of Theorem 5.1.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the proofs are self-contained and the only self-citation is a non-load-bearing exposition credit.

full rationale

Each derivation is carried out against an external mathematical benchmark rather than assuming its own conclusion. Section 3 proves the normal-operator formula R*R = 4π|D|^{-1} directly from the Fourier slice theorem and Parseval, with no circular input. Section 4 proves Theorem 4.1 by first establishing the integral identity (Lemma 4.3), then constructing WKB/quasimode solutions in Proposition 4.4, and finally taking λ→∞; the theorem is attributed to Rakesh–Symes [RS88], and the proof does not use the equality of the DN maps except in the intended way. Section 5 proves boundary determination through the Alessandrini identity, explicit oscillatory quasimodes, and Borel summation; the result is attributed to Kohn–Vogelius [KV84] and Sylvester–Uhlmann [SU88], and again the proof is self-contained rather than imported. The only self-citation is the sentence 'The treatment in this section follows [FSU]' at the start of Section 5, where FSU is an unpublished draft by Feldman, Salo and Uhlmann. This is not load-bearing: Theorem 5.1 is explicitly attributed to KV84/SU88, and the present notes supply the proof, including the flattening argument and the approximate-phase construction. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation. A possible technical defect in the printed formula (4.4) involving the phase of the two quasimodes would be a correctness gap in an exposition of a classical theorem, not a circularity, since the proof does not reduce to its conclusion. Overall, the paper has at most a harmless self-citation and is otherwise derivatively self-contained against classical external results.

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

The paper is a review of classical results, so the ledger is light. There are no free parameters fitted to data and no invented entities; the choices ε(λ) = λ^{-1/(n+8)} (Proposition 4.4) and the cutoff function ψ (Section 1) are proof devices that vanish in the relevant limits and do not affect the statements. The theorems rest on standard well-posedness for wave and elliptic equations (Evans), the Borel summation lemma (Hörmander), and the classical microlocal facts of Sections 1-2 (Hörmander), all cited in the text. The microlocal background is the largest unpaid input, but it is standard, properly referenced, and mostly used for motivation and interpretation rather than for the main proofs.

assumptions (5)

domain assumption Well-posedness of the wave equation with Dirichlet boundary data, plus the energy estimate bound for the correction term in the quasimode construction.
Invoked in the proof of Proposition 4.4 to bound the remainder r, with ‖r‖_{L∞((0,T);H^1(Ω))} ≲ ‖F_λ‖_{L^2((0,T);L^2(Ω))}; cited to Evans [Ev10, Theorems 5 and 7 in §7.2.3]. Without this estimate the concentration limits (4.3)-(4.4) would not follow.
domain assumption Well-posedness, uniqueness and energy estimates for the elliptic boundary value problem div(γ∇u) = 0 with u|∂Ω = f.
Used in Section 5 to define the Dirichlet-to-Neumann map Λ_γ and to bound the correction terms r_j by O(λ^{-N}) in the proof of Theorem 5.1; cited to Evans [Ev10, Section 6.2].
standard math Borel summation lemma: given a sequence of Taylor data at a hypersurface, there exists a smooth compactly supported function realizing those data to infinite order.
Used in the proof of Proposition 5.5 to construct amplitudes b_{-j} whose transport equations hold to infinite order at {x_n = 0}; cited to Hörmander [Hö85, Theorem 1.2.6].
standard math Standard microlocal facts: ΨDO symbol calculus and composition, pseudolocality, elliptic regularity, FIO canonical relations, and propagation of singularities along null bicharacteristics.
Sections 1-2 state these (Theorems 1.2, 2.3, 2.4, 2.7) with 'no proofs whatsoever', citing Hörmander [Hö85]. They motivate the paradigm but are not used in the elementary proofs of Theorems 4.1 and 5.1.
domain assumption The Radon transform R is an elliptic FIO of order -1/2 with the stated singularity correspondence.
Theorem 3.6 is stated without proof ('We will not spell out the precise relationship here'), attributed to Guillemin [Gu75]; it supports the visible/non-visible singularity dichotomy in Section 3.2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Applications of microlocal analysis to inverse problems." pith.science (2026). https://pith.science/paper/M5CGXXVD

@misc{pith2026190803041,
  author       = {Pith},
  title        = {Pith review of: Applications of microlocal analysis to inverse problems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M5CGXXVD}},
  note         = {Machine review of arXiv:1908.03041}
}

read the original abstract

These are lecture notes for a minicourse on applications of microlocal analysis in inverse problems, given in Helsinki and Shanghai in June 2019.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 21 canonical work pages

[1]

Pure Appl

J.J.\ Kohn, L.\ Nirenberg, An algebra of pseudo-differential operators , Comm. Pure Appl. Math. 18 (1965), 269--305

work page 1965
[2]

2nd edition, AMS, 2010

L.C.\ Evans, Partial differential equations. 2nd edition, AMS, 2010

work page 2010
[3]

Book in progress (draft available on request)

J.\ Feldman, M.\ Salo, G.\ Uhlmann, The Calder\'on problem - an introduction to inverse problems. Book in progress (draft available on request)

work page
[4]

report, MIT, 1975

V.\ Guillemin, Some remarks on integral geometry, Tech. report, MIT, 1975

work page 1975
[5]

I, Acta Math

L.\ H\"ormander, Fourier integral operators. I, Acta Math. 127 (1971), 79--183

work page 1971
[6]

Springer-Verlag, Berlin Heidelberg, 1983--1985

L.\ H\"ormander, The analysis of linear partial differential operators, vols.\ I--IV. Springer-Verlag, Berlin Heidelberg, 1983--1985

work page 1983
[7]

R.\ Ramlay, O.\ Scherzer), de Gruyter, 2019

J.\ Ilmavirta, F.\ Monard, Integral geometry on manifolds with boundary and applications , chapter in The Radon transform: the first 100 years and beyond (eds. R.\ Ramlay, O.\ Scherzer), de Gruyter, 2019

work page 2019
[8]

Monographs and Surveys in Pure and Applied Mathematics 123, Chapman Hall/CRC, 2001

A.\ Katchalov, Y.\ Kurylev, M.\ Lassas, Inverse boundary spectral problems. Monographs and Surveys in Pure and Applied Mathematics 123, Chapman Hall/CRC, 2001

work page 2001

Show all 21 references

[9]

Pure Appl

R.\ Kohn, M.\ Vogelius, Determining conductivity by boundary measurements, Comm. Pure Appl. Math. 37 (1984), 289--298

1984
[10]

V.\ Krishnan, E.T.\ Quinto, Microlocal analysis in tomography, chapter in Handbook of Mathematical Methods in Imaging (ed.\ O.\ Scherzer), 2015

2015
[11]

B.\ Sirakov, P.\ Ney de Souza), vol

M.\ Lassas, Inverse problems for linear and non-linear hyperbolic equations , in Proceedings of ICM 2018 (eds. B.\ Sirakov, P.\ Ney de Souza), vol. III

2018
[12]

E.T.\ Quinto, An introduction to X-ray tomography and Radon transforms, Proceedings of Symposia in Applied Mathematics 63 (2006), 1--23

2006
[13]

PDE 13 (1988), no

Rakesh, W.W.\ Symes, Uniqueness for an inverse problem for the wave equation, Comm. PDE 13 (1988), no. 1, 87--96

1988
[14]

P.\ Stefanov, Y.\ Yang, The inverse problem for the Dirichlet-to-Neumann map on Lorentzian manifolds, Analysis & PDE 11 (2018), 1381--1414

2018
[15]

Pure Appl

J.\ Sylvester, G.\ Uhlmann, Inverse boundary value problems at the boundary - continuous dependence, Comm. Pure Appl. Math. 41 (1988), 197--219

1988
[16]

G.\ Uhlmann, Inverse problems: seeing the unseen , Bull. Math. Sci. 4 (2014), 209--279

2014
[17]

Duoandikoetxea: Fourier analysis

J. Duoandikoetxea: Fourier analysis. Graduate Studies in Mathematics 29 . AMS, Providence Rhode Island, 2001

2001
[18]

Rudin: Functional analysis

W. Rudin: Functional analysis. McGraw-Hill, New York, 1973

1973
[19]

Schwartz: Th\'eorie des distributions

L. Schwartz: Th\'eorie des distributions. Hermann, Paris, 1966

1966
[20]

Stein, R

E. Stein, R. Sharkarchi: Fourier analysis: an introduction. Princeton Lectures in Analysis I . Princeton University Press, Princeton, 2003

2003
[21]

Strichartz: A guide to distribution theory

R. Strichartz: A guide to distribution theory. Studies in Advanced Mathematics. CRC Press, Boca Raton, 1994

1994

Pith tools

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

[1] [1]

Pure Appl

J.J.\ Kohn, L.\ Nirenberg, An algebra of pseudo-differential operators , Comm. Pure Appl. Math. 18 (1965), 269--305

work page 1965

[2] [2]

2nd edition, AMS, 2010

L.C.\ Evans, Partial differential equations. 2nd edition, AMS, 2010

work page 2010

[3] [3]

Book in progress (draft available on request)

J.\ Feldman, M.\ Salo, G.\ Uhlmann, The Calder\'on problem - an introduction to inverse problems. Book in progress (draft available on request)

work page

[4] [4]

report, MIT, 1975

V.\ Guillemin, Some remarks on integral geometry, Tech. report, MIT, 1975

work page 1975

[5] [5]

I, Acta Math

L.\ H\"ormander, Fourier integral operators. I, Acta Math. 127 (1971), 79--183

work page 1971

[6] [6]

Springer-Verlag, Berlin Heidelberg, 1983--1985

L.\ H\"ormander, The analysis of linear partial differential operators, vols.\ I--IV. Springer-Verlag, Berlin Heidelberg, 1983--1985

work page 1983

[7] [7]

R.\ Ramlay, O.\ Scherzer), de Gruyter, 2019

J.\ Ilmavirta, F.\ Monard, Integral geometry on manifolds with boundary and applications , chapter in The Radon transform: the first 100 years and beyond (eds. R.\ Ramlay, O.\ Scherzer), de Gruyter, 2019

work page 2019

[8] [8]

Monographs and Surveys in Pure and Applied Mathematics 123, Chapman Hall/CRC, 2001

A.\ Katchalov, Y.\ Kurylev, M.\ Lassas, Inverse boundary spectral problems. Monographs and Surveys in Pure and Applied Mathematics 123, Chapman Hall/CRC, 2001

work page 2001