General singular symbols for the biest: $L^p$ boundedness beyond product structure and its necessity

The biest of Muscalu–Tao–Thiele is the trilinear Fourier multiplier on $\mathbb{R}$ with symbol $\chi_{\xi_1<\xi_2<\xi_3}$, singular simultaneously on the two adjacent hyperplanes $\xi_1=\xi_2$ and $\xi_2=\xi_3$; its Hölder-type $L^p$ estimates underlie WKB (Wentzel–Kramers–Brillouin) asymptotic expansions of eigenfunctions of one-dimensional Schrödinger operators. The paper that introduced it closed by deferring the natural generalization: prove theorems for more general multiplier symbols, where a product-structure condition in the sense of the bilinear Marcinkiewicz theory of Grafakos–Kalton was expected to occur. Jung later resolved the sufficiency direction: arbitrary product-structured symbols $m_1(\xi_1,\xi_2)\,m_2(\xi_2,\xi_3)$, each satisfying Marcinkiewicz-type derivative bounds relative to its own hyperplane, define operators bounded on the same range as the biest. What remains open is the boundary of that theorem: whether general symbols adapted only to the union of the two hyperplanes — in particular non-factoring symbols such as two-variable Marcinkiewicz symbols $a(\xi_1-\xi_2,\xi_2-\xi_3)$ — satisfy analogous bounds, and whether product structure is necessary, that is, whether an adapted but non-product symbol can be unbounded. The problem asks to determine the boundary of the biest $L^p$ theorem for the non-smooth trilinear multipliers with this two-scale singularity structure, together with the precise role and necessity of product structure.

Background

For Schwartz functions $f_1,\dots,f_n$ on $\mathbb{R}$ with Fourier transform $\hat f$, and a bounded measurable symbol $m$ on $\mathbb{R}^n$, the $n$-linear Fourier multiplier operator is

$$T_m(f_1,\dots,f_n)(x)=\int_{\mathbb{R}^n} m(\xi_1,\dots,\xi_n)\prod_{j=1}^n \hat f_j(\xi_j)\,e^{2\pi i x(\xi_1+\cdots+\xi_n)}\,d\xi.$$

The classical Coifman–Meyer paraproduct theory covers symbols that are smooth away from the origin with $|\partial^\alpha m(\xi)|\lesssim |\xi|^{-|\alpha|}$ and yields Hölder-type estimates $T_m:L^{p_1}\times\cdots\times L^{p_n}\to L^q$ whenever $1<p_1,\dots,p_n\le\infty$, $1\le q\le\infty$ and $1/p_1+\cdots+1/p_n=1/q$. Much of modern multilinear harmonic analysis concerns non-smooth symbols whose singular sets are larger than a point; there, boundedness depends on the geometry of the singularity and is proved by time-frequency analysis (wave-packet decompositions organized into trees, almost-orthogonality, and interpolation), the technique created by Lacey and Thiele for the bilinear Hilbert transform, whose symbol $\chi_{\xi_1<\xi_2}$ is singular on the hyperplane $\xi_1=\xi_2$. Gilbert and Nahmod then treated general bilinear symbols with singularities along cone edges and vertices, unifying the Coifman–Meyer and bilinear Hilbert transform theories, and Grafakos and Kalton analyzed bilinear multipliers of Marcinkiewicz type, whose derivatives are controlled by powers of the separate coordinate variables: they proved that classical Marcinkiewicz multipliers on $\mathbb{R}^{2n}$ do not always give bounded bilinear operators, found necessary and sufficient conditions for boundedness, and showed that a logarithmic modification of the Marcinkiewicz condition suffices. In the $n$-linear setting, Muscalu, Tao and Thiele proved $L^p$ estimates for symbols that are smooth away from a non-degenerate subspace $\Gamma\subset\mathbb{R}^n$ of dimension less than $(n+1)/2$ and satisfy $|\partial^\alpha m(\xi)|\lesssim \mathrm{dist}(\xi,\Gamma)^{-|\alpha|}$. The biest sits just outside that theorem. It is the trilinear multiplier $T$ on $\mathbb{R}$ with symbol $\chi_{\xi_1<\xi_2<\xi_3}=\chi_{\xi_1<\xi_2}\cdot\chi_{\xi_2<\xi_3}$, singular simultaneously on the two adjacent hyperplanes $\Gamma_1=\{\xi_1=\xi_2\}$ and $\Gamma_2=\{\xi_2=\xi_3\}$, and it can be viewed as a nested (iterated) pair of bilinear Hilbert transforms. The operator arose — through an observation of Mike Christ recorded in the source paper — in the Christ–Kiselev WKB (Wentzel–Kramers–Brillouin) analysis of generalized eigenfunctions of one-dimensional Schrödinger operators. After a Walsh-model prequel, Muscalu–Tao–Thiele proved that $T$ maps $L^{p_1}\times L^{p_2}\times L^{p_3}$ boundedly into $L^q$ whenever $1<p_1,p_2,p_3\le\infty$, $1\le q<\infty$, $1/p_1+1/p_2+1/p_3=1/q$ — more precisely on an open polytope region $\mathcal D$ of exponent quadruples defined in that paper, which also contains sub-Hölder outputs such as $L^2\times L^2\times L^2\to L^{2/3}$. The proof decomposes the symbol by cones into two pieces, each a nested composition treatable by time-frequency analysis, plus a smooth error that is singular only on the diagonal $\{\xi_1=\xi_2=\xi_3\}$ and hence falls under the earlier one-scale theorem. The paper then explicitly defers the natural generalization, remarking that modifications of the argument, as in the work of Gilbert–Nahmod, give theorems for more general multiplier symbols, but that a product structure condition, as discussed in the work of Grafakos–Kalton, seems to occur. The sufficiency direction of that expectation is now a theorem of Jung: every product-structured symbol $m(\xi)=m_1(\xi_1,\xi_2)\,m_2(\xi_2,\xi_3)$, with each factor $m_i$ smooth away from the line $\{\xi_i=\xi_{i+1}\}$ in the plane of its own two variables and satisfying the corresponding two-variable condition $|\partial^\alpha m_i(\xi)|\lesssim \mathrm{dist}(\xi,\{\xi_i=\xi_{i+1}\})^{-|\alpha|}$ for every multi-index $\alpha$, defines an operator bounded on the same range $\mathcal D$ as the biest, and finite sums of such products are equally bounded by linearity. What is not known is whether the product hypothesis can be removed, and whether it is in any sense necessary.

Problem Statement

Call a symbol $m$ on $\mathbb{R}^3$ adapted to the biest hyperplane pair $\Gamma_1\cup\Gamma_2$, where $\Gamma_1=\{\xi_1=\xi_2\}$ and $\Gamma_2=\{\xi_2=\xi_3\}$, if $m$ is bounded, smooth away from $\Gamma_1\cup\Gamma_2$, and satisfies the two-scale Marcinkiewicz-type estimates $|\partial^\alpha m(\xi)|\lesssim \mathrm{dist}(\xi,\Gamma_1\cup\Gamma_2)^{-|\alpha|}$ for every multi-index $\alpha$ (the direct analog, for this singular set, of the symbol conditions in the Coifman–Meyer, Gilbert–Nahmod, and Muscalu–Tao–Thiele tradition). Finite sums of Jung-type products $m_1(\xi_1,\xi_2)\,m_2(\xi_2,\xi_3)$, with each $m_i$ satisfying the analogous two-variable adaptation condition relative to the line $\{\xi_i=\xi_{i+1}\}$ in its two variables, are examples, as are two-variable Marcinkiewicz symbols $a(\xi_1-\xi_2,\xi_2-\xi_3)$ with $|\partial^\alpha a(u,v)|\lesssim |u|^{-\alpha_1}|v|^{-\alpha_2}$, which need not factor. The problem, left open by this remark of the biest paper, is to determine the boundary of the biest $L^p$ theorem inside this class, and thereby the precise role of product structure. It is a single prove-or-disprove question with one resolution standard, so that either branch settles it: prove, or disprove, that every symbol adapted to $\Gamma_1\cup\Gamma_2$ defines an operator bounded on the same range $\mathcal D$ as the biest — the open polytope region of exponent quadruples defined in the biest paper, comprising the Hölder tuples $1<p_1,p_2,p_3\le\infty$, $1\le q<\infty$ with $1/p_1+1/p_2+1/p_3=1/q$ together with sub-Hölder outputs such as $L^2\times L^2\times L^2\to L^{2/3}$ ($q=2/3<1$). A positive resolution proves boundedness for the full adapted class on $\mathcal D$, showing that product structure is not necessary; a negative resolution exhibits an adapted symbol that is not a finite sum of Jung-type products modulo symbols already known to be bounded (such as symbols singular only on the diagonal $\{\xi_1=\xi_2=\xi_3\}$, covered by the one-scale theorem of Muscalu–Tao–Thiele) and for which $T_m$ fails to be bounded for at least one exponent tuple in $\mathcal D$, confirming that adaptation to $\Gamma_1\cup\Gamma_2$ alone does not suffice.

References

1. Camil Muscalu, Terence Tao, Christoph Thiele, 'L^p estimates for the biest II. The Fourier case', Mathematische Annalen 329 (2004), no. 3, 427–461. DOI: 10.1007/s00208-003-0508-8, https://doi.org/10.1007/s00208-003-0508-8.
2. Camil Muscalu, Terence Tao, Christoph Thiele, 'L^p estimates for the biest I. The Walsh case', Mathematische Annalen 329 (2004), no. 3, 401–426. DOI: 10.1007/s00208-004-0518-1, https://doi.org/10.1007/s00208-004-0518-1.
3. Michael Lacey, Christoph Thiele, 'L^p estimates on the bilinear Hilbert transform for 2 < p < ∞', Annals of Mathematics (2) 146 (1997), no. 3, 693–724. DOI: 10.2307/2952458, https://doi.org/10.2307/2952458.
4. John E. Gilbert, Andrea R. Nahmod, 'Boundedness of bilinear operators with nonsmooth symbols', Mathematical Research Letters 7 (2000), no. 6, 767–778. DOI: 10.4310/mrl.2000.v7.n6.a9, https://doi.org/10.4310/mrl.2000.v7.n6.a9.
5. Loukas Grafakos, Nigel J. Kalton, 'The Marcinkiewicz multiplier condition for bilinear operators', Studia Mathematica 146 (2001), no. 2, 115–156. DOI: 10.4064/sm146-2-2, https://doi.org/10.4064/sm146-2-2.
6. Camil Muscalu, Terence Tao, Christoph Thiele, 'Multi-linear operators given by singular multipliers', Journal of the American Mathematical Society 15 (2002), no. 2, 469–496. DOI: 10.1090/S0894-0347-01-00379-4, https://doi.org/10.1090/S0894-0347-01-00379-4.
7. Michael Christ, Alexander Kiselev, 'WKB asymptotic behavior of almost all generalized eigenfunctions for one-dimensional Schrödinger operators with slowly decaying potentials', Journal of Functional Analysis 179 (2001), no. 2, 426–447. DOI: 10.1006/jfan.2000.3688, https://doi.org/10.1006/jfan.2000.3688.
8. Joeun Jung, 'Iterated trilinear Fourier integrals with arbitrary symbols', Journal of Functional Analysis 273 (2017), no. 10, 3027–3060. DOI: 10.1016/j.jfa.2017.07.005, https://doi.org/10.1016/j.jfa.2017.07.005.
9. Camil Muscalu, 'Paraproducts with flag singularities I. A case study', Revista Matemática Iberoamericana 23 (2007), no. 2, 705–742. DOI: 10.4171/rmi/510, https://doi.org/10.4171/rmi/510.
10. Camil Muscalu, Terence Tao, Christoph Thiele, 'Multi-linear multipliers associated to simplexes of arbitrary length', in: Advances in Analysis: The Legacy of Elias M. Stein, Princeton Mathematical Series 50, Princeton University Press, 2014, pp. 346–401. DOI: 10.1515/9781400848935-015, https://doi.org/10.1515/9781400848935-015.
11. Camil Muscalu, 'Some remarks on the n-linear Hilbert transform for n ≥ 4', Mathematical Research Letters 21 (2014), no. 5, 1155–1176. DOI: 10.4310/MRL.2014.V21.N5.A11, https://doi.org/10.4310/MRL.2014.V21.N5.A11.
12. Robert Kesler, 'Mixed estimates for degenerate multi-linear operators associated to simplexes', Journal of Mathematical Analysis and Applications 424 (2015), no. 1, 344–360. DOI: 10.1016/j.jmaa.2014.11.005, https://doi.org/10.1016/j.jmaa.2014.11.005.
13. Robert Kesler, 'Generic trilinear multipliers associated to degenerate simplexes', Collectanea Mathematica 70 (2019), no. 1, 153–186. DOI: 10.1007/s13348-018-0224-z, https://doi.org/10.1007/s13348-018-0224-z.
14. Robert Kesler, 'Unboundedness theorems for symbols adapted to large subspaces', Studia Mathematica 244 (2019), no. 2, 109–158. DOI: 10.4064/sm8750-7-2017, https://doi.org/10.4064/sm8750-7-2017.
15. Robert M. Kesler, 'L^p estimates for semi-degenerate simplex multipliers', Revista Matemática Iberoamericana 36 (2020), no. 1, 99–158. DOI: 10.4171/rmi/1123, https://doi.org/10.4171/rmi/1123.
16. Polona Durcik, Christoph Thiele, 'Singular Brascamp–Lieb: A Survey', in: Geometric Aspects of Harmonic Analysis, Springer INdAM Series, vol. 45, Springer, 2021, pp. 321–349. DOI: 10.1007/978-3-030-72058-2_9, https://doi.org/10.1007/978-3-030-72058-2_9.
17. Wei Dai, Guozhen Lu, 'L^p estimates for bilinear and multiparameter Hilbert transforms', Analysis & PDE 8 (2015), no. 3, 675–712. DOI: 10.2140/apde.2015.8.675, https://doi.org/10.2140/apde.2015.8.675.
18. Jing Qiao, Lu Zhang, 'L^p and H^p estimates for N-linear Fourier multipliers with flag structure', Mathematische Zeitschrift 310 (2025), no. 4, article 66. DOI: 10.1007/s00209-025-03771-1, https://doi.org/10.1007/s00209-025-03771-1.
19. A. Martina Neuman, 'L^2 × L^2 × L^2 → L^{2/3} boundedness for trilinear multiplier operator', preprint, arXiv:2004.08622 (2020), https://arxiv.org/abs/2004.08622.