The Muscalu–Tao–Thiele theorem on multi-linear operators given by singular multipliers established Hölder-range LpL^p bounds for (n−1)(n-1)-linear multiplier operators whose multiplier satisfies symbol estimates outside a kk-dimensional non-degenerate subspace Γ′\Gamma' of the frequency hyperplane Γ={ξ1+⋯+ξn=0}\Gamma=\{\xi_1+\dots+\xi_n=0\}, in the rank regime k<n/2k<n/2, unifying the Coifman–Meyer paraproduct theory (k=0k=0) and the Lacey–Thiele bilinear Hilbert transform (n=3n=3, k=1k=1). The source paper explicitly leaves open the analogue for k≥n/2k\ge n/2 — the regime of the trilinear Hilbert transform (n=4n=4, k=2k=2), for which no LpL^p bounds are known and for which the authors state that new methods “would probably require radically different techniques than the ones developed to date” — and the behavior of the bounds as Γ′\Gamma' degenerates toward configurations failing the non-degeneracy condition. The rank threshold stands: a March 2025 work on fractional-rank multipliers confines itself strictly below rank n/2n/2 and records that no bound is known for the triangular Hilbert transform (rank n/2n/2) and only partial progress for the trilinear Hilbert transform. Progress toward the open core includes Wiener-algebra inputs, dyadic models with one structured function, logarithmic power-saving cancellation estimates, curvature (non-flat) variants, and a low-exponent unboundedness result (unboundedness for input exponents close to 11, with quasi-Banach output); uniform bounds under degeneration have been settled for the bilinear Hilbert transform family (Grafakos–Li; Uraltsev–Warchalski; Fraccaroli–Saari–Thiele) but not in general multilinear generality. The problem — LpL^p bounds for multipliers singular on subspaces with k≥n/2k\ge n/2, with the trilinear Hilbert transform as benchmark, together with the quantitative geometric dependence of the bounds on Γ′\Gamma' — remains open.

Background

Multi-linear multiplier operators. Fix an integer n≥2n\ge 2 and let

Γ={ξ∈Rn:ξ1+⋯+ξn=0}\Gamma=\{\xi\in\mathbb{R}^n:\xi_1+\dots+\xi_n=0\}

be the frequency hyperplane. A function mm on Γ\Gamma, called a multiplier, defines an (n−1)(n-1)-linear operator TmT_m on functions on R\mathbb{R} through the nn-linear form

Λm(f1,…,fn)=∫δ(ξ1+⋯+ξn) m(ξ) f^1(ξ1)⋯f^n(ξn) dξ,\Lambda_m(f_1,\dots,f_n)=\int \delta(\xi_1+\dots+\xi_n)\,m(\xi)\,\hat f_1(\xi_1)\cdots\hat f_n(\xi_n)\,d\xi,

where f^\hat f denotes the Fourier transform and Λ(f1,…,fn)=∫RTm(f1,…,fn−1)(x) fn(x) dx\Lambda(f_1,\dots,f_n)=\int_{\mathbb{R}} T_m(f_1,\dots,f_{n-1})(x)\,f_n(x)\,dx; the case n=3n=3 (bilinear operators) is typical. The target of the theory is the Hölder-type bound

Tm:Lp1×⋯×Lpn−1→Lpn′T_m:L^{p_1}\times\dots\times L^{p_{n-1}}\to L^{p_n'}

for 1<pi≤∞1<p_i\le\infty, 1/(n−1)<pn′<∞1/(n-1)<p_n'<\infty and 1/p1+⋯+1/pn−1=1/pn′1/p_1+\dots+1/p_{n-1}=1/p_n'; the target exponent pn′p_n' may lie below 11, which is meaningful for these operators.

When mm is a symbol of order zero, smooth on Γ\Gamma away from the origin (∣∂αm(ξ)∣≲∣ξ∣−∣α∣|\partial^\alpha m(\xi)|\lesssim|\xi|^{-|\alpha|}), TmT_m is a multi-linear paraproduct — for m≡1m\equiv 1 it is pointwise multiplication, so the target bound generalizes Hölder's inequality from products to paraproducts — and the classical Coifman–Meyer theory gives the above bounds. Many natural operators are not of this type: their multipliers are smooth not off the origin but off a lower-dimensional linear subspace Γ′⊂Γ\Gamma'\subset\Gamma on or near which the multiplier is constant or jumps. Such an mm satisfies symbol estimates with respect to the distance to Γ′\Gamma',

∣∂αm(ξ)∣≲dist(ξ,Γ′)−∣α∣,|\partial^\alpha m(\xi)|\lesssim\mathrm{dist}(\xi,\Gamma')^{-|\alpha|},

and TmT_m is then a singular multi-linear operator with a large group of modulation symmetries (invariance under translating the frequency variables along Γ′\Gamma'). The model example is the bilinear Hilbert transform

B(f1,f2)(x)=p.v.∫f1(x−t)f2(x+t) dtt,B(f_1,f_2)(x)=\mathrm{p.v.}\int f_1(x-t)f_2(x+t)\,\frac{dt}{t},

whose multiplier πi sgn(ξ2−ξ1)\pi i\,\mathrm{sgn}(\xi_2-\xi_1) is singular along a line Γ′⊂Γ⊂R3\Gamma'\subset\Gamma\subset\mathbb{R}^3 (here n=3n=3 and k=dim⁡Γ′=1k=\dim\Gamma'=1); Lacey and Thiele proved its boundedness in the full Hölder range with the sharpened condition pn′>2/3p_n'>2/3.

Muscalu, Tao and Thiele unified the paraproduct and bilinear Hilbert transform regimes. Their Theorem 1.1 states: if Γ′\Gamma' is a subspace of dimension kk with 0≤k<n/20\le k<n/2 which is non-degenerate in the sense that for every 1≤i1<⋯<ik≤n1\le i_1<\dots<i_k\le n the space Γ′\Gamma' is a graph over the variables ξi1,…,ξik\xi_{i_1},\dots,\xi_{i_k} (its projection onto those kk coordinates is a bijection), and mm satisfies the symbol estimates above on Γ\Gamma, then the Hölder-range bounds hold, together with a dimension-sensitive refinement of the admissible exponents that improves as kk decreases. The proof requires a frequency-translation invariant time-frequency analysis — wave-packet “tiles” with arbitrary frequency location, organized into collections called trees — since for k>0k>0 no frequency origin is distinguished. In higher-dimensional generalizations (functions on Rd\mathbb{R}^d, singular subspace of dimension mm) the governing parameter is the rank m/dm/d, and the known generalizations cover only rank <n/2<n/2; the same threshold appears in the translation-invariant case, where TmT_m reduces to a Calderón–Zygmund convolution-type operator (a standard translation-invariant singular integral) on the orthogonal complement of Γ′\Gamma' in Γ\Gamma, bounded when n−1≤2 dim⁡(Γ′⊥∩Γ)n-1\le 2\,\dim(\Gamma'^{\perp}\cap\Gamma) — the same threshold k<n/2k<n/2, since dim⁡(Γ′⊥∩Γ)=n−1−k\dim(\Gamma'^{\perp}\cap\Gamma)=n-1-k.

Two limits of this theorem were explicitly flagged by the authors as open directions. First, the rank threshold. The trilinear Hilbert transform

H3(f1,f2,f3)(x)=p.v.∫f1(x−t)f2(x+t)f3(x+2t) dttH_3(f_1,f_2,f_3)(x)=\mathrm{p.v.}\int f_1(x-t)f_2(x+t)f_3(x+2t)\,\frac{dt}{t}

has n=4n=4 and k=2=n/2k=2=n/2, which violates the strict inequality and is not covered; the authors write that an analogue of their theorem when the rank condition fails “would probably require radically different techniques than the ones developed to date”. The only positive statement in this direction in the source is a Wiener-algebra substitute (Theorem 13.1): if enough inputs lie in the Wiener algebra A=F−1L1A=\mathcal{F}^{-1}L^1 of functions with integrable Fourier transform, then bounds hold — for instance H3H_3 maps Lp×Lq×A→LrL^p\times L^q\times A\to L^r whenever 1<p,q≤∞1<p,q\le\infty, 1/p+1/q=1/r1/p+1/q=1/r and 2/3<r<∞2/3<r<\infty. It is conjectured that the kk-linear Hilbert transforms HkH_k (here the index kk denotes the arity of the operator, i.e. its number of inputs, and is not the rank k=dim⁡Γ′k=\dim\Gamma' used elsewhere) satisfy the Hölder-range bounds for all 1<p1,…,pk<∞1<p_1,\dots,p_k<\infty with 1/p1+⋯+1/pk=1/p1/p_1+\dots+1/p_k=1/p in the sharp output range, whose threshold is p>2/3p>2/3 for k=2k=2 and p>1/2p>1/2 for k=3k=3; this is known for k≤2k\le 2 and open for k≥3k\ge 3. Other prominent operators at rank n/2n/2 include the triangular Hilbert transform — the trilinear singular-integral form p.v.∫R3F0(x,y) F1(y,z) F2(z,x) (x+y+z)−1 d(x,y,z)\mathrm{p.v.}\int_{\mathbb{R}^3}F_0(x,y)\,F_1(y,z)\,F_2(z,x)\,(x+y+z)^{-1}\,d(x,y,z) on functions on R2\mathbb{R}^2, which specializes to several well-studied objects (among them the Carleson operator, the maximal operator governing pointwise convergence of Fourier series, and the bilinear Hilbert transform) — and the simplex multipliers, the multi-linear singular forms associated with the configurations of a simplex, of which the triangular Hilbert transform is the three-function case. Second, degeneracy. Theorem 1.1 requires the graph-type non-degeneracy of Γ′\Gamma'; the authors state that it would be interesting to study the behavior of the bounds as Γ′\Gamma' degenerates in the sense of Theorem 1.1 — for instance as the singular line of a bilinear Hilbert transform approaches a position where it fails to be a graph over some coordinate variable — and note that this issue is not treated in the paper.

Problem Statement

The problem is the two-fold completion of the Muscalu–Tao–Thiele LpL^p theory for multi-linear operators given by singular multipliers, as explicitly left open by the source; the two components share one objective, namely the dependence of the mapping bounds on the geometry of the singular subspace Γ′\Gamma'.

(A) Beyond the rank threshold. Establish, or disprove, bounds of the form

Tm:Lp1×⋯×Lpn−1→Lpn′T_m:L^{p_1}\times\dots\times L^{p_{n-1}}\to L^{p_n'}

in the Hölder range 1<pi≤∞1<p_i\le\infty, 1/2<pn′<∞1/2<p_n'<\infty, 1/p1+⋯+1/pn−1=1/pn′1/p_1+\dots+1/p_{n-1}=1/p_n' for multipliers mm satisfying the symbol estimates ∣∂αm(ξ)∣≲dist(ξ,Γ′)−∣α∣|\partial^\alpha m(\xi)|\lesssim\mathrm{dist}(\xi,\Gamma')^{-|\alpha|} on Γ\Gamma, where Γ′⊂Γ\Gamma'\subset\Gamma is a non-degenerate linear subspace of dimension k≥n/2k\ge n/2 in the sense of Theorem 1.1. The bound 1/2<pn′<∞1/2<p_n'<\infty is the conjectured sharp threshold for this rank-n/2n/2 regime, sharpening the generic (automatic) Hölder condition 1/(n−1)<pn′<∞1/(n-1)<p_n'<\infty; the sub-range (1/(n−1),1/2)(1/(n-1),1/2) is already known to fail in the trilinear Hilbert transform benchmark and is not part of the conjectured target, and no sharp threshold is known for general nn. The benchmark instance is the trilinear Hilbert transform (n=4n=4, k=2k=2), conjectured to satisfy the Hölder-range bounds for all 1<p1,p2,p3<∞1<p_1,p_2,p_3<\infty and 1/2<p<∞1/2<p<\infty with 1/p1+1/p2+1/p3=1/p1/p_1+1/p_2+1/p_3=1/p; no LpL^p bound with three general inputs is currently known. The triangular Hilbert transform and the simplex multipliers are further rank-n/2n/2 instances. Prior results in this direction require regularizing extra structure — curvature, or restrictions on the input algebra — or lie outside the conjectured exponent range.

(B) Degeneracy and uniformity. Determine how the bounds of Theorem 1.1 — and of its translation-invariant extension — depend quantitatively on the geometry of Γ′\Gamma': in particular, whether the operator norms and admissible exponent ranges remain uniform as Γ′\Gamma' degenerates toward configurations that fail the non-degeneracy (graph) condition, for the natural degenerating families (such as the family of bilinear Hilbert transforms BHTβ(f,g)(x)=p.v.∫f(x−t) g(x+βt) dt/t\mathrm{BHT}_\beta(f,g)(x)=\mathrm{p.v.}\int f(x-t)\,g(x+\beta t)\,dt/t, 0<β≤10<\beta\le 1, whose singular line becomes coordinate-degenerate as β→0\beta\to 0, and its multi-linear analogues), and more generally whether an analogue of Theorem 1.1 holds with constants depending only on quantitative non-degeneracy parameters of Γ′\Gamma' rather than on its exact position.

(A) has the trilinear Hilbert transform conjecture as its benchmark instance, so settling (A) settles that conjecture. Component (B) is known at present only for bilinear families; its open core is the multi-linear (k≥2k\ge 2) degeneration regime and the quantitative geometric dependence of the general theorem. What remains open across both components is the behaviour of the mapping bounds for multi-linear singular multipliers as a function of the dimension and position of the singular subspace, on both sides of the rank threshold and at the degenerate boundary of the non-degeneracy condition.

References

  1. 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 (arXiv:math/9910039, https://arxiv.org/abs/math/9910039)
  2. R. R. Coifman, Y. Meyer. Wavelets: Calderón–Zygmund and multilinear operators. Cambridge Studies in Advanced Mathematics 48, Cambridge University Press, 1997.
  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. Michael Lacey, Christoph Thiele. On Calderon's conjecture. Annals of Mathematics (2) 149 (1999), no. 2, 475–496. DOI 10.2307/120971. https://doi.org/10.2307/120971
  5. C. E. Kenig, E. M. Stein. Multilinear estimates and fractional integration. Mathematical Research Letters 6 (1999), no. 1, 1–15. DOI 10.4310/mrl.1999.v6.n1.a1. https://doi.org/10.4310/mrl.1999.v6.n1.a1
  6. Ciprian Demeter, Malabika Pramanik, Christoph Thiele. Multilinear singular operators with fractional rank. Pacific Journal of Mathematics 246 (2010), no. 2, 293–324. DOI 10.2140/pjm.2010.246.293. https://doi.org/10.2140/pjm.2010.246.293
  7. Ciprian Demeter. Divergence of combinatorial averages and the unboundedness of the trilinear Hilbert transform. Ergodic Theory and Dynamical Systems 28 (2008), no. 5, 1453–1464. DOI 10.1017/s0143385707001101. https://doi.org/10.1017/s0143385707001101
  8. Jianghao Zhang. Non-Degenerate Multilinear Singular Multipliers with Fractional Rank. arXiv:2503.14635 (2025). https://arxiv.org/abs/2503.14635
  9. Terence Tao. Cancellation for the multilinear Hilbert transform. Collectanea Mathematica 67 (2016), no. 2, 191–206. DOI 10.1007/s13348-015-0162-y. https://doi.org/10.1007/s13348-015-0162-y
  10. Pavel Zorin-Kranich. Cancellation for the simplex Hilbert transform. Mathematical Research Letters 24 (2017), no. 2, 581–592. DOI 10.4310/mrl.2017.v24.n2.a16. https://doi.org/10.4310/mrl.2017.v24.n2.a16
  11. Polona Durcik, Vjekoslav Kovač, Christoph Thiele. Power-type cancellation for the simplex Hilbert transform. Journal d'Analyse Mathématique 139 (2019), no. 1, 67–82. DOI 10.1007/s11854-019-0052-4. https://doi.org/10.1007/s11854-019-0052-4
  12. Vjekoslav Kovač, Christoph Thiele, Pavel Zorin-Kranich. Dyadic triangular Hilbert transform of two general functions and one not too general function. Forum of Mathematics, Sigma 3 (2015), e25. DOI 10.1017/fms.2015.25. https://doi.org/10.1017/fms.2015.25
  13. Loukas Grafakos, Xiaochun Li. Uniform bounds for the bilinear Hilbert transforms, I. Annals of Mathematics (2) 159 (2004), no. 3, 889–933. DOI 10.4007/annals.2004.159.889. https://doi.org/10.4007/annals.2004.159.889
  14. Xiaochun Li. Uniform bounds for the bilinear Hilbert transforms, II. Revista Matemática Iberoamericana 22 (2006), no. 3, 1069–1126. DOI 10.4171/RMI/483. https://doi.org/10.4171/rmi/483
  15. Gennady Uraltsev, Michał Warchalski. The full range of uniform bounds for the bilinear Hilbert transform. arXiv:2205.09851 (2022). https://arxiv.org/abs/2205.09851
  16. Marco Fraccaroli, Olli Saari, Christoph Thiele. Uniform bounds for bilinear symbols with linear K-quasiconformally embedded singularity. Analysis & PDE 18 (2025), no. 9, 2293–2323. DOI 10.2140/apde.2025.18.2293. https://doi.org/10.2140/apde.2025.18.2293
  17. Xiaochun Li, Lechao Xiao. Uniform estimates for bilinear Hilbert transforms and bilinear maximal functions associated to polynomials. American Journal of Mathematics 138 (2016), no. 4, 907–962. DOI 10.1353/ajm.2016.0030. https://doi.org/10.1353/ajm.2016.0030
  18. Michael Christ, Polona Durcik, Joris Roos. Trilinear smoothing inequalities and a variant of the triangular Hilbert transform. Advances in Mathematics 388 (2021), 107863. DOI 10.1016/j.aim.2021.107863. https://doi.org/10.1016/j.aim.2021.107863
  19. Bingyang Hu, Victor Lie. On the curved Trilinear Hilbert transform. arXiv:2308.10706 (2023). https://arxiv.org/abs/2308.10706
  20. Bingyang Hu, Victor Lie. On the boundedness of the curved trilinear Hilbert transform and the curved n-linear maximal operator in the quasi-Banach regime. arXiv:2511.07373 (2025). https://arxiv.org/abs/2511.07373
  21. Michael Christ. On implicitly oscillatory quadrilinear integrals. arXiv:2204.03780 (2022). https://arxiv.org/abs/2204.03780