The first eigenvalue of embedded minimal hypersurfaces in the unit sphere
Published 21 Mar 2026 in math.DG | (2603.20890v1)
Abstract: In this article, we prove that for an embedded minimal hypersurface Σ<sup>m in S<sup>m+1, the first eigenvalue λ<em>1 of the Laplacian operator on Σ satisfies: $$λ_1> \frac{m}{2}+G(m, |A|</em>{\max}, |A|<em>{\min} ) ,$$ where ∣A∣</em>max and ∣A∣<em>min denote the maximum and minimum of the norm of the second fundamental form on Σ, respectively; G(m,∣A∣</em>max,∣A∣<em>min) is a positive constant that depends only on m,∣A∣</em>max,∣A∣min. In particular, when the norm ∣A∣ of the second fundamental form is constant, we can obtain a gap depending only on m, i.e., $$λ_1>\frac{m}{2} \left(1+ c \right) ,$$ where c is a positive absolute constant. This improves the previous result of Choi and Wang \cite{chw1983first}, which gave λ1≥2m. Our result shows that one can skip proving Chern's conjecture to directly improve Choi-Wang's result. This also generalizes Tang and Yan's work \cite{tangyan2013isoparametric}. Based on the proof of the result above, using the lower bound of the Steklov eigenvalue, we prove that if the norm ∣A∣ of the second fundamental form is constant, then ∣A∣≤Volume(S<sup>m)C(m)Volume(Σ), where C(m) is a constant that depends only on m. This provides a uniform estimate for the scalar curvature of embedded minimal hypersurfaces with constant norm of the second fundamental form. Moreover, this may useful for Chern's problem.
The paper improves the Choi–Wang estimate from λ₁ ≥ m/2 to an explicit strict lower bound whose gap remains nonzero as maximum curvature grows.
The proof uses higher-order Reilly identities, third-derivative estimates, tubular-neighborhood traces, and boundary curvature inequalities to retain information discarded in earlier arguments.
For constant second-fundamental-form norm, the result gives λ₁ > (1/2+c)m with an asymptotic gap near 0.042m and connects Steklov bounds to volume-based curvature control.
Overview and main results
This paper by Yuhang Zhao establishes a quantitative improvement of the Choi–Wang lower bound for the first Laplacian eigenvalue of embedded minimal hypersurfaces in the unit sphere. For a compact minimal embedding F:Σm→Sm+1 that is not totally geodesic, the main theorem asserts
where ∣A∣max and ∣A∣min are the maximum and minimum of the norm of the second fundamental form. The bracketed quantity is positive, so this strictly improves the classical bound λ1≥m/2 of Choi and Wang (2603.20890), which had remained the best general result toward Yau's conjecture (λ1=m) despite a sequence of refinements by Zhao (Zhao, 2023), Duncan–Spruck–Sire, and Jiménez–Tapia–Zhou (Jiménez et al., 2024). A key structural point is that all previous gaps decay to zero as ∣A∣max→∞, and doublings provide infinitely many examples with arbitrarily large ∣A∣max; the new estimate retains a nonzero gap uniformly in ∣A∣max.
In the special case where ∣A∣ is constant (so λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],0), the gap becomes dimension-dependent only:
with an absolute constant λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],2; asymptotically as λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],3 the gap is approximately λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],4. This is significant because Tang and Yan proved Yau's conjecture for isoparametric hypersurfaces only via their classification; here the author shows one can bypass Chern's conjecture entirely and still improve Choi–Wang under constant λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],5, using only basic properties of minimal hypersurfaces.
Method: higher-order Reilly computation
The proof departs from Choi–Wang at a specific point. Their argument applies Reilly's formula to harmonic extensions λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],6 of a first eigenfunction λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],7 on the two regions λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],8 cut out by λ1>2m+48∣A∣maxm2−1[(1110−mm+1)∣A∣min+21113m−m+1m+2],9, obtaining
∣A∣max0
and discards the Hessian term. The present work instead computes ∣A∣max1 and ∣A∣max2 explicitly. Exploiting the ambient sphere, the author applies Reilly's formula to each component function ∣A∣max3 rather than performing a general tensorial Bochner computation, then integrates over both regions and adds the results so that boundary terms cancel. The resulting identity (Theorem 3 of the paper) relates third-order interior integrals ∣A∣max4 to boundary integrals involving ∣A∣max5, the normal derivatives of ∣A∣max6 and ∣A∣max7, and the pullback Hessians ∣A∣max8, ∣A∣max9.
Two auxiliary inputs complete the scheme:
Tubular neighborhood trace estimate: using Howard's rolling theorem, the rolling radius equals the focal distance ∣A∣min0, bounded below via Simons' inequality ∣A∣min1. A cutoff-function argument on the tubular neighborhood yields ∣A∣min2 with ∣A∣min3 arbitrarily small.
Boundary algebra: pointwise identities on ∣A∣min4 decompose ∣A∣min5 into pieces involving ∣A∣min6, the sums ∣A∣min7, differences of tangential Hessian data, and the term ∣A∣min8. Young-type inequalities with parameters ∣A∣min9 and a free positive function λ1≥m/20 absorb the cross terms; the constraint λ1≥m/21 (a Lagrange multiplier extremum under λ1≥m/22, λ1≥m/23) controls the quartic curvature terms.
Choosing λ1≥m/24 and minimizing a one-variable concave function λ1≥m/25 over λ1≥m/26 produces the final explicit gap. The author notes candidly that several estimates in the chain are rough, and poses whether the gap can be pushed close to λ1≥m/27.
A Steklov eigenvalue bound and uniform scalar curvature estimate
The second half of the paper derives a volume-controlled upper bound on λ1≥m/28 when it is constant. From the main computation with a different choice of λ1≥m/29 and λ1=m0, one obtains
λ1=m1
The left side dominates the first nonzero Steklov eigenvalue λ1=m2 of the summed Dirichlet-to-Neumann mapλ1=m3. The paper proves
λ1=m4
where λ1=m5. The proof glues the two harmonic extensions into a globally Lipschitz function on λ1=m6, establishes a mean value formula valid for Lipschitz functions — requiring an additional boundary integral over λ1=m7 absent from the smooth case — and closes via the monotonicity estimate λ1=m8, adapted from Brendle and Colding–Minicozzi to the spherical setting. Only volume growth of λ1=m9 enters, not finer geometry.
Combining the two bounds gives, for constant ∣A∣max→∞0,
∣A∣max→∞1
with ∣A∣max→∞2 expressed explicitly through ∣A∣max→∞3 and ∣A∣max→∞4. Together with the area bound ∣A∣max→∞5, this also yields two-sided control relating ∣A∣max→∞6 and volume. The implication for Chern's problem — whether ∣A∣max→∞7 universally when ∣A∣max→∞8 is constant — is direct: under the embedding hypothesis, a uniform volume upper bound would resolve it. The author frames volume as the more tractable quantity, though no such volume bound is established here.
Limitations and open questions
Several restrictions qualify the results. The eigenvalue theorem requires embeddedness throughout; immersed examples fall outside its scope. The gap, while uniform in ∣A∣max→∞9, shrinks like ∣A∣max0 unless ∣A∣max1 is comparable to ∣A∣max2, and the constant-∣A∣max3 case yields only the modest additive term ∣A∣max4 rather than anything approaching ∣A∣max5 — the author attributes this to rough intermediate estimates and to the limited structural knowledge of constant-∣A∣max6 hypersurfaces in dimensions above four. The curvature bound depends on the unresolved question of uniform volume control, and the Steklov comparison between ∣A∣max7 and ∣A∣max8 is left open except when dependence on ∣A∣max9 is permitted, where Colbois–Girouard–Hassannezhad already suffices.
Conclusion
The paper converts the discarded Hessian term in the Choi–Wang argument into a genuine spectral gap, yielding the first improvement of ∣A∣max0 that survives as ∣A∣max1 and a dimension-only gap in the constant-∣A∣max2 case without recourse to isoparametric classification. Its secondary contribution — a Steklov lower bound leading to ∣A∣max3 — reduces Chern's problem, under embeddability, to a uniform volume estimate. Whether the eigenvalue gap can be sharpened toward ∣A∣max4, and whether the required volume bound holds, remain open.