Every closed surface metric in four-dimensional Euclidean space

← All digests

Family 333 · Differential geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-26

1. The question

Imagine a surface equipped with a complete set of instructions for measuring lengths and angles. Can that geometry be realized as a surface sitting in ordinary Euclidean space, without stretching any tangent direction?

A round sphere has an obvious realization in three dimensions. A general metric on a general surface need not resemble that familiar example. Its curvature may change sign, its topology may be complicated, and the lengths prescribed in neighboring coordinate patches must fit together globally.

Family 333 claims that four ambient dimensions always suffice for a smooth isometric immersion of a closed surface. Here “closed” means compact and without boundary. “Immersion” allows self-intersections: different points of the abstract surface may occupy the same location, although each small patch remains a regular surface. The manuscript includes nonorientable and disconnected surfaces, with no restriction on Gaussian curvature.

The question is about preserving the geometry measured along the surface. It does not ask for the ambient straight-line distance between every pair of image points to equal their intrinsic surface distance. Even a round sphere would fail that stronger requirement: its chords are shorter than its surface arcs.

2. Background and essential definitions

A smooth surface $M$ is a space locally described by two real coordinates with smooth changes of coordinates. A Riemannian metric assigns an inner product $g_p$ to the tangent plane at each point $p$, varying smoothly with $p$.

In local coordinates $(u,v)$, write

$$g=g_{11}\,du^2+2g_{12}\,du\,dv+g_{22}\,dv^2.$$

The symmetric matrix $(g_{ij})$ is positive definite. It determines lengths of tangent vectors, angles, areas, and lengths of curves. Gaussian curvature is an intrinsic quantity derived from this metric; it describes how the local geometry differs from the flat plane.

A smooth map $F:M\to\mathbb R^4$ is an isometric immersion when its derivative preserves every tangent inner product:

$$\langle dF_p(a),dF_p(b)\rangle=g_p(a,b).$$

In coordinates, this becomes the three equations

$$\langle F_u,F_u\rangle=g_{11},\qquad \langle F_u,F_v\rangle=g_{12},\qquad \langle F_v,F_v\rangle=g_{22}.$$

Positive definiteness ensures that the two derivative vectors are linearly independent. An embedding adds a global requirement: the surface must be realized without identifying distinct points. The theorem here only asserts immersion.

An elementary four-dimensional example is a product of circles:

$$F(u,v)=(a\cos u,a\sin u,b\cos v,b\sin v).$$

For positive $a,b$ and periodic coordinates, this realizes a flat torus with metric $a^2du^2+b^2dv^2$. It illustrates how an extra ambient direction can accommodate geometry that is not represented by the familiar curved doughnut in three-space. It does not by itself address arbitrary surface metrics.

Finally, $C^\infty$, or smooth, means that derivatives of every order exist and are continuous. This is much stronger than $C^1$, which controls only the map and its first derivatives. The difference is central because the metric uses first derivatives, while classical curvature and bending involve second derivatives.

3. Development history and previous partial results

The local analytic theory is an early landmark. Janet and Cartan showed that an analytic two-dimensional Riemannian metric has a local isometric realization in three-dimensional Euclidean space. Analytic means locally representable by convergent power series. Both the regularity hypothesis and the local conclusion matter: neither can simply be replaced by an arbitrary smooth closed-surface statement.

Nash’s smooth embedding theorem established a far-reaching global existence principle in higher dimension. His original compact-manifold bound is $n(3n+11)/2$, which gives seventeen ambient dimensions for surfaces. This provides an embedding, a stronger global conclusion, but in substantially more room than the present four-dimensional immersion problem.

Gromov obtained smooth isometric immersions of closed surfaces into five-space. He also recorded the four-space question and later attributed it to Chern around 1950, while noting uncertainty about the original compactness qualification. M1 specifically addresses closed surfaces, so its scope should not be enlarged to every possible noncompact formulation of that historical question.

There was already a smooth four-dimensional theory for suitable planar domains. Poznyak’s work gives isometric immersions on compact portions of simply connected domains, with corresponding finite-regularity statements. Lewicka’s modern account explains the smooth version and develops finer flexibility results for metrics of low Hölder regularity. These theorems do not automatically assemble into a realization of an arbitrary closed surface: corrections in one patch must preserve the geometric conditions needed in later patches.

A different line of progress comes from Nash and Kuiper. Their oscillatory constructions produce $C^1$ isometric immersions approximating strictly short immersions. A map is strictly short when its induced metric is strictly smaller than the desired metric in every tangent direction. The flexibility available at $C^1$ regularity is striking, but it does not settle a $C^\infty$ problem.

Family 334 supplies a complementary claimed obstruction in three-space: a smooth metric with no local smooth isometric immersion near one point. Together, the two families concern a sharp contrast between the smooth local three-dimensional problem and the smooth global closed-surface four-dimensional problem.

4. What the new manuscript claims

M1 states that every closed smooth Riemannian surface $(M,g)$ admits a smooth map $F:M\to\mathbb R^4$ satisfying

$$F^*\delta_4=g, \qquad \delta_4=dx_1^2+dx_2^2+dx_3^2+dx_4^2.$$

The notation $F^*\delta_4$ means the Euclidean metric pulled back by $F$: evaluate Euclidean inner products after applying its derivative. Thus the equality is exact, not an approximation or a limiting assertion about separate maps without a smooth limit.

There are no orientability, connectedness, or Gaussian-curvature restrictions. In particular, the statement covers smooth metrics on nonorientable closed surfaces. The resulting map can have self-intersections, and M1 does not claim that every such surface metric admits an isometric embedding into four-space.

The construction uses a finite list of positive local metric additions. Each addition includes an infinite smoothing iteration that converges in every smooth norm. Calling the global assembly finite therefore does not mean that an exact immersion is obtained after finitely many elementary numerical operations.

A central analytical ingredient is an exact correction theorem under a condition on the second fundamental form, which records how tangent planes bend into the normal directions. The manuscript requires this vector-valued tensor to be nonzero at every point in the relevant correction step. It does not assume the stronger rank conditions that would supply a simple global inverse for the linearized metric equation.

That intermediate condition is arranged and preserved by the construction. It is not an additional hypothesis on the surface metric in the main theorem.

5. Why the advance matters

Importance: 5/5. The claim reaches a central low-dimensional target in the theory of smooth isometric immersions, lowering the known closed-surface immersion dimension from five to four without curvature or orientability assumptions.

The ambient dimension is not a superficial parameter. A two-dimensional surface in four-space has only two normal directions in which to bend. Metric correction must use that limited freedom while controlling the compatibility equations and all derivative orders. A general high-dimensional existence theorem does not explain how to accomplish this in codimension two.

The result would also clarify the division between local flexibility and global geometry. Having an immersion on each coordinate disk does not ensure that those disks can be joined isometrically. The manuscript’s claimed advance is a construction that keeps later corrections possible while modifying earlier regions.

The importance score concerns the theorem’s stated significance if correct. It does not certify the analytical argument or compare its reliability with independently checked results.

6. The main idea

Start with a smaller metric

The construction begins with an ordinary smooth immersion into a small round three-sphere inside $\mathbb R^4$. After scaling, its induced metric $g_0$ is strictly below $g$. The inward radial direction provides a globally defined normal vector field, even when the surface itself is nonorientable.

The positive deficit is decomposed into finitely many local rank-one pieces:

$$g-g_0=\sum_{j=1}^{J}a_j^2\,dx_j^2.$$

Each coordinate function $x_j$ belongs to a disk, and the smooth amplitude $a_j$ vanishes outside that disk. A rank-one piece adds length in one selected direction. The decomposition is prepared with conditions on the intermediate metrics, because the order of the additions affects what the next step can do.

Add one direction of length by oscillation

For one piece $a^2dx^2$, choose a complementary coordinate $y$. The manuscript places a periodic velocity loop in a plane perpendicular to both $F_y$ and $F_{yy}$. Its mean is prescribed, while its circular motion supplies the desired extra contribution to the metric.

The idea resembles adding a fine corrugation to increase length along one direction, but the argument must preserve far more than a picture of corrugations suggests. Orthogonality to the second derivative $F_{yy}$ removes a leading transverse error. Further finite corrections cancel oscillatory errors to a chosen high order; separate slower-scale corrections handle the remaining mean terms.

Make the metric exact while retaining smoothness

High-frequency oscillation can make a map small in position while making its higher derivatives large. This is why an approximation in the visible shape is not enough. The source develops small oscillatory corrections with controlled second derivatives and solves the linearized metric equations mode by mode to high finite accuracy.

Smoothing between successive corrections controls the loss of derivatives. Increasing the required accuracy in successive blocks yields convergence in every $C^m$ norm, for every fixed derivative order $m$. The endpoint is one smooth map with the exact metric, not merely a $C^1$ realization.

Preserve the geometry for later disks

A large oscillation can rotate or degenerate normal directions that a later patch needs. M1 therefore keeps track of specified boundary curves, crossings, preferred normals, and transverse derivatives. It chooses the velocity loop at its turns so that the necessary directions remain usable. The subsequent small exact correction preserves those conditions.

This last stage is the global compatibility mechanism. It allows the finite sequence of primitive additions to continue until their sum fills the entire metric deficit. The paper separates the large operation that adds length from the small operation that removes error, assigning each a different role in controlling second-order geometry.

7. Scope, limitations, and open questions

The theorem concerns compact boundaryless surfaces with smooth positive-definite metrics and smooth immersions. It does not assert a general embedding theorem in four-space, a boundary-value theorem with prescribed boundary data, or a corresponding result for every complete noncompact surface.

Nor does it claim that every metric can be realized smoothly in three-space. Analytic local results and low-regularity flexibility theorems have different hypotheses and conclusions; the companion smooth local obstruction is part of that distinction.

The construction is existential and analytical. Its finite global decomposition still relies on limiting corrections. The statement does not provide a practical algorithm with effective complexity bounds or a way to visualize arbitrary exact realizations without self-intersections.

Questions about global injectivity, additional geometric constraints on the immersion, effective estimates, and other domain classes remain separate from the claim discussed here. This digest reports the manuscript’s theorem and mechanism without independently verifying the proof.

8. Sources and further reading

• Nash, The imbedding problem for Riemannian manifolds, for the smooth high-dimensional existence theorem.
• Gromov, Geometric, algebraic and analytic descendants of Nash isometric embedding theorems, for historical context and the five-dimensional surface result.
• Lewicka, Full flexibility of isometric immersions in Poznyak’s dimension, version 3, for planar-domain results and the regularity distinction.
• Family 334 discusses the claimed smooth local obstruction in three dimensions.

Source manuscripts

• M1. OpenAI, Smooth isometric immersions of closed surfaces into Euclidean four-space (September 23, 2026).

Sources consulted on 26 October 2026 at repository revision 1581d4c. Download the manuscript citations.