Regular Value and the Preimage Theorem

How do scientists and mathematicians make sense of complex landscapes, from the energy states of a physical system to the configuration space of a robot? A powerful approach is to trace the contours—the sets where some key property remains constant. These "level sets" slice a complex space into simpler, more comprehensible pieces. However, this raises a crucial question: when are these slices smooth, well-behaved surfaces, and when do they have sharp corners, cusps, or other problematic singularities? The answer lies in a foundational concept in differential geometry: the distinction between regular and critical values. This article delves into this pivotal idea, revealing how it provides a definitive guarantee of smoothness.

The following chapters will guide you through this elegant theory and its far-reaching consequences. First, in ​​"Principles and Mechanisms,"​​ we will explore the core concepts of regular values, critical points, and level sets. We will uncover the Preimage Theorem, a mathematical contract that guarantees smoothness, and Sard's Theorem, which assures us that troublesome critical values are incredibly rare. We will also touch upon Morse Theory to see how passing through critical values systematically builds the topology of a space. Then, in ​​"Applications and Interdisciplinary Connections,"​​ we will put this theory to work, demonstrating how it is used to construct fundamental geometric objects like spheres and rotation groups, and to describe the very arenas in which the laws of physics unfold, such as the constant-energy surfaces of classical mechanics.

Imagine you are a cartographer, tasked with drawing a map of a mountain range. Your most important tool is the contour line, a line connecting all points of the same elevation. Where the mountain slope is gentle, the contour lines are spread out and gently curving. Where the slope is steep, they are tightly packed. But what happens at the very peak of a mountain, or at the bottom of a valley, or at that curious place called a saddle point, a pass between two mountains? At these special locations, the ground is momentarily flat. A tiny step in any direction might lead you up or down, but at that exact spot, the slope is zero. These are the places where the character of the landscape is defined.

In mathematics, we do something remarkably similar. For a function, say $f(x, y)$ which assigns a "height" to every point $(x, y)$ on a plane, the sets of points where the function takes a constant value, $f(x,y)=c$, are called ​​level sets​​. These are the contour lines of our mathematical landscape. The study of these level sets reveals the deep structure of the function, and by extension, the structure of countless phenomena in physics, engineering, and beyond.

Let’s start with a simple one-dimensional landscape, a function from the real number line to itself, like the one in problem, $p(x) = x^3 - 3x$. The "slope" is just the derivative, $p'(x) = 3x^2 - 3$. The special points where the ground is flat are where this slope is zero. Setting $p'(x)=0$ gives us $x=1$ and $x=-1$. These are the ​​critical points​​ of our function. They correspond to a local valley (minimum) and a local peak (maximum).

The heights at these special points are called the ​​critical values​​. For our function, they are $p(1) = -2$ and $p(-1)=2$. All other height values are called ​​regular values​​. This distinction between regular and critical values is not just a piece of terminology; it is the fundamental organizing principle for understanding the geometry of functions.

Why is this distinction so important? Let's move to a two-dimensional landscape. Think of the function $f(x,y) = x^2 - y^2$, which describes a saddle shape. To find the critical points, we need the "slope" to be zero in every direction. This is captured by the ​​gradient​​, $\nabla f$, which is a vector of the function's partial derivatives: $\nabla f = (\frac{\partial f}{\partial x}, \frac{\partial f}{\partial y})$. For our saddle function, the gradient is $\nabla f = (2x, -2y)$. The only point where this vector is $(0,0)$ is the origin, $(0,0)$. So, we have just one critical point. The height at this point, the critical value, is $f(0,0)=0$.

Now, let's look at the contour lines. For any regular value $c \neq 0$, the level set $x^2 - y^2 = c$ is a hyperbola—a perfectly smooth, well-behaved curve. In the language of geometry, it’s a smooth 1-dimensional manifold. But what happens at the critical value $c=0$? The level set becomes $x^2 - y^2 = 0$, which factors into $(x-y)(x+y)=0$. This isn't one curve, but two straight lines, $y=x$ and $y=-x$, that cross at the origin. At that crossing, the level set is not a smooth curve. It has a sharp corner. We see a dramatic change in the topology—the very nature of the shape—of the contour line precisely at the critical value.

This is not a coincidence. This observation is captured by one of the most powerful and elegant tools in differential geometry: the ​​Preimage Theorem​​ (also known as the Regular Value Theorem). In simple terms, the theorem says:

For any smooth function, the level set corresponding to a ​​regular value​​ is always a nice, smooth manifold.

This theorem is a profound guarantee. It's a contract. If a value $c$ is regular—meaning the gradient of the function is non-zero at every point on the level set $f(p)=c$—then the level set $f^{-1}(c)$ is guaranteed to be a smooth surface (or curve, or hypersurface in higher dimensions), with no corners, cusps, or self-intersections. This theorem gives us a powerful method to construct smooth manifolds. If we can write a set of points as the level set of a regular value for some smooth function, we've proven it's a manifold.

Consider the unit sphere $S^2$, defined by the equation $x^2+y^2+z^2 = 1$. We can see this as the level set $g^{-1}(1)$ for the function $g(x,y,z) = x^2+y^2+z^2$. The gradient is $\nabla g = (2x, 2y, 2z)$, which is only zero at $(0,0,0)$. Since the origin is not on the sphere (i.e., not a point in the level set), the gradient is non-zero everywhere on the sphere. Thus, $c=1$ is a regular value, and the theorem confirms that the sphere is a smooth 2-dimensional manifold.

Let's look at another beautiful example: the simple height function on this sphere, $f(x,y,z)=z$. Where are the critical points, the places where the "slope" is zero? Intuitively, they are the very top (the North Pole, $(0,0,1)$) and the very bottom (the South Pole, $(0,0,-1)$). At these points, the tangent plane is horizontal. The corresponding critical values are the heights $z=1$ and $z=-1$. For any other height $c$ between $-1$ and $1$, the level set $f^{-1}(c)$ is just a circle of latitude. The Preimage Theorem guarantees that these circles are smooth 1-manifolds, which is obviously true. At the critical values, the level sets are single points—the poles—which are 0-dimensional, not 1-dimensional. The theorem holds perfectly; the character of the level set changes at the critical values.

A crucial word in the theorem's guarantee is "smooth". The function itself must be smooth (infinitely differentiable). Consider the function $f(x,y) = |x|$. This function is not smooth; it has a sharp crease along the y-axis where the derivative is not defined. The Preimage Theorem does not apply. And indeed, for the value $c=0$, the level set is the y-axis, a smooth line. However, the theorem can't be used to conclude this, because its fundamental hypothesis is violated. The mathematical contract is void if you don't meet its conditions.

When the Preimage Theorem's guarantee doesn't apply—that is, at a critical value—the level set can become a rather interesting "zoo" of shapes. We've already seen two lines crossing. Another famous example comes from the function $F(x,y) = y^2 - x^3$. The only critical point is at $(0,0)$, making $c=0$ the only critical value. The level set $y^2 - x^3 = 0$ forms a shape called a ​​cusp​​, which has a sharp, pointed tip at the origin. It's a single, connected curve, but it's not smooth at that critical point. For another function, $f(x,y) = x^2 y^2$, the critical points are all the points on both the x- and y-axes, yet they all map to a single critical value, $c=0$. The level set for this value is the union of the two axes, again forming a cross.

Given that critical values lead to these interesting, and sometimes problematic, behaviors, you might worry that they are common. But here, nature hands us another astonishingly beautiful and deep result: ​​Sard's Theorem​​.

Sard's Theorem tells us that for any smooth function you can write down, the set of its critical values is "small" or "rare". In mathematical terms, it has ​​measure zero​​. What does this mean, intuitively? Imagine the number line representing all possible output values of your function. If you were to throw a dart at this number line completely at random, the probability that you would hit a critical value is zero. Absolutely zero.

This is a profoundly comforting thought for scientists and engineers. It means that most situations are the "regular" ones. The level sets of your system are almost always going to be well-behaved, smooth manifolds. The singularities—the cusps, the crossings, the collapses in dimension—are the exceptions. They are infinitely rare, even if they are interesting!

So, if the topology of our landscape only changes when we cross the "rare" critical values, what kind of change happens? This is the subject of a beautiful extension of these ideas called ​​Morse Theory​​. Let's go back to our analogy of raising the water level on an island landscape.

As the water level $t$ rises, the shoreline, which is the level set $f^{-1}(t)$, changes.

• When the water level reaches the bottom of a valley (a critical point of ​​index 0​​), a new island (shoreline) appears out of nowhere. We have changed the topology by adding a disconnected piece. This is like attaching a "0-handle" (a disk).
• As the water rises further, it might reach a saddle point between two peaks (a critical point of ​​index 1​​). At that moment, two previously separate shorelines merge into one. The topology has changed by connecting two pieces. This is like attaching a "1-handle" (a strip or a bridge).
• Finally, as the water submerges a peak (a critical point of ​​index 2​​), an island's shoreline shrinks to a point and disappears. The topology has changed by filling a hole. This is like attaching a "2-handle" (a cap).

This amazing idea shows that the entire, possibly complex, topology of a manifold can be understood by breaking it down into these simple attachments, all governed by the critical points of a function defined on it. And even more wonderfully, the type of "handle" you attach ($k$-handle) is determined by the "index" of the critical point—the number of independent directions that go "downhill" from that point. For a critical point of index $k$, it turns out the change in a topological quantity called the Euler characteristic is simply $m(-1)^k$, where $m$ is the number of such critical points you cross.

This journey, from the simple idea of a contour line to the construction of entire universes by attaching handles, reveals the deep and beautiful unity of mathematics. The local behavior of a function—its simple slope—dictates the global shape of the spaces it defines. The distinction between regular and critical values is not just a definition to be memorized; it is the key that unlocks this profound connection between analysis and geometry, a principle that echoes throughout science.

In the last chapter, we acquainted ourselves with a remarkable piece of machinery: the Preimage Theorem. It’s a sort of universal blueprint for geometers. You hand it a smooth function $F: M \to N$, it inspects a value $c$ in the target space $N$, and if $c$ passes the crucial "regular value" test, the machine outputs a perfect, newly-minted submanifold, the level set $F^{-1}(c)$. We have tinkered with the gears and levers of this machine, understanding its internal logic. Now, the real fun begins. We are going to take it for a spin. Where can it take us? What can it build? The answer, you will see, is that it can build entire worlds, from the familiar shapes that surround us to the abstract arenas where the laws of physics play out.

Let’s start with an old friend: the sphere. You can describe a sphere of radius $R$ as the set of points $(x, y, z)$ satisfying $x^2 + y^2 + z^2 = R^2$. But with our new tool, we can see this differently. Consider the smooth function $f: \mathbb{R}^3 \to \mathbb{R}$ given by $f(x,y,z) = x^2+y^2+z^2$. This function defines a kind of "height" at every point in space, measuring the squared distance from the origin. A sphere is simply a level set of this function! For any positive value, say $c > 0$, the set $f^{-1}(c)$ is a sphere of radius $\sqrt{c}$. Our theorem confirms what we intuitively know: the sphere is a lovely, smooth 2-dimensional manifold, because every value $c>0$ is a regular value of $f$.

But what happens at $c=0$? The level set $f^{-1}(0)$ is just a single point, the origin. This is a 0-dimensional manifold, not the 2-dimensional one our rule $\dim(\mathbb{R}^3) - \dim(\mathbb{R}) = 2$ would suggest. What went wrong? Nothing! The theorem simply doesn't apply because $c=0$ is a critical value. The origin is the one and only critical point of $f$, where its differential (represented by the gradient) vanishes. This is a beautiful illustration of the theorem's power and its limits: it works wonders, but only when its conditions are respected.

The true beauty of this approach is its effortless generalization. We can’t visualize a 17-dimensional sphere living in $\mathbb{R}^{18}$, but we don't have to. We can define it abstractly as the level set of $F(x) = \langle x, x \rangle = 1$ for $x \in \mathbb{R}^{18}$. The Preimage Theorem immediately assures us that this object, $S^{17}$, is a well-defined 17-dimensional manifold. What's more, it gives us a concrete handle on its tangent spaces. The condition for a vector $v$ to be tangent to the sphere $S^n$ at a point $p$ turns out to be wonderfully simple: the vector $v$ must be orthogonal to the position vector $p$, i.e., $\langle p, v \rangle = 0$. So, while we cannot see the surface, we can understand with perfect clarity the space of all possible velocities one could have while moving upon it.

This "level set" construction method becomes truly profound when we apply it to spaces whose "points" are more abstract entities than mere points in space.

Consider the set of all possible rotations in three-dimensional space. Every rotation can be represented by a $3 \times 3$ matrix $Q$ with two properties: it preserves lengths and angles (making it an orthogonal matrix, $Q^{\top} Q=I$) and it preserves orientation (its determinant is 1). This collection is called the Special Orthogonal Group, $SO(3)$. Is this set of matrices just a jumble, or does it have a geometric structure of its own?

Let's build a function $\Phi(A) = A^{\top} A$ from the space of all $n \times n$ matrices ($\mathbb{R}^{n \times n}$) to the space of symmetric matrices. The orthogonal matrices are precisely the preimage of the identity matrix $I$. By showing that $I$ is a regular value of this map, the Preimage Theorem strikes again! It tells us that the set of all $n$-dimensional rotations, $SO(n)$, is itself a smooth manifold. This is a breathtaking result. It means we can do calculus on the space of rotations itself. We can talk about a "smoothly varying rotation," a concept essential for describing the motion of rigid bodies and for the gauge theories that form the foundation of modern particle physics. The theorem, in one clean stroke, builds the very stage on which the physics of symmetry is performed. The same logic allows us to construct other fundamental objects, like the Stiefel manifolds, which represent the space of all possible orthonormal frames and are indispensable in fields from physics to statistics.

The theorem's reach extends deep into the heart of classical mechanics. The complete state of a physical system—the positions and momenta of all its particles—can be represented as a single point in a high-dimensional space called phase space. For many systems, the total energy is conserved. This energy is given by a smooth function on the phase space, the Hamiltonian $H$. The law of energy conservation isn't just an algebraic rule; it's a geometric constraint. It dictates that the system's state must always lie on the surface of constant energy, $H=E$. And what is this surface? For any energy $E$ that is not a critical value of the Hamiltonian, the Preimage Theorem guarantees that this constant-energy surface is a smooth, embedded submanifold of the phase space. The entire evolution of the universe, under the laws of classical mechanics, is nothing more than a trajectory traced out on one of these magnificent energy manifolds.

A good craftsman must know their tools, and that includes knowing their limitations. The "regular value" condition is not a fussy bit of mathematical fine print; it is the absolute heart of the matter. What happens if we try to build a manifold from a critical value?

Let's investigate the set of all $2 \times 2$ matrices $A$ that are idempotent, meaning that applying the transformation twice is the same as applying it once: $A^2 = A$. These represent projection operations. We can write this as a level set problem: $F(A) = A^2-A=0$. Is the set of idempotent matrices a smooth manifold? The answer is no. If we analyze the differential of the map $F$, we find that for certain idempotent matrices (those that are not the zero or identity matrix), the differential fails to be surjective. This means $0$ is not a regular value. The consequence? The set of idempotent matrices is not a single, clean manifold. It has "singularities"—sharp corners or intersections where different components of the set meet. Our manifold-making machine has jammed precisely because we fed it a critical value. This is a powerful cautionary tale: regularity is the guardian of smoothness.

We can see this distinction more visually in other examples. Consider a map from a torus (a donut's surface) to a sphere. We can construct such a map where the "bad" points on the torus—the critical points where the map's Jacobian determinant vanishes—all get sent to just two points on the sphere: the North and South poles. These two poles are the critical values. Every other point on the sphere, from the equator to anywhere in between, is a "good" regular value, whose preimages are nice, well-behaved collections of curves on the torus.

So what if a problem—perhaps a constrained optimization problem in engineering or economics—hands us a function whose level set is defined by a critical value? Are we doomed to work with a singular, ill-behaved object?

Here, we witness one of the most elegant plays in modern mathematics. A remarkable result called Sard's Theorem comes to our aid. In essence, it tells us something profound about the universe: ​​regularity is the norm, and singularity is the exception.​​ The collection of all "bad" critical values, while it may exist, is vanishingly small. In the space of all possible values, the set of critical values has "measure zero." They are like a sprinkle of dust in a vast room.

This is not merely an abstract comfort; it's a practical license to nudge our problem. Suppose you need to find the minimum of a function $f$ on a constraint set $h(x)=c$, but $c$ happens to be a critical value, making your set singular. Don't despair! Sard's Theorem guarantees that there are regular values $c'$ arbitrarily close to $c$. You can simply solve a slightly perturbed, well-behaved problem on the smooth manifold $h(x)=c'$. The solution you find will be an excellent approximation to your original problem.

Furthermore, once you are on one of these "nice" nearby manifolds, the full power of calculus, including tools like the Implicit Function Theorem, becomes available. One can then analyze how the solutions (the constrained critical points) change smoothly as you vary the constraint value $c'$, providing a deep understanding of the problem's stability and structure. This technique of perturbing a problem from a singular to a regular setting is a cornerstone of variational calculus, global analysis, and numerical methods. It is the ultimate "get out of jail free" card, and its existence is guaranteed by our simple, beautiful theory of regular values.

From the humblest sphere to the grandest arenas of physics and the subtle art of mathematical perturbation, the Preimage Theorem provides a unified language for describing and creating the geometric structures that underpin our world. It is a testament to the power of a single, well-posed idea to illuminate a vast and interconnected landscape of scientific thought.