Classical Mechanics Series Part 1

Part 1: Escaping Newton's Cage: Configuration Space and Manifolds

As one ventures deeper into theoretical physics, the standard inclined-plane problems of first-year lecture halls give way to the fundamental symmetries of nature and the geometric structures woven by those symmetries. At this point, the true elegance of classical mechanics lies in escaping Newton's restrictive cage of arrows—Cartesian vectors—and discovering the geometric language of the universe.

1 Breaking Free of Coordinates: The Universe as an Affine Space

The most fundamental principle of physics is this: there is no privileged point in the universe. In other words, the universe has no "origin" at the point $(0,0,0)$ . If we define space directly as the vector space $\mathbb{R}^3$ , we unjustly confer a sacred status on the zero vector—the origin—inherent to that space. Arnold resolves this philosophical problem by defining space as an affine space.

You can think of an affine space as "a vector space whose starting point has been forgotten." Points and vectors are distinct concepts. Our space $\mathbb{A}^3$ contains only "points," such as $A, B, C$ . Subtracting one point from another gives a vector, but points cannot be added together—adding two positions has no physical meaning!

Mathematically, our space is an affine space $\mathbb{R}^3$ acted upon by a vector space $\mathbb{A}^3$ . Thus, adding a vector $A$ to a point $\mathbf{v}$ gives a new point $B$ :

B = A + \mathbf{v} \implies \mathbf{v} = B - A \tag{1}

This simple definition encodes the homogeneity of space—the fact that the same rules apply everywhere—directly into geometry. There is no origin, only relations between points. Space is a blank page without direction or hierarchy.

The Bridge Between Two Spaces: Choosing and Forgetting the Origin

How, then, do we pass from this directionless, nonhierarchical affine space $\mathbb{A}^3$ to the familiar, calculable vector space $\mathbb{R}^3$ , or vice versa? The answer is very simple, but it carries a profound meaning in physics: by choosing an origin, or reference point.

We defined affine space as "a vector space that has forgotten its origin." If we choose an entirely arbitrary point $\mathbb{A}^3$ on the blank page $O$ and call it the "origin," every point $P$ in the affine space corresponds one-to-one with a position vector drawn from $O$ to $P$ , $(\mathbf{r} = P - O)$ . At that instant, points become vectors and our affine space collapses into the vector space $\mathbb{R}^3$ . We now have a foundation on which to place the standard Cartesian axes.

Conversely, if we want to pass from the vector space $\mathbb{R}^3$ to the affine space $\mathbb{A}^3$ , we "forget" the origin. We erase the absolute positions of the vectors and describe space only through the distances and directions of objects relative to one another—that is, through difference vectors.

The philosophy behind Arnold's vision of mechanics is this: the true fabric of the universe is $\mathbb{A}^3$ ; the universe itself has no origin. We mortal physicists, however, place an observer—a reference frame—in space so that we can write equations and make measurements, thereby temporarily reducing space to $\mathbb{R}^3$ . The true power of classical mechanics comes from the fact that the physical laws we discover never depend on the arbitrary point $O$ that we later drive into space. How, then, does this independence, or invariance, manifest itself in spacetime?

2 Absolute Time and Galilean Spacetime

Space alone is not enough; we must also know when events occur. In Newtonian, or Galilean, mechanics, time is a one-dimensional affine space, $\mathbb{A}^1$ , entirely independent of space. Time, too, has no beginning or origin; saying "10,000 BC" or "three seconds after the Big Bang" does not turn time into a vector space, but merely selects reference points.

What, then, is the thing we call spacetime? Arnold defines spacetime as a four-dimensional affine space $\mathbb{A}^4$ whose points are "events." This is not just any four-dimensional space, however. Within it is a very special projection called the "time" function:

t : \mathbb{A}^4 \to \mathbb{A}^1 \tag{2}

Let $a$ and $b$ be two events in spacetime. The time difference between them is given by $t(b) - t(a)$ .

We now arrive at the most critical concept in Arnold's construction of classical mechanics: simultaneity. Two events occur at the same time if $t(a) = t(b)$ . In the Galilean universe, distance is defined only and exclusively between simultaneous events. The question "What is the distance between my present location and an explosion that will occur on Mars ten seconds from now?" is meaningless in classical mechanics, because spatial distance changes for a moving observer. If two events occur at the same time, however, all observers in the universe agree on the spatial distance—the Euclidean metric—between them.

Arnold-style spacetime can be imagined as a stack of pancakes made up of simultaneity slices, each one an $\mathbb{A}^3$ . When you later move on to the theory of relativity or quantum field theory, you will appreciate far better how Einstein bent and melted this "stack of pancakes": dismantling absolute time and combining distance and time in a single Minkowski metric.

3 The Galilean Group: The Symmetry That Determines the Rules of Physics

The stage is set. What "camera movements" are allowed on this stage? The transformations between reference frames under which the laws of physics remain unchanged form the Galilean group. This group includes:

Spatial Translations: $\mathbf{r} \to \mathbf{r} + \mathbf{a}$ (homogeneity of space)
Time Translations: $t \to t + \tau$ (homogeneity of time)
Spatial Rotations: $\mathbf{r} \to R\mathbf{r}$ (isotropy of space, or independence of direction)
Inertial Transformations (Galilean Boosts): $\mathbf{r} \to \mathbf{r} + \mathbf{v}t$ (equivalence of frames moving at constant velocity)

4 Inertial Reference Frames and the Inadequacy of Vectors

To study mechanical events, we must first choose a reference frame, since space will not be homogeneous and isotropic in an arbitrary frame. According to Galileo's principle of relativity, however, it is always possible to find "inertial frames" in which space is homogeneous—all of its points are equivalent—and isotropic—all of its directions are equivalent—while time is homogeneous. In an inertial frame, a free body moves with a velocity whose magnitude and direction remain constant.

What happens, however, when the system before us consists not of free particles but of bodies subject to constraints? At this point, Newtonian mechanics leads us into a difficult process. If we try to describe the motion of a pendulum confined to the surface of a sphere using the vectors $\mathbb{R}^3$ in $(x, y, z)$ , we must account for constraint forces that prevent the body from leaving the surface and whose directions and magnitudes vary continuously. Yet these reaction forces arise from the geometric boundaries of the system rather than from a physical interaction. To analyze systems with ideal holonomic constraints, we must free the equations describing the system from the burden of these invisible and unhelpful forces.

5 Why Do We Abandon $\mathbb{R}^3$ and Take Refuge in Tori and Spheres?

Newton gave us a starting point, but he confined us to the flat, dull, Cartesian space $\mathbb{R}^3$ . Under universal constraints, however—consider a pendulum moving on the surface of a sphere or a train moving along a track—Cartesian coordinates become helpless. The equations overflow with constraint forces. They are called constraint forces because, conceptually, they bind the system to its configuration manifold and geometrically restrict it from leaving that manifold, whether a surface or a curve.

In Mathematical Methods of Classical Mechanics, Arnold describes the innovation introduced in response to these constraints as follows: redefine the space in which the physical system lives using the system's own degrees of freedom.

If our system contains $N$ particles and $k$ holonomic—that is, integrable—constraints, the dimension of the space in which the system can move freely is $n = 3N - k$ . We call the abstract space defined by these $n$ independent coordinates $(q_1, q_2, \ldots, q_n)$ the configuration manifold ( $M$ ). Consider a double pendulum: although two masses move through space, the entire system can be described completely by only two angles, $(\theta_1, \theta_2)$ . The configuration space of this double pendulum is therefore a torus, $T^2 = S^1 \times S^1$ . In Arnold's geometric language, every instantaneous configuration of the system is a point on this manifold.

A point $q \in M$ on the manifold represents, by itself, the complete configuration of the system at that instant—the positions of all its particles. As time $(t)$ passes, this point traces a smooth curve on the manifold:

\gamma : \mathbb{R} \to M, \quad t \mapsto q(t) \tag{3}

What about velocity? Velocity is the vector tangent to this curve on the manifold. This vector does not live in the manifold itself, however; it lives in the tangent space ( $T_qM$ ) of the manifold at that point. The tangent space at a point is the set of all possible velocity vectors that can pass through that point. If we want to consider positions and velocities together, we join the tangent spaces at every point of the manifold to construct an entirely new and enormous $2n$ -dimensional space: the tangent bundle ( $TM$ ).

TM = \bigcup_{q \in M} T_q M \tag{4}

The legendary function $L$ that we call the Lagrangian is neither a formula handed down from the sky nor an arbitrary scalar. It is a smooth, differentiable function from this tangent bundle and time to the real numbers:

L : TM \times \mathbb{R} \to \mathbb{R}

(q, \dot{q}, t) \mapsto L(q, \dot{q}, t) \tag{5}

You give it the system's position $L$ , velocity $(q)$ , and time $(\dot{q})$ , and it gives you a scalar number. All the physics is contained in this scalar function.

6 Example: The Helical Roller-Coaster Pendulum

To internalize this abstract geometric structure completely, let us devise a system in which Newtonian forces would create an intractable chaos, but which analytical mechanics transforms into a work of art: the helical roller-coaster pendulum.

Our system is as follows. We have a fixed wire in the shape of a helix in space. A bead of mass $(t)$ can slide frictionlessly along this wire. A second particle of mass $m_1$ is attached to $m_1$ by a massless rod of length $l$ , forming a pendulum. Let the pendulum be allowed to swing only in the radial–vertical plane, directed outward from the center of mass.

Defining the Coordinates (the Manifold):

Our system has two degrees of freedom. Its manifold is formed by the one-dimensional structure of the helix, with a topology similar to $m_2$ , and the angular motion of the pendulum, also $S^1$ . Let us define the coordinates:

$S^1$ : The azimuthal angle determining the position of mass $q_1$ on the helix.
$m_1$ : The angle that the pendulum makes with the vertical.

Let the helix have radius $q_2$ and pitch parameter $R$ . The position of mass $c = h/2\pi$ in space is

\mathbf{r}_1 = (R\cos q_1,\ R\sin q_1,\ cq_1) \tag{6}

Relative to the location of $m_1$ , the pendulum mass $m_2$ is displaced radially outward by $m_1$ and downward along the z-axis by $l\sin q_2$ :

\mathbf{r}_2 = \big((R + l\sin q_2)\cos q_1,\ (R + l\sin q_2)\sin q_1,\ cq_1 - l\cos q_2\big) \tag{7}

Velocities and Kinetic Energy:

Kinetic energy is always a quadratic function of the generalized velocities and is expressed as $l\cos q_2$ . We find the squared velocities by differentiating the coordinates with respect to time.

\dot{\mathbf{r}}_1^2 = (R^2 + c^2)\dot{q}_1^2 \tag{8}

For the term $T = \frac{1}{2}\sum a_{ik}(q)\dot{q}_i\dot{q}_k$ , differentiating and summing the squares produces cross terms arising from the curvature of the manifold:

\dot{\mathbf{r}}_2^2 = \dot{q}_1^2\left[(R + l\sin q_2)^2 + c^2\right] + l^2\dot{q}_2^2 + 2cl\sin q_2\,\dot{q}_1\dot{q}_2 \tag{9}

After rearranging the total kinetic energy $\dot{\mathbf{r}}_2^2$ , we obtain

T = \frac{1}{2}\underbrace{\left[m_1(R^2+c^2) + m_2\left((R+l\sin q_2)^2+c^2\right)\right]}_{a_{11}}\dot{q}_1^2 + \frac{1}{2}\underbrace{\left[m_2 l^2\right]}_{a_{22}}\dot{q}_2^2 + \underbrace{m_2 cl\sin q_2}_{a_{12}=a_{21}}\dot{q}_1\dot{q}_2\tag{10}

Metric Tensor (Riemannian Metric):

According to Arnold, the positive-definite quadratic form defined on every tangent space of the configuration space of an inertially moving system creates a Riemannian metric $T = \frac{1}{2}m_1\dot{\mathbf{r}}_1^2 + \frac{1}{2}m_2\dot{\mathbf{r}}_2^2$ . This metric is directly the coefficient matrix of the kinetic energy:

g_{ij} = \begin{pmatrix} m_1(R^2+c^2) + m_2\left((R+l\sin q_2)^2+c^2\right) & m_2 cl\sin q_2 \\ m_2 cl\sin q_2 & m_2 l^2 \end{pmatrix} \tag{11}

This is the triumph of our escape from $(ds^2)$ . Instead of struggling with complicated reaction forces, we have obtained the "metric tensor" of the two-dimensional manifold on which the system lives. As the pendulum swings—as $\mathbb{R}^3$ changes—the $q_2$ and $g_{11}$ components of the metric tensor change; in other words, the curvature of the manifold is shaped by the configuration of the system itself. The system's free motions are now nothing other than geodesics in this curved space.

There is, however, a very elegant detail that we must not overlook. The definition of "free motion" applies to situations in which no physical potential energy acts on the system, $g_{12}$ . In the system we constructed, gravitational potential energy inevitably enters because of the helical slope of the manifold along the $V(q) = 0$ -axis and the pendulum's oscillation in the vertical plane. Taking the system's absolute $z_1 = cq_1$ coordinates, $z_2 = cq_1 - l\cos q_2$ and $g_{ij}$ , into account, its actual potential energy is

V(q) = m_1 g(cq_1) + m_2 g(cq_1 - l\cos q_2) = (m_1 + m_2)gcq_1 - m_2 gl\cos q_2 \tag{12}

In the presence of potential energy, the particle's trajectory no longer obeys only the geodesics of the pure metric $E$ arising from the mass distribution; gravity begins to bend the geometry of the configuration manifold around itself. If we want to recast motion under this force in a geometric framework as a smooth "geodesic problem," we must redefine the metric tensor obtained from the kinetic energy using the Maupertuis–Jacobi metric, which incorporates the system's total energy $(\dot{q})$ and potential:

ds^2 = 2(E - V(q))\sum a_{ik}(q)\,dq_i\,dq_k \tag{13}

Classical mechanics thereby geometrizes even potential energy, turning it into the "curvature of space" instead of representing it by force vectors. To reach the true heart of dynamics, all that remains is to transform these velocities $(p)$ in the tangent bundle into momenta $T^*M$ and step into phase space ( $A$ ).

7 Nature's Laziness: The Principle of Least Action

We have constructed our manifold, defined the Riemannian metric generated by kinetic energy, and seen how the potential bends space. But when a particle travels from point $B$ to point $TM$ , how does it decide which path to take through this complex geometry? Do particles "see" the obstacles ahead and plot a route accordingly, or do they move under the blind push of instantaneous forces?

Here is the incisive answer that brings all of classical mechanics, quantum mechanics, and general relativity under a single roof: the principle of least action.

When the Lagrangian $L(q, \dot{q}, t)$ that we defined on the tangent bundle ( $t_1$ ) is integrated along a trajectory $t_2$ followed by the system from a specified time $(\gamma)$ to $S$ , the result is a scalar called the action, $S$ :

S(\gamma) = \int_{t_1}^{t_2} L(q(t), \dot{q}(t), t)\,dt \tag{14}

This is not a function but a functional—a function of functions. There are infinitely many possible trajectories that the system can trace through configuration space. Nature, however, is exceedingly economical, even "lazy." The universe chooses the particle's actual trajectory in such a way that the action integral $\delta S = 0$ takes an extremal value, generally a minimum. In other words, the system follows the route whose variation vanishes, $\delta S = 0$ . It is as though the particle "sniffs out" all possible paths at once, selects the unique geodesic that minimizes the action, and glides along it.

What Is a Variation? The Virtual Dance of Trajectories

We have said that the route is the one whose variation vanishes, $\delta$ , but we must clarify the physical and mathematical meaning of this operator $(d/dt)$ , the variation.

In ordinary differential calculus, $(\delta)$ examines the change in a particle's position as time passes. In the calculus of variations, $q(t)$ describes something entirely different: it freezes time, takes the entire smooth trajectory drawn by the system in configuration space, and bends it hypothetically.

Suppose that the perfect, actual trajectory chosen by nature is $\delta q(t)$ . Imagine a neighboring trajectory obtained by adding a very small virtual displacement to it:

\tilde{q}(t) = q(t) + \delta q(t) \tag{15}

The function $(t_1)$ is called the variation of the trajectory. Nature imposes a very strict rule on these virtual trials, however: the point from which the system departs at $(t_2)$ and the point at which it arrives at $\delta S = 0$ are fixed. The endpoints are nailed to the wall, so the variations must vanish at the initial and final times:

\delta q(t_1) = 0, \quad \delta q(t_2) = 0 \tag{16}

This is the profound truth expressed by the principle $\delta q$ , the stationarity of the action: every infinitesimal imaginary bending—every variation $S$ —of the actual trajectory that leaves its endpoints fixed produces no first-order change in the action integral $df = 0$ . Just as the slope of the tangent is zero at the minimum of a function in standard calculus, $L = T - V$ , the variation of nature's chosen path vanishes in the vast functional space over the configuration manifold. The particle finds the extremal point where the topography of the action seems to flatten and glides through it.

8 Why $L = T - V$ ? The Rigid Demand of Symmetry

In textbooks on analytical mechanics or the first lectures in theoretical-physics courses, we generally begin with a decree from above: the Lagrangian is the kinetic energy minus the potential energy, $L = T + V$ . But why? Why has the universe chosen this specific structure instead of a formula such as $L = T^2/V$ or $L = T - V$ ?

Here lies the greatest stroke of genius in Lev Landau's construction of mechanics. Landau proves that the equation $(\mathbf{r})$ did not simply fall from the sky; it is a mathematical necessity imposed by the symmetries of the Galilean group that we defined at the beginning of this article.

Let us see this mathematically, step by step. Imagine a completely free particle—one on which no force acts—in an inertial reference frame. The Lagrangian describing this particle may initially be an arbitrary function of position $(\mathbf{v} = \dot{\mathbf{r}})$ , velocity $(t)$ , and time $L(\mathbf{r}, \mathbf{v}, t)$ : $\Delta\mathbf{r}$ .

The symmetries of nature immediately begin to constrain this function:

Homogeneity of Space: There is no privileged point in the universe. Translating the particle by $L$ in space changes no physical law. The Lagrangian therefore cannot depend on the position vector $(\mathbf{r})$ : $\frac{\partial L}{\partial \mathbf{r}} = 0$ .
Homogeneity of Time: The rules of physics are the same today as they were yesterday. Since time has no absolute beginning, $L$ cannot depend explicitly on time: $\frac{\partial L}{\partial t} = 0$ .
Isotropy of Space: There is no privileged direction in space. Therefore, $L$ can depend not on the direction of the velocity vector $(\mathbf{v})$ but only on its scalar magnitude, or more precisely on the square of that magnitude, $(v^2 = \mathbf{v} \cdot \mathbf{v})$ .

As a result of all these geometric constraints, the Lagrangian of a free particle simplifies considerably: $L = L(v^2)$ .

What is the exact form of the function? To find it, we must use the Galilean principle of relativity. The rules must be exactly the same in two inertial frames, $\boldsymbol{\epsilon}$ and $K$ , moving relative to each other with the constant velocity $K'$ . The velocity in frame $K'$ is $\mathbf{v}' = \mathbf{v} + \boldsymbol{\epsilon}$ . For a very small, infinitesimal velocity $\boldsymbol{\epsilon}$ , expanding the new Lagrangian in a Taylor series gives

L(v'^2) = L\!\left((\mathbf{v}+\boldsymbol{\epsilon})^2\right) = L(v^2 + 2\mathbf{v}\cdot\boldsymbol{\epsilon} + \epsilon^2) \approx L(v^2) + \frac{\partial L}{\partial v^2}(2\mathbf{v}\cdot\boldsymbol{\epsilon}) \tag{17}

For the two reference frames to yield exactly the same physical equations of motion, the difference between these two Lagrangians must be a total time derivative of position and time, $\left(\frac{dF}{dt}\right)$ . Let us examine the difference term more closely:

\delta L = 2\frac{\partial L}{\partial v^2}\mathbf{v}\cdot\boldsymbol{\epsilon} = 2\frac{\partial L}{\partial v^2}\frac{d\mathbf{r}}{dt}\cdot\boldsymbol{\epsilon} \tag{18}

This expression can be written in the pure form $\frac{d}{dt}(\ldots)$ if and only if the term $\frac{\partial L}{\partial v^2}$ is a constant independent of velocity. Calling this constant $\frac{1}{2}m$ , where $m$ is the particle's mass, and integrating the derivative back, we arrive with remarkable certainty at

L(v^2) = \frac{1}{2}mv^2 = T \tag{19}

As we can see, the form of the kinetic energy—its proportionality to $v^2$ —is dictated entirely by the isotropy of space and Galilean transformations.

If an interaction field in space breaks the particle's freedom and the homogeneity of space, as gravity does, that field must enter the action as a scalar function of the coordinates. By nature's principle of simplicity, this interaction term enters the Lagrangian only as an addition $-V(\mathbf{r})$ depending on position. The Galilean principle of relativity and the homogeneity of the universe thus lead us to the inevitable formula

L = \frac{1}{2}mv^2 - V(\mathbf{r}) = T - V \tag{20}

9 From Geometry to Dynamics: Deriving the Euler–Lagrange Equations

We have defined the philosophy of minimizing the action, $\delta S = 0$ , and virtual displacements, $(\delta q)$ . Let us now turn this geometric idea into rigorous mathematics. Taking the variation of the action integral gives

\delta S = \delta \int_{t_1}^{t_2} L(q, \dot{q}, t)\,dt = \int_{t_1}^{t_2} \left(\frac{\partial L}{\partial q}\delta q + \frac{\partial L}{\partial \dot{q}}\delta\dot{q}\right)dt = 0 \tag{21}

We must now take a crucial step. The term $\delta\dot{q}$ is in fact the time derivative of the variation, $\delta\dot{q} = \frac{d}{dt}(\delta q)$ . Substituting this into the integral and applying integration by parts to the second term gives

\int_{t_1}^{t_2} \frac{\partial L}{\partial \dot{q}}\frac{d}{dt}(\delta q)\,dt = \left[\frac{\partial L}{\partial \dot{q}}\delta q\right]_{t_1}^{t_2} - \int_{t_1}^{t_2} \frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}}\right)\delta q\,dt \tag{22}

Recall that when nature allowed virtual variations, it nailed the endpoints to the wall. At times $t_1$ and $t_2$ , the variation is zero: $\delta q(t_1) = 0$ and $\delta q(t_2) = 0$ . Thanks to this strict rule, the boundary term on the right—the expression in square brackets—vanishes completely. Substituting the remaining integral into the original equation gives

\delta S = \int_{t_1}^{t_2} \left[\frac{\partial L}{\partial q} - \frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}}\right)\right]\delta q\,dt = 0 \tag{23}

Here the fundamental lemma of the calculus of variations enters. The virtual displacement $t_1$ that we choose between $t_2$ and $(\delta q)$ is entirely arbitrary. If an integral multiplied by an arbitrary function $\delta q$ always evaluates to zero, the expression in brackets must itself vanish everywhere. This yields the famous differential equations that are entirely independent of the coordinates of the configuration manifold:

\frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}_i}\right) - \frac{\partial L}{\partial q_i} = 0 \quad (i = 1, 2, \ldots, n) \tag{24}

In Newtonian mechanics, $F = ma$ , changing from Cartesian coordinates to spherical or cylindrical coordinates is a nightmare; the acceleration terms fill with bewildering cross products. The Euler–Lagrange equations we have derived, however, are covariant. Whether the coordinates $q_i$ used to describe our space are Cartesian coordinates, angles, or a length along a helix as in our roller-coaster example, the form of this differential equation never changes. No matter how contorted the metric of configuration space is, this equation finds the correct trajectory in that space perfectly.

10 Awakening the System: Solving the Equations of Motion

We now have everything we need. It is time to "awaken" the complex helical roller-coaster pendulum that we constructed in Section 6. We found its kinetic energy $(T)$ and potential energy $(V)$ . The system's enormous Lagrangian is

L = \frac{1}{2}\left[m_1(R^2+c^2) + m_2((R+l\sin q_2)^2+c^2)\right]\dot{q}_1^2 + \frac{1}{2}m_2 l^2\dot{q}_2^2 + m_2 cl\sin q_2\,\dot{q}_1\dot{q}_2 - V(q_1, q_2) \tag{25}

Finding the equations of motion for this system requires no complicated free-body diagrams or three-dimensional vector projections. All we have to do is feed the function $L$ into the Euler–Lagrange machine. Consider, for example, the coordinate $q_2$ , which describes the pendulum's vertical oscillation:

\frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}_2}\right) - \frac{\partial L}{\partial q_2} = 0 \tag{26}

First, let us find the partial derivative with respect to the velocities—the generalized momentum—and differentiate it with respect to time:

\frac{\partial L}{\partial \dot{q}_2} = m_2 l^2\dot{q}_2 + m_2 cl\sin q_2\,\dot{q}_1 \implies \frac{d}{dt}(\ldots) = m_2 l^2\ddot{q}_2 + m_2 cl\cos q_2\,\dot{q}_1\dot{q}_2 + m_2 cl\sin q_2\,\ddot{q}_1 \tag{27}

Now take the partial derivative with respect to the position $(q_2)$ , the generalized force. Note that the derivative arising from the potential energy is $-\frac{\partial V}{\partial q_2} = -m_2 gl\sin q_2$ :

\frac{\partial L}{\partial q_2} = m_2 l(R + l\sin q_2)\cos q_2\,\dot{q}_1^2 + m_2 cl\cos q_2\,\dot{q}_1\dot{q}_2 - m_2 gl\sin q_2 \tag{28}

When we subtract these two expressions, the wonderful mathematical "cancellation" occurs: the cross terms containing $\dot{q}_1\dot{q}_2$ cancel, leaving the pure equation of motion:

m_2 l^2\ddot{q}_2 + m_2 cl\sin q_2\,\ddot{q}_1 - m_2 l(R + l\sin q_2)\cos q_2\,\dot{q}_1^2 + m_2 gl\sin q_2 = 0 \tag{29}

Everything that we could never see so easily in Newton's force diagrams appears before us at once: the pendulum's moment of inertia $(m_2 l^2\ddot{q}_2)$ , the interaction coupling caused by acceleration along the helix $(m_2 cl\sin q_2\,\ddot{q}_1)$ , the centrifugal term that throws the pendulum outward as the system rotates $(m_2 l(R + l\sin q_2)\cos q_2\,\dot{q}_1^2)$ , and gravity's delicate projection $(m_2 gl\sin q_2)$ . All of it arose from pure geometry!

A Geometric Inference: Vector Fields and Phase Flow

Arnold does not regard the acceleration equations $(\ddot{q})$ we obtained as merely "an algebra problem to be solved." This system of differential equations is in fact an enormous vector field defined on the $2n$ -dimensional tangent bundle $(TM)$ .

The evolution of the system from time $t_1$ to $t_2$ is the instantaneous state $(q, \dot{q})$ being caught in the currents of this vector field and carried along like a drop of fluid. Arnold calls this the phase flow. The fundamental laws of nature weave through space from end to end as the integral curves of this flow.

11 The Birth of Energy and a First Look at Hamilton: The Jacobi Integral

Is there nothing that remains unchanged as the system is carried through this vector field? At this point, Arnold uses the homogeneity of time to prove the conservation of energy not through formulas handed down from above, but directly through Euler–Lagrange.

If the external conditions of the system do not change with time, $\left(\frac{\partial L}{\partial t} = 0\right)$ , let us expand the total time derivative of the Lagrangian using the chain rule:

\frac{dL}{dt} = \sum_i \frac{\partial L}{\partial q_i}\dot{q}_i + \sum_i \frac{\partial L}{\partial \dot{q}_i}\ddot{q}_i \tag{30}

Using the Euler–Lagrange equation to replace $\frac{\partial L}{\partial q_i}$ with $\frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}_i}\right)$ gives

\frac{dL}{dt} = \sum_i \left[\frac{d}{dt}\!\left(\frac{\partial L}{\partial \dot{q}_i}\right)\dot{q}_i + \frac{\partial L}{\partial \dot{q}_i}\ddot{q}_i\right] = \frac{d}{dt}\!\left(\sum_i \frac{\partial L}{\partial \dot{q}_i}\dot{q}_i\right) \tag{31}

Gathering all derivatives on one side reveals a magnificent mathematical invariant, the Jacobi integral:

\frac{d}{dt}\!\left(\sum_{i=1}^{n} \dot{q}_i\frac{\partial L}{\partial \dot{q}_i} - L\right) = 0 \implies E = \sum_{i=1}^{n} \dot{q}_i\frac{\partial L}{\partial \dot{q}_i} - L = \text{Sabit} \tag{32}

Through the homogeneity of time, nature conserves this immense function $E$ . The mathematical form hidden within this conservation law, $(\sum \dot{q}p - L)$ , is far more than a simple energy. It is a hidden bridge that will carry us away from the tangent world of velocities and into the cotangent bundle $(T^*M)$ —in other words, into phase space, where the true heart of mechanics beats.

To name this bridge and see how it propels us into that entirely new universe, in the next part we will set sail toward Hamilton's flawless geometric revolution.

References

Arnol'd, V. I. (1989). Mathematical Methods of Classical Mechanics (2nd ed.). Springer-Verlag.
Landau, L. D., & Lifshitz, E. M. (1976). Mechanics (Vol. 1, 3rd ed.). Butterworth-Heinemann