Potential Energy Diagrams and Equilibrium

A million and a half kilometres from Earth, the James Webb Space Telescope orbits an empty point in space. Nothing is there — no mass, no station, nothing to orbit around in the ordinary sense. What holds the telescope loosely in place is the shape of an energy landscape: the Sun-Earth L2 point is a saddle of an effective potential, a mountain pass in an invisible terrain, and the spacecraft's tiny thruster burns are the physics of balancing near an unstable equilibrium, performed as a trillion-dollar engineering routine. That is the power of a potential energy diagram. Plot $U(x)$ against position for any one-dimensional conservative system and the entire dynamical story is encoded in one curve: the force is the negative slope, the speed anywhere is set by the gap down from the conserved-energy line, motion reverses at the shoreline where the line meets the curve, valleys trap and oscillate, peaks repel, and the curvature at the bottom of any well dictates the rhythm of small vibrations. Molecular bonds, pendulums, satellites, and buckling beams all become the same picture read at different scales. This article teaches you to read that picture fluently — turning points, stable and unstable equilibria, the harmonic approximation $\omega = \sqrt{U''(x_0)/m}$ that makes every smooth valley a simple harmonic oscillator, and the separatrix energies where the character of motion changes qualitatively. No differential equation gets solved until the diagram has already told you the answer.

Key takeaways

For 1D conservative motion the force is the negative slope of the potential curve: $F(x) = -\mathrm{d}U/\mathrm{d}x$ — the particle is always pushed toward lower $U$ .
Energy conservation $E = \tfrac12 m\dot x^2 + U(x)$ makes the vertical gap between the horizontal energy line and the curve equal to the kinetic energy; the motion is confined to $U(x)\le E$ .
Turning points sit where $U(x) = E$ ; between two of them the motion is bounded and periodic, with period given by a quadrature over the diagram.
Equilibria are the flat points $U'(x_0) = 0$ : minima are stable, maxima unstable, and the curvature $U''(x_0)$ decides which.
Near any stable minimum, every smooth potential is a harmonic oscillator with $\omega = \sqrt{U''(x_0)/m}$ — the single most-used fact extracted from these diagrams.
Separatrix energies (energy lines through a local maximum) divide qualitatively different motions — oscillation from escape, libration from rotation — and the period diverges there.
Key idea: tilt your eye to the curve $U(x)$ as if it were a frictionless hill — a ball on it rolls downhill (force $=-U'$ ), oscillates in valleys (stable equilibria), balances precariously on peaks (unstable equilibria), and stops where the curve rises to meet the energy line (turning points).

The core insight

Tilt your eye to the curve $U(x)$ as if it were a frictionless hill: a ball placed on it rolls downhill (force $=-U'$ ), oscillates in valleys (stable equilibria), balances precariously on peaks (unstable equilibria), and stops where the curve rises to meet the energy line (turning points).

Where this lives today

The largest instrument humanity has ever parked at an unstable equilibrium is the James Webb Space Telescope, orbiting the Sun-Earth L2 point. In the co-rotating frame, L2 is a saddle of the effective potential — stable to sideways nudges, exponentially unstable along the Sun-Earth line, with an e-folding time of about 23 days (Derivation C's $e^{\gamma t}$ , at astronomical scale). Mission design embraces the diagram: rather than sit on the saddle point, JWST flies a wide halo orbit around it and spends a few m/s of thruster budget per year canceling the runaway mode — balancing the pencil, forever, by gentle taps. At the opposite scale, potential energy diagrams are the daily language of chemistry and materials science. A reaction rate is set by a barrier height on an energy landscape (the Arrhenius picture); a molecule's infrared spectrum is equation (8) applied to the curvature of its bond potential; and computational drug design amounts to mapping multidimensional $U$ surfaces and hunting their minima and saddle points with algorithms. When a machine-learning force field predicts a new battery electrolyte, the deliverable is, quite literally, a better potential energy diagram.

The physical picture

Picture the potential curve as the cross-section of a landscape and the total energy $E$ as a flood level. Wherever the water sits above the land, the particle can roam; where the land pokes above the water, motion is forbidden. The shoreline — where $E = U$ — is exactly the set of turning points. Raise the flood and shorelines migrate, pools merge, and a particle trapped in one valley may suddenly gain the run of the whole terrain: qualitative changes in motion are geometric events in the diagram. The slope of the land is the force, $|F| = |U'|$ , so the particle accelerates hardest where the terrain is steepest and coasts through flat regions. At a valley bottom the slope vanishes — no force, an equilibrium — but the curvature is positive, so any displacement is met by a push back toward the bottom. On a hilltop the slope also vanishes, but the curvature is negative and any displacement is amplified. Stability is a second-derivative question, and the diagram answers it at a glance.

Figure

Reading a double well — equilibria, turning points, and two flood levels with different fates

Curvature is destiny for small motions. Zoom into the bottom of any smooth valley and it looks parabolic — $U \approx U_0 + \tfrac12 U''(x_0)(x-x_0)^2$ — so the particle oscillates sinusoidally, with a sharper well (larger $U''$ ) giving a faster rhythm. This is why the harmonic oscillator is everywhere in physics: it is not an assumption about the world but a theorem about Taylor series. The one caution, developed below, is the marginal case $U'' = 0$ , where the parabola degenerates and higher-order terms take over.

Figure

Curvature sets the clock — two wells of equal depth, different stiffness

The core equations

The dictionary between curve and dynamics starts with the defining property of a conservative force in one dimension. The potential stores minus the work the force does, $\mathrm{d}U = -F\,\mathrm{d}x$ , so differentiating recovers the force from the landscape:

F(x) = -\frac{\mathrm{d}U}{\mathrm{d}x} \qquad (1)

where: $F$ — the (one-component) force; $U(x)$ — the potential energy. The minus sign is the geometry of the whole subject: positive slope means a force in the $-x$ direction. The particle is always pushed downhill on the graph — which, beware, is a statement about the energy axis, not about physical height. Multiplying Newton's second law $m\ddot x = -U'(x)$ by $\dot x$ and recognizing total derivatives yields the conserved quantity that the horizontal line represents:

E = \tfrac{1}{2}m\dot{x}^2 + U(x) = \text{constant} \qquad (2)

where: $E$ — the total mechanical energy, fixed by initial conditions. Every horizontal line you draw on the diagram is a possible motion; the vertical gap from line down to curve is the kinetic energy at that position. Solving (2) for the velocity turns the picture into numbers:

v(x) = \pm\sqrt{\frac{2\bigl(E - U(x)\bigr)}{m}} \qquad (3)

where: the sign tracks the direction of travel. The square root exists only in the allowed region $U \le E$ ; the particle moves fastest where the curve dips lowest beneath the line. The shoreline condition deserves its own display, because it controls the qualitative structure of every motion:

U(x_t) = E \quad\Rightarrow\quad v(x_t) = 0 \qquad (4)

where: $x_t$ — a turning point. If the slope $U'(x_t) \neq 0$ there, the force is nonzero and the particle reverses; a particle caught between two turning points shuttles between them forever (in the conservative idealization). Where can the particle simply sit? Only where the force vanishes — the flat points of the landscape:

\left.\frac{\mathrm{d}U}{\mathrm{d}x}\right|_{x_0} = 0 \qquad (5)

where: $x_0$ — an equilibrium position. Whether it can stay near $x_0$ after a nudge is the stability question, and the second derivative answers it:

U''(x_0) > 0 \ \text{stable}, \qquad U''(x_0) < 0 \ \text{unstable}, \qquad U''(x_0) = 0 \ \text{higher-order test} \qquad (6)

where: the three cases are valley, hilltop, and marginal. Derivation C makes the classification rigorous by solving the linearized motion in each case. Near a stable minimum, Taylor's theorem hands us the most useful approximation in physics. The linear term dies by (5), and what remains to leading order is a parabola:

U(x) \approx U(x_0) + \tfrac{1}{2}U''(x_0)\,(x - x_0)^2 \qquad (7)

where: the curvature $U''(x_0)$ plays the role of a spring constant $k$ . Newton's law in this parabola is simple harmonic motion, with

\omega = \sqrt{\frac{U''(x_0)}{m}}, \qquad T_{\text{small}} = 2\pi\sqrt{\frac{m}{U''(x_0)}} \qquad (8)

where: $\omega$ — the angular frequency of small oscillations, amplitude-independent in this limit. Measure the curvature of any well — molecular, mechanical, orbital — and you have predicted its vibrational frequency. Away from the small-amplitude limit, the exact period is still a pure diagram quantity — an integral over the allowed region between the turning points:

T(E) = \sqrt{2m}\,\int_{x_1(E)}^{x_2(E)} \frac{\mathrm{d}x}{\sqrt{E - U(x)}} \qquad (9)

where: $x_1, x_2$ — the turning points at energy $E$ . In general $T$ depends on $E$ (equivalently on amplitude); the harmonic well is the special case where it does not. As $E$ approaches a barrier top the integral diverges — the particle takes forever to creep up to an unstable equilibrium.

Which equation answers which question

you want	use	notes
the force at a glance	slope, eq. (1)	force points downhill on the graph
speed at a position	gap, eq. (3)	biggest where the curve dips lowest
where motion reverses	shoreline, eq. (4)	intersections of line and curve
where a particle can rest	flat points, eq. (5)	then classify with (6)
vibration frequency of a well	curvature, eq. (8)	the universal harmonic limit
exact period at finite amplitude	quadrature, eq. (9)	diverges at the separatrix

Derivation APeriod of bounded motion as a quadrature

We are deriving equation (9): the exact oscillation period of a particle trapped between two turning points, expressed as an integral over the potential diagram — no solution of Newton's equation required. The strategy is separation of variables: energy conservation gives speed as a function of position, and time is distance over speed, integrated. Two points need care: the integrand blows up at both turning points (the speed vanishes there), so we must check the integral still converges; and the half-period must be argued from time-reversal symmetry, not assumed.

Step 1. Rearrange the energy law (2) for the speed on the outbound leg:

\dot{x} = \frac{\mathrm{d}x}{\mathrm{d}t} = \sqrt{\frac{2}{m}\bigl(E - U(x)\bigr)}

Why the positive root: we follow the half of the motion from the left turning point $x_1$ to the right one $x_2$ ; on this leg the particle moves in the $+x$ direction throughout.

Step 2. Separate variables — time becomes an integral over position:

\mathrm{d}t = \sqrt{\frac{m}{2}}\;\frac{\mathrm{d}x}{\sqrt{E - U(x)}}

Step 3. Integrate across the allowed region and identify the result as half the period:

\frac{T}{2} = \sqrt{\frac{m}{2}}\int_{x_1}^{x_2}\frac{\mathrm{d}x}{\sqrt{E - U(x)}}

Why half: conservative 1D dynamics is time-reversal symmetric, so the return trip from $x_2$ to $x_1$ retraces the same speeds in reverse and takes exactly the same time as the outbound trip.

Step 4. Double it to obtain equation (9), and check convergence at the endpoints. Near a generic turning point $U'(x_t)\ne0$ , so $E - U \approx |U'(x_t)|\,|x - x_t|$ and the integrand behaves as $|x - x_t|^{-1/2}$ — integrable. But if a turning point sits at a maximum of $U$ (so $U' = 0$ there), then $E - U \sim \tfrac12|U''|(x - x_t)^2$ , the integrand behaves as $|x - x_t|^{-1}$ , and the integral diverges logarithmically. Why the divergence is physical, not a defect: at the separatrix energy the particle approaches the unstable equilibrium ever more slowly, taking infinite time to arrive — the period smoothly grows without bound as $E$ approaches the barrier top. *As a check: for $U = \tfrac12 kx^2$ the integral evaluates to $T = 2\pi\sqrt{m/k}$ , the familiar spring period, independent of $E$ — confirming that the harmonic well is the one case where the period does not depend on amplitude.*

Go deeper · Derivation B · 1 minEvery smooth well is a harmonic oscillator

Derivation BEvery smooth well is a harmonic oscillator

The goal is the harmonic approximation, equations (7) and (8): sufficiently small oscillations about any stable minimum are simple harmonic with frequency set by the curvature alone. This single result is why one solvable system — the harmonic oscillator — dominates all of physics. The route is a Taylor expansion whose linear term is killed by the equilibrium condition, followed by Newton's law on what survives. Watch the hypotheses: $U$ must be smooth at $x_0$ , and $U''(x_0)$ must be strictly positive — the marginal case falls outside this derivation and is handled separately.

Step 1. Shift coordinates to the equilibrium: let $\xi \equiv x - x_0$ where $U'(x_0) = 0$ .

Step 2. Taylor-expand the potential about $\xi = 0$ :

U = U(x_0) + U'(x_0)\,\xi + \tfrac{1}{2}U''(x_0)\,\xi^2 + \mathcal{O}(\xi^3)

Why the linear term vanishes: that is precisely the equilibrium condition (5) — expanding about anywhere else would leave a constant force term.

Step 3. Differentiate to get the force and insert into Newton's second law:

m\ddot{\xi} = -U''(x_0)\,\xi + \mathcal{O}(\xi^2)

Why we may drop the higher terms: for amplitude $a$ , the cubic force correction is smaller than the linear term by a factor of order $a\,|U'''|/U''$ — small oscillations are precisely those with $a$ small enough that this ratio is negligible.

Step 4. Recognize the equation of simple harmonic motion and read off the frequency:

\ddot{\xi} + \omega^2\xi = 0, \qquad \omega = \sqrt{\frac{U''(x_0)}{m}}, \qquad \xi(t) = A\cos(\omega t + \phi)

The frequency depends only on curvature and mass — not on amplitude. Anharmonic corrections enter as the amplitude grows, which is why a pendulum's period lengthens at large swing. Two quick checks: for $U = \tfrac12 kx^2$ , $U'' = k$ gives $\omega = \sqrt{k/m}$ , the standard spring result; for the pendulum $U = -mgL\cos\theta$ , the effective curvature at $\theta = 0$ is $mgL$ and the moment of inertia is $mL^2$ , giving $\omega = \sqrt{g/L}$ — both as expected.

Go deeper · Derivation C · 1 minStability decided by the sign of the curvature

Derivation CStability decided by the sign of the curvature

Here we prove the classification (6) — minima stable, maxima unstable — by actually solving the linearized motion near an equilibrium, rather than waving at the picture. The strategy: the same expansion as Derivation B, but now keeping track of the sign of $U''(x_0)$ , which flips the character of the solutions from oscillatory to exponential. The marginal case $U'' = 0$ is where linear analysis honestly fails, and we treat it by example rather than pretend the theorem covers it.

Step 1. Linearize about the equilibrium as before:

m\ddot{\xi} = -U''(x_0)\,\xi

Why this equation covers both cases: the Taylor expansion never asked whether $x_0$ was a valley or a hilltop — only the sign of the curvature differs.

Step 2. Positive curvature. With $U'' > 0$ , the solutions are bounded oscillations $\xi = A\cos(\omega t + \phi)$ : a small nudge stays small forever. The equilibrium is stable (a valley).

Step 3. Negative curvature. With $U'' < 0$ , define $\gamma = \sqrt{-U''(x_0)/m}$ ; the general solution is

\xi(t) = Ae^{\gamma t} + Be^{-\gamma t}

Why instability is generic: any perturbation with $A \neq 0$ — that is, all of them except a perfectly tuned set of measure zero — grows exponentially with e-folding time $1/\gamma$ . A pencil on its tip falls not because balance is impossible but because the required tuning is infinitely precise.

Step 4. Marginal curvature. With $U''(x_0) = 0$ the linear term gives no verdict; the first non-vanishing derivative decides. For $U \propto \xi^4$ the equilibrium is stable (a flat-bottomed valley, with amplitude-dependent frequency); for $U \propto \xi^3$ it is half-stable — restoring on one side, runaway on the other — the saddle-node situation that governs bifurcations. This result is the content of the Lagrange-Dirichlet theorem: an isolated minimum of $U$ is Lyapunov-stable.

Limiting cases

Case	Result	Meaning
$E \to U_{\min}^{+}$	harmonic motion, $T = 2\pi\sqrt{m/U''}$	tiny oscillations in the parabolic bottom; amplitude-independent period
$E \to U_{\max}^{-}$ (separatrix)	$T \to \infty$ logarithmically	the particle creeps up to the unstable equilibrium and never quite arrives
$E$ above all barriers	no turning point on one side	unbounded escape; the diagram predicts scattering, not oscillation
$U = \text{const}$	$F = 0$ everywhere	neutral equilibrium at every point; uniform motion (Newton's first law)
large amplitude in an anharmonic well	$T = T(E)$	period picks up energy dependence; the pendulum's swing slows as it widens
$U''(x_0) = 0$	linear test silent	quartic minimum: stable but non-isochronous; cubic inflection: half-stable

Common trapThe particle rolls along the U curve like a ball on a hill

The graph of $U(x)$ is not a track in space, and the particle does not move along it. The vertical axis is energy, not height: the physical motion is entirely along the horizontal $x$ -axis, with the curve silently dictating force and speed. The rolling-ball analogy is useful for intuition about equilibria — and it is quantitatively wrong for dynamics: a bead on a literal frictionless wire shaped like $U(x)$ obeys a different equation of motion, because part of its kinetic energy lives in the vertical velocity component (the discrepancy grows with slope). Trust the analogy for where things settle and which way they are pushed; trust only equations (1)-(9) for how fast. And keep the sign straight: positive slope means force in the $-x$ direction.

Worked examples

1 · Reading a spring's diagram (confidence builder)

A $0.50\ \mathrm{kg}$ glider on a frictionless air track is attached to a spring with $U(x) = \tfrac12 kx^2$ , $k = 200\ \mathrm{N/m}$ . Its total energy is $E = 1.0\ \mathrm{J}$ . Find the turning points, the maximum speed, and the oscillation frequency.

Step 1. Turning points from eq. (4): $\tfrac12 kx_t^2 = E$ gives $x_t = \pm\sqrt{2E/k} = \pm\sqrt{2(1.0)/200} = \boxed{\pm 0.10\ \mathrm{m}}$ — the shoreline sits 10 cm either side of the origin.

Step 2. Maximum speed where the gap $E - U$ is largest, i.e. at $x = 0$ where $U = 0$ : $v_{\max} = \sqrt{2E/m} = \sqrt{2(1.0)/0.50} = \boxed{2.0\ \mathrm{m/s}}$ .

Step 3. Frequency from the curvature, eq. (8): $U'' = k$ , so $\omega = \sqrt{k/m} = \sqrt{200/0.50} = 20\ \mathrm{rad/s}$ , i.e. $f = \omega/2\pi = \boxed{3.2\ \mathrm{Hz}}$ , period $T = 0.31\ \mathrm{s}$ .

Step 4. Consistency check: for SHM, $v_{\max} = \omega A = 20 \times 0.10 = 2.0\ \mathrm{m/s}$ ✓ — the diagram's gap reading and the oscillator formula agree exactly, as they must for a parabolic well.

2 · A double well, fully read (exam level)

A particle of mass $m = 2.0\ \mathrm{kg}$ moves in $U(x) = x^4 - 2x^2$ (joules, with $x$ in metres). Find and classify all equilibria; find the small-oscillation frequency in a well; and for total energy $E = -0.50\ \mathrm{J}$ , locate the turning points and describe the motion.

Step 1. Equilibria from eq. (5): $U'(x) = 4x^3 - 4x = 4x(x^2 - 1) = 0$ , so $x_0 = 0, \pm 1$ .

Step 2. Classify with $U''(x) = 12x^2 - 4$ . At $x = \pm1$ : $U'' = +8 > 0$ — two stable minima, depth $U(\pm1) = 1 - 2 = -1.0\ \mathrm{J}$ . At $x = 0$ : $U'' = -4 < 0$ — unstable maximum (the barrier), $U(0) = 0$ .

Step 3. Small oscillations in either well, eq. (8): $\omega = \sqrt{U''(\pm1)/m} = \sqrt{8/2.0} = \boxed{2.0\ \mathrm{rad/s}}$ , period $T = 2\pi/\omega = 3.1\ \mathrm{s}$ .

Step 4. Turning points at $E = -0.50\ \mathrm{J}$ : solve $x^4 - 2x^2 = -0.50$ , i.e. $x^4 - 2x^2 + 0.50 = 0$ , a quadratic in $x^2$ : $x^2 = \dfrac{2 \pm \sqrt{4 - 2}}{2} = 1 \pm \dfrac{\sqrt{2}}{2}$ , giving $x^2 = 1.707$ or $x^2 = 0.293$ .

Step 5. So $x_t = \pm 1.31\ \mathrm{m}$ and $\pm 0.54\ \mathrm{m}$ . Since $E = -0.50\ \mathrm{J}$ lies below the barrier top ( $U = 0$ ), the particle cannot cross $x = 0$ : started in the right well, it oscillates between $\boxed{x = 0.54\ \mathrm{m}\ \text{and}\ x = 1.31\ \mathrm{m}}$ , asymmetrically about the minimum at $x = 1$ .

Step 6. Epilogue the diagram gives for free: raise the energy to any $E > 0$ and the two turning-point pairs merge into one ( $x_t = \pm\sqrt{1 + \sqrt{1+E}}$ ), the pools connect, and the particle sweeps both wells in a single slow oscillation — passing over the old barrier with speed $v = \sqrt{2E/m}$ at $x=0$ . Exactly at $E = 0$ the period diverges: the separatrix.

Origins · 1 minJoseph-Louis Lagrange and how this was discovered

Joseph-Louis Lagrange

1736–1813 · Turin, Berlin, Paris

Lagrange rebuilt mechanics as pure analysis — his masterpiece Mécanique analytique (1788) famously contains not a single diagram, replacing Newton's geometry with the calculus of a single scalar function. It is a delicious irony that the potential-energy pictures in this article are the fastest road into his own results: Lagrange proved that a configuration where the potential energy is at a minimum is stable (the rigorous proof was completed by Dirichlet in 1846, and the result is now the Lagrange-Dirichlet theorem), which is exactly the valley-versus-hilltop reading of the diagram. Born Giuseppe Lodovico Lagrangia in Turin, self-taught from a library book at seventeen, he succeeded Euler in Berlin at Frederick the Great's invitation and later survived the French Revolution to help design the metric system. The five equilibrium points of the restricted three-body problem — including the L2 saddle where the James Webb Space Telescope now station-keeps — carry his name: unstable equilibria he classified with pen and paper, two and a half centuries before spacecraft learned to balance on them.

History

1687

— Newton's Principia frames dynamics in terms of forces; energy landscapes are still a century away.

1788

— Lagrange's Mécanique analytique recasts mechanics around scalar functions and proves stability at potential minima.

1834

— Hamilton reformulates dynamics so that the energy function itself generates the motion — the diagram becomes the dynamics.

1846

— Dirichlet supplies the rigorous proof of the Lagrange stability criterion, closing a sixty-year gap.

1892

— Lyapunov's doctoral thesis founds modern stability theory, generalizing the curvature test far beyond mechanics.

Going further

Texts. Morin, Introduction to Classical Mechanics, ch. 5 (Conservation of Energy and Momentum) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §3.3 (the equivalent one-dimensional problem) · Landau & Lifshitz, Mechanics, §11-§14 · Taylor, Classical Mechanics, ch. 4 · Kleppner & Kolenkow, An Introduction to Mechanics, ch. 5. All results are analytic; no empirical constants are quoted. The worked-example numbers — turning points $\pm0.10$ m, $v_{\max} = 2.0$ m/s, $f = 3.2$ Hz for the glider; equilibria at $0, \pm1$ with $U'' = 8$ and $-4$ , $\omega = 2.0$ rad/s, and turning points $0.54$ m and $1.31$ m at $E = -0.50$ J for the double well — are illustrative arithmetic computed directly from the boxed formulas.

Builds on: Newton's Second Law · One-Dimensional Kinematics · Velocity and Acceleration · Gravitational Force Near Earth's Surface. Leads to: Simple Harmonic Motion · Central Forces and the Effective Potential · Phase Space and Stability · Lagrangian Mechanics.

Is the particle actually rolling along the U(x) curve?

No — this is the single most common misreading, and the article's warning box exists for it. The particle moves only along the horizontal $x$ -axis; the curve is a bookkeeping device whose height at each $x$ records stored energy. The rolling-ball picture survives as an analogy because a ball on a gentle physical hill happens to feel a force proportional to the slope, just like $F = -U'$ — so intuitions about settling into valleys and toppling off peaks transfer correctly. What does not transfer is the quantitative motion: a bead on a real wire shaped like the graph carries kinetic energy in its vertical motion too, so its equation of motion contains extra slope-dependent terms. Use the landscape to see; use equations (1)-(9) to compute.

How can I read the speed off the diagram without solving anything?

Vertical distances. At position $x$ the kinetic energy is the gap from the curve up to the energy line, $T = E - U(x)$ , so $v = \sqrt{2[E - U(x)]/m}$ — equation (3). The particle is fastest where the curve dips lowest (largest gap), slows wherever the terrain climbs toward the line, and stops exactly at the shoreline. This is why you can rank speeds at different points, locate the fastest point, and identify forbidden zones literally at a glance, before any algebra. What the diagram alone cannot give you is the time anything takes — that needs the quadrature (9), because time depends on how the speed is distributed along the path, not just its values.

What is special about the energy exactly at a barrier top — the separatrix?

It is the watershed between qualitatively different motions, and strange things happen there. For $E$ just below the barrier, the particle oscillates in one well; just above, it sweeps both wells (or escapes). Exactly at $E = U_{\max}$ , the turning point coincides with the unstable equilibrium, and Derivation A's integral diverges: the particle approaches the top asymptotically, slowing exponentially, and formally never arrives. A pendulum given exactly the energy to reach the inverted position exhibits this — swinging up ever more slowly toward the top. In phase space this energy traces the separatrix, the curve separating libration from rotation; near it the period's logarithmic divergence makes the motion exquisitely sensitive, which is one gateway to chaos in driven systems.

Why does every textbook reduce everything to a harmonic oscillator? Is nature really that quadratic?

Nature isn't quadratic — Taylor series are. Any smooth potential, expanded about a stable minimum, has no linear term (that's what equilibrium means) and generically a nonzero quadratic term, so sufficiently small oscillations are harmonic regardless of the system: a molecular bond, a ship's roll, a star's pulsation mode. That is Derivation B, and it is a theorem, not a modeling choice. The fine print matters, though: "sufficiently small" is controlled by the ratio of the cubic correction to the quadratic term, and wells with $U'' = 0$ (quartic bottoms) never become harmonic — their frequency depends on amplitude all the way down. Anharmonicity is not a nuisance either; it is measurable physics, from thermal expansion (asymmetric wells) to the pendulum clock's amplitude sensitivity.

Can a particle sit at an unstable equilibrium if it is placed there perfectly at rest?

In the mathematics, yes: $x_0$ with $U'(x_0) = 0$ and $v = 0$ is an exact solution — zero force, zero velocity, forever. In practice, never: the solution is unstable in the precise sense of Derivation C — a displacement $\epsilon$ grows as $\epsilon\,e^{\gamma t}$ , so any imperfection (thermal jitter, air currents, the finite precision of placement) is amplified exponentially with e-folding time $1/\gamma = \sqrt{m/|U''|}$ . A pencil balanced on its tip has $1/\gamma$ of a fraction of a second; JWST at L2 has about 23 days, which is why station-keeping is a monthly chore rather than a millisecond panic. The deeper point: "equilibrium exists" and "equilibrium is occupiable" are different questions, and the diagram answers both — the first by slope, the second by curvature.

What changes in two or three dimensions, where U is a surface instead of a curve?

The concepts generalize cleanly; the pictures get richer. Force becomes the negative gradient, $\mathbf F = -\nabla U$ , pointing down the steepest descent. Equilibria are points where the gradient vanishes, and stability is governed by the matrix of second derivatives (the Hessian): all curvatures positive gives a bowl (stable), all negative a summit (unstable), mixed signs a saddle — stable along some directions, unstable along others, with no 1D analogue. Saddles are where the action is: chemical reactions cross energy surfaces at saddle points (transition states), and Lagrange points L1-L3 are saddles of the effective potential. One honest 1D casualty: in higher dimensions, energy conservation alone no longer confines motion to neat intervals, and trajectories can wander a surface in genuinely complicated ways.

The pendulum has infinitely many equilibria on its diagram — is that real?

Yes, and it is a good consistency check on your reading. The pendulum's potential $U(\theta) = -mgL\cos\theta$ is periodic: minima at $\theta = 0, \pm2\pi, \pm4\pi,\dots$ (all the same physical state — hanging down) and maxima at $\theta = \pm\pi, \pm3\pi, \dots$ (inverted). The diagram over one period contains the whole story: small-energy lines cut one valley (ordinary swinging, with $\omega = \sqrt{g/L}$ from the curvature $mgL$ against inertia $mL^2$ ); the line grazing the maxima is the separatrix (just barely reaching inverted); higher lines never intersect the curve at all — no turning points — which is over-the-top rotation, with $\theta$ increasing forever. Three regimes, one glance. The repetition of equilibria is just the coordinate wrapping around, but it becomes physically meaningful for systems like a particle in a crystal lattice, where each well is a genuinely distinct site.

How do I handle a diagram when friction is present?

The diagram stops being a strict conservation statement and becomes a descent picture. With weak dissipation, the energy line is no longer horizontal — it drifts slowly downward at the rate $\mathrm{d}E/\mathrm{d}t = -bv^2$ (for linear drag), so the particle spirals down its well: each oscillation slightly narrower, until it settles at the minimum. This is why stable equilibria are where things are actually found in nature — friction funnels everything into the valleys. Quantitatively you treat the motion as conservative over one cycle and let $E$ decrease adiabatically between cycles (good when the energy lost per cycle is small compared to $E - U_{\min}$ ). An unstable equilibrium stays unoccupiable; friction does not rescue the pencil, it only slows the fall.