<?xml version="1.0" encoding="UTF-8"?>

<record version="1" id="1275">
 <title>Eccentricity Vector</title>
 <name>EccentricityVector</name>
 <created>2026-09-25 00:37:31</created>
 <modified>2026-09-25 00:37:31</modified>
 <type>Definition</type>
 <creator id="1" name="bloftin"/>
 <modifier id="1" name="bloftin"/>
 <comment>rearranged to have more definition of eccentricity vector up front</comment>
 <author id="1" name="bloftin"/>
 <classification>
	<category scheme="pacs" code="45.50.Pk"/>
 </classification>
 <keywords>
	<term>eccentricity vector</term>
	<term>Laplace-Runge-Lenz vector</term>
	<term>Kepler orbit</term>
	<term>conic section</term>
	<term>periapsis</term>
	<term>angular momentum</term>
	<term>semilatus rectum</term>
	<term>true anomaly</term>
	<term>orbital elements</term>
	<term>celestial mechanics % License intent: CC BY-SA 4.0</term>
 </keywords>
 <preamble>% this is the default PhysicsLibrary preamble.  as your knowledge
% of TeX increases, you will probably want to edit this, but
% it should be fine as is for beginners.

% almost certainly you want these
\usepackage{amssymb}
\usepackage{amsmath}
\usepackage{amsfonts}

% used for TeXing text within eps files
%\usepackage{psfrag}
% need this for including graphics (\includegraphics).
% If you use \includegraphics{figure} with no file extension, PhysicsLibrary
% will prefer figure.png before falling back to figure.eps.
\usepackage{graphicx}
\DeclareGraphicsExtensions{.png,.jpg,.jpeg,.pdf,.eps}
% for neatly defining theorems and propositions
%\usepackage{amsthm}
% making logically defined graphics
%\usepackage{xypic}

% there are many more packages, add them here as you need them

% define commands here
\usepackage{booktabs}
\usepackage{array}
\usepackage{hyperref}
\hypersetup{hidelinks}
\usepackage{enumitem}
\setlist{nosep}
\setlength{\parindent}{0pt}
\setlength{\parskip}{0.72em}
\newcommand{\vect}[1]{\mathbf{#1}}</preamble>
 <content>\section{What problem does the eccentricity vector solve?}

For ideal two-body motion,
\begin{equation}
\boxed{
\ddot{\vect r}=-\frac{\mu}{r^3}\vect r
}
\label{eq:newton}
\end{equation}
with
\begin{equation}
 r=\|\vect r\|,
 \qquad
 \hat{\vect r}=\frac{\vect r}{r},
 \qquad
 \mu=G(M+m)\approx GM
\end{equation}
for a satellite whose mass is negligible relative to the central body.

Angular momentum immediately gives one conserved vector,
\begin{equation}
\boxed{
\vect h=\vect r\times\vect v,
}
\label{eq:hdef}
\end{equation}
where $\vect h$ is the \emph{specific} angular momentum. Because the gravitational acceleration is parallel to $\vect r$,
\begin{equation}
\frac{d\vect h}{dt}
=\frac{d}{dt}(\vect r\times\vect v)
=\vect v\times\vect v+\vect r\times\vect a
=\vect 0.
\end{equation}
Thus $\vect h$ is constant.

This tells us something extremely important: $\vect h$ fixes the orbital plane and the direction of positive orbital rotation. But it does \emph{not} tell us where periapsis lies inside that plane. Rotating an ellipse about its focus within the same plane leaves $\vect h$ unchanged.

We therefore seek a second conserved vector with two properties:
\begin{enumerate}
\item it should lie \emph{in} the orbital plane rather than normal to it;
\item it should identify the preferred direction of the conic, which for an ellipse is the periapsis direction.
\end{enumerate}

The eccentricity vector is precisely that object.

\begin{center}
\includegraphics[width=0.92\textwidth,keepaspectratio]{EV_eccentricity_geometry.png}

{\small Figure 1. The eccentricity vector points from the occupied focus toward periapsis. Its magnitude is the dimensionless scalar eccentricity $e$; its drawn arrow length should not be interpreted as the physical focus-to-periapsis distance.}
\end{center}

The distinction in Figure 1 is important. The magnitude
\begin{equation}
\|\vect e\|=e
\end{equation}
is dimensionless. It is not the distance from the focus to periapsis. For an ellipse that physical distance is
\begin{equation}
 r_p=a(1-e).
\end{equation}

\section{The ``discovery'' idea: build another conserved vector}

The usual formula looks mysterious if we start at the end,
\begin{equation}
\vect e
=\frac{\vect v\times\vect h}{\mu}-\hat{\vect r}.
\label{eq:epreview}
\end{equation}
A more useful way to understand it is to ask how someone could have been led to this combination.

The angular-momentum vector $\vect h$ is perpendicular to the orbital plane. The velocity $\vect v$ lies in the orbital plane. Therefore
\begin{equation}
\vect v\times\vect h
\end{equation}
is itself in the orbital plane. That makes it a natural candidate for constructing an in-plane conserved vector.

The question becomes:

\begin{quote}
What is the time derivative of $\vect v\times\vect h$ under an inverse-square central force?
\end{quote}

If that derivative can be written as the derivative of some other simple vector, the difference between the two quantities will be constant.

This is exactly what happens.

\section{A needed kinematic identity: radial distance and radial direction}

Before differentiating $\vect v\times\vect h$, we need two facts about the position vector. Write
\begin{equation}
\boxed{
\vect r=r\hat{\vect r}.
}
\label{eq:rdecomp}
\end{equation}
Differentiating gives
\begin{equation}
\vect v
=\dot r\,\hat{\vect r}
+r\frac{d\hat{\vect r}}{dt}.
\label{eq:vdecomp}
\end{equation}

Because $\hat{\vect r}$ has unit magnitude,
\begin{equation}
\hat{\vect r}\cdot\hat{\vect r}=1.
\end{equation}
Differentiating this identity gives
\begin{equation}
2\hat{\vect r}\cdot\frac{d\hat{\vect r}}{dt}=0,
\end{equation}
so
\begin{equation}
\boxed{
\hat{\vect r}\cdot\frac{d\hat{\vect r}}{dt}=0.
}
\label{eq:rhatperp}
\end{equation}
Thus the derivative of a unit vector is perpendicular to the unit vector itself: $\frac{d\hat{\vect r}}{dt}$ changes direction, not magnitude.

Dot Eq.~\eqref{eq:vdecomp} with $\hat{\vect r}$:
\begin{align}
\hat{\vect r}\cdot\vect v
&amp;=\dot r\,(\hat{\vect r}\cdot\hat{\vect r})
+r(\hat{\vect r}\cdot\frac{d\hat{\vect r}}{dt})\\
&amp;=\dot r.
\end{align}
Therefore
\begin{equation}
\boxed{
\hat{\vect r}\cdot\vect v=\dot r.
}
\label{eq:rdot}
\end{equation}
This says that the scalar radial rate is simply the component of velocity along the radial unit vector.

Solving Eq.~\eqref{eq:vdecomp} for the unit-vector derivative,
\begin{equation}
\boxed{
\frac{d\hat{\vect r}}{dt}
=\frac{1}{r}
\left(
\vect v-\dot r\,\hat{\vect r}
\right).
}
\label{eq:rhatdot}
\end{equation}
Only the transverse part of velocity rotates $\hat{\vect r}$.

\section{The key cross-product identity}

Now evaluate
\begin{equation}
\hat{\vect r}\times\vect h.
\end{equation}
Using $\vect h=\vect r\times\vect v$,
\begin{equation}
\hat{\vect r}\times\vect h
=
\hat{\vect r}\times(\vect r\times\vect v).
\end{equation}
Apply the vector triple-product identity
\begin{equation}
\vect A\times(\vect B\times\vect C)
=\vect B(\vect A\cdot\vect C)
-\vect C(\vect A\cdot\vect B).
\end{equation}
Then
\begin{align}
\hat{\vect r}\times\vect h
&amp;=
\vect r(\hat{\vect r}\cdot\vect v)
-\vect v(\hat{\vect r}\cdot\vect r)\\
&amp;=
(r\hat{\vect r})\dot r-r\vect v\\
&amp;=-r\left(\vect v-\dot r\hat{\vect r}\right).
\end{align}
Using Eq.~\eqref{eq:rhatdot},
\begin{equation}
\vect v-\dot r\hat{\vect r}
=r\frac{d\hat{\vect r}}{dt},
\end{equation}
so
\begin{equation}
\boxed{
\hat{\vect r}\times\vect h
=-r^2\frac{d\hat{\vect r}}{dt}.
}
\label{eq:keycross}
\end{equation}
This identity is the algebraic hinge of the eccentricity-vector derivation.

\section{Deriving the conserved vector from Newton's law}

Differentiate $\vect v\times\vect h$:
\begin{align}
\frac{d}{dt}(\vect v\times\vect h)
&amp;=\frac{d\vect v}{dt}\times\vect h
+\vect v\times\frac{d\vect h}{dt}.
\end{align}
Because $\frac{d\vect h}{dt}=0$,
\begin{equation}
\frac{d}{dt}(\vect v\times\vect h)
=\vect a\times\vect h.
\label{eq:vdh}
\end{equation}

Newton's inverse-square acceleration is
\begin{equation}
\vect a=-\frac{\mu}{r^2}\hat{\vect r}.
\end{equation}
Therefore
\begin{align}
\vect a\times\vect h
&amp;=-\frac{\mu}{r^2}
(\hat{\vect r}\times\vect h).
\end{align}
Substitute Eq.~\eqref{eq:keycross}:
\begin{align}
\vect a\times\vect h
&amp;=-\frac{\mu}{r^2}
\left(-r^2\frac{d\hat{\vect r}}{dt}\right)\\
&amp;=\mu\frac{d\hat{\vect r}}{dt}.
\end{align}
Thus
\begin{equation}
\boxed{
\frac{d}{dt}(\vect v\times\vect h)
=\mu\frac{d\hat{\vect r}}{dt}.
}
\label{eq:cancelderivs}
\end{equation}

Now the construction is no longer mysterious. Move the right-hand derivative to the left:
\begin{equation}
\frac{d}{dt}(\vect v\times\vect h)
-
\mu\frac{d\hat{\vect r}}{dt}
=0,
\end{equation}
or
\begin{equation}
\boxed{
\frac{d}{dt}
\left(
\vect v\times\vect h
-\mu\hat{\vect r}
\right)=\vect 0.
}
\label{eq:conservedA}
\end{equation}
Therefore the quantity in parentheses is a conserved vector.

Divide by the constant $\mu$ and define
\begin{equation}
\boxed{
\vect e
=\frac{\vect v\times\vect h}{\mu}
-\hat{\vect r}.
}
\label{eq:evec}
\end{equation}
This is the \textbf{eccentricity vector}.

\begin{center}
\includegraphics[width=0.93\textwidth,keepaspectratio]{EV_conservation_mechanism.png}

{\small Figure 2. The eccentricity vector emerges because Newton's inverse-square law makes the derivatives of $\vect v\times\vect h$ and $\mu\hat{\vect r}$ identical. Their difference is therefore conserved.}
\end{center}

The dimensional check is also revealing. Since
\begin{equation}
[\vect v\times\vect h]
=\frac{L}{T}\frac{L^2}{T}
=\frac{L^3}{T^2},
\end{equation}
and
\begin{equation}
[\mu]=\frac{L^3}{T^2},
\end{equation}
the ratio $(\vect v\times\vect h)/\mu$ is dimensionless, just like $\hat{\vect r}$. Therefore $\vect e$ is dimensionless.

\section{Why does this vector point toward periapsis?}

Conservation alone does not yet prove that $\vect e$ points toward periapsis. That fact comes from projecting Eq.~\eqref{eq:evec} onto the radial direction.

Let $\alpha$ be the instantaneous angle between $\vect e$ and $\hat{\vect r}$. Then
\begin{equation}
\vect e\cdot\hat{\vect r}
=e\cos\alpha.
\label{eq:edotrhatgeom}
\end{equation}
Dot Eq.~\eqref{eq:evec} with $\hat{\vect r}$:
\begin{equation}
 e\cos\alpha
=\frac{(\vect v\times\vect h)\cdot\hat{\vect r}}{\mu}-1.
\label{eq:edotrhat1}
\end{equation}
Use the scalar triple-product identity,
\begin{equation}
(\vect v\times\vect h)\cdot\hat{\vect r}
=\vect h\cdot(\hat{\vect r}\times\vect v).
\end{equation}
But
\begin{equation}
\vect h
=\vect r\times\vect v
=r(\hat{\vect r}\times\vect v),
\end{equation}
so
\begin{equation}
\hat{\vect r}\times\vect v
=\frac{\vect h}{r}.
\end{equation}
Therefore
\begin{equation}
(\vect v\times\vect h)\cdot\hat{\vect r}
=\frac{h^2}{r}.
\end{equation}
Equation~\eqref{eq:edotrhat1} becomes
\begin{equation}
 e\cos\alpha
=\frac{h^2}{\mu r}-1.
\end{equation}
Rearrange:
\begin{equation}
1+e\cos\alpha
=\frac{h^2}{\mu r},
\end{equation}
and solve for $r$:
\begin{equation}
\boxed{
 r
=\frac{h^2/\mu}{1+e\cos\alpha}.
}
\label{eq:orbitpre}
\end{equation}

Define the semilatus rectum
\begin{equation}
\boxed{
 p=\frac{h^2}{\mu}.
}
\label{eq:p}
\end{equation}
Then
\begin{equation}
\boxed{
 r=\frac{p}{1+e\cos\alpha}.
}
\label{eq:conic}
\end{equation}
This is the polar equation of a conic with the central body at the focus.

For $e&gt;0$, $r$ is smallest when the denominator is largest, which occurs at
\begin{equation}
\cos\alpha=1
\qquad\Longrightarrow\qquad
\alpha=0.
\end{equation}
Thus the direction $\alpha=0$, namely the direction of $\vect e$, is the direction of closest approach. That is periapsis.

Therefore
\begin{equation}
\boxed{
\text{$\vect e$ points from the occupied focus toward periapsis.}
}
\end{equation}
Once that fact is established, the angle $\alpha$ is conventionally called the true anomaly $\nu$, so Eq.~\eqref{eq:conic} becomes
\begin{equation}
\boxed{
 r=\frac{p}{1+e\cos\nu}.
}
\label{eq:conicnu}
\end{equation}

\begin{center}
\includegraphics[width=0.72\textwidth,keepaspectratio]{EV_orbit_equation_bridge.png}

{\small Figure 3. Dotting the eccentricity vector with the radial unit vector produces the conic equation and identifies $p=h^2/\mu$ as the semilatus rectum.}
\end{center}

This is one of the most useful results in celestial mechanics: the same vector derived from Newton's differential equation simultaneously encodes the orbit's \emph{shape} and its \emph{orientation within the orbital plane}.

\section{Why is its magnitude the conic eccentricity?}

The symbol $e$ in Eq.~\eqref{eq:conicnu} is exactly the geometric eccentricity parameter of the conic. Since that equation was obtained using
\begin{equation}
 e=\|\vect e\|,
\end{equation}
the magnitude of the eccentricity vector is the scalar conic eccentricity.

Its value classifies the Kepler orbit:
\begin{equation}
\begin{array}{ccl}
0\le e&lt;1 &amp;:&amp; \text{ellipse},\\
e=1 &amp;:&amp; \text{parabola},\\
e&gt;1 &amp;:&amp; \text{hyperbola}.
\end{array}
\end{equation}
A circle is the special elliptical case $e=0$.

\begin{center}
\includegraphics[width=0.92\textwidth,keepaspectratio]{EV_circular_vs_eccentric.png}

{\small Figure 4. For a circular orbit the eccentricity vector vanishes, so no periapsis direction exists. For an eccentric orbit its direction selects periapsis and its magnitude measures the conic eccentricity.}
\end{center}

This circular limit explains an important singularity in classical orbital elements. If
\begin{equation}
\vect e=\vect 0,
\end{equation}
then the argument of periapsis $\omega$ cannot be defined uniquely because there is no distinguished periapsis direction.

\section{Connection with orbital energy}

The eccentricity magnitude can also be related directly to the specific mechanical energy
\begin{equation}
\mathcal E=\frac{v^2}{2}-\frac{\mu}{r}.
\label{eq:energy}
\end{equation}
Square Eq.~\eqref{eq:evec}:
\begin{align}
 e^2
&amp;=\left\|
\frac{\vect v\times\vect h}{\mu}
-\hat{\vect r}
\right\|^2\\
&amp;=\frac{\|\vect v\times\vect h\|^2}{\mu^2}
+1
-\frac{2(\vect v\times\vect h)\cdot\hat{\vect r}}{\mu}.
\end{align}
Because $\vect v$ lies in the orbital plane while $\vect h$ is normal to it,
\begin{equation}
\vect v\cdot\vect h=0,
\end{equation}
so
\begin{equation}
\|\vect v\times\vect h\|^2=v^2h^2.
\end{equation}
We already found
\begin{equation}
(\vect v\times\vect h)\cdot\hat{\vect r}
=\frac{h^2}{r}.
\end{equation}
Hence
\begin{equation}
 e^2
=1+\frac{v^2h^2}{\mu^2}
-\frac{2h^2}{\mu r}.
\end{equation}
Factor $h^2/\mu^2$:
\begin{equation}
 e^2
=1+\frac{h^2}{\mu^2}
\left(v^2-\frac{2\mu}{r}\right).
\end{equation}
Using Eq.~\eqref{eq:energy},
\begin{equation}
 v^2-\frac{2\mu}{r}=2\mathcal E,
\end{equation}
so
\begin{equation}
\boxed{
 e^2=1+\frac{2\mathcal E h^2}{\mu^2}.
}
\label{eq:eenergy}
\end{equation}
This formula connects the two principal invariants of Kepler motion: angular momentum controls the orbital plane and transverse scale, while energy controls whether the conic is bound or unbound. Their combination determines eccentricity.

For an ellipse,
\begin{equation}
\mathcal E=-\frac{\mu}{2a}.
\end{equation}
Substituting into Eq.~\eqref{eq:eenergy},
\begin{equation}
 e^2
=1-\frac{h^2}{\mu a}.
\end{equation}
Therefore
\begin{equation}
\frac{h^2}{\mu}=a(1-e^2),
\end{equation}
and with Eq.~\eqref{eq:p},
\begin{equation}
\boxed{
 p=\frac{h^2}{\mu}=a(1-e^2).
}
\label{eq:pall}
\end{equation}
Thus the semilatus rectum is where Newtonian dynamics, conic geometry, and classical orbital elements meet.

\section{An equivalent Cartesian formula}

Equation~\eqref{eq:evec} is compact, but software often uses an equivalent expression containing only $\vect r$, $\vect v$, and scalar dot products. Start with
\begin{equation}
\vect v\times\vect h
=\vect v\times(\vect r\times\vect v).
\end{equation}
Apply the triple-product identity:
\begin{equation}
\vect v\times(\vect r\times\vect v)
=\vect r(\vect v\cdot\vect v)
-\vect v(\vect v\cdot\vect r).
\end{equation}
Hence
\begin{equation}
\vect v\times\vect h
=v^2\vect r-(\vect r\cdot\vect v)\vect v.
\end{equation}
Since
\begin{equation}
\hat{\vect r}=\frac{\vect r}{r},
\end{equation}
Eq.~\eqref{eq:evec} becomes
\begin{equation}
\boxed{
\vect e
=\frac{1}{\mu}
\left[
\left(v^2-\frac{\mu}{r}\right)\vect r
-(\vect r\cdot\vect v)\vect v
\right].
}
\label{eq:ecart}
\end{equation}
Both forms are mathematically identical:
\begin{equation}
\boxed{
\frac{\vect v\times\vect h}{\mu}-\hat{\vect r}
\equiv
\frac{1}{\mu}
\left[
\left(v^2-\frac{\mu}{r}\right)\vect r
-(\vect r\cdot\vect v)\vect v
\right].
}
\end{equation}
The cross-product form is often better for geometric understanding; the expanded form makes the dependence on radial velocity $\vect r\cdot\vect v=r\dot r$ explicit.

\section{A physical reading of the two terms}

The eccentricity vector can be read as a competition between
\begin{equation}
\frac{\vect v\times\vect h}{\mu}
\qquad\text{and}\qquad
\hat{\vect r}.
\end{equation}
The radial unit vector always has magnitude one and continually rotates as the body moves around the focus. The first term also changes as the velocity direction changes. For an inverse-square force, Eq.~\eqref{eq:cancelderivs} says these two changes are exactly synchronized:
\begin{equation}
\frac{d}{dt}
\left(
\frac{\vect v\times\vect h}{\mu}
\right)
=
\frac{d\hat{\vect r}}{dt}.
\end{equation}
Their difference is therefore frozen in inertial space.

That is the deepest intuition behind Eq.~\eqref{eq:evec}: it is not a lucky algebraic combination. It is a cancellation created by the special $1/r^2$ central force.

If the force law is perturbed, for example by Earth's oblateness, third-body gravity, or atmospheric drag, the osculating eccentricity vector generally changes slowly with time. Its changing direction describes apsidal precession, and its changing magnitude describes changes in orbital shape.

\section{The Laplace--Runge--Lenz vector}

A closely related conserved quantity is
\begin{equation}
\boxed{
\vect A
=\vect v\times\vect h
-\mu\hat{\vect r}.
}
\label{eq:lrl}
\end{equation}
This is commonly called the Laplace--Runge--Lenz vector, or simply the Runge--Lenz vector in many mechanics texts. Comparing Eqs.~\eqref{eq:evec} and \eqref{eq:lrl},
\begin{equation}
\boxed{
\vect A=\mu\vect e.
}
\end{equation}
The two vectors therefore have exactly the same direction. The normalized form $\vect e$ is particularly convenient in astrodynamics because its magnitude is immediately the dimensionless orbital eccentricity.

\section{Worked GPS-like numerical example}

Consider the Earth-centered inertial state
\begin{equation}
\vect r=
\begin{bmatrix}
-13130.4247\\
8498.2557\\
21350.9782
\end{bmatrix}\ {\rm km},
\end{equation}
\begin{equation}
\vect v=
\begin{bmatrix}
-2.70186285\\
-2.74572481\\
-0.52359229
\end{bmatrix}\ {\rm km/s},
\end{equation}
with
\begin{equation}
\mu=398600.4418\ {\rm km^3/s^2}.
\end{equation}

First compute the specific angular momentum,
\begin{equation}
\vect h=\vect r\times\vect v,
\end{equation}
which gives approximately
\begin{equation}
\vect h=
\begin{bmatrix}
54174.2894\\
-64562.4039\\
59013.6542
\end{bmatrix}\ {\rm km^2/s},
\end{equation}
with
\begin{equation}
 h\approx102887.1665\ {\rm km^2/s}.
\end{equation}

The position magnitude is
\begin{equation}
 r\approx26466.8221\ {\rm km},
\end{equation}
so
\begin{equation}
\hat{\vect r}=\frac{\vect r}{r}.
\end{equation}
Now evaluate
\begin{equation}
\vect e
=\frac{\vect v\times\vect h}{\mu}
-\hat{\vect r}.
\end{equation}
The result is
\begin{equation}
\boxed{
\vect e\approx
\begin{bmatrix}
0.00479070\\
0.00776363\\
0.00409576
\end{bmatrix}.
}
\end{equation}
Its magnitude is
\begin{equation}
\boxed{
 e=\|\vect e\|\approx0.0100000.
}
\end{equation}
The orbit is therefore a low-eccentricity ellipse.

The specific mechanical energy is
\begin{equation}
\mathcal E
=\frac{v^2}{2}-\frac{\mu}{r}
\approx-7.5037734\ {\rm km^2/s^2},
\end{equation}
which corresponds to
\begin{equation}
 a=-\frac{\mu}{2\mathcal E}
\approx26560.000\ {\rm km}.
\end{equation}
The semilatus rectum obtained dynamically is
\begin{equation}
 p=\frac{h^2}{\mu}
\approx26557.3439\ {\rm km},
\end{equation}
while the orbital-element expression gives
\begin{equation}
 a(1-e^2)
\approx26557.3439\ {\rm km}.
\end{equation}
The agreement provides a useful implementation check.

The angle between $\vect e$ and $\vect r$ is
\begin{equation}
\nu
=\cos^{-1}
\left(
\frac{\vect e\cdot\vect r}{er}
\right)
\approx70.0000^\circ,
\end{equation}
which recovers the true anomaly used to generate the state.

Finally, the radial speed is
\begin{equation}
\dot r
=\hat{\vect r}\cdot\vect v
\approx0.0364051\ {\rm km/s}.
\end{equation}
The positive sign indicates that the spacecraft is moving outward, away from periapsis toward apoapsis.

\section{Implementation recipe from a Cartesian state}

Given $\vect r$, $\vect v$, and $\mu$, the eccentricity-vector calculation is short:
\begin{align}
 r&amp;=\|\vect r\|,\\
 \vect h&amp;=\vect r\times\vect v,\\
 \vect e&amp;=\frac{\vect v\times\vect h}{\mu}-\frac{\vect r}{r},\\
 e&amp;=\|\vect e\|.
\end{align}
For $e&gt;0$, the unit periapsis direction is
\begin{equation}
\boxed{
\hat{\vect P}=\frac{\vect e}{e}.
}
\end{equation}
This is exactly the $P$ axis of the perifocal $PQW$ frame.

In numerical software, one should treat very small $e$ carefully. As $e\rightarrow0$, the magnitude of $\vect e$ becomes small and its normalized direction becomes sensitive to numerical noise because a circular orbit has no physically unique periapsis.

\section{What to remember}

The eccentricity vector can be understood through a compact chain of ideas:
\begin{equation}
\boxed{
\text{central inverse-square force}
\Longrightarrow
\frac{d\vect h}{dt}=0
\Longrightarrow
\frac{d}{dt}(\vect v\times\vect h)
=\mu\frac{d\hat{\vect r}}{dt}
\Longrightarrow
\vect e=\text{constant}.
}
\end{equation}
Its final form is
\begin{equation}
\boxed{
\vect e
=\frac{\vect v\times\vect h}{\mu}
-\hat{\vect r}.
}
\end{equation}
Its magnitude is
\begin{equation}
\boxed{e=\|\vect e\|,}
\end{equation}
its direction is toward periapsis, and its projection onto $\hat{\vect r}$ immediately generates the orbit equation
\begin{equation}
\boxed{
 r=\frac{p}{1+e\cos\nu},
 \qquad
 p=\frac{h^2}{\mu}.
}
\end{equation}
For an ellipse,
\begin{equation}
\boxed{
 p=\frac{h^2}{\mu}=a(1-e^2),
}
\end{equation}
and energy supplies the equivalent relation
\begin{equation}
\boxed{
 e^2=1+\frac{2\mathcal E h^2}{\mu^2}.
}
\end{equation}

The vector therefore compresses an unusually large amount of orbital information into one conserved quantity: the orbit's eccentricity, periapsis direction, conic equation, and connection between angular momentum and energy.

\section*{References}
\begin{thebibliography}{9}

\bibitem{bate}
R. R. Bate, D. D. Mueller, and J. E. White,
\emph{Fundamentals of Astrodynamics},
Dover Publications, 1971.

\bibitem{curtis}
H. D. Curtis,
\emph{Orbital Mechanics for Engineering Students},
Elsevier, 4th ed., 2020.

\bibitem{vallado}
D. A. Vallado,
\emph{Fundamentals of Astrodynamics and Applications},
Microcosm Press, 4th ed., 2013.

\bibitem{goldstein}
H. Goldstein, C. Poole, and J. Safko,
\emph{Classical Mechanics},
Addison-Wesley, 3rd ed., 2002.

\end{thebibliography}</content>
</record>
