Often, hyperbolic functions are defined in terms of exponential functions. Here, these definitions will be derived through basic differential equations.
To begin these derivations, a relation between the hyperbolic angle and coordinates of an arbitrary point, $(x_1, y_1) = (\cosh{a}, \sinh{a})$, on the unit hyperbola must be found. Consider a right triangle formed by the vertices $(0,0)$, $(x_1, y_1)$, and $(x_1, 0)$, having an area of $x_1y_1/2$. Notice that this triangle's area is also given by the subtraction of $a/2$ from $\int_0^{y_1} x dy$. Therefore, the following relation holds between $x_1$, $y_1$, and $a$:
$$\frac{1}{2}x_1y_1 = \int_0^{y_1} x dy - \frac{1}{2}a$$
Using the equation of the unit hyperbola, $a$ can be solved for in relation to $y_1$.\begin{equation}
a = \begin{cases} 2\int_0^{y_1} \sqrt{1+y^2} dy - y_1\sqrt{1+{y_1}^2}, & x_1 \geq 1 \\ -2\int_0^{y_1} \sqrt{1+y^2} dy + y_1\sqrt{1+{y_1}^2}, & x_1 \leq -1\end{cases} \end{equation}
a = \begin{cases} 2\int_0^{y_1} \sqrt{1+y^2} dy - y_1\sqrt{1+{y_1}^2}, & x_1 \geq 1 \\ -2\int_0^{y_1} \sqrt{1+y^2} dy + y_1\sqrt{1+{y_1}^2}, & x_1 \leq -1\end{cases} \end{equation}
Hyperbolic Sine
To derive the formula for the hyperbolic sine function, $\sinh{a} = y_1$, equation $(1)$ must be solved for $y_1$ in relation to $a$. Unfortunately, integrating the integrands of equation $(1)$ proves to be extremely difficult, so differentiation is used here instead for an easier derivation.\begin{equation}
\frac{da}{dy_1} = \begin{cases} \frac{2 \left(1+{y_1}^2\right) - \left(1+{y_1}^2\right) - {y_1}^2}{\sqrt{1+{y_1}^2}} , & x_1 \geq 1 \\ \frac{-2 \left(1+{y_1}^2\right) + \left(1+{y_1}^2\right) + {y_1}^2}{\sqrt{1+{y_1}^2}} , & x_1 \leq -1\end{cases}
\end{equation}
\begin{equation}
\frac{da}{dy_1} = \begin{cases} \frac{1}{\sqrt{1+{y_1}^2}} , & x_1 \geq 1 \\ -\frac{1}{\sqrt{1+{y_1}^2}} , & x_1 \leq -1\end{cases}
\end{equation}
\begin{equation}
\left(\frac{dy_1}{da}\right)^2 = 1+{y_1}^2
\end{equation}
Differentiating implicitly with respect to $a$ gives:\frac{da}{dy_1} = \begin{cases} \frac{2 \left(1+{y_1}^2\right) - \left(1+{y_1}^2\right) - {y_1}^2}{\sqrt{1+{y_1}^2}} , & x_1 \geq 1 \\ \frac{-2 \left(1+{y_1}^2\right) + \left(1+{y_1}^2\right) + {y_1}^2}{\sqrt{1+{y_1}^2}} , & x_1 \leq -1\end{cases}
\end{equation}
\begin{equation}
\frac{da}{dy_1} = \begin{cases} \frac{1}{\sqrt{1+{y_1}^2}} , & x_1 \geq 1 \\ -\frac{1}{\sqrt{1+{y_1}^2}} , & x_1 \leq -1\end{cases}
\end{equation}
\begin{equation}
\left(\frac{dy_1}{da}\right)^2 = 1+{y_1}^2
\end{equation}
\begin{equation}
2\frac{dy_1}{da}\frac{d^2y_1}{{da}^2} = 2y_1\frac{dy_1}{da}
\end{equation}
\begin{equation}
\frac{d^2y_1}{{da}^2} = y_1
\end{equation}
The above equation is a homogeneous second-order linear differential equation and has the general solution:2\frac{dy_1}{da}\frac{d^2y_1}{{da}^2} = 2y_1\frac{dy_1}{da}
\end{equation}
\begin{equation}
\frac{d^2y_1}{{da}^2} = y_1
\end{equation}
\begin{equation}
y_1 = Ae^{r_Aa} + Be^{r_Ba}
\end{equation}
where $A$ and $B$ are constants while $r_A$ and $r_B$ are the two solutions to the equation $r^2=1$.y_1 = Ae^{r_Aa} + Be^{r_Ba}
\end{equation}
\begin{equation}
y_1 = Ae^{a} + Be^{-a}
\end{equation}
To solve for $A$ and $B$, apply equation $(4)$ and the initial condition stating that $y_1=0$ when $a=0$.y_1 = Ae^{a} + Be^{-a}
\end{equation}
\begin{align}
A^2e^{2a} - 2AB + B^2e^{-2a} &= 1 + A^2e^{2a} + 2AB + B^2e^{-2a} \\
0 &= A + B
\end{align}
\begin{align}
A &= \pm \frac{1}{2} \\
B &= \mp \frac{1}{2}
\end{align}
By convention, hyperbolic functions parameterize only the right side of the unit hyperbola, for $x\geq1$, meaning $y_1$ and $a$ should share the same sign. Therefore, A^2e^{2a} - 2AB + B^2e^{-2a} &= 1 + A^2e^{2a} + 2AB + B^2e^{-2a} \\
0 &= A + B
\end{align}
\begin{align}
A &= \pm \frac{1}{2} \\
B &= \mp \frac{1}{2}
\end{align}
\begin{align}
A &= \frac{1}{2} \\
B &= - \frac{1}{2}
\end{align}
\begin{equation}
y_1 = \frac{1}{2}\left(e^{a} - e^{-a}\right)
\end{equation}
Since $\sinh{a}=y_1$, the hyperbolic sine function, written in terms of exponential functions, is thus derived to be:A &= \frac{1}{2} \\
B &= - \frac{1}{2}
\end{align}
\begin{equation}
y_1 = \frac{1}{2}\left(e^{a} - e^{-a}\right)
\end{equation}
\begin{equation}
\sinh{a} = \frac{1}{2}\left(e^{a} - e^{-a}\right)
\end{equation}
\sinh{a} = \frac{1}{2}\left(e^{a} - e^{-a}\right)
\end{equation}
Hyperbolic Cosine
Although the hyperbolic cosine function can also be derived in a similar manner through differential equations, a much easier derivation is possible using the already found hyperbolic sine function. Since $\sinh{a}=y_1$ and $\cosh{a} = x_1$, the equation of the unit hyperbola can be used to relate $\cosh{a}$ to the exponential expression that gives $\sinh{a}$.\begin{equation}
\cosh^2{a} - 1 = \frac{1}{4}\left(e^{a} - e^{-a}\right)^2
\end{equation}
\begin{equation}
\cosh^2{a} = \frac{1}{4}e^{2a} + \frac{1}{2} + \frac{1}{4}e^{-2a}
\end{equation}
\begin{equation}
\cosh{a} = \pm\frac{1}{2}\left( e^{a} + e^{-a}\right)
\end{equation}
Once again, hyperbolic functions, by convention, parameterize the unit hyperbola only for $x\geq1$, meaning $\cosh{a} = x_1$ should always be greater than or equal to $1$. \cosh^2{a} - 1 = \frac{1}{4}\left(e^{a} - e^{-a}\right)^2
\end{equation}
\begin{equation}
\cosh^2{a} = \frac{1}{4}e^{2a} + \frac{1}{2} + \frac{1}{4}e^{-2a}
\end{equation}
\begin{equation}
\cosh{a} = \pm\frac{1}{2}\left( e^{a} + e^{-a}\right)
\end{equation}
\begin{equation}
\cosh{a} = \frac{1}{2}\left( e^{a} + e^{-a}\right)
\end{equation}
\cosh{a} = \frac{1}{2}\left( e^{a} + e^{-a}\right)
\end{equation}
Hyperbolic Tangent
The hyperbolic tangent function, analogous to the trigonometric tangent function, is equal to $y_1/x_1 = \sinh{a}/\cosh{a}$. Therefore, the hyperbolic tangent function is found to be:\begin{equation}
\tanh{a} = \frac{e^{a} - e^{-a}}{e^{a} + e^{-a}}
\end{equation}
or, written in another form,\tanh{a} = \frac{e^{a} - e^{-a}}{e^{a} + e^{-a}}
\end{equation}
\begin{equation}
\tanh{a} = \frac{e^{2a} - 1}{e^{2a} + 1}
\end{equation}
\tanh{a} = \frac{e^{2a} - 1}{e^{2a} + 1}
\end{equation}
Reciprocal Hyperbolic Functions
The hyperbolic cosecant, secant, and cotangent functions are simply the reciprocals of the hyperbolic sine, cosine, and tangent functions respectively and are respectively thus found to be:\begin{align}
\,\text{csch}\,{a} &= \frac{2}{e^{a} - e^{-a}} \\
\,\text{sech}\,{a} &= \frac{2}{e^{a} + e^{-a}} \\
\,\coth{a} &= \frac{e^{a} + e^{-a}}{e^{a} - e^{-a}} \\
&= \frac{e^{2a} + 1}{e^{2a} - 1}
\end{align}
\,\text{csch}\,{a} &= \frac{2}{e^{a} - e^{-a}} \\
\,\text{sech}\,{a} &= \frac{2}{e^{a} + e^{-a}} \\
\,\coth{a} &= \frac{e^{a} + e^{-a}}{e^{a} - e^{-a}} \\
&= \frac{e^{2a} + 1}{e^{2a} - 1}
\end{align}
You say
ReplyDelete>a major difference lies in the arguments these two types of functions take. While trigonometric functions take the central angle of the unit circle as their argument, hyperbolic functions instead take the hyperbolic angle. The hyperbolic angle, a, is double the area enclosed by x2−y2=1, the x-axis, and a line segment between the origin and a point, (x1,y1)=(cosha,sinha), on the unit hyperbola.
But how is that a difference? Isn't the central circular angle also twice the area of the enclosed sector of the unit circle?
You make an interesting point. I had simply meant that the hyperbolic angle to a point on the unit hyperbola is not just a usual angle from the $x$-axis to the line segment connecting that point to the origin, as one might naively assume by direct analogy to the argument taken by trigonometric functions. I have updated the appropriate section of the article. Thank you for your observation.
ReplyDeleteI prefer taking the opposite approach. Instead of wrestling with areas, ignore them. By the way, the area which is identified as half the hyperbolic angle, is an artifact of which hyperbola to use as a reference. Specifically, I refer to the unit hyperbola, xy = 1. You will note that the semi-major axis of this hyperbola is √2 in magnitude. This affects both axes, so the area corresponding to the integral you defined is √2² = 2 times greater than the integral. But this is exactly the hyperbolic angle.
ReplyDeleteHowever, I want to bring to your attention an alternative treatment. About the semi-major axis. Physicists have an exotic name for it. They call it a relativistic invariant. You know they call the Lorentz Transform a hyperbolic rotation? It is explained as sliding a point on the hyperbola from one location to another on the same hyperbola. And of course, every point on a given hyperbola satisfies the same equation, x²-y²=r², where r is the semi-major axis. If a pair of coordinates represents a Lorentz transformable property, physics says they have a relativistic invariant, x²-y²=r². See any difference? I didn't think so. The reason it is invariant is that the semi-major axis of any hyperbola is the shortest distance to the origin. Regardless of where the test point is on the hyperbola, the semi-major axis is always on the axis of symmetry. But the point does have coordinates that vary, (r cosh(w), r sinh(w)). These are the Cartesian projections of the position of the point. Implied is that the slope of the ray to the hyperbola is tanh(w). Although w is a hyperbolic angle, every hyperbolic angle has a unique gudermannian, φ, such that tanh(w) = sin(φ) = v/c. In any case, the atanh(v/c) = w, the hyperbolic angle (or rapidity when discussing velocity). No annoying factors of 2.
The reason I have belabored the case for the hyperbola is the inverse case can be treated identically. In fact, every point on the unit semicircle maps to a unique point on the unit hyperbola. The arc from the tip of the Real unit IS the same arc, point-by-point, as the circular arc. The first has negative curvature, the second has positive curvature. We can treat it the same way. The semi-major axis is the same, either way. But instead of picturing it as a radius vector that sweeps a central angle, it is an arc that reaches out from the tip of the semi-major axis to some destination point on the circumference. If the Real coordinate is r, the normal coordinate is irφ. So as the Imaginary component grows, the Real projection shrinks, even as the semi-major axis remains invariant. The Cartesian projections of this position are (r cos(φ), r sin(φ)). This can also be written in terms of the hyperbolic angle, w, as (r sech(w), r tanh(w)). If we compare these coordinates to the coordinates of the corresponding point on the hyperbola, (r cosh(w), r sinh(w)), If we assign γ = cosh(w), this can be written in terms of the point on the semicircle: (γ² r sech(w), γ tanh(w)). In any case, the slope of the ray to the point on the circumference is tan(φ) = sinh(w) = βγ. The ray to the hyperbola is tied to a ray to the circumference by the gudermannian function, so as the hyperbolic ray approaches a slope of magnitude 1, the slope of the circular ray approaches infinity. This is interesting, as the circular angle asymptotically approaches π/2 as the hyperbolic angle approaches infinity. In sum, under these definitions, the area is described by the space between the radius vector to the origin and the fixed unit and the arc. This applies to both extremes. In fact it also applies at the geometric mean of the two x coordinates, x = r. Both rays intersect this vertical axis. To reach the axis, the coordinates of the circular point must be scaled by γ, (r, r tan(φ)). The intersection point of the hyperbolic ray is scaled back by 1/γ, (r, r tanh(w)). Both of these points are on the tangent line, which has zero curvature.