7

To find the area of triangle formed by three straight lines Given is combined equation of three straight lines $$x^2 y -y^3-x^2+5y^2-8y+4=0$$

This can easily be factorized into three straight lines

  • $y=1$
  • $y=x+2$
  • $y=2-x$

Now it will be easy to calculate intersection points and calculate the area as 1 sq units.

However, I was wondering if it were possible to calculate the area from the combined equation of the lines without actually finding the equations of the straight lines by making use of the coeffecients of the terms in the combined equation.

  • 1
    When you find the triangle in the picture, you're adding some information not really present in the equation. Given just the equation, what is the region whose area you're trying to find? If you ask about the region where this expression is $\le 0$, that's an infinite region that just happens to include the triangle. – Misha Lavrov Oct 11 '17 at 04:33
  • @MishaLavrov - that is incorrect. The equations (no picture!) determine the lines, and then the intersection points, and then the area, completely. There is nothing "pictorial" in the whole thing. The question makes perfect sense. –  Oct 11 '17 at 04:34
  • The coordinates of the intersection points determine the area. The equations of the lines determine the coordinates of the intersection points, so the area can (in principle) be written in terms of the coefficients of the line equations. So, the question is - can you express the coefficients of the line equations in terms of the coefficients of the original equation? Obviously you can in the opposite direction; can you solve for the coefficients of the line equations in terms of the coefficients in the original equation, given the "other direction" equations? –  Oct 11 '17 at 04:37
  • I don't mean that the area isn't well-specified, merely that it's a quantity that seems to lack algebraic meaning. If I gave you a generic cubic polynomial in $x$ and $y$, how would you specify which region you wanted to know the area of, in terms of that polynomial alone? – Misha Lavrov Oct 11 '17 at 04:56
  • 1
    You can try it in reverse. Take $A=(a_x,a_y)$, $B=(b_x,b_y)$, $C=(c_x,c_y)$, and express the area ($T$) of the triangle (up to sign) using the standard determinant formula. Then, find the equations for the connecting lines, and expand the cubic, and define, say, $k_{ij}$ for the coefficient of $x^i y^j$. Then "simply" eliminate $a_x$ and friends to leave only polynomial in $T$ and the $k$s. (This is always possible using resultants, although the polynomial is bound to be super-complicated and of high degree.) That there are ten $k$s and only six coordinates complicates matters. (con't) – Blue Oct 11 '17 at 06:30
  • (con't) So ... Maybe try something closer to this particular example, in which the points are $(0,2)$, $(1, 1)$, $(-1,1)$. To generalize a bit, use $A=(0,a)$, $B = (b,d)$, $C = (c,d)$. When you examine the cubic, you'll find that there are lots of ways to recover the coordinates from the coefficients. (If my calculations are correct, you get stuff like $k_{20}/k_{21}= -d$ and $k_{02}/k_{03} = -2a-d$.) The fact that there are more coefficients than coordinates to find means that there are dependencies involved that make it difficult to know exactly which approach to take. But it's do-able. – Blue Oct 11 '17 at 06:40
  • 2
    If I didn't make any mistake, if the cubic has the form $$f(x,y) = y^3 + by^2x + cyx^2 + dx^3 + ey^2 + fyx + gx^2 + hy + ix + j = 0$$ then the area is given by the formula $$\frac{12ch-4b^2h-3f^2+4bef-4ce^2}{2\sqrt{-27d^2+18bcd-4b^3d-4c^3+b^2c^2}}$$ Please note that what's inside the square root in numerator is the discriminant for the cubic polynomial $y^3 + by^2 + cy +d$. – achille hui Oct 11 '17 at 08:45

1 Answers1

8

Before we start, let us recall a formula for the area of a triangle in terms of the lines bounding it.

Let $\hat{z}$ be the unit vector $(0,0,1)$. For any line $\ell : \alpha x + \beta y + \gamma = 0$ on the plane, we will use the same symbol $\ell$ to denote the affine function $\ell(x,y) = \alpha x + \beta y + \gamma$ and $\vec{\ell}$ to denote the vector $(\alpha,\beta,\gamma)$.

Given any three lines $\ell_i : \alpha_i x + \beta_i y + \gamma_i = 0$ for $i = 1,2,3$, the area of the triangle bounded by them is given by the formula: $$\verb/Area/ = \frac{D^2}{2C_1C_2C_3}$$ where

$$ D = \vec{\ell}_1 \cdot (\vec{\ell}_2 \times \vec{\ell}_3) = \left|\begin{matrix} \alpha_1 & \alpha_2 & \alpha_3 \\ \beta_1 & \beta_2 & \beta_3 \\ \gamma_1 & \gamma_2 & \gamma_3 \end{matrix}\right| \quad\text{ and }\quad \begin{cases} C_1 &= |\hat{z} \cdot (\vec{\ell}_2 \times \vec{\ell}_3 )| = \left\|\begin{matrix}\alpha_2 & \alpha_3\\ \beta_2 & \beta_3\end{matrix}\right\|\\ C_2 &= |\hat{z} \cdot (\vec{\ell}_3 \times \vec{\ell}_1 )| = \left\|\begin{matrix}\alpha_3 & \alpha_1\\ \beta_3 & \beta_1\end{matrix}\right\|\\ C_3 &= |\hat{z} \cdot (\vec{\ell}_1 \times \vec{\ell}_2 )| = \left\|\begin{matrix}\alpha_1 & \alpha_2\\ \beta_1 & \beta_2\end{matrix}\right\| \end{cases} $$ For a derivation of this result, please refer to my answer to a related question.

Let's back to our original problem. For simplicity, we will only consider the case where all $\beta_i = 1$. The union of the three lines is the zero set of a cubic polynomial:

$$\begin{align}f(x,y) &= \ell_1(x,y)\ell_2(x,y)\ell_3(x,y)\\ &= y^3 + by^2x + cyx^2 + dx^3 + ey^2 + fyx + gx^2 + hy + ix + j \end{align} $$ Partial differentiate against $y$, we obtain

$$\begin{align}\partial_y f(x,y) &= \ell_1(x,y) \ell_2(x,y) + \ell_2(x,y) \ell_3(x,y) + \ell_3(x,y) \ell_1(x,y)&\tag{*1a}\\ &= 3y^2 + 2bxy + 2ey + cx^2 + fx + h\\ &= \begin{bmatrix}x \\ y \\ 1\end{bmatrix}^T \begin{bmatrix}c & b & \frac{f}{2} \\ b & 3 & e\\ \frac{f}{2} & e & h\end{bmatrix} \begin{bmatrix}x \\ y \\ 1\end{bmatrix} & \tag{*1b} \end{align}$$ Let $\Delta$ be the $3 \times 3$ matrix having $\vec{\ell}_i$ at $i^{th}$ column and $\Lambda$ be the $3 \times 3$ matrix appear in $(*1b)$.
If we rewrite the RHS in $(*1a)$ in matrix form, we obtain

$$2\Lambda = \left(\vec{\ell}_1 \vec{\ell}_2^T + \vec{\ell}_2 \vec{\ell}_1^T\right) + \left(\vec{\ell}_2 \vec{\ell}_3^T + \vec{\ell}_3 \vec{\ell}_2^T\right) + \left(\vec{\ell}_3 \vec{\ell}_1^T + \vec{\ell}_1 \vec{\ell}_3^T\right) = \Delta \begin{bmatrix}0 & 1 & 1 \\ 1 & 0 & 1 \\ 1 & 1 & 0\end{bmatrix} \Delta^T $$ This leads to

$$D^2 = (\det\Delta)^2 = \frac12\det(2\Lambda) = 12ch-4b^2h-3f^2+4bef-4ce^2\tag{*2}$$

Since all $\beta_i = 1$, we have

$$(C_1C_2C_3)^2 = (\alpha_2 - \alpha_3)^2(\alpha_3 - \alpha_1)^2(\alpha_1-\alpha_2)^2$$

Notice $-\alpha_i$ are roots of the cubic polynomial $\lambda^3 + b\lambda^2 + c\lambda + d = 0$. The expression on RHS is nothing but the discriminant of this cubic polynomial. i.e.

$$(C_1C_2C_3)^2 = b^2c^2 - 4c^3 - 4b^3d - 27d^2 + 18bcd\tag{*3}$$

Combine $(*2)$ and $(*3)$, we obtain

$$\verb/Area/ = \frac{12ch-4b^2h-3f^2+4bef-4ce^2}{2\sqrt{b^2c^2 - 4c^3 - 4b^3d - 27d^2 + 18bcd}}$$

As an example, consider the case

$$\begin{align} & x^2 y - y^3 - x^2 + 5y^2 - 8y + 4 = 0\\ \iff & y^3 - x^2 y - 5y^2 + x^2 + 8y - 4 = 0\\ \iff & (c,e,g,h,j) = (-1,-5,1,8,-4), b = d = f = i = 0 \end{align} $$ Since $b = d = f = i = 0$, the area reduces to $$\verb/Area/ \leadsto \frac{12ch-4ce^2}{2\sqrt{- 4c^3}} = \frac{12(-1)(8)-4(-1)(-5)^2}{2\sqrt{-4(-1)^3}} = \frac{4}{2\sqrt{4}} = 1$$ as expected.

achille hui
  • 125,323