2
$\begingroup$

Show that the orthocentre of triangle formed by the striaght lines $$ax^2 + 2hxy + by^2 = 0$$ and $lx + my = 1$ is point $(x^\prime, y^\prime)$ such that $${x^\prime \over l} = {y^\prime \over m} = {a+ b \over am^2 - 2hlm + bl^2} \tag{+}$$


Let the two lines be $y - \alpha x = 0, y - \beta x = 0$

enter image description here

The altitude from top vertex is $mx - ly = 0 \qquad \qquad (1)$

Now we need to compute the intersection of $y - \alpha x = 0$ and $lx + my - 1 = 0$ which is $\displaystyle \left( {1\over l + m\alpha}, {\alpha \over l + m \alpha} \right)$.

From this, the equation of altitude from intesection of lines $y - \alpha x = 0$ and $lx + my - 1 =0$ can be computed.

I did the labour and got $$\beta \gamma y + \gamma x - (\alpha\beta + 1) = 0\tag 2$$ where $\gamma = l + m \alpha$

The intersection (1) and (2) is the orthocentre which I get as $$(x^\prime, y^\prime) = \left( {m \over \gamma}\left[ {\alpha \beta + 1 \over m + l\beta}\right], { l \over \gamma } \left[ {\alpha \beta + 1 \over m + l\beta} \right]\right) \tag{3}$$

Now from here computing $\displaystyle {x^\prime\over l}$ and $\displaystyle {y^\prime \over m}$

$${y^\prime \over m} ={l\over m} \left({1+ \alpha \beta \over lm + l^2\beta + m^2\alpha + ml\alpha\beta}\right)\tag{4}$$

$${x^\prime \over l} = {m\over l} \left({1+ \alpha \beta \over lm + l^2\beta + m^2\alpha + ml\alpha\beta}\right)\tag{5}$$


We know that $b(y - \alpha)(y - x\beta) := ax^2 + by^2 + 2hxy = 0$

$$\therefore b(y - \alpha)(y - x\beta) = by^2 + byx(-\beta - \alpha) + b\alpha\beta x^2 = 0$$

We get the values of $a,b,h$ in terms of $\alpha, \beta$ from this equation and plugging those in (+) we get $$1+ \alpha \beta \over \alpha\beta m^2 + (\alpha + \beta)(ml) + l^2 \tag{6}$$

From (4), (5) and (6) I get
$${x^\prime \over l} \ne {y^\prime \over m} \ne {a+ b \over am^2 - 2hlm + bl^2}$$

I did this question same way at least twice but still get the same result.


  • What went wrong in my answer ?
  • 0
    See here: http://www.acadblock.com/co-ordinate-geometry/orthocentre-dgjd/2017-02-21
  • 0
    @Rohan Doing as told in the first comment will raise my medical bills and I did the same thing as told in the second comment but did not get the results $\ddot\frown$. Please help me this problem is really frustrating.2017-02-21
  • 2
    Shouldn't (1) be mx - ly = 0?2017-02-21
  • 0
    @Mick Yes thanks.2017-02-21
  • 0
    You’re again neglecting to cover the case that one of the lines is vertical.2017-02-22
  • 0
    Proving the first equality should be straightforward. The origin is a vertex of the triangle, so by definition the orthocenter lies somewhere on the perpendicular through the origin to $lx+my=1$, I.e., $(x',y')=c(l,m)$ for some $c$, and so $x'/l=y'/m=c$. I’d go back and try to figure out why those first two ratios aren’t equal in your proposed solution.2017-02-22
  • 0
    @amd I did not consider slope this time. Still why I missing vertical lines ?2017-02-22
  • 0
    The line $x=0$ can’t be written in the form $y-\alpha x=0$.2017-02-22
  • 0
    Yes you are correct but this time I took the general equation of line $y = mx + c$, since it passes through origin $c = 0$.2017-02-22
  • 0
    You *are* using slopes. That form of equation doesn’t include vertical lines—what would $m$ be for a vertical line? As I already mentioned, you can’t write the equation $x=0$ in that form. Similarly, what values of $\alpha$ and $\beta$ yield $xy=0$ (which is the equation of the coordinate axes) from $b(y-\alpha x)(y-\beta x)=0$? The most inclusive form of equation for a line through the origin is $ax+by=0$.2017-02-22
  • 0
    @amd Yes I would love to use $ax + by = 0$ but that will make the problem very messy.2017-02-22
  • 0
    I think I can do something like $$b(y-\alpha x)(y-\beta x)=0 \implies b(y-\alpha x)(y/\beta - x) = 0$$ If I take $\alpha =0$ then $xy =0$ can be represented as $$\lim_{\beta \to \infty}b(y)(y/\beta - x) = 0$$. Am I correct ?2017-02-22
  • 0
    I suppose you could do it that way, if you could assign a reasonable meaning to this limit, but why? You then have to show that all of the algebraic manipulations that you did are also valid in the limit. Seems more straightforward to either start with the general case in the first place or handle vertical lines as a special case.2017-02-23

2 Answers 2

1

This is a fairly straightforward, though mildly tedious, calculation if you use homogeneous coordinates. Doing so lets you represent lines in $\mathbb R^2$ as cross products of points and by duality compute their intersection as a cross product. (The vectors that represent lines transform differently than the ones that represent points, but that’s not going to come into play here). You can use the model of the projective plane as the plane $z=1$ in $\mathbb R^3$ to understand how this works: the vector that represents a line is just the normal to the plane defined by the line and the origin.

It turns out that the homogeneous vector that represents the line with equation $ax+by=c$ is simply $a:b:-c$. So, taking two generic lines through the origin, we have for our three lines $$\begin{align}L_0&=l:m:-1\\L_1&=\alpha:\beta:0\\L_2&=\gamma:\delta:0.\end{align}$$ We want $L_1$ and $L_2$ to be distinct, so $\alpha\delta-\beta\gamma\ne0$. For our two points of intersection we then have $$\begin{align}P_1&=L_0\times L_1=-\beta:\alpha:m\alpha-l\beta\\P_2&=L_0\times L_2=-\delta:\gamma:m\gamma-l\delta.\end{align}$$ The point at infinity that is on a line is also its direction vector, so the two altitude lines through these points are $P_1\times L_2$ and $P_2\times L_1$, and thus the orthocenter is $$\begin{align} H&=(P_1\times L_2)\times(P_2\times L_1) \\ &=([-\beta,\alpha,m\alpha-l\beta]\times[\gamma,\delta,0])\times([-\delta,\gamma,m\gamma-l\delta]\times[\alpha,\beta,0]) \\ &=[(m\alpha-l\beta)\delta,-(m\alpha-l\beta)\gamma,\alpha\gamma-\beta\delta]\times[(m\gamma-l\delta)\beta,-(m\gamma-l\delta)\alpha,\alpha\gamma-\beta\delta] \\ &=[l(\alpha\gamma+\beta\delta)(\beta\gamma-\alpha\delta),m(\alpha\gamma+\beta\delta)(\beta\gamma-\alpha\delta),(m\alpha-l\beta)(m\gamma-l\delta)(\beta\gamma-\alpha\delta)] \\ &=l(\alpha\gamma+\beta\delta):m(\alpha\gamma+\beta\delta):(m\alpha-l\beta)(m\gamma-l\delta)\end{align}$$ from which we can read $$\frac{x'}l=\frac{y'}m={\alpha\gamma+\beta\delta\over(m\alpha-l\beta)(m\gamma-l\delta)},$$ which looks pretty close to the desired result. Setting $(\alpha x+\beta y)(\gamma x+\delta y)=ax^2+2hxy+by^2$, we see that the denominator is simply this expression evaluated at $x=m$, $y=-l$, while the numerator is $a+b$, so $$\frac{x'}l=\frac{y'}m={a+b\over am^2-2hlm+bl^2}.$$

Postscript: Although the above computation produces the required result, I find it somewhat unsatisfactory because the last ratio seems to pop up out of nowhere. The left-hand equality has an obvious geometric interpretation: it says that the orthocenter lies somewhere on the line $\lambda(l,m)$, which is the normal to $L_0$ and so also the altitude line through the origin. The geometric meaning of the other equalities escapes me, although I can see that the right-hand fraction is a ratio of dot products of various relevant normals. The trilinear coordinates of a triangle’s orthocenter are $\sec A:\sec B:\sec C$ so perhaps there’s some connection to those ratios.

  • 0
    P.P.S.: Writing $H$ in terms of the three lines has a pleasing and suggestive symmetry: $((L_0\times L_1)\times L_2)\times((L_0\times L_2)\times L_1)$.2017-02-23
  • 0
    I have always wondered what is homogeneous coordinates ? I looked it up on wikipedia but only thing I was able to understand was that this type of coordinates produces more symmetric formulae than cartesian coordinates.2017-02-23
  • 0
    @A---B When you use homogeneous coordinates for $\mathbb R^n$, you effectively move to the projective space $\mathbb{RP}^n$. This can have several advantages. For instance, in $\mathbb{RP}^2$, parallel lines then intersect at a point at infinity, so you don’t need to treat them as special cases. Other benefits of working in a projective space: duality is a very powerful tool, a perspective projection is just the identity map on a certain subspace, and so on.2017-02-23
  • 0
    @A---B In $\mathbb{RP}^2$, the duality principle says that if you swap the terms “line” and “point” in any definition or theorem, the resulting definition is still significant and the resulting theorem is also valid.2017-02-23
  • 0
    The duality principle seems like the transfer principle from non-standard analysis. Can you tell when will I learn all this ?2017-02-23
1

I calculated with line equation of altitude from $(0,0)$ incorrectly. Doing everything from scratch gives correct results.

Correct equation is $mx - ly = 0$ which gives first eqaulity with $\displaystyle {y\over m} ={x \over l}$.

  • 1
    Good catch!. Amazing how an early sign error can throw things way off, isn’t it?2017-02-23
  • 0
    I did this question twice before your comment under the question made me realise my mistake.2017-02-23
  • 0
    I believe it was Mick’s comment, not mine. I had spotted that error, but didn’t bother to comment since it was already there.2017-02-23
  • 0
    @amd I saw Mick's comment but error did not struck me but this comment http://math.stackexchange.com/questions/2154601/a-problem-relating-orthcentre-and-pair-of-equation/2155749?noredirect=1#comment4433852_2154601 made me realize the error.2017-02-23