Proof without words of the Quadratic Formula?

901 Views Asked by At

As suggested by @Moti and @YvesDaoust in this post, a simple way to identify the roots (red dots) of a parabola (given focus and directrix, blue) by means of straightedge and compass is to draw the circle with center in the focus and radius the distance $\overline{HI}$ between the $x$-axis and the directrix.

enter image description here

Now, consider the Quadratic Formula

$$ \color{red}{x_{\pm}}=\frac{-b\pm\sqrt{b^2-4a \cdot \mathbf{c} }}{2a}. $$

In the above image (a part the roots), it is easy to spot the term $\mathbf{c}$, i.e. the intercept of the parabola with the $y$-axis.

My question is:

How to geometrically illustrate the other various algebraic terms of the quadratic formula by means of this construction, is such a way that the algebraic relation results immediately evident?

With geometrically, I mean some visual intuition, based on such plot (or something similar), in the spirit of a "proof without words".

Thanks for your help!

4

There are 4 best solutions below

3
On BEST ANSWER

Here's a slight re-packaging of notions from my previous answer.


enter image description here

$$|OQ_{\pm}| \;=\; |BB_{-}| \pm |MQ_{+}| \;=\;-\frac{b}{2a} \pm \sqrt{\frac{b^2}{4a^2}-\frac{c}{a}} \;=\; \frac{1}{2a} \left(\;-b \pm \sqrt{b^2-4ac}\;\right)$$


  • The figure represents the scenario in which $a>0$, $b\leq 0$, $c\geq 0$ (and thus that $|OM|\geq |MQ_{\pm}|$). Adjustments to accommodate various sign changes should be clear.

  • Figure labels and calculations incorporate the fact that the latus rectum has length $1/a$.

  • That $\overline{OM} \cong \overline{BB_{-}}$ is my previous answer's Property 2. That these segments' common signed length is $-b/(2a)$ follows, as before, from the equation of the represented parabola, by calculating the difference in $y$-coordinates for points with $x$-coordinates $\pm 1/(4a)$.

  • The relation between the highlighted areas follows immediately from my previous answer's Property 1. Stripping away the trappings of the specific problem, we can state the area property as a general principle that @Andrea should appreciate:

Property 1a. If squares are erected upon axis-perpendicular semi-chords of a parabola, then the difference in their areas is the area of the rectangle bounded by those chords and the extremities of the parabola's latus rectum.

enter image description here


By the way, here's a proof-without-words for my previous question's Property 2, using Property 1a above.

enter image description here

3
On

As the Wikipedia article Power of a point indicates, the tangent to the circle from the origin distance squared is the product of the two roots, but this is just $\,c/a.$

4
On

This solution isn't nearly as self-evident as I like my illustrations to be, but there are some interesting ideas here.

I'll preface this by noting, in something of an echo of @Rahul's comment, that geometricizing $y=ax^2+bx+c$ is a little tricky, in that $a$, $b$, $c$ are dimensionally distinct. In the approach described below, we take $x$ and $y$ (and thus also the roots of the quadratic equation) to be represented by ($1$-dimensional) lengths; necessarily, we see that $c$ must also be $1$-dimensional, $b$ must be $0$-dimensional (a ratio), and $a$ must be ... $(-1)$-dimensional!


Suppose that the graph of $y=ax^2+bx+c$ represents an upward-facing parabola with vertex $V= (h,-k)$; that is, we take $a$ positive and $b$ non-positive. Let $f$ be the vertex-to-focus distance, $f := |VF|$. Let the parabola cross the $y$-axis at $C$, of distance $c\geq0$ from the origin (although it's less problematic here to allow $c<0$), and let the parabola cross the $x$-axis at $R_{\pm}$, at distances $h\pm s$ from the origin.

Some auxiliary points: Let the $x$-axis and parabola axis meet at $M$ (the midpoint of $R_{+}$ and $R_{-}$). Let the horizontal line through $V$ meet the $x$-axis at $k$, and let $S$ be the projection of $R_{+}$ onto that line (so $|VK|=h$, $|VS|=s$, and $|OK|=|SR_{+}|=k$). Also, let the lines $x=\pm f$ meet the parabola at $B_{\pm}$, and let $B$ complete the right triangle with hypotenuse joining those points. Points $A$ and $G$ are on the $y$-axis and parabola axis such that $|KA|=|VG|=4f$.

Given the above, the below happens to be an illustration of the Quadratic Formula:

enter image description here

As I mentioned: not nearly as self-evident as I like. The illustration relies on two interesting properties of parabolas that derive from the reflection property; I'll prove them later.

Property 1. If $P$ is a point on the ("vertical") parabola, then its horizontal displacement from the vertex is the geometric mean of $4f$ and its vertical displacement from the vertex.

The illustration includes two instances of this property in the form of a classical right-triangle construction of the geometric mean.

$$\begin{align} \triangle AVC: &\quad \frac{|KV|}{|KA|} = \frac{|KC|}{|KV|} \quad\to\quad |KV|^2=|KA||KC|\quad\to\quad h^2=4f(c+k) \tag{1} \\[6pt] \triangle GSM: &\quad \frac{|VS|}{|VG|} = \frac{|VM|}{|VS|} \quad\to\quad |VS|^2=|VG||VM| \quad\to\quad s^2=4fk \tag{2} \end{align}$$

From these, we may conclude $s^2 = h^2 - 4fc$, so that the $x$-coordinates of $R_{\pm}$ ---that is, the roots of the quadratic polynomial--- have the form $$h\;\pm\;\sqrt{h^2-4fc} \tag{3}$$

(As an aside: Let the circumcircle $\bigcirc R_{+} R_{-} C$ meet the $y$-axis again at, say, $D$. Then the power of a point theorems, applied to the origin with respect to this circle, imply $$|OR_{+}||OR_{-}| = (h+s)(h-s) = c\cdot 4f = |OC||OD|$$ If we could show independently that $|OD| = 4f$, then we could reason conversely to get $(3)$ without the separate geometric means. I don't see an obvious way to make that association, however ... although little about this approach is obvious.)

Now, $(3)$ bears a bit of a resemblance to the Quadratic Formula. To get it closer, we invoke another property:

Property 2. If $P$, and distinct points $Q_{+}$ and $Q_{-}$, are on a ("vertical") parabola, such that the horizontal displacement from $P$ to each $Q$ is $f$, then the vertical displacement between the $Q$s is the distance from $P$ to the axis of the parabola.

In the figure above, $C$ plays the role of $P$, and $B_\pm$ the roles of $Q_{\pm}$. Since our parabola represents $y=ax^2+bx+c$, we have that $B_{\pm}$ is at (signed) distance $af^2\pm bf+c$ from the $x$-axis; thus, the vertical displacement between them is simply the difference of these expressions. By Property 2, we can write $$h = \left(\;af^2-bf+c\;\right) - \left(\;af^2+bf+c\;\right) = -2bf \tag{4}$$ (Recall that $b$ is non-negative here.) Therefore, $(3)$ becomes $$-2bf\;\pm\;\sqrt{4b^2f^2-4cf} \tag{5}$$ which we can write as $$2f\left(\;-b \pm \sqrt{b^2-\frac{c}{f}}\;\right) \tag{6}$$ In light of the "known" observation that $a = \dfrac{1}{4f}$ (there's that $(-1)$-dimensionality we needed!), we see $$\frac{1}{2a}\left(\;-b\pm\sqrt{b^2-4ac}\;\right) \tag{7}$$ so that we do, in fact, have the Quadratic Formula. $\square$

I'm a little disappointed in the algebraic manipulations required in this demonstration. Perhaps a second pass at the argument, drawing from some more sophisticated geometric properties of parabolas, will streamline things.


Here are proofs of the Properties ...

Property 1.

enter image description here

Here, $\overline{DW}$ is the directrix of the parabola, so that $\triangle PFD$ is isosceles. The reflection property of parabolas implies that the tangent at $P$ bisects the angle at $P$; it therefore also bisects base $\overline{FD}$ at a point $M$ that, by a simple similarity argument, also serves as the midpoint of $\overline{BV}$. From similar subtriangles within $\triangle PMD$, we have $$\frac{|BM|}{|BD|}=\frac{|BP|}{|BM|} \quad\to\quad \left(\frac12 q\right)^2=fp \quad\to\quad q^2 = 4f\cdot p$$ giving the result. $\square$

Property 2.

enter image description here

Again, $\overline{DW}$ is the directrix. This time, we use the reflection property relative to $P$ to conclude that the tangent at $P$ is perpendicular to $\overline{FD}$. It is "known" that chord $\overline{Q_{+}Q_{-}}$ is parallel to that tangent. With a little angle chasing, we find that we may conclude $\triangle Q_{+}QQ_{-}\cong \triangle FWD$, and the property follows. $\square$

0
On

The coefficients $a,b,c$ of the quadratic equation $ax^2+bx^2+c=0$ are not very geometric, so let's work with some slightly different variables that do have a geometric meaning: \begin{align} \alpha &= -\frac b{2a}, & \beta &= -\frac cb, & \gamma &= c. \end{align} In reverse order, $C=(0,\gamma)$ is the $y$-intercept of the parabola, $B=(\beta,0)$ is the point where the tangent through $C$ meets the $x$-axis, and $A=(\alpha,0)$ is the point on the $x$-axis with the same $x$-coordinate as the parabola's focus. The parabola is specified via $\alpha,\beta,\gamma$, and we need to find the points $P$ and $Q$ where it crosses the $x$-axis.

enter image description here

Blue: given data, gray: constructed, green: equal quantities, red: desired roots

Denote the focus by $F$ and the intersection of the directrix and the $y$-axis by $D$.

  1. Construct the line $CF$ using the property of the parabola that the tangent $CB$ bisects $\angle OCF$. Obtain $F$ as the intersection of $CF$ and the vertical through $A$.

  2. Obtain $D$ using the fact that $C$ is equidistant from $F$ and $D$. The directrix is the horizontal through $D$, and is at distance $|OD|$ from the $x$-axis.

  3. Obtain $P$ and $Q$ as the points on the $x$-axis at distance $|OD|$ from $F$.

$P$ and $Q$ are equidistant from $F$ and the directrix, and so lie on the parabola.


To derive the quadratic formula from this, we take an additional step, which may or may not be acceptable from a pure Euclidean-geometry perspective: We note that moving $C$ along the $y$-axis doesn't change the location of the roots, since it just scales the parabola vertically about the $x$-axis. Therefore, we may choose $C$ freely to simplify the construction.

In particular, let us take $C=(0,\beta)$. Then $\angle OCB=45^\circ$, so the line $CF$ is horizontal, and $F=(\alpha,\beta)$. Now $|CD|=|CF|=\alpha$, so $|OD|=\alpha-\beta$. The right triangle $\triangle AFP$ has hypotenuse $|FP|=|OD|=\alpha-\beta$ and vertical side $|AF|=\beta$, so the horizontal side is $|AP|=\sqrt{(\alpha-\beta)^2-\beta^2}=\sqrt{\alpha^2-2\alpha\beta}$; the same is true for $|AQ|$. Therefore, \begin{align} \{|OP|,|OQ|\} &= |OA| \pm |AP| \\ &= \alpha \pm \sqrt{\alpha^2-2\alpha\beta}. \end{align}

enter image description here

Plug in the values of $\alpha$ and $\beta$ from above, and you obtain the quadratic formula.