2

Problem statement:

Let $\triangle ABC$ be a right triangle with $B=\frac{\pi}{2}$ and a unit square sitting inside the triangle as shown in the image below. Find the length of the side $AB$

The only solution I could think of was:

By the Pythagorean theorem $$AB^2=AC^2-BC^2\iff(y+1)^2+(x+1)^2=4^2$$ Notice that by the slope of the hypotenuse we get $\frac1x=\frac y1\iff xy=1$. The problem becomes equivalent to solving the quartic : $$x^4+2x^3-14x^2+2x+1=0$$

My question: Is there any other way to solve this that doesn't involve solving a fourth degree equation?

Thanks to anyone who took the time to read.

Antony Theo.
  • 4,716

5 Answers5

10

You have: $(y+1)^2 + (x+1)^2 = 4^2$

Expand out:

$y^2 + x^2 + 2y + 2x + 2 = 4^2$

Since you have $xy=1$, you can set $2=2xy$:

$y^2 + x^2 + 2y + 2x + 2xy = 4^2$ $\implies (y + x)^2 + 2(x+y) = 4^2$

Let $k = (y+x)$; solve the quadratic in $k$ (taking the positive solution): $k = \sqrt{17} - 1$

Hence $x+y = (\sqrt{17} - 1)$; substituting $y=\frac{1}{x}$ you have a quadratic in $x$ to solve.

DrM
  • 1,323
4

Alternative approach:

The equation

$$x^4+2x^3-14x^2+2x+1=0, \tag1 $$

can be directly converted into a quadratic equation by recognizing that the equation in (1) above is a symmetric quartic equation. That is, the $~x^4~$ and $~x^0~$ coefficients are equal, and the $~x^3~$ and $~x^1~$ coefficients are equal.

Therefore, the method employed in this answer may be used.

First, the equation in (1) above is transformed into

$$\left( ~x^2 + \frac{1}{x^2} ~\right) + 2\left( ~x + \frac{1}{x} ~\right) - 14 = 0. \tag2 $$

Then, you set $~w = \displaystyle \left( ~x + \frac{1}{x} ~\right),~$ and note that
$\displaystyle w^2 - 2 = \left( ~x^2 + \frac{1}{x^2} ~\right).$

Therefore, the equation in (2) above becomes

$$( ~w^2 - 2) + 2w - 14 = 0 \implies w^2 + 2w - 16 = 0 \implies $$

$$w = \frac{1}{2} \left[ -2 \pm \sqrt{ ~4 + 64 ~} ~\right] = -1 \pm \sqrt{17}.$$

Since $~w = \displaystyle \left( ~x + \frac{1}{x} ~\right),~$ which must be positive, the negative root above must be rejected.

Therefore,

$$w = \left( ~x + \frac{1}{x} ~\right) = -1 + \sqrt{17} \implies $$

$$x^2 + x \left( ~1 - \sqrt{17} ~\right) + 1 = 0 \implies $$

$$x = \frac{1}{2} \left[ ~\sqrt{17} - 1 \pm \sqrt{~18 - 2\sqrt{17} - 4 ~} \right]$$

$$= \frac{1}{2} \left[ ~\sqrt{17} - 1 \pm \sqrt{~14 - 2\sqrt{17} ~} \right]. \tag3 $$

What the equation in (3) above represents is that there are $~2~$ different solutions, depending on whether $~x < y,~$ or $~y < x.$


$\underline{\text{Sanity Checking}}$

The solution given in (3) above may be routinely sanity checked to verify that both of the following constraints are satisfied:

  • $xy = 1.$

  • $(x + 1)^2 + (y+1)^2 = 4^2 = 16.$

$$\left\{ ~\frac{1}{2} \left[ ~\sqrt{17} - 1 + \sqrt{~14 - 2\sqrt{17} ~} \right] ~\right\} \\ \times \left\{ ~\frac{1}{2} \left[ ~\sqrt{17} - 1 - \sqrt{~14 - 2\sqrt{17} ~} \right] ~\right\} $$

$$= \frac{1}{4} \left\{ ~ \left[ ~18 - 2\sqrt{17} ~\right] - \left[ ~14 - 2\sqrt{17} ~\right] ~\right\} = \frac{4}{4}.$$

Note that $~(R + S)^2 + ~(R - S)^2 = 2(R^2 + S^2).$

Therefore,

$$\left\{ ~\frac{1}{2} \left[ ~\sqrt{17} + 1 + \sqrt{~14 - 2\sqrt{17} ~} \right] ~\right\}^2 \\ + \left\{ ~\frac{1}{2} \left[ ~\sqrt{17} + 1 - \sqrt{~14 - 2\sqrt{17} ~} \right] ~\right\}^2 $$

$$= 2 \times \frac{1}{4} \times \left\{ ~\left[ ~18 + 2\sqrt{17} ~\right] + \left[ ~14 - 2\sqrt{17} ~\right] ~\right\}$$

$$= \frac{2}{4} \times (18 + 14) = 16.$$

user2661923
  • 42,303
  • 3
  • 21
  • 46
1

Another way to solve leads to a quartic too and it is because of the unknowns are irrational of four degree. For example we have $$x^2+1=(2+h)^2\\y^2+1=(2-h)^2\\xy=1$$ which yields $$(h^2+4h+3)(h^2-4h+3)=1\implies h^4-10h^2+8=0$$ or $$(h^2-5)^2-17=0$$ so we can make $$\overline{AD}=2-\sqrt{5-\sqrt{17}} \text{ and }\overline{DC}=2+\sqrt{5-\sqrt{17}}$$ In this case the resultant equation is simpler than yours but it is also irreducible over $\mathbb Q$ and the unknowns are of the four degree. You can verify that solutions of your quartic are also irrational of the four degree.

Ataulfo
  • 32,657
  • 2
    "The unknowns are of fourth degree" - You can set $h^2 = k$ to get a quadratic there – BlueRaja - Danny Pflughoeft Jan 30 '25 at 09:19
  • It is impossible that the unknowns $x, y$ or $AD$ and $DC$ be roots of a quadratic equation with rational coefficients because all of them are irrational of the four degree. Completely other thing is that these unknowns be roots of a quadratic equation but with irrational (of second degree) coefficients. – Ataulfo Jan 30 '25 at 15:30
0

Couldn't you just use simple observational geometry, and the Pythagorean Theorem?

If $BEDF$ is a unit square, then $AED$ and $DFC$ are both right triangles of the same size. And that means point $D$ lies in the center of line $AC$, given by sheer geometry: angle $ADE$ must equal the angle $DCF$, and $CDF$ must equal $DAE$, since every right triangle's angles add up to $180°$, line $ADC$ is $180°$, and angle $EDF$ is $90°$ (unit SQUARE) no matter how long $x$ or $y$ are.

Since $AED$ is a right triangle, and $D$ is the midpoint of $AC$ (making line $AD$ , $2$ units long), then you have a Pythagorean Theorem:

$$y^2 + 1^2 = 2^2$$

Simplifying this, you get:

$$y^2 = 4 - 1 = 3$$

$$y = \sqrt{3}$$

Since $y$ is the length of line $AE$, and you're given the length of line $EB$ as $1$ unit, then the length of line $AB$ is $\left(1 + \sqrt{3}\right)$ units

Thinh Dinh
  • 8,233
  • D is not the midpoint of AC unless x=y=1, i.e. unless all of the triangles are 90°-45°-45° triangles. However, all of the triangles are necessarily similar, meaning that EAD = FDC, ADE = DCF, and the ratios AE:ED and DF:FC are equal. – Jivan Pal Jan 31 '25 at 23:10
  • Furthermore, if x=y=1, then AC = 2√2 rather than 4, so this cannot be the case. Hence, you have concluded y=√3, a contradiction. – Jivan Pal Jan 31 '25 at 23:13
  • D must be the midpoint, because the sum of all angles around a line is 180, and the sum of all angles around a triangle is also 180. So, if we sum up the angles around the triangle, we get EAD+ADE+DEA=180. Since we know DEA is 90 (unit square), it simplifies to: ADE=90-EAD. And, blessed with geometry, we can also calculate the angles around the line at point D: ADE+EDF+CDF=180. And, once again, we have a united square, so EDF is 90, and we can simplify to: ADE=90-CDF. And since ADE is not changing, we can infer that CDF=EAD, making D the midpoint. – stillborn86 Feb 02 '25 at 00:11
  • And not to mention the law of parallel lines and angles. Since AE and DF are parallel lines, and they both intersect line ADC, then: EAD=FDC. Same goes for the bottom angles: DCF=ADE. And obviously the third angles are right, due to the unit square: AED=DFC. Since all of the characteristic angles are the same, the two triangles are the same... including the lengths of their hypotenuse: AD=DC. This makes D the midpoint of line AC. – stillborn86 Feb 02 '25 at 00:46
  • And x and y can never equal 1, they're both sqrt(3), which makes your contradiction moot? – stillborn86 Feb 02 '25 at 00:48
  • "we can infer that CDF=EAD". This is indeed correct. Likewise, ADE = DCF. However, all we know about these two pairs of equal angles in relation to each other is that they sum to 90°, e.g. CDF+ADE = 90°. We cannot conclude e.g. CDF = ADE = 45°, which is what your "solution" requires. – Jivan Pal Feb 02 '25 at 16:30
  • Let B be the origin (0, 0), and D be the point (1, 1). Consider the case where A is the point (0, 1.5). What are the coordinates of C (hint: use the equation of the line AD to determine this), and thus what are the lengths of AD and DC? – Jivan Pal Feb 02 '25 at 16:32
  • Said another way: You are assuming that the triangle is isosceles from looking at the diagram, but nothing about the problem statement requires this. Consider y being very small, then x must be very large. – Jivan Pal Feb 02 '25 at 16:40
  • "they're both sqrt(3)" — If that were the case, then the two smaller triangles would not be isosceles, and thus ADC wouldn't be a single straight line, but two connected line segments, with angle ADC > 180°. – Jivan Pal Feb 02 '25 at 16:45
  • No, my solution DOES NOT require the angles to be 45°. Like I said, the side is sqrt(3), which makes them 30-60-90 triangles, with sides 1-2-sqrt(3)... It's the unit circle, my guy. And with one triangle a 30-60-90, the other triangle is a 60-30-90. This makes all three angles around point D 30+60+90=180, or a straight line. I'm not sure where you're getting 45°, or isocoles triangles, but that's all wrong, and the basis of why your "proof" is wrong. – stillborn86 Feb 03 '25 at 17:29
  • "Let B be the origin (0, 0), and D be the point (1, 1). Consider the case where A is the point (0, 1.5). What are the coordinates of C" A wouldn't be at 1.5, it would be 1+y, and the coordinates of C would be (1+x, 0). The coordinates of A would ultimately be (0, 1+sqrt(3)), or (0, 2.732). – stillborn86 Feb 03 '25 at 17:40
  • I am asking you to consider the case where y=0.5. Either deduce the value of x, or prove that y cannot be 0.5. – Jivan Pal Feb 04 '25 at 17:48
  • Said another way: yes, your triangles are 30-60-90 triangles, but the angles around point D are 60 in one triangle, 90 in the square, and 60 in the other triangle, and these angles' sum exceeds 180. – Jivan Pal Feb 05 '25 at 15:59
-1

The two triangles are similar. y=1/x=p/q (two hypotenuses) So y=p/q and p+q=4 p-qy=0, q(y+1)=4 1<q<3 (hypotenuse) So y(1/3;3) Analogical for x