The magic is that B2−4AC is a rotation invariant: rotate the axes to kill the xy term, and this number doesn't change. So it captures something intrinsic to the shape, not to the tilt.
Step 1 — Rotate the axes by angle θ. Replace
x=x′cosθ−y′sinθ,y=x′sinθ+y′cosθ.Why this step? A tilted conic becomes axis-aligned in some rotated frame; we want to find that frame.
Step 2 — Substitute into Ax2+Bxy+Cy2. Collecting terms gives new coefficients A′,B′,C′:
A′=Acos2θ+Bcosθsinθ+Csin2θ,C′=Asin2θ−Bcosθsinθ+Ccos2θ,B′=Bcos2θ−(A−C)sin2θ.Why this step? We want to see how A,B,C transform, so we can hunt for a quantity that stays fixed.
Step 3 — Choose θ to eliminate the cross term. Set B′=0:
Bcos2θ=(A−C)sin2θ⇒tan2θ=A−CB.Why this step? Once B′=0, the equation has no xy: it's a standard, un-tilted conic we recognize instantly.
Step 4 — Compute B′2−4A′C′. After algebra (verified below):
B′2−4A′C′=B2−4AC.Why this step? This proves B2−4AC is invariant under rotation — it's a property of the shape itself, independent of tilt.
Step 5 — In the rotated frame (B′=0) read off the shape. Equation looks like
A′x′2+C′y′2+⋯=0.
A′C′>0 (same sign): A′x′2+C′y′2=const → ellipse. Then B′2−4A′C′=−4A′C′<0. ✓
A′C′<0 (opposite signs): → hyperbola. Then −4A′C′>0. ✓
A′C′=0 (one is zero): only one squared term → parabola. Then −4A′C′=0. ✓
Since B2−4AC=B′2−4A′C′=−4A′C′, the sign of B2−4AC tells us the sign of A′C′ — hence the shape. QED.
A conic can collapse: an ellipse into a point, a hyperbola into two crossing lines, a parabola into parallel lines. To catch these, use the full 3×3 determinant:
Δ=AB/2D/2B/2CE/2D/2E/2F.
Recall Predict before revealing (Forecast-then-Verify)
Q: Sign of B2−4AC for 4x2+4xy+y2+x=0?
Forecast, then check: B2−4AC=16−16=0 → parabola (indeed 4x2+4xy+y2=(2x+y)2).
Recall Feynman: explain to a 12-year-old
Imagine a flashlight. Shine it straight at a wall — you get a round/oval blob (ellipse). Tilt it just so the edge of the beam skims the wall — the blob opens up forever on one side (parabola). Tilt more and the light splits into two curved patches (hyperbola). The messy equation with x2,xy,y2 is just describing where the light lands. Instead of drawing it, we compute one magic number, B2−4AC: negative = oval, zero = the just-skimming case, positive = the two-patch case. The xy term only means the flashlight is held at a slant — it changes the number but never changes which of the three shapes you get.
Dekho, kisi bhi conic ka general equation hota hai Ax2+Bxy+Cy2+Dx+Ey+F=0. Ismein sabse zyada important part hai degree-2 wala group yaani Ax2+Bxy+Cy2 — yehi shape decide karta hai. Linear part Dx+Ey sirf curve ko idhar-udhar shift karta hai, aur F position adjust karta hai. Shape janne ke liye ek hi magic number nikalte hain: discriminantB2−4AC.
Rule bilkul simple hai: agar B2−4AC<0 hai to ellipse (aur agar A=C aur B=0 ho to circle), agar =0 hai to parabola, aur agar >0 hai to hyperbola. Yaad rakhne ke liye: "Negative Loops, Zero Parks, Positive Splits." Dhyan rahe — yahan B matlab xy ka coefficient hai, wo school wale b2−4ac waala b (jo x ke saath aata hai) nahi hai. Yeh sabse common galti hai.
Ab sawaal — yeh ek number kaam kaise karta hai chahe conic tilted (tedha) ho? Kyunki B2−4ACrotation invariant hai. Matlab agar aap axes ko ghuma kar xy term hata do (jiske liye tan2θ=B/(A−C)), tab bhi yeh number same rehta hai. Isliye yeh curve ki asli shape batata hai, tilt ka effect ignore karke.
Ek aur important baat: kabhi-kabhi conic "degenerate" ho jaati hai — jaise ellipse ek point ban jaaye, ya hyperbola do crossing lines ban jaaye. Isko pakadne ke liye 3×3 determinant Δ check karo (yaad rahe off-diagonal mein B/2,D/2,E/2 aata hai). Agar Δ=0 to conic degenerate hai. Toh strategy: pehle Δ se dekho genuine conic hai ya nahi, phir B2−4AC se type nikaalo. Bas — solve karne ki zaroorat hi nahi, sirf coefficients padho aur ek number nikaalo!