How to Actually Use the Double Angle Formulas for Cosine
The three forms of the cosine double angle identity are cos(2x) = cos²(x) - sin²(x), cos(2x) = 2cos²(x) - 1, and cos(2x) = 1 - 2sin²(x). Most textbooks present them as interchangeable options, but in practice they live in completely different parts of your workflow. You pick one based on what you already have or what you're trying to isolate, and picking wrong just adds a step you didn't need. I learned this the hard way during a signal processing project where I was simplifying a modulated waveform with terms like cos²(t) and sin²(t) mixed together. I stuck with the first form for twenty minutes because it felt most "balanced," then realized the entire expression collapsed into a single line once I switched to the 2cos²(x) - 1 version. The math was the same either way. The time cost was not.
Deriving the Double Angle For Cosine from Scratch
It starts from the angle addition formula: cos(A + B) = cos(A)cos(B) - sin(A)sin(B). Set A = B = x and you immediately get cos(2x) = cos²(x) - sin²(x). That's the base form. The other two come from substituting the Pythagorean identity sin²(x) = 1 - cos²(x) or cos²(x) = 1 - sin²(x) into that first result. One substitution gives you 2cos²(x) - 1. The other gives you 1 - 2sin²(x). That's it. Four lines of algebra, nothing mysterious. Common confusion point: people sometimes flip the signs and write cos(2x) = 1 - 2cos²(x) by accident. It's an easy mistake because the sine version and cosine version look nearly identical structurally. I've caught myself doing it twice in a row during exams. The trick that finally worked for me was memorizing which one has the positive leading term: the cosine-squared version always starts with +2cos²(x), and the sine-squared version always starts with +1.
When to Use Which Form
If your problem contains cos²(x) and you need to eliminate the square, use cos(2x) = 2cos²(x) - 1. Rearrange it to cos²(x) = (1 + cos(2x))/2 and you've just linearized a quadratic term. This is the form behind half-angle formulas in integration, and it's the one you reach for when converting a squared trig term into something integrable. If your problem is all sines and you need to remove a sin²(x) term, use cos(2x) = 1 - 2sin²(x). Rearranged, sin²(x) = (1 - cos(2x))/2. Same idea, different sign. This appears constantly in Fourier analysis when you're computing power in a signal over a period. The original form cos(2x) = cos²(x) - sin²(x) is useful almost exclusively when you need to keep both sin and cos present, like in product-to-sum derivations or when converting between rectangular and polar representations in complex number problems.
Get the Full Details
Integration Applications
Integrating cos²(x) dx over [0, ] is a textbook example. Without the double angle formula you'd be stuck. With it: substitute cos²(x) = (1 + cos(2x))/2, integrate term by term, and you get /2. That's one line if you know which form to grab. Two pages if you try to work around it. Similarly, sin²(x) dx from 0 to equals /2 as well, using sin²(x) = (1 - cos(2x))/2. The fact that both give the same result over a full half-period isn't coincidence—it's a direct consequence of sin² and cos² being phase-shifted versions of each other, and the double angle formula makes that relationship explicit rather than hidden.
Pitfalls and Edge Cases
One thing that trips people up is the domain. The double angle formulas are identities, so they hold for all real x. But if you're working in a computational environment where angles are in degrees versus radians, cos(2x) will give completely wrong numerical results if x is interpreted in the wrong unit. I once spent an hour debugging a MATLAB script where the double angle was being evaluated in degrees while the rest of the code assumed radians. The formula itself was fine. The units weren't. Another edge case: numerical stability at extreme values. When x is very close to /2, cos²(x) and sin²(x) are both near zero and subtracting them in the original form cos(2x) = cos²(x) - sin²(x) can lose precision due to floating point cancellation. In that regime, the 2cos²(x) - 1 form is more stable because you're not subtracting two nearly equal numbers—you're multiplying and subtracting from 1, which is better conditioned. This matters in financial modeling or physics simulations where you chain hundreds of these evaluations together. A limitation worth noting: the double angle formulas don't help you when you have cos(x) + cos(2x) mixed with unrelated angles like cos(3x) in the same expression. They reduce the argument to a multiple, but they don't simplify sums of different frequencies. For that you need sum-to-product or product-to-sum formulas, which are a separate family of identities. People sometimes try to force the double angle into situations where it doesn't belong and end up going in circles.
Numerical Example
Find cos(2·15°) using only cos(15°). cos(15°) = (6 + 2)/4 0.9659. Square it: cos²(15°) 0.9330. Apply 2cos²(x) - 1: 2(0.9330) - 1 = 0.8660. Check against a calculator: cos(30°) = 3/2 0.8660. Matches. Now do the same with the sine form. sin(15°) = (6 - 2)/4 0.2588. Square it: sin²(15°) 0.0670. Apply 1 - 2sin²(x): 1 - 2(0.0670) = 0.8660. Same answer, different path. Both paths work. Pick the one with less arithmetic in front of you. The formulas are simple enough that the main challenge isn't memorization—it's recognizing which version your problem actually needs at the right moment. Most mistakes I see aren't calculation errors. They're picking the wrong form and then not noticing until you're three steps deeper into a derivation that could have been two steps shorter.
