Find T, N, and K for a Space Curve: A Step-by-Step Guide
You're staring at a problem set, and there it is — "find T, N, and κ for the space curve r(t) = ...". That said, your professor mentioned the Frenet-Serret frame back in lecture, but honestly, it went by fast. Now you're trying to piece it together from your notes and a textbook that seems to assume you already know what you're doing Still holds up..
Sound familiar?
Here's the thing — finding the tangent, normal, and curvature for a space curve is one of those skills that looks way more intimidating than it actually is. Once you see the pattern, it's actually pretty straightforward. Let me walk you through it.
What Is the Frenet-Serret Frame?
When you're dealing with a curve in three-dimensional space — not just a flat 2D line, but something that twists and turns in 3D — you need a way to describe how it bends and twists at every point. That's where the Frenet-Serret frame comes in.
It's a collection of three perpendicular unit vectors that travel along with you as you move along the curve:
- T (Tangent) — points in the direction you're moving
- N (Normal) — points toward the direction the curve is bending
- B (Binormal) — perpendicular to both T and N, showing the "twist" of the curve
Together with curvature (κ) and torsion (τ), these give you a complete picture of how a space curve behaves locally.
Why does this matter? Because in physics, engineering, computer graphics, robotics — anywhere you need to describe motion along a curved path — these quantities show up. The curvature tells you how sharply the curve bends. The torsion tells you if it's twisting out of a plane. The T and N vectors give you a moving coordinate system that follows the curve.
How to Find T, N, and κ for a Space Curve
Alright, let's get into the actual process. I'll walk you through each step with a concrete example so you can see how it works in practice.
The Setup
Suppose you have a space curve given in parametric form:
r(t) = ⟨t, t², t³/3⟩
That's a nice curve to work with — simple enough to follow, but with enough complexity to show real 3D behavior It's one of those things that adds up. And it works..
Step 1: Find the Derivative r'(t)
The first thing you need is the velocity vector. Just differentiate each component:
r'(t) = ⟨1, 2t, t²⟩
This vector points in the direction the curve is moving at each point. But it's not necessarily a unit vector — its length changes depending on how fast t is changing and where you are on the curve.
Step 2: Compute the Unit Tangent T
To get the unit tangent vector T, you divide r'(t) by its magnitude:
T = r'(t) / |r'(t)|
First, find the magnitude:
|r'(t)| = √(1² + (2t)² + (t²)²) = √(1 + 4t² + t⁴)
So:
T = ⟨1, 2t, t²⟩ / √(1 + 4t² + t⁴)
This vector always has length 1 and always points along the direction of motion.
Step 3: Find the Curvature κ
Curvature measures how fast the direction is changing. There are a couple of equivalent formulas, but here's the most useful one for computation:
κ = |r'(t) × r'(t)| / |r'(t)|³
Actually, that's not quite right — let me fix that. The formula is:
κ = |r'(t) × r''(t)| / |r'(t)|³
First, find r''(t) by differentiating r'(t):
r''(t) = ⟨0, 2, 2t⟩
Now compute the cross product r'(t) × r''(t):
r'(t) × r''(t) = | i j k | | 1 2t t²| | 0 2 2t|
= i(2t · 2t - t² · 2) - j(1 · 2t - t² · 0) + k(1 · 2 - 2t · 0) = i(4t² - 2t²) - j(2t) + k(2) = ⟨2t², -2t, 2⟩
Now find the magnitude:
|r'(t) × r''(t)| = √((2t²)² + (-2t)² + 2²) = √(4t⁴ + 4t² + 4) = 2√(t⁴ + t² + 1)
And we already know |r'(t)| = √(1 + 4t² + t⁴), so |r'(t)|³ = (1 + 4t² + t⁴)^(3/2).
Putting it together:
κ = 2√(t⁴ + t² + 1) / (1 + 4t² + t⁴)^(3/2)
That's the curvature. At t = 0, this simplifies to κ = 2/1 = 2. As t gets larger, the curvature changes.
Step 4: Find the Principal Normal N
The principal normal vector N points in the direction the curve is bending. To find it, you take the derivative of T with respect to arc length, then normalize:
N = (dT/ds) / |dT/ds|
Here's the trick — instead of reparameterizing everything in terms of arc length s (which can be messy), you can use this formula:
N = (dT/dt) / |dT/dt|
Or equivalently:
N = (r''(t) - (r'(t) · r''(t))/|r'(t)|² · r'(t)) / κ|r'(t)|
That's a bit heavy, so let me show you the simpler approach. First, find dT/dt by differentiating T:
T = ⟨1, 2t, t²⟩ / R, where R = √(1 + 4t² + t⁴)
Using the quotient rule (or product rule, thinking of it as (1/R) times the numerator):
dT/dt = [⟨0, 2, 2t⟩ · R - ⟨1, 2t, t²⟩ · (dR/dt)] / R²
This gets messy fast. Here's the more practical approach for most problems:
N = (r'(t) × r''(t)) × r'(t) / (|r'(t) × r''(t)| · |r'(t)|)
Let me compute that for our example.
We already have r'(t) × r''(t) = ⟨2t², -2t, 2⟩
Now compute (r' × r'') × r':
= | i j k | | 2t² -2t 2 | | 1 2t t²|
= i((-2t)(t²) - 2(2t)) - j(2t²(t²) - 2(1)) + k(2t²(2t) - (-2t)(1)) = i(-2t³ - 4t) - j(2t⁴ - 2) + k(4t³ + 2t) = ⟨-2t³ - 4t, -(2t⁴ - 2), 4t³ + 2t⟩ = ⟨-2t³ - 4t, 2 - 2t⁴, 4t³ + 2t⟩
Now divide by |r' × r''| · |r'| = 2√(t⁴ + t² + 1) · √(1 + 4t² + t⁴)
So:
N = ⟨-2t³ - 4t, 2 - 2t⁴, 4t³ + 2t⟩ / [2√(t⁴ + t² + 1) · √(1 + 4t² + t⁴)]
At t = 0, this simplifies to:
N = ⟨0, 2, 0⟩ / (2 · 1) = ⟨0, 1, 0⟩
That makes sense — at the origin, the curve is bending in the y-direction.
Step 5: Find the Binormal B (Optional But Useful)
If you need the binormal vector B, it's easy — just take the cross product of T and N:
B = T × N
This gives you a vector perpendicular to both the tangent and the normal. It points in the direction of the "twist" of the curve.
Common Mistakes People Make
Here's where most students get tripped up:
Forgetting to normalize. The tangent vector r'(t) is not T — it's only the direction. You have to divide by its magnitude to get the unit vector. Same goes for N. If your T doesn't have length 1, everything downstream will be wrong.
Using the wrong curvature formula. There are several equivalent formulas for curvature, but make sure you're using one that actually works for your situation. The formula κ = |dT/ds| is always valid but requires arc length parameterization. The cross product formula κ = |r' × r''| / |r'|³ is easier if you have a general parameter t.
Confusing N with the derivative of r. The normal vector N is not r''(t). It's the direction that T is changing — which involves both r' and r'' Small thing, real impact..
Forgetting the chain rule. When you compute dT/ds, remember that dT/ds = (dT/dt) / (ds/dt) = (dT/dt) / |r'(t)|.
Practical Tips That Actually Help
-
Start with a simple example. Before tackling your homework problems, work through r(t) = ⟨cos t, sin t, 0⟩ — that's just a circle in the xy-plane. You know the answer should be κ = 1 everywhere, T rotates around, and N points toward the center. If your formulas give you something different, you know to check your work That alone is useful..
-
Check your answers with geometry. At any point on the curve, ask yourself: does the direction of N make sense? Is the curve bending toward N? Does the curvature seem reasonable — sharp bends should have high κ, gentle curves should have low κ?
-
Organize your work. Keep r', r'', r''' in a row. Calculate magnitudes separately. One small algebra mistake early on will cascade through everything else.
-
Use technology to verify. Once you've done a problem by hand, plug it into WolframAlpha or Mathematica to check. Just make sure you understand what the software is showing you — the output format can be different from what your professor expects.
-
Memorize the key formulas. Not everything — just the core relationships: T = r'/|r'|, κ = |r' × r''|/|r'|³, N = (r' × r'') × r' / (|r' × r''| · |r'|), B = T × N.
FAQ
What's the difference between curvature and torsion?
Curvature (κ) measures how sharply a curve bends — how fast it turns away from a straight line. Practically speaking, torsion (τ) measures how much the curve twists out of a plane. A circle has constant curvature but zero torsion (it stays flat). A helix has constant curvature and constant torsion — it bends and twists uniformly.
Do I always need to find all three vectors?
It depends on the problem. Plus, read the problem carefully — if it asks for "find T, N, and κ," you need all three. Sometimes you only need T and κ. Sometimes you need the full Frenet-Serret frame. If it just asks for curvature, you can often use a simpler formula.
What if the curve is parameterized by arc length s?
If your curve is already given in terms of arc length s, things get simpler. In practice, then T = dr/ds, and κ = |dT/ds|. No division by |r'| needed because |dr/ds| = 1 by definition. Most textbook problems don't give you arc length parameterization directly though — it's usually something you compute after the fact.
Can curvature ever be zero?
Yes — for a straight line, curvature is zero everywhere. Worth adding: if your calculation gives κ = 0 for a curve that clearly bends, something went wrong. Check your cross product computation Turns out it matters..
Why does the normal point in the direction of bending?
Think about it this way: the tangent T tells you where you're going. dT/ds points in the direction T is turning — which is exactly the direction the curve is bending. The derivative of T tells you how that direction is changing. That's why N is parallel to dT/ds.
The Bottom Line
Finding T, N, and κ for a space curve is really just a systematic process: differentiate, compute magnitudes, normalize, repeat. The formulas can look intimidating when they're written all at once, but if you break them into steps — find r', find |r'|, find T; find r'', find the cross product, find κ; then build N from there — it's entirely manageable Worth keeping that in mind. Which is the point..
The key is understanding what each piece actually represents. N is which way you're turning. On the flip side, κ is how fast you're turning. T is which way you're going. Once that clicks, the algebra is just the language for expressing those ideas precisely Easy to understand, harder to ignore..
Work through a few problems, check your answers, and you'll get the hang of it. It's one of those things that feels strange the first time but becomes automatic with practice.
