Questions: Even and Odd Extensions in Fourier Series

Question 1

You are solving the heat equation on a rod [0, L] with boundary conditions u(0,t) = 0 and u(L,t) = 0 (zero temperature at both ends). Which extension should you use for the initial condition f(x), and why?

• A. Even extension, because reflecting across the y-axis preserves positivity of temperatures
• B. Odd extension, because it produces a sine series whose terms automatically vanish at x = 0 and x = L
• C. Either extension works; the choice only affects how many Fourier coefficients you must compute
• D. Even extension, because cosine functions are smoother and converge faster for heat problems

Answer: (B)

The boundary conditions u(0,t) = 0 and u(L,t) = 0 are Dirichlet conditions — the function value is zero at the endpoints. Sine functions sin(nπx/L) are exactly zero at x = 0 and at x = L for all integers n, so the sine series automatically satisfies these conditions by construction. The odd extension produces a sine series. The even extension produces a cosine series, whose terms are *not* zero at x = 0 and x = L — so using an even extension would violate the boundary conditions. Option C is the key misconception: the choice is not arbitrary.

Question 2

Why does extending a function f on [0, L] evenly to [−L, L] produce a Fourier series with only cosine terms and no sine terms?

• A. Because cos(0) = 1, which ensures the series converges at the boundary x = 0
• B. Because the even extension creates an even function, and the Fourier integral of an even function times an odd function (sine) is zero
• C. Because cosines have lower frequency and are therefore more suitable for smooth extensions
• D. Because the even extension doubles the period, eliminating the need for sine basis functions

Answer: (B)

The key is symmetry. Sine functions are odd: sin(−x) = −sin(x). Cosine functions are even: cos(−x) = cos(x). The Fourier coefficient bₙ involves integrating f(x)·sin(nπx/L) over [−L, L]. If f is even and sin is odd, their product is odd, and the integral of an odd function over a symmetric interval [−L, L] is exactly zero. So all bₙ = 0, leaving only the cosine coefficients aₙ. This is not a computational trick — it follows directly from the symmetry structure of the basis functions.

Question 3

The choice of even versus odd extension for a half-range function affects which boundary conditions the resulting Fourier series automatically satisfies.

• T. True
• F. False

Answer: True

This is the core reason the choice of extension is not arbitrary in PDE applications. The even extension produces a cosine series; cosines have zero derivative at x = 0 and x = L (they satisfy Neumann conditions: f'(0) = f'(L) = 0). The odd extension produces a sine series; sines have zero value at x = 0 and x = L (they satisfy Dirichlet conditions: f(0) = f(L) = 0). When solving a PDE, the extension must match the physical boundary conditions — otherwise the series solution will not satisfy those conditions at the boundary, making it invalid.

Question 4

Either the even or the odd extension can be used for any half-range Fourier series problem; the choice primarily affects computational convenience, not the correctness of the solution.

• T. True
• F. False

Answer: False

This is the most important misconception to avoid. In PDE problems, the extension must encode the boundary conditions. Using an odd extension (sine series) for a problem with Neumann (insulated) boundary conditions — where the derivative, not the value, is zero at the endpoints — will produce a series that does not satisfy those conditions. The solution will be mathematically valid as a Fourier series of the extended function but will be wrong as a solution to the physical problem. The extension is a choice with real consequences, not mere computational convenience.

Question 5

Explain why using an odd extension is appropriate for the heat equation with zero-value (Dirichlet) boundary conditions, and what would happen if you used an even extension instead.

Model answer: The odd extension produces a sine series: f(x) = Σ bₙ sin(nπx/L). Each term sin(nπx/L) equals zero at x = 0 and at x = L for all positive integers n. So the series automatically satisfies u(0,t) = 0 and u(L,t) = 0 by construction. If you used an even extension instead, you would get a cosine series: f(x) = a₀/2 + Σ aₙ cos(nπx/L). Cosines are generally nonzero at x = 0 (cos(0) = 1) and at x = L, so the cosine series would not satisfy the Dirichlet boundary conditions. You would be solving a different (and physically wrong) problem.

The deeper insight is that the boundary conditions determine the appropriate basis functions, and the extension determines which basis functions appear. This is not just a calculation trick — it is the mathematical encoding of physics. Choosing the wrong extension means choosing basis functions that are incompatible with the physical constraints of the problem, yielding a series that cannot represent the correct solution regardless of how accurately the coefficients are computed.