I have been trying to solve a particular problem that establishes both the existence and uniqueness of a cubic hermite interpolant on some generic interval $[a,b]$. Briefly, for a function $f$ we require a cubic polynomial $p$ that satisfies

$$p(a) = f(a), \ p^{\prime}(a) = f^{\prime}(a), $$ $$p(b) = f(b), \ p^{\prime}(b) = f^{\prime}(b). $$

I am aware that there are several approaches for this but the approach that my book recommends is to use the fact (which I have successfully proven) that the differential equation $u^{(4)}(x) = f(x)$ with $u(a) = u(b) = u^{\prime}(a) = u^{\prime}(b) = 0$ has a unique solution. Using this one can then easily show that the cubic hermite interpolant is unique but I am baffled as to how this proves existence and therefore I would appreciate some help.

I should mention that this is an exercise from Prenter's "Splines and Variational Methods".

Thank you in advance.


The problem of existence and uniqueness of a cubic polynomial passing through two given points with two given slopes reduces to a linear system of four equations in four unknowns, as follows. Take the most general cubic, $P(x)=\alpha x^3+\beta x^2 + \gamma x + \delta$. Then $P'(x)=3 \alpha x^2 + 2\beta x + \gamma$, and you want: $$ \begin{aligned} \alpha a^3 + \beta a^2 + \gamma a + \delta &= f(a)\\ \alpha b^3 + \beta b^2 + \gamma b + \delta &= f(b)\\ 3\alpha a^2 + 2\beta a + \gamma &= f'(a)\\ 2\alpha b^2 + 2\beta b + \gamma &= f'(b)\\ \end{aligned} $$ The corresponding matrix of coefficients is $$ A= \begin{pmatrix} a^3 & a^2 & a & 1\\ b^3 & b^2 & b & 1\\ 3a^2 & 2a & 1 & 0 \\ 3b^2 & 2b & 1 & 0 \end{pmatrix} $$

A straighforward calculation shows that the determinant of this matrix is $-(a-b)^4$, which is nonzero if we assume that $a\neq b$, and of course we assume that. From Linear Algebra we know that the above linear system of equations will have a unique solution. That is, there exists a solution, and it is unique. Moreover, using standard techniques it is possible to express the coefficients of the interpolating polynomial as functions of the input data $(f(a),f(b),f'(a),f'(b))$.

  • $\begingroup$ Yes I know all this but my question is about the specific exercise. How does it prove existence, as Prenter claims? $\endgroup$ – JohnK Apr 11 '19 at 10:24
  • $\begingroup$ I don't see how existence of an interpolating cubic follows from the uniqueness of the solution to the IVP $u^{(4)}(x)=f(x)$. The way things are usually done, is that if you know that this differential equation has a unique solution you look at $f-P$ where you already have proven that $P$ exists. Maybe you are just reading too much into the exercise. $\endgroup$ – uniquesolution Apr 11 '19 at 10:41
  • $\begingroup$ Perhaps you are right, thanks for your contribution. $\endgroup$ – JohnK Apr 11 '19 at 11:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.