Error analysis for frequency-dependent interpolation formulas using first derivatives (Q544070)

Let \(f\) be an oscillatory function with a frequency \(\omega\), which can be represented in the form \(f(x)=f_1(x)\cos(\omega x)+f_2(x)\sin(\omega x)\). For three equally spaced nodes \(c-h,c,c+h\) \((h>0)\), the author considers the following interpolation formula: \(f(c+th)\approx I(t)=\alpha_1 f(c-h)+ \alpha_2f(c)+\alpha_3(c+h)+ h(\beta_1f'(c-h)+\beta_2f'(c)+\beta_3f'(c+h))\), \(t\in[-1,1]\). If the values of \(f\) and \(f'\) are known at the nodes, using the functional \(L(f(x),h,\zeta) =f(x+th)-\alpha_1f(x-h)-\alpha_2f(x)-\alpha _3(x+h)-h[\beta_1f'(x-h) +\beta_2f'(x)+\beta_3f'(x+h)]\) and imposing the conditions \(L(x^{n-1}\exp(\pm i\omega x),h,\zeta)=0\) \((n=1,2,3)\) one obtains the coefficients \(\alpha_1,\alpha_2,\alpha_3,\beta_1,\beta_2, \beta_3\). A similar formula \(\widetilde{I}(t)\) is given in the case of unequally spaced nodes \(c-h_1,c\), \(c+h_2,(h_1,\) \(h_2>0)\). The interpolation formulas \(I(t)\) and \(\widetilde{I}(t)\) have the same values and same slopes as \(f\) at the considered three nodes. The author uses the interpolating formula \(I(t)\) to the function \(f(x)=x\sin x\cos(\omega x)+x\cos x\sin(\omega x),\) \(x\in[0,3]\) (in fact a composite -- \(I(t)\) formula on \([0,1]\cup[1,2]\cup[1,3]\)). One concludes that the obtained result is more accurate and stable than the result obtained by the classical composite Hermite interpolating formula and composite exponential-fitting interpolation formula.