Skip to main content
Post Migrated Away to math.stackexchange.com
added 55 characters in body
Source Link
Carlo Beenakker
  • 188.2k
  • 18
  • 448
  • 651

The first term in Watson's expansion amounts to the saddlepoint approximation (expansion of the exponent to second order around $t=1$), which gives $$I_{sp}=\frac{1}{2}e^{-x}\int_{-\infty}^\infty e^{-(x/2)(1-t)^2}\,dt=\frac{\sqrt{\pi } e^{-x}}{\sqrt{2 x}}.$$ This is quite accurate, see the plot (gold: exact; blue: saddlepoint approximation)

The saddlepoint approximation (expansion of the exponent to second order around $t=1$) gives $$I_{sp}=\frac{1}{2}e^{-x}\int_{-\infty}^\infty e^{-(x/2)(1-t)^2}\,dt=\frac{\sqrt{\pi } e^{-x}}{\sqrt{2 x}}.$$ This is quite accurate, see the plot (gold: exact; blue: saddlepoint approximation)

The first term in Watson's expansion amounts to the saddlepoint approximation (expansion of the exponent to second order around $t=1$), which gives $$I_{sp}=\frac{1}{2}e^{-x}\int_{-\infty}^\infty e^{-(x/2)(1-t)^2}\,dt=\frac{\sqrt{\pi } e^{-x}}{\sqrt{2 x}}.$$ This is quite accurate, see the plot (gold: exact; blue: saddlepoint approximation)

Source Link
Carlo Beenakker
  • 188.2k
  • 18
  • 448
  • 651

The saddlepoint approximation (expansion of the exponent to second order around $t=1$) gives $$I_{sp}=\frac{1}{2}e^{-x}\int_{-\infty}^\infty e^{-(x/2)(1-t)^2}\,dt=\frac{\sqrt{\pi } e^{-x}}{\sqrt{2 x}}.$$ This is quite accurate, see the plot (gold: exact; blue: saddlepoint approximation)