In 11.0,

Series[(1 - Sqrt[1-4z])/2, {z, 1/4, 1}]


1/2 - I Sqrt[z-1/4] + O[z-1/4]^(3/2)

as one would expect. From 11.1 (and still in 11.3), the output is

(1/2 - 1/2 Sqrt[1-4z]) + O[z-1/4]^2

Even worse,

SeriesCoefficient[(1 - Sqrt[1-4z])/2, {z, 1/4, 0}]

now returns 1/2 - 1/2 Sqrt[1-4z] which is nonsense, rather than the correct answer of 1/2. Similarly,

SeriesCoefficient[(1 - Sqrt[1-4z])/2, {z, 1/4, 1/2}]

now gives 0 rather than -I.

This is presumably related to this, but is not yet fixed and seems like a more fundamental failure in the handling of Puiseux series to me.

This has broken a lot of code I regularly use.

  • 3
    $\begingroup$ Getting Series to behave at branch points has been something of a challenge. The issues indicated here have been addressed, although obviously not in time for version 11.3. My apologies for any inconveniences this may have caused. $\endgroup$ – Daniel Lichtblau Apr 23 '18 at 15:59
  • $\begingroup$ This appears to be a relative of the issue from here and others linked to in a comment there. $\endgroup$ – Daniel Lichtblau Apr 23 '18 at 16:01
  • $\begingroup$ @DanielLichtblau: Thanks for the update; glad to know a fix is on its way; I'll keep using 11.0 until 11.4 appears. $\endgroup$ – David Bevan Apr 23 '18 at 17:47

Your Answer

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

Browse other questions tagged or ask your own question.