vibra precision: double -> single complex -> double complex

[Nick Papior]

In vibra the dc variable is build the dynamical matrix build with phases.

However, it is only a complex variable (not complex(dp)). Hence the dynamical matrix is build using single precision, then copied to a double precision array.
This is particularly important for the low energy eigenmodes.

This patch fixes this:

=== modified file 'Util/Vibra/Src/vibra.f'
--- Util/Vibra/Src/vibra.f 2016-01-25 16:00:16 +0000
+++ Util/Vibra/Src/vibra.f 2019-02-20 09:01:01 +0000
@@ -78,7 +78,7 @@
       real*8 zero(3,3,maxa), zeroo(3,3)

 c Work space for diagonalization.
- complex dc(maxd,maxd),phase,IRtrm,vecmw(3)
+ complex*16 dc(maxd,maxd),phase,IRtrm,vecmw(3)
       real*8 work(maxd),work2(2,maxd)
       real*8 dd(maxd,maxd),zr(maxd,maxd),zi(maxd,maxd),omega(maxd)

@@ -517,7 +517,7 @@
             enddo
 c qr = q(1)*r(1) + q(2)*r(2) + q(3)*r(3)
             do in = 1,neq
- phase = cos(qr(in))*(1.0,0.0) + sin(qr(in))*(0.0,1.0)
+ phase = cmplx(cos(qr(in)), sin(qr(in)), 8)
               do ii=1,3
               do ij=1,3
                 ix = (i-1)*3+ii
@@ -542,8 +542,8 @@

         do ix=1,3*natoms
           do jx=ix,3*natoms
- dd(ix,jx)=imag(dc(jx,ix))
- dd(jx,ix)=real(dc(jx,ix))
+ dd(ix,jx)=aimag(dc(jx,ix))
+ dd(jx,ix)=real(dc(jx,ix), 8)
           enddo
         enddo

@@ -571,7 +571,7 @@
 c Mass weight eigenvectors
               rmass = 1.0d0/sqrt(xmass(j))
               do ii = 1,3
- vecmw(ii) = cmplx(zr(ind+ii,i),zi(ind+ii,i))
+ vecmw(ii) = cmplx(zr(ind+ii,i),zi(ind+ii,i), 8)
                 vecmw(ii) = vecmw(ii)*rmass
               enddo

[Nick Papior (nickpapior)]

I'll fix this in 4.0 and push up through. Probably wait for next week ;)