In spite of large concentrations of phosphorus being incorporated into diamond, the material often remains insulating. It is argued that this occurs through the formation of phosphorus-vacancy complexes which are deep acceptors and compensate any donor. The complex is analyzed using a first principles cluster method. In the ionized state, the defect is diamagnetic and cannot give rise to any internal optical transitions although broad band donor-acceptor transitions are expected - and observed - in material co-doped with nitrogen.
There    has been a considerable    amount  of  work   carried out  on
incorporating  phosphorus in diamond with the  aim of making a shallow
donor defect.  However, almost  all  studies have succeeded in  making
highly resistive material   despite  substitutional P    (P 
 ) being
expected   to  behave   as    a  donor [1, 2, 3].
Electron   paramagnetic resonance (EPR)    provides some  evidence for
P 
  [4], and a second P defect is found in P doped diamond
grown by the high pressure method [5].  The latter defect is
trigonal with very little spin density at the P nucleus.
P has  also  been incorporated into   chemical vapur  deposition grown
(CVD) diamond by adding  PH 
  to the  gas phase [6, 7].
Nitrogen apparently increases the incorporation  of P within the films
yielding concentrations   as high  as 3 
  cm 
  (as
measured via SIMS).  In spite of these large concentrations, the films
were highly resistive.    Photoluminescence (PL) experiments  revealed
broad intense  bands around 1.7 eV   which increased with  [P].  These
were   assigned  to donor-acceptor  pair  recombination  although  the
identities of the   defects are not  known.    Significantly, no sharp
optical lines due to P or the vacancy were detected.
Success at making conducting, n-type diamond has been achieved through a cold implantation of P ions followed by rapid thermal annealing [8]. It may be that this process rapidly removed vacancies resulting from the implantation.
CVD diamond usually  contains  a high  concentration, up  to   
  cm 
  of vacancies [9].  The  lack
of an optical signal due to vacancies in the  P rich material suggests
that they are bound up  with P  in  some form.  Now,  impurity-vacancy
complexes   are   important optical   centers  in  diamond.   The best
understood  is the N-V   center whose electronic  structure  is easily
understood      in       terms        of       the   ideal     neutral
vacancy [10, 11, 12].   This has  the   electronic
configuration  
 .  In   the N-V  defect, where  a   C atom
bordering the vacancy is replaced by N,  the  
  level is split into
an    
 -level   lying below an  e-doublet.    For  (N-V) 
 , the
e-level contains two electrons  in an S=1  state [13].  The
partial occupancy of the e level permits  
  optical
transitions which  are assigned  to  the 1.945 eV optical  center  for
[N-V] 
   [14, 15]   and    2.156 eV  for   neutral
N-V  [16, 13].
It might be thought that the P-V defect  would be similar  to N-V.  In
that case we would expect it to act as an acceptor, with spin 1 in the
ionized state,  and to possess a  sharp luminescence line arising from
the   
   internal transition.  However, we shall show
below  that this is  not  the case and [P-V] 
   has  spin S=0 and no
internal optical transitions are possible.  However, P-V does act as a
deep  acceptor and compensates  substitutional  P  and N. This   would
explain the   insulating properties  of P-rich   diamond, the lack  of
P-related sharp  luminescent bands, as well  as  the presence of broad
donor-acceptor optical bands.
We  have carried out spin-polarized,  ab initio calculations on
P-V  complexes using the  AIMPRO [17] code.   We used 86 atom
clusters, C 
 H 
 , having trigonal symmetry  from which two C
atoms  were removed and  a P atom  added.  This generates  a P-V pair.
The basis used  is similar to recent  investigations  on N-V  and Si-V
centers [13].   Previously [3],  we  have investigated
substitutional phosphorus  finding it to be  an on-site  defect with a
localized vibrational mode around 380 cm 
 .
In the first case the P atom was placed as far away from the vacancy and the surface as possible. The structure was relaxed in the neutral charge state. Then the P atom was moved to border the vacancy and the cluster re-relaxed. The energy in this case was lower by 1.5 eV than the case where P was separated from the vacancy. This shows that substitutional P is unstable in the presence of vacancies and readily forms the P-V complex.
In the P-V complex, the P atom did not  remain at its lattice site but
drifted  away until  it lay mid-way  between two  lattice sites.  This
structure   is   quite distinct  from   N-V  but  is similar    to the
split-vacancy Si-V complex [13].  Fig. 1  shows the  inner core
of the relaxed 86 atom cluster showing its  
  symmetry.  The six
P-C lengths  are 2.00 Å.    The  reason for  the stability   of  the
split-vacancy is partially connected with  the large atomic size of  P
although  Ge-V does not appear to   assume the same structure.  Fig. 2
shows the spin-polarized Kohn-Sham energy levels  of the P-V defect in
the vicinity of  the band gap, together with  those of a similar sized
cluster without any defects.  These  reveal a mid-gap  
 -level
(even) containing three electrons and an  
 -level just
above the valence band top, resulting in an effective spin of S=1/2.
The position of  this mid-gap level  is qualitatively similar to  that
found in the Si-V defect and thus the defect would be expected to trap
electrons arising from say substitutional N or indeed P.  Thus the P-V
defect acts as a deep acceptor.   However, the  
  levels
are then  filled and  the defect  must be diamagnetic  and no internal
optical transitions are possible.
When the Fermi-level is low, such as for  example when the material is
Irradiated or  the  absence of  substitutional N, then  a neutral  P-V
defect  would be formed   having  S=1/2.  The wave-functions of  the
 
  gap levels are combinations of  
  orbitals on the six C
atoms surrounding P and  possess very little  amplitude on P itself as
shown in Fig. 3.   The trigonal symmetry  of the  defect and  the fact
that the spin density is very low on P suggests that this defect might
be the EPR center investigated in Ref. [5].
Although no internal optical  transitions are possible for the ionized
defect, neutral pairs of  N 
  and (P-V) 
  defects  could luminesce
and be the source of  the broad band donor-acceptor recombination band
observed in Ref. [7].  The luminescence energy would then be:
 
 
With   the N donor   level  and  the  P-V  acceptor level  (Fig. 2) at
 
  eV and  
  eV respectively, then for r  about
5 Å,  this transition would be around  1.5 eV, close to the observed
value.
In conclusion, the calculations  suggest a strong  binding energy of P
with vacancies.  In  the P-V center, the  P atom lies  mid-way between
two    vacancies.  The  center     has   
  symmetry  with   spin
S=1/2   in the  neutral charge  state,    and there is  little
spin-density on  the P  atom consistent  with  EPR  experiments.   The
defect acts as  a deep acceptor and together  with N would explain the
pronounced broad PL bands seen around  1.7 eV in P-doped diamond.  The
defect could also  compensate remaining substitutional  P atoms.  This
would  explain  the  difficulties  in  making phosphorus  electrically
active in diamond.
 
  present address: Department of Physics, University of
Exeter, Exeter, EX4 4QL, UK.