Indium phosphide (InP) is a compound semiconductor widely used in the optoelectronic industry. The reason for that lies in the direct bandgap and in the lattice parameter which is suitable for the lattice-matched growth of ternary and quaternary III-V alloys. The active regions of various active (LEDs, Lasers) and passive (photodetectors, PIN diodes, modulators) devices are in fact made of epitaxial layers based on the In, Ga, As, P elements. These layers are deposited by metalorganic vapor phase epitaxy (MOVPE) or molecular beam epitaxy (MBE) on semiconducting n-type and p-type InP wafers. More recently, InP has received increasing attention as a good material for the fabrication of opto-electronic integrated circuits and high­ speed devices such as metal-insulator field effect transistors (MISFETs) wafers. For such applications, it is necessary to have semi-insulating (SI) substrates in order to isolate the different components and reduce the parasitic capacitances. Unfortunately, otherwise than gallium arsenide, undoped SI InP cannot be obtained as no native deep levels are present is sufficient concentration. The most common way to prepare SI InP is therefore based on doping with transition elements, in particular iron. Nowadays, commercial SI InP is prepared by adding pure iron to the starting charge, before growing the crystal from the melt. [1,2] The Fe atoms substitutional to In give rise to deep acceptor levels (0.64 eV from the conduction band) and compensate the residual shallow donors responsible for the n-type conductivity of as-grown undoped InP crystals. A concentration of Fe atoms > 1 x 1016 cm - 3 in the crystal is normally required in order to obtain SI materials. There are however a few drawbacks in the use of Fe doping: i) the distribution coefficient Kpe is much smaller than one (% 0 .0 0 1 ), which results in segregational problems and strongly non-uniform axial doping profiles, ii) a relatively large fraction of the iron incorporated is not electrically active, therefore the total Fe content is higher that the one strictly necessary to compensate the shallow levels, iii) iron can out-diffuse from substrates to epilayers and degrade the device performances; note that the higher the Fe concentration the heavier the diffusion phenomena; in particular it was found to accumulate at the substrate/epilayer interface [3], iv) it reduces the electric activation of the ionic implants [4], v) its presence in the channel of the FET’s greatly reduces the carrier lifetime and mobility [5]. For these reasons, some attempts have been made to produce semi-insulating undoped InP by annealing at 900-920°C, under a pressure of phosphorus [6,7] or vacuum [8 ], very pure InP wafers (with a residual carrier concentration n < 4 x 1015 cm-3 ). The dramatic resistivity increase was explained, in the case of the annealing under P pressure, by taking into account the Fe contamination and outdiffiision/annihilation of intrinsic shallow donors [9,10]; while, for the vacuum annealing, both Fe contamination and generation of new shallow

acceptors were considered [11,12]. There is, however, a general agreement in recognising that the reproducibility of the anneal-related conversion of undoped InP is rather poor, as it depends on an uncontrolled entity such as Fe contamination. Hence, attention has been turned out to the preparation of SI InP with a low, but well defined, Fe content. Two approaches have been followed: 1) annealing of lightly Fe pre-doped wafers and; ii) annealing of undoped InP in the presence of a Fe source [13].