Crystallization of Amorphous Chalcogenide Nano-Regions and Test-Structure Fabrication for Non-Volatile Memories

In this thesis we have studied the crystallization behavior of amorphous samples at different scales, pointing towards layouts similar to those ones occurring in PCM-based devices. In the first part we have clarified the relevance of the amorphous local structures produced by ion implantation and we have detailed the crystallization kinetics of GeTe thin film. Nucleation and growth phenomena have been observed by optical microscopy in a large region and corresponding rates and velocities have been found. From these data, the activation energies of each process have been calculated and compared with those obtained by a JMAK analysis of independent TRR measurements. The nucleation rate and growth velocity of the implanted films increased by a factor ten and three, respectively, with respect to the as deposited samples, suggesting the occurrence, during implantation, of a local atomic arrangement which enhances the crystallization kinetics. This evidence is in agreement with Raman spectroscopy data, suggesting that implantation, proving kinetics energy by collision cascades, produce a reduction of wrong bonds formed during sputter deposition. In the second part we have analyzed the crystallization of amorphous nanostructures of Ge2Sb2Te5 (GST) obtained by ion implantation, both isolated or embedded in its crystalline environment (fcc or hcp). Hcp Embedded nano amorphous dot (100nm in diameter) crystallization has been observed to start by the rearrangement of the a/c interface in a defective crystalline layer. At 90°C, this initial step proceeds with a fast growth velocity of 6.4 pm/s, about four times higher than that one measured in the subsequent steps. The slower rate is very similar to that one measured in the fcc embedded nanostructures. A similar behavior has been observed for annealing at 75°C, with an initial faster rate of 0.58 pm/s. We have also observed the behavior of the smaller nanostructures (20nm) that completely crystallize very fast, after only two hours at 75°C. Since the active amorphous region in most devices is surrounded by the hexagonal phase this result suggests that particular attention should be addressed to the data retention characteristics of sub-50nm GST based phase change memories. Isolated amorphous regions of different sizes and shapes has been observed after isochronal annealing from 90°C to 150°C. Crystallization has been observed to occur at 145 °C from the boundaries of the structures, while nucleation from the bulk of star-shaped regions takes place at 150°C. The higher crystallization temperature observed in this isolated structures suggests that similar systems could be used to improve data retention. Finally we have demonstrated a new procedure to fabricate maskless amorphous region using a controlled Ga+ beam at low energy (30 keV) and low fluencies (1E14 ions/cm2). This procedure allows to eliminate the presence of resist residuals during TEM observation. In the third part we have presented a simplified approach to fabricate and characterize a phase change test-structure memory with reduced active size. By fast pulses, as short as 300 ns, cells has been successful switched between SET and RESET states. Moreover, programming current has been found to be very low, in the 100 µA range. In most cases a cyclability of 100÷200 has been obtained. Although this result must be improved, a mechanism to recover the cell from a stuck reset fail has been shown, giving an important perspective about enhancing the endurance of the PCRAM. The cell fabricated permits direct morphological, structural and chemical analyses, because of the easy access to the active region of the device (that is under a unique passivation film, instead of several layers). Correlation of these analyses to the electric characteristics of the test-structures (and their variations) could represent a key point in understanding the behavior of PCRAM