the µ ring structure
Some of the PCM structures presented before (i.e. “Wall”, confined), have already been integrated in final industrial products. But even if their scaling have been demonstrated almost down to the 10 nm technology node, to achieve higher device density and lower power consumption, new concepts and new design approaches for the cell structure are still one of the major topic of investigation. In this light, we proposed and studied an original different solution, that can be suitable also for the analytical study of new phase-change materials: the µring structure.
The µring structure is similar to theµtrench regarding the first steps of the process flow, till the CMP process to leave the “ring” surface exposed. Then after:
- lithographically masking of part of the ring (made in TiN) surface;
- recess of the exposed ring by TiN etching;
- stripping of the residual resist mask;
- deposition by CVD of a dielectric (SiN) layer, with the main goal to isolate the recessed cavity;
- CMP process to remove superficially the dielectric, with the goal to expose the ring surface not recessed (µring);
- phase-change material deposition;
- top electrode deposition.
In Fig. 2.14 we report the description of the process flow, and a SEM top view image of the µring device after the ring masking, where the final dimensions of the plug (called µring because it is obtained by the reduction of the ring surface) are highlighted. In particular we defined with w and t respectively the width, the thickness of the plug, and with h the height of the plug, equivalent to the effective length of the recessed region.
In order to understand the main features of this structure, we performed 3D sim-ulations with FlexPDE tool, solving the heat diffusion equation and the continuity equation in the stationary case (the main equations and parameters are analyzed in section 2.7), but without the implementation of the crystallization mechanism. The main goal was to have a first idea of the impact of the scaling of this structure on the final programming performances. The phase-change material considered is the GST.
2.4 New PCM concepts to scale the dimensions: the µring structure 53
Fig. 2.14. Description of the process flow to fabricate the µring structure. On the right the SEM top view image of theµring cell, after the ring masking, where are highlighted the final dimensions of the plug (µring).
Fig. 2.15. The output of the 3D simulation of the temperature profile in the µring structure.
In Fig. 2.15 is reported an example of the simulations performed, showing the profile temperature achieved in the device during the RESET operation. As we can see we supposed a uniform recess of the ring.
The thermal isolation of the cell, as already observed before, is important to reduce the programming power. We simulated the device, considering two different dielectrics:
SiO2 (kth = 1.4 WK−1m−1) and SiN (kth = 22 WK−1m−1). The results on the current density to RESET the cell are reported in Fig. 2.16. Here the thickness and the height of the plug are kept constant, while the width of the plug is varied. First of all, we observe an increase of the current density, when the surface area scales down (as already observed before in other structures). At the same time, the use of a dielectric with higher thermal conductivity (i.e. the SiN), increases the final current density needed to achieve the same temperature variation in the phase-change material. Moreover, as the device area scales, the temperature peak reached in the phase-change material, approaches the plug/phase-change material interface (Fig. 2.17), allowing more efficient programming operations.
The optimization of the temperature control in the phase-change material, can be achieved increasing the µring height. In Fig. 2.18 we show that keeping constant the
Fig. 2.16. Simulated current density as a function of theµring surface, for two different dielectrics used for the device fabrication.
Fig. 2.17. Displacement of the peak of the temperature over the plug surface (inside the phase-change material) as a function of the plug surface resulted from the simulations.
Fig. 2.18. Simulated maximum temperature generated at the plug/phase-change material interface, as a function of the voltage applied on the cell, forw = 50 nm,t = 5 nm, and for different values of the ring recessed height h (from 5 nm up to 500 nm). The melting tem-perature of the GST is evidenced (dashed line).
Fig. 2.19. Power to melt the phase-change material in the µring device, as a function of the plug heighth. The plug thickness and the plug width are kept constant. The higher is the plug, the higher is the power dissipated in the device.
plug area (250 nm2), the increase ofh, enables a more gradual raise of the temperature in the device. In particular, thinking to a 3 V powered device, and considering as an example a threshold voltage of 1.2 V (like in GST based PCM), the choice ofhshould be of∼ 200 nm (in the graph reported with the blue curve). In fact, for this height of the plug, the temperature excursion from the ambient temperature (300 K) and the melting temperature of the considered phase-change material (dashed line), is perfectly covered from 0 V up to 3 V. If this solution can make the device “self-controlled”, without the use of an external current limitation, at the same time, the increase of the height of the plug, rises the final power dissipation in the cell. The latter, is confirmed in Fig. 2.19, where the plug height is varied, keeping constant the plug area,. The y-intercept calculated from the linear interpolation of the data (y-intercept = 0.45 mW), represents the real