3D Nanomagnetic Logic
269
5. Dataflow directionality. The ANC is breaking field coupling symmetry, [23]. By
this, properties of magnetic islands are locally changed in order to achieve signal
directionality. As the ANC can be in the 10 nm-range when ultimately scaled, this
method is prospective for high-density integration.
To the best of our knowledge, all fundamental properties for digital computation
are fulfilled. But following the achievements in logic devices of the last decades
it is obvious, that many more requirements are necessary to be listed as potential
beyond CMOS candidate. Even worse, a single shortcoming might be sufficient
not to be selected. To be more specific, due to [35] further technological requirements are: (1) room temperature or higher temperature operation, (2) low sensitivity to parameters (e.g. fabrication variations), (3) operational reliability, (4) CMOS
architectural compatibility, (5) CMOS process compatibility and (6) comprehending
intrinsic/extrinsic parasitic and their interface to interconnect.
For those requirements more detailed benchmarking in future is needed, in order
to decide on pNML applicability in products. Our recent experimental investigations
project, that all the requirements (1–6, see above) claimed in [35] can be fulfilled,
but it is also clear, that they are hardly tackled by pure university research. A lot
of engineering is needed, to achieve a mature computing systems for adding further
functionality to CMOS circuits. At the bottom line, it becomes obvious that further physical needs, namely size (scalability), switching time (speed) and switching
energy decide on whether pNML is competitive or not [35].
3 Device and Circuit Modeling on Different Levels
of Abstraction
So far, the concept of 2D pNML was introduced and the experimental demonstration
of the fundamental logic devices was summarized. In order to benchmark device
performance and to extrapolate for larger circuits and systems, simulations on different levels of abstraction are needed. For that, a three step approach was found
to be well suited. First, micro-magnetic finite difference simulations provide insight
in the physics of magnetization reversal dynamics and domain-wall propagation of
single pNML islands and devices. Second, together with magneto-optically characterized test-structures, switching distributions of magnets are extracted and simplified
behavioral models are formulated. Third, the parameterized compact-models form
the basis of Verilog-A simulations, utilizing the well-known strength of abstraction
in system level simulation.
269
5. Dataflow directionality. The ANC is breaking field coupling symmetry, [23]. By
this, properties of magnetic islands are locally changed in order to achieve signal
directionality. As the ANC can be in the 10 nm-range when ultimately scaled, this
method is prospective for high-density integration.
To the best of our knowledge, all fundamental properties for digital computation
are fulfilled. But following the achievements in logic devices of the last decades
it is obvious, that many more requirements are necessary to be listed as potential
beyond CMOS candidate. Even worse, a single shortcoming might be sufficient
not to be selected. To be more specific, due to [35] further technological requirements are: (1) room temperature or higher temperature operation, (2) low sensitivity to parameters (e.g. fabrication variations), (3) operational reliability, (4) CMOS
architectural compatibility, (5) CMOS process compatibility and (6) comprehending
intrinsic/extrinsic parasitic and their interface to interconnect.
For those requirements more detailed benchmarking in future is needed, in order
to decide on pNML applicability in products. Our recent experimental investigations
project, that all the requirements (1–6, see above) claimed in [35] can be fulfilled,
but it is also clear, that they are hardly tackled by pure university research. A lot
of engineering is needed, to achieve a mature computing systems for adding further
functionality to CMOS circuits. At the bottom line, it becomes obvious that further physical needs, namely size (scalability), switching time (speed) and switching
energy decide on whether pNML is competitive or not [35].
3 Device and Circuit Modeling on Different Levels
of Abstraction
So far, the concept of 2D pNML was introduced and the experimental demonstration
of the fundamental logic devices was summarized. In order to benchmark device
performance and to extrapolate for larger circuits and systems, simulations on different levels of abstraction are needed. For that, a three step approach was found
to be well suited. First, micro-magnetic finite difference simulations provide insight
in the physics of magnetization reversal dynamics and domain-wall propagation of
single pNML islands and devices. Second, together with magneto-optically characterized test-structures, switching distributions of magnets are extracted and simplified
behavioral models are formulated. Third, the parameterized compact-models form
the basis of Verilog-A simulations, utilizing the well-known strength of abstraction
in system level simulation.