Echoes of the past in silicon chips
Illustration of the silicon atom positions near the interface (the horizontal line) of crystalline silicon and amorphous thermal oxide (SiO2) for a crystal structure. The small dots in the thermal oxide (above the horizontal line) represent where the silicon atoms would be if the crystal structure had expanded without disordering. The outline of the four silicon unit cells is shown below the interface. The outline of four expanded lattice cells in the oxide is shown above the interface. (Graphic courtesy of SSRL)
Thermal oxide is the real on-off switch for your computer. The nanometers-thick film on the surface of silicon transistors helps turn on and off the flow of electricity through the transistor, providing the 0 and 1 binary signals modern electronics run on. There are several million transistors on each computer chip.As technology produces smaller chips that require thinner oxides, the ability of thermal oxide to act as the basis for integrated circuits is starting to break down.
"We're pushing the fundamental limits," said materials researcher Sean Brennan (ESRD). "Anything you can do to learn more about the thermal oxide is a huge plus."
Thermal oxide is 'grown' on the surface of silicon wafers by diffusing oxygen atoms into the silicon's crystal lattice. The oxygen atoms break silicon-silicon bonds and form silicon-oxygen bonds, in the process disrupting the perfectly repeated and regular crystal structure. This layer of oxidized silicon (SiO2) is thermal oxide, and was long believed to be completely amorphous, in other words, an unpredictable structure without long-range order.
New evidence from Brennan and former SSRL graduate student Anneli Munkholm, now at Lumileds Lighting, is overturning that assumption. Their research, recently published in Physical Review Letters, shows that thermal oxide holds "weak crystalline 'echoes' of the silicon's former self buried within the non-crystalline oxide," said Munkholm.
X-ray scattering at SSRL revealed that thermal oxide has faint memories of the former position of the silicon atoms. Each silicon chip is a single crystal, meaning it follows a regular three-dimensional pattern. Think of a three-dimensional chess board--all the black squares are in predictable locations (forward 1, over 1, up 1). In contrast, in thermal oxide the oxygen atoms randomly attaches to silicon in any direction, so the structure is not a crystal.
So Munkholm and Brennan were surprised to find their scattering patterns show the silicon atoms are relatively close to the positions they held before being disrupted by oxidation. The new model based on the data also shows that the crystal memory is stronger closer to the pure silicon (at the interface between the silicon and thermal oxide, where oxygen atoms are less dense), and fades closer to the surface of the thermal oxide. The silicon lattice also expands as oxygen diffuses in, but expands less at the interface.
"It is only through the use of an intense synchrotron x-ray beam from SPEAR that we were able to observe the residual order," Munkholm said.
They found residual order in a wide range of oxidation recipes with oxide thicknesses from 6 nanometers (nm) up to 100 nm and on silicon with different surface orientations (the crystal sliced at different angles).
"We have seen evidence that different recipes result in different amounts of disorder," Brennan said.
This suggests researchers may be able to relate certain amounts of order with specific electrical properties in thermal oxide that could be better for running new integrated circuits.
For more information, see: http://www-ssrl.slac.stanford.edu/research/highlights_archive/oxidizedsi.html