TOKYO, Japan, January 5, 2011 — Renesas Electronics Corporation (TSE: 6723), a premier provider of advanced semiconductor solutions, today announced its successful development of a new analysis method for a time-dependent variability phenomenon, called random telegraph noise (RTN, Note 1). The method can be used to analyze statistical behavior of traps (Note 1) in detail, to realize predictive simulation necessary for designing highly reliable next-generation system LSIs.
RTN is a phenomenon in which the current flowing in a transistor changes discontinuously with time due to the capture or emission of a single elementary charge in a trap. Since the influence of this phenomenon increases as the size of a transistor decreases, it has been pointed out that RTN may possibly cause malfunction of system LSIs in the near future. It is determined that RTN cannot be completely eliminated simply by improving the fabrication processes, and therefore it is necessary to design system LSIs that can prevent malfunction even if the RTN phenomenon occurs. Since the characteristics and number of the traps are determined randomly, RTN is a type of random variation phenomenon, and thus statistical methods are indispensable for its analysis.
Renesas Electronics already developed a simulation method that can quickly determine whether malfunction resulting from RTN will occur or not within a product's lifetime. Furthermore, Renesas Electronics has explicated the statistical distributions of the number of traps and the amount of current change (amplitude) caused by each trap, which are necessary information for performing this simulation. However, the information on the statistical distribution of the trap time constants (Note 2) was still lacking. Traps with a small time constant frequently capture or emit a charge, whereas traps with a large time constant rarely capture or emit a charge. Without the information of the time constants, simulations must be performed by arbitrarily assuming some of the parameters, which may degrade the precision of the predictions.
The variations in the time constants arise from the variations in the energy level and position (the distance from the surface of the silicon substrate) of each trap. A measurement method for extracting the energy level and position of each trap was already known by the experts in the relevant fields. However, since the procedure is complex, only results for one or a few traps are published, and no statistical information with respect to the time constants has been available until now. Renesas Electronics has now created a data processing program that can perform this procedure automatically, and has extracted the statistical distributions of the energy levels and positions by performing this operation efficiently for a large number of transistors.
Key features of the newly-developed technologies:
(1) Transistors that are appropriate for analysis (transistor that has only a limited number of traps) are selected and their current waveform is measured at multiple gate voltage conditions, to determine the energy levels, amplitudes, and trap positions.
(2) The above procedure is performed automatically for multiple transistors, and the distributions and correlations of the energy levels, amplitudes of the current variations, and trap positions are determined.
As a result of applying the new method, it became clear that the trap energy levels (and thus time constants) spans over a wide range of several orders. Furthermore, Renesas Electronics also confirmed that the time constants vary independently of the trap positions. Previously, there had been different theories about the relationship between the time constants and the trap positions. One theory stated that a trap closer to the surface of the silicon substrate causes faster switching RTN (the time constants become smaller), since the speed is dependant on the distance of the trap from the surface On the other hand, another theory stated that the time constants show no relationship with the position as the speed is dependant on the chemical nature of the traps. According to the first theory, the current fluctuation amplitude should be correlated with the time constants, while no correlation is expected according to the second. The data obtained by Renesas Electronics support the second theory, and confirmed that the time constants and current variation amplitudes vary independently.
Renesas Electronics' newly developed methods can achieve reliability predictions with even higher accuracy than before by incorporating the information explicated in this work into existing simulations, thereby making it possible to realize even more optimized circuit design that assure high reliability with respect to RTN.
To continue providing products that achieve both high performance and high quality, Renesas Electronics intends to engage in research and development related to variability, including RTN and other phenomena.
Renesas Electronics presented the results of this research at the International Electron Devices Meeting 2010 (IEDM 2010), held from December 6 through December 8, in San Francisco, U.S.
(Note 1) RTN (random telegraph noise), trap
Traps are microscopic defects in the gate dielectric layer that capture and emit a single charge. Traps alternate between two states where a charge is captured (the trap is filled) or emitted (the trap is empty), and as a result, the transistor conductivity varies in time between two discrete values. This phenomenon is called RTN. The number of traps in a transistor, the amplitude of the change in the transistor current between the captured and emitted states for a specific trap, and the interval at which each trap changes the states (the time constants) are all determined by chance. Therefore the RTN behavior varies in a statistical manner.
(Note 2) Time constant
The time that a trap stays in the captured state, or in the emitted state, is determined probabilistically and changes every time; the average value of the stay time is called the time constant. There are two time constants: one for the captured state, and the other for the emitted state. The smaller the time constant is, the higher is the probability that the state will switch to the other state within a certain time.
(Remarks)
All registered trademarks or trademarks are the property of their respective owners.
About Renesas Electronics Corporation
Renesas Electronics Corporation (TSE: 6723) delivers trusted embedded design innovation with complete semiconductor solutions that enable billions of connected, intelligent devices to enhance the way people work and live. A global leader in microcontrollers, analog, power and SoC products, Renesas provides comprehensive solutions for a broad range of automotive, industrial, infrastructure, and IoT applications that help shape a limitless future. Learn more at renesas.com. Follow us on LinkedIn, Facebook, Twitter, and YouTube.
The content in the press release, including, but not limited to, product prices and specifications, is based on the information as of the date indicated on the document, but may be subject to change without prior notice.