Comparison of Spintronics and Nanoelectronics for Information Processing

Kang L. Wang, Igor V. Ovchinnikov, Alex Khitun and Ming Bao Device Research Laboratory (DRL) University of California, Los Angeles Los Angeles, CA, 90095-1594, USA wang@ee.ucla.edu

Abstract
To date, electronics uses electron charge as a state variable which is often represented as voltage or current. In this representation of state variable in today's electronics, carriers in electronics devices work independently even to a few and single electron cases. As the scaling continues to reduce the feature size, power dissipation and variability become two major challenges among others as identified in ITRS. This paper presents the exposition that spintronics as a collective effect may be favorably used as state variables in the near future information processing beyond conventional electronics for room temperature. An example is presented to compare electronics and spintronics in terms of variability, quantum and thermal fluctuations. This example shows the benefit of scaling to smaller sizes in the case of spintronics (nanomagnetics), which will have a much reduced variability problem as compared with today's electronics. Finally, spin wave bus is used to illustrate the potential use as a state variable for logic application. Prototype logic devices using the spin wave bus concept have been demonstrated. The requirements and benchmarks for choosing a state variable are also discussed in terms of its interaction strength for the energy efficiency.

where n is the number of electronics and is the reliability factor, or reciprocal of the probability of failure. The variability comes from the result of independent electrons in the device; the scaling to small feature sizes makes the quantum fluctuation of electrons important. However, in spintronics, there are two scenarios: one is single spin case and the other the collection of interaction spins (with exchange interaction, which is the consequence of many-electron effect of Coulomb interaction). For the former, it will be similar to that of electronics. In this paper, we will focus on the case of correlated electron systems, e.g., ferromagnetism. We will show that there are advantages in energy and in variability when using correlated electrons, in particular, in the case that the correlated energy is higher than kT for room temperature applications.
It may be this collective variable, which may become the Information carriers or the representation of the Information for beyond independent electron devices (CMOS) in the near future. From the point of view of the concept of Information, the physical representationthe collective variable is what it matters. From the information processing point of view, however, a minimum set is needed for the unique identification of the information processing device, and this minimum set includes: the state variable, outer interaction and inner interaction. The Outer interaction is the interaction, with which different units of the computational device interact among each other and with which the computational device is connected with the outside world including input and output. The Inner interaction is the interaction, which is responsible for the inner physics of the state variable itself. There are three technological limits (connected to each entity to the three elements), which put constraints for us in making a choice among different information processing bases. (i) The first limit is in the sensitivity of our devices (outer Interaction). We, humans, live and think on a macroscopic level. Even if we gethold of a sub-nuclei-size computer, somehow, which works perfectly from information processing point of view, we will never know what this computer

1. Spintronics for Information Processing

There has been increasing interest in considering spintronics as an alternate to today scaled CMOS for next generations of information processing beyond today's electronics. In the past, scaled CMOS, electron charge and its representations as e.g., voltage and current are used as a state variable for device to perform logic functions. Upon further scaling to the nanometer scale, the use of electron charge and its long-range growing strong Coulomb interactions resulted in two major problems: power dissipation per unit area and variability of the device. The first comes from the strong Coulomb interaction. It is shown that the minimal energy of a switch is kT In for n electrons, the energy required will be

r; nkT In r if

electrons work independently,

978-1-4244-2186-2/08/$25.00 2008 IEEE

calculates until we perform a read function; nor will we be able to ask this computer to calculate something for us unless we provide input. Thus, there is a problem of bringing the microscopically represented information to the interface that we can comprehend and thus control. The stronger the outer interaction of the State Variable with our every-day world, the easier it is for us to control the information processing and read-out. The first information processing device was mechanical one. (For those mechanical devices one had to spin a lever to multiply numbers.) The lever of the mechanical multiplying-numbers machine was one of the easiest use perform control or computing. With the advent of semiconductor industry more than a half a century ago, the information processing changed the physical basis from the mechanical one to the semiconductor electronics (ii) The second limit is the heat dissipation or power dissipation (in Inner and outer Interactions). The inner Interaction, unlike Outer Interaction, is the Interaction responsible for the physical state of an (information) State Variable. It also describes how much energy goes in and out of a State Variable for controlling the process of computing. The stronger the inner interaction is, the more power will probably be dissipated into the environment during computing. Therefore, the weaker the inner interaction, the better it is from the heating problem point of view. The outer interaction is the interaction of the State Variable with the input and the output. This will also give rise to power dissipation. Consequently, how well one can take the heating energy away from an information processing device is another technology limit. (iii) The third limit is in the size of the State Variable that we can manufacture and operate at. We want to store and operate as many operations as possible, so that the smallest-size of the state variable is preferred for increasing the information density and information throughout. At the same time, we have to be crafty enough to assess among different small-size information units so that the fluctuations will not be the problem for room temperature operation. This is another technological issue. Using these three limitations as the benchmarks, we can compare the State Variable-Interactions and choose the candidate for information processing. Today's electronics is ideal for accommodating limits (i) and (iii). The electronic (semiconductor) circuits are well developed by now and the charge state variable can be scaled down to nanometers. The form of the Coulomb interaction (Ifr) produces a common bound for the limits

(ii) and (iii). The attempts to satisfy each requirement (controlled by the Inner Interaction and the Size) become controversial. In this case, our objective will be to minimize the canonical numbers as the result of the limits (ii) and (iii), which is defined as (voltage per charge i.e. capacitance) size. The bound from the limits of (ii) and (iii) together approaches a common fundamental limit. Magneto-electronic circuits: The information is represented by the collective spin degree of freedom as magnetic polarization, but the connection between different units of the computational circuit is realized just as in the conventional electronics (MTJ devices). The inner interaction (Exchange) unlike that of Coulomb is local. This means that the performance is not bounded by limits (ii) and (iii). Rather, it is bounded by limits (i) and (iii). That is, the sensibility as the result of interaction of spintronics devices does not allow the scaling of a collective spin state variable down to a single spin. The progress in lowering the limit (i) is technologically behind for spin. Thus magneto-electronic-circuits use Coulomb interaction as outer interaction can be only a temporary solution for (i). 3) Multi-Ferroics: Again, we see the use of multiferroic materials as a temporary solution in resolving limit (i) on the sensibility of spins, by turning to an electric force (via ferroelectric polarization this time).

- - - r -- - -' - - T8: - - - _.. - tlf .............."'""'-'J.....-.-.......""""-'-'...-................- . . . . . . . - - 10' Ie' 10' 10' 10' 10' ~o' 10'

Ni

Fig.l Domains of quantum fluctuations for three quantum numbers: N, S, and Sz. The curves, TN, TS and Tsz separate the areas of quantum (to the left and below) and thermal (to the right and above) behaviors of the corresponding quantum number. At 300 K, the fluctuations in magnetic quantum number S give the limiting number of 102 atoms vs. the electronics counterpart of 107 atoms, 4) Single Spin: Today's Spintronics using a single electron spin level is not realizable for room temperature operation because of limit (i) for spins as their

interactions are too weak. Thermal noise or fluctuation seems to overwhelm the interaction (Zeeman) energy for the single spin case. 5) Molecular shuttle- or rotor like device: Limit (i) also does not allow an easy observation and control of the state collective variable describing the electron configuration in a single molecule. Though the limitation (ii) and (iii) are well satisfied. From the above analysis, we see that there is a high probability that collective variables will be the new state variables for new generations of logic devices. Nevertheless, the control of the collective variables and interconnects between them at present must still be accomplished by the conventional electronics tools, even though it is possible in the future that some more appropriate methods (without implication of the Coulomb interaction) will be developed. 2. Variability of Electronics and Spintronics As we discussed before, in the continuing scaling of the feature size of semiconductor devices, there are two major issues for further scaling: power dissipation and variability [1]. In search of low power and variability devices, spintronics [2] has emerged as a potential candidate in addition to others. In this section, we address the variability issue of nanoscaled electronic and spintronics devices. There are different sources of variability due to thermal and quantum fluctuations, resulted from fluctuations of the number of electrons, for example, and process control such as pattern size and dopant variations in semiconductor FET. We will, however, in this paper look at the fundamental of quantum fluctuations [3]. In order to assess electronic and spin based circuits for their variability, we address the problems of the fluctuations in the elemental units - FET and without loss of generality Spin-FETe The input-output fluctuations in a generic transistor can be understood from rather general thermodynamics reasoning. When the retardation (delay) effects are neglected, the output variable can be assumed to be in an "on-time" correspondence with the state variable of the gate (the charge Q or the magnetization M), so that Vout = f(Q). In tum, the state variable is controlled by the input parameter (Vin or Sin), which is effectively its Legendre conjugate in a thermodynamic potential (e.g. the free energy F), governing the behavior of the system, which is thermally open to heat bathes). The effective couplings in the thermodynamical potential are QVin and SinM, Depending on which variable is fixed, the corresponding potential must be used. Thus, just as for any Legendre conjugate pair (e.g., pressure-volume, temperatureentropy, chemical potential-number of particles), fixing one of the variable (the input) we leave the other variable

(output) loose and let it fluctuate - due to thermal fluctuations at high temperatures and/or due to the quantum discreteness of the state variable's eigenvalues at low temperatures. In Fig. 1, we summarized the results of theoretical analysis showing the comparison of charge and spin based minimal feature sizes due to fluctuations, and illustrating the feature size on top and the number of atoms for a given devices using a spherical gate. The room temperature line is also shown. Clearly the exact geometry of the gate electrode will change the capacity and also affect the number. In addition, for realistic devices, the gate capacitance will also depend on the dielectric layer. But the number obtained should be still close to the same order of magnitude. These numbers can be plotted together with the limitation based on energy consideration discussed above. Different materials can be used and there will be different sizes as a result of using atomic sizes. The fact that the spin degree of freedom is more advantageous than the charge degree of freedom from the quantum fluctuations point of view has its origin in the locality of the interaction responsible for the magnetism (exchange interaction). On the energy scale the energy per "quantum unit" scales as llr for electronics (Coulomb repulsion), whereas it is size independent for spins (exchange). In the above discussion, we only limit ourselves to electron and spin based devices. For other cases, as in molecules and ferroelectric materials, there will, more or less, give similar geometric dependences of the scaling rules of correlated effects or many-body (collective) effects in nanoscale systems. 3. Logic Devices with Spin Wave Bus Finally, we would like to present as example of using collective phenomena- spin wave as a state variable for logic application. Spin wave is a collective oscillation of spins in an ordered spin lattice studied for decades in a variety of magnetic materials and nanostructures [6-10]. The near-neighbor spins in the lattice are connected via the exchange interaction. Potentially, it is possible to use ferromagnetic films as a conduit for spin wave propagation, referred to as "Spin Wave Bus" [11], where the information can be coded into a phase or the amplitude of the spin wave as a state variable. Spin wave, as a highly correlated system of spins, possesses quite As it was experimentally long coherence length. demonstrated in spin wave Mach-Zehnder-type interferometer, spin wave signals remain coherent after propagation of millimeters distance in yttrium iron garnet film [12]. In the permalloy films [8], the coherence length is of the order of tens of microns in room temperature. Earlier, spin-wave based logic circuit has been

experimentally demonstrated by M. Kostylev et al. [12]. The authors built a spin wave Mach-Zehnder interferometer, which consists of a spin wave generator, a splitter that divides the input spin wave pulses into two ferrite channels made of Yttrium-Iron-Garnet (YIG), two controllable phase shifters attached to the branches, and a mixer where the signals modified by the phase shifter. The phases of the propagating spin waves are controlled by the magnetic fields produced by electric currents in the conducting wires under the waveguides. Depending on the phase shift, the device can operate as NOT or XOR logic gate. The same group has also demonstrated exclusive-not-OR and not-AND gates based on a similar Mach-Zehnder-type spin-wave interferometer structure [14]. Another prototype spin wave logic device has been recently realized by our group [13]. In Figure 2 (a), we show the general view of the device. The core of the structure from the bottom to the top consists of a silicon
a)
Input -1 Output Input- 2

b)

g
Q)

$'
co +-'
C)

12 In-phase Out of phase 10 8 6 4 2

Frequency

. ....

=3GHz

(5

>

"5 c.. "5 0

receiver to detect the inductive voltage produced by two spin wave signals. The distance between the microstrips is 4f.lm. In Figure 2(b), it is shown the experimental data on the output inductive voltage measured at the central ACPS line at different values of the external magnetic field (excitation frequency 3GHz). The red and black curves depict the output power for the in-phase (L1~=O) and the out-of-phase (L1~= Jr) cases, respectively. The initial phase difference between two input signals is defined by the direction of the current flow in the excitation lines. The phase difference is zero if the direction of current flow is the same (clockwise or counter-clockwise wise) in both lines. If the directions in the excitation lines are such as one loop is clockwise and the other is counter-clockwise, the spin wave signals receive a 1t relative phase difference. These experimental data show a prominent (about 4 times) output power difference. All measurements are carried out at room temperature. 4. Summary In summary, we showed that collective effects may be used as state variables for scaled room temperature information processing. The collective state variables seem to have the benefits of lower variability and high efficiency in energy needed for performing computing. For room temperature operation, the collective effect state variables are the choice, in particular, the use of collections of spins as nanomagnets. For single charge electronics, it was shown that the quantum fluctuation will dictate and result in the variability limit due to the large Coulomb interaction energy. On the other hand, for single spin, thermal energy is much higher than the energy of single particle systems at room temperature, i.e., Zeeman splitting, rendering room temperature operation unlikely. We also introduced the spin wave bus concept to illustrate a potential of using collective spins for constructing logic devices and circuits. Acknowledgments The authors acknowledge the research support of FCRP Center on Functional Engineered Nano Architectonics - FENA and Western Institute of Nanoelectronics WIN under the support of Nanoelectronics Research Initiative (NRI). References: G. E. Moore, "Gramming more components [1] onto integrated circuits," Electronics, vol. 38, pp. 114-117, 1965. [2] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S. von Molnar, M. L. Roukes, A. Y. Chtchelkanova, and D. M. Treger, "Spintronics: a spin-based electronics vision for the future," Science, vol. 294,pp.1488-95,2001. [3] I. V. Ovchinnikov and K. L. Wang,

100 200 300 400 500 600 700 800

Magnetic Field (De)

Fig.2 (a) General view of the three-tenninal spin wave logic device. The core of the structure from the bottom to the top consists of a silicon substrate, a 100nm thick CoFe film, and a 300nm thick silicon dioxide layer. Three ACPS lines on the top are used as the two input (edge) and one output (in the middle) ports. (b) Experimental data. Output signal amplitude for two spin waves coming in-phase (red curve) and out of phase (black curve).

substrate, a C0 30Fe70 IOOnm thick film, and a 300nm thick silicon dioxide layer. There are three asymmetric coplanar strip (ACPS) transmission lines on the top of the structure. The edge ACPS lines are the transducers to excite spin waves, and the line in the center is the

"Variability of electronics and spintronics nanoscale devices," Applied Physics Letters, vol. 92, pp. 093503-1-3, 2008. [4] L. Michalak, C. M. Canali, and V. G. Benza, "Electron-magnon coupling and nonlinear tunneling transport in magnetic nanoparticles," Physical Review Letters, vol. 97, pp. 096804/1-4, 2006. [5] E. Abe, K. M. Itoh, 1. Isoya, and S. Yamasaki, "Electron-spin phase relaxation of phosphorus donors in nuclear-spin-enriched silicon," Physical Review B, vol. 70, Jul 2004. [6] T. 1. Silva, C. S. Lee, T. M. Crawford, and C. T. Rogers, "Inductive measurement of ultrafast magnetization dynamics in thin-film Permalloy," Journal ofApplied Physics, vol. 85, pp. 7849-62, 1999. [7] B. A. Kalinikos, N. G. Kovshikov, M. P. Kostylev, P. Kabos, and C. E. Patton, "Observation of the amplification of spin-wave envelope solitons in ferromagnetic films by parallel magnetic pumping," JETP Letters, vol. 66, pp. 371-5,1997. [8] M. Covington, T. M. Crawford, and G. J. Parker, "Time-resolved measurement of propagating spin waves in ferromagnetic thin films," Physical Review Letters, vol. 89, pp. 237202-1-4, 2002. [9] M. Bailleul, D. Olligs, C. Fermon, and S. Demokritov, "Spin waves propagation and confinement in conducting films at the micrometer scale," Europhysics Letters, vol. 56, pp. 741-7,2001. [10] M. Bailleul, R. Hollinger, and C. Fermon, "Microwave spectrum of square permalloy dots: quasisaturated state," Physical Review B (Condensed Matter and Materials Physics), vol. 73, pp. 104424-1-14, 2006. A. Khitun and K. Wang, "Nano scale [11 ] computational architectures with Spin Wave Bus," Superlattices & Microstructures, vol. 38, pp. 184-200, 2005. M. P. Kostylev, A. A. Serga, T. Schneider, B. [12] Leven, and B. Hillebrands, "Spin-wave logical gates," Applied Physics Letters, vol. 87, pp. 153501-1-3,2005. A. Khitun, B. M., Y. Wu, 1.-Y. Kim, A. Hong, [13] A. Jacob, K. Galatsis, and K. L. Wang, "Logic Devices with Spin Wave Buses - an Approach to Scalable Magneto-Electric Circuitry," Procedings of the Material Research Society (in press), vol. spring meeting, 2008, 2008. T. Schneider, A. A. Serga, B. Leven, B. [14] Hillebrands, R. L. Stamps, and M. P. Kostylev, "Realization of spin-wave logic gates," Appl. Phys. Lett., vol. 92, pp. 022505-3, 2008.