Thursday, November 25, 2010

Fifty Year Electronics Speed Barrier Overcome

They have figured out how to speed up metal – insulator – metal diodes.  It is a case of picking super smooth metals.

There is no indication here over how much better the speed will be, but I would not be surprised if we are talking about several orders of magnitude.  Been related to smoothness suggests far better geometric control than ever before.

Moore’s law continues to rule.  Understanding that law made the present world apparent to those informed a full forty years ago.

Revolutionary diode design cracks 50 year-old electronics speed barrier
22:57 November 8, 2010

Metal-insulator-metal (MIM) diodes might just be the technology that allows electronics achieve the next big leap in processing speed. Research into diode design conducted at the Oregon State University (OSU) has revealed this week cheaper and easier to manufacture MIM diodes that will also eliminate speed restrictions of electronic circuits that have baffled materials researchers since the 1960's.

“Researchers have been trying to do this for decades, until now without success. It’s a basic way to eliminate the current speed limitations of electrons that have to move through materials,” OSU materials chemistry professor, Douglas Keszler said. “This is a fundamental change in the way you could produce electronic products at high speed on a huge scale at very low cost, even less than with conventional methods,” he said.

“For a long time, everyone has wanted something that takes us beyond silicon. This could be a way to simply print electronics on a huge size scale even less expensively than we can now. And when the products begin to emerge the increase in speed of operation could be enormous,” Keszler said.

Traditional silicon-based materials used in electronics work by limiting the flow of electrons using transistors, a process that restricts how quickly electrons can move across a circuit – and therefore how quickly your computer or iPhone can load a program. MIM diodes allow almost instantaneous electron transfer through the insulator surface – a major step in eliminating the 50 year-old speed barrier found in transistor based electronics.
The Breakthrough
While MIM diodes have been around for a while now, the biggest problem has been controlling electron flow to make it even across the entire surface of the diode. The breakthrough was to use a super smooth metal ZrCuAlNi in thin film form as opposed Aluminium (Al) which is comparatively rough. The roughness of the Al electrodes used in previous MIM diode research had resulted in poor consistency in electron flow. Now with the flatter ZrCuAlNi electrodes this flow of electrons can be controlled much more easily. So the diode is constructed with two ZrCuAlNi electrodes with a SiO2 insulator layer in between. A patent has been applied for on this new technology.

“When they first started to develop more sophisticated materials for the display industry, they knew this type of MIM diode was what they needed, but they couldn’t make it work,” Keszler said. “Now we can, and it could probably be used with a range of metals that are inexpensive and easily available, like copper, nickel or aluminum. It’s also much simpler, less costly and easier to fabricate.”

University scientists will initially apply the work to innovations in electronic displays, but they say many applications are possible. High speed computers and electronics that aren't limited by transistors are possibilities.

The Paper Advancing MIM Electronics: Amorphous Metal Electrodes is published inAdvanced Materials.

Advancing MIM Electronics: Amorphous Metal Electrodes

1.                  E. William Cowell III1
2.                  Nasir Alimardani1,
3.                  Christopher C. Knutson2
4.                  John F. Conley Jr.1
5.                  Douglas A. Keszler2
6.                  Brady J. Gibbons3
7.                  John F. Wager1,*

Article first published online: 25 OCT 2010
DOI: 10.1002/adma.201002678

Copyright © 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Effectively controlling quantum mechanical tunneling through an ultrathin dielectric represents a fundamental materials challenge in the quest for high-performance metal-insulator-metal (MIM) diodes. Such diodes are the basis for alternative approaches to conventional thin-film transistor technologies for large-area information displays,12 various types of hot electron transistors,2–6 ultrahigh speed discrete or antenna-coupled detectors,7–14 and optical rectennas.15 MIM diodes have been fabricated by anodization,1 thermal oxidation,8–1114plasma oxidation,101213 or plasma nitridation16 of crystalline metal films. Diodes fabricated using these approaches have invariably exhibited poor yield and performance. These problems are to a large extent a consequence of the roughness of the surface of the crystalline metal film, which is often larger than the thickness of the MIM insulator. As a result, the electric field across a MIM device will be highly nonuniform, making the control of quantum mechanical tunneling problematic. In this contribution, we describe the use of an amorphous metal contact as a critical component for circumventing the surface roughness and field uniformity roadblocks that have precluded the realization and utility of MIM electronics for applications requiring high device current rectification ratios (e.g. display applications).

The MIM diode is the fundamental building block of metal-insulator electronics. The device is characterized by a high degree of nonlinearity in its current-voltage characteristics as a result of a large difference in conductivity between on and off states. The operational theory of this diode, based on Fowler-Nordheim tunneling, has been described in detail by Simmons.1718 The probability of quantum mechanical tunneling depends exponentially on the thickness of the insulator between a pair of metal electrodes. Hence, the performance of the diode is critically dependent on the thickness uniformity of the tunnel-dielectric layer across the entire device. Interfacial roughness and dielectric imperfections give rise to alternate conduction mechanisms, e.g. Frenkel Poole emission, that can dominate at low voltages and reduce the device rectification ratio. The inability to create and effectively control a uniform electric field across the whole device area has been the primary limitation in producing reliable MIM devices. Here, we demonstrate that the necessary field control can readily be achieved by integrating the atomically smooth-surface of an amorphous metal electrode with high-quality insulators. This combination provides a rich materials and processing palette for development of MIM electronics, enabling new strategies for device design and fabrication.

Amorphous metals have been primarily investigated as bulk materials, addressing diverse applications that range from micromachines and hinges for digital light processors to golf clubs and transformer cores.19–21 They have also been deposited in thin-film form, primarily for study and development of micro-electromechanical systems.22–27 Relative to crystalline metals, they are more electrically resistive by approximately two orders of magnitude.2829 While this resistivity limits their use as interconnects, it is not an impediment to their use as electrodes. To date, however, there have been no reports involving the use of amorphous metal films as electrodes in electronic devices.

Our investigation of amorphous metal films as electrode materials was stimulated by the smooth surfaces reported for the amorphous metal ZrCuAlNi in thin-film form.2526 We hypothesized that the availability of such a smooth surface would provide the basis for a flat metal-insulator interface, producing an MIM device with a uniform and homogeneous electric field over large areas. Initial characterization of a ZrCuAlNi film deposited via DC sputtering revealed an rms roughness of 0.2 nm with maximum excursions to 1.7 nm in an area of 5 μm × 5 μm (Figure1a). For comparison, the rms roughness of a thermally-evaporated crystalline Al film is 5 nm with excursions as high as 70 nm (Figure 1b); this degree of surface roughness is not surprising given the tendency for elemental metals to crystallize, even when deposited at room temperature.30 Clearly, the ZrCuAlNi amorphous metal film exhibits a much smoother surface than that of a typical crystalline metal. The initial MIM diodes fabricated for electrical characterization are schematically represented in Figure 1c. The devices are built from a blanket coat of 200 nm of a ZrCuAlNi amorphous metal film electrode (M1) on SiO2, an approximately 10-nm thin film of Al2O3 produced by atomic-layer deposition (ALD), and upper electrodes (M2) as arrays of 1-mm diameter dots of ZrCuAlNi or Al metal. As seen in TEM images (Figure 1d and e), the selected materials and processes afford uniform films and well-defined metal/insulator interfaces. Note from Figure 1d and e that a native oxide 1.5 nm thick is present at the M1/insulator interfaces for both MIM diodes. X-ray photoelectron spectroscopy assessment of this native oxide indicates it to be predominantly ZrO2.31

Figure 1. a) An AFM image of the surface of a 100-nm ZrCuAlNi amorphous metal film. b) AFM sectional analysis of the surface of a 150-nm Al blanket. c) Schematic representation of an MIM diode. d) TEM image of an MIM diode; M1 = ZrCuAlNi, I = Al2O3, M2 = Al. e) TEM image of an MIM diode; M1 = ZrCuAlNi, I = Al2O3, M2 = ZrCuAlNi.

The MIM diode, constructed using ZrCuAlNi amorphous metal films for both M1 and M2 electrodes, is a symmetric device; the associated equilibrium energy-band diagram is shown in Figure2b. This diode has equivalent barriers (φb1 = φb2) at the metal/insulator interfaces, giving rise to a symmetric current density-applied electric field (J–ξ) curve (Figure 2c). Symmetric MIM diodes with ZrCuAlNi amorphous metal electrodes have been fabricated with yields >70%. (In contrast, we were unable to fabricate even a single working MIM diode when using 10 nm Al2O3 and symmetric evaporated Al electrodes, even though we employed the same high-quality, conformal ALD Al2O3dielectric.) In the amorphous-metal diodes, negligible current flow is observed until the magnitude of the electric field exceeds 3.5 MV/cm. These devices are quite stable as repeated voltage sweeps reveal no hysteresis. This performance is consistent with the high quality of the individual films and the resulting uniform electric fields. These characteristics are particularly noteworthy, considering the relatively large areas of the individual devices (1 mm2).

Figure 2. a) TEM image of a symmetric MIM diode fabricated with M1 = ZrCuAlNi amorphous metal film, I = 10 nm Al2O3, M2 = ZrCuAlNi amorphous metal film. b) Equilibrium energy-band diagram for a symmetric MIM diode. c) Measured and simulated current density-field (J–ξ) curves for the MIM diode shown in (a). d) TEM image of an asymmetric MIM diode fabricated with M1 = ZrCuAlNi amorphous metal film, I = 10 nm Al2O3, M2 = Al. e) Equilibrium energy band diagram for an asymmetric MIM diode. f) Measured and simulated J–ξ curves for the MIM diode shown in (d). A positive electric field corresponds to application of a positive voltage to electrode M2.

In contrast, the MIM diode with ZrCuAlNi as the lower electrode (M1) and Al as the upper electrode (M2) possesses asymmetric barriers (φb1> φb2; Figure 2e), leading to the asymmetric J–ξ curve shown in Figure 2f. As seen in Figure 2f, tunneling currents on the order of microamps occur at fields above 4 MV/cm for negative polarity and above 5 MV/cm for positive polarity, confirming the expected asymmetric behavior. The barrier height asymmetry for this device is approximately 0.8 eV, since the work functions of the ZrCuAlNi and Al metal films are measured via Kelvin probe to be 4.8 and 4.0 eV, respectively. These asymmetric MIM diodes possess desirable yield and performance properties similar to those of the symmetric MIM devices discussed previously, even though the upper Al electrode is crystalline with a normally rough surface (Figure 1b). The roughness of this surface is not relevant, as the insulator/Al interface is smooth due to the planarity of the ALD Al2O3. Thus, it appears that the key to obtaining a high-performance MIM diode is to insure that the initial metal surface is ultrasmooth. The TEM image (Figure 2d) reveals sharp and uniform interfaces.

Measured J–ξ curves for symmetric and asymmetric MIM tunnel diodes, as shown in Figure 2c and 2f, can be accurately simulated by using the theory of Simmons,1718 confirming that Fowler-Nordheim tunneling indeed dominates at both positive and negative fields. The agreement between measured and simulated J–ξ curves for the asymmetrical MIM tunnel diode (Figure 2f) is particularly good. The simulated symmetrical MIM diode J–ξ curve (Figure 2c) is less satisfying, since a noticeable deviation occurs for the positive polarity, corresponding to additional or greater than predicted electron tunnel injection from the top Al2O3/ZrCuAlNi interface. We attribute this discrepancy to the use of DC magnetron sputtering for deposition of the upper ZrCuAlNi electrode. In this process, depositing species have energies on the order of 2–7 eV,30 which appears to be sufficient to slightly thin and roughen the tunnel insulator (see Figure 2a). In contrast, thermal evaporation, a low-energy process that imparts only thermal energy to the Al2O3 tunnel-barrier surface, is employed in the construction of the upper Al electrode in the asymmetric MIM tunnel diode. Here, a much smaller positive polarity difference between measured and simulated J–ξ curves (Figure 2f) is witnessed. The associated TEM image shows the insulator/M2 interface to be pristine (Figure 2d).

Current–voltage (I–V) curves for a series of asymmetric ZrCuAlNi–Al2O3–Al MIM diodes with variable tunnel-barrier thicknesses are presented in Figure3. The insulator thickness is specified in terms of ALD pulse cycles, i.e., alternating purge-separated pulses of trimethlyaluminum and deionized water. As a point of reference, 112 pulse cycles has been measured to correspond to an Al2O3thickness of approximately 10 nm. At a small number of pulse cycles, the thickness of the ALD-deposited Al2O3 is difficult to measure. The pulse cycles/thickness relationship becomes less linear as the number of pulse cycles is reduced below 30. In addition, the thickness of the native oxide present on the amorphous metal electrode for the ultrathin deposited layers is an appreciable fraction of the total MIM insulator thickness.

Figure 3. Current-voltage (I–V) curves of asymmetric MIM diodes fabricated with M1 = ZrCuAlNi amorphous metal film, I = Al2O3, and M2 = Al. a) I–Vcharacteristics for MIM diodes with ultrathin tunnel barriers (< 2 nm). b) I–Vcurves for 30, 56, 112 ALD pulses of Al2O3. Asymmetry parameters for devices are given in parentheses. A positive voltage corresponds to application of a positive voltage to electrode M2.

MIM diode applications involving rectification or detection require I–V curves to be asymmetric, if they are to be operated without an offset bias.1115 As shown in Figure 3a, diodes with ultrathin Al2O3 insulators exhibit nonlinear behavior but very little I–V curve asymmetry. (For comparison, I–V asymmetry is quantitatively defined as the ratio of the positive-to-negative polarity current at an electric-field magnitude of 4 MV/cm. The asymmetry evaluated in this manner is specified in parentheses in Figure 3.) These I–V curves are nearly symmetric, because their insulators are so thin that current from direct tunneling across the insulator dominates at very low applied voltages. (Direct tunneling occurs through a trapezoidal barrier, in contrast to Fowler-Nordheim tunneling through a triangular barrier.) I–Vasymmetry is expected to be pronounced only if the insulator is thick enough to stand off significant direct tunneling current up to the onset of Fowler-Nordheim tunneling. Figure 3b shows that this changeover occurs between 30 and 56 pulse cycles, where the asymmetry abruptly jumps from 7.4 to 499. Direct tunneling occurs at low voltages for all diodes, regardless of the insulator thickness, and it determines the magnitude of the zero bias resistance. For MIM tunnel diodes fabricated using 112 and 12 ALD pulse cycles, the zero bias resistance is > 1011 Ω and 20 kΩ, respectively.

To extend the functionality of amorphous metal films and to simplify MIM diode fabrication, two approaches were explored, leveraging the direct oxidation of the amorphous metal electrode and the ease of solution processing. First, a blanket amorphous metal electrode was annealed in air to thicken the native Zr(IV) oxide (at 300 °C, a temperature that does not promote crystallization of the amorphous metal). After the anneal, Al electrodes were thermally deposited to complete the structure. A representative I–V curve for these devices is presented in Figure4. The I–V asymmetry of the resulting structure is measured to be 1.2, which is less than that expected from consideration of the differences in electrode work functions. The origin of this discrepancy is under investigation. A second oxidation-based asymmetric MIM diode was fabricated via the incorporation of a solution-deposited aluminum phosphate (AlPO) film32 as a portion of the tunnel barrier. An MIM structure was realized by spin coating the AlPO film onto the ZrCuAlNi electrode, annealing in air at 300 °C, and then depositing an Al upper electrode. The device exhibits the onset of significant tunneling currents (Figure 4) at voltages higher than those of the devices employing Al2O3 films produced via ALD. The asymmetry is 0.0015, indicating that the polarity dependence of the current has switched from that seen for the asymmetric diodes fabricated with the Al2O3 tunnel barriers. We have recently reported significant interdiffusion between the interfaces of ZrCuAlNi and AlPO bilayers in nanolaminated structures.31 The presence of the AlPO layer and its interaction with the native ZrO2, formed during annealing, creates an opportunity for controlling the I–V polarity dependence of the MIM diode.

Figure 4. I–V curves for MIM diodes with M1 = ZrCuAlNi, M2 = Al, and I = ZrO2or AlPO + ZrO2. The ZrO2 diode was annealed in air to form the tunnel dielectric. The AlPO + ZrO2 diode had 10 nm of AlPO deposited onto the surface of the ZrCuAlNi lower electrode and was subsequently annealed in air. Asymmetry parameters for devices are given in parentheses. A positive voltage corresponds to application of a positive voltage to electrode M2.

In summary, high-performance, high rectification ratio MIM diodes employing amorphous metal electrodes have been demonstrated. ZrCuAlNi amorphous metal film lower electrodes have been coupled with high-quality insulators and ZrCuAlNi and Al upper electrodes to produce uniform electric fields for the successful operation of both symmetric and asymmetric diodes. All diodes were fabricated by using relatively low temperatures (≤300 °C), rapid DC sputtering of the amorphous metal, and ALD or solution processing for deposition of high-quality insulators. Together, these methods provide opportunities for device fabrication on a variety of substrates, extending to large areas. In addition, the exceptionally broad compositional space of amorphous metal films provides unique opportunities to modify work functions and tune barrier heights for control of electron tunneling and device operation. Hence, this approach to MIM electronics presents an intriguing new means both for designing very high-performance electronic devices and integrating them across multiple technology platforms.

Experimental Section

Thin Films: ZrCuAlNi amorphous metal thin films were deposited onto Si/SiO2 (100 nm SiO2) substrates using DC magnetron sputtering with no intentional substrate heating at a power of 60 W, a pressure of 3 mTorr, and an Ar flowrate of 20 sccm. A 3-inch diameter, 0.25-inch thick vacuum arc-melted metal target (with an atomic composition Zr40Cu35Al15Ni10) fabricated by Kamis Inc. was used for all ZrCuAlNi depositions.

Atomic layer deposition (ALD) of Al2O3 was carried out in a Picosun SUNALE R-150B ALD reactor using trimethlyaluminum (TMA) and de-ionized water at a temperature of 300 °C. The pulse durations for both TMA and water were 0.1 s with a 2-s nitrogen purge between pulses. MIM diode structures were completed by shadow masking 1 mm diameter top contacts deposited by either thermal evaporation of Al or DC magnetron sputter deposition of ZrCuAlNi films using the deposition parameters described above.

The solution-based dielectric depositions were carried out using an aluminum oxide phosphate solution containing a 0.10 M metal concentration at an aluminum-to-phosphorus ratio of 30:18.32 The solution was deposited via spin-coating for 30 s at a speed of 3000 rpm, followed by rapid heating at 300 °C for 1 min on a hotplate in air. Films were annealed in air at 300 °C.

Atomic force microscope (AFM) measurements were made using a Digital Instruments 3 instrument with silicon-nitride tips; images were acquired over 5 μm × 5 μm areas. Transmission electron microscopy (TEM) images were obtained with a JEOL 2500 TEM from samples prepared with a Dual Beam FEI 235 focused ion beam. Work function measurements were performed in air by using a KP Technology SKP5050 scanning Kelvin probe with a 2-mm tip calibrated against a gold standard. The work function analysis was carried out over an area of approximately 1 mm × 1 mm.
Electrical measurements were performed by using a Hewlett-Packard 4156C semiconductor parameter analyzer. The blanket lower ZrCuAlNi electrode was held at ground potential with bias applied to the upper electrode. A dual-sweep measurement was employed to allow for an assessment of current-voltage curve hysteresis. The magnitude of the applied voltage bias was scaled according to the tunnel-barrier thickness to target maximum current levels in the μA range.
Current Density-Electric Field Simulation: Simulation of current density-electric field (J–ξ) curves was performed with Matlab, using the Fowler-Nordheim tunneling equations developed by Simmons.1718 Optimized fits to measured J–ξ curves, as shown in Figure 2, were obtained by using an effective mass between 0.45 to 1.0 of the free electron mass, φbZnCuAlNi = 2.2 ± 0.3 V, φbAl = 1.3 ± 0.1 V, and adjusting the insulator thickness.


This material is based upon work supported by the National Science Foundation under grant numbers CHE-0847970 (EWC, DAK, JFW) and DMR-0805372 (JFC) and by the Army Research Laboratory under contract W911NF-07–2-0083 (NA, CCK). The authors thank Peter Eschbach, William Stickle, Ronald Kelley, and Randy Burgess of the ADL at Hewlett Packard, Corvallis, Oregon for the TEM analysis; and Iain Balke of KP Technologies, LTD. for the Kelvin Probe analysis.
W. den Boer, in Active Matrix Liquid Crystal Displays, Elsevier, Amsterdam, 2005.
K. K. Ng, in Complete Guide to Semiconductor Devices, 2nd Edition, Wiley-InterscienceNew York, 2002.
J. L. Moll, IEEE Trans. Electron Devices 1963, 10, 299.
S. M. Sze, in Physics of Semiconductor Devices, 1st Ed, Wiley-InterscienceNew York, 1969.
High-Speed Semiconductor Devices (Ed: S. M.Sze), Wiley-InterscienceNew York, 1990.
S. M. Sze, K. K. Ng, in Physics of Semiconductor Devices, 3rd Edition, Wiley-Interscience, Hoboken, New Jersey, 2007.
S. P. Kwok, G. I. Haddad, G. Lobov, J. Appl. Phys. 1971, 42, 554.
G. M. Elchinger, A. Sanchez, C. F. Davis, A. Javan, J. Appl. Phys. 1976, 47, 591.
M. Heiblum, S. Wang, J. R. Whinnery, T. K. Gustafson, IEEE J. Quantum Electron. 1978, 14, 159.
M. Brunner, H. Ekrut, A. Hahn, J. Appl. Phys. 1982, 53, 1596.
E. N. Grossman, T. Harvey, C. D. Reintsema, J. Appl. Phys. 2002, 91, 10134.
P. C. D. Hobbs, R. B. Laibowitz, F. R. Libsch, Appl. Optics 2005, 44, 6813.
S. Krishnan, E. Stefanakos, S. Bhansali, Thin Solid Films 2008, 516, 2244.
J. A. Bean, B. Tiwari, G. H. Bernstein, P. Fay, W. Porod J. Vac. Sci. Technol. B 2009, 27, 11.
B. Berland, National Renewable Energy Laboratory Final Report, 2003, see:
G. Lewicki, C. A. Mead, Phy. Rev. Lett. 1966, 16, 939.
J. G. Simmons, J. Appl. Phys. 1963, 34, 1793.
J. G. Simmons, J. Appl. Phys. 1963, 34, 2581.
H. Warlimont, Mater. Sci. Eng. A 2001, 304, 61.
J. Tregilgas, Adv. Mater. Proc. 2005, 163, 46.
A.Greer, E.Ma, MRS Bull. 2007, 32, 611.
S. Hata, K. Sato, A. Shimokohbe, Proc, SPIE 1999, 3892, 97.
Y. Liu, S. Hata, K. Wada, A. Shimokohbe, Jpn. J. App. Phys. 2001, 40, 5382.
J. P. Chu, C. T. Liu, T. Mahalingam, S. F. Wang, M. J. O’Keefe, B. Johnson, C. H. Kuo, Phys. Rev. B 2004, 69, 113410.
P. Sharma, W. Zhang, K. Amiya, H. Kimura, A. Inoue, J. Nanosci. Nanotechnol. 2005, 5, 416.
P. Sharma, N. Kaushik, H. Kimura, Y. Saotome, A. Inoue, Nanotechnology 2007, 18, 035302.
F. X. Liu, P. K. Liaw, W. H. Jiang, C. L. Chiang, Y. F. Gao, Y. F. Guan, J. P. Chu, P. D. Rack, Mater. Sci. Eng. A 2007, 468, 246.
N. F. Mott, in Conduction in Non-Crystalline Materials, Oxford University Press, Oxford, 1993.
J. Dugdale, in The Electrical Properties of Disordered Metals, Cambridge University Press, Cambridge 2005.
M. Ohring, in Materials Science of Thin Films, 2nd Ed., Academic Press, San Diego, 2001.
E. W. Cowell III, C. C. Knutson, J. F. Wager, D. A. Keszler, Appl. Mater. Interfaces 2010, 2, 1811
S. T. Meyers, J. T. Anderson, D. Hong, C. M. Hung, J. F. Wager, D. A. Keszler, Chem. Mater. 2007, 19, 4023.

No comments: