APPLIED PHYSICS LETTERS
VOLUME 77, NUMBER 20
13 NOVEMBER 2000
Highly parallel data storage system based on scanning probe arrays M. I. Lutwyche,a) M. Despont, U. Drechsler, U. Du¨rig, W. Ha¨berle, H. Rothuizen, R. Stutz, R. Widmer, G. K. Binnig, and P. Vettiger IBM Research, Zurich Research Laboratory, Sa¨umerstrasse 4, CH-8803 Ru¨schlikon, Switzerland
共Received 31 July 2000; accepted for publication 18 September 2000兲 This letter discusses an alternative storage approach to conventional magnetic data storage. The approach uses a 32⫻32 array of scanning probe microscopes working in parallel to read and write data as small indentations in a polymer storage medium. The results have densities of 100–200 Gbit/in.2. At such densities, it is shown that well over half the array works, and at lower densities more than 80% of levers are working. © 2000 American Institute of Physics. 关S0003-6951共00兲05246-3兴
The magnetic hard disk drive 共HDD兲 is todays most widely used mass data storage technique. This letter discusses an alternative storage approach using scanning probe microscope 共SPM兲 techniques. The read/write process is a slow thermomechanical rather than a fast electromagnetic process. However, the speed disadvantage is reclaimed by operating many probes in parallel. The data is written into a poly共methylmethacrylate兲 storage medium. An alternative storage technology must compete in five key areas: areal density, data rate, access time, error rate, and reliability. Recent experiments indicate that parallel SPMbased storage systems could compete with HDD technology in these areas, especially in the emerging field of small-formfactor devices for mobile computing applications. The areal density of HDD products is growing at an annual rate of 60–⬇100%, and is now about 6–⬇15 Gbit/in.2 for products. The superparamagnetic effect of current magnetic media means that areal density growth is uncertain beyond 100 Gbit/in.2. At current rates, this could ocHowever, SPM-based cur soon after 2004.1,2 thermomechanical writing and reading already shows potential for densities of 400 Gbit/in.2.3 This is almost ten times the current HDD world record lab demonstrations and well into the regime where the superparamagnetic effect will cause problems. The current data rate, the speed at which data can be read off the disk, for HDD is 20–30 MByte/s. The best data rate demonstrated for a single-lever SPM is only about 50 kbit/s for writing and 1 Mbit/s for readback.4 Hence, it is clear that some parallel form of operation will be needed to achieve sufficient data rates. This letter describes the operation of a system that allows parallel, time-multiplexed access to a total of 1024 cantilevers, arranged in a 32⫻32 array, with parallel operation of a row of 32 cantilevers. By using this simple system, parallel operation data rates comparable to or far exceeding those of HDD technology may one day be achieved. The thermomechanical method used was pioneered on a polycarbonate disk. An atomic force microscope 共AFM兲 lever is used to write data by heating the tip to make indentaa兲
Author to whom all correspondence should be addressed; electronic mail:
[email protected].
tions in the plastic film. Data readback was then carried out using standard AFM techniques. This thermomechanical method has evolved over the past decade, but four recent developments have been particularly important. The first is that thin films of polymer spun on a hard substrate allow much finer bits to be written because the hot tip penetrates only as far as the hard substrate. The second innovation was the use of the heated lever to read back the data. The heat lost to the substrate is affected by the proximity of the heater to the surface, which depends on whether the tip is in an indentation. Third was the discovery that by heating the polymer film, bits could be erased either by heating the entire medium or by scanning over the medium with a hot lever to achieve the same result locally. Finally, the understanding that if the array and the medium are sufficiently flat, only global array approaching is needed, simplifying the control system considerably. The latter three developments together make for a very efficient read/write system in which an array of passive levers with no individual z actuator can be made. Each lever needs only two wires to connect it and so it is easy to build a multiplexed system that works rather like a wired ROM. The manufacture and characterization of these chips has already been described.5,6 During the read/write process, the medium is scanned across the array in a raster fashion using two magnetic actuators. The medium is held at the correct distance from the array by a control system. This system senses the distance between the tip array and medium using four additional levers at the corners of the 32⫻32 array. These levers are identical to the array levers but have no tip. Similar systems have been used before for 5⫻5 arrays.7 The distance is measured by using the heat loss to the medium. The control system keeps the heat loss constant by adjusting four magnetic z actuators attached to the medium. A simplified 3⫻3 array circuit is shown in Fig. 1 to explain the read/write process used for this demonstration. Figure 1 also shows the current circuit symbol we used for the lever, a resistor with a lever attached. To write one bit to each of a row of levers first requires the column switches to be set. If a ‘‘1’’ 共indentation兲 is required on the column then the switch is connected to the positive voltage V c . If a ‘‘0’’ 共no indentation兲 is required, the switch is connected to the ground. Current values for V c are 2 to 4 V. Once the scanner
0003-6951/2000/77(20)/3299/3/$17.00 3299 © 2000 American Institute of Physics Downloaded 15 Nov 2002 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
3300
Appl. Phys. Lett., Vol. 77, No. 20, 13 November 2000
Lutwyche et al.
FIG. 1. Circuit for driving the chip.
is in the correct place the row switch is closed for 20 s to connect the row to the negative voltage V r . Current flows through all the levers but only those levers connected to V c receive sufficient power to write a bit. Levers in other rows receive no current due to their series and integrated diodes being in a reverse bias. The reason for this approach of heat-
FIG. 3. 共a兲 1024 images, one from each lever at 100–200 Gbit/in.2. 共b兲 Enlarged view of typical images from 共a兲. Numbers in the images indicate the row and column of each lever.
ing all the levers is to avoid too high a reverse bias of the diodes, which break down at between 3 and 4 V. To read back data, the column switches are set to the ground and the row switch is closed. The voltage V r is set to make the levers hot enough to sense the bits without damaging them. As the medium is scanned, the levers are cooled and heated as they move in and out of the bits. This changes the resistance of the heater. The change of resistance is measured by sensing the voltage across the sense resistors R s . Depending on the array chip, the values for V r are from ⫺7 to ⫺8 V for reading and from ⫺9 to ⫺11 V for writing. The value of R s is between 2.5 and 5 k⍀. With these values, the readback signal, ⌬R/R, is about 10⫺5 /nm and the 40 nm deep bits produce a signal of about 0.8 mV. The results of writing and reading in this fashion can be seen in Fig. 2, which shows 1024 images written by the levers then read back. Of the 1024 levers, 834 were able to write and read back data, which is more than 80%. The sequence is as follows. First, a bit pattern is written to each of
FIG. 2. 共a兲 1024 images, one from each lever at 15–30 Gbit/in.2. 共b兲 Enlarged view of typical images from 共a兲. Numbers in the images indicate the row and column of each lever. Downloaded 15 Nov 2002 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
Lutwyche et al.
Appl. Phys. Lett., Vol. 77, No. 20, 13 November 2000
the levers in row 1 simultaneously then read back simultaneously, followed by row 2, etc., through row 32. The images sent to the levers are different, each lever writing its own row and column number in the array. The bit pattern is 64⫻64 bits, but odd bits are always 0. In this case, the area used is 6.5⫻6.5 m2. The image read back is a gray scale bit map of 128⫻128 pixels. The interlever distance is 92 m so the images in Fig. 2 are also 92 m apart. A working storage system would fill the entire space between levers with data. Data in Fig. 2 correspond to 15–30 Gbit/in.2, depending on whether the coding system allows adjacent bits to run together. Allowing the bits to run together means two 1’s in a row would be an extra long indentation. Figure 3 shows data written with a different chip. Here, fewer levers are performing well, but the density is much higher at 100–200 Gbit/in.2. Individual levers in this array have achieved up to 150–300 Gbit/in.2. The range of densities from 15 to 150 with the two chips used may seem very large, but it should be noted that in terms of bit diameter it is only a factor of 冑10, i.e., about 3. Those levers that did not read back failed for one of four reasons: 共i兲 a defective chip connector like that in column 25 made that column unusable, 共ii兲 a point defect occurred, meaning that a single lever or tip is broken, 共iii兲 nonuniformity of the tip contact due to tip/lever variability or storage substrate bowing due to mounting, or 共iv兲 thermal drifts, with the latter two being the most likely and major sources. At present, there is clearly a tradeoff between the number of working levers and the density, which will most likely be resolved by a better substrate/chip mounting technique and lower thermal drifts. The writing and readback rates achieved with this system are 1 kbit/s per lever, thus the total data rate is about 32 kbit/s. This rate is limited by the rate at which data can be transferred over the personal computer industry standard architecture 共PC ISA兲 bus, not by a fundamental part of the read/write process. In conclusion, a very large two-dimensional array of local probes have been operated in a semiparallel fashion, and write/read storage operation in a thin polymer medium has been successfully demonstrated at densities at, or signifi-
3301
cantly higher than, those achieved with magnetic storage systems. Densities and yield of operation achieved with this demo are very encouraging, although considerable improvements are possible in both areas. Storage densities comparable to or higher than the 400 Gbit/in.2 demonstrated with single levers3 will be possible, whereas the high operating yield confirms the concept of global array approaching. Faster electronics will allow the levers to be operated at higher rates. The system needs to reach 60 Mbit/s to be useful for video applications, for example, which with 32 levers working in parallel would mean 2 Mbit/s per lever. With integrated electronics beside each lever all 1024 could be operated in true parallel fashion, allowing the rate per lever to drop to 60 kbit/s. Current and future investigations are focused on issues such as bit stability in polymer medium, tip/polymer wear as a function of lever loading force, system reliability and shock resistance as well as exploring the limits for storage density, signal-to-noise ratio, data rate, and power consumption. The authors are pleased to acknowledge the technical contributions, stimulating discussions, and encouraging support of our colleagues R. Beyeler, G. Cross, G. Genolet, R. Germann, and P. F. Seidler of the IBM Zurich Research Laboratory, as well as C. F. Quate, K. Goodson, and W. P. King of Stanford University, USA. E. Grochowski and R. F. Hoyt, IEEE Trans. Magn. 32, 1850 共1996兲. D. Weller and A. Moser, IEEE Trans. Magn. 35, 4423 共1999兲. 3 G. K. Binnig, M. Despont, U. Drechsler, W. Ha¨berle, M. Lutwyche, P. Vettiger, H. J. Mamin, B. W. Chui, and T. W. Kenny, Appl. Phys. Lett. 74, 1329 共1999兲. 4 R. P. Ried, H. J. Mamin, B. D. Terris, L. S. Fan, and D. Rugar, J. Microelectromech. Syst. 6, 294 共1997兲. 5 M. Despont, J. Brugger, U. Drechsler, U. Du¨rig, W. Ha¨berle, M. Lutwyche, H. Rothuizen, R. Stutz, R. Widmer, G. Binnig, H. Rohrer, and P. Vettiger, Sens. Actuators 80, 100 共2000兲; in Proceedings of the Electromechanical Systems Conference ‘‘MEMS ’99’’ 共IEEE, Piscataway, NJ, 1999兲, p. 564. 6 P. Vettiger, M. Despont, U. Drechsler, U. Du¨rig, W. Ha¨berle, M. I. Lutwyche, H. E. Rothuizen, R. Stutz, R. Widmer, and G. K. Binnig, IBM J. Res. Dev. 44, 323 共2000兲. 7 M. I. Lutwyche, C. Andreoli, G. Binnig, J. Brugger, U. Drechsler, W. Ha¨berle, H. Rohrer, H. Rothuizen, P. Vettiger, G. Yaralioglu, and C. F. Quate, Sens. Actuators A 73, 89 共1998兲. 1 2
Downloaded 15 Nov 2002 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
