Read-Write Heads Follow Steep Progress Curve
John Baliga, Associate Editor -- Semiconductor International, 9/1/2000
 |
At a Glance | | |
|
| ||
| |||
The write part of the head is still an inductive coil. The width of the top pole determines the areal density that can be recorded (Fig. 1). The limiting factor in areal density, though, is the ability to read the data. Three things must be done to improve this capability: reduce the distance between the read sensor and the disk, reduce the size of the sensor, and use material sets that can be scaled down and still sense the data reliably.
Read-write heads are made using technology similar to semiconductormanufacturing technology, but there are some important differences. One is that heads are most often made on alumina titanium carbide substrates instead of silicon. The presence of magnetic films that require orientation adds different variables to the process. On top of this, the die, or slider, must be given the appropriate three-dimensional shape, since it must literally fly over the disk.
The sensing mechanisms
The most popular sensing mechanism is the magnetoresistive effect (Fig. 2).1 The sensing layer takes on the magnetization of the sensed bit on the disk, while the pinned layer maintains its magnetization. The copper spacer layer makes the whole sandwich conductive. When the magnetizations are aligned, electrons moving from one magnetic layer to the other are less likely to scatter. A current is forced through the sandwich, and resistance changes are read as voltage changes.
Detection capability is highest when the magnetizations of the two magnetic (ferromagnetic) layers are forced to be either parallel or antiparallel. For giant magnetoresistive (GMR) read structures, this is accomplished by making the copper spacer thin enough to allow weak coupling between the two magnetic layers. The magnetization of the pinned layer is maintained by the antiferromagnetic exchange layer. The interface between the exchange and pinned layers usually is such that electrons generally do not cross the boundary, so the exchange layer plays no direct part in the GMR effect.
|
|
|
|
In all these structures, the spacer or barrier film is usually only a few nanometers thick.
Material choices
The reduction of bit sizes requires an increase in the sensitivity of the read head. Also, as the feature size decreases, magnetic materials have more difficulty holding a magnetization. For these reasons, new magnetic materials are under investigation for upcoming generations of read-write heads.
Antiferromagnetic films are used to stabilize the pinned layer partly because the pinned layer is so thin. Making the pinned layer thick enough to hold its magnetization stable on its own would require too much material.
Permalloy (Ni80Fe20) and cobalt are typical ferromagnetic materials. Alloys of manganese with metals such as platinum, iron and iridium are typical antiferromagnetic materials, as are oxides of metals such as nickel and cobalt.
Many have expressed the opinion that the magnetic properties of materials will impose the ultimate limit in areal density, rather than lithography or patterning capabilities. This may be true, but many think this limit is far from being reached.
"There is a huge universe of magnetic materials that have not been investigated as yet," said Jim McKibben, vice president, worldwide marketing and sales for Tegal (Petaluma, Calif.). "They (head manufacturers) have moved to steppers from scanners and other approaches that they used in the past, so they are indeed following a scaling path in lithography, but that's not a limitation. Holding a spin state on magnetic materials is an issue; but the head manufacturers have many new materials that they're investigating, and the answer probably lies in that universe somewhere."
Material deposition
The two most prevalent methods for depositing magnetic materials are sputtering and ion beam deposition. Unlike semiconductor manufacturing, it is not enough to deposit a film with good thickness uniformity. Some of the magnetic layers also must be oriented.
Typically, the wafer stage in the deposition chamber will have electromagnets to provide a magnetic field to orient the film as it is deposited. For the pinned layer, this ensures that all of the layer ends up being pinned in the appropriate direction. "One of the layers is going to end up being pinned by an antiferromagnet underneath, but it tends to stabilize the magnetic film if you give it an orientation prior to doing any sort of other stabilization to it," said Kurt Williams, process technology manager at Veeco. "In other words, if you let it break up into its own domain structure, it causes other issues such as electromagnetic noise due to the switching of the individual domains."
On the subject of orienting the free layer during deposition, Williams said: "When it sees a magnetic field such as a bit on the rotating media, and the sensing film switches to another direction, the entire film switches, and not just certain domains of the film or certain domains of the film that are switching." Orienting during deposition optimizes the MR effect of the stack, which buys some capability to make them smaller.
Not only do the films have to be oriented; they have to be oriented uniformly. Williams added: "Generally you don't need more than about 100 oersteds across the wafer, but the problem is that it has to be uniform, less than two degrees of skew across the wafer." Small variations of the applied field can cause significant variations in the MR response of the heads.
In some cases, annealing is done in the presence of a magnetic field to properly orient the film. However, since the antiferromegnetic film must be oriented at 90° to the pinned layer, magnetic biasing under temperature usually is not used. "They sometimes like to deposit a whole stack of layers simultaneously in the same process, and they like to change the orientation of the magnetization of separate layers, so the stage has to be able to switch magnetization by 90 degrees," said Williams. "Once you've deposited the whole film in one direction, it tends to stay in that direction due to the shape anisotropy, so it's difficult for it to switch completely (during deposition) unless of course you change deposition conditions like temperature that will cause it to want to switch."
The metallic and insulating materials for the spacers and shields also are typically sputtered. These layers can be extremely thin. IBM has reported research in which an aluminum oxide layer is made by sputtering two atomic layers of aluminum, then oxidizing.
The layer between the sensing stack and the shared pole is very important in allowing the bit width to be decreased, and its thickness is approaching 10 nm for areal densities above 100 Gb/in.2. At the same time, this layer needs to meet very stringent requirements in breakdown voltage, uniformity and step coverage. "At this thickness the advantages of ion beam deposition have been found to have tremendous potential to achieving high device yields," according to Mervyn Davis, technical director of Nordiko (Havant, U.K.). "Present head designs require simple alumina," he added. "However, this technology can be extended to the more complex dielectric oxy/nitride materials thought to be needed for improved thermal properties as the bit size shrinks even further." Davis went on to say that using IBD to deposit the TMR barrier layer with the required uniformity is straightforward.
Since these films can be so thin, companies providing atomic layer deposition technology have expressed interest in the read-write head market. This type of technology has been used to deposit extremely thin aluminum oxide layers.2 It also is possible to deposit elemental films such as copper, though the process requires at least two different precursors.
Layer control
The thickness of the layers deposited for read-write heads is comparable to, if not thinner than, those found in semiconductor manufacturing. Also, metrology is being used in more places, adding shorter control loops in much the same way as semiconductor manufacturing does. Many companies supplying film stack metrology equipment to the semiconductor industry also supply the read-write head industry.
On the subject of the value of inspection and metrology, Chris Wootan, general manager of KLA-Tencor's Data Storage Group, commented: "The cost to produce recording heads at wafer level is on the order of 50 to 80 cents each. Another one to two dollars are added during the slider area process, where they are cut, lapped and the air bearing surface (ABS) is defined. By the time a head is mounted and tested magnetically and ready for drive assembly, the cost accelerates to four to six dollars each. Bad devices with wafer-level root causes that aren't found until mag test and drive test hurt these manufacturers' product margins substantially." He went on to say some defects can be detected at the wafer level that would not show up in probe or intermediate testing stages.
According to Roger Ingalls, vice president and director of marketing for Nanometrics (Sunnyvale, Calif.), these exotic multilayer films can be more difficult to analyze than films used in semiconductor processing. Nanometrics extended the range of its spectroscopic ellipsometers into the deep-UV range to simultaneously analyze the components of film stacks.
Rudolph Technologies (Flanders, N.J.) provides film stack characterization equipment using its MetaPULSE ultrasonic sonar technology, which determines the thickness and materials properties of each layer in the multilayer metal film stack and can detect buried defects. George Collins, director of marketing at Rudolph, said this has allowed head manufacturers to achieve increases in yield of more than 10%.
n&k Technology (Santa Clara, Calif.) provides broadband photospectroscopy equipment for characterizing film stacks, using light from the UV to the near IR parts of the spectrum. The company has demonstrated accurate and repeatable capabilities for measuring the overcoat thickness, which can be 20 to 30 Å thick.
X-ray fluorescence (XRF) has been used for years as an impurity analysis tool, but it also has started seeing use as a thickness monitoring method for ultrathin films. Alec Reader, marketing manager for Philips Analytical (Almelo, Netherlands), explained XRF's capabilities for determining both thickness and composition of the thin layers in a GMR stack. "XRF tools can measure the 25 angstrom copper film in the stack center with a relative standard deviation less than 0.2%," Reader said. "The stoichiometry of layers such as PtMn or NiMn also can be measured with a similar precision." Reader added that XRF is non-destructive, and the results are operator-independent.
In addition to layer thicknesses, orientation and field strength of the magnetic layers also must be controlled within the wafer and wafer to wafer. ADE Technologies (Newton, Mass.) provides vibrating sample magnetometers for process development, and a wafer mapping line that measures the magnetic properties of all the layers within a stack.
Patterning
Patterning capabilities required for read-write head manufacturing are comparable to those required in semiconductor manufacturing. State-of-the-art production hard disks have heads with 0.4 µm features, and research devices have features approaching 0.10 µm. All major lithography exposure tool suppliers in the semiconductor industry supply tools for the critical dimensions in read-write head manufacture.
|
|
In the past few years, minimum feature size requirements for read-write heads have progressed on a sharper slope than for semiconductor manufacturing (Fig. 3). A cross-over or merge of the requirements generally is estimated to occur in the middle of 2001.
The two critical dimensions on a read-write head are the width of the MR stripe and the width of the top write pole. For the top pole, the most common method for patterning is to use thick photoresist, open holes where the poles are to be deposited, then plate them on. After the photoresist is removed, the pole can be milled to the proper dimension using standard ion mills or focused ion beams (Fig. 4).
According to Jay Lindquist, general manager, MicroElectronics Product Group, FEI (Hillsboro, Ore.): "Recent advances in throughput and precision have made FIB trimming economically viable for head manufacturing. Furthermore, FIB trimming supports the next several generations of head technology with existing process equipment, eliminating the need to retool and develop new processes with each generation. FIB also is used in combination with high-resolution SEM to deliver in-process measurements of sub-surface features for control of GMR structures as well as write pole trimming, notching and plating."
 4. The pole of this head was milled to a thickness of 0.1 µm using a focused ion beam tool. (Source: FEI) |
Another method that is starting to find use is to deposit the material, pattern the photoresist, then use reactive ion etching (RIE) to remove all but the pole. The MR stripe is patterned using standard semiconductor manufacturing techniques. Wet etching is giving way to RIE and other dry etching methods.
Formation of the air bearing surface (ABS) requires the removal of large amounts of material from a surface that is no longer flat. This can be done lithographically with etching, or with ion mills or focused ion beams. Tegal's Jim McKibben believes ion mills will be used less in the future for the fine features as well as bulk removal on the ABS. "I'm not forecasting the death of ion milling, but we do believe the market, as it pursues the scaling challenge, is going to need to move to dry etching techniques," he said.
McKibben believes the shallow angle required for argon ion milling ultimately will limit its capabilities. "The issue becomes how shallow an angle can you use before you can no longer achieve the device size that you want to achieve," he said. "And then the secondary issue is the redeposition of the sputtered material."
Air bearing surface (ABS)
The air bearing surface is unique to read-write heads. The head literally flies over the surface of the disk as the disk moves. In addition to being a device using submicron features, the head must be on an aerodynamically shaped slider that flies over the disk at a well-controlled distance. This distance is on the order of tens of nanometers, but it is very important to note that it is non-zero. The head can rest on the disk; but when the disk moves, contact would lead to a literal crash. The shape required for the ABS can be very slight, but very important. In modern "pico" sliders, the curvature parameters can be a few nanometers, while the width and length of the slider are ~1 mm.3 The three-dimensional shape of the slider is very important. After the write and read elements have been created, the wafer is cut into strips, or rowbars. These pieces are then ground or lapped to create the desired front-to-back shape (Fig. 5). Afterward, the rowbars are placed in a carrier and patterned. The patterning process can have a couple of lithographic steps followed by ion milling or reactive ion etching steps.
|
|
The slider and disk each have a protective overcoat, which protects against a crash when the disk starts or stops. This overcoat must be thick enough to protect both surfaces, but no thicker. A layer of diamond-like carbon deposited on an amorphous silicon layer is typically used, and its thickness can be tens of angstroms.
Interplay between head and media
The read-write head and disk medium have taken turns as the limit for areal recording density. Since the grains in the medium are randomly oriented, 200 to 300 grains are required to make up one bit to ensure enough grains are oriented in the appropriate direction. Recent formulations have made it possible for the grains to be on the order of 5 nm in diameter, so the medium no longer imposes the limit. Minimum feature size on the heads now is the limiting factor.
The medium ultimately will impose a limit, known as the superparamagnetic limit. The stability of a grain's magnetization depends on its size. The smaller the grain, the more likely it is to flip its magnetization due to thermal fluctuation at room temperature. A 2× reduction in grain size could change the expected frequency of this flipping from 100 years to 100 nanoseconds.4 Although the material is magnetic, it acts like a non-magnetic (paramagnetic) material at the macroscopic ("super") level.
Areal densities have increased to the point that the arm holding the head cannot possibly position the head accurately over a chosen track on the disk. A 20 Gb/in.2 areal density requires positioning within one micron. One solution for this problem is the use of microactuators. The arm gets the head within a few tracks of the appropriate position, and the head itself distorts or adjusts to position the read or write elements over the appropriate track. In addition to being aerodynamic, the slider also must be attached to a MEMS device to achieve these high areal densities.
Conclusion
Read-write head manufacturing is similar to semiconductor manufacturing, though it has important differences. In addition to film thickness precision and uniformity, magnetic properties also must be controlled. After all of that is done, the die, or slider, must be turned into an aerodynamic device to take advantage of the head's high-density capability.
The lithographic and other technical requirements of head manufacture are fast approaching those of semiconductor manufacturing. Opinions vary on whether the lithographic progress curve will cross over or merge with that of semiconductor technology, and the differences in opinion hinge on economic factors rather than technology. •
REFERENCES
- J. Baliga, "GMR Read-Write Heads Yield Data Storage Record," Semiconductor International, February 1998, p. 38.
- P. Singer, "Atomic Layer Deposition Targets Thin Films," Semiconductor International, September 1999, p.
40.
- A.C. Tam, C.C. Poon, L. Crawforth, P.M. Lundquist, "New Laser Curvature Adjust Technique (LCAT) for Fine Adjustment of the Crown and Camber of Magnetic Head Sliders," INSIGHT,
May/June 2000, pp.8-12.
- D.A. Thompson, J.S. Best, "The Future of Magnetic Data Storage Technology," IBM Journal of Research & Development, volume 44, number 3 (2000).
