Issue



Structure and performance of TGMR heads for next-generation HDDs


09/01/2004







This article will provide a general view and the current status of the new tunneling magnetoresistive technology to achieve recording performance suitable for next-generation hard disk-drive (HDD) products.

Ultrahigh-density magnetic HDD recording needs high sensitivity and stability for digital information sensing. The technologies used during the past 40 years have evolved from thin-film heads to anisotropic magnetoresistive (AMR) and giant magnetoresistive (GMR) heads.


Figure 1. a) CIP recording head structure. b) CPP head structure, including CPP-GMR, CPP-SV, and CPP-TGMR.
Click here to enlarge image

It has been a considerable effort to extend GMR/spin-valve (SV) head technology beyond 100Gbit/in.2, and the technical challenges of such scaling are significant [1, 2]. Soon current-perpendicular-to-place (CPP) heads will replace historical current-in-place (CIP) heads. Both are based on the current feed-through method in the sensing area of a readback head. Figures 1a and 1b show the general layout of CIP and CPP heads. Several potential CPP readback heads (CPP-SV, CPP-GMR, and CPP-TMR) are proposed for 100Gbit and beyond [3–6]. Among them, the spin-tunneling [also known as tunneling magnetoresistive (TMR), or tunneling giant magnetoresistive (TGMR)] head appears to be the only type that demonstrates acceptable recording bit-error rate performance beyond an areal density of 100Gbit/in.2 [3, 6–9].


Figure 2. Comparison of CPP-SV and CPP-TMR stacks. In the tunneling head, the spacer layer of Cu is replaced by an oxide barrier.
Click here to enlarge image

The TGMR head stack reported here is basically a bottom type with a pinning layer of PtMn to achieve thermal stability due to the high blocking temperature. This multilayer stack is similar to traditional spin-valve materials, except that the spacer layer of Cu is replaced by an oxide barrier (Fig. 2). The oxide barrier process requires special oxidation, which can be done in a commercial multiple-chamber deposition tool. Various oxidation methods (e.g, natural, plasma, UV light-assisted, radical oxidation, and more) can be used to optimize the process. The pinned layer is typically formed by a synthetic-antiferromagnet (SAF) structure consisting of a CoFe binary alloy. The free layer is also composed of CoFe/NiFe soft magnetic layers. Adding an oxide layer to the stack introduces new tunneling physics that will guide the transport properties, so a magnetoresistance as high as 70% can be achieved. An oxide layer also poses some unique physical properties, such as a negative temperature coefficient of resistance, and a nonlinear bias-voltage dependence (I-V curve) that will affect the head performance differently than in GMR/SV devices. Due to the high magnetoresistance ratio, this is also used in MRAM devices, which have made significant progress in the past year [10]. The only difference between an MRAM and a TGMR recording head is the barrier width target, which affects the product of area and resistance (Ra). For an MRAM tunneling material, the Ra is typically in the range of 1000Ω-µm2 vs. <10Ω-µm2. The metal layer thickness before oxidation is typically around 12Å for MRAM and 6Å for TGMR heads, respectively.

The TMR stack is built on a bottom electrode of NiFe that also serves as a magnetic shield. A ceramic substrate (AlTiC) is used below the magnetic shield to provide the mechanical support for building the final slider. The TGMR stack is patterned by a standard photolithographic process. An ion mill step is used to form the reader width, which is typically <0.1µm.


Figure 3. ABS view of a merged TGMR head with PM bias layer. An inductive writer is on top with an insulation layer of alumina.
Click here to enlarge image

Since the TMR is a CPP device, the junction needs to be isolated from the permanent hard bias layers, with a layer of alumina. Thickness and profile control of the layer is important. After TGMR junction nanofabrication, it is capped by a top electrode of NiFe, which also serves as the top shield. To complete a recording head, an inductive writer (either ring head or probe head for longitudinal or perpendicular recording) is built on top of the TMR reader with an electrical insulation layer between the reader and writer, as shown in Fig. 3 for a longitudinal recording head viewed from the air-bearing-surface (ABS). Figure 4 is a cross-sectional view of the finished TGMR head with a ring inductive head. Figure 5 is a detailed view of a TGMR head with abutted hard bias permanent-magnet layers.


Figure 4. Cross-sectional SEM image of a finished TGMR head with an inductive writer for longitudinal recording.
Click here to enlarge image

These processes can be carried out on existing production lines, indicating the compatibility of this new technology with current GMR production. Volume manufacturing of this new technology presents still more engineering challenges that must be resolved before a product is finally realized [3, 6]. For example, a better deposition system can improve the uniformity and reproducibility of very thin oxide barriers [11, 12] and an improved junction definition process can significantly boost yield [13]. It appears that a uniform process across several million junctions can be achieved [10]. A TGMR tool enabling atomic engineering and interface control is needed to commercialize this new technology. Besides advanced sputtering tools, process optimization is also important for volume production of TGMR heads. The junction fabrication can be quite different, for example, depending upon the ion mill details and interaction with the mask.


Figure 5. ABS view of a 130Gbit/in.2 TGMR head with a permanent-magnet longitudinal hard bias layer separated from the free layer by an oxide.
Click here to enlarge image

Current TGMR heads have an electrical reader width of about 4µin. to reach high track densities. Finished heads have stable and linear transfer curves, achieved through optimized PM stabilization. With a spinstand test using a standard inductive writer, one can optimize the write current for the best on-track bit-error rate and conduct a 747 curve test using standard two-side squeeze (a test conducted routinely in the HDD industry) [7]. Using the TGMR head in longitudinal recording mode, in which the magnetization in the media is aligned in the disk plane, the best linear density achieved is at 890kbpi (kilobits per inch). One can test the TMR head using a one-side squeeze method for a reader capability test, showing that a track density of 171ktpi (kilotracks per inch) is attained and 152Gbit/in.2 areal density is possible. The head also typically shows a few decibels better S/N ratio over similar areal density heads using the SV GMR design. Rolloff curves show that even at 1000kbpi, the TGMR head still has 5dB gain over the media.

We further explored the application for TGMR in perpendicular recording mode, considered the next recording technology for HDD products. Due to the narrower shield-to-shield spacing, a TMR head is suitable for high linear-density recording in the perpendicular mode. It is speculated that the new TGMR head technology may be combined with perpendicular recording, so testing of a TGMR head in this mode will help answer the question.


Figure 6. TGMR/probe writer structure for perpendicular recording.
Click here to enlarge image

Figure 6 shows the head layout of a TGMR reader combined with a single top probe writer [11]. A very similar test of the 747 curve shows that on perpendicular recording, an areal density of 170Gbit/in.2 is obtained. This is more than double the current 60–70Gbit/in.2

Conclusion

There are no fundamental drawbacks for TGMR head use in perpendicular recording. In general, the TGMR head has proven to be a viable candidate for next-generation head applications, and engineering hurdles appear to be surmountable, paving the way for commercialization of this technology.

References

  1. S. Mao, et al., "Materials Properties of Spin Valve Stacks for 100Gbit/in.2 Head Applications," invited paper T1.1, MRS Spring Meeting, 2001.
  2. Z. Zhang, et al., "Magnetic Recording Demonstration over 100Gbit/in.2," invited paper BA02, Intermag, 2002.
  3. S. Araki, et al., invited paper CA-02, Abstracts of MMM, 2002.
  4. K. Nagasaka, et al., invited paper CA-04, Abstracts of MMM, 2002.
  5. M.A. Seigler, et al., invited paper CA-05, Abstracts of MMM, 2002.
  6. S. Mao, et al., invited paper DC-01, Abstracts of MMM, 2002.
  7. R. Swanson, et al., paper DC-02, Abstracts of MMM, 2002.
  8. S. Mao, et al., invited paper E4, Digests of TMRC, 2003.
  9. E. Murdock, et al., invited talk, "170 Gbit/in.2 Recording Demonstration," Lake Arrowhead Meeting, Dec. 2003.
  10. B. Engel, et al., " 4-Mbit MRAM Based on a Novel Bit and Switching Method," invited paper GE-05, Abstracts of 9th Joint MMM/Intermag Conf., 2004.
  11. S. Mao, Veeco Symposium on Data Storage, Sept. 2003.
  12. S. Xue, et al., "Advanced GMR/TMR Reader in Magnetic Recording Heads," invited talk, U. of Minnesota, Feb. 21, 2003.
  13. E. Chen, "Magnetic Tunnel Junction Pattern Technique," J. Appl. Phys. 93, p. 8379, 2003.

Sining Mao is R&D director of advanced device development in the recording head division at Seagate Technology, 7801 Computer Ave., Minneapolis, MN 55347; ph 952/402-7909, fax 952/402-8074, e-mail [email protected].