Advanced fabrication of some of magnetic and electronic
structures of interest to the program will be performed in the
clean room facility of MCNC. Patterning of magnetic multilayers,
containing materials such as Cu, Au, Cr and other transition metals,
will be performed using the metal lift-off technique. For these
materials, dry etching techniques are not readily available due
to the lack of high vapor pressure compounds of the metals with
common dry etching gases, such as Cl2, Br2,
and F2. In lift-off, an inverse pattern is
first formed in a sacrificial layer deposited on a substrate,
using lithographic techniques. Next, the metal film is deposited
over the layer and in the openings of the pattern. Those portions
of the metal film which are deposited on the sacrificial layer
are removed (lifted-off) when the substrate is immersed in a suitable
solvent, leaving behind the desired metal pattern.
2. Lift-off process for the I-line stepper
During the first reporting period, we have therefore begun qualification of lift-off processes for MCNC's I-line (365 nm) stepper (expected I-line stepper resolution is 0.6 µm). Because of safety concerns, Shipley's 1400-31 and 1400-37 resists for which MCNC's original positive resist lift-off process had been developed, were withdrawn from production. We have therefore decided to investigate suitability of the JSR SC1827 material for the positive resist lift-off process, as well as attempt to develop a negative resist lift-off process which has fewer steps than the positive resist lift-off, and is therefore more cost- and time-effective. The JSR NFR016D2-55cp resist is a candidate material for the negative lift-off process. To qualify the new resists, we need to verify that the process utilizing the resists works reliably for the desired resolution.
Figure 1 shows a sequence of steps in the lift-off process in which a negative photoresist is used. Regions which are exposed to the light are insoluble in the developer solution. Note the negative slope of the photoresist profile following the development. This negative slope facilitates the subsequent metal deposition and lift-off since it prevents metal to deposit on sidewalls of the resist, and allows for the solvent to reach the sacrificial layer.
Figure 2 shows a processing sequence utilizing a positive resist. In order to reverse the tone of the image in the positive resist so that it functions as a negative resist, a so-called image reversal method is used in which after exposure the resist layer is subjected to amine vapors. The amine diffuses through the resist, and reacts with the dissolution enhancer (carboxylic acid) photoproduct in the exposed areas. The process is conducted in a vacuum bake chamber, which allows the amine vapors to be uniformly delivered to the resist, and for the temperature to be uniformly controlled. Following the vapor treatment, the resist is flood exposed by UV light, and developed in a conventional manner.
As part of the qualification effort, we have processed one batch of test wafers using a 2.3 µm thick SC1827 positive resist layer and one batch using a 2.3 µm thick NFR016D2 negative resist layer. A 500Å Cr/500Å Ta/4000Å Pt metal multilayer deposited by e-beam evaporation was chosen for the test. The lift-off was performed in the NMP solvent (1-methyl-2-pyrrolidinone). Completed wafers were inspected using scanning electron microscope (SEM). The inspection focused on examination of metal edges (smooth/ragged) for both the large (tens of µm) and small (0.6 µm - 2 µm) features. The latter had the form of L-bars with varying widths of the bars, and the pitch between the bars being equal to their width.
The edge of the metal in the process utilizing the negative resist looked ragged. On the other hand, a very thin resist ribbon was observed at the metal edge in the process utilizing the positive photoresist, but this ribbon was easily removed using a 3 min. O2 plasma. After the plasma strip, the metal edge looked very clean and smooth.
Neither positive or negative resist tests, however, gave adequate
uniformity for lift-off on the small L-bar features. The test
will therefore be repeated with focus and exposure adjustments.
If this does not improve the resolution and uniformity, a resist
layer thinner than the one used in the first round of tests will
be examined This will require identifying a negative resist which
has a lower viscosity than the NFR016D2 material, and can therefore
be spinned to a smaller thickness.
3. Lift-off process for the deep UV laser stepper
In the past, MCNC developed a so-called tri-level lift-off process to form metal layers embedded in a patterned dielectric layer. Figure 3 shows a proposed process sequence for patterning of free standing (non-embedded) metal layers, based on the tri-level lift-off. The advantage of the process shown in Fig. 3 is its compatibility with a positive resist process for MCNC's deep UV laser stepper. In principle, this would allow lift-off patterning of structures down to 0.5 µm.
The modified tri-level lift-off process, shown in Figure 3, begins with spinning of a 0.2 µm - 1 µm soluble polyimide layer. In the original tri-level lift-off, other materials such as PMGI and polystyrene, were used successfully. Following the polyimide, a thin (500 Å - 1000 Å) template is deposited. The template needs to be resistant to the dry etch which is used to open holes in the polyimide layer. Candidate materials include evaporated Si, Al or SiOx layers. Next, a layer of positive photoresist is spinned on top of the polyimide-template sandwich, followed by the exposure of the photoresist using the laser stepper. Then the photoresist layer is developed, and the pattern created in the photoresist is transferred to the template via dry etching (e.g., in the case of polycrystalline silicon Cl or F containing gases are used). This is followed by opening holes in the polyimide layer using a dry etching oxygen plasma based process. A critical aspect of the etching process is the creation of the undercut in the polyimide layer with respect to the template. This undercut facilitates the subsequent metal deposition and lift-off since it prevents metal deposition on sidewalls of the polyimide. The lift-off is performed in a suitable solvent bath.
We will need to test the process described above to verify its
usefulness for patterning of specific materials of interest to
the program. Materials compatibility and process integration
issues will be examined as part of the planned process development.
Figure 1. Metal lift-off process sequence utilizing negative
Figure 2. Metal lift-off process sequence utilizing positive
Figure 3. MCNC's tri-level lift-off process modified for patterning of free standing metal layers.