AZ1512 is a positive tone photoresist commonly used for nanofabrication. A common approach to process development is to simply follow a canned recipe found online. When the patterns don't come out as expected, it is useful to determine what exactly what needs to be improved.


Develop a process for AZ1512 to print a pattern comprising of an array of 7 μm hexagons spaced 40 μm apart. The resist needs to be as thick as possible.

Design of experiment

A reported dose for AZ1512 is 70 mJ so we can design a 9 point dose series that covers this range.

  1. Use a 4" Silicon wafer
  2. Develop spin speed curve
    1. Spin coat at speed of 1000, 2000, 3000 and 5000 rpm
    2. Bake at 100 °C for 1 minute on hotplate
    3. Scratch film and measure thickness with profilometer
  3. Spin on 1.7 μm of AZ1512
  4. Bake at 100 °C for 1 minute on a hotplate
  5. Remove edge bead with acetone and wipes
  6. Develop dose curve
    1. Record g-line intensity (Channel B)
    2. Expose test patterns from 0.4 seconds to 3.6 seconds at 0.4 seconds increment
    3. Develop in AZ300 for 1 minute
  7. Expose pattern at 53 mJ
  8. Develop in AZ300 for 2 minute
  9. Inspect pattern in microscope
  10. Inspect pattern in electron microscope


Spin speed curve

The viscosity of AZ1512 makes it difficult to completely spread the photoresist across a wafer. Failure to cover the wafer with AZ1512 before spin coating will result in uncoated regions. We use a two step spin coating process:

  1. Spin at 200 rpm at 10 rpm/s for 60 seconds
  2. Spin at desired rpm at 1250 rpm/s for 60s
First dispense 1 mL of AZ1512 at the center of the wafer. Then run the spin program. During the first step watch for an even spread on the wafer. If the spread is not even, abort the spin program and dispense AZ1512 to cover the gaps. Then run the spin program again to completion.

Spin coating at a speed below 3000 rpm will result in thicker resist near the edge of the 3" wafer. Figure 1 show images of 3" wafers coated with AZ1512 at various spin speeds. At 1000 rpm, the edge bead is a few millimeters wide and it is ~2.18 μm thick compared to the nominal thickness of ~1.6 μm. The edge bead can be removed with a wipe and acetone, as shown in Figure 2.

Figure 1: Edge beads are apparent at spin speeds below 3000 rpm

Figure 1: AZ1512 photoressit piles up at the edge of the wafer at spin speeds below 3000 rpm.

Figure 2: Edge beads can be removed by wiping with acetone

Figure 2: Edge beads can be removed by wiping with acetone. The defect circled in red occurs when the photoresist does not completely coat the wafer before spinning.

Figure 3: Spin speed curve for AZ1512.

Height vs Dose (HD) Curve

The objective of the HD curve is to determine the amount of AZ1512 dissolves at a particular dose. To determine the optimal process, a series of parameters should be tested such as dose, developer type, developer time and developer temperature. To keep this simple, we use the recommended development process.

A dose series is conducted using a test mask that contains 9 identical patterns arranged in a 3 x 3 grid spaced 2 cm apart. An aluminum foil with a small square cut out if placed on top of the mask. The aluminum foil allows us to selectively expose any one pattern on the mask. After each pattern is exposed, the wafer is developed in AZ300 for 1 minute. Printing the entire dose series on the sample wafer ensures the only variable in the process is dose.

Figure 4: HD curve for AZ1512 reveals a critical dose of ~50 mJ/cm2.


These patterns are quite small, so good contact is required for photolithography to be successful. The edge beads must be removed prior to printing and the mask should be clean. The chuck should be planarized and vacuum contact should be used. If there are any gaps between the mask and the sample, the patterns may not be present on the sample after development. A good print is shown in Figure 5, where the printed pattern (6.2 μm) is slightly smaller than the mask pattern (7 μm). In Figure 6, an SEM image of the printed pattern shows straight vertical sidewalls. After printing this pattern many times, we have never gotten 7 μm hexagons, but frequently get ~6 μm hexagons.

Figure 5: The (left) mask contains 7 μm hexagons spaced 40 μm apart and the (right) printed pattern contains 6.2 μm hexagons spaced 40 μm apart.

Figure 5: The (left) mask contains 7 μm hexagons spaced 40 μm apart and the (right) printed pattern contains 6.2 μm hexagons spaced 40 μm apart.

Figure 6:  An SEM image of the printed pattern shows vertical side walls in the AZ1512 photoresist.

Figure 6: An SEM image of the printed pattern shows vertical side walls in the AZ1512 photoresist

Remaining Issues

This process works on a very specific lot of wafers where the adhesion of AZ1512 is very good. Unfortunately, we purchased this lot from Ebay with the intention of using it as a carrier wafer. These Ebay wafers are 100 mm Si(111) and single side polished (SSP). The polished side has some dot shaped defects that has no significant effect on the process. The adhesion of AZ1512 is not very good on brand new 100 mm, Si(100), 10 Ohm*cm, SSP wafers. After printing and developing, very few hexagonal patterns remain on the wafer.

Adhesion can be improved by spin coating HMDS on the wafer before the photoresist. The additional HMDS layer appears to have a significant effect on the dose and development for AZ1512 as shown in Figure 4. Although HMDS fixes the patterning process, it ruins the Bosch process. A method needs to be developed to completely remove HMDS without removing too much AZ1512. According to MicroChemicals, HMDS should be applied using a vapor coating method. Applying HMDS via spin coating is not recommended because it will interfere with photoresist development and contaminate the spin coater.