Why they make the best ALD moisture barriers
In our ALD-CAP® case study, we discussed why atomic layer deposition (ALD) is the best choice for depositing encapsulation layers for RF, MEMS and photonic devices over other methods, like PECVD. But there’s more to the moisture barrier story than simply running an ALD process over PECVD.
The astute reader will have noticed that our ALD-CAP solution is based on nanolaminate films, or films made up of alternating layers of material. This lamination is intentional and crucial to creating a robust, long-lasting barrier film.
In this application note, we will explore:
- Why nanolaminates perform better as barriers than monolithic films
- Performance of a Forge Nano nanolaminate in extreme environments
- Benefits of using Forge Nano processes for your moisture barrier solution
Why do nanolaminates perform better than single-material films?
To understand this, let’s recall how ALD functions to prevent gas ingress.
Gases diffuse through materials via defect sites in the film. These defects may be an artifact of the process – a continuous-growth process, like CVD, will have more defect sites than a self-limiting process like ALD. Defects are also the result of the crystallinity of the material – amorphous films have fewer defect sites due to a lack of grain boundaries.
With fewer defect sites, a moisture barrier provides what we refer to as a “torturous path” for molecules to move from the surface to the underlying device. The more difficult it is to get through the barrier, the better the barrier performs. It’s like the difference between a hallway and a maze. It’s much easier to get from one end to the other via the former (Figure 1).
Figure 1. Schematic showing the difference between a A) non-torturous path and a B) torturous path for the diffusion of gas molecules. Note: This is only meant to depict the path and not be a representation of defect density.
There are a few explanations for how nanolamination improves the performance of moisture barriers. The simplest is that the introduction of a new material “resets” the torturous path. By periodically adding a new material to the stack, the path that gas molecules can use to move through the material is interrupted, making it more difficult to proceed. In essence, the defects present in the new layer are decoupled from those in the previous layer and gas molecules would need to move laterally across the interface to find a new path of entry. Some early work on gas diffusion models in dyad structures showed that diffusion is limited by long path lengths, rather than time lag.1
Another explanation is that the new material has a passivating effect on the first material. Here, the addition of the material fills interfacial defects, decreasing the overall defect density. A decrease in the defect density will translate to an increase in the barrier performance.
These interfacial effects can be extended even further. Some have shown that the addition of a new material at the interface can suppress extended crystalline formation of the previous material.3,5 Oftentimes, the interface will be a mixed oxide and can be a denser, less permeable section of the barrier. Reducing crystallite formation is another excellent strategy to decrease defect density in the barrier.
Add on top of these reasons that the thicker ALD films necessary for single layer materials tend to incur stress buildup and risk cracking, it is easy to see why nanolaminates are the desired strategy for ALD-made diffusion barriers.4 Many have experimented with using different types of nanolaminates to improve barrier performance, often showing orders of magnitude decreases in water vapor transmission rates (WVTR) with some as low as 10-6 g/m2/day.6
Forge Nano ALD-CAP® Performance
How does Forge Nano’s ALD-CAP perform? Let’s take a look!
Figure 2 shows XPS depth profiling data of the oxygen content of a SiNx substrate that had been coated with different moisture barriers before and after Highly Accelerated Stress Testing (HAST).
Figure 2. Oxygen concentration of a SiNx substrate coated with A) single 24.5 nm layer of SiO2, and B) a 24.5 nm nanolaminate made of alternating Al2O3 and SiO2 films following HAST exposure.
After 100 hours in a HAST environment (125 °C/85% RH), the SiNx substrate with the SiO2 monolith barrier showed an approximately 5% increase in oxygen, meaning some gas was able to penetrate through the single material barrier. In contrast, the SiNx substrate coated with a nanolaminate film showed no increase in its oxygen content following HAST.
To further validate the barrier performance, bow measurements were performed before and after HAST testing. Increases in the wafer bow, typically above 0.5µm, are another indication of moisture penetration. Following HAST exposure, the single material SiO2 barrier showed a bow increase of 0.6 µm, corroborating the XPS data. In contrast, the nanolaminate only showed a bow of 0.2 µm, well within the error of the measurement.
This was a simple, yet elegant, experiment to show the performance improvement of a multi-material, nanolaminate film over a monolith. While the two barriers are the exact same thickness, and both deposited by ALD at the same temperature, the introduction of lamination drastically increases the barrier performance. In fact, the thickness of the additional Al2O3 layers is a fraction of the SiO2 layers, showing that even small interruptions are enough to improve barrier performance.
So why choose Forge Nano for your moisture barrier solutions?
One advantage of Forge Nano’s moisture barrier process is our ability to deposit materials with our catalyzed thermal processes; and SiO2 is a great example. SiO2 is a go-to choice for moisture barriers for its high corrosion resistance. Si has no empty p-orbital and is amorphous when deposited, making it less reactive and with fewer grain boundaries for gas ingress. However, high-quality SiO2 has traditionally only been achievable using plasma or at high temperatures.
If you use plasma, the time penalty for performing nanolaminates is high due to the need to move your wafer back and forth between a thermal and plasma chamber. Since our SiO2 is deposited with a catalyzed thermal process, the entire film stack can be deposited in a single chamber, saving valuable process time and increasing throughput.
On the other hand, a high-temperature thermal process may allow one to perform the entire deposition in a single chamber, but the temperature difference between the SiO2 and complementary oxide (e.g., Al2O3) might be large, resulting in long wait times to change the process temperature. Our catalyzed processes offer much wider temperature windows than typical ALD processes, eliminating temperature compatibility as a concern.
Figure 3. Different ALD chamber configurations for depositing nanolaminate films. A) Dual-chamber thermal/plasma chambers and B) traditional single chamber configurations incur a time penalty. C) Forge Nano single chamber configurations do not require wafer transfers or large temperature changes.
Figure 4. Compound semiconductor photonic integrated circuit (PIC) coated with a nanolaminate moisture barrier using ALD. High AR and 3D structures are easily handled by Forge Nano’s thermal processing.
Forge Nano TEPHRA™ 200mm Cluster Tool
Forge Nano’s dedicated manufacturing tool is called TEPHRA. It is a single wafer, thermal-only cluster system equipped with proprietary valving and chamber design to ensure high quality ALD films at batch throughputs. The tool is designed to be modular with configurations available with up to 6 process modules to accommodate wafer volumes large and small. For information about our cluster system, please read our blog highlighting everything you need to know about TEPHRA.
Message our experts to learn more about an ALD-CAP solution for your devices!
References
- Graff et al. J. Appl. Phys. 96, 1840–1849 (2004)
- Park et al. 2011 Semicond. Sci. Technol. 26 034001
- Raghavan et al. Appl. Phys. Lett. 100, 191912 (2012)
- Behrendt et al. ACS Appl. Mater. Interfaces 2016, 8, 6, 4056–4061
- Meyer et al. Appl. Phys. Lett. 96, 243308 (2010)
- Dameron et al. J. Phys. Chem. C 2008, 112, 12, 4573–4580