3D InCites Podcast
3D InCites Podcast
The Role of Material Modeling in Semiconductor Packaging Innovation
Unlock the secrets of semiconductor packaging materials with insights from industry experts Dariush Tari and Rose Guino of Henkel's Semiconductor Packaging Materials Division. This episode promises a deep dive into the processes behind developing materials that are both reliable and high-performing, crucial for the ever-evolving demands of AI, machine learning, and quantum technologies.
Dariush and Rose share their wealth of knowledge on material characterization, modeling, and application engineering, and discuss how Henkel maintains its pivotal role within the semiconductor ecosystem. From underfills to thermal interface materials, discover how comprehensive material offerings are shaping the future of high-performance computing.
Explore the fascinating advancements in material modeling for semiconductors, where physics-based simulations are transforming reliability testing and development cycles.
Gain a clearer understanding of capillary underfill materials and their vital role in enhancing solder joint reliability under thermal stress.
This episode delves into the collaboration between material developers and modeling experts, underscoring the importance of early customer engagement to tailor innovative solutions.
If you are interested in learning how cutting-edge materials are propelling the semiconductor industry forward, this conversation is a must-listen.
Contact the Speakers on LinkedIn
Dariush Tari, Henkel Semiconductor Packaging Materials
Rose Guino, Semiconductor Packaging Materials
Henke's advanced materials elevate semiconductor packaging to meet power, performance, area and cost
Disclaimer: This post contains affiliate links. If you make a purchase, I may receive a commission at no extra cost to you.
Become a sustaining member!
Like what you hear? Follow us on LinkedIn and Twitter
Interested in reaching a qualified audience of microelectronics industry decision-makers? Invest in host-read advertisements, and promote your company in upcoming episodes. Contact Françoise von Trapp to learn more.
Interested in becoming a sponsor of the 3D InCites Podcast? Check out our 2024 Media Kit. Learn more about the 3D InCites Community and how you can become more involved.
This episode of the 3D InCites podcast is brought to you by Henkel, a global materials innovator for industrial and consumer businesses In the electronics sector. Henkel is renowned for its semiconductor solutions for wire bond and advanced packaging applications, consumer device assembly ingenuity and leading-edge thermal management materials. Founded in 1876, the company employs about 48,000 people worldwide and has a long tradition of sustainability leadership. Discover more at Henkel. com.
Francoise von Trapp:Hi there, I'm Francoise von Trapp, and this is the 3D InCites Podcast.
Francoise von Trapp:Hi everyone, have you ever thought about what goes into developing novel materials for your advanced packaging processes? You know, much of the same consideration that goes into the design of packages themselves also goes into materials development. So in this episode we're speaking with Dariush Tari and Rose Guino, who are subject matter experts at Henkel's Semiconductor Packaging Materials Division, and they're going to tell us how they approach the process by looking specifically at capillary underfill development, and we're going to learn how they're developing new methods to address new challenges and development of these important materials as semiconductor packaging evolves. So welcome to the podcast!.
Dariush Tari:Thank you.
Rose Guino:Hello.
Francoise von Trapp:So it's so great to have you both on here Before we dive in. Can you each share a little bit about your background and your specific roles at Henkel Dariush, why don't you start?
Dariush Tari:Yes, hello everyone. My name is Dariush Tari. I'm a modeling manager in central R&D team. I've been with Henkel for the past five years. I have a PhD in solid mechanics from University of Waterloo in Canada. My expertise are material characterization, material modeling, structural modeling, and I've been working in metal forming, automotive, aerospace and semiconductor modeling space for the past 17 years. I'm very happy to be here with you.
Francoise von Trapp:Okay, and Rose, how about you?
Rose Guino:Thank you. Hello everyone, my name is Rose Guino and I'm the application engineering manager at Henkel. Been with Henkel for 17 going 18 years now, have a PhD in polymer chemistry and have developed more than a dozen products. Thank you for having us.
Francoise von Trapp:Henkel manufactures a lot of different materials for different markets, so how does it operate within the semiconductor ecosystem? Rose?
Rose Guino:Well, Henkel works with many customers across the value chain and semiconductor right. So we don't operate in the wafer front end but we do, starting from the back end right. So that is the assembly house down to the modules and the end user device manufacturing. So we work with them, we work with IDMs, we work with OSATs, we work with even the foundries, from the very beginning of design all the way to their high volume manufacturing.
Francoise von Trapp:Okay, you talked about mostly being in the advanced packaging space. Can you specifically talk about your semiconductor portfolio?
Rose Guino:From different material sets, whether it's wire, bond or advanced packaging. We have material offerings. In particular, for advanced packaging. We have different types of underfills. You're familiar with capillary underfill, which we're going to talk about. We also have the thermal compression bonded non-conductive paste and the non-conductive film. And within wafer level packaging, we have different films for processing. We also have liquid compression mold and when we talk flip chip BGAs or 2.5D architectures, we also have the lid or stiffener attached materials and developing thermal interface material. So you see, we're a one-stop shop where we can help you with your packaging needs.
Francoise von Trapp:From what I'm hearing, it sounds like Henkel's material sets are really focused on the reliability and thermal aspect of the advanced packaging process, not so much the interconnect.
Rose Guino:Correct, we don't do wires. It is the integration, so assembly of those, the protection of those.
Francoise von Trapp:So now, what are you seeing as some of the latest trends in advanced packaging that's driving your development and your decisions?
Rose Guino:A lot of cool stuff continues to evolve. We hear AI, machine learning, now quantum technology, all high-performance compute, and that is mainly driven by us, the consumers. We want more, more data, faster and reliable gadgets and reliable gadgets In our materials. We have to make sure that we enable all these architectures and everybody's doing their own way 2.5D integration there are different interfaces, different materials being combined and we need to be compatible with all of those. As an underfill, we need to make sure adhesion is good in all different surfaces you have metal, silicon and inorganic and in terms of reliability, whether it goes in the car or outer space, we have to make sure our material continues to support the interconnects.
Francoise von Trapp:So would you say that in some cases, the materials that you work with on for your customers are really some of their secret sauce?
Rose Guino:Yes, of course, the design right, that's the hardware and the software, but to make it reliable and for them to put them together, that is with our help.
Francoise von Trapp:Right, okay. So now, as the industry evolves and we see more high performance, compute and processes are changing. We're seeing, like thermocompression bonding, maybe at some point giving way to hybrid bonding. How are all of these advancements changing what people require from their materials?
Rose Guino:Depending on the spaces that we need to fill. That will dictate what materials we put in the CTE mismatches. We need to make sure our material will help with mitigation of those thermal expansions. The interfaces also I mentioned right. We have to make sure all are intact and held together throughout the harsh testing or the real life harsh environments.
Francoise von Trapp:So, for instance, substrate materials we're hearing a lot about glass core substrate. Is that impacting your materials development?
Dariush Tari:Yes, that would definitely. That means that interacting with different interfaces dealing with different CTE mismatches, thermal effects would be different, moisture effects could be different. So changing materials is not as simple as it may sound. That changes a lot of development for us as well. That means that we have to take into account the challenges that arise with that material change. In this case you could have fracture, for example, happening in the substrate where you didn't expect it before, or the load transfer will be very different. Those are the challenges that arises with these materials and combinations of materials, because we are putting more brittle materials with different CTE mismatches next to each other and we're creating these very complex packages. Of course, the features in the packages are getting finer and finer and that creates more stress concentration. Maybe more brittle materials are involved. So there's a lot of things we have to go through to design these new materials and develop them.
Francoise von Trapp:Do you think materials are now becoming more important than ever then? Because, as feature sizes continue to shrink, it really comes down to what you're building them with.
Dariush Tari:Yes, that material is always part of it and it plays a role. Of course that means that optimized materials is becoming more important, because we are not designing material in a vacuum space. I imagine we are not just designing a cuff material to withstand the harshest conditions. New packages means that we have to do sacrifices at different points and there's an optimum that maybe is not the best for cuff but maybe for the low dielectric material at the end of line, that maybe it has to transfer the load, manage the load at different locations a little bit better. The interfaces with that maybe it has to transfer the load, manage the load at different locations a little bit better. The interfaces with the solder maybe has to be managed better.
Francoise von Trapp:And when you say cuff, you're meaning that's CUF capillary underfill right.
Dariush Tari:Yes, the capillary material is correct, okay.
Francoise von Trapp:So I was thinking about this how long does it generally take to qualify a new material into the advanced packaging process flow?
Rose Guino:Yes, I think it depends. If it's a minor tweak from an existing, it could be fast six to a year. It's totally something new and it will require a series of optimization, fine tuning, and it could take more than a year to do Okay.
Francoise von Trapp:Let's talk about what it takes to actually develop a new material. Now, traditionally, if a customer came to you asking for a new underfill material, for instance, how would you approach the development process?
Rose Guino:Traditionally we'll get their requirement and most of the time they'll have a specific material property, target, TG, CTE, modulus, and we'll go down into details. What are the gaps, the pitch, and that will help us design what feature sizes, for example on the fillers to use, and in some cases they'll require thermal conductivity for heat dissipation. So then when we get that, our product development chemist will start the formulation until they get that targeted material property set formulation. Until they get that targeted material property set, Our team, application engineering, will then do the actual physical test on our test vehicles, flip chips, chip on wafer and then see and get that experimental data, most especially on reliability. So that will go through a series of cycles until we get that final golden material and then sampled to the customer.
Dariush Tari:And, if I can add about the CTQs, maybe the flow is something that you have to look at. The reliability cycles, crack resistance of the material Maybe the customer has a preference to that, maybe they want to manage stress in their devices could be an aspect of material sustainability or material homogeneity after cure, keep out zones, modulus and CT, as Rose mentioned. So there's a lot of aspects of material that we have to optimize for when we're developing new material.
Francoise von Trapp:Dariush, your role is in modeling. Can you explain a little bit more about how you use modeling to develop capillary endophyll?
Dariush Tari:Yes, so modeling allows you to basically study a device before you actually make it or study a material before you actually make the material. In this case, if there's a CTQ, then they ask for a modulus temperature profile. You can try that in modeling in the models that is available to us with the test vehicles that we have internally, to understand how the stress will change as a result of that CTQ, and that helps us to understand the customer requirements a little better and find the basically design space that we have to work in and optimize the material in that space and also modeling. One other aspect is that it can help our customers. Of course this opens the room for discussions and collaborations a lot earlier on than if we wait and make a material and send it to customer for testing and receive the results. So this iterative process is unlocked by power of modeling and this is something that we benefit from, both us and the customer.
Francoise von Trapp:Are these actual physical models or are they simulation software models?
Dariush Tari:It is simulation of the real physics, so it's a physics-based modeling. Usually, we use a method such as finite element modeling for structural analysis, and it's an example of it.
Francoise von Trapp:So you actually model the device plus the material.
Dariush Tari:Correct the combination, as is in the design. We have internal devices that we can model and that helps us to actually validate and correlate our models to the actual experimentation and that gives us more confidence of what we are seeing. In the modeling we try to capture all the complexities of the geometry of the device as much as possible. Of course, there are always simplifications in the modeling. In any case, model allows you to look into things that in testing you cannot go see. For example, you cannot see how the forces are transferred within different materials and which are the critical areas that you should be worried about. If you change the modulus here or TG there, how does it affect the fatigue in the solder ball now? So all that, basically things that gets unlocked by modeling- and so modeling is a new approach to materials development.
Dariush Tari:Not new. However, with more computational power we can advance our modeling techniques. That means that we can add a complexity to the mathematics behind the engine of our software. Basically, we can utilize more complex material models material models in this case. Imagine the math behind that simulation. Maybe before it would take a supercomputer to run a simulation like that, but now a normal you know commercial cluster of computational nodes can give you results in a few days or a few hours, depending on what you're solving for.
Francoise von Trapp:So we've been talking a lot about lately about digital twins and how digital twins are being used at the factory level or even used to design advanced packages. Is this similar to a digital twin approach?
Dariush Tari:It could be the input to a digital twin. So imagine if you have a physics-based model. It usually takes longer time to run than a digital twin, because when you think about a digital twin you're thinking about instantaneously getting the output from the model. But in this case a physics-based simulation can be used as points of input to calibrate a digital twin. That means that with deep learning and neural networks, you can fit the output from a physics-based model and after you're done with the calibration of a meta model or a deep learning neural network that allows you to instantaneously interpolate between the points that you have given the neural network, in this case as input. And also an interesting point is that with digital twins you can also use experimental data as input, and so if you have physics-based modeling and you have on the other side these experimental points, you can combine them and make hybrid models. So things that you cannot actually test you can simulate, and then that combined basically creates new opportunities to see more advanced behavior in that design space that you're looking at.
Francoise von Trapp:So can you even do like reliability testing in the simulation.
Dariush Tari:Yes, you can do. There are methods for fatigue and fracture that you can do in the simulation. They tend to be very expensive. There are approaches that help us reduce the cost. For example, we can do the technique called sub-modeling. That means that you capture the details of complexity of your device in a sub-model, the boundary conditions of which comes from a more simple model. So you run the simple model, you get the boundary conditions that you need for the next model, which is called submodel, and that submodel really dives into the detail of complexities that you have in the device. And that allows us then, for example, to model a crack, crack propagation through the material. Let's imagine a low-key material at the end of line and then you can maybe put a crack there and see if the crack grows or not.
Francoise von Trapp:And then you can maybe put a crack there and see if the crack grows or not. So how important is it to work simultaneously along with the package designer?
Dariush Tari:It is very close because the package designer. Of course they work on the design, but at the same time, from my point of view, they need to run simulations on their side to come up with design features that manages the stress and reliability and aspects that they're concerned about in their design. What they need as input to their models are material models. Basically, material models are a mathematical calibrated model that describes how each part of this simulation, the materials inside the simulation, should stretch and deform while you're loading it. Imagine if the temperature is changing. How should the cuff soften with temperature? What's the modulus at each point of geometry? That's given by characterizing material, calibrating material model, and that's why it's important that a modeling engineer from the device design perspective works with a company like Henkel that understand the material space very well and to understand material modeling and material characterization very well.
Francoise von Trapp:Yeah, because I would imagine doing those things in collaboration and designing the device and the materials used can make it possible to do things you didn't think you would be able to do, For instance, if you were just designing and you didn't know that there was a material available that might help you achieve the performance goals.
Dariush Tari:Yeah, in this case in modeling you can try any idea and failure cost basically is so low you're missing a few days of simulation time as opposed to in prior. You have to come up with techniques of assembly, actually how it works in the manufacturing, and then run it through a month of testing to see whether it would fail or not. But that, then run it through a month of testing to see whether it would fail or not. But that instantaneously knowing changing the modulus here or there and material properties, how does it really affect? Is it in the right direction or not? That is the insights you get with this modeling techniques.
Francoise von Trapp:So what is this approach done for the materials development cycle versus traditional approaches?
Dariush Tari:development cycle versus traditional approaches. You can do techniques such as what we call virtual material modeling. We can assume the properties and build the material model with those assumed properties, instead of actually making a material. If, imagine, your product development for the material development side has an idea and says okay, if I have a TG like this, this modulus range, would it help or not? Well, you can do virtual material modeling or calibrate the mathematical formulation, basically, which I described before, to these imaginary properties, use this and even give it to the customers, and we try this. This mimics what would be the outcome.
Dariush Tari:Of course, one thing you have to take note of is that the assumptions have to be physically possible. You cannot make any combination of properties because you change something in the formulation, something else may change opposite to how you want. So, considering that, which is exactly what the expertise of a formulation engineer is, you can then come up with virtual material models that then you can put it into simulation. Interestingly, you can even run the simulation multiple times. Let's say you create outputs using the virtual material modeling. Imagine you cover all the design space that you can play with in the material design space. Now you have inputs and outputs. Then, with that input and output, it's a perfect scenario for calibrating a meta model or neural network or a response surface method, and that allows you to even run these cycles faster and understand how these inputs will affect the output.
Francoise von Trapp:So all of the new developments in AI must be impacting this in some way. Are you using AI for?
Dariush Tari:your models.
Dariush Tari:Yes, we started to see AIs actually coming in and helping in many different ways.
Dariush Tari:You can see that in the software development side the tools we use, of course, there's more and more user assistant features in the softwares.
Dariush Tari:There are GPTs that are released that help you understand your questions better, ask the questions better. Also, the trends that I see with AI is that across multi-physics, the model starts probably to get more vertically integrated, going from thermal to structural, and so these products are emerging. And with the acquisitions that are happening in the space of software for simulation, you can see that they are packing these solutions more closely and bringing in as a harmonious ecosystem for design of semiconductors. And the other aspect is that robotics a dance of robotics, help us collect test information in scale, in larger scale than a test lab engineer could do with better quality data, and that, combined with the power of neural networks and machine learning, faster simulation coming from the more advanced chips unlocks more simulation power and combined with this mass test data, mass simulation data, that allows us to build even larger, more complicated AI, machine learning driven models in the future, and that's something that we're looking at for the future.
Francoise von Trapp:Okay, Okay, so we've been talking in a lot of like high level generalizations, but one of the things we did want to talk about specifically is how you're using these models to develop capillary underfill. So let's start a little bit by explaining what capillary underfill is and what it does.
Dariush Tari:Yes, great question. So capillary underfill cuff is a material that we dispense under the dye to predict the solder balls. Basically, without this material we have a lot of shear experience through the reliability cycle, especially with the solder balls, because of the warpage. In the package you have CTE mismatches and that really causes parts of your device expanding with temperature more than the other parts. Like dye, for example, has a low CTE but substrates organic substrates have larger CTE, then that causes the package to warp and during this warpage the substrate will transfer the load through the solder balls to dye. That means that they have to shear a lot Each cycle of reliability.
Dariush Tari:These materials are accumulating damage inside them, microscopic damages, and eventually this damage will cause fatigue of the problem and this is a phenomenon well known in mechanical engineering. So cuff materials we dispense this as a liquid form and then we cure it. It forms a radius around the dye and goes under the dye, basically surrounds all the solder balls and then after the cure it stiffens. Of course it's a polymeric, usually epoxy-baked material with some fillers inside, and it actually helps carry the load that should be transferred from the substrate to dye and that saves the solder balls. That allows them to survive the thousand cycles that you expect them during the reliability cycle.
Francoise von Trapp:So there are different types of underfill and, rose, maybe this is a good question for you when would you choose a capillary underfill over, say, a liquid mold underfill? When would you determine which to use?
Rose Guino:Right. So, as you mentioned, there are many different types. You have your molded underfill, the capillary underfill and the pre-applied underfill. So it all boils down to what interconnects and what types of interconnects and the spaces that you want to fill, what types of interconnects and the spaces that you want to fill. Dhirush mentioned solder balls. Now we've seen copper pillar with solder cap, You're seeing smaller versions of them, the micro bumps, and even down to copper. To copper, you know whether it's hybrid or just. You know pure metal bonding.
Rose Guino:In terms of the assembly of the chip to the substrate or to an interposer or another dye, if it's via mass reflow or if the approach is form the joints first before underfill, that's your post-applied. So that would be your capillary underfill or your molded underfill. But if you want to protect the joints while you're forming them, that would be your pre-applied. So that is where your thermocompression NCP or NCF come in. So from, let's say, copper pillar or solder balls that are still wide gap and not so tight pitch, reflow works. But as you saw from the evolution into the TSVs and you have your die stacking, your HBMs, I think your gaps now are less than 10 micron and the pitches are less than 30 micron. The pre-applied may work or is preferred, whether it's the NCP or the NCF for stacking, and we're able to use molded underfill for that application. So again, you have many choices. The goal of these underfills are the same it's to protect. They're all reliable and it's now a matter of technical and economical.
Francoise von Trapp:Basically, it depends on how high density the device is.
Rose Guino:Correct. I mean, in theory you can use all, but do you really need it? For that simple device, right, a simple cuff or a simple muff might work, but for that high end, where you really need extra protection and assembly is critical, then you might go with the higher end version, okay.
Francoise von Trapp:And Henkel provides all of these options.
Rose Guino:Yeah, so we do have the capillary underfill, the NCP, the NCF and now the liquid molded underfill.
Francoise von Trapp:When you're working on developing new advanced capillary underfills. Advanced capillary underfills how does working with Dairush's modeling team impact?
Rose Guino:your ability to provide your customer with what they're looking for. Yeah, good question. Dairush and his team, the modeling team, really helped us develop faster and understand our material better, especially in terms of reliability performance. We actually recently launched a new underfill for advanced silicon nodes and, with the help of the modeling team, we've showcased to our customers why this material is good. It shows where the stresses are and how it helps with the reliability of those interconnects.
Francoise von Trapp:Yeah, so what do you want listeners to understand about the benefits of modeling?
Dariush Tari:So I want them to understand that through modeling we accelerate the development process. It unlocks opportunities for early engagement. It basically creates an environment between us and the customer who work on a project earlier with a more chance of success. Those are the benefits of our modeling work.
Francoise von Trapp:So, as we're wrapping this up, what would you like the audience and customers to know?
Dariush Tari:We're always happy to hear from them. We would like to engage with them, especially on the modeling side, as soon as possible, when they're on the modeling side, as soon as possible when they're in the design phase, because that really, based on our discussion, unlocks opportunities to develop and provide them and serve them better, set them for the success. We have good material understanding of semiconductor space. Also the testing and characterization. Maybe the customer requires this testing and input information for the design. We have very good, equipped and knowledgeable people in Henkel for that. Also, material modeling we are experts in materials. Of course. We understand material models and customer needs. Based on that, we also do modeling. We understand the pain points and what they need to improve their models. We are very eager to actually collaborate with our customers, co-develop these methods, unlock new challenges, basically, and new opportunities for co-development. That's my take.
Francoise von Trapp:So engage early and often.
Dariush Tari:Yes, early and often.
Francoise von Trapp:Rose, do you have any final thoughts?
Rose Guino:Yes, I think Henkel. We've been working with many of our customers. We hope that we continue to do that as we address new challenges. We have global presence, so wherever they are, we have Henkel there and we have innovation labs, application engineering labs to help with feasibility, proof, of concept right. So from beginning we can help and we can do it in parallel to test. They're more than welcome to visit our labs and they have I mean, some of them have utilized that Continuous collaboration, development, understanding failures and even helping with their high volume manufacturing, how to improve UPH and all that.
Francoise von Trapp:All right. Where can people go to learn more?
Dariush Tari:At Henkelcom they can find the material portfolio and in the adhesive section portfolio and in the adhesive section we have mentions of semiconductor space and that industry.
Francoise von Trapp:They can go and find out about our materials okay, well, we will be sure to put links to that in the show notes. We'll also put links to each of your linkedin profiles so people can contact you directly. Thank you, everybody for joining me today. It was a pleasure, thank you, thank you very much.
Francoise von Trapp:This wraps up Season 4 of the 3D Insights Podcast. We hope all of our listeners enjoy a relaxing holiday season and we invite you to catch up on the episodes you may have missed by subscribing to the 3D Insights Podcast on Apple Podcasts, spotify, amazon or wherever you get your podcasts. We'll be back in 2025 with all new episodes. Until then, thanks for listening. There's lots more to come, so tune in next time to the 3D Insights Podcast. The 3D Insights Podcast is a production of 3D Insights LLC.