Mean time particle failure vs Defect Count

As a CVD process engineer one has to accustomed oneself with the habit of looking at trend which is been supported by the SPC system. The Western electric rule is one of the most common method to identify certain drift in the process. This trend is always vivid from process with normal distribution for instance a thickness of the thinfilm measurement.

A thickness measurement trend is easy to look for since it has a certain causality characteristic whereby changes in the trend might be attributed to a chamber maintenance that is taking place. You can see a thickness measurement gradually increases after a certain number of wafers processing which is caused by the seasoning layers that the equipment actually sees and eventually changes the chamber impedance characteristic.

But now, how do one actually sees a trend in a Defect count of particle failure in our SPC trend. I have been thinking of this problem for sometime whereby finding a way to see 'trend' in random distribution. One of the problem that I see is that the count failure sometimes does not lead to any significant information since a failure of 20 counts might just be the same to a 15 counts failure since the number of particles drop is also a random number as the particle might fall 15 this time and would drop more count in the next failure event.

Hence after thinking for sometime I deduce that the best method of capturing a particle trend is by implementing a binomial distribution method. A pass and failure might be the best way of anticipating a failure trend. Thats how I came about with the concept of mean time particle failure. The mean time particle failure actually captures the frequency of the failure itself. The hypothesis is that as the failure of the equipment becomes imminent hence the frequency failure itself regardless of the defect count. After developing the algorithm and testing this method on our particle SPC, i was able to see a trend in some of the equipment as it nears a catastrophic failure or as the system actually nears a certain Preventive maintenance wafer count. There are further work that can be done on the system such as by developing a trend chart between this failure and maybe we can see some different slope magnitude between those failures. A bigger slope might indicate a faster equipment failure.

Engineer vs Scientist

Something crosses my mind a few days ago while thinking what is the function and responsibility that an engineer has to carry out which differs from a scientist. Are engineers required to publish their findings similar to what scientist do. After giving a serious thought in the matter, what I can conclude is that engineers are basically hired to solve problem that the company faces. For instance in the semiconductor industry we might have problem with the capacity of the wafer that is required to be moved through a certain processing steps. As an engineer we are required to provide plausible solution to the impregnable problem without the company to have to invest in the additional tools in order to support the higher move requirement.

A scientist is someone who actually delves into a known problem in the society and hence finding a solution to the problem that the society is facing. So what a scientist normally does is that he or she will eventually perform a literature review in order to ascertain whether the problem that he or she is planning to work into has not been studied by anyone else or he might be looking for an alternate method in finding the solution.

So as engineer we are been paid to find solution to the problem that the company is facing and while we do that maybe find some interesting aspect in our path towards the solution so that we will not get bored with what we are doing in the company.

Controlling your tool better

Semiconductor industry has seen and explosive growth in terms of its market capitalization and its sales volume. The progress of the technology development has been spurred by the ever increasing consumer market trend.  As more consumer electronics has been designed for the ease and benefit of humankind, technology trend has so seen an aggressive push towards more circuit density per a single chip which has been predicted by Gordon Moore which been recapitulate in his Moore's law.

The shrink in the circuit density has also prompt in a more sensitive devices which requires a more stringent control on its semiconductor process. If for an older technology a defect of the size of 0.2um will not cause and detrimental effects on the overall processing circuit, now with a technology node of 65nm and beyond this defect size has catastrophic effect on the chip which translate to yield loss on the wafer.

Hence ushered the world of FDC or Fault Diagnostic and Control. FDC has provided the semiconductor player with a tool of controlling the processing tool by monitoring the processing parameter. This will help process engineer or the equipment engineer to stop the tool in the event that that the tool detected a drift before more wafers are exposed to the process. Previously control of the wafers was done based on a post processing data such as thickness drift or based on the qual wafer monitoring. This creates a big potential impact in the event that a major excursion happened in between the qualification wafers.

FDC has evolved from merely a tool to detect simplistic drift to conducting multivariate analysis to see correlation between multiple variables. By having multivariate analysis capability we are able to monitor multiple parameters at once and reduce any false alarm that might be triggered due to false alarm in the condition of a single parameter drift. There has been extensive study conducted on the modelling of tool parameters in order to determine the health of the tool and detecting abnormal drift and stopping the tool before excursion even occurs.

FDC is a wonderful tool only if a process engineer or equipment engineers utilizes it well by ensuring proper parameter are been monitored and specific rule are been generated.

Determining your net worth to the team

In any organization, moving forward as a team is important as it will the goal set is achieved together in concert. A disarray in the team will lead to a breakdown in communication and make it harder for the leader in the team to direct and bring the team towards achieving the goals set by the management.

Being part of a team has elevated my reputation within the team as someone who carries certain experience and knowledge. But how does a leader appreciate his member that actually brings with him a wealth of knowledge and experience. Experience is something that is commonly sought over by an employer. A person of experience might not be able to complete the task or project that he has on hand due to the fact that he is normally preoccupied with assisting and guiding junior engineers in troubleshooting the problem that they have. Having experience means that the solution that we are seeking will eventually find its way faster. So how does the management evaluate this type of guidance that an experience engineer actually brings to the team or company.

What i sincerely believe, a manager should set a certain goal and compare on how an experience engineer assist in finding a root cause towards a certain problem that the team encounters and measure how quickly the problem is able to be resolved or the plausible path that the experience engineers actually propose that brings the teams closer to the solution that is required.

Understanding Thinfilm radiation mechanics

HDP chambers has been extensively use in the area of dielectric deposition. Among its known advantage compared to the PECVD chambers is that it offers better gap filling capability by providing an insitu dep and sputter cycle during the whole deposition process. By having this cycle the process is able to ensure that the oxide valley or trench remains open by sputter and removing a portion of the film before the two opposing edges remain open throughout the whole process. By having this insitu sputtering on the wafers causes the wafer to heat up and again monitoring the temperature of the wafers are critical to ensure we do not exceed the stipulated thermal budget.

Then comes the radiometry which uses the radiation from the wafers as it gets heated up transmitting radiation through the process chamber. The radiation emitted from this heating process can be use to establish the actual temperature reading of the wafers. In the NTM (Non Contact Temperature Measurement) system that we used, it actually utilizes two distinct signal in order to decouple the two information required to obtain an accurate reading of the wafer temperature. One of it is the wafer emission and another one is the emissivity of the wafers itself. The emissivity is determined by emitting a signal of photons which actually bounces or reflected back by the wafer surface and hence derived the information of the wafer emissivity.

Since the emissivity is a function and direct correlation of the reflectivity of the wafers itself then, the number of the photon reflected back by the signal shows how reflective the wafer surfaces are and hence the emissivity of the said wafers. The emissivity calibration is done by moving wafers and heating the wafers up with polished surface to indicate an emissivity value of 0 while when the wafers are transferred out this situation is considered as a blackbody condition or an emissivity value close to 1. But the thing that bothers me is running the wafer at an elevated temperature will surely induce a different emissivity compared to when the plasma is turn off. How are we able to plot a straight line based on this two condition for the emissivity calibration. This is something that I need to ponder through in order to understand further the mechanics of how does this radiometry actually works

Analyzing Data Quickly

Data analysis is one of the important aspect in a daily engineer's life. Whether we like it or not, most probably we are been tasked to analyze certain data and elucidate the meaning behind the summary of the data at hand. In the life a CVD engineer some of the data that we have to analyze is look at the how does the performance of certain process are compared to an existing process that we already have. Here statistical analysis plays a critical role towards establishing a conclusion whether the new process that we are qualifying is having better performance or lesser compared to our current baseline process.

The most common statistical software that I have used is JMP. It offers varies analytical and statistical tool which assist me in establishing correlation or making a statistical comparison whether between two processes. The software helps us to extract information and hence assist in providing a stronger argument in bringing to our final decision.

Therefore I strongly believe that an engineer should be well verse in the statistical tool so that we are able to make decision better by looking at statistic rather than relying on our gut feeling.

Temperature mismatch due to different Photo Equipment

Temperature measurement in the semiconductor industry has become critical in many processing area. The need to ensure that temperature budget is maintained within the required range necessitates the necessary instrument to monitor the temperature distribution in the semiconductor equipment. One of the critical area that requires close temperature monitoring is the backend dielectric deposition step. The importance of temperature monitoring at this step is due to the fact that a shift in the temperature might induce a change in the metal grain which might affect reliability performance of the circuit or might even lead to metal distortion which will eventually lead to wafer scrap incurring loss to the company.

One of the most popular and common used temperature probe uses the optical pyrometry method. In the optical pyrometry method, the system utilizes infrared radiation that is been emitted by the substrate to determine the temperature of the wafers. This phenomena was then discovered by planck and eventually lead to the planck theorem which stipulates that electromagnetic radiation emitted by a black body in thermal equilibrium at a certain temperature. Basically what it says is that the amount of radiation emitted by a a black body has a direct correlation with the temperature of that body.

From this law, optical pyrometry uses the radiation from the body to actually determine the temperature of the body based on certain wavelength which normally is near the infrared region. Further improvement in the temperature measurement eventually creates an emissivity measurement reading of the wafer substrate as well. Since the emission that is been emitted by a certain body is relatively link to the emissivity of the wafers itself. Some of the factors that affects this emissitivy is the doping on the wafer, the roughness of the wafers which will lead to light scattering and thus creating multiple reflection of the radiation and also the distance and cavity between the wafers and the temperature probe. The temperature probe normally is made of sapphire rod which is able to transmit signal with minimum loss. The optical pyrometry reads the wafer emission which is the function of the wafer emissivity and temperature based on the wavelength.

One of the problem that we encountered so far is that we are seeing a temperature mismatch between wafers that is running in DUV compared to wafers that is been processed in ILINE. Further experiment shows that by removing the BARC from the DUV process we were able to match the temperature during the oxide deposition during the IMD layer. But what we are still yet to decipher is how does the BARC is able to affect our subsequent temperature since eventually the BARC is supposed been removed during the etch process and should not leave any trace of the layer before the IMD deposition. If the phenomena that we are seeing is true then the only reason that the temperature is going to be different is because the heat transfer from the residual barc layer causes a different temperature reading.

Hence what we are going to do is to try to understand the difference between a BARC process and a non BARC process. Few of the suspect that we have are, the BARC actually changes the chemistry or surface of the metal itself. To verify this a surface roughness measurement will be conducted to look at the surface of the metal after post etch process. Another possible hypothesis is that there is a residual BARC that is shifting the temperature and causing the radiation to be different compared to the non BARC process.

This is an interesting problem that I would like to understand. Hopefully i would be able to find the answer to this perplexing problem before the year end....

Solving that elusive problem

As an engineer, the most frequent work that we have to work through is solving problems. Engineers are there to find solution to an existing problem or to work around a problem and finding a better and improved ways dealing with them. I have been in the industry for almost 12 years now and have seen numerous issue that requires attention and solution. Some of the problem can take years to solve and some of them might just take a few minutes. The key to all this is i believe starts with the problem identification. Once we are able to find that, then eventually the solution is somewhat easier to conjure.

So what is the golden rule in identifying a root cause, what I sincerely believe is that we have to look at the problem and find any theory or hypothesis that can actually explain the problem that we have on hand. In doing so what I normally do is to browse through the internet finding similar problem that others might have already faced. From there we at least has a foundation theory to back our claim and able to scientifically explain the phenomena that we have seen in our problem.

There are numerous ways that people do to actually brainstorm to search for ideas relating to the problem at hands. The most popular methods i believe is the fish bone diagram. What the diagram does is that it helps us to visualize the plausible causes based on a few category. Looking at those category will help us to generate ideas which eventually might lead to the actual root cause.

Most of the times we have to work backwards by looking at the symptoms that our problem is showing to us, much like a doctor diagnosing a patient by asking them the symptom. From all these symptoms we will be able to deduce what are the cause that actually fit all these symptoms and rejecting the other hypothesis that does not generate all the symptoms that we see.

Sometime we just have to be stubborn and determined in order to not let our effort diminishes away as time flies by.

Writing that matters

As what people have said, the spoken words reach thousands but written words last eternity. We have been taught to write since at early age as a way to deliver our message across. As an engineer and in particular doing research, we do have to try to articulate our research ideas so that our findings are shared among peers and validated. Writing is actually an art with its own nuance. Finding the perfect words will paint a beautiful picture like a painter selecting their brushes and the mixture of colors. As an engineer I have always wanted to be able to write good technical papers. So where do one start in order to achieve this vision.

Looking at the vast resources that is available in the web, what I can say is that in order to write here are some pointers so that we are able to conjure a meaningful writing that brings benefit to our targeted audience.

First, the concept of memory muscle, the more that a person is trained, the more well verse and skillful will he be in a subject. The same applies for writing, in order to write a good scientific article, practice is something that we have to pay attention to. I do remember the time when it is very difficult to actually deliver a presentation in front of a full audience. As time goes by, we are been programmed to learn and connects those neurons in our brain and generate the perfect synapses which actually been manifest as our personal skill.

Second is a junk in and junk out, you have to expand your reading horizon so that you have a wealth and depth of information at hand. The more you read about a certain subject, it will be much easier to spew out letter of wisdom and thought. So always spent time to explore and dwell deeper into your subject. The more you read, the better your brain in keeping meaningful information that can be used latter when you are writing about a certain subject. Trying picking up methods or style from other writers so that you can pickup new words and how to use them in your future writings.

The third, begins with the end in mind, a good writers should always be able to envision to the end of his or her story, the same thing applies in technical writing, you should already be able to visualize how do you want to bring your reader to the conclusion that you are going to make at the end of your writing. Organizing your thought by scribbling down the important facts and the arrangement that it will follow will help you to entice your reader to stick through the whole story that you have written for them.

So to recapitulate, in order to write perfectly start typing those word and start writing....

Becoming an expert in your field

Albert Einstein, Gordon Moore, Michio Kaku, Stephen Hawkings are a few names that is been considered as expert in their respective field. The innovation that has been brought by those name has inherently ushered a new era in science and technology.  In the electronic industry you can be sure that any engineer would have heard of Moore's Law either during the tertiary education or in the initial period of their career. I believe every engineer would have dream to become an expert in their respective field.

So how does one considered to be an expert. This is some of things that crosses my mind when word expert comes to being. An expert normally have a in depth understanding of the foundation theory that govern his or her work. For example a good electrical design engineer should already mastered the Kirchhoff law which is an essential foundation in any circuit design. A person who is an expert in their field should know by heart all the foundation theory that revolves around his or her field. Hence when posed with a theoretical question an expert should be able to explain and deliver the answer instinctively

An expert is someone that is known in his field by the number of technical papers and researchers that he or her has published and he is more than capable of substantiating his research with a resounding proof. As a CVD engineer the foundation theory that governs my field i guess would be the physics of the plasma and the chemical reaction that follows, how heat is been transferred and the mass limiting or heat limiting condition are an important aspect that an expert in CVD should be aware of. Plasma is been generated by the energy that is been provided by an external generator which provides the necessary energy for the atoms to reach its activation energy for the electron to be released and with that the generation of photon which varies in terms of color which depends on the plasma material itself. An expert in the field should be indicated by the number of citation that his or her technical papers is able to garnered which means that his research does carry a lot of weight among his peers in the same field.

This is some of the things that i can think of, and it will always been my vision to encapsulates my knowledge with the foundation theory that revolves around this field so that one day i will be able to be listed among the expert of our field today.

Particle Issue

As the semiconductor industry move towards ever shrinking it gate size to smaller spacing, the technology hence becomes more sensitive towards defect contaminating the wafer surface. In the dielectric deposition area, any location that is been affected with particulate will eventually lead to yield losses which eventually translates to the loss in the company revenue due to the customer return or wafer compensation.

Particle contamination has been one of the biggest challenge that any CVD engineer will ever face. To make matters worst the problem of particle contamination is normally a random event. Many factors can actually lead to the generation of this particulate contamination. From the equipment perspective, any degraded oring or parts lifetime normally are the two most common factors contributing to the particle generation inside the chambers. Although a certain lifetime has been set, there is more often that not that there will be some erratic behavior on one of the parts that is not able to reach its lifetime cycle. Other common factors might be due to the misalignment of parts causing two separate bodies that is been separated by a small distance scratches each others and hence generating those particles.

In the aspect of a process induced defects certain studies has shown that some of the factors that is likely to induce unwanted particulates are the regions of the gas flow. This studies indicates that if there is only a small gas been flowed then its likely to generate a localized entrapment area where the particulate will eventually agglomerate. Therefore a certain threshold of minimum flow is required in order for the process to be able to stay away from this process regions. Another factors that can contribute to particles generation are the temperature itself, certain study conducted reveals that the lower temperature of the wafers the more likely that the particulate will be attracted towards the wafer itself, hence running a process with lower temperature will create a process that is vulnerable towards particulate generation.

Moving forwards, there are a substantial amount of research been conducted to study the behavior of the particles in CVD chambers. Nonetheless it remains a very hotly debated subject since process to its equipment pose a different condition and region which guarantees a good and stable particle performance. As for myself, learning to develop a new particle-free process has always been a dream that I have, but so far the only way I can see through this development is by running many wafers to determine its failure rate since defectivity is something more of the probability problem rather than an empirical problem. Its rather a tossing of a coin problem rather than digital problem with a determined on and off switch.

Being inspired

The mazy road ahead has always been tough to unravel with many obstacle that might cause us to trip over. In the line as an engineer i have sometimes encounter problems that is very difficult to resolve, but one of the inspiration that i got is from watching aircrash investigation series from discovery channel, it is really inspiring to see the tenacity of the investigator in finding the rootcause which might take them at least a few month or years to find the rootcause/s, in my line as semiconductor engineer our challenge is to deal with something that is very small and minute, a sharp eyes are required to finding the small failure that is not visible to our naked eyes. Tenacity is something as an engineer we have to build within ourselves, because finding a solution might take you through a long journey with no guarantee of finding the end of the trail

Making a significant contribution

As an engineer, i have always dreamt that someday i will be able to mark my name in the technology history timeline as someone who actually made a noteworthy contribution towards the society and towards the advancement of our existing technology, but working here i have always been bogged down but trivial issue that is important to the company but does not have a good milestone in the scientific progress, alas eventually i will be stuck with doing routine things that will not justify any significant contributions towards this scientific world, again when we look at those people who are actually able to mark their name in the history of science i would tend to believe that only geniuses are able to do such thing, to come out with something significant, most of them are either working in research institute or research company that pays a lot of emphasis towards developing something important and significant, hence working in profit oriented company will not give you the luxury of extending your contribution further than the company's goal which is making a lot of money

Between two goals

Looking for that breakthrough

2016 has ushered a new aspiration on my behalf, as an engineer it has always been my goal and aim to achieve a breakthrough something that i believe all engineers should aspire to, looking at the problem that we have in our process it has always been my goal that i can achieve a major reduction in our particle problem in our IMD layer tools, we have received feedback from our tool supplier that the low gas flow region that we are running at is susceptible to high particle formation, from there on i have set to optimize the process by running at a higher gas flow, but the problem that i have is after running a few runs it is very difficult to statistically show that a higher gas flow provides a better process in terms of particle performance. So for 2016 i would set a goal to setup a method on how to quantify a process that has a better particle performance process

Challenges in wafer count and equipment correlation analysis

I have given myself a task of looking into the possibility of correlating the failure particles in our processing chambers with the number of wafers the equipment has processed, using a standard one way analysis shows that they is no coherent correlation between the two, when i actually review again the chart i do realized that there might be some factors that actually masked the analysis that was made.

1) The number of cycles each pm has gone through in the analysis of the equipment, when we do not consider this in our analysis what actually transpired is that we might assume that a repeating failure in a single pm cycles masks away the subsequent pm cycles correlation

2) Certain failure within a certain pm cycles might in effect contributed by another pm failure which might also mask the relevance of the analysis for that particular pm type

3) A repeating failure might actually indicates a catastrophic failure instead of giving you the actual lifetime of the pm itself

These are some of the things that came across my mind at this time

I will share further when i manage to have a better intepretation of the data at hand

A beginning

After graduating from a local university back in 2003, i was fortunate enough to be offered a job in a semiconductor company in a northern region. From there on i have worked in the same company and the same department since. Through the 12 years that i have been here, i have seen the ups and down of the volatile market in the semiconductor industry, the company so far has provided me with ample of oppurtunity for me to progress further acquiring new knowledge and enhancing my skill as an engineer, no one i guess would have thought that the volatile market in the electronic industry could eventually be felt by the oil and gas industry, this industry which was so often than not was considered the most sought after place to work with, due to its high paying salary and allowances. So here i am being in a wafer fabrication facility and focusing on dielectric department. The reason that i was thinking about blogging about my life's experience as an engineer with the hope that eventually someone out there might benefit from the things that i'm going to write and share

The journey has begun

The first step towards something bigger