MDZ Detailed Information

Oxygen & The Basic Problem

The control of oxygen precipitation behavior in all silicon wafer products is a crucially important component of the silicon wafer business because it is important to the successful manufacture of integrated circuits.

There is a degree of unpredictability in the performance of wafers in regard to the behavior of oxygen. Uncontrolled behavior is completely unacceptable. Control of oxygen concentration with the intent of controlling oxygen behavior is not sufficient to deliver optimum performance.

Device manufacturers are concerned with two things related to the behavior of oxygen in the wafers they use to manufacture their devices with:

  • First and foremost, they are concerned that no oxygen-related defect (precipitate) be the root cause of a measurable yield loss in their finished good. 
  • Assuming the first, necessary requirement is met, they are generally interested in utilizing the internal gettering (IG) potential of bulk oxygen precipitates as a yield maintenance tool.

In other words, not only should the silicon wafer itself not be a yield-limiting factor, but it should be a yield-enhancing factor for our customer.

Over the years, the silicon industry has collected a great deal of costly, excess baggage along the road to satisfying these two conditions.

SunEdison Semiconductor’s patented MDZ® process and silicon wafer is the ideal solution for the problem of controlling oxygen precipitation in all IC processes. MDZ® wafers are programmed for optimum, robust, and uniform performance. MDZ® greatly simplifies the use of silicon wafers while at the same time bringing a new, and until now, unheard of precision to their performance.

Conventional Approach to Internal Gettering

Oxygen & The MDZ Solution

SunEdison Semiconductor’s patented MDZ® process and silicon wafer is the ideal solution for the problem of controlling oxygen precipitation in all IC processes. MDZ® wafers programmed for optimum, robust and uniform performance. MDZ® greatly simplifies the use of silicon wafers while at the same time bringing a new, and until now, unheard of precision to their performance.

The source of all unpredictability – and hence complication – in standard silicon wafers can be traced to the difficulty in controlling all the coupled oxygen reactions which take place during crystal growth and wafer processing. MDZ sweeps this all aside by wrestling away the control of these reactions and giving it to a new, much easier to control component -– the vacancy.

Reaction-controlling vacancies are installed into silicon wafers by a rapid and positive RTP process which replaces the standard thermal donor anneal step. The installed profile of vacancy concentration effectively and robustly programs the wafer to behave in a certain predetermined way. The MDZ process installs a profile that insures optimum performance for all applications.

By sweeping aside all the oxygen-related complications which have stood for many years as a important barrier to the rational use of silicon wafers in IC processes, the MDZ process removes a huge number of oxygen related complications. The MDZ wafer is simplicity itself. It does not require tight oxygen concentration control to give high performance results; it does not require specific tailoring or tuning of silicon wafer to meet the needs of specific processes; it does not require additional costly heat treatments to insure denuded zones. The details of the crystal growth process is no longer important to oxygen behavior. MDZ simplifies the use of silicon to the degree that one specification -– MDZ -– covers all IC applications.

An MDZ wafer is one simple answer to what used to be a million questions. And it is the best answer.

Oxygen: Q & A

  1. Why do vacancies enhance the precipitation of oxygen in silicon?
  2. Is MDZ® really independent of oxygen concentration?
  3. Why is precipitation suppressed in the surface layer of an MDZ® wafer?
  4. What bulk precipitate densities are needed to ensure effective gettering?
  5. What are the sources of unpredictability in standard material?
  6. Under what conditions could the vacancy profile of an MDZ wafer be erased before it has time to nucleate the desired bulk oxygen precipitates?
  7. How can it be that the behavior of MDZ wafers are not dependent on the details of the application to which they are submitted?


1. Why do vacancies enhance the precipitation of oxygen in silicon?

The simple answer is that oxygen precipitates need room to grow inside a silicon crystal. They are larger than the silicon lattice and thus introduce strain as they grow. This strain is the main limiting factor to the nucleation of oxygen precipitates in silicon. Vacancies are effectively represent open space in the lattice and offer a means to relieve this strain.

A more detailed answer:

The relief of strain by vacancies is clearly the basic reason underlying the MDZ effect. But the details are more complicated. A more subtle explanation of the strain relief of the ‘magic’ effect of vacancies (oxygen-independent production of precipitates!) is that oxygen clustering, during the nucleation anneal, proceeds by the agglomeration of the vacancy-oxygen species (O2V). The nucleation rate is controlled by the species concentration Cv, and thus independent of the oxygen content. The nucleation rate is a rapidly increasing function of the species concentration Cv. Below some critical concentration Cv*, the nucleation rate becomes negligible.

In the range of low Cv, the agglomerating species are oxygen atoms, and the nucleation rate is then controlled by the oxygen content C. The produced oxygen clusters are strongly strained (unlike the clusters of O2V species). The free energy gain per an atom is m-w, where m=kTlog(C/COi*) is the oxygen chemical potential (COi* is the oxygen solubility) and w is the strain energy. The difference m-w is positive at the nucleation temperature (such as 650oC) but becomes negative at higher T (at 1000oC), since m is then considerably reduced. Thus the nucleated clusters will all dissolve during the precipitation anneal, regardless of their size. The only way for the clusters to survive is to relieve the strain by emission of self-interstitials. The emission rate becomes fast only if the clusters are larger than some characteristic size. This is considered to be the main reason for a long incubation. During the nucleation stage, the clusters need to grow to a size large enough for subsequent efficient emission of self-interstitials during the precipitation anneal. In the case of clusters of the vacancy-oxygen species O2V, there is no problem of strain: the space for a cluster is ‘carried’ by the agglomerating species themselves. Hence there is a negligible incubation time for this vacancy-assisted nucleation.

Return to Questions

2. Is MDZ® really independent of oxygen concentration?

Vacancies really do control the show in the MDZ business. Oxygen is important only as a "construction material". In CZ silicon there is always enough of it around. It never limits the process. MDZ really is independent of oxygen concentration.

Yet, at low oxygen concentrations (below about 10 pmma) some experiments may seem to show that the MDZ effect rolls off and that there is an oxygen concentration dependence. But this is not true. Simple treatments such as the NEC1 treatment (4 hours at 800°C followed by 16 hours at 1000°C) simply do not tell the true story at low oxygen concentrations. The reason for this experimental error is that at very low oxygen concentrations, the growth treatment (the 1000°C part) necessary to make the precipitates visible is not completely effective. To see the true picture (MDZ is completely independent of oxygen concentration) you have to increase the driving force for precipitate growth (go to lower temperatures) for longer times.

Just because a precipitate is not visible after a precipitation test does not mean it is not an effective gettering site (or for that matter a danger if located near the wafer surface). Quite the contrary.

Oxygen Independence of MDZ

Even at very low oxygen concentrations but beware detection limits

Return to Questions

3. Why is precipitation suppressed in the surface layer of an MDZ® wafer?

The surface region of an MDZ wafer has the same character as a Tabula Rasa wafer. A Tabula Rasa wafer is a wafer in which all the oxygen clusters "grown-in" by the crystal growth process have been dissolved by a brief high temperature wafer treatment. In the case of the MDZ wafer this comes for free during the first phase of the process. It happens on the way up to the process temperature - during the heating-up phase of the process.

The surface region of an MDZ wafer has the same character as a Tabula Rasa wafer. A Tabula Rasa wafer is a wafer in which all the oxygen clusters "grown-in" by the crystal growth process have been dissolved by a brief high temperature wafer treatment. In the case of the MDZ wafer this comes for free during the first phase of the process. It happens on the way up to the process temperature - during the heating-up phase of the process.

A Tabula Rasa wafer not only eliminates the possibility of oxygen precipitates originating at crystal growth but also – in all but extreme and highly unlikely situations – effectively suppresses the future nucleation of oxygen precipitates in process.

In doing so it relies on the phenomenon of "incubation". A long time (typically more than many hours) at low temperatures is required to produce oxygen clusters large enough to grow at a higher temperature.

A very important fact to keep in mind is that the time required to do so is not cumulative. Any high temperature treatment following a potentially nucleating low temperature treatment acts itself a Tabula Rasa treatment - erasing all that came before. The incubation process must then start again from zero during the next low temperature phase.

A more detailed answer:

The MDZ wafer delivers a "real" denuded zone in IC processes in the sense that there are really no precipitates in the near surface region – not just precipitates that you cannot see by common analysis techniques.

How does it do this and how does it ensure that no precipitates form in the near surface region without oxygen out-diffusion?

The major source of oxygen precipitates in the near surface region of a silicon wafer is the existence of small oxygen clusters which form during the growth of CZ crystal. They form in a complex manner as the crystal cools. The usual critical temperature range in which they form is between about 400 and 800°C. Their density distribution is a complicated function of the crystal’s (or rather crystal segment’s) oxygen content, it’s carbon content, the cooling rate through this temperature range, and even – due to more subtle effects – the cooling rate in other, higher temperture – ranges too. The complex interaction of all these effects (together with their interaction with the specific details of the application to which they are submitted) is the main reason for the need of most of the complicated wafer tailoring and specification baggage which has arisen over the last 20 years of applying silicon to advanced IC applications.

When the crystal is cut into wafers these clusters are uniformly distribution throughout the thickness of the wafer. These clusters grow into precipitates during wafer processing. If proper care is not taken, they may form in the near surface region of the wafer resulting in device failure.

Tabula Rasa:
In modern silicon, carbon levels have been reduced to the point where it no longer has an effect as a catalyist to oxygen clustering. Largely as a result of this, oxygen cluster sizes formed during growth are generally restricted. This happy event paved the way for the "Tabula Rasa" process. In today’s low carbon silicon, grown in oxygen clusters may be completely dissolved at the wafer level by short heat treatments at temperatures greater than about 950°C. Such a treatment dissolves the clusters and sends the oxygen contained in them back into solution. It is called "Tabula Rasa" (lat: blank slate) since it effectively erases the "thermal history" or that the crystal has undergone during crystal growth. It’s memory is wiped clean.

Such a wafer produces no oxygen precipitates in standard precipitation test anneals such as the "NEC1 test" (4 hours at 800°C followed by 16 hours at 1000°C). Without a Tabula Rasa anneal the density of oxygen precipitates produced by such a treatment could easily vary over about 5 orders of magnitude – even within the same crystal, and even if the oxygen content were well controlled.

This has long been recognized as a problem. It is, in fact, the main reason why many wafer road maps show the need for oxygen content to drop to very low values in the future. The hope is that a very low oxygen contents the oxygen cluster density produced during crystal growth will be small.

But this is an expensive and really rather silly solution to the problem. Tabula Rasa easily ensures that, regardless of oxygen content, the grown-in cluster density will be zero.

OK, so you’ve eliminated grown-in oxygen clusters from the wafer, how can we be sure that they do not form again?

In the vast majority of real practical cases they will not form again. The Tabula Rasa wafer really does ensure effective suppression of the formation of oxygen precipitates in process. The reason is does this is due to a phenomenon well know to materials science, namely the "incubation effect" in time-dependent nucleation.

During heat treatments, distributions in size of clusters form by adding one atom at a time. In general, these distributions are broad at low temperature (can reach large sizes) and narrow at high temperature (are restricted to small sizes). However, the build up of the distribution over time is strongly dependent on the mobility of the solute atoms at the temperature of interest. At high temperature things move faster than at low temperatures. At high temperatures, the narrow (steady state) distributions are reached quickly, while at low temperatures, the potentially wider size distributions take much longer to form.

It is the speed of reaching high temperature narrow size distributions which ensures rapid Tabula Rasa action: the broad distribution of oxygen clusters formed during crystal growth are rapidly (in a matter of seconds) relaxed to the narrow high temperature distributions. Oxygen precipitates cannot be produced from narrow high temperature size distributions.

But it is possible to produce sufficiently wide distributions of oxygen cluster sizes at low temperatures such that when the wafer is heated to high temperatures the largest of them grows. But this takes a long time. And in fact, the time it takes at low temperature to produce clusters of size large enough to grow at high temperatures is what we call the "incubation time".

This whole process is furthermore limited by the fact that it is only at temperatures below about 750°C that clusters large enough to grow upon subsequent treatment can even possibly form.

As long as the time spent at these temperatures prior to a high temperature treatment is less than the incubation time, the ALL of the clusters produced at low temperatures will relax to the narrow, and harmless, high temperature distribution.

The times required to exceed this are, in general, very long: greater than about 4 hours at 650°C (an "ideal" nucleation temperature) are required for this to happen.

This time is strongly oxygen concentration dependent. Should such long times exist in a given application, then an oxygen concentration sufficiently low to ensure the effective suppression of oxygen precipitation can always be found.

Tabula Rasa wafers have been in widespread and successful use for many years now in a wide variety of applications.

MDZ and Tabula Rasa

A Tabula Rasa wafer is a wafer in which the precipitation of oxygen is everywhere suppressed. Such a wafer meets our goal of not getting in the way of high yields. But it does not add anything to the enhancement of yield. To do this the wafer must contain bulk precipitate sites for effective gettering.

MDZ combines the benefits of Tabula Rasa with precipitation enhancement in the bulk. The Tabula Rasa effect comes for free in the MDZ wafer. As the wafer is heated on its way up to the high process temperatures (and before a vacancy concentration profile is installed) the grown-in clusters dissolve by the Tabula Rasa effect.

After the installation of the vacancy profile, the near surface region contains a low concentration of vacancies. As such it acts as a "normal" wafer. In fact, it acts as a normal Tabula Rasa wafer.

Oxygen precipitation there is suppressed and stays suppressed.

Tabula Rasa material is dead and stays dead.

Return to Questions

4. What bulk precipitate densities are needed to ensure effective gettering?

Much experience has shown that efficient gettering is reached at a bulk oxygen precipitate density of about 108 cm-3.

And precipitates do not have to be big to be effective. In fact they’re effective long before you can even see them by etching tests.

Return to Questions

 5. What are the sources of unpredictability in standard material?

The major source of the unpredictability of oxygen precipitation in standard silicon wafers is the existence of small oxygen clusters which form during the growth of CZ crystal. They form in a complex manner as the crystal cools. The usual critical temperature range in which they form is between about 400 and 800°C. Their density distribution is a complicated function of the crystal’s (or rather crystal segment’s) oxygen content, it’s carbon content, the cooling rate through this temperature range, and even – due to more subtle effects – the cooling rate in other, higher temperture – ranges too. The complex interaction of all these effects (together with their interaction with the specific details of the application to which they are submitted) is the main reason for the need of most of the complicated wafer tailoring and specification baggage which has arisen over the last 20 years of applying silicon to advanced IC applications

In addition to this there are other effects which further couples the details of the crystal process to the resulting oxygen precipitation behavior. One often important complication relates to the concentrations of residual vacancies grown in to crystal as a result of a completely different phenomenon – that of the formation of voids in silicon crystals. Such an effect is responsible for the commonly observed increase in oxygen precipitation in crystal tail end sections.

In short, the details of the crystal growth process are hugely coupled to the oxygen precipitation performance of "normal" wafers. The constraints this adds to the silicon business is unacceptable in the modern world.

Return to Questions

6. Under what conditions could the vacancy profile of an MDZ wafer be erased before it has time to nucleate the desired bulk oxygen precipitates?

During the installation of the MDZ vacancy profile, the final stage is reached during cooling when the temperature of the wafer drops below 900°C. At this point all the hitherto mobile vacancies are frozen in place through a binding process. The immobile complex O2V is formed. By 900°C all the vacancies in the wafer are bound up in this state. The wafer vacancy profile is fixed in place.

A second "fixing" of our latent precipitate image occurs at temperatures lower than about 800°C. In this range the bound up vacancies are consumed into the oxygen clusters which subsequently grow upon further treatment.

Should a virgin MDZ wafer be heated rapidly (in an RTP furnace) to temperatures greater than 900°C and go through the very rapid low temperature nucleation phase to quickly, then the frozen in vacancies will become free again and will diffuse to the wafer surfaces, relaxing the profile and erasing the MDZ effect.

This erasure effect is strictly limited to the first heat treatment the wafer sees. Following the stabilization phase, the profile can no longer be affected by high temperature treatments.

Return to Questions

7. How can it be that the behavior of MDZ wafers are not dependent on the details of the application to which they are submitted?

The catalysis effect of the vacancies on oxygen clustering is very, very fast. The vacancy profile installed in the wafer is transformed into oxygen precipitate clusters as the wafers are being heated up during the first heat treatment they receive. In the process of transforming into clusters, the vacancies themselves are consumed. Their life and job is over at this point. The clusters that form during this stage merely grow with further processing. Nothing in all but the most extreme (and unlikely) process environment will alter their distribution.

Return to Questions

Oxygen & How MDZ is Made

An MDZ wafer relies on the installation of a vacancy concentration profile in a wafer. Where the vacancy concentration is large, a very rapid catalysis of the oxygen precipitation process takes place. Where the vacancy concentration is low, oxygen precipitation is suppressed. A simple inverted U-shaped profile front to back (with the concentration high in the center, low at the surfaces) results in a layered oxygen precipitate structure which is ideally suited for advanced IC manufacture.

A straightforward and reliable process gets the job done in seconds.

When a wafer is heated to high temperatures increasing numbers of silicon atoms leave their lattice sites in the crystal. The numbers that do so are governed by the laws of thermodynamics. When a silicon atom leaves its lattice site it becomes an "interstitial" atom. It leaves behind a "vacancy". Interstitials and vacancies are created in pairs (and thus in equal numbers) in this way -- naturally and predictably simply by heating silicon to a given temperature. Call this "step 1."

In silicon, it turns out, the vacancy is slightly more energetically preferred than the interstitial. Thus in equilibrium there should be more vacancies than interstitials. This is rapidly corrected. Interstitials diffusion phenomenally fast in silicon and the excess interstitials diffuse out to the surfaces of the wafer. After only one or two seconds at temperatures around 1200°C, the wafer has established true temperature dependent equilibrium concentrations of both its vacancies and interstitials. Thus at this stage there is a slight imbalance in the wafer with more vacancies than interstitials. This is critical to the working of the process. Call this stage "step 2". The higher the temperature, the higher the concentration of vacancies. Their concentration is dependent only on temperature. Simple.

As it turns out, the problem is keeping them in the wafer as it cools from the process temperature. As a wafer cools from high temperatures, it tries to keep up with its ever-changing equilibrium state. At lower temperatures, equilibrium requires lower concentrations of vacancies and interstitials. There are two ways the wafer can try to keep up with equilibrium (that is: get rid of the excess vacancies and interstitials) as the wafer cools.

One way is for the intersititial atoms to fall back into the vacancy sites (or "recombine"). But recall that in step one, we created an imbalance in the concentrations of the vacancies and interstitials. There are simply not enough interstitials around anymore to fill all the vacancies. Thus As the wafer cools, the vacancy concentration is reduced in accordance with the dictates of equilibrium by interstitials falling back into vacancies – but only up to the point at which we run out of interstitials. This occurs rapidly - after about 50 degrees of cooling, typically. Cooling further the remaining vacancies become "supersaturated". These are the vacancies that will later be doing our job of enhancing the precipitation of oxygen.

But this is not the end of the story. There is another avenue for some of the vacancies to take in trying to maintain equilibrium. Those vacancies which are sufficiently close to the surface can diffuse there and recombine. The surface of a wafer may be considered to be an infinite sink for excess vacancies (and interstitials). But diffusion from the interior of the wafer to the surface takes time and is limited by how fast the vacancies can be transported there. Thus, depending on the cooling rate, various depths of the silicon beneath the surface are "cleaned out" of vacancies by transport to the surfaces.

If the cooling rate is too slow – such as is always the case with furnace heat treatments, then effectively all of the excess vacancies are cleaned out of the wafer through the surfaces. This is in fact the usual case for a furnace heat treatment. You pretty much end up with what you started with - not many vacancies in your wafer. No matter how many you put in during the high temperature treatment, by the end of the cooling phase they’ve all leaked out again.

But, if the cooling rate is "just right" then the leaking out of vacancies is limited to a useful depth (about 80 microns or so). Call this step 3. This clean-out or low vacancy (and tabula rasa-ed) region becomes the denuded zone. Below the depth at which vacancies can reach the surface during the cooling, rather high concentrations of vacancies can be quenched into the wafer. These become the wildly effective catalysers of subsequent oxygen precipitation and result in the formation of the future gettering layer.

RTP is used to produce the desired vacancy profiles in silicon wafers. The tempertures required to achieve sufficiently large vacancy concentrations are above about 1180°C, and the cooling rates required to keep them there are greater than about 20K/s. This is well within RTP operating ranges.

Features & Benefits

  • Oxygen precipitation density and depth distribution are preprogrammed to ideal targets
  • Superior to other gettering processes and products such as Hydrogen annealing and Nitrogen doping

Process for Reliable Internal Gettering

MDZ® is a rapid thermal processing (RTP) method of achieving reproducible and reliable internal gettering which is nearly independent of:

  • Initial oxygen concentration
  • Thermal history effects:
    • Wafer position within an ingot
    • Details of the crystal pulling process
  • IC application

Comparison of Precipitate-Free Zone

The MDZ® process produces an ideal density of oxygen precipitates and a deep precipitate-free zone. This eliminates the need for additional, costly outdiffusion, nucleation and growth thermal cycles in the customers’ manufacturing lines.