Epitaxy - Science topic

Main question:

If you look at the basic structures fcc and hcp, you find that fcc(111) and hcp(001) are both maximum density packed surface structures with triangles. Therefore these two facets match with each other and you can do epitaxy there. Please be aware that, since fcc has an abcabc stacking and hcp just ababab, an fcc-derived layer provides one additional "structure variable" about the stacking direction, so you can expect epitaxial growth without any texture issues of hcp(001) on fcc(111), but if you try it the other way round, you can expect the fcc layer to stack both abcabc and cbacba (symmetry-equivalent to a 60° or 180° rotated domain), so if you check the texture with an XRD pole figure, you can expect twice as many spots as on a single crystal.

Secondary question:

The adatoms can be bound to the substrate by any known kind of bond, be it van der Waals, covalent, ionic, metallic or quite often a mix of them. That's not a question of the crystal structure but more a question of what the specifice atoms prefer when they react with each other.

Thank you Prof. Jürgen Weippert for mentioning the possible points.

Dear Arpan Ghosh ,

A change in temperature, for any kind of study, implies a change in the phonon modes. In Raman spectroscopy, an increase of the temperature always induces a shift of the Raman bands to lower frequencies (see DOI: 10.1038/srep32236). This affact is attributed to thermal expansion and anharmonic effects of phonon processes

Best regards,

Elhoucine

Well, epitaxy means that the grown film has a structure which is predefined by the substrate. Therefore, you XRD reflexes in the "most epitaxial" case should correspond to exactly one in-plane orientation, otherwise you have multi-domain or twinned growth. EPSD would also be a method to check that, but when you already have XRD and there is no differently oriented phase, you already know you're fine.

There are also other criteria like thickness uniformity or a low defect&unwanted dopant density, but these are general coating parameters and not epitaxy-specific.

Thanks. Though it seems that any oxidation operation is a dead end.

Kindly see also the following useful RG link:

Len Leonid Mizrah Thank you for your generous share. Your considerate suggestions brings me a lot of ideas. In fact, heteroepitaxial diamond growth and even single crystal growth is not a technical problem for me; however, it is an important task for me to understand why epitaxy can occur. Thanks again.

For the best application of epitaxial thin film, ideally zero lattice mismatch among thin film and substrate is necessary.

Otherwise, gradually strain energy is developed at the thin film/substrate interface and ends up with formation of defects viz. dislocation.

Engineering of control over lattice mismatch is one of the most challenging task !

I totally agree with you

Mikhail Pylnev

The nanoparticles on the surface have an unfinished crystal lattice. If the synthesis conditions (temperature, pressure) favor the formation of another allotropic modification, then it appears.

If you are interested in details write a letter to Pugachevsky. He knows this problem.

for growing epitaxial films on need a single crystal substrate

MAUD is not a very appreciated software for Rietveld refinement. For beginners or for basic profile fitting, it works well; but if you want to fit multiphase materials, complex structures, preferred orientations, strained materials, magnetic and nanomaterials, non-stoichiometric compounds etc; you might go through problems for precise fitting. In fact, high impact journals and renowned journals of physics prefer FULLPROF more.

I'll personally suggest FULLPROF, it is a bit harder, but it is the most powerful software, which will offer you 20-30 parameters for finest refinement.

There are a number of tutorials present in youtube as well, you can learn it by a week.

Dear Kamal

Please read the following articled. Hope that you will get the best answer for your queries.

https://www.blue-scientific.com/large-area-reciprocal-space-mapping/

https://www.malvernpanalytical.com/en/products/measurement-type/reciprocal-space-analysis

https://www.rigaku.com/applications/bytes/xrd/smartlab/1119947204

you can also have a look at:

and related topics

Bests

Cristian

Just a working protocol in fabricating a device:

You have to assess every processing step and judge if satisfy the requirement for the device or not

- You can start the following process if the current process fail to perform as required.

-You have to be knowledgeable in semiconductors and electron devised in addition to their technology.

-One of the main precaution is the to avoid contamination absolutely

- You have sufficient measuring methods and instruments

Best wishes

The quality of the epitaxial layer is determined by the crystallographic defects contained it. As the crystallographic defects increase the quality will be deteriorated. The tolerance to the crystallographic defects is dictated by the application. As the mismatch between layers deposited layer increases, the greater will be the crystallographic defects. Also the deposition temperature and the rate of deposition affects the crysatllographic defects.TH inclusion of contamination on the initial substrate affects much the formation of defects.

One of the major defect is the dislocation, dislocation line and stacking faults.

You need to delineate them by the suitable etching and counting them by an optical or electron microscope.

There are books in the epithelial growth and fault formation and detection in the web.

Epitaxial growth is a complicated process needing elaboration both in theory and practice.

Best wishes

I think you can use Ge instead of silicon. It matches GaAs to a great extent according to the paper in the link:

As a proposal you can dope silicon with germanium heavily such that you can reduce the matching between silicon and GaAs.

Either you can use GaP or AlP instead of Ga As. This solution is extracted from the figure in the paper of the link:

Band gap and lattice constant for various III–V and group-IV material alloys. 

Best wishes

Hi Alaa, the problem is not associated whether it is copper grid or nickel grid. It may be associated with the process. I am still getting the better Contrast with Cu grid in TEM. Here the problem is disintegration of the film while vacuum annealing on Cu grid. Go with the explanation of the Krishnan sir. Hope it will sounds good for you. Thanks for your response.

I agree with the other responses. I'll add that the word epitaxy indicates the objective of crystal growth versus simple deposition. The goal is to maintain the underlying crystal structure of the substrate and have it continue through the various MBE grown layers even as the material layers change. Thus, the various layers of material must have the same crystalline structure and similar lattice constant as the underlying substrate. If two layers with a large difference in lattice constant are desired, then transition layers might be used. The advantages of MBE include the crystal purity and ability to control film thickness to the monolayer. MBE layers can be made so thin that quantum mechanical effects can be engineered into a material. Since the crystal structure must closely match, there is a fundamental limitation to the available materials for a given substrate or material system. MBE is also slow, so it requires an ultra-high vacuum environment.

Sputtering involves dislodging molecules from a sputtering target using a high-energy source beam. The goal is a homogeneous film. Sputtering is not epitaxial - there is no attempt at maintaining the crystalline structure of the film or matching it with the substrate. The film is deposited not grown. For this reason, there are fewer fundamental limitations on the type of materials that can be sputtered on a given substrate. Sputtering is also faster than MBE, so it does not require the same high vacuum levels. Since the films a largely amorphous, sputtering offers the advantage of tailoring the density by changing the deposition parameters. The refractive index of a sputtered film, for example, may be tailored. Sputtering is also capable of creating very dense films compared with other thin film deposition techniques.

More information is needed to answer your question. What is the phase scale? Is it in attractive or repulsive regime? Such phase contrast may be due to switching between attractive and repulsive regimes. Check this video lecture https://www.youtube.com/watch?v=Y8Sr9PHL_oU&t=0s&list=PLtkeUZItwHK78uGpH76_hhsehADDWnKQB&index=12

I guess XRD would be good choice and you can also confirm it with AFM. Please see this paper.

Are both decent for small quantities and have fairly wide selection.

The most important property for the epitaxial growth is the lattice matching between the substrate and the deposited material. Mathing means complete coincidence between the atomic arrangement at the interface of the film and substrate. This condition can be realized easily in homoepitaxy. In heteroepitaxy of two different material the lattice mismatch must be limited to a small percentage deviation while keeping the thickness of the epitaxial film under some critical thickness otherwise dislocations begins to appear in the deposited film. For more information please follow: users.wfu.edu/ucerkb/Nan242/L12-Epitaxy.pdf

Best wishes

Hi,

Try to set laser power to 10%. Also after acquiriy the data, perform a despike and smoothing. Also make sure your stage is not moving and you are well focused in the sample. Another thing, increase the data accumulation for every data point. Use 25 to 30 accumulation for each data point.

Dear Amine,

welcome,

The epitaxial growth rate depends on deposition temperature and the transport rate of the reacting gases to the wafer surface. For a given temperature there a maximum rate of deposition other wise the grown layer may contain many defects and ultimately becomes polycrystalline.

In addition to the paper brought by Daniela please see the thesis in the link: https://www.mobt3ath.com/uplode/book/book-2978.pdf

Best wishes

Dear Anyanwu,

Epitaxy means to grow oxide layer on the substrate in crystalline form.

This implies that both the grown layer and the substrate must be crystallographically matched. If the film and substrate materials are the same this will be homoepitay other wise it will be heteroepitaxy.

I assume that you have homo epitaxy.

There will be important parameters for the epitaxial growth:

- the substrate temperature. As the temperature increases, the grown film will tend to be single crystal.

- The second important factor if the growth or deposition rate. As the deposition rate increases, the deposited film will tend to be poly crystalline.

So, the deposition rate must be as low as possible to give the molecules chance to order themselves in a crystallographic form.

- Th e last important factor the surface of the substrate must be very clean and free from any contamination to ease the crystallographic building process.

To adjust the evaporation rate by the LASER beam please refer to the link:

Best wishes

Dear Daniela,

Thank you for your answers. The substrate is glass. Unfortunately, the films are too thin to get an accurate methods of weight measurement. There is no epitaxy at all, just a preferential orientation along 111 direction, but the film are polycristalline.

My question are:

1. Is there relation between the density of a film and its texturation?

2. Is the bandgap of a film similar for a textured or polycristalline film?

3. Is the bandgap dependent on the texturation level?

If you have any paper related to this subject, could you give me some reference?

Thank you and best regards ,

Emile HAYE

SNR to COD relationship for the above data set at location "800" on topograph.

You can calculate the density of dislocations by X-ray diffraction w-scan rocking curve of the (002) plains for screw dislocations and the w-scan of asymetric plains such as (101) plains. Check the next reference:

Comparative study of threading dislocations in GaN epitaxial layers by nondestructive methods , Volume 57, January 2017, Pages 32-38

The pits can be estimated by counting its density by AFM.

The tramps can be studied by Deep-level transient spectroscopy.

Also luminescence can give you an idea of the crystal quality and deep defects.

Moreover you can measure the concentration of impurities such as Oxigen and Carbon by SIMS.

Finally Hall effect mobility can tell you the scattering mechanisms such as dislocations related to the defects in your film.

I hope this will be useful for you.

regards

Dear Hans, meanwhile Michael Mauk published a newer review on LPE within the Second Edition of Elsevier Handbook of Crystal Growth, Vol. IIIA, Ch. 6 (2015). There, on page 244, Fig. 6.2, are shown the 8 methods and Michael added related comments. Is that, what you mean ? Best regards, Peter.

We offer such epi service, please contact me directly at

Best regards,

At first, whatever the crystalline system is, atomic position for a and b should be nearly matched each other in various way, with off-axis angle.

please refer below

1. C. M. Carlson, P. A. Parilla, T. V. Rivkin, J. D. Perkins, and D. S. Ginley, Appl. Phys. Lett. 77, 3278 (2000)

2. L. Vegard, Z. Phys. 5, 17 (1921)

3. Landolt-Bőrnstein handbook

Ravi, 

"This would reveal only those defects that terminate on the exposed sample surface"

Yes, and that would not be a problem because misfit dislocations in epitaxial thin film usually terminated at the film surface. 

"This would reveal only those that have been exposed by the EPD method, correct?"

No. The epitaxial film growth process usually includes substrate heating or thermal annealing. These heating processes will cause thermal etching and pits will be formed at the surface terminated points of dislocations without any additional etching process. 

It's depending on the balance of surface energy and line tension of dislocations, but according to my experience, surface pits is easily formed at 1000 K.

So by limiting the discussion to the NiO epitaxial thin film on SrTiO₃, surface topography measurement with AFM will be a simple and fast way to check dislocations.

That means each crystalline structures have at least one common plane over which lattice planes belonging both crystals can match  with a minimum mismatch strain to  produce a coherent interface. This is a simple and operational definition.  This mismatch strain can be relaxed  due time by the formation entangled dislocations and/or quantum dot pattern formation in thin film. This mismatch strain energy is then becomes the main driving force. 

Thank you very much!

You can have InBi clasters, even as small as several atoms, in InP lattice. Small clusters should still be coherent with host lattice (but they may introduce strain to surrounding host lattice). So, not directly Bi_n clusters, when Bi is on both In and P sublattices, but (InBi)_n clusters, when Bi is still on P sublattice in InP, but locally aggregated into some larger molecules (several atoms) rather than nanoclusters (hundreds or thousands atoms). They may be hard to identify, as it is in case of high electrical doping. 

Epitaxially growth does not mean just single crystal, you can grow columnar grains in the direction of substrate grains.please read this articles.

Dear Zhimin Qi

Generally, the following factors are needed to be considered in epitaxial growth:

   - Growth mechanism or growth mode

   - Strain due to lattice mismatch and thickness of epitaxial layer(s)

   - Thermal lattice expansion

   - Interdiffusion between atoms in epitaxial layer(s) and substrate

Apart from these, the growth condition is very essential. We can effectively control the growth mechanism and interdiffusion through the growth condition such as growth temperature and growth rate.

so this dates quite a while back. Have your questions been clarified?

for the moment just a comment on the following you wrote:

"I agree with the hardness comes from the magnetostriction (shape anisotropy). But the magnetostriction induce the anisotropy due to the dipolar interaction."

This is actually partly incorrect: Shape anisotropy derives from dipolar interactions, this is true. In a thin film, this contribution will favor in-plane magnetization vs. out of plane magnetization.

Magnetostriction, however, relates to mechanical (usually elastic) deformations which occur as magnetization is rotated. At the very heart of this is local spin orbit coupling (at the site of each moment). It can occur in ferromagnets and in antiferromagnets. CoO is a famous example for this: as magnetic order is established at the Nèel temperature, the crystal structure changes from cubic to tetragonal. Now, if you want to rotate the (staggered) magnetization from one easy axis to another, you also have to rotate the tetragonal distortion. This may give rise to a tremendous energy barrier having to be overcome.

Good Morning Mam,

I am not finding attached link.

With Best Regards

Rupesh 

I would agree with the colleagues in that my first question would be: how many times have you reproduced the result and is it the same when you perform the loop in inverted sequence [one from pos. field to neg. and back to positive; another one the other way around]. It is very unlikely that what you observe is real unless you sample has changed its properties while doing the magnetization loop.

As a second remark, a single magnetization loop (even if reproduced and free of artifacts) will most likely not be able to tell you much physics about your sample. Especially for composite materials and exchange bias systems (which yours could be) there is a host of intelligent protocols worth studying in the pertinent literature.

Finally, also the negative field side does not necessarily look normal, the "up" and "down" branches do not really superimpose. If that is real, then your maximum negative field is smaller than the so-called irreversibility field and it would be useful to extend the measurement range. If not real, then there is a drift here, too, as indicated by Greig.

Hello again Mahmood,

To compare substrate temperature in your machine with rheed surface reconstruction maps or transition temperatures published in literature I would suggest you (in our corresponding lesson of elementary MBE  ;-)   - I am not sure  how  elementary it should be) to try  e.g. :

(1) first heat up fresh GaAs(001) substrate to around 450-500 oC read by TC in your machine, and then go on slowly in steps like 10 oC  every ~5 min to recognize oxides removal temp in rheed  (relatively fast increase of brightness of spots) which occurs at ~580-600 oC real substrate temp, as is commonly regarded, and then you will have an estimation how big is the difference between  TC and real surface temp in your MBE. It may be as high as +100 oC or -100 oC or even more depending on the machine. Usually above ~400 deg C I heat up GaAs under some As flux, like 1e-6 Tr.

(2) At real surface temp like 580-600 oC under As flux around  2e-6 Tr try to start growing  GaAs with Ga flux at around 1e-7 to 2e-7 Tr as measured by flux ion gauge, which should give you growth rate probably around 200-400 nm/h   and safe As/Ga BEP ratio of 10-20  which guarantee stable GaAs growth with As-rich conditions. After 50-100 nm of GaAs you should see (2x4) streaky basic rheed pattern rotating the substrate. And such surface with streaky pattern is good starting point  to play  with substrate temp.

(3) instead of changing Ga source temp you have mentioned  I suggest to change substrate temp still growing GaAs (again try to do it in steps like 10 oC/5 min ). At around 650 oC surface temp you should see a transition to (1x1) rheed pattern, then increasing surface temp more to around 680 oC you should see a change to (3x1) pattern. Again you can compare those published temps in rheed maps with readings of your TC to learn the difference in your machine.

(4) then you can try decreasing substrate temp still growing GaAs. Around 500 degC surface temp you should see a change of pattern to (2x1) or (2x2) -this last one is also named c(4x4) as you can see at 45 deg angle from the angular direction of seeing (2x2) pattern rotating the substrate.  You can decrease GaAs surface temp to as low as say 200 oC still growing GaAs, this is called low temperature MBE growth and is used sometimes eg in extreme highly doped GaAs.

Try to compare the temp reading from your TC with surface reconstruction maps published as this will give you good feeling where you are with real surface temp at a given TC set point . Such feeling is essential for future growth of various materials in your machine and comparing your  temperatures with published values in literature. But probably we all have some inaccuracy in the temperatures we give in reports, so own experience and trials are most important, at least for me.

(5) finally you can return to surface temp 580-600 oC and try to slowly decrease As flux  (As/Ga BEP ratio) as low as you can but still having (2x4) As rich pattern. This is usually the bottom line for stable growth of GaAs with good surface morphology.  This bottom line can be at BEP As/Ga ratio of say 3 to 10 depending on the machine, as I have written yesterday, it depends on surface temp used. 

(6) You can try similar temp trip with no growth, ie only under As flux or with different As/Ga ratio. There should be some difference at rheed pattern transition temperatures with one seen above.

(7) In older machines the difference between TC reading and real surface temp may unfortunately also depend on substrate holders. So you may check this too to know better the machine.

I hope those details may be useful, sorry if this all is obvious for you. I am not sure this is best practice but it works in our lab.

So, good luck for playing, 

Regards,

Tom

I would like to ask you what is the purpose of measuring the mobility in the vertical direction. If your device is lateral then all what you need is to measure the mobility in the lateral direction as normally done and as you described. This is because the carrier transport will be in this lateral direction.

If your material is isotropic like silicon, then its mobility will be independent on the direction of the movement so long as you acquire the bulk properties. So, for isotropic materials the bulk mobility will be the same in all directions.

In case of thin film the the measured mobility mu will be an effective mobility, of the surface and volume such that 1/mu = 1/mub  + 1/mus, with mub is the bulk mobility and mus is the surface mobility.

However if you want to measure the mobility in the vertical direction along the epithelial film thickness you can make the substrate acts as a minority carrier injecting 

emitter in the film and then you supply the top of the film be collecting contact of the minority carrier, that is you form an npn or pnp structure from the substrate, the film and the added collector layer. The you can inject minority carrier pulse and collect it from the collector. The delay time between the injected and the collected  pule will be related to the the transit time which is in turn related to the mobility of carriers.

You can invert the arrangement and make the injector on the top and the collector at the substrate. In this case you can use laser pulse for initiation of the injection.

This method is based on time of flight measurement.

For more information, please revert to the link:

wish you success

Dear All,

Thank you for your replies. Ravi, I have these papers but that anyway.

I am working on growing ZnO films on Al2O3 substrates with CVD. From literature they will have 6/7 or 5/6 domain matching epitaxy. I have read many reports claiming that you can get better quality films with DME as defect are limited to near substrate-film interface regions. I agree with you Vladimir that people are using DME approach to explain how their films grow with larger mismatch with substrate and still have good film quality. But there must be an explanation or growth conditions under which this DME is occurring. I agree with Gavin that if we can choose conditions where it encourage to have more misfit dislocations at film-substrate interface you may be able to get intentional DME growth. I am just curious to know how this works. As far I know ALE or ALD tends to grow polycrystalline or amorphous films. I am interested in epitaxial films.

Thank you for your help.

Regards

Nitin

Dear Hongwei Liu, I am very grateful to you for really clear answer about what is going on mentioned TEM image. It seems to be Si fragments that are displayed on the image.

Could you please help with analysis for the image I attached this time? The fragment situated in the central area in my opinion has some kind of several flat facets. Also, FFT demonstrate some extra spots that hardly correspond to second-diffraction phenomenon.

Yours sincerely,

Shevlyagin Alexander,

juniour researcher

No damage, Si is very stable material.

Growth on (111) is very different from (11-1) as the latter has a low growth rate and the (111) a 100 times higher growth rate surface in GaAs, see L. Hollan and C. Shiller, J. Crystal Growth 13/14, 319 (1972) and W. Shaw, J. Electrochem. Soc. 115, 405 (1968). The reason is the stability of the different surface reconstructions on these orientations.

Hence there in practive is no stable (111) surface obtained. Because the higher growth rate results from a strong sticking and thus a very low mobility on (111). However, a (111) surface can be stable (thermodynamically) at low V/III ratios (i.e. a very low growth rates for near equilibrium and low arsenic pressure). See the calculated WUlff plot (from A. Kley at FHI-Berlin 1997) In practice this is almost impossible to achieve in MOVPE but maybe possible via MBE.

About double side substrates: As written above those are first polished on one side, then mounted (glued) on that polished side and the final surface CMP is done; this is also done using a much finer polishing paste. While both side may look shiny, one side is much flatter and has much less polishing damage than the other. For epitaxy one the backside (and generally for any GaAs growth) you could try to etch the surfaces. An epi-ready ozone treatment tend to fail on high-index GaAs surface. If you attempt growth with MOVPE, a long anneal in H2 and AsH3 at 650°C may help before growth.

Dear Wang

The trick depends on the method which you are using. If you growing with PLD then minimized the gas pressure. The sample to target distance should not be grater than 7cm. It is also depends on the Laser power.

In X-ray scattering, f' and f'' are the resonant or anomalous X-ray scattering factors,  expressing the non-Thomson part of the scattering form factor.  It is due to the X-ray energy being near (or, strictly speaking, below) any of the binding energies of the core electrons of any of the atoms.

See  https://en.wikipedia.org/wiki/Atomic_form_factor#X-ray_form_factor or https://en.wikipedia.org/wiki/Anomalous_scattering or http://skuld.bmsc.washington.edu/scatter/

I have no idea what the  PANalytical wants for these, but would imagine that the values at http://physics.nist.gov/PhysRefData/FFast/html/form.html would be a good start.

MgO is hygroscopic. With H2O from ambient air, or else as a consequence of the cleaning treatment, hole defect can form in the substrate surface. This is particularly pronounced on the (111) surface but can also occur on the (100) surface. We had trouble with this issue many years ago and finally ended up by re-polishing the substrates just a few hours before use in thin film growth. You may want to have a look at the following publications in which the same issue was apparent with Ti growth on MgO:

Titanium thin film growth on small and large miscut substrates

M. Huth, C. P. Flynn

Appl. Phys. Lett. 71, 2466-68 (1997)

Also in our old work on Co/Pt trilayers and multilayers we had this issue:

MgO Surface Microstructure and Crystalline Coherence of Co/Pt Superlattices

P. Haibach, J. Köble, M. Jourdan, M. Huth, H. Adrian

Thin Solid Films 336, 168-171 (1998)

Step-edge induced anisotropic domain-wall propagation

P. Haibach, M. Huth, H. Adrian

Phys. Rev. Lett. 84, 1312-1315 (2000)

Dear Phanikumar,

With the help of High Resolution X-Ray Diffraction (HRXRD) technique and Rocking curve analysis & Reciprocal lattice map of epitaxial film,  you can estimate the misfit value of your epitaxially grown film and the substrate.

May be you can contact Dr. CLAUDIO FERRARI , Head (Gruppo di Diffrazione X), IMEM – CNR, Parma, ITALY. Email: claudio.ferrari@cnr.it, 

The instrument is Rigaku smart lab diffractometer with following settings ..

Thanks for helping out .. let me know if any other specific details are required 

Thank you for your answers.

I tried many of different methods, but the one working the best, was to use razor blade broken in half, hit with a small 75g hammer.

Hereby, I close the topic.

Seems like your best shot would be olivine on olivine.  What technique are you using to grow the olivine?  How big (measured parallel to the epitaxial surface) do you want the crystals to be? 

Detectors with very small effective pixel size are typically indirect detectors, i.e. a scintillator screen converts x-rays into visible light, which is then projected onto a larger pixel CCD or CMOS detector by a (visible light) microscope.

You may care to look at

Enhanced nitrogen diffusion in 4H-SiC

Applied Physics Letters Vol 80 No. 2,

and,

Enhanced Dopant Diffusion Effects in 4H Silicon Carbide

Materials Science Forum Vols. 389-393 (2002),

and,

Field Enhanced Diffusion of Nitrogen and Boron in 4H Silicon Carbide

Applied Physics Letters Vol 94 No 7 2003

I hope this helps a little.

I don't think that it is possible to grow STO directly on Si at that background pressure. Si is easily oxidized and forms an amorphous SiOx layer preventing epitaxy.

Usually YSZ or CeO2/YSZ is used as buffer layer for epitaxial SrTiO3 growth on Si.

MBE allows to grow SrTiO3 on Si, but ultra high vacuum, low temperature deposition and recrystallization steps are needed (see for example Li et al. J. Appl. Phys. 93, 4521 (2003))