Sunday, April 24, 2016

Oxide interfaces for fun and profit

The so-called III-V semiconductors, compounds that combine a group III element (Al, Ga, In) and a group V element (N, As, P, Sb), are mainstays of (opto)electronic devices and condensed matter physics.  They have never taken over for Si in logic and memory like some thought they might, for a number of materials science and economic reasons.  (To paraphrase an old line, "GaAs is the material of the future [for logic] and always will be.")  However, they are tremendously useful, in part because they are (now) fortuitously easy to grow - many of the compounds prefer the diamond-like "zinc blende" structure, and it is possible to prepare atomically sharp, flat, abrupt interfaces between materials with quite different semiconducting properties (very different band gaps and energetic alignments relative to each other).  Fundamentally, though, the palette is limited - these materials are very conventional semiconductors, without exhibiting other potentially exciting properties or competing phases like ferroelectricity, magnetism, superconductivity, etc.

Enter oxides.  Various complex oxides can exhibit all of these properties, and that has led to a concerted effort to develop materials growth techniques to create high quality oxide thin films, with an eye toward creating the same kind of atomically sharp heterointerfaces as in III-Vs.  A foundational paper is this one by Ohtomo and Hwang, where they used pulsed laser deposition to produce a heterojunction between LaAlO3, an insulating transparent oxide, and SrTiO3, another insulating transparent oxide (though one known to be almost a ferroelectric).  Despite the fact that both of those parent constituents are band insulators, the interface between the two was found to play host to a two-dimensional gas of electrons with remarkable properties.  The wikipedia article linked above is pretty good, so you should read it if you're interested.   

When you think about it, this is really remarkable.  You take an insulator, and another insulator, and yet the interface between them acts like a metal.  Where did the charge carriers come from?  (It's complicated - charge transfer from LAO to STO, but the free surface of the LAO and its chemical termination is hugely important.)  What is happening right at that interface?  (It's complicated.  There can be some lattice distortion from the growth process. There can be oxygen vacancies and other kinds of defects.  Below about 105 K the STO substrate distorts "ferroelastically", further complicating matters.)   Do the charge carriers live more on one side of the interface than the other, as in III-V interfaces, where the (conduction) band offset between the two layers can act like a potential barrier, and the same charge transfer that spills electrons onto one side leads to a self-consistent electrostatic potential that holds the charge layer right against that interface?  (Yes.)

Even just looking at the LAO/STO system, there is a ton of exciting work being performed.  Directly relevant to the meeting I just attended, Jeremy Levy's group at Pitt has been at the forefront of creating nanoscale electronic structures at the LAO/STO interface and examining their properties.  It turns out (one of these fortunate things!) that you can use a conductive atomic force microscope tip to do (reversible) electrochemistry at the free LAO surface, and basically draw conductive structures with nm resolution at the buried LAO/STO interface right below.   This is a very powerful technique, and it's enabled the study of the basic science of electronic transport at this interface at the nanoscale.

Beyond LAO/STO, over the same period there has been great progress in complex oxide materials growth by groups at a number of universities and at national labs.  I will refrain from trying to list them since I don't know them all and don't want to offend with the sin of inadvertent omission.  It is now possible to prepare a dizzying array of material types (ferromagnetic insulators like GdTiO3; antiferromagnetic insulators like SmTiO3; Mott insulators like LaTiO3; nickelates; superconducting cuprates; etc.) and complicated multilayers and superlattices of these systems.   It's far too early to say where this is all going, but historically the ability to grow new material systems of high quality with excellent precision tends to pay big dividends in the long term, even if they're not the benefits originally envisioned.


Anonymous said...

Thanks for the write up!

If I understand it correctly, the LTO/STO interface hosts a two-dimensional electron gas. What makes it different from other well known systems that host 2D electron gases?

Ted Sanders said...

"Despite the fact that both of those parent constituents are band insulators, the interface between the two was found to play host to a two-dimensional gas of electrons with remarkable properties. The wikipedia article linked above is pretty good, so you should read it if you're interested.

When you think about it, this is really remarkable. You take an insulator, and another insulator, and yet the interface between them acts like a metal."

Sometimes I worry that the LAO/STO people oversell this "remarkable" fact. I'd call STO a semiconductor rather than an insulator (it can be doped n-type with impurities, after all), so viewed from that perspective, I don't think it's super surprising that it can be modulation-doped to host an n-type 2DEG. Anyway, I suppose it's all a matter of perspective.

P.S. I wrote the Wikipedia article on LAO/STO and I'm glad you liked it.

Anonymous said...

It is indeed very simple to “convert” SrTiO3 to an electron “conductor”: A short annealing under vacuum conditions or in reducing atmospheres generates oxygen vacancies which act as donors (induced self-doping by reduction).

Douglas Natelson said...

Anon@10:53, a couple of things. There are multiple bands involved including d-bands, while conventional III-V 2deg or group IV 2deg is based only on s and p. Also, the dielectric function of STO is pretty wild (very large in the dc limit, tending in the direction of divergence at T = 0).

Ted, cool. Yeah, STO can be doped (as can other oxide semiconductors, of course, like TiO2 or ZnO), and you have a point about overselling. Still, you don't get a 2deg at the GaAs/AlGaAs interface without gating unless you deliberately dope, where the LAO/STO case seems to happen without deliberate doping (b/c of the need to avert the polar catastrophe? This isn't really my direct area of research, so I haven't kept up on this as much as I should have.).

Anon@7:27, yes, to echo Ted. Indeed, this has been a major challenge in understanding these systems - you want to make sure you can tell the difference between intrinsic effects and effective doping due to unintentional defect formation at or near the interfaces.

Anonymous said...

Prof. Natelson, it's true that you don't get a 2-DEG at an AlGaAs/GaAs interface without gating or doping the AlGaAs layer but that is not necessarily true for AlGaN/GaN. You end up with a 2-DEG at that interface due to polarization effects and some unintentional doping from defects. Would you say that the III-N system is similar to the LAO/STO system in that case? The 2-DEG density is certainly lower for III-N systems (1-2e13).

Anonymous said...

This paper explains a bit more about the nature of doping in the III-N system:

Ryan Comes said...

The LAO-STO interface is amazingly complex, but it's also ripe with inconsistencies. How you prepare the samples matters a ton, because it's been shown pretty conclusively IMO that the conductivity is all defect driven. Papers by Lane Martin's group, Darrell Schlom's group and various others have looked at the role of the synthesis process in getting the properties and shown that it only happens in the case of LaAlO3 films that don't have La:Al 1:1 ratios. Levy's group and many others have done some remarkable science on the defect-induced 2DEGs, but I think the polar catastrophe model has been pretty conclusively disproven.

Anonymous said...

Anonymous said...