How Scientists can Manipulate Individual Atoms with Scanning Tunneling Microscopy and the Question of Determinism
Commonly used as microscopy tool to study materials at the nanoscale, STM microscopy has another, more interesting hidden function. The ability to manipulate individual atoms in almost entirely deterministic way.
For a machine whose function is quite advanced, the basic design and operation of an STM are relatively simple, and can be made by hobbyists and amateurs. Though a machine with the stability to manipulate individual atoms is a more difficult but still accessible endeavor which requires ultra high vacuum systems and some vibration damping such as magnetic levitation.
First let’s discuss how an STM in normal microscope operation works. It starts by carefully controlling an atomically sharp tungsten tip with piezoelectric servos.
Then it begins it’s scanning maneuver by moving the tip vertically and latterly so that it is close enough to “read” the position of the atom by sending a bias voltage that tunnels though the energy barrier (quantum tunneling) but far enough away so that it does not crash into the surface, producing a undesired event which is known as a “tip crash”. A feedback loop regulates the position of the tip and compensates for vibrations.
Once a full scan has been obtained, an output is sent which contains the full positions of the atoms. It’s important to note here that the images sent out can vary widely based on both the algorithms used to decipher the data and the method used to scan the image.
What I have described so far is the basic operation for the imaging of atoms, but how do we pick up and move them? This is where things get complicated, because there are number of different methods each with a different purpose and each with there own advantages and disadvantages. There is actually only one method which allows you to pick up and move an individual atom 100% percent or essentially 100% of the time but even scientists cannot explain why this mechanism works(I will get to this later).
The more common mechanism works by using the same “tip crash” we talked about earlier. The tungsten tip is lowered into the surface while a high bias voltage(4v) is applied and the resulting atoms stick to the tip. These atoms can now be deposited in a new area to form nanowires or other atomic structures and circuits. Or they can also be analyzed independently if desired.
The tip crash method, while being relatively versatile and simple to perform has limitations which prevent from it from being useful as a nanofab technique. This is because there is an inherent randomness to the process which makes it difficult to control both the exact amount of atoms and their placement. While this accuracy is still very high (within the angstroms), it is still not the best and not the limit which has been achieved.
The deterministic method[2], works in a completely different way. It works by lowering the tip and hovering near the surface with zero bias voltage, for roughly about 10ms. This is quite strange when you think about it. Why does this work? And ten milliseconds is quite a long time for an individual atom. The researchers who first confirmed this in 1998 were confused as well. Here’s an excerpt from the paper.
“It is remarkable that, in more than 98% cases, the manipulation with the tip either results in the extraction of a single adatom or has no visible effect. Only on very rare occasions s, 2% does the manipulation result in the pickup of two or more adatoms or in the deposition of some atoms on the surface. We are sure that the created modification, as shown in Fig. 1, is due to the extraction of a single Ge adatom. At room temperature, the created vacancy is not stable and slowly diffuses on the surface [20]. The resulting surface rearrangements are compatible only with that of a single atom vacancy.”
I should also note here, that for this method to be useful there needs to be another material for the atom to be “dropped” on so that you can use to make transistors and other electronic components. This would likely require the material to be prepared by some other form of nanofab such as liquid ion implantation or molecular beam epitaxy or both.
Let’s get back to the questions I posed earlier, Why does this happen from a physical perspective? To answer this question the researchers tried to mathematically model the phenomena, but in order to do this they had to use a whole series of approximations and assumptions which I think greatly limited it’s utility and explanatory power.
To understand how a Ge adatom is picked up from a Ge(111) surface with the tip apex, we have performed detailed calculations, using the standard atom superposition and electron delocalization molecular orbital (ASED-MO) method [23], of the “tip apex-Ge(111) surface” atomic structure at each 0.1 Å step of the approach and retraction of the tip. This semi-empirical calculation technique was preferred over first principles calculations [24] because a full optimization of this STM junction atomic structure at each step is actually way beyond the accessible computation time. Furthermore, the ESQC technique, which provides the tip-surface distance z is compatible with the Hamiltonian delivered by ASED. Finally, we have recently studied the quality of the ASED predictions on other semiconductor surfaces such as the Si(100) surface for chemisorption [25] and lateral atom manipulation [26] problems.
What is very strange here is that the present theory of quantum mechanics which assumes superposition of atomic orbitals, struggles explain this basic incidence where positions are actually deterministic and in fact the calculations were so long that they had to be published in another place.
Anyone that has read Thomas Kuhn’s famous Structure of Scientific Revolutions, would know that there is a word for this sort of thing. An anomaly, something that doesn’t quite match up with the present theory. Or which the present theory fails to explain. At first these anomalies are difficult to spot because often the error is so subtle, and the nature of scientific research especially in this age so fragmented.
But once you find them and realize the origin, it is like putting on a new pair of glasses. Everything looks different. Computer Scientists also have a word for this type of flaw, a cascading error. A small mistake that has huge ramifications. All it takes is for a single line code to be wrong, a single equation or number to be misinterpreted. The possibilities for failure are nearly endless.
What is actually happening?
This is difficult question to answer exactly, but I will do my best to explain what is actually happening. The reason it takes ten milliseconds for the atom to be pulled is because when the tip is lowered the atom is held by the collective strength of all of the neighboring atoms and this strength must be overcome.
This “wave” propagates through the crystal lattice in the form of Schrödinger’s equation.
The calculations for the researchers are almost certainly incorrect because of this simple fact, this is because their model only considered the relationship between the tip height and the atom. A more correct model would have included the relationship between a single atom and the rest of the neighboring lattice.
