Introduction: Ivar Giæver, a Norwegian American physicist, was born in Norway in 1929. He worked as an engineer in the Norwegian army for one year in his early years and as a patent examiner in the Norwegian government for one year. He immigrated to Canada in 1954 and joined the

My task is to study thin films, and to me "films" means photography. But I was lucky enough to work with John Fisher, who obviously had more ideas than I did. Fisher was also initially a mechanical engineer, but recently he turned his attention to theoretical physics. He believes that thin film technology can be used to make practical electronic devices. Soon after, I began to study metal films separated by thin insulating layers and tried to conduct tunneling experiments. I can be sure that Fisher knew Leo Esaki's tunneling experiment at that time, but I didn't. For me who has worked hard to learn some quantum mechanics at Rensselaer Polytechnic in Troy, the concept that particles can pass through obstacles seems a bit strange. If you throw a tennis ball on the wall many times, it will eventually pass through without damaging the wall or itself, which is definitely as difficult as winning the Nobel prize! Of course, the trick is to use very small balls, as well as a lot of balls.Therefore, if we can place two metals very close without short circuiting, the electrons in the metal can be considered as balls, and the "walls" here are the intervals between the metals. These concepts are shown in Figure 1. Although classical mechanics correctly predicts the behavior of large objects (such as tennis balls), in order to predict the behavior of small objects such as electrons, we must rely on quantum mechanics. Our physical insights are related to daily experiences with large objects, so we should not be too surprised that electrons sometimes act unexpectedly.

Figure 1 A. If a person throws the ball on the wall, the ball will bounce back. The laws of physics allow the ball to penetrate or pass through a wall, but since the ball is a macroscopic object, the possibility is slim. B. Two metals separated by vacuum are similar to the above. The electrons in metal are "balls" and the vacuum represents the wall. C. Diagram of energy diagrams of these two metals. The electrons do not have enough energy to escape into the vacuum. However, these two metals can exchange electrons through tunneling. If the distance between metals is very close, the possibility of tunneling is high, because electrons are a microscopic particle.

Since neither Fisher nor I have much background in experimental physics, we made a lot of mistakes from the beginning. To be able to measure the tunneling current, the two metals must not be spaced more than about 100 Å. Considering the impact of vibration, we decided at the beginning not to use the air or vacuum between the two layers of metal as an insulating layer. After all, we have all received training in mechanical engineering! We tried to keep the two metals separated using various thin insulating layers made of Langmuir film and polymethylethylene acetate fat. But these films always have pinholes, and the mercury-assisted electrodes we use can cause short circuits. So we spent a lot of time measuring the very interesting but non-repeatable volt-ampere characteristic curve — we call it a miracle because each curve only appears once. After a few months we finally came up with the right way to use a evaporated metal film and separate them through a naturally grown oxide layer.

Figure 2 Schematic diagram of the vacuum system used to deposit metal films. If the aluminum is heated with resistance on the tantalum crucible, the aluminum will melt first, then boil and evaporate. The aluminum vapor then solidifies on the cold substrate in the steam stream. The most common substrate is a regular microscope slide. By using the shielding effect of the metal mask, patterns can be formed on the slide.

To achieve our idea we needed a coating machine, so I purchased my first piece of experimental equipment. I was very worried while waiting for it to arrive, I was worried that I would fall into experimental physics related to this expensive machine, and my plan was to switch to theory after learning enough. The hunch is right, I did get stuck on the coating machine, but not because it is expensive, but because it fascinates me. Figure 2 shows a schematic diagram of the coating machine. To make the tunneling junction, we first evaporated an aluminum tape onto the slide, then removed it from the vacuum and heated to quickly oxidize the surface. Then we plated several crossed aluminum strips on the first film and formed several connection points at the same time. The sample preparation steps are shown in Figure 3. This method solves two problems, first there is no pinhole in the oxide, because it is self-healing, and secondly we get rid of the mechanical problems caused by mercury-assisted electrodes.

Figure 3 A. A microscope slide with a layer of vapor-deposited aluminum tape in the middle. Once the aluminum film is exposed to air, a protective oxide insulating layer will form on the surface. The thickness of the oxide depends on factors such as time, temperature and humidity. B. After the oxide layer is formed, an aluminum tape intersecting therein is vapor-deposited on the first layer of film, so that the oxide layer is sandwiched between the two metal films. The current will pass up through the oxide layer along one aluminum film and then flow out through another aluminum film, while we measure the pressure drop on the oxide layer. C. Circuit diagram. The figure measures the volt-ampere characteristics of a capacitive device formed by two aluminum films and an oxide layer. When the thickness of the oxide layer is less than 50 Å, a significant DC current passes through the oxide layer.

By around April 1959, we had conducted several successful tunneling experiments. The voltammetry characteristics of the sample are quite repeatable and are well in line with the theory. Typical results are shown in Figure 4.We performed several checks, such as changing the area of ​​the junction and the thickness of the oxide layer and changing the temperature. Everything looked good and our lab even held a seminar. By this time, I've solved enough Schrödinger equations to convince myself that the electrons sometimes behave like waves, and I'm no longer worried about this.

Figure 4 Voltammetry characteristic curves of five tunneling junctions with the same thickness but varying areas. The current is proportional to the area of ​​the junction. This is the first clue that we can achieve tunneling rather than short circuit. In early experiments, the oxide layer we used was relatively thick, so only a small current could pass through at low voltages.

However, there are many real physicists in the laboratory who have raised reasonable doubts about my experiments. How do I make sure it is a tunneling effect instead of a metal short circuit, an ionic current or a semiconductor? Of course, I don't know that despite the same theory and experiments, doubts about correctness are always in my mind. I spent a lot of time thinking about impossible solutions, such as tunnel transistors or cold cathodes, all of which were intended to confirm my explanation of the tunneling effect. At the time I thought it was strange to do what I thought was interesting and get paid for it, and my conscience bothered me. But just like quantum mechanics, just get used to it. And now I often defend the opposite view: we should support more people in doing pure scientific research.

Figure 5 A. Energy band diagram of two metals separated by barriers. Due to the potential difference between the two metals, their Fermi levels are different. Only electrons with energy higher than the highest energy level of the metal on the right can tunnel to the right, because only the energy level of these electrons is in a hollow state on the right. Pauli's principle makes it possible to have only one electron in each quantum state. B. The metal on the right is now in the superconducting state , so there is a superconducting energy gap in its energy band, and there is no single electron energy state in the energy gap. The electrons on the left can still pass through the barrier, but if the applied voltage gives them less than half of the energy gap, there is no state in the right superconductor that matches its energy, so they will not be able to continue to enter the right side. Current only occurs when the energy provided by the applied voltage to the tunneled electrons is greater than half of the superconducting energy gap. C. Schematic diagram of voltammetry characteristic curve. When both metals are in normal state, the current is proportional to the voltage. When a metal is in a superconducting state, the current voltage characteristics change greatly. The exact shape of the voltammetry characteristic curve depends on the electronic band in the superconductor.

I continued to try my ideas, while John Fisher went to study elementary particles with his unique optimism and enthusiasm. Also, I got more and more suggestions and guidance from Charles Bean and Walter Harrison. Just give these two physicists a piece of chalk and a blackboard and they can unbelievably sort things out. Meanwhile, I continued to take formal courses at RPI, and one day we learned superconductivity in a solid state physics course taught by Huntington. Well, I don't believe that the resistance drops to zero accurately, but what really caught my attention was the energy gap in the superconductor, which is the heart of the new BCS (Bardeen-Cooper-Schrieffer) theory. If the theory is reasonable, and if my tunneling experiment is reasonable, combined with both should have something very interesting to happen, as shown in Figure 5. When I returned to GE's lab, I tried to tell my friends about this simple idea, and I remember they didn't seem to think it was that good. Energy gap is indeed a multibody effect and cannot be simply explained in my way, but even with quite a lot of doubts, everyone encouraged me to continue trying. Then I realized that under the electronic volt unit system that I could understand, I didn't know what the size of the energy gap was. But this is an easy problem, just use the method I use: first ask Bean and Harrison. When they all say it's a few millielectron volts, I'm glad because it's in a voltage range that's easy to measure.

Figure 6 A standard device for low temperature experiments.It consists of two dewars, the outer layer is filled with liquid nitrogen and the inner layer is filled with liquid helium . The boiling point of helium at atmospheric pressure is 4.2K. By decompressing the liquid helium, the temperature can be reduced to about 1K. The sample is simply suspended in liquid helium through the measurement line.

I have never done an experiment that requires low temperature and liquid helium, which seems to be a complicated thing. However, the huge advantage of a large lab like GE is that you have knowledgeable people around you in almost any field, and even better, they are happy to help you out. In my case, all I had to do was walk to the end of the hall, where Warren DeSorbo was doing experiments related to superconductors. I can't remember how long it took me to install the borrowed liquid helium dewar, but it may not be more than a day or two. People who are not familiar with low temperature work will think that the entire low temperature field is very profound, but the real thing to do is actually obtain liquid helium, which is easy to do in the laboratory, and the experimental device is shown in Figure 6. Then I made the sample using my familiar aluminum- alumina , but I added a lead strip to the top. Both lead and aluminum are superconductors, and lead is superconducting at 7.2K, so to make lead superconducting you only need to use liquid helium with a boiling point of 4.2K. Aluminum can only superconduct in the presence below 1.2K. In order to reach this temperature, we need a more complex experimental device.

Figure 7 Voltammetry characteristic curve of aluminum-alumina-lead sample. When lead enters the superconducting state, the current is no longer proportional to the voltage. There is a large change between 4.2K and 1.6K because the superconducting energy gap varies with temperature. When the potential energy provided by the applied voltage is less than half of the energy gap, there is still some current because of the thermal excitation of electrons in the conductor.

The first two experiments I tried failed because the oxide layer I added was so thick that the current was not large enough to be reliable measurements with the instruments I used, which were just standard voltmeters and standard ammeters. It feels strange whenever I think back to this, because just 13 years later, the labs are already full of complex x-y recorders. Of course, we had a lot of oscilloscopes at the time, but I wasn't very familiar with how to use them. In my third attempt, instead of deliberately oxidizing the first aluminum strip, I only exposed it to the air for a few minutes before putting back the cross strip of lead deposited in the coating. This way the oxide thickness will not exceed 30Å, and I can easily measure the volt-ampere characteristics with existing equipment. For me, the greatest moments in experiments always come before I understand whether a particular idea is good or bad, so even failure is exciting. Of course, most of my ideas are wrong. But this time it worked! When the electrode changes from the normal state to the superconducting state, the voltammetry characteristic curve changes significantly, as shown in Figure 7. This is quite exciting! I immediately repeated this experiment with different samples - everything looked good! But how to finalize this conclusion? As we all know, , superconductivity will be destroyed by magnetic fields, but the simple Dewar device I used is impossible to do this experiment. This time I had to walk through the entire hall to Israel Jacobs' laboratory for cryogenic magnetism. I was once again lucky to be able to use an experimental device directly, which can control both temperature and magnetic fields. Using this device, I can quickly complete all the experiments. The basic results are shown in Figure 8, and all the results are in line with very good results. I remember the whole group was very excited, especially Bean, who enthusiastically spread the news in our lab and patiently explained to me the importance of the experiment.

Figure 8 Voltammetry characteristic curve under different external magnetic fields at 1.6K. At 2400 Gauss, the lead film is in a normal state, and at 0 Gauss, the lead film is in a superconducting state. The change between 800 Gauss and 0 Gauss is due to the change of superconducting energy gap with the applied magnetic field.

Figure 9 Informal discussion of coffee time. From left: Ivar Giaever, Walter Harrison, Charles Bean, and John Fisher

Of course, I was not the first to measure superconducting energy gaps, and I soon discovered a beautiful experiment done by M. Tinkham and his students using infrared transmission. I still remember that I was worried that the energy gap size I measured was inconsistent with previous measurements.But Bean told me straight that from then on, others will have to endorse my results; my experiments will become the standard, I feel happy and for the first time I feel like I am a physicist.

Figure 10 Tunneling between two superconductors with different energy gaps when the temperature is greater than 0 K. (A). No voltage is applied between the two conductors. (B). When voltage is applied, more and more thermally excited electrons will flow from superconductors with smaller energy gaps to superconductors with larger energy gaps. At the voltage shown in the figure, all the excitation electrons can find empty state on the right. (C). As the voltage increases further, no more electrons will play a role, and the current will decrease with the increase of voltage as the number of states that can be accepted for tunneling electrons decreases. When the voltage is high enough, the electrons below the energy gap of the superconductor on the left correspond to the empty state on the right, and the current will increase rapidly. (D). Schematic diagram of expected voltammetry characteristics.

That was the most exciting time of my life; we had some great ideas to improve the experiment and expand it to various materials such as ordinary metals, magnetic materials and semiconductors. I remember we had a lot of informal discussions about what to try next during coffee time, one of which was recorded in a 1960 photo taken, as shown in Figure 9. To be honest, the photos are posed, our clothes are generally not so formal, and I myself would hardly be responsible for deducing them on the blackboard! Most of our ideas are not very effective, and Harrison quickly published a theory to prove that life is complicated and difficult to understand after all. But superconducting experiments are attractive and always effective. It seems that the probability of tunneling is proportional to the superconducting density of states. Now, if this is considered strictly correct, it is not difficult to realize that the tunneling between the two superconductors should reflect negative resistance characteristics, as shown in Figure 10. The negative resistance characteristic of course means amplifiers, oscillators and other devices. But I don't have equipment around that to make aluminum superconducting by reducing pressure by liquid helium. This time I had to leave our building and re-activate an old low-temperature device in the next building. Sure enough, once the aluminum superconducting occurs, negative resistance will appear. In fact, the view that the tunneling probability is proportional to the density of state is experimentally correct. Figure 11 shows a typical experimental feature.

Figure 11 Negative resistance characteristics obtained in tunneling experiments between two different superconductors.

Things are going very well because using this effect we can make all kinds of electronic devices, but of course, they can only work at low temperatures. We should remember that semiconductor devices were not that advanced in 1960, and we think superconducting junctions have great hope of competing with them (such as Ezaki diodes). The basic question I face is which path to take, engineering or science? I decided to do science first and got the full support of my boss, Roland Schmitt.

Looking back now, I realize how tempting it is for Schmidt to study this entirely new field, especially with so many experienced physicists around us. In contrast, Schmidt introduced me in due course a colleague, Karl Megerle, who joined our lab as a training researcher. I worked very well with McGell, and soon we published a paper discussing a lot of basic effects.

Figure 12 Normalized differential conductance of lead junction at low temperature. Simple BCS theory can predict that as energy increases, the differential conductance should asymptotic to one unit. But on the contrary, we can observe several swings between 4x superconducting energy gap and 8x superconducting energy gap. These swings are related to the phonon spectrum in lead.

It has always been important for physics to expand experiments to higher energy, stronger magnetic fields, or for our case to lower temperatures. So we worked with Howard Hart, who just built a helium 3 refrigerator that was able to lower the temperature to 0.3K. At the same time, McGell made a phase-locked amplifier, which we can use to directly measure the differential conductance.It was a pretty beautiful instrument in which the magnet rotated around the pickup coil at 8 turns per second, and of course it was far inferior to modern phase-locking amplifiers. We have long known that there are abnormalities in lead's voltampere characteristic curve, and now we can finally determine this through the additional swing measured on the differential conductance curve. As shown in Figure 12. This result makes us happy because all the tunneling experiments we are doing so far are to prove the BCS theory, and this is not something an experimental physicist really wants to do. Our dream is to prove that a famous theory is incorrect, and now we finally poked a hole in this theory. We speculated at the time that these swings were related to phonons, which were believed to be the source of electron-electron attraction interactions in superconductors. But as always, theoretical physicists turned the situation around. They cleverly use these swings to properly expand the theory and prove that the BCS theory is indeed correct. Professor Badin explained the result in detail in his recent Nobel speech.

So far, I mainly talk about General Electric's research situation at that time. Sometimes it is difficult for me to realize that Schenectady is not the center of the world. Several other groups have also begun to study tunneling. Here are a few brief introductions: J.M. Rowel and W.L. McMillan elucidate the phonon structure in superconductors; of course, W.J. Tomasch insists on finding his own discovery; while we were doing the experiment, S. Shapiro and his colleagues were also conducting tunneling experiments between two superconductors; J. Bardeen and later M.H. Cohen and others were responsible for most of the theoretical research.

Figure 13 The influence of bound magnetic field on tunneling characteristics. Curve 1 is an initial curve, with a medium-sized magnetic field in curve 3, and the magnetic field in curve 2 has been removed. There is a small dissipation-free current in the curve l, which we believe is caused by a metal short circuit. But now looking back, this is actually due to the Josephson effect!

At the same time, I completed my own course in RPI and decided to work with Professor Huntington on a theoretical topic about orderly-disordered alloys, because we have basically understood the tunneling effect in superconducting. Later, I was reminded that Brian Josephson published a short article in Physics Express—What did I think at the time? Well, I didn't understand. But I had the opportunity to meet Josephson himself in Cambridge and was impressed by his work. One of the effects Josephson predicted is that when both metals are in superconducting states, there is a possibility that supercurrent with zero voltage drops can pass through the oxide barrier. This phenomenon is now called the DC Josephson effect. We have measured this behavior many times, and in fact, it is difficult to see this current when measuring a "tin-tin oxide-tin" junction or a "lead-lead oxide-lead" junction. Early tunneling junctions usually use alumina, while alumina is generally thicker, so thermal fluctuations suppress the DC current. In McGell and my first paper there is a curve, as shown in Figure 13, showing such a supercurrent, and it is strongly dependent on the magnetic field. However, at that time I had a ready-made explanation for this phenomenon, that is, the short circuit of the metal electrode. I was confused at that time because, in theory, such a small contact should not have such a magnetic field, , and 7 degrees, but no one knows what will happen if a contact with a length of 20Å and 20Å wide. If I have learned something as a scientist, it is that when a simple explanation can solve the problem, things should not be complicated. Therefore, all samples we made that observed the Josephson effect were discarded as short-circuited samples. But obviously I thought it was too simple this time! Later I was asked many times, do I feel sad about missing this effect? The answer is of course no, because just observing some phenomena cannot be called experimental discoveries. We must also recognize the physical significance behind the phenomena, and in this example I am not even close to this.Even after I really understood the DC Josephson effect, I still felt that it was indistinguishable from short circuits, so I had mistakenly believed that only by observing the communication Josephson effect could the Josephson effect be confirmed or falsified.

In short, I hope this rather private narrative can provide some simple insight into the nature of scientific exploration. My own belief is that the path to scientific discovery is often less direct and it does not necessarily require high expertise. In fact, I believe that newcomers in a field often have great advantages because they are ignorant and do not know the complex reasons behind a specific experiment. However, when you need some suggestions for , it is important to be able to get help from experts in different fields. The most important thing for me is that I was in the right place at the right time and I had many selfless friends who supported me both inside and outside GE.