Direct Measurement of Itinerant Magnetism & Interface States in Semiconductors using Time-varying Magnetic Fields

Abstract

Magnetism in a solid | dia, para, ferro, or of other forms | originates majorly from its electrons; one could, infact, ignore the nuclear contribution. There are two types of electrons in a solid: bound, and free (also called itinerant). It is interesting to note that although several experimental techniques exist that measure the total magnetization/ susceptibility of a solid, no experiment directly probes the individual magnetic contributions from the bound and the itinerant electrons. In the past couple of decades, owing to the advent of sophisticated fabrication facilities, certain man-made, (ferro)magnetic materials have come into existence whose carrier concentrations can be tuned extrinsically: doped semiconductors like DMS (diluted magnetic semiconductors) and hexaborides are two such examples. However, whether the (ferro) magnetism in these materials originate from their itinerant carriers is still an open question. A conclusive answer to this question is eagerly awaited by the scientific community; the answer is not only supposed to solve debates related to the physics of ferromagnetism, but, also, should lend a helping hand in selecting right materials to build devices for upcoming exotic technologies such as Spintronics. A novel experimental technique is proposed in this work that directly measures the itinerant carrier magnetism of a solid. The technique is practically demonstrated on the bulk semiconductor: n-type GaAs. A Landau-Peierls itinerant (dia)magnetic susceptibility as low as 1 10 8 cm 3/mol | which is 10 3 times smaller than the magnetic background stemming from the bound electrons in the GaAs host lattice, and 10 times lower than the sensitivity limit of the SQUID | was clearly, and reproducibly detected from samples having carrier concentrations as low as 5 10 15 cm 3. The technique relies on measurements with MIS capacitors fabricated out of the given semiconductor. Unfortunately, as an artifact, such MIS fabrication processes unintentionally, but unavoidably, introduce certain energy levels in the semiconductor band-gap that unwantedly communicate with its bands by trapping and releasing carriers. Such traps lie along the interface of the semiconductor and the oxide. Though clear signals, which match with theoretically estimated signals within acceptable accuracy, have been measured from the itinerant electrons in GaAs, this work demonstrates theoretical calculations showing that the signals decrease in magnitude owing to the presence of such interface traps. Quantifying this decrement comes as an added advantage of this work, because such measurements can then directly probe the MIS interface and find the concentration of the interface traps (Dit) more accurately and precisely than what is done at present. Thus, the experimental technique this work proposes can also probe a given MIS interface, using time-varying magnetic fields, and reveal a more accurate and precise measure of Dit. Otherwise, the existing techniques for measuring Dit su er from imprecision caused by several theoretical assumptions. A more general technique which can extract Dit accurately and precisely, without needing to know the particular physical model that the interface traps follow for a given MIS capacitor, is what one requires at present, to give CMOS technology the direction and impetus it needs to cross-over to the non-Silicon territory. Such a technique is theoretically developed in this work. How a magnetic field a effects the MIS Energy Band Diagram is also derived in the process. The technique that is developed and demonstrated in this thesis, capable of directly probing both the itinerant magnetism and the MIS interface of a given semiconductor, depends on successfully measuring a very small voltage drop across a MIS capacitor when the latter is externally subjected to a high, time-varying magnetic field. This voltage signal originates because the semiconductor's electronic density of states depends on the magnetic field, thus rendering the semiconductor's electron chemical potential, i.e. the Fermi level, magnetic field dependent. The idea of detecting such magnetic field dependence of electron chemical potential was theoretically proposed more than five decades back, but an experimental detection of the phenomenon, in any bulk (i.e. three dimensional) solid, had remained elusive despite numerous trials. Virtually, the topic had been `dead' for the past couple of decades with very few reports (of trials) getting published on it. The primary reason behind such a failure is an interesting spurious effect that arises and overshadows the signal otherwise coming from the magnetic shift of the electron chemical potential. This is the spurious Hall voltage caused by the time-varying magnetic field and the eddy current it induces in the semiconductor following Faraday's Law of Electromagnetic Induction. Unless this Hall voltage can be reduced below a threshold, there is no hope of successfully measuring the sample signal. In this work, we have discussed about this spurious effect in details and have given experimental recipes to avoid it from interfering with the data. Infact the data we publish for n-GaAs is free from any such spurious effects. From that viewpoint, this work becomes the first to report the experimental detection of the magnetic field dependence of a Fermi level in any bulk solid. A common pulse magnet capable of producing high magnetic field pulses, lasting for only some tens of milliseconds, was built and used for the purpose of this work. For certain samples other than GaAs, however, the spurious Hall voltage may be larger and the proposed technique may fail as one may not be able to rule out the spurious effect with the simple recipe demonstrated here for GaAs. In such a case, measurements are encouraged, instead, in a special magnet uniquely developed to rule out the Hall voltage. This magnet was constructed in-house, and can sit on a table-top and generate magnetic fields as high as a few Teslas that can, further, be `temporally shaped' by the user. Such a class of pulse magnets whose pulse waveforms can be programmed over time are called controlled waveform magnets (CWMs) and the work presented in this thesis also demonstrates the construction and calibration of such a CWM.