This note is about uniform, plane, singly connected, regular Hall-plates with an arbitrary number of contacts exposed to a uniform magnetic field of arbitrary strength. In practice, the regular symmetry is the most common one. If the Hall-plates are mapped conformally to the unit disk, regular symmetry means that all contacts are equally large and all contact spacings are equally large, yet the contact spacing may be different from the size of the contacts. The indefinite conductance matrices of such Hall-plates are circulant matrices, whose complex eigenvalues are computed in closed form.
Definitions
In electrically linear Hall-plates the currents into the terminals are linear combinations of the potentials at the terminals. Suppose that the Hall-plate has N terminals. We group all N currents to a vector I and all N voltages to a vector V. Then we express the linear combination as a matrix multiplication I=iGV, with the indefinite conductance matrix iG. Thereby, the reference potential (ground node) is arbitrary – none of the terminals needs to be grounded. If we ground the ℓ-th terminal, we write I=GV, where we delete the ℓ-th current and voltage in I,V, respectively, and we delete the ℓ-th row and column in iG to get G. The indefinite conductance matrix is positive semi-definite – its determinant vanishes and therefore it cannot be inverted. The definite conductance matrix of any passive (= dissipative) system is positive definite, its determinant is positive, and an inverse exists: V=RI with the resistance matrix R=G−1.
In the presence of the Hall-effect Ohm’s law is
E=ρJ−ρμHJ×(Bznz),(1)
with the electric field E, the current density J, the unit vector nz orthogonal to the plane Hall-plate, with the z-component of the magnetic flux density Bz, and with the Hall-mobility μH. Thus, E and J are not colinear – there is the Hall-angle θH in-between, with tan(θH)=μHBz. We use the following abbreviations,
Rsheet=tHρ,Rsq=cos(θH)Rsheet,(2)
whereby tH is the thickness of the Hall-plate. If we have a square plate with two contacts fully covering two opposite edges, the resistance between these contacts is the sheet resistance Rsheet at zero magnetic field, and it is the square resistance Rsq in the presence of a magnetic field1.
Regular Hall-plates
Suppose we have a regular boundary with N contacts of equal size, the spacings between the contacts are all equal, yet they are not necessarily identical to the size of the contacts (see Figure 1). Suppose we connect an external circuit to
Figure 1. Examples of strictly regular and regular Hall-plates with four contacts.
this Hall-plate. If we keep the Hall-plate fixed and move the external circuit by one terminal, say clockwise, then the resulting potentials and currents into the external circuit must remain the same, because the Hall-plate is regular symmetric in shape and the applied magnetic field is parallel to the rotation axis. This means that the entries in the current and voltage vectors move one instance downwards, and the entries in the indefinite conductance matrix move one row down and one column to the right. Both the original and the new iG matrices must be identical. We can repeat this procedure N times. The conclusion is that the elements along diagonals parallel to the main diagonal are identical, which gives the following patterns for iG and G,
iG is an N×N matrix, G is an (N−1)×(N−1) matrix. We get G from iG by deleting the row and column which corresponds to the grounded contact. G and iG are Toeplitz matrices, which have the property that Gℓ,m=gℓ−m – it depends only on ℓ−m. Moreover, iG is a circulant matrix with gℓ−m=gℓ−m+N. Third, the sum of entries per row in any indefinite conductance matrix has to vanish, because no currents flow if all contacts are at identical non-zero potential.
If we map a regular Hall-plate to the unit disk, its contact Cm ranges from azimuthal angle
αm=2πNm−1toβm=αm+χNπ,(4)
for χ∈(0,2) and m=1,2,…,N. Contacts are small for $\chi \to 0, whereas contact spacings are small for χ→2. The case χ=1 gives strictly regular Hall-plates, whose contacts are equally large as the contact spacings (see Figure 1).
We can write down the eigenvalues and eigenvectors of a real-valued circulant matrix2
iG=QiΓQCwithiΓ=diag(iγ1,iγ2,…,iγN),
(Q)k,ℓ=N1exp(N2πikℓ)∀k,ℓ∈{1,2,…,N}.(5)
QC is the conjugate complex of Q, and i=−1. iΓ is a diagonal matrix, which comprises all eigenvalues iγℓ of iG. It holds Q=QT and QQC=QCQ=1. The eigenvectors of iG are the columns of Q, which are coefficient vectors of the discrete Fourier transform (DFT). It generally holds
iγN−ℓ=iγℓC,(6)
for the eigenvalues of any real circulant matrix, where zC is the conjugate complex of z.
How to compute the eigenvalues of iG
We use Ref. 2. Their equation (17) reads in our notation
whereby ϕm is the potential on Cm (= the m-th peripheral contact), and ψm is the stream function on the insulating peripheral segment between Cm and Cm+1, with CN=C0. In Ref. 2 the matrices A,B are defined as
Note that in (7) the sums extend over N terms, whereas in Ref. 2 they had only N−1 terms. Hence, in contrast to Ref. 2, we do not require ϕN=0 and ψN=0. This enables us to work with the indefinite conductance matrix, which is circulant, and therefore it lends for a simpler mathematical treatment than the definite conductance matrix. Therefore, our matrices A,B have N columns and N−1 rows. Equation (7) must hold, if we add an arbitrary constant to all ϕk or to all ψk. Consequently it must hold
m=1∑NAk,m=0andm=1∑NBk,m=0for k∈{1,…,N−1}.(9)
Multiplying the left equation of (9) with ψN/cos(θH) and adding it to (7) gives
This equation is valid for arbitrary voltage vectors ϕ, therefore we can skip ϕ on both sides of (13). Next we introduce the matrix
1^=00⋮0110⋮0001⋮00⋯⋯⋱⋯⋯00⋮10.(14)
1^ is a circulant N×N matrix, which is obtained from the unit matrix 1 by shifting all entries up once, in a rolling way. It holds 1^−1=1^T, det(1−1^)=0, and det(1−1^T)=0. Multiplying (13) from right with (1−1^T) gives
whereby we used (9) in the last equality of (16). Equation (16) is remarkable, because AL1(1−I^T)=A even though L1(1−I^T)=1. Inserting (16) into (15) gives
The properties of these eigenvalues will be discussed elsewhere. They can be used in (5) to compute the conductance matrix, which gives all electrical properties of the Hall-plate (impedances, response to magnetic field, thermal noise, signal-to-thermal-noise ratio over power).
Inserting (4) into (8) and using the product formula3 gives
Finally, this gives the complex eigenvalues iγm,m∈{1,2,…,N} of the indefinite conductance matrix iG of a regular symmetric Hall-plate with N contacts as a function of the Hall-angle θH and of the contact size χ,
The sum over index k vanishes, except for m−1−ℓ−q=−pN with p=0,1,2,…, hence it equals Nδq,m−1−ℓ+pN. We re-arrange the double sum over indices ℓ,q as a double sum over indices a,b with a=ℓ−q,b=ℓ+q (see Figure 2). With
Figure 2. Transformation of indices for a double sum.
[1]1H.J. Lippmann, F. Kuhrt, "Der Geometrieeinfluss auf den transversalen magnetischen Widerstandseffekt bei rechteckförmigen Halbleiterplatten," Zeitschrift für Naturforschung A 13, 462 (1958). (in German)
[2]2D. Homontcovschi, B.T. Murray, "Explicit resistance matrix for a Hall disk with multiple peripheral contacts: Application to a van der Pauw type method for extended contacts," Sensors and Actuators A: Physical 294, 1 (2019).
[3]3U. Ausscherlner, "The classical Hall effect in multiply-connected plane regions part I: topologies with stream function," Journal of Applied Mathematics and Physics 7(9), 1968 (2019).
[4]4R.M. Gray, "Toeplitz and circulant matrices: A review," Foundations and Trends in Communications and Information Theory 3(2), 155 (2006).DOI
[5]5I.M. Ryshik, I.S. Gradstein, Tables of Series, Products, and Integrals, (VEB Deutscher Verlag der Wissenschaften Berlin, 1957), page 34, equation 1.392.