Simple Proofs of Upper and Lower Envelopes of Van Der Pauw’s Equation for Hall-Plates with an Insulated Hole and Four Peripheral Point-Contacts ()
1. Introduction
The sheet resistance of a plane conductive layer is of prime importance in thin layer technologies. It is used pervasively in micro-electronic manufacturing to monitor the properties of thin conductive layers. It is given by
, where
is the conductivity and
is the thickness of the layer. Van der Pauw showed that it is possible to derive the sheet resistance from purely electrical measurements of currents and voltages, no other geometrical parameters are necessary [1] [2]. Yet some general requirements have to be granted: the conductive layer has to be plane, its resistivity and thickness must be homogeneous, the contacts must be small (point-sized), and the contacts must be on the circumference (=peripheral contacts) of a singly-connected region (no holes). Homogeneous resistivity also means a linear material law of electric conduction where resistivity is constant versus electric field, without self-heating, and without self-magnetic field. The resistivity is allowed to be anisotropic with a symmetric resistivity tensor. In this case we can apply an isotropization procedure as shown in [3] [4] [5]. At the beginning we rule out anti-symmetric resisitivity tensors as they occur if magnetic fields are applied, but Section 4 extends the range of validity to include the Hall-effect.
The plane conductive region of a conventional Hall-plate has four peripheral point-sized contacts with consecutive labels 0, 1, 2, 3 in a positive mathematical direction (i.e. counter-clockwise). Thus, if we move along the boundary from contact 0 via 1 and 2 to 3 (in ascending order) the conductive region is on the left hand side. If current is forced to flow between two contacts and the voltage is tapped between two contacts, then van der Pauw used the ratio
(1)
are the electric potentials at the m-th and n-th contacts, and
is the current entering the conductive region through contact k and leaving it through contact
. The quantity
has the dimension “voltage over current”, which is a resistance. The contacts for the current may be the same or they may be different from the contacts for tapping the voltage, in the latter case we call
a trans-resistance. Depending on the sequence of the indices, trans-resistances may be positive or negative. Van der Pauw chose the sign in the definition (1) such that for rising sequence of contacts and their cyclic permutations the trans-resistances are positive,
. For point-sized contacts, the trans-resistances are the only finite resistances.
For the sake of brevity, I define van der Pauw’s function
(2)
X and Y are abbreviations for the exponential terms. Then the basic result for Hall-plates without a hole in [1] is van der Pauw’s equation,
(3)
In the van der Pauw plane
Equation (3) is a straight line
. The peculiarity of (3) lies in the fact that it relates measurable electrical quantities
to the sheet resistance
irrespective of any geometrical details, neither the shape of the Hall-plate nor the locations of the contacts are specified. Having measured the two trans-resistances, one can solve the nonlinear Equation (3) to get the sheet resistance. This is van der Pauw’s method to determine the sheet resistance. It works well in many practical cases.
However, occasionally one faces the problem that some of the above given requirements are not fullfilled. Then van der Pauw’s method gives inaccurate or even wrong results for the sheet resistance. This is an inherent problem in materials science, where one needs to characterize novel materials. Often these samples have poor quality and poor homogeneity due to limitations in the manufacturing process, especially when the fabrication on a small laboratory scale is not yet mature [6] [7] [8]. From a practical standpoint, one would like to have a procedure that detects poor sample homogeneity and that gives error bounds for the derived sheet resistance. Inhomogeneous conductivity is supposed to have a similar effect to small voids. This is the motivation to study Hall-plates with holes.
The topic was pioneered over the last decade in a couple of papers by Szymański and coworkers [9] [10] [11] [12]. They introduced the concept of upper and lower envelopes (u.e., l.e.) for conductive samples with a hole,
(4)
For constant hole size, the lower envelope depends only on X. Further contributions came from [13]. For nearly one decade (4) was just a conjecture, while a strict mathematical proof was missing, until only recently a thesis solved this problem (with its potential generalization to more than one hole) at a fairly elaborate mathematical level [14] [15]. The present article gives new and simpler proofs for samples with a single hole with less sophisticated mathematics and closer to the physical intuition of an electrical engineer. The employed mathematical tools are series expansions and conformal transformations which lead to Jacobi functions and elliptic integrals.
There is a certain similarity of the current topic with another topic called the Hall/Anti-Hall bar [16] [17]. No Hall voltage appears there between any two points on the hole boundary if current flows between two points on the outer boundary of the Hall-plate. In van der Pauw’s measurement current flows between neighboring contacts and voltages are also tapped between neighboring contacts, whereas in common Hall-plates current flows between non-neighboring contacts and voltages are tapped between other non-neighboring contacts. In the Hall/Anti-Hall bar we cannot speak of neighboring contacts anymore, because current and voltage contacts are on different boundaries. The focus of interest in the Hall/Anti-Hall bar lies on the case of applied magnetic field (i.e., the Hall-effect with non-reciprocal conductivity tensor), whereas the focus of the present article lies on the case of zero magnetic field (i.e., simple ohmic conduction with scalar conductivity).
This article starts with the easier case of a small hole, which leads us straight to the star-configuration and the minimum of the van der Pauw function. Then we compute the trans-resistances for arbitrary hole size with conformal transformations, and we prove the upper and lower envelopes. We discuss some properties of the trans-resistances and how they are affected by a magnetic field. Finally, we check the derived formulae with numerical simulations.
2. Hall-Plates with a Small Hole
2.1. Series Expansion of the Potential
Let us start with a plane irregular ideal Hall-plate with a single irregular hole of arbitrary size. The entire inner and outer boundary is insulating except for four point-sized contacts
. Current
is injected by an ideal current source at
and extracted at
while the voltage from
to
is measured. We know from Riemann that a conformal map exists, which maps the irregular Hall-plate onto the unit disk with a central hole of radius
, whereby
is the Riemann modulus of the singly-connected domain. Let us rotate the disk such that the current contacts
are symmetrical to the real axis. Then, the azimuthal locations of the contacts are (see Figure 1)
(5)
with
being the azimuthal angle of the location of
. The electrostatic potential at zero magnetic field,
, is given in [17]. At the outer perimeter, it holds
(6)
(6) is derived from a Fourier series which solves the Laplace equation of the potential in the annular region with insulating boundary conditions at the perimeter and at the hole. With
and
(7)
(see (102) in Appendix A) this is
(8)
With (105) we get
Figure 1. Plane annular Hall-plate with insulating boundaries and point-sized peripheral contacts
. Current flows from
to
. Voltage is tapped between
and
. The potentials in
are identical to the potentials in
, respectively (see Section 3.2).
(9)
Rearranging this gives
(10)
which is equivalent to
(11)
The term
corresponds to the singly-connected Hall region with
. The measured van-der-Pauw voltage is
. With
this gives
(12)
If we inject the current at
, extract it at
and measure
we can re-use (12) if we replace
(13)
With
this gives
(14)
A Taylor series for small holes
keeps only the terms
,
(15)
with
(16)
are defined in (2),
are the values for
, and
are the lowest order terms in
.
(17)
In this equation, the coefficient of
is positive, because
and
. This is the simple proof that a small insulating hole reduces the van der Pauw function below 1. Equation (17) is also derived in [9].
2.2. Derivation of the Star-Configuration of Contacts
A general contact arrangement is defined by three parameters
. For a specific set
a certain value for
follows. Yet, according to (16) there are many other sets
which give the same
. Which of all these sets causes the steepest drop of
for small holes? In other words, for fixed trans-resistances of a singly-connected Hall-plate, how do we have to place the contacts such that the van der Pauw function becomes most sensitive to a small nucleating hole? Keeping
fixed implicitly defines the azimuthal position
as a funcion of the other two positions,
. From
it follows
and
, which gives
(18)
The minimum of
means the largest negative slope of
versus
for
. There it holds
and
, which gives
(19)
We solve (18) for
and
and insert this into (19). With (16) and after some manipulation this gives
(20)
with the only meaningful solution
(21)
Let us call this specific pattern of contacts a star-configuration—the contacts are in the vertices of a rectangle inscribed into the perimeter of the Hall-plate.
For fixed trans-resistances
and
the drop in the van der Pauw function caused by a small hole gets largest, if the contacts are in a star-configuration.
Inserting (21) into (17) gives
(22)
From all star-configurations the one with the steepest decline of
versus
is for
(23)
where all four contacts are equidistant, i.e., they are in the vertices of a square inscribed in the perimeter of the Hall-plate. This configuration gives the smallest possible
for a given hole of small size
,
(24)
Inserting (21) into (16) into (15) gives
(25)
Eliminating
from (25) gives a curve in the van der Pauw plane
, which holds for star-configurations with small holes,
,
(26)
This is the small-hole approximation of the lower envelope as it will be explained in Section 3.3.
For holes of arbitrary size a strict proof of
appears to be difficult, because the trans-resistance
may increase or decrease versus hole radius
for holes of small and moderate size.
Example: For the Hall-plate in Figure 1 set
,
. Then
(and consequently
) increases for
while it decreases for
when the hole grows from
(see also curves 1, 4 in Figure 9(a)).
The decrease of
may come a bit surprizingly: a trans-resistance may get smaller if one cuts out a bigger hole of the conductive medium. For the explanation we consider an elongated asymmetric hole in radial direction extending from the center of the disk up to very close to the perimeter, see Figure 2. This asymmetric geometry can be mapped onto a symmetric one with a central circular hole (Any plane domain with a single hole can be mapped conformally to a circular ring with inner radius
and outer radius 1 [18]). The radial slit may be placed in-between the two voltage taps. This increases the trans-resistance. However, it may also be placed outside the two voltage taps. Then, a larger fraction of the total supply voltage drops outside the voltage taps, and therefore the trans-resistance becomes smaller. Thus, by the placement of the hole one can make the voltage
smaller or larger.
Next we have a look at the trans-resistances of Hall-plates with contacts in a star-symmetry as defined in (21) and shown in Figure 3(a). Without loss of generality, the restriction
is used, if
we only have to shift the indices of all contacts by one instance further to pull
again inside
. From (12) and (14) we get
(27)
In contrast to the example given above, both trans-resistances increase with growing hole if the contacts are in a star-configuration. This can be readily seen in (27). The plot in Figure 3(b) visualizes this fact. The inequality (17) holds for small holes, and according to (27) both trans-resistances increase for larger holes. Therefore the inequality
holds also for large holes in the case of a
-symmetry.
(a) (b)
Figure 2. A radial slit may increase or decrease a trans-resistance depending on its location. Yet, if it increases the first trans-resistance
, it decreases the second trans-resistance
. (a) Circular Hall-plate with a radial slit between the voltage taps
increases
when the hole grows; (b) Circular Hall-plate with a radial slit outside the voltage taps
decreases
when the hole grows.
(a) (b) (c)
Figure 3. A symmetrical annular Hall-plate with four symmetrical point contacts has three degrees of freedom,
,
, and
. Its transresistance
is a concave surface above the
-plane. Two values of trans-resistances,
and
, give two curves in the
-plane, which intersect in a unique point
. (a) A symmetric annular Hall-plate (
) with four point-contacts in the vertices of an inscribed rectangle; (b) Its normalized trans-resistance
versus
and
is a concave surface subtending all values
; (c) Two intersecting curves are generated, when the surfaces
and
are cut through at different heights
and
.
The two functions in (27) have a couple of useful properties. Not only are they monotonic in
, they are also monotonic in
in the relevant interval
.
strictly increases with
, whereas
strictly decreases. Both functions are mirrored at
(note that the terms
and
are swapped in the two equations in (27)). For a fixed value of
the trans-resistance
goes from
if
goes from
. Conversely,
goes from
if
goes from
. Hence,
goes up monotonically if one moves radially away from the origin in the
-plane. The surface in Figure 3(b) is concave. If the left hand sides in (27) are given, each of the two equations gives a curve in the
-plane in the domain
. The first curve encircles the origin while the second curve encircles the point
(because of the mirror symmetry of both surfaces, see also Figure 3(c)). Each curve gives
as a strictly monotonic function of
. If we intersect the
-surface at two different heights we get two curves, which have no common point (they do not intersect and they do not touch). If we mirror the second curve at
this reflects the measurement of the second trans-resistance
. If the two curves cross, due to their monotonicity they have to cross in a single uniquely defined point, which gives
and
as shown in the example of Figure 3(c). However, if both trans-resistances are very small this would shift the
-curve left of
and the
-curve right of
, such that the two curves would not cross at all. In this case the
-curve starts at
at
in Figure 3(c) while the
-curve starts at
. With (27) it follows
(28)
because it holds
for
. Equation (28) contradicts the classical van der Pauw Equation (3) for singly-connected plates (
), and therefore we can rule out this case. Thus, we have proven that…
…for any
-arrangement of contacts with fixed sheet resistance the measurement of both trans-resistances
defines two curves like in Figure 3(c), which intersect in exactly one point. This point specifies the hole radius
and the locations
of the contacts.
For any doubly-connected Hall-plate with arbitrary
we can find a
-configuration of contacts with
, which has the same trans-resistances
, but generally different hole size
and different sheet resistance
.
The proof goes like this:
Suppose we have a general asymmetrical contact placement, which gives two measurement results
. If accidentally the measurement returns
we simply swap the two trans-resistances by moving all contacts one instance further. Now we consider a hypothetical Hall-plate with a
-symmetry as in Figure 3(a). We choose its sheet resistance
(29)
Then it follows from (27) that
, if this Hall-plate had no hole,
. If it has a hole,
is smaller, but at this moment we do not know anything about the hole. If we set
we get a first curve in the
-plane which starts at
and encircles the origin counter-clockwise. We also set
, which gives a second curve that encircles the point
in the
-plane clock-wise. Since
the second curve starts at a point on the
-axis, which is left of
. Therefore, the two curves must intersect. Since all curves are strictly monotonic in
, they must intersect in a single point only. This gives the unique solution of a hypothetical
-Hall-plate, which has the same trans-resistances as our original Hall-plate, albeit it has a different hole and a different sheet resistance.
This argument clearly shows that one cannot determine the sheet resistance of a doubly-connected Hall-plate with the measurement of both trans-resistances as in the singly-connected case, unless one has additional information about the hole or the contacts placements. In general it holds
, either one can be larger than the other one. For the
-case the value of the van der Pauw function
is bounded: we insert (29) into (2)
(30)
which is fulfilled due to our assumption
.
3. Hall-Plates with a Large Hole
3.1. Conformal Mapping of the Annular Hall-Plate
Next we apply conformal mapping to the general ring-shaped Hall-plate from Figure 1. Since the current contacts are symmetric to the real axis it is clear that all points
and
on the real axis are at the same potential, say 0 V. There we can insert a contact. We can further cut the ring apart at the positive real axis, apply contacts at both cut edges, and short them with a wire (see Figure 4(a)), without affecting the potential in the annular region. From the discussion in [17] we know that the fraction
of the current flows through the shorted wire, independent of the size of the hole. The conformal transformation
(31)
maps the annulus in the z-plane to the rectangle in the w-plane shown in Figure 4(b). The width of this rectangle is
and its height is
. The outer perimeter of the ring in the z-plane appears at the right edge of the rectangle in the w-plane, whereas the hole boundary appears at the left side of the rectangle. A Schwartz-Christoffel transformation maps this rectangle from the w-plane onto the upper half of the
-plane in Figure 4(c),
(32)
In the w-plane the current contacts
are placed symmetrically to the large contacts
. Thus, also in the
-plane they are symmetrically to the large contacts. From the sequential order of the points on the rectangular boundary in the w-plane it follows the same order in the
-plane,
(33)
It holds
(34)
where
are the locations of the points
in the w-plane, respectively,
is the complementary complete elliptic integral of the first kind (see Appendix B), and
. It also holds
(35)
Combining (34) and (35) gives
(36)
with the modular lambda elliptic function
(see Appendix B). Inserting the right side of (36) into (35) gives the scaling constant
of the mapping (32). The locations
of the point current contacts
in the
-plane follow from
(37)
with the Jacobi-sine function
from the Appendix B. In an analogous way we find the locations of the voltage taps
in the
-plane,
(38)
A final transformation maps the upper half of the
-plane onto the infinite stripe in the t-plane in Figure 4(d),
(39)
The point of the input current is at
, the point of the output current is at
. The structure is folded in such a way that the large contacts
and
are placed back to back: current
exits the right upper part of the stripe through contact
and it enters the left upper part of the stripe through contact
. The hole degenerates to a slit
with zero width. The slit is aligned in current flow direction. The points K and M are the stagnation points of the current flow pattern. The exterior angles at points
are
, at points
they are
, which brings the terms
to the numerator and the terms
to the denominator of the integrand in (39). The ultimate goal of all these transformations is to achieve homogeneous current density in the stripe in the t-plane. Then the distance between points
and
gives the voltage
. There are still two unknowns
to be determined. With
we make the width of the slit zero,
(40)
With the substitution
and with [19] this gives
(41)
with the complete elliptic integral of the third kind
(see Appendix B). The scaling constant
follows from
(42)
whereby the integration path is an infinitely small semi-circle around
, i.e.,
with
and
(see Figure 4(c)). We arbitrarily choose the width of the stripe equal to 1. With
it follows
(43)
from which we get
. The measured voltage is
with
(44)
which is split up in two integrals
(45)
(the equality at the right side comes from (113) and (38)) and
(46)
Both integrals (45) and (46) are solved by substituting
.
is the incomplete elliptic integral of the third kind (see Appendix B). Summing up the results of (43) - (46) gives the trans-resistance as a function of parameters in the
-plane,
(47)
The hole size is reflected by
(see (36)), and the three azimuthal positions of the point contacts are given by
(see (37), (39)). Expressing the trans-resistance in terms of the physical parameters
gives
(48)
In (48) sn, cn, dn are Jacobi functions (see Appendix B). The modular lambda elliptic function L simply scales
in a highly non-linear way. For the second trans-resistance we can use the replacements (13) in (48). With these formulae for the trans-resistances
and
we will proof two basic properties of doubly-connected plates with peripheral point contacts in the following sections.
3.2. Proof of the Upper Envelope
The upper envelope was first conjectured in [9]. It reads
(49)
for arbitrary placement of the point-contacts on the outer perimeter of a Hall-plate with one insulated hole of arbitrary size. The inequality (49) was proven recently in [14] by arguments using the prime function and Fay’s trisecant identity. This Section presents an alternative proof based on the conformal mapping in Figure 4(d). It is short and elegant and it needs no numerical computations.
We start with a general contact arrangement in Figure 4(a),
(50)
If accidentally
we shift all contacts by one instance to get
. The current splits in two parts, one flowing left around the hole and the other one flowing right around the hole. Thus, there must be a point F' right of the hole, which has the same potential as point F (=contact
) left of the hole. There must also be a point D' right of the hole, which has the same potential as point D (=contact
) left of the hole. Let us call the potential in point F
, in point D
, in point F'
, and in point D'
. Then it holds
and this means
. In Figure 4(d) we can easily localize points F' and D'. Point F' has the same horizontal position as point F, however, point F' is on the upper edge of the stripe, whereas point F is on the lower edge. The same applies to points D and D'.
When the second trans-resistance
is measured, current flows between points C (=contact
) and D (=contact
) and the voltage is measured between points G (=contact
) and F (=contact
). Analogously, for
current flows between points C and
and the voltage is tapped between points G and F'. However, Figures 4(a)-(d) do not apply in this case, because now the potential distribution is asymmetric. Hence the potential along the straight line
is not constant and therefore we are not allowed to insert an extended contact there. In fact we have to step back to (31) which maps the annulus of Figure 1 without a cut and without large contacts
to an infinite stripe made up of rectangles like in Figure 4(b) lined up along the
-direction yet without the extended contacts. Instead of the annulus we can think of a helical track that winds around the out-of-plane axis of Figure 1 infinitely often, whereby all four contacts repeat after every full revolution. This is shown in Figure 5(a), where we have infinitely many current and voltage contacts, each ones shorted with a pole, and the potential is periodic in each turn of the spiral. The first turn of the spiral for azimuthal angles
is called the Riemann sheet #0. It is followed by Riemann sheet #1 for azimuthal angles
and it is preceded by Riemann sheet
for azimuthal angles
. This trick extablishes an equivalence between the doubly-connected domain in Figure 1 and the infinite singly-connected domain in Figure 5(a) (in
fact we may also consider it doubly-connected because it closes at infinity, thus we have shifted the closure to infinity). Applying the transformation (31) to Figure 1 gives an infinite stripe made up of infinitely many replications of the rectangle of Figure 4(b) with all its contacts. The Schwartz-Christoffel transfor-mation (32) maps this infinite stripe to infinitely many Riemann sheets, which all look like in Figure 4(c), yet the potentials along
and
are not homogeneous. Instead, Riemann sheet #0 is connected to Riemann sheet #1 along
and it is connected to Riemann sheet
along
. The final mapping (39) gives a structure like in Figure 5(b), which comprises infinitely many storeys. Each storey represents one Riemann sheet. Each storey is connected to the upper one along
and to the lower one through
of Figure 4(d). The voltage and current contacts are at identical positions in all storeys, and the potential is identical in all storeys. This justifies our last step, where we collaps all storeys to a single one, whereby we can join the loose ends
with
in such a way that A coincides with J and B coincides with H. This final domain is identical to the one in Figure 4(d) with the only difference that the potential along line
is not homogeneous, and therefore the contacts
are deleted and the edges
are glued together. This is shown in Figure 6.
Now we consider the measurement of
in Figure 6. Thereby, current flows between
and
. We may choose the polarity of the current arbitrarily, for physical intuition it might be simpler to inject the current in point D instead of point C and extract it at point C instead of D. Analogously,
is measured by injecting current at point
, extracting it at point C, and measuring the voltage between points
and G. Now the slit plays a decisive role: since we started with
the slit in Figure 6 is closer to the lower edge with points
than to the upper edge with points
. The slit represents an obstacle to the current flow, and therefore the voltage between points
must be larger than the voltage between points
. Hence, it holds
.
The reciprocity principle [20] says that at zero magnetic field the voltage between
for current flowing between GC is identical to the voltage between
Figure 6. Annular Hall-plate in the t-plane with longitudinal slit. Current may flow between arbitrary contacts
, in this example the current flows from G to C, which gives a homogeneous current density with
.
GC for current flowing between
. In fact this also holds in the presence of a magnetic field as long as the entire boundary is insulating with all point-sized current contacts on the same boundary with G and C being neighbours as well as
and
, see Section 4.
To sum up, we have two sets of contacts in Figure 1, the original points
and the new points
, whereby the first trans-resistances are identical,
, but the second trans-resistance is smaller for the new points,
.
I call this transformation
a contraction, because the new points are closer together than the old ones.
Let us repeat the contraction process infinitely often, until all four contacts are infinitely close together.
In this limit the contacts are so close together that the current arcs between the current contacts are tiny. Then the hole is comparatively distant and it does not affect the current distribution any more. Thus the potentials at the voltage contacts become identical to the potentials in a singly-connected Hall plate.
However, for singly-connected Hall plates the van der Pauw Equation (3) holds. Since the second trans-resistance decreased during the contraction process, the inequality (49) must hold before contraction. This completes the proof.
The essential step in the proof was to show that
holds. To this end we used the arguments of the multi-storey Hall-plate in Figure 5 to justify Figure 6, in which the slit was a bigger obstacle for
than for
. We can avoid the use of multi-storey Hall-plates by the following line of arguments. We use Figure 1. In the measurement of
current
flows from
to
and voltage
is tapped. Thereby a first current
flows clockwise around the hole [17]. In the measurement of
we inject the same current
into
and extract it at
and we tap the voltage
. Thereby a second current
flows counter-clockwise around the hole, whereby
is the azimuthal position of
. If
is in the lower half of the z-plane it holds
and then the current is
. We have to prove that
. Per definition, the point
was obtained from
by a contraction process, therefore it holds
. If
is in the lower half of the z-plane it holds
. Consequently, in any case the first current is larger than the second current, because
is more distant from
than
is from
. If we superimpose both measurements, identical currents
flow simultaneously from
to
and from
to
and a positive net current flows clockwise around the hole in a direction from
towards
. Since there are no current sources except in
the potential drops monotonically along the clockwise current streamline on the outer perimeter from
to
. This means
, which means
, which again completes this proof.
3.3. Derivation of the Lower Envelope
The upper envelope theorem implies that for any doubly-connected Hall-plate we can find a
-configuration which has the same trans-resistances
and the same sheet resistance. This is not a specific property of
-configurations. Also other contact patterns have the same property: e.g. contacts with
and
, let us call them type 2 configurations, are also able to assume any physically meaningful pairs of values for the two trans-resistances (see Figure 7).
In the van der Pauw plane of Figure 7, a specific Hall-plate is represented by a dot. During the contraction process (c.p.) this point moves vertically up in the van der Pauw plane until it finally is on the straight line
, which is the upper envelope (u.e.). If we reverse the contraction process we can expand the contact arrangement, whereby the point moves down in the van der Pauw plane. However, this expansion process (e.p.) stops when the spacing between the current contacts,
, is larger than the spacings of all other neighbouring contacts,
, and the voltage contacts are in the obtuse angle of the current contacts. This brings us in a natural way to the question of the smallest possible Y and the smallest possible
for fixed X. For the
-configuration of contacts we know that the van der Pauw function decreases with larger holes and for
. On the other hand, the van der Pauw function tends to its maximum of 1 if only the contact arrangement is contracted sufficiently, or if
in a
-configuration. Then, one trans-resistance goes to infinity and the other one to zero. Thus, the question arises, what is the minimum van der Pauw function, if the hole and one trans-resistance are fixed. In other words, what are the maximum second trans-resistance and its associated contact locations? For every arbitrarily chosen first trans-resistance we get a maximum second trans-resistance. The set of all these pairs is called the lower envelope (l.e), because all Hall-plates with a fixed Riemann modulus are repesented by points between upper and lower envelopes.
Figure 7. In the van der Pauw plane
each Hall-plate is represented by a dot. It moves up in the contraction process (c.p.) and down in the expansion process (e.p.). The upper envelope (u.e.) is the line
. The lower envelope (l.e) is given by (69). Type 2 and type 3 contact arrangements give the curves 2 and 3, respectively. The three curves
assume a hole of size
.
[9] conjectured that the lower envelope is given by
-contact arrangements and this was also proven in [14]. For small holes, our derivation of (21) leads straight to the same conjecture. A precise statement of the lower envelope reads:
For a fixed trans-resistance
the arrangement of the point contacts for largest
is a
-arrangement.
3.4. A Proof of the Lower Envelope
In mathematical terms the lower envelope is defined like this:
Equation (48) gives
as a function of the contacts’ positions
. If
has to remain constant, this defines
as an implicit function of
and
. If we compute
analogous to (48) and insert the implicit function of
, this gives a function of two degrees of freedom,
. We want to prove that this function assumes a unique maximum if
and
, which is the
-configuration as it is defined in (21).
Put in another way,
(51)
whereby the first line of (51) reflects the constancy of trans-resistance
and the second line defines the extremum of
. The proof gets simpler if we apply the following transformations. Instead of the free parameters
we use (37), (38) and (113) to introduce new parameters
,
(52)
with
and
like in (48). From (50) it follows
(53)
The replacements (13) were used to compute
with the same formula as
. In terms of the new parameters
the replacement rules become
(54)
Analogous to (21) the
-configuration is specified by
(55)
We define the following function
(56)
where we skipped the second argument in the Jacobi functions, e.g.,
. With (48) it holds
(57)
The first part of the upper envelope theorem requires constant
, which means
with the implicit function
. This gives
(58)
in
-configuration, which is in
,
,
according to (55). From (56) it follows
(59)
Inserting this into (58) and adding up both equations gives
(60)
Inserting (59) into the second line of (58) gives
(61)
The second part of the upper envelope theorem says that
has an extremum, which means
in
-configuration with
according to (55). This gives
(62)
in
-configuration. Adding both equations and using (59) and (60) gives
(63)
Re-inserting this into the first equation of (62) gives
(64)
Combining (64) and (61) gives
(65)
To sum up, we have to proof the validity of (63) and (65). From the contraction process we know that the extremum of
cannot be a minimum, it must be a maximum. The nice feature is that both equations have an identical shape, they differ only in the test point
. Thus we only have to prove
(66)
From the reciprocity principle in [20] we know that
remains constant if we swap current and voltage contacts. This gives
(67)
Combining (67) and (59) gives
(68)
For a Hall-plate with contacts in
-configuration it holds
, see (55). Inserting this into (68) gives (66), which completes the proof. An alternative proof of (66) is given in Appendix C.
3.5. The Minimum of the Van Der Pauw Function
With (57) and (55) the lower envelope curve in the van der Pauw plane is parametrized in a closed formula as follows,
(69)
The lower envelope is identical to general
-configurations with
(insert the first line of (52) into (69)). For the specific
-configuration with
the van der Pauw function has its minimum,
(70)
whereby
depends only on the Riemann modulus
(see (48)). Figure 8 shows this function. It is close to 1 for
and it is very close to 0 for
.
The lower envelope theorem marks the outstanding position of the
-Hall-plates: if the two values
,
are given, we can find a
-arrangement that fits to them, and we can be sure that there is no other
Figure 8. The plot shows the minimum of the van der Pauw function
versus hole radius
. All possible arrangements of the four point contacts were varied until the minimum was obtained for a specific
-configuration of contacts which is specified by
, then the contacts are in the vertices of a square inscribed into the unit circle. The blue curve 1 is the exact Formula (70), the red dashed curve 2 is the small hole approximation (24), and the green dashed curve 3 is the large hole approximation (83). Curve 1 is behind curves 2, 3. Both approximations are very accurate. They have identical values at
.
contact arrangement with a smaller hole which could give the same values
,
. Thus, the star-arrangement determines the minimum required hole size to give the measured deviation of the van der Pauw function from 1. Other contact arrangements like the type 2 configuration are also able to produce the same trans-resistances, but they may need larger holes to do so. Conversely, contact arrangements with
, which I call type 3 configuration, cannot give very large
and very small
at fixed hole radius
, as it is shown in the red curve (3) in Figure 7. The type 3 curves go through the point
for
for all hole sizes
. Type 3 configuration and
-configuration are similar near
.
3.6. Some Properties of
and
The function
may be expressed in various formulae. We can eliminate the complete elliptic-Pi function in (56) with the help of [21]. We can also pull out a logarithm from the incomplete elliptic-Pi integrals with [22],
(71)
Here we use again the short-hand writing
. This gives
(72)
is the Jacobi-zeta function defined in (111). For vanishing hole,
, only the logarithmic term in (72) remains. With [23],
(73)
and with the addition theorems of the Jacobi-zeta function [24] and of the Jacobi-sn function [30] it follows
(74)
Figure 9 shows how the trans-resistances and the van der Pauw function change when the size of the hole grows from zero to full size while the contacts positions remain constant. In the van der Pauw plane of Figure 9(a) curves 1, 3 and 4 show that one of the two coordinates
may increase initially before it decreases (the directions of growing
are indicated by the arrows on the curves). Along the other curves, both coordinates
decrease monotonically for all hole sizes
. In the limit of infinitely thin annular regions, all curves end in the origin
. In all cases, the van der Pauw function
decreases monotonically versus
,
, see Figure 9(b). I have no rigorous proof of this conjecture. The plots in Figure 9(b) also show that the van der Pauw function may change only little for
in curve 3
(a) (b)
Figure 9. Six Hall-plates with different contacts positions for centered circular holes with increasing size
. Curves 1, 2, 4 have identical
,
, yet
is 270˚ for curve 1, 180˚ for curve 2, and 90˚ for curve 4. Curve 3 has
,
,
. Curves
,
are star-configurations with
for curve
and
for curve
. Curve
is identical to
from (70). (a) Representation of the six Hall-plates in the van der Pauw plane
for
.
are indicated for curve 1; (b)
-function of the six Hall-plates versus
.
or for
(see curves 1, 2, 4) or for
(see curve 6). In these cases, the van der Pauw function is not a very sensitive measure to detect holes.
A distinct feature of the curves in Figure 9(a) is the angle
under which they start from the line
. Let us call it the small-hole-angle. This angle is between the tangent on the parametric curve
in
and the vector
. It holds
with
(75)
with
from (16). The small-hole-angle
vanishes for
(76)
which has two meaningful solutions
(77)
In words, if two non-neighbouring peripheral contacts lie on a straight line through the center of the annular Hall-plate, the curves in Figure 9(a) start perpendicularly from the upper envelope. This comprises all star-configurations, but it is more general than star-configurations. From (22) we know that for
-configurations in the asymptotic limit of a small hole the van der Pauw function
has the steepest decline versus hole size, as a quantity to detect small holes,
becomes most sensitive if the contacts are in a
-configuration. Therefore, for small
the red curves
,
are below the blue curves 1, 2, 3, 4 in Figure 9(b). If (77) is fullfilled the curves
start perpendicularly from the upper envelope. Yet, there exist contact configurations, which are not
-configurations, but which still fullfill (77). Their curves
also start perpendicularly from the upper envelope, but for them the slope of the van der Pauw function differs from (22),
(78)
For
the curves
start tangentially from the upper envelope. Then it holds
(79)
This condition is fullfilled only if three or all four contacts approach infinitely closely. From Figure 9(a) I surmize that both trans-resistances are monotonic versus
as long as
. Then it holds
(80)
Inequality (80) is fullfilled for
(81)
For the blue curves 1, 2, 3, 4 in Figure 9(a) we get
, 18.08˚, 89.74˚, 77.33˚, respectively, whereby I define the sign of
equal to the sign of
, this is identical to the sign of
.
3.7. The Asymptotic Limit of a Very Large Hole
In the limit
the annular region of the Hall-plate degenerates to an infinitely thin ring. Then the trans-resistances grow unboundedly. We use
in (57) with (72) to compute the limit of
. With (117) and (118) it follows
(82)
Inserting (82) into (2) gives
(83)
which is plotted as the green curve 3 in Figure 8. Let us define
(84)
Inserting (82) into (84) with the definitions in (2) leads to
(85)
in the asymptotic case
. This tells us at which angle the curves in the van der Pauw plane of Figure 9(a) approach the origin. It holds
(86)
Let us define the angle
between the tangent on the parametric curve
in
and the vector
. I will call it the large-hole-angle. It holds
with
(87)
This gives
(88)
whereby I define the sign of
equal to the sign of
, this is identical to the sign of
. For the curves 1 - 6 in Figure 9(a) we get
, respectively. In general, star-arrangements have
for
,
for
, and
for
. (For a star-configuration
means
and
.) The interesting case
corresponds to
(89)
This holds for a wide class of contact arrangements, including the specific star-configuration with
. Inserting (89) into the first line of (16) and into (75) gives
(90)
A numerical inspection shows that we can find solutions
of (90) for arbitrary
. They define curves in the van der Pauw plane, which start from any point on the upper envelope and go towards the origin
with
. An example is curve 3 in Figure 9(a), which has
and
. Interestingly, in Figure 9(b) curve 3 remains at
for
and only for very large holes
the van der Pauw function drops sharply. The numerical computation of curve 3 in Figure 9(a) and Figure 9(b) is tricky, it needs 5000 digits.
Inserting (21) into (82) into (2) gives the large-hole approximation for star-configurations
(91)
which fails if
is close to 0 or
. Eliminating
in (91) gives the large-hole approximation of the lower envelope,
(92)
which holds well for
and X and Y larger than
.
3.8. Checks for Correctness of the Derived Formulae
The formulae of the Section 2 are consistent with [9] for
(93)
where the quantities on the left hand sides of (93) are from this article and the quantities on the right hand sides are from [9].
In Figure 1 the potential
at the outer perimeter for the azimuthal coordinate
is given by
with
,
A. In the limit of vanishing hole size,
it follows from (48)
and
. From (52) it follows
. Next we use (57) and (56). With
and
and
and
(94)
it follows
(95)
which is identical to (A11b) in [17]. Thus, (57) holds in the limit of singly-connected Hall-plates. Moreover, a series expansion of (57) for small k (small
) leads to (17). (It is lengthy and arduous and therefore I do not report it in detail here.)
For hole sizes of
and 0.9 I computed the potential in
analogous to the preceding paragraph (i.e., via
with
) and compared it with results of a finite element simulation with COMSOL Multiphysics. There I used a plane two-dimensional model in application mode “emdc” (static conductive media). Thickness and conductivity were set to 1 m and 1 S/m, respectively. Due to symmetry, only the upper half of the annular ring was modelled with a fine mesh of 917,504 elements. All boundaries were set insulating, except for the segments on the real axis, which were grounded to 0 V. A current of 1 A/m was extracted from contact
at position
. Figure 10 shows the potential along the perimeter for these four cases and the relative error between analytical and numerical results. The relative errors
Figure 10. Potential
(full symbols) and relative error of
between analytical formula and FEM-simulation (open symbols) for annular Hall-regions with holes of radius
. The current contacts are at
, and the test points are on the outer perimeter at azimuthal positions
.
are in the order of 10−8 which is plausibly due to the finite mesh size around the point-sized current contact. Exemplary numbers for the potential on the unit circle at azimuthal position
are
(96)
As a second numerical check I modelled the Hall-plate in the
-plane of Figure 4(c) for the case
. Equation (36) gives
. Position
of contacts
corresponds to
(see (37)).
gives
.
gives
. From (41) it follows
for the stagnation points in the
-case. The finite element model (FEM) uses a handle between
and
where current flows from Riemann sheet #0 to sheets #1 and # (−1), respectively (see Figure 11). The handle is exactly semi-circular and has anisotropic conductivity
(in radial direction it is zero, and in tangential direction it is ×106 bigger than the conductivity of the Hall-plate with
),
(97)
whereby
and
. The purpose of the handle is to make a short between points
and
for
, but not to short any two points inside
.
• In the
-case only the right half of the symmetric geometry was modelled with a mesh of 1.8 million elements, see Figure 11(a). A current of
was injected into point G and the edges at
were grounded. According to theory [17], the current through the handle should be
, the FEM result deviates by 486 ppm. The reason might be insufficient meshing and insufficient shorting by the handle (in the FEM, points H and J are not exactly at identical potentials, they are at 68 μV and 8 μV). Point D is at −0.26575 V, which corresponds to
, it is 839 ppm larger than the value obtained from (48).
• In the
-case the full geometry was modelled with a mesh of 1.1 million elements, see Figure 11(b). A current of 1 A was injected at point C and extracted at point D. Point D was also grounded to 0 V. According to the theory in [17], the current through the handle should be
, the FEM result deviates by 1120 ppm. Again the handle did not short perfectly: the potential at point B was 2.56193 V, whereas it was 39 μV lower at point H; the potential at point A was 2.48041 V, and it was 2.3 μV lower at point J. The FEM-results for the potentials in point E and G were 2.26454 V and 2.44880 V, respectively. This gives
, which is 462 ppm larger than the result from Formula (48).
(a)(b)
Figure 11. Hall-plate in the
-plane with potentials and current streamlines in two operating conditions for
and
, respectively. The geometry corresponds to
,
,
, and
. (a)
: Zoom into the region near the short-circuit handle (only the right half is modelled). Vertical edges at left side are grounded. 1 A current is injected into point G; (b)
: Zoom into the region near the short-circuit handle. 1 A current is injected into C and extracted at D. D is grounded. Current streamlines inside the handle are not drawn.
The analytical formulae give
and the FEM-simulation gives
. The large vdP-value is due to the small hole, yet for larger holes the point contacts are closer to the handle and the numerical accuracy of the FEM gets even more challenging. This is also a strong indication that in reality the validity of point-sized contacts has to be questioned.
As a by-product of this paper we get a closed form expression of the infinite product
(98)
Equation (98) follows from a comparison of the first lines of (27) and (57). It also relates to the prime function of the annular Hall region, see [15].
4. The Hall-Plate at Applied Magnetic Field
As it was shown in [17] the current density does not change when magnetic field is applied, whereas the potential indeed depends on the magnetic field. Thus we can compute the potential at zero magnetic field,
, derive its current density
and its stream function
, and compute the potential
at arbitrary Hall angle
. Furthermore, in [17] it is derived that the Hall potential is constant along a current streamline (the Hall potential is the difference in electric potential at positive and negative applied magnetic field, it comprises only terms of odd order of the magnetic field). Because of the point-contacts the potential comprises only linear terms of the applied magnetic field, there are no even order terms of the magnetic field (no magneto-resistance terms). Since a current streamline flows from the input contact
to the output contact
along the insulating outer boundary via both voltage contacts
it follows that the voltage between
does not depend on the magnetic field. Thus the trans-resistance
does not depend on the applied magnetic field. The same holds for
, regardless if there is a hole or not. Therefore, Equations (3) and (4) still hold if magnetic field is applied to the Hall-plate. The situation changes if current flows between the hole boundary and the outer boundary or if there is an extended contact on a boundary, but this goes beyond the scope of this paper.
5. Singly-Connected Hall-Plate with Extended Contacts in Star-Configuration
If a Hall-plate has no hole and if its contacts have finite size the van der Pauw function also deviates from 1. Thereby the extra degree of freedom from the hole is replaced by the additional parameters for the finite sizes of the contacts. For the simplified case of a star-arrangement of contacts (also called odd symmetry in [25]) at zero magnetic field closed analytical formulae are available in the literature (combine [25] with (C24), (C25) in [26]),
(99)
whereby the angles
are defined in Figure 12(a). It follows
(100)
(a) (b)
Figure 12. Hall-plates with no hole and with extended symmetric contacts. (a) Star-arrangement of extended contacts (=odd symmetry, [25]); (b) Complementary star-arrangement of extended contacts (=even symmetry, [25]).
for all
. This is readily proven by plotting
versus its two variables
or versus
. The limits are found by the asymptotic limits
and
. For small contacts
, which is consistent with (3). For large contacts
.
Swapping contacts and isolating boundaries gives the complementary star-configuration (also called even symmetry in [25]), see Figure 12(b). There it holds
(101)
with
from (99).
have the following physical meaning: In Figure 12(b) the resistance between contacts
, with
not connected, is equal to
. The resistance between contacts
, with
not connected, is equal to
. Again, for small contacts
tends to 1, and for large contacts it tends to 2.
In summary, large contacts increase the van der Pauw function (at least for the symmetric cases of Figure 12), whereas a hole reduces it.
6. Conclusions and Suggestions
In this paper, I studied the case of an annular Hall-plate with insulating boundaries and four point-contacts on the perimeter. A conformal transformation
maps the ring-domain onto itself, thereby swapping inner and outer boundaries. All resistances remain constant under conformal mapping. Hence, upper and lower envelopes also hold if all contacts are on the boundary of the hole.
In practice, one may equip a sample with several point-sized contacts. Four contacts of a first group should be close together, four contacts of a second group should be spaced equidistantly along the full outer boundary. With the first group, one can measure the local sheet resistance via van der Pauw’s original method (3). Using this value for the sheet resistance one can use the second group of contacts to determine their respective trans-resistances, from which one can derive
with (2). If this value is close to 1 it means that the sample has homogeneous conductivity without hidden holes. If the value obtained for
differs markedly from 1, the conductivity is strongly inhomogeneous and there should be at least one hole inside the sample. With Figure 8 we can assess a lower bound for the size of this hole.
The main results of this article are new proofs of the upper and lower envelopes and closed form expressions for the trans-resistances and the lower envelope in van der Pauw’s measurement. This simple geometry of a circular annulus led to a surprisingly complicated Formula (48) for the trans-resistances. Asymptotic limits were derived for small and large holes and specific properties of symmetric contact arrangements were highlighted. The new concepts of contraction and expansion were introduced as well as the small-hole-angle
and the large-hole-angle
. Yet, several questions are still open for future inquisitions: Is van der Pauw’s function
monotonously falling with the size of the hole for arbitrary contacts positions? How the general behavior of the trans-resistances versus hole size is? What happens, if not all contacts are on the same boundary? Is there a qualitative difference for contacts of finite size? What happens if the hole boundary is conducting such that the hole is short instead of a void? And finally, what happens if the Hall-plate has more than one hole?
Acknowledgements
Sincere thanks to Michael Holliber for his help in Figure 5(a).
Appendix A
The geometric series is
(102)
With (102) it follows
(103)
whereby
or
. We can integrate the very left and right sides of (103), whereby the series on the left side can be integrated term-wise, because integration improves the convergence. This gives
(104)
for
or
. With (104) it holds
(105)
for
or
.
Appendix B
Definition of the incomplete elliptic integral of the first kind
(106)
with
.
is the Mathematica notation of
. This function is strictly monotonic in u and k. The complete elliptic integral of the first kind is
. Its Mathematica notation is
. The complementary elliptic integral of the first kind is denoted by a prime
. For
Equation (106) gives
(107)
For
this gives
(108)
The incomplete elliptic integral of the second kind is
(109)
with
. The complete elliptic integral of the second kind is
. The Mathematica notations of
and
are
and
, respectively. Definition of the incomplete elliptic integral of the third kind
(110)
with
with
. The complete elliptic integral of the third kind is
.
and
are the Mathematica notations of
and
, respectively. The Jacobi-zeta function is
(111)
with the Mathematica notation
. Frequently we are interested in aspect ratios of rectangles from conformal maps of Hall-plates. Then the ratio
shows up. This function is monotonic. Thus, its inverse exists, this is the modular lambda elliptic function
[27].
(112)
The Mathematica notation is
. Several properties of
are explained in [25].
Inversion of (106) gives the Jacobi-sn function and the Jacobi amplitude
(113)
The Mathematica notation is
. Thus it holds
with further Jacobi functions like
and
. Like
also
is odd in u and even in k and strictly monotonic in u and k as long as
and
. It follows
(114)
Yet, for
the Jacobi-sn function has a real-valued period
and a half-period
.
(115)
It holds
(116)
It also holds
(117)
for
and
. From (111), (117), and (113) it follows
(118)
Appendix C
An alternative proof of (66) is by direct computation of the partial derivatives
. Thereby
is a long expression, which is most conveniently computed with an algebraic program like e.g. Mathematica. It comprizes elliptic-Pi functions,
and
, which are multiplied by terms that can be shown to vanish. There is also a term
, which is multiplied by
, which also vanishes, since F is the inverse function of
. The remainder is
(119)
and
are incomplete and complete elliptic integrals of the second kind (see Appendix B). For the other partial derivative one gets
(120)
We use the following identities
(121)
The first equation can be found in [21]. The other two equations follow from the addition formulas for elliptic F- and E-functions [28] [29], and for the Jacobi-sn function [30]. Equations (121) are at least valid in
. Finally we subtract twice (120) from (119) and insert (121). Then all terms cancel out, which completes the proof of (66).
Appendix D
Table A1. Notation list of all the variables of this work.