---
title: Hall Effect
module: Magnetic Field
moduleNumber: 6
lessonNumber: 2
order: 602
summary: >
  Current tells you charge is moving, but not whether the movers are positive or
  negative, nor how many there are. A magnetic field settles both questions. Push
  current through a strip in a transverse field and the carriers pile up on one edge
  until a transverse electric field just balances the magnetic deflection; the sign
  of the resulting Hall voltage names the carrier's charge and its size counts the
  carriers per volume. We derive the balance $q\vec E+q\vec v_d\times\vec B=0$, read
  off $V_H=IB/(nqt)$, and see why field-and-current reversal is what separates the
  real Hall signal from the offsets that mimic it.
topics: [Magnetic Field]
draft: false
sources:
  - book: Tipler & Mosca
    ref: "Ch. 26 — The Magnetic Field; Hall effect"
---

## Transverse force balance in a conducting strip

The Hall effect appears when a current-carrying conductor or semiconductor is placed
in a magnetic field perpendicular to the current. Consider a rectangular sample with
conventional current along x, magnetic field along z, and voltage measured across y.
Mobile carriers have a drift velocity $\vec v_d$, so each carrier experiences
the transverse magnetic force $q\vec v_d\mathbin{\times}\vec B$. Charges
therefore accumulate on one side of the sample. Their separation establishes an
electric field across the width, called the Hall field. Accumulation stops when the
electric force balances the magnetic force on an additional carrier:

$$
q\left(\vec E_H+\vec v_d\mathbin{\times}\vec B\right)=0.
$$

The charge $q$ cancels from the balance equation but its sign remains essential for
the direction of charge accumulation. In magnitude, with drift velocity perpendicular
to the field, $E_H=v_dB$. The resulting voltage is measured between two side
contacts; it is not a voltage drop along the direction of current. A Hall voltage can
be small even while the longitudinal resistive voltage is large, so contact placement
and lead routing matter.

$$
% caption: Hall-bar geometry. Conventional current runs along the strip and the magnetic field points out of the page; carriers deflect across the width and the resulting Hall voltage is read between the two side contacts.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[black,fill=black!8] (0.6,0.7) rectangle (4.8,2.0);
\draw[->,very thick] (1.0,1.35)--(2.3,1.35) node[above] {current};
\foreach \x in {1.1,2.1,3.1,4.1} {\draw[black] (\x,2.55) circle (0.10); \fill[black] (\x,2.55) circle (0.028);}
\node[right] at (4.35,2.55) {$B$ out};
\filldraw[black] (3.05,2.0) circle (1.4pt);
\filldraw[black] (3.05,0.7) circle (1.4pt);
\draw[<->,acc,thick] (3.05,0.74)--(3.05,1.96) node[midway,right] {Hall voltage};
\draw[<->,black] (0.6,0.42)--(4.8,0.42) node[midway,below] {length};
\draw[<->,black] (5.05,0.7)--(5.05,2.0) node[midway,right] {width};
\end{tikzpicture}
$$

For one dominant carrier type, the current density is $J=nqv_d$, where $n$ is
the signed-carrier density conventionally taken positive and $q$ carries the
carrier sign. If the sample width is $w$ and its thickness along the magnetic field
is $t$, then $I=Jwt$. Combining this relation with $V_H=E_Hw$ gives

$$
V_H=\frac{IB}{nqt}.
$$

This form makes the geometry clear: width cancels, whereas a thinner specimen gives a
larger Hall voltage for the same current and field. The Hall coefficient is
$R_H=E_H/(JB)=1/(nq)$, so its sign identifies the sign of the dominant mobile
carrier after the current, field, and voltage-lead conventions have been stated.

> **Worked example (Hall voltage in a copper strip).** A copper strip of thickness
> $t=0.10\ \mathrm{mm}$ carries $I=10\ \mathrm{A}$ in a field $B=1.0\ \mathrm{T}$
> normal to its face. Copper's free-electron density is
> $n=8.5\times10^{28}\ \mathrm{m^{-3}}$, so
>
> $$
> V_H=\frac{IB}{nqt}
> =\frac{(10)(1.0)}{(8.5\times10^{28})(1.60\times10^{-19})(0.10\times10^{-3})}
> =7.3\ \mu\mathrm{V}.
> $$
>
> Only a few microvolts, despite a full tesla and 10 amps, because a metal packs so
> many carriers that each drifts slowly. This is why practical Hall probes use
> semiconductors, where $n$ is smaller by many orders of magnitude and $V_H$ grows in
> proportion.

## Carrier sign, density, and measurement limits

The sign interpretation should be made from the force direction, not from a remembered
label. Fix a right-handed coordinate system, state which voltage terminal is listed
first, and use $q\vec v_d\mathbin{\times}\vec B$ for the actual carrier.
Electrons drift opposite conventional current; positive holes drift with it. Reversing
the field reverses the Hall voltage, and reversing the current also reverses it.
Reversing both leaves the Hall voltage unchanged. These symmetries provide immediate
checks on lead polarity and field orientation.

$$
% caption: Transverse force balance in a Hall bar. The magnetic force drives one carrier sign to one edge; the charge that piles up sets a Hall field $E$ whose force opposes further deflection, and its measured polarity names the dominant carrier.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[black,fill=black!8] (0.5,0.5) rectangle (5.0,2.5);
\node[above] at (1.4,2.5) {charge builds here};
\node[below] at (3.9,0.5) {opposite charge};
\filldraw[draw=black,fill=black!8] (2.2,1.5) circle (0.09);
\draw[->,thick] (2.2,1.5)--(3.3,1.5) node[midway,below] {motion};
\draw[->,black,thick] (2.2,1.5)--(2.2,2.15) node[left] {force};
\draw[->,acc,thick] (4.1,2.1)--(4.1,0.9) node[midway,right] {Hall $E$};
\foreach \x in {1.4,2.4,3.4,4.4} {\draw[black] (\x,2.95) circle (0.09); \fill[black] (\x,2.95) circle (0.026);}
\node[right] at (4.65,2.95) {$B$ out};
\end{tikzpicture}
$$

Once the sign is known, the magnitude of the Hall coefficient gives an effective
carrier density, $n=IB/(|q|t|V_H|)$, in the simple single-carrier model. Metals
often have a low Hall voltage because their carrier density is high. Semiconductors
can give a much larger signal and may switch Hall sign when the dominant carrier type
changes. The inferred density is an effective value: materials with electrons and
holes conducting at the same time, anisotropic bands, or field-dependent scattering
need a more complete transport model. A measured Hall coefficient should therefore
be reported with temperature, field range, and material state.

> **Worked example (Carrier density from a Hall reading).** A conducting film of
> thickness $t=0.20\ \mathrm{mm}$ carries $I=5.0\ \mathrm{A}$ in $B=0.50\ \mathrm{T}$
> and shows a corrected Hall voltage $|V_H|=6.0\ \mu\mathrm{V}$. Inverting the
> single-carrier relation,
>
> $$
> n=\frac{IB}{|q|\,t\,|V_H|}
> =\frac{(5.0)(0.50)}{(1.60\times10^{-19})(0.20\times10^{-3})(6.0\times10^{-6})}
> =1.3\times10^{28}\ \mathrm{m^{-3}}.
> $$
>
> The sign of $V_H$, not its magnitude, names the carrier; the magnitude counts them.
> This value is one carrier per few atoms, consistent with a good metal. A density
> orders of magnitude smaller would instead flag a semiconductor or a two-carrier
> material whose effective $n$ should not be read literally.

Practical Hall measurements separate the transverse signal from larger unwanted
voltages. Misaligned side contacts pick up part of the longitudinal resistive drop;
thermoelectric offsets and amplifier offsets can remain even at zero magnetic field.
Measure at both positive and negative field and take the antisymmetric component,
$[V(+B)-V(-B)]/2$, to reject many field-independent offsets. Sweep current at
fixed field as a second test: the Hall voltage should be linear in current within the
ohmic regime. A sweep of field checks linearity in $B$ and detects saturation or
magnetoresistive contamination. Keep current low enough to avoid heating, verify the
actual sample thickness, and use a high-input-resistance differential measurement so
the voltage contacts do not draw transverse current. Those controls turn a signed
side voltage into a defensible carrier measurement.

## Semiconductor Hall effect, mobility, and calibration

In a semiconductor, Hall data connect carrier sign with transport strength. The
drift mobility $\mu$ is the magnitude of drift velocity per applied longitudinal
electric field, $\mu=|v_d|/|E_x|$. For one carrier population, conductivity is
$\sigma=n|q|\mu$, while the Hall coefficient is $R_H=1/(nq)$. Combining them
gives the ideal Hall mobility $|R_H|\sigma$. A negative Hall coefficient identifies
electrons as the dominant carriers; a positive coefficient identifies holes. The
word dominant matters: a semiconductor can contain both carrier types, and the Hall
coefficient weights their densities and mobilities rather than simply counting every
mobile charge. Scattering can also introduce a Hall factor, so Hall mobility is an
effective transport quantity unless the relevant scattering model has been established
for the material and temperature range.

The geometry of a thin sheet makes Hall measurements particularly sensitive. With
current flowing along a sheet of thickness $t$, the Hall voltage remains
$V_H=R_HIB/t$; it does not depend on the width between the Hall contacts in the
ideal uniform-current limit. Longitudinal resistance, by contrast, is often reported
as a sheet resistance $R_s=\rho/t$, measured in ohms per square. Combining a Hall
coefficient with sheet resistance and thickness separates carrier density from
mobility. This separation fails if current crowds near a contact, the film thickness
is nonuniform, or side contacts draw enough current to disturb the transverse field.

Temperature changes both factors in $\sigma=n|q|\mu$. In doped material, carrier
density can rise as thermal excitation becomes important, while mobility commonly
falls when lattice scattering becomes stronger. The Hall coefficient and resistance
can therefore move in opposite directions with temperature. Bias matters as well:
excessive current heats the sheet, changes carrier density or mobility, and can make
an apparently linear Hall-voltage sweep curve. Establish the ohmic current range by
checking that longitudinal voltage and Hall voltage are each proportional to current
at fixed field. Report sample temperature or use controlled-temperature measurements
when values are compared across devices.

Calibration should use all reversal symmetries. Let $V_{sB}$ be the measured side
voltage for current sign $s=+1$ or $-1$ and field sign $B=+1$ or $-1$. The
combination

$$
V_H=\frac{V_{++}-V_{+-}-V_{-+}+V_{--}}{4}
$$

keeps the term odd in both current and field while rejecting many offsets and
longitudinal pickup terms. Measure the actual field with a calibrated probe at the
sample position; the magnet-supply setting is insufficient by itself. A reference
specimen with known Hall coefficient checks the product of current, field, thickness, and voltage
calibration before an unknown sample is interpreted.

$$
% caption: Reversal control for the Hall voltage. The true Hall signal changes sign with either drive current or field, so the two field polarities give oppositely sloped lines; offsets that survive reversal are rejected by the odd-in-both combination.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0.45,0.4)--(5.4,0.4) node[right] {drive current};
\draw[->,black] (0.45,0.4)--(0.45,2.95) node[above] {side voltage};
\draw[black,dashed] (0.55,1.65)--(4.9,1.65);
\draw[acc,thick] (0.7,0.7)--(4.85,2.6) node[right] {$B$ positive};
\draw[black,thick] (0.7,2.6)--(4.85,0.7) node[right] {$B$ negative};
\filldraw[draw=acc,fill=acc!15] (2.78,1.65) circle (0.05);
\end{tikzpicture}
$$

Uncertainty in carrier density combines uncertainty in Hall voltage, current, field,
and thickness; thin films improve signal but make thickness error relatively more
important. Finite contact size, nonuniform field, magnetoresistance, and mixed
conduction are model limits rather than random scatter. A credible result gives the
reversal procedure, field calibration, thickness method, current range, temperature,
fit residuals, and uncertainty budget alongside the reported carrier sign, density,
and mobility. That record makes clear whether the measurement supports a
single-carrier interpretation or only an effective transport parameter.

## Hall sensors, magnetic-field mapping, and failure modes

A Hall sensor converts the magnetic-field component normal to its active layer into
a voltage. At fixed bias current, its current-related sensitivity is
$S_I=V_H/(IB)$, commonly stated in volts per ampere-tesla. Devices operated from a
fixed bias voltage are often specified by a voltage-related sensitivity instead.
Neither number is universal: temperature changes carrier density and mobility, while
package stress and bias current can change the zero-field output. The calibrated field
estimate is obtained from a calibrated sensitivity and an explicitly measured offset,
not from the raw sensor voltage alone.

The output represents one field component. If the active-layer normal
$\hat n$ makes angle $\theta$ with the local field, the ideal output is
proportional to $\vec B\mathbin{\cdot}\hat n=B\cos\theta$. A small
tilt produces a cosine scale error and can couple an unwanted transverse field
component into the reading. Mark the probe normal physically, approach the sample
with a repeatable orientation, and reverse or rotate the probe as a check. Vector
mapping requires three calibrated, mutually oriented component measurements.

$$
% caption: Orientation of a Hall probe. Only the field component along the layer normal produces the ideal output, so a tilt of angle $\theta$ scales the reading by $\cos\theta$ and couples in a cross-axis component.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[black,fill=black!8] (1.2,0.75)--(3.95,1.2)--(3.55,2.1)--(0.8,1.65)--cycle;
\draw[->,acc,thick] (2.35,1.4)--(2.35,2.8) node[above] {layer normal};
\draw[->,black,thick] (2.35,1.4)--(3.85,2.65) node[right] {applied $B$};
\draw[black] (2.35,2.05) arc[start angle=90,end angle=40,radius=0.65];
\node at (2.74,2.3) {tilt};
\end{tikzpicture}
$$

Offset is the first practical failure mode. Contact asymmetry, mechanical stress,
thermoelectric voltages, amplifier input offset, and finite longitudinal pickup can
all produce output at zero field. Measure zero-field output before and after a map,
and use field or current reversal to extract the odd magnetic component. Temperature
drift requires either a temperature-compensated sensor, repeated zero checks, or a
calibration model over the operating range. Rapidly changing offset cannot be cured
by simply subtracting one value recorded at the start of a long experiment.

Saturation also needs a specific diagnosis. The Hall material may cease to be linear
at high field or high bias, but more commonly the sensor amplifier or analog-to-
digital converter reaches its output range first. In current-biased devices, excessive
bias can heat the active area and alter sensitivity before any electronic rail is
reached. Sweep a known field through positive and negative values and inspect both
the slope and residuals. Flattening, unequal positive and negative slopes, or clipped
waveforms identifies a range problem; it is not evidence that the field itself has
stopped changing.

Mapping introduces spatial limits. A Hall element averages the normal field over its
finite active area and over its separation from the sample. Near a narrow conductor,
magnet edge, or small defect, this averaging can hide peak fields and blur gradients.
Record the active-area dimensions, stand-off distance, scan step, and probe path.
Use a step smaller than the spatial feature to be resolved, but do not mistake a
dense grid of overlapping averages for higher intrinsic resolution. Repeat scans in
opposite directions to expose position backlash and thermal drift.

Calibration against a known field closes the measurement chain. A well-characterized
electromagnet or calibrated coil produces several positive and negative reference
fields at the actual probe position. Fit the odd output component to field, retain
the intercept as an offset diagnostic, and repeat after changing temperature or bias.
Include reference-field uncertainty, sensitivity fit uncertainty, orientation error,
offset drift, spatial averaging, and digitizer resolution in the final map. A map
with this information can distinguish a real local magnetic feature from a probe
artifact or an uncorrected change in sensor response.

## Multicarrier transport, anomalous Hall response, and model scope

The relation $R_H=1/(nq)$ is a single-carrier, low-field model. It works when one
mobile carrier population dominates the conductivity, the Hall response is linear in
field, and the sample can be treated as an isotropic ohmic sheet. Semiconductors near
compensation, semimetals, and many metals do not always meet these conditions. Both
electrons and holes can carry longitudinal current while their Hall deflections have
opposite signs. A small measured Hall coefficient can then mean high carrier density,
nearly cancelling electron and hole contributions, or both; it is not by itself a
unique density measurement.

In the simple two-carrier low-field limit, the conductivity and Hall coefficient are

$$
\sigma=e(n\mu_e+p\mu_h),
\qquad
R_H=\frac{p\mu_h^2-n\mu_e^2}{e(n\mu_e+p\mu_h)^2}.
$$

Here $n,p$ are electron and hole densities and $\mu_e,\mu_h$ their mobilities.
The squared mobility weighting explains why a relatively sparse, high-mobility carrier
population can control the Hall sign while another population dominates conductivity.
At stronger fields, the Hall curve can become nonlinear because the carrier deflection
is no longer in the low-field limit. Fitting such data requires longitudinal
magnetoresistance and Hall data together; fitting a Hall slope alone leaves density
and mobility parameters strongly correlated.

The low-field requirement is often expressed as $\mu B\ll1$ for every carrier
population included in the approximation. It is a dimensionless transport condition.
A very high-mobility semiconductor can leave the low-field regime at a field where an
ordinary metal remains effectively linear.

$$
% caption: Ordinary Hall response. One dominant carrier gives a straight odd-in-field line, while several carrier populations bend the trace; sweeping both field signs shows whether a single Hall coefficient applies.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0.4,1.55)--(5.4,1.55) node[right] {$B$};
\draw[->,black] (2.9,0.35)--(2.9,2.95) node[above] {Hall signal};
\draw[acc,thick] (0.85,0.65)--(4.95,2.5);
\draw[black,thick] plot[smooth] coordinates {(0.85,1.02)(1.5,1.2)(2.2,1.4)(2.9,1.55)(3.6,1.82)(4.3,2.32)(4.95,2.9)};
\node[right] at (4.55,1.95) {single carrier};
\node[black,right] at (3.35,2.7) {curved};
\filldraw[draw=black,fill=black!8] (2.9,1.55) circle (0.05);
\end{tikzpicture}
$$

Ferromagnetic conductors add a distinct effect. Their transverse resistivity is often
written $\rho_{xy}=R_0B+R_s\mu_0M$, where the first term is the ordinary Hall
response and the second is an anomalous Hall contribution associated with
magnetization and spin-orbit-coupled transport. The anomalous term need not be
proportional to applied field because magnetization can saturate or reverse
hysteretically. Treating the full transverse signal as $R_0B$ can therefore assign
an implausible carrier density or even the wrong sign to the ordinary Hall component.

$$
% caption: Anomalous Hall response in a ferromagnet. The transverse signal is hysteretic: a magnetization-linked term adds to the ordinary field-linear Hall term, so the switching loop and the high-field slope must be separated before reading carrier parameters.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0.4,1.55)--(5.4,1.55) node[right] {$B$};
\draw[->,black] (2.9,0.35)--(2.9,2.95) node[above] {transverse signal};
\draw[acc,thick] plot[smooth] coordinates {(0.7,0.7)(1.4,0.9)(2.1,1.1)(2.45,1.85)(3.2,2.05)(4.0,2.28)(5.05,2.5)};
\draw[acc,thick] plot[smooth] coordinates {(5.05,2.4)(4.0,2.05)(3.2,1.8)(2.45,1.1)(2.1,0.95)(1.4,0.72)(0.7,0.5)};
\draw[->,acc,thick] (2.2,1.05)--(2.5,1.72);
\node[right] at (3.7,2.62) {high slope};
\node[left] at (2.12,1.0) {switching};
\end{tikzpicture}
$$

Diagnostic measurements should establish model scope before extracting microscopic
numbers. Sweep positive and negative field far enough to observe the high-field
slope, measure longitudinal resistance simultaneously, and repeat at several
temperatures or gate biases when available. A linear, reversible Hall trace with
stable sheet resistance supports the single-carrier approximation over that range.
Curvature, sign reversals with temperature, strong magnetoresistance, or hysteresis
requires a multicarrier or magnetic transport model. Field reversal still removes
many offsets, but it cannot remove a genuine anomalous Hall signal because that term
is itself transverse and can be odd under field reversal through the magnetization.
Record magnetic history, sweep direction, and saturation state whenever a
ferromagnetic sample is measured.

Uncertainty also changes with the model. In a single-carrier analysis, thickness,
field, current, and voltage dominate the density estimate. In multicarrier fitting,
parameter covariance and the field range become central; apparently precise values
can be nonunique. Report the chosen equations, fit residuals, parameter correlations,
and the observations that justify excluding competing carriers or anomalous response.
That is the boundary between a Hall coefficient serving as an effective material
descriptor and a Hall measurement overinterpreted as a literal carrier count.

## Worked Hall measurement and inverse-problem checks

> **Worked example (Hall coefficient of an n-type film).** A rectangular n-type film
> has thickness $t=0.500\ \mathrm{mm}$, width $w=2.00\ \mathrm{mm}$, and longitudinal
> contact spacing $L=10.0\ \mathrm{mm}$. A current $I=12.0\ \mathrm{mA}$ flows along
> $+x$ while a field $B=+0.800\ \mathrm{T}$ points along $+z$. Take the side voltage
> as the lower-$y$ contact minus the upper-$y$ contact.
>
> Misaligned side contacts add a field-even offset, so the instrument reads
> $-1.20\ \mathrm{mV}$ at $+B$ and $+1.80\ \mathrm{mV}$ at $-B$. The odd-in-field part
> is the Hall voltage,
>
> $$
> V_H=\frac{V(+B)-V(-B)}{2}=\frac{-1.20-1.80}{2}\ \mathrm{mV}=-1.50\ \mathrm{mV},
> $$
>
> and the $+0.300\ \mathrm{mV}$ average is the offset. The negative sign agrees with
> negative carriers: electron drift is opposite the current, and the deflection makes
> the lower-$y$ contact negative. The Hall coefficient follows,
>
> $$
> R_H=\frac{V_Ht}{IB}
> =\frac{(-1.50\times10^{-3})(0.500\times10^{-3})}{(12.0\times10^{-3})(0.800)}
> =-7.81\times10^{-5}\ \mathrm{m^3/C},
> $$
>
> giving carrier density $n=1/(|R_H|e)=8.00\times10^{22}\ \mathrm{m^{-3}}$. Thickness
> enters directly; width does not. A longitudinal drop $V_x=0.240\ \mathrm{V}$ gives
> resistance $20.0\ \Omega$, resistivity $\rho=2.00\times10^{-3}\ \mathrm{\Omega\,m}$,
> conductivity $500\ \mathrm{S/m}$, and Hall mobility
> $|R_H|\sigma=3.91\times10^{-2}\ \mathrm{m^2/(V\,s)}$.
>
> The force balance checks independently: $J=I/(wt)=1.20\times10^4\ \mathrm{A/m^2}$
> gives drift speed $v_d=J/(ne)\approx0.94\ \mathrm{m/s}$, so
> $v_dB\approx0.75\ \mathrm{V/m}$ matches the measured Hall field
> $|V_H|/w=0.75\ \mathrm{V/m}$. Disagreement would point to a lead-order error, a wrong
> thickness, or a model outside its stated scope.

An uncertainty test uses relative standard uncertainties of three percent in the
corrected Hall voltage, two percent in thickness, one percent in field, and one-half
percent in current. Independent quadrature combination gives about 3.8 percent for
both $|R_H|$ and the inferred single-carrier density. This number is incomplete if
the field-even correction is unstable, contact alignment changes between reversals,
or the Hall curve is nonlinear. Inspect residuals of a multi-field linear fit and
compare repeated reversal cycles. A result that remains consistent under current and
field reversal, satisfies the longitudinal force-balance check, and has a documented
geometry correction is an inverse problem with tested assumptions rather than a
single substituted voltage.

## Design tradeoffs and a Hall-probe test plan

Hall-probe design begins with the field range, spatial scale, bandwidth, and allowed
uncertainty rather than with sensitivity alone. A semiconductor Hall element usually
has a much larger Hall coefficient than a metal film, giving more output voltage per
ampere-tesla at modest bias. The tradeoff is greater temperature dependence, stronger
resistance variation, and often more sensitivity to stress or bias heating. Metal
elements have lower raw sensitivity but can be stable over wide ranges. An integrated
sensor adds amplifier noise, offset-cancellation circuitry, and output-range limits;
its data sheet sensitivity is a system specification, not a material constant.

Geometry sets both signal and disturbance. For a fixed current and field,
$V_H$ increases as active-layer thickness decreases. A thin layer also has higher
longitudinal resistance, so the same current produces more Joule heating. Increasing
bias current raises Hall signal linearly but heating quadratically through $I^2R$.
The operating current must remain below the level at which sensitivity or
offset changes during a measurement. Check this with an up-and-down current sweep
and compare zero-field output before and after the sweep. Width does not enter the
ideal Hall voltage directly, but it affects current density, contact geometry, and
the ease of avoiding local hot spots.

Noise and bandwidth must be chosen together. Johnson noise of the element resistance,
amplifier voltage and current noise, digitizer resolution, and external magnetic
interference set a minimum resolvable field. Narrowing the electronic bandwidth
reduces uncorrelated noise, but it also prevents the probe from following a changing
field. A field map of a static magnet can use long averaging and field reversal;
measurements of pulsed currents or moving machinery require a bandwidth high enough
to preserve the relevant waveform. State the effective bandwidth and integration
time with every sensitivity claim. Sampling faster than the amplifier bandwidth adds
data points but does not recover information that filtering has removed.

Spatial response is another bandwidth-like limit. The active area averages the normal
field over its footprint and stand-off distance. A smaller element resolves steeper
gradients but gives less signal and may demand higher bias or lower-noise electronics.
Near a conductor edge, compare maps at two stand-off distances and scan directions.
If the apparent peak changes strongly with either, report a spatially averaged field
instead of presenting it as a point value. Choose scan spacing from the intended
feature size and active-area dimensions, not from the smallest motion increment of
the translation stage.

A traceable calibration plan uses a known, uniform reference field over the expected
range, including positive and negative values and a zero-field reading. At the actual
probe orientation, fit the odd field response, retain the intercept as an offset
check, and record reference-field uncertainty, probe temperature, bias current,
bandwidth, and digitizer settings. Validate the calibration at an intermediate field
not used in the fit. Before and after the unknown-field scan, repeat at one reference
point to test drift. The final uncertainty budget combines reference-field scale,
sensitivity fit, offset drift, orientation, spatial averaging, noise, and position.
This chain makes a Hall map reproducible and shows which improvement—cooler operation,
smaller active area, lower noise, or better field reference—would materially reduce
the stated uncertainty.

Repeat one reference-field measurement after removing and remounting the probe. A
change in the recovered field can reveal an orientation or stand-off error that is
invisible in an electrical zero check. Record the mounting datum used for every map.

## Dynamic Hall measurements and frequency response

Low-level Hall signals are often measured with a modulated bias and a lock-in
amplifier rather than as a static dc voltage. If the sensor is biased with
$I(t)=I_0\cos\omega_m t$ in a steady field, its ideal Hall output is proportional
to the same modulation frequency. Multiplying the measured signal by a phase-matched
reference and low-pass filtering retains the component coherent with the bias while
rejecting much of the broadband noise and dc offset. The resulting in-phase lock-in
output is proportional to field; the quadrature output is a diagnostic for phase
error, capacitively coupled pickup, or a response that is no longer instantaneous.

Modulation does not remove every unwanted voltage. A resistive longitudinal pickup
also follows the bias current, so current reversal alone cannot distinguish it from
the Hall signal. Magnetic-field reversal, contact symmetry, or a second modulation
of the field is needed to isolate the term with the required current-field symmetry.
Bias frequency should avoid mains harmonics, mechanical vibration frequencies, and
the region where cable capacitance or sensor electronics attenuate the current drive.
The lock-in time constant sets measurement bandwidth: a longer time constant lowers
noise but averages away genuine field variation.

Sensor bandwidth is set by the active element, bias network, amplifier, cable, and
the chosen detection filter. A Hall element can respond rapidly at the carrier level,
yet a high-resistance bias source and cable capacitance can form a low-pass network
that rounds the measured waveform. Determine amplitude and phase response with a
known sinusoidal field or a calibrated drive coil across the intended frequency
range. Reference the measured phase to a field monitor or coil current after its own
delay has been characterized. A fixed electronic delay can be calibrated out; a
frequency-dependent phase curve is part of the transfer function and must be retained
when reconstructing transient field timing.

Quote the modulation frequency, reference phase, equivalent noise bandwidth, and
settling time so that the reported dynamic sensitivity can be independently repeated.

For pulsed or rapidly changing fields, record the full waveform with sample rate and
analog bandwidth above the highest frequency content needed for the rise time and
ringing. Do not infer a sharp field edge from a trace limited by the probe response.
Compare the Hall waveform with an independent current monitor or pickup coil, taking
their bandwidth and orientation into account. Saturation of the sensor amplifier,
current source, or digitizer can mimic a flat-topped field pulse, so range checks are
part of transient analysis rather than a separate instrument setting.

Lead construction matters because a changing field induces voltage in any enclosed
loop area, while electric-field coupling drives current through cable capacitance.
Use short, closely paired differential leads, a shield connected according to the
instrument grounding scheme, and a defined return path. A shield does not eliminate
magnetic pickup in a large loop, and grounding both ends indiscriminately can create
circulating-current pickup. Verify shielding by repeating a zero-field dynamic test
with the same bias and nearby drive current. Any residual coherent signal sets a
floor on the transient field measurement and belongs in the uncertainty budget.

## Temperature dependence, calibration, and error budgets

The Hall coefficient is a transport quantity, so its temperature trend carries
physical information but can also undermine a calibration. In an ideal single-carrier
material, $R_H=1/(nq)$. A doped semiconductor may have an approximately constant
carrier density over part of its extrinsic temperature range, yet its Hall response
can still shift because mobility and the Hall scattering factor change. At lower
temperature, carrier freeze-out can reduce the mobile density; at higher temperature,
intrinsic electrons and holes can both contribute, changing magnitude, curvature, or
even sign. Metals usually show smaller fractional Hall changes over ordinary
laboratory ranges, but contact stress and thermal gradients can still produce larger
apparent drift than the intrinsic coefficient change.

Temperature control begins with the bias. A Hall element dissipates longitudinal
power while it is driven, and self-heating changes both resistance and sensitivity.
Record sensor temperature near the active region as well as ambient air temperature.
At each calibration temperature, take a zero-field reading before the field sweep,
repeat it after the sweep, and allow enough time for the element to return to thermal
equilibrium. A changing zero reading is evidence of offset drift; it cannot be
treated as random voltage noise simply because the field was held fixed.

A traceable sensitivity calibration uses a reference field that is known at the Hall
element's actual position and orientation. A Helmholtz pair with coil radius $a$
and $N$ turns per coil gives the central field

$$
B_{\rm ref}=\mu_0\left(\frac45\right)^{3/2}\frac{NI}{a},
$$

provided coil geometry, current, and positioning are controlled. A calibrated field
probe or magnet standard can serve the same role. Sweep reference field through both
signs, fit the odd component of sensor output to field, and retain the fitted slope
as sensitivity. The intercept diagnoses residual offset rather than being silently
absorbed into the scale factor. Check at one intermediate reference value withheld
from the fit; this catches a gain error or a nonlinear response that a two-point
calibration would miss.

Offset removal uses symmetry as well as thermal control. With current and field both
reversed, the desired Hall term is odd in both variables. The four-state combination
removes many fixed amplifier, thermoelectric, and contact-misalignment terms. Repeat
the four states in an interleaved order rather than collecting every positive-field
point first: slow temperature drift can otherwise masquerade as a field-odd signal.
If field reversal changes magnet temperature, mechanical position, or magnetic history,
those effects belong in the calibration model rather than in an offset subtraction.

For the single-carrier density $n=IB/(|q|t|V_H|)$, uncorrelated relative standard
uncertainties combine approximately as

$$
\left(\frac{u_n}{n}\right)^2=
\left(\frac{u_I}{I}\right)^2+
\left(\frac{u_B}{B}\right)^2+
\left(\frac{u_t}{t}\right)^2+
\left(\frac{u_{V_H}}{V_H}\right)^2.
$$

This expression propagates uncertainty within the stated model. Thickness may
dominate in a thin film; voltage noise and offset drift may dominate at low field;
field-scale and orientation error can dominate in a mapped nonuniform field. Add
covariance terms when the same current source or calibration fit contributes to more
than one quantity. Report each contribution before combining them so improvement
work targets the limiting measurement.

Physical limits remain after a careful statistical budget. A finite Hall area averages
field gradients; excessive current heats the element; high field can reveal
multicarrier curvature or anomalous Hall response; and amplifier range can clip the
output. A calibration is valid only over the stated temperature, bias, bandwidth,
field, orientation, and spatial range. Outside those conditions, uncertainty bars on
the old sensitivity do not restore accuracy: the sensor model itself must be tested
again.

**Semiconductor Hall sensors and multidimensional transport limits.**

A semiconductor Hall element is not characterized by carrier concentration alone.
For one carrier population, longitudinal conductivity is $\sigma=n|q|\mu$, whereas
the Hall coefficient is more accurately written $R_H=r_H/(nq)$, with a Hall factor
$r_H$ that depends on the scattering mechanism and band structure. In a simple
n-type material, donors supply electrons and the Hall coefficient is negative; in a
p-type material, acceptors supply holes and the coefficient is positive. The common
identification $n=1/(|qR_H|)$ therefore returns an effective density unless
$r_H$ is known or close to unity. Hall mobility $\mu_H=|R_H|\sigma$ likewise need
not equal the drift mobility used in a conductivity model.

Temperature exposes why these distinctions matter. In the freeze-out regime, dopants
are not all ionized and mobile carrier concentration changes rapidly with temperature.
Across an extrinsic plateau, carrier density can be comparatively stable while
mobility changes because ionized-impurity and phonon scattering redistribute the
relaxation time. At higher temperature, thermally generated electrons and holes add
ambipolar transport. A Hall sensor calibrated at one temperature can then have a
different sensitivity at another even if the bias current and applied field remain
unchanged. Current-induced heating shifts the active element along the same transport
curves; a dc calibration that ignores self-heating can assign an apparent field drift
to the magnet when the sensor temperature was the changing variable.

The operating bias range is bounded from both sides. At low current, the Hall signal is
small compared with Johnson noise, amplifier noise, offset drift, and digitizer
resolution. At high current, $I^2R$ heating changes mobility and contact resistance,
and high electric field can make transport non-ohmic. Pulsed bias can reduce average
heating, but its duty cycle, rise time, and thermal recovery must be stated; a pulse
still deposits energy during its on-time. An operating-point test measures Hall
voltage, longitudinal voltage, and zero-field offset over an up-and-down bias sweep.
The acceptable region has proportional Hall voltage, stable normalized sensitivity,
and no hysteresis between increasing and decreasing current.

$$
% caption: Semiconductor Hall sensitivity varies with temperature. Across the extrinsic plateau the carrier density is roughly fixed while mobility falls; in the intrinsic regime both electrons and holes contribute, so a single room-temperature calibration need not hold elsewhere.
\begin{tikzpicture}[>=stealth,font=\footnotesize]
\definecolor{acc}{HTML}{4A6FA5}
\draw[->,black] (0.45,0.4)--(5.4,0.4) node[right] {temperature};
\draw[->,black] (0.45,0.4)--(0.45,2.95) node[above] {transport scale};
\draw[acc,thick] plot[smooth] coordinates {(0.65,0.62)(1.35,1.18)(2.15,1.7)(3.05,1.82)(3.9,2.0)(4.95,2.7)};
\draw[black,thick] plot[smooth] coordinates {(0.65,2.55)(1.35,2.35)(2.15,2.1)(3.05,1.78)(3.9,1.4)(4.95,0.85)};
\draw[dashed,black] (2.2,0.4)--(2.2,2.4);
\draw[dashed,black] (3.6,0.4)--(3.6,2.4);
\node[acc,right] at (3.95,2.2) {density};
\node[black,right] at (3.95,1.15) {mobility};
\node[below] at (1.3,0.4) {freeze-out};
\node[below] at (2.9,0.4) {extrinsic};
\node[below] at (4.5,0.4) {intrinsic};
\end{tikzpicture}
$$

The single-carrier Hall model has a precise domain. When electrons and holes both
conduct, their longitudinal currents add while their ordinary Hall contributions
oppose. In the classical two-carrier model, the field-dependent conductivity tensor
contains

$$
\sigma_{xx}=e\left[\frac{n\mu_e}{1+(\mu_eB)^2}+
\frac{p\mu_h}{1+(\mu_hB)^2}\right],
$$

$$
\sigma_{xy}=eB\left[\frac{p\mu_h^2}{1+(\mu_hB)^2}-
\frac{n\mu_e^2}{1+(\mu_eB)^2}\right].
$$

The measured resistivity tensor is the inverse of this conductivity tensor, not a
single Hall slope substituted into $1/(nq)$. A minority carrier population with
high mobility can dominate low-field Hall response, while a different population
dominates longitudinal conductivity. Curvature of transverse voltage with field,
large magnetoresistance, or a temperature-driven Hall-sign reversal is evidence that
one density and one mobility are insufficient. Such observations do not identify a
unique two-carrier fit by themselves: densities and mobilities can trade off over a
limited field range. Simultaneous fits to longitudinal and transverse data over both
field signs, with physically constrained parameters, provide much stronger tests.

The classical tensor model also has material and field limits. It assumes a spatially
uniform, isotropic response described by scalar mobilities and ignores quantization of
orbital motion. In a high-mobility two-dimensional electron system at sufficiently
low temperature and high field, Landau-level structure can produce oscillatory or
quantized transverse response; a density obtained from the low-field Drude formula
is then not a description of every field interval. Strong disorder, inhomogeneous
doping, and parallel conducting layers create a different problem: the measured
voltage is an effective mixture of paths that may not be represented by any single
set of bulk electron and hole parameters. Residual structure, thickness dependence,
and reproducibility across contact pairs establish whether a classical multicarrier
fit is warranted.

Hall-bar geometry adds nonidealities before any transport model is selected. The
ideal derivation assumes uniform current density, negligible current through side
contacts, voltage probes centered on equipotential edges, uniform thickness, and a
field normal to the active layer. Finite current pads crowd current near their edges.
Large side contacts average the local potential over a finite region; a contact pair
displaced along the current direction acquires longitudinal voltage. Nonuniform film
thickness changes both current density and the conversion from Hall coefficient to
Hall voltage. A tilted bar senses the normal field component, and a field gradient
across the active area produces an area average rather than the point field at its
center.

Place current contacts far enough from the Hall cross that the central region has
nearly parallel current flow. Use small, high-impedance voltage probes, and measure
thickness at several positions along the active region. Irregular thin films require
an appropriate sheet-resistance method and an explicit contact layout. Field reversal
removes a fixed longitudinal pickup only when contact potentials and current paths
remain unchanged; current reversal and contact-pair interchange provide additional
tests. A discrepancy between nominally equivalent contact pairs indicates a geometry
or contact problem and must be retained for inspection.

Validation begins with the active geometry, thickness map, contact dimensions, and
probe orientation. Establish the linear bias range from longitudinal and transverse
sweeps at zero and nonzero field. Collect interleaved four-state data at several
field magnitudes and temperatures. Retain every raw reading with its current,
field-monitor value, and sample temperature. Inspect residuals of the odd-in-field
and odd-in-current components, compare alternate voltage pairs, and fit longitudinal
and transverse responses before applying a multicarrier model. Reserve one field or
temperature point as a prediction test. Agreement with that withheld measurement
within the geometry, calibration, and noise budget supports the model. The same
record identifies the limiting improvement: active-layer thickness, thermal control,
contact size, field range, or the transport model.

A reported Hall coefficient needs the complete measurement convention. State the
current direction, field direction, named voltage contacts, thickness definition,
temperature, field range, reversal sequence, and the sign assigned to a positive
transverse voltage. Give the odd-in-field slope and intercept separately, together
with the geometry correction and its uncertainty. A one-number carrier-density
estimate is appropriate only when the single-carrier model has passed the field,
temperature, and contact-pair checks. Curvature, contact disagreement, or a
temperature-dependent offset should remain in the record. Those observations
determine whether the result supports a carrier-density inference, a mobility trend,
or only a calibrated magnetic-field measurement.