Materials Science · Magnetism and coupled material responses

Magnetic moments and collective order

Useful background: Atoms, bonding, structure and defects; Energy, stability and environmental context; Measuring and checking material claims. Each worked case explains the quantities it uses.

Back to the chapter

After reading, you can

Distinguish local magnetic moments from collective magnetic order.
Add signed magnetic moments and distinguish total moment from magnetisation per volume.
Distinguish domain cancellation, sublattice cancellation and a thermally disordered local-moment response.
Use a bounded two-state equilibrium model to test a field/temperature comparison.
Identify the additional spatial and condition evidence needed to infer magnetic order.

Moments can exist without a large net magnetisation

Electronic spin and orbital contributions give magnetic moments. Their collective arrangement can reinforce, oppose or fluctuate. Ferromagnetic and antiferromagnetic order therefore describe different organisations, not simply the presence or absence of magnetic atoms.

Two specimens can both give a near-zero net magnetic signal for different reasons. One can contain locally parallel ordered regions pointing in different directions. Another can have an atomic-scale alternating arrangement. A third can retain local moments whose directions fluctuate without long-range order. Start by asking what quantity was summed and over what space and time.

An electronic magnetic moment is a vector: it has magnitude and direction. Electron spin is an intrinsic quantum angular momentum, not a small charged ball literally rotating. Orbital contributions come from electronic states associated with orbital angular momentum, not a measured planetary path around an atom. Pairing, bonding and the surrounding crystal environment affect the resultant moment. The local-moment picture below is useful for this lesson; itinerant electrons in metals and induced magnetic responses require additional descriptions.

The magnetic dipole moment m has SI units A·m², equivalently J/T. Magnetisation M is moment per specimen volume, in A/m. For a uniform-volume average, M⃗ = (Σ m⃗_i)/V. The sum is a vector sum; equal magnitudes can cancel if their directions oppose. With collinear moments use signed components along a declared axis. Dividing A·m² by m³ gives A/m. M is not itself the magnetic induction B, measured in tesla (T).

Ordering specifies a relationship between moments

A ferromagnetically ordered region has a spontaneous aligned component of magnetisation in zero applied field. Collective order comes from interactions among electrons and moments, not merely a list of atoms that possess moments. Exchange interactions have a quantum origin in electronic states, Coulomb interactions and the symmetry constraints on electron wavefunctions. Depending on the electronic structure and connections between sites, they can favour parallel or antiparallel arrangements. They are not simply the ordinary attraction between two miniature bar magnets.

For a simple collinear antiferromagnet, divide the magnetic sites into two interpenetrating sublattices A and B. Neighbouring sites can belong to different sublattices; these are not two macroscopic blocks. Equal opposing sublattice sums cancel the uniform magnetisation while retaining a repeated magnetic pattern. In the illustrative ferrimagnetic case, the sublattices also oppose but their contributions differ, leaving a net moment. Counts as well as moment magnitudes matter.

The new arrangement diagram uses six selected sites. Equal +u moments give +6u for a parallel region. Three +u and three −u give zero for the alternating case. Three +3u and three −2u give +3u for the unequal opposing case. Here u is a supplied unit of moment, not a unit of magnetisation. The drawn arrow lengths encode these stipulated moment magnitudes; site spacing is schematic.

Real magnetic structures can be noncollinear, have more sublattices or carry moments distributed over bonds and electronic bands. A zero at one temperature also need not uniquely identify the entire order type. The three simple arrangements demonstrate different sums; they are not an exhaustive classification obtained from one bulk measurement.

Original schematic comparison of six selected collinear magnetic moments along a common positive vertical z direction. Horizontal site spacing and row separation are schematic, not atomic distances. Arrow length is 24 pixels per supplied moment unit u; u is moment, not magnetisation per volume. In the parallel row, six positive one-u arrows sum to plus six u. In the alternating row, three positive one-u A-site arrows and three negative one-u B-site arrows sum to zero. In the unequal opposing row, three positive three-u A-site arrows and three negative two-u B-site arrows sum to plus three u. A sites use blue circles and B sites violet squares; arrow direction and shape distinguish contributions as well as colour. Each row is a supplied ideal arrangement, not an inference from net signal alone. Sublattices are interpenetrating groups of sites, not separate large domains. The original four-domain schematic elsewhere in the lesson provides the different spatial contrast. No named material, atomic spacing, moment density, temperature or magnetic field is assigned to these rows.

Conditions control order

Temperature, composition and structure can alter magnetic states. A specimen with domains oriented differently can show small overall magnetisation even when each region is magnetically ordered. Distinguish local order from the net signal of the complete specimen.

A magnetic domain is a region sharing an ordered orientation. Oppositely directed ferromagnetic domains can each have a large local magnetisation yet cancel when averaged over the whole specimen. By contrast, sublattice cancellation occurs within the repeating atomic-scale pattern of an antiferromagnet. The retained four-region diagram shows domain cancellation, not alternating atomic sublattices and not a measurement of domain size.

Interactions can lower energy for an ordered pattern, while many thermally accessible arrangements favour disorder. In a local-moment material, warming can destroy long-range orientation correlations without making every local moment disappear. Ferromagnetic and ferrimagnetic ordering transitions are commonly described by a Curie temperature; an antiferromagnetic ordering transition by a Néel temperature. Their values and the possible sequence of magnetic states depend on the particular material. A transition need not be melting or loss of the crystal lattice.

Paramagnetic response can occur when a field biases moments that have no spontaneous long-range magnetic order in the state under study. Thermal motion opposes that bias. Some paramagnetic states retain short-range correlations; the independent-moment model below makes a stronger approximation by neglecting interactions altogether. Diamagnetic response, in which the field induces an opposing contribution, is another mechanism and is not described by the permanent two-state moments in that model.

Direction, field, temperature, specimen geometry and preparation therefore belong with a magnetic observation. Domains, hysteresis and magnetic losses develops changes of domains and history-dependent loops; Soft and hard magnetic behaviour develops reversal and operating-cycle comparisons. Here we need only the distinction between the ordered arrangement and its net sum.

Four schematic domains have oppositely directed local magnetisation. Their sum can be small even when each region is ordered. No domain size or measured magnetisation is assigned.

Add contributions before dividing by volume

Consider an original invented ordered-state record. Each repeating magnetic cell has one A-site moment +3.00u and one B-site moment −2.00u along the same axis, where u = 1.00 × 10⁻²³ A·m². There are 2.00 × 10²⁰ such cells in V = 1.00 cm³ = 1.00 × 10⁻⁶ m³. All cells have the same supplied orientation and no additional magnetic contribution is included.

The cell moment is +1.00u, so total moment is m_tot = (2.00 × 10²⁰)(1.00 × 10⁻²³) = +2.00 × 10⁻³ A·m². Magnetisation is M = m_tot/V = +2000 A/m. The A and B contributions separately are +6000 and −4000 A/m. Adding their magnitudes would incorrectly give 10000 A/m.

If the B-site moment were instead −3.00u at another stipulated state, equal opposing contributions would give M = 0 with nonzero sublattice moments. That cancellation alone does not establish why the state occurred or a complete phase assignment. If instead the original unequal cells form two oppositely oriented regions occupying fractions f and 1 − f of the same material volume, M_avg = f(2000) + (1 − f)(−2000) = (2f − 1)2000 A/m. At f = 0.50 the net is zero; at f = 0.75 it is +1000 A/m.

These calculations supply the arrangement and local contributions; they do not predict them from composition. Increasing the total volume of an otherwise identical uniformly oriented specimen increases its total moment proportionally while leaving M unchanged. A measured moment must not silently become a magnetisation without the appropriate magnetic-material volume.

A field can bias independent moments without producing collective order

Use a separate fictional model of identical independent moments in thermal equilibrium at positive absolute temperature T. Each moment has only two allowed projections along the chosen z-axis: +μ or −μ. These are two quantum projections, not a claim that every classical dipole is restricted to two directions. B is the prescribed uniform induction along that axis, in tesla. Mutual interactions, feedback of the specimen on B and all other magnetic contributions are neglected.

The two energies are E_+ = −μB and E_− = +μB. The aligned moment has lower energy when B is positive. For equilibrium, an energy E has a relative Boltzmann weight exp[−E/(k_B T)], where k_B is Boltzmann’s constant and exp is the exponential function. Only relative weights matter here: divide each weight by their sum to obtain probabilities that add to one.

Define x = μB/(k_B T). Then p_+ = eˣ/(eˣ + e⁻ˣ), p_− = e⁻ˣ/(eˣ + e⁻ˣ) and q = p_+ − p_− = (eˣ − e⁻ˣ)/(eˣ + e⁻ˣ). q is also written tanh(x); the explicit ratio defines it without requiring prior hyperbolic-function notation. The average moment is μq. With number density n = N/V, M_z = nμq. Since μ has units J/T, μB and k_B T are both energies, so x is dimensionless.

At B = 0, both probabilities are 1/2 and the mean is zero for any positive T in this model. Every local moment still has the supplied projection magnitude μ. At nonzero B the mean becomes nonzero, but independent moments do not thereby acquire spontaneous cooperative order. The two-state equilibrium mean also obeys |M_z| ≤ nμ; it approaches its limiting value gradually, rather than reaching it exactly at one finite x.

For |x| much smaller than one, the exponential ratio is approximately x, giving M_z ≈ nμ²B/(k_B T). The leading response therefore varies with B/T if n and μ stay fixed and the assumptions apply. This is a limited weak-field/high-temperature approximation, not a universal law for every paramagnet or every temperature. Its derivative with respect to B has different units from the dimensionless SI susceptibility defined using H; do not substitute B for H in another convention.

Test the approximation against the complete model

Supply μ = 1.38 × 10⁻²³ A·m², n = 1.00 × 10²⁷ m⁻³ and the rounded k_B = 1.38 × 10⁻²³ J/K. Thus nμ = 13800 A/m. The numeric equality of the two given constants has different units: x = B/T only when B is inserted in tesla and T in kelvin for this particular record. The constants, fields and temperatures here are hypothetical model inputs, not a named material specification or an experimental operating instruction.

At B = 1.00 T and T = 10.0 K, x = 0.100, q ≈ 0.099668 and M_z ≈ 1375.42 A/m. At the same B and T = 20.0 K, x = 0.0500 and M_z ≈ 689.426 A/m. The weaker thermal bias is almost half the first response, but the exact ratio is about 0.501248, not exactly 1/2.

At B = 10.0 T and T = 10.0 K, x = 1.00. The complete model gives q ≈ 0.761594 and M_z ≈ 10510.0 A/m, while the linear approximation gives 13800 A/m. Relative to the complete-model value, that linear estimate is about 31.3% too high. The graph deliberately shows the extrapolated straight line as a comparison, not as a valid prediction throughout its range.

Raising temperature in this independent model reduces a field-induced bias; it is not an ordering transition because the model contains no interaction that creates collective order. In a real interacting specimen, a temperature change can instead modify the ordered arrangement, moment magnitudes, domains or phase fractions. A curve that decreases with temperature is therefore evidence needing a model, not a unique diagnosis of disappearing magnetic atoms.

Original independent two-state model, with μ, n and rounded k_B as stated. M_z values are ensemble means; the linear column is an approximation to be tested, not measured data.
B (T)	T (K)	x	Complete M_z (A/m)	Linear estimate (A/m)
1.00	10.0	0.100	1375.42	1380.00
1.00	20.0	0.0500	689.426	690.000
10.0	10.0	1.00	10510.0	13800.0

Original independent two-state equilibrium model, not an interacting ordered magnet or measured material dataset. The prescribed uniform induction B along z runs from zero to 10 tesla. The mean magnetisation M_z is in amperes per metre, with 13800 amperes per metre as n times the supplied moment magnitude. Blue solid curve with circle markers is the complete 10-kelvin model; violet dashed curve with square markers is the complete 20-kelvin model. A dotted orange straight line is the 10-kelvin weak-bias approximation extrapolated for comparison; it is valid only when the dimensionless energy bias is much smaller than one. The plot maps B at 48 pixels per tesla and M_z at 0.035 pixels per ampere per metre. At one tesla, the complete values are about 1375.42 and 689.426 amperes per metre. At 10 tesla and 10 kelvin, the complete value is about 10510.0, whereas the extrapolated line gives 13800. A separate grey long-dashed line marks the limiting mean n times μ; the complete finite curves do not reach it exactly. The origin is zero mean in both positive-temperature independent models, with nonzero individual moment projections retained. Lines/markers distinguish the series without relying only on colour. No Curie or Néel transition, domain motion, hysteresis, actual apparatus or experimental operating recommendation is depicted.

Net signal and spatial order answer different questions

A bulk moment measures a signed sum; it does not state how neighbours are arranged. To illustrate the missing information, take three original zero-field records of 100 sampled neighbouring pairs with signs + or − along one fixed axis. These are supplied illustrative joint counts, not neutron counts or a reconstruction of a named material. In each record each position is + in 50 pairs and − in 50, so its sampled mean sign is zero.

Yet the pair relationships differ. Record P has 96 equal-sign and 4 opposite-sign pairs; record A has 4 equal-sign and 96 opposite-sign pairs; record U has 50 of each. Assign a pair product +1 to equal signs and −1 to opposite signs. Their mean products are therefore +0.92, −0.92 and 0. The same zero mean moment hides very different local correlations.

These nearest-neighbour counts establish only the stated local relationship. They do not alone demonstrate long-range spontaneous order; short-range correlations can survive when long-range order is absent. Longer-distance correlations, direction-dependent measurements and compatible magnetic diffraction can constrain an ordered pattern. A single photograph or one near-zero moment cannot supply that spatial information.

Magnetic neutron diffraction is sensitive to the arrangement and orientation of atomic moments through their scattering pattern. Structural and magnetic contributions, specimen state and candidate magnetic structures must be considered; the absence of one peak is not a universal proof of no moments or no order. As a real material example, antiferromagnetic order in fully dehydrated black dioptase, Cu₆Si₆O₁₈, was investigated using single-crystal neutron diffraction together with magnetisation, with ferromagnetic and antiferromagnetic couplings in different structural directions. It illustrates why one net sum cannot describe every neighbouring relationship; no numerical property of that crystal is used in the teaching records.

A defensible interpretation reports the sample volume, direction, temperature, field definition, preparation and measurement timescale, then states what spatial or local evidence supports the proposed order. Compare matched conditions before assigning a changed mechanism. Measuring an applied-field response is not by itself evidence of a permanent magnet, and a vanished net response is not proof that local moments have vanished.

Original illustrative neighbouring-pair records at zero applied field. Each row has 100 pairs and balanced signs at either position. Counts are not an actual diffraction dataset or a proof of long-range order.
Record	++	+−	−+	−−	Mean pair product
P	48	2	2	48	+0.92
A	2	48	48	2	−0.92
U	25	25	25	25	0.00

Worked examples

Work through the case

A magnetically ordered specimen has nearly zero net magnetisation. Is local order ruled out?

Different domains or sublattices can oppose.
The bulk measurement sums contributions.
Investigate the arrangement and field history.

No. Cancellation can hide local order in the net signal.

A cell sum, a volume average and a domain average differ

Use the original unequal opposing-cell record: N_cell = 2.00 × 10²⁰, V = 1.00 cm³, moments +3.00u and −2.00u, u = 1.00 × 10⁻²³ A·m². Find total moment and M. Then consider equal volumes of oppositely oriented regions.

One cell contributes (+3.00 − 2.00)u = +1.00u. Add signed moments, not their magnitudes.
Total moment is +2.00 × 10⁻³ A·m². Convert 1.00 cm³ to 1.00 × 10⁻⁶ m³ before dividing: M = +2000 A/m.
Equal opposite regions average to 0.50(2000) + 0.50(−2000) = 0 A/m, although each region still has its supplied local order.
This domain cancellation is distinct from equal opposite sublattice contributions within every cell. A zero bulk signal alone cannot choose between them.

Uniform orientation gives +2.00 × 10⁻³ A·m² and +2000 A/m. Equal opposite regions give zero bulk M without erasing their local order.

A thermal change does not automatically identify an ordering transition

For the separate independent two-state model, compare the exact response at B = 1.00 T for 10.0 and 20.0 K. Does the difference establish a loss of spontaneous order?

At 10.0 K the dimensionless energy bias is x = 0.100. Weight the two states by e⁰·¹ and e⁻⁰·¹, giving q ≈ 0.099668 and M_z ≈ 1375.42 A/m.
At 20.0 K, x = 0.0500 gives q ≈ 0.0499584 and M_z ≈ 689.426 A/m. More thermal energy reduces the field bias.
The response ratio is about 0.501248. The linear B/T approximation gives exactly 1/2, but the complete two-state response is slightly nonlinear.
This model has no inter-moment interaction and no spontaneous collective order at either state. At zero B its mean is zero at both positive temperatures while local projections remain.

The mean response drops from about 1375.42 to 689.426 A/m through thermal biasing of independent moments. The calculation is not an ordering-transition diagnosis.

Zero averages can hide opposite correlations

All three illustrative pair records have zero sampled mean at either site. What additional information distinguishes them, and what remains unresolved?

For P, equal signs occur 48 + 48 = 96 times and opposite signs 2 + 2 = 4 times. The mean product is (96 − 4)/100 = +0.92.
For A, the balance reverses: (4 − 96)/100 = −0.92. For U, (50 − 50)/100 = 0.
These records distinguish local parallel, antiparallel and zero sampled pair correlation despite equal zero means.
Only neighbouring pairs were supplied. Without evidence at larger distances and compatible conditions, none of these finite local tables alone establishes macroscopic long-range order.

Pair relationships distinguish the records; one-point averages do not. Long-range order and a unique material mechanism remain unresolved by these tables alone.

Try the reasoning

Does every magnetic material behave as a permanent magnet?
Show the answer and reasoning
No.
Order, domain response and history differ.
Why report temperature?
Show the answer and reasoning
It can change collective order.
The same composition may enter another magnetic state.
A second uniformly oriented specimen has the same cells and number of cells per volume as the unequal opposing-cell example, but V = 2.50 cm³. Find N_cell, total moment and M. Does the larger total moment establish stronger local order?
Show the answer and reasoning
N_cell = 5.00 × 10²⁰; total moment is +5.00 × 10⁻³ A·m²; M remains +2000 A/m. The greater amount alone does not establish stronger local order.
Both cell count and moment scale by 2.50, as does the volume denominator. An extensive total moment and moment per volume answer different comparisons.
At one supplied collinear state, a specimen contains 3.00 × 10²⁰ A sites with +2.00u each and 2.00 × 10²⁰ B sites with −3.00u each, where u = 1.00 × 10⁻²³ A·m². Is the net moment zero? Does zero prove equal individual moment magnitudes or establish the entire order type?
Show the answer and reasoning
The net is zero: +6.00 × 10⁻³ − 6.00 × 10⁻³ = 0 A·m². It proves neither equal per-site moments nor a unique order classification.
Add population times signed moment for each contribution. Unequal populations can compensate unequal magnitudes; local arrangement, temperature dependence and other evidence remain necessary.
In the independent model at 10.0 K, change the prescribed B from +1.00 to −1.00 T. Find the mean M_z. Does this single equilibrium relation imply a hysteresis loop or retained magnetisation?
Show the answer and reasoning
M_z changes from about +1375.42 to −1375.42 A/m. This equilibrium independent model supplies no hysteresis or retained mean at B = 0.
Changing the sign of x exchanges the two state weights and reverses q. At B = 0 the weights are equal. A history-dependent loop requires additional physics and observations.
Use the complete independent model at 10.0 K and 20.0 T. A worksheet extrapolates the weak-bias expression to M_z = 27600 A/m. Check the maximum allowed mean and calculate the complete-model value.
Show the answer and reasoning
The maximum magnitude is nμ = 13800 A/m; the complete-model mean is about 13303.6 A/m. The extrapolated 27600 A/m is incompatible with this model.
x = 2.00, so q = (e² − e⁻²)/(e² + e⁻²) ≈ 0.964028. An average of projections ±μ cannot exceed μ in magnitude. This is a model validity check, not a claim about the maximum magnetisation of every material.
A new 200-pair zero-field record has 90 ++, 10 +−, 10 −+ and 90 −− pairs. Calculate the mean at either position and mean pair product. Do these finite nearest-neighbour counts prove long-range ferromagnetic order?
Show the answer and reasoning
Each mean sign is zero and the mean product is (180 − 20)/200 = +0.80. Long-range ferromagnetic order is not established by this record alone.
Each position is + in 100 and − in 100 cases. The excess of equal signs demonstrates the supplied local correlation. Larger-distance evidence and a suitable state/model are still needed.
A bulk magnetometer gives a small net moment and one selected diffraction position shows no resolved magnetic peak. May the report conclude that the material has no local moments?
Show the answer and reasoning
No. Those two observations do not establish the absence of local moments.
Cancellation, a different magnetic pattern, short-range fluctuations, limited sensitivity or the selected scattering geometry can give weak signals. Compare the full relevant pattern/local evidence with the proposed structure and stated conditions; do not treat a missing single peak as a universal test.

What to carry forward

Separate moments, order and net magnetisation.
Include conditions and history.
Add vector moments before dividing by the stated magnetic-material volume.
Domain cancellation and sublattice cancellation describe different spatial organisations.
Thermal loss of long-range order need not remove each local moment; a field can bias independent moments without cooperative order.
A bounded two-state response tests thermal weighting and approximation limits, not every magnetic material.
Use spatial, field/temperature and condition evidence together; one zero or one nonzero bulk signal does not determine the order.

References

Teaching resources and research supporting this lesson’s explanations and examples.