To understand the origin of these occasional spurious signals, it is necessary to consider both the P.I. technique and the mechanism of magnetic phenomena.

In the P.I. method, pulses of current are repeatedly sent through a coil in the search head, and produce a pulsed magnetic field which propagates to the target. At the instant of switch-off of this primary field, a large transient back-emf voltage is induced in the coil which however only lasts for a few microseconds after which there is no voltage across the coil until the switch-on of the next pulse.
The switch-off will also induce eddy-currents to flow in any conductive target present; since there is no source of energy to maintain them, they will decay away but nevertheless persist for a time of several to a few hundred microseconds, i.e. for longer than the primary switch-off transient. These eddy-currents generate a secondary magnetic field which propagates back to the search head and induces a voltage in the same coil that generated the primary field pulse. The electronics of the receiver circuit samples the coil voltage after a delay which is long enough to miss the primary back-emf spike, but short enough to include the eddy-current signal (if present).

Poorly-conducting materials such as thin foil and alloys such as stainless steel produce a signal with a very rapid decay time comparable to the sampling delay in the receiver circuit: careful selection of this delay time (and of other related pulse-widths) enables an instrument to be set to either detect or ignore such objects. The ionic conductivity of salt or brackish water is so low (by comparison) that any signals generated decay away completely within the delay time, and so generate no response.
A purely magnetic non-conductor, such as ferrite and most magnetic minerals, will become magnetised by the primary field and will demagnetise immediately on the removal of the primary field, and so will not induce any signal in the coil during the delayed sample. A conductor which is also magnetic (ferrous) will produce a signal in exactly the same way as a non-magnetic conductor, but the strength of the response will be magnified by the effective permeability of the target (which depends on shape and orientation as well as on the absolute value of the relative permeability of the material).

The above list of examples should account for all target types, but there remains one further phenomenon: magnetic viscosity. Here the material is non-conductive, so eddy-currents are not generated. The material exhibits a magnetic permeability (or susceptibility), but this should not generate any signal during sampling. However the term "viscous" refers to the fact that the induced magnetisation does not vary instantaneously with applied field, but changes sluggishly, so that at a finite time after the primary switch-off the magnetisation is still present and reducing; this produces a signal of much the same time-characteristics as an eddy-current signal and so does generate a response in the detector.

Unfortunately, magnetic viscosity is not described in physics textbooks, presumably because
(a) it is not well understood, and (b) it is a rare occurrence.

Atoms of all elements exhibit either diamagnetism or paramagnetism, but even in bulk material both of these are very weak effects, and so materials exhibiting only one of these are normally considered 'non-magnetic'. A magnetic material occurs only when neighbouring paramagnetic atoms "join forces" and line up their magnetic moments to produce ferromagnetism or ferrimagnetism. In the simplest case, the bulk material is comprised of small grains, and each grain has one or two magnetic "domains". Each domain is a tiny permanent magnet as all the atomic magnetic moments are parallel to each other. A single-domain grain is also a magnet; in a two-domain grain the two domains are anti-parallel and the boundary between them divides the grain into two equal halves, so the grain has zero net magnetic moment. In the bulk material, the single-domain grains are oriented in random directions, so the bulk material has no net magnetisation.
When a magnetic field is applied to such a material, both types of grain are affected. The single-domain grains rotate their magnetisation vector very slightly towards the direction of the applied field. In the two-domain grains, the boundary moves sideways very slightly such that the domain more nearly in the direction of the field grows in size and moment at the expense of the anti-parallel domain. The net result of both effects is a significant magnetisation in the direction of the applied field; when the field is removed, all grains return immediately to their former state. This is "induced magnetisation", measured as susceptibility or permeability.

If a sufficiently strong field is applied, the torque on the domains may be sufficient to wrench whole domains round to a new direction and stay there when the field is removed; the bulk material will then be permanently magnetised (this technique is often used for producing the magnets in loudspeakers). The more commonly-described method of producing a permanent magnet is to heat and then cool the material whilst in a steady magnetic field. Here the thermal energy agitates the grains and domains so that they can be easily rotated by even a modest magnetic field; when the temperature is reduced, the domains remain "frozen" in their new positions. For bulk ferrous materials with uniform grain sizes there is a critical temperature for this effect to occur known as the "Curie point"; this is "thermo-remanence".

If however the grains are not uniform, there is no one single Curie point; instead each individual grain has its own "blocking temperature", and there will be a wide distribution of these blocking temperatures. Similarly, although the average temperature of the bulk material may be well defined, at the microscopic level the thermal energy of a tiny grain may instantaneously be either more or less than the overall average, and there is a statistical distribution of thermal energies extending to temperatures well above ambient and having a (very small) tail right up to the blocking temperatures.
If a magnetic field is suddenly applied or removed from such a material, some grains will "flip" almost immediately, others will take longer to do so, and even after a considerable period of time there will still be a few changing direction. The distribution of "relaxation times" has been observed to extend from microseconds to literally centuries! This effect is viscous magnetisation, and is always accompanied by induced magnetisation though the observed induced magnetisation is really the short-term component of the viscosity.

As stated earlier, the subject is not widely understood, and most of the studies on the occurrence and mechanism of magnetic viscosity have arisen in the archaeological field where magnetic methods are used to locate buried pottery kilns and fired-clay structures, and even to date then by measuring their remanence recording the past direction and strength of the earth's magnetic field. Although the above techniques rely on "conventional" induced and thermo-remanent magnetisation, viscous magnetisation is observed to occur; and a mineralogical study has shown the minerals responsible for all three forms of magnetism.

Most soils and all clays contain a significant amount of iron oxide. Conventional chemical analyses normally presents the iron content as Fe₂0₃ regardless of the actual oxidation state or crystalline mineral present. In fact there are three main iron oxide minerals:- haematite, magnetite and maghaemite; uncultivated soil and raw clay usually only contain predominantly haematite, which is weakly ferrimagnetic. Magnetite, as its name suggests, is strongly ferrimagnetic. Maghaemite has the same chemical constitution as haematite, but a totally different crystal structure; not only is it fairly strongly ferrimagnetic, it also exhibits magnetic viscosity.

The conversion of haematite to maghaemite appears to be a reduction-oxidation reaction:-

All this ties in well with the three main instances when magnetic viscosity may cause spurious responses in a pulse induction metal detector:-

(a) soil effect when searching the ground for buried pipelines etc; the low iron content of most soils and the low likelihood of conversion mean that this is rarely serious, except where outcrops of "ironstone" are met or road surfaces have used quantities of slag (which causes a metallic signal anyway);

(b) signals from brickwork when searching for wall-ties; although a conversion of haematite to maghaemite is quite likely during the brick-firing, in practice only a small number of brick types such as "blue" engineering bricks exhibit a troublesome effect;

(c) aggregates in concrete when measuring reinforcing bars; only pfa- and ggbs-bearing concretes have been observed to be prone to the effect, and only Lytag shows a measurable signal which still does not prevent the Elcometer P350 & P351 and Elcometer P330 cover meters from reading accurately.

The significance of the fact that the P.I. technique is immune to induced magnetisation and only rarely affected by viscous magnetisation should be emphasised by comparison with other techniques. The continuous-wave metal detector types (balanced coil, induction-balance and b.f.o.), and cover meters using the "magnetic reluctance" technique, are highly sensitive to induced magnetisation, often to the point of unusability; magnetic viscosity is never noticed because it is always swamped by induced magnetisation signals. Even metal detectors with provision for minimising "ground effects" suffer a drastic loss of sensitivity to ferrous metal when used in that mode; or else employ a continuous auto-zero circuit which requires the search head to be in continuous motion for metal to be detected, and lose the metal signal whenever the head is stopped thus making exact location impossible.