Directions and specification are available for all possible methods of sampling from all possible materials. For sampling metals and in particular molten metals such as pig iron, cast iron and steel, reference is made to /226/, and for non-ferrous metal /227/. Selected sampling procedures for precious metals are described in /228/.

    These samples are used to monitor the molten metal and must be taken, prepared and examined quickly; we speak of preliminary and final samples.

    The sampling method for spectrometers with SDAR can be simplified and standardized by meeting the following requirements:

   a)  The sample must be representative of the melt.

   b)  The elements must be evenly distributed so that the effective sample of about 1 mg will give a reliable value for the composition. The sample must be suitable for use as an electrode in the discharge gap ie. It should have few and small precipitates. Geometric dimensions must be suited to the sample stands designed for cast samples.

   c)  Samples should not have slag inclusions or cracks.

   d)  Chill mould handing should be simple.

   e)  Sample preparation should be simple.

   Re a) The correct time for sampling is sampling is important. After master alloys have been added. It is necessary to wait until a homogeneous melt has formed. When oxidized scrap is added it is necessary to wait until the oxygen has left the metal bath in the form of oxides.

   Re b ) This requirement is often met by casting minimum-weight analysis samples at cooling rates up to 50°C/s. As the sample mass grows with the square of diameter, it should be as small as possible. A diameter of 25-30 mm is sufficient for 4 burn spots. The sample is usually taken in two steps: the liquid sample is removed from the melt and transferred to the mould.

   During sample-taking, the spoon is coated with the slag floating on the bath in order to prevent reaction between the spoon and the liquid metal and to heat the spoon in order to avoid premature solidification of the sample. The spoon can also be coated with mould wash, dried and preheated; the slag coating can then be dispensed with. The sample material should be hot when poured into the mould in order to ensure a reliable flow. The mould should be near the sampling point, and pouring should take place quickly in order to avoid heat loss. The spoon must not be set down during pouring as this will give rise to cold shuts which require additional sample preparation work. It is also assumed threat cold shuts cause breaking-up at the burn spots since heat conduction to the dense part of the sample is made more difficult. To ensure a reliable flow into the mould the riser should be poured full to increase the inlet pressure.

   The interior of the mould must be metallically bright using a brass brush. Cu moulds with oxidized inert surfaces become ineffective for rapid heat conduction. In pig iron and cast iron samples this may lead to precipitation of free C. The interior must be cleaned after about 20 samples.

   It is recommended that there should be no water cooling, because of condensation. At each sampling point there should be number of moulds which can be used in succession.

   With moulds made of soft or haematite iron or cast iron. Because of poor heat dissipation local melting of the mould material occurs where sample and mould become welded together. When the sample is removed the mould splits at these points. These moulds are suitable for non-ferrous metals with lower melting points. Low melting point metals may have segregations from high melting point metals due to slow cooling, so special care must be taken to have a high cooling rate to obtain homogeneous samples.

   Re c) This requirements can only be met by the sampler’s skill. The slag floating on the sample material in the spoon can be largely removed by pouring back some of the spoon contents with a jerky movement into the melt or by skimming it with a slag knife.

   The sampler should wear blue goggles during sampling so as not to make excessive demands on his eyes. Because of its higher optical emission coefficient he can distinguish the slag from the metal and prevent it from flowing into the mould. Fig. shows Al samples with and without inclusions. Sample a cannot be used for spectrochemical analyses.

   Unkilled steels must be deoxidized. After the slag skin has been removed, Al or Zr is added to be spoon in unit lengths. Deoxidation in the mould is not advisable since the agent may be dissolved before the melt solidifies.

   Rapid cooling is desirable for reasons connected with homogeneity and to obtain small precipitates.

   Oxygen-containing copper with ＞0.05% oxygen leads to porous samples during casting and to incorrect analyses with SDAR. Effective killing can be achieved by adding CuP10-1/200^th of sample weight. Care must be taken to ensure that contamination introduced to the sample by impurities from the deoxidizing agent does not cause interference. If they do cause interference this must be taken into account during calibration.

   When not in use, moulds should lie on one side or be covered to that no dirt can enter them.

   Some qualities of materials become cracked become cracked or break up if they are quenched in water from deep red heat. Cast iron samples with phosphorus concentration ＞0.15% must be removed from the mould after ＜10s and cooled in air or by compressed air. If they are left 15s in the mould like other cast iron samples they often break up because of high internal stresses. Sample transport to the laboratory can be included in the cooling time if steel containers are used.

  Mould shape and sample dimensions once differed widely. The shapes most widely used for ferrous and non-ferrous metals were cylindrical with a diameter of about 35mm and a height of 50-100mm. Sample weights were 0.5-1kg and the mean cooling rate was a few ℃/ s.

   Steel samples with relatively large quantities of precipitates eg. MnS  were cut to expose the bottom third using expensive cutting machines. In this region the distribution of the precipitates is, to some degree. homogeneous. The size distribution of the MnS precipitates from the base upwards is as shown in Fig. 5.5. The chemical concentration of manganese and sulphur below the solidification shrinkage cavity is the same all over. Since intensities were formerly measured before the SS was reached, they increased with increasing MnS Ø. For many years the “best part”of the sample was cut off, namely the bottom part with small precipitates.

    For pig iron and cast iron a wider range of mould shapes was used than for steel.  All efforts were concentrated on obtaining small analysis samples with high cooling rates, as was and still is obligatory because of the risk of free carbon precipitation.

    Samples from non-ferrous metal were often cast in a “mushroom-shape” but the diameter of the analysis was often large than necessary.

On the basis of these requirements and explanations, sampling can be simplified as follows:

    1. Metal with small precipitates and little tendency towards segregation can be cast in Mould. A ring 25 mm high with an internal diameter of 30mm lies on a copper plate at least 1 cm thick with a metallically bright surface. The ring is made of high-alloy steel or ceramic so that heat is mainly conducted into the copper plate. This gives rise to directed structure with good homogeneity over the surface of the sample. Vertical homogeneity must be tested.

Mould is suitable for the following metals:

Pure metals

Genuine hypoeutectic metal alloy eg. CuZn, CuNi, ZnAl.

   Test on steel with 0.06% S have shown that samples from Mould 1 are homogeneous over at least 10mm from the base. These steel samples, after grinding, are ready for analysis. A cutting machine is not required. Much the same applies to the most alloys on Ni and Co bases. Fig 5.8 also shows 4 samples taken from this mould. The samples can be removed from the ring after 25s.

   2. Metal with large precipitates and a tendency towards segregations should be cast . The samples thus obtained are 6mm thick and about 30mm Ø.  The upper part can be cast iron. The riser and underside of the upper part should have oxidized surfaces so that the liquid samples flows into the analysis sample shape as far as possible without cooling. The analytical sample mould must be brushed bright by regular use of a brass brush in order to ensure rapid directional solidification. The cooling rate is about 25°C/s. The samples can be removed from the mould after about 15 s is necessary to reach the SS. If the same material is cast into Mould fine droplets are obtained and a pre-spark time of 30 s is sufficient. The samples with large precipitates give poor calibration and excessively high intensity. Samples with precipitation sizes different from the foregoing lead to incorrect analysis results. 

   All types of sample in group 1 can be cast in the Mould , but manipulation of Mould 1 is easier.

   Samples which require a high cooling rate in order to suppress precipitates can be cast. Their thickness is 3-4mm, Ø about 30 mm and their weight is 1.5 times less than the weight of samples from Mould. The analysis sample has a cooling rate of about 50°C/s and can be removed from the mould after about 15s.

    Pig iron and mixer metal, cast iron, malleable iron and nodular cast iron.

    Most samples have solidification hollows on their underside which, if the riser is at the centre, should be removed by grinding. If the hollow is at the edge it need not be ground-out as there is still sufficient surface for 4 burn spots. There is less risk of cold shuts occurring with this mould than with top pouring. Cooling is sufficient. With a C equivalent of:

C eq=%C +1/3(%Si+%P)＜5.0,

equivalent to a degree of saturation S =5.0/4.2=1.2, to prevent grey solidification. In nuclei-rich iron grey solidification is promoted, but is retarded if carbide-forming agents are present. Grey solidification can be recognized in the fracture or form the C burn-off curves. Metallically bright mould interiors and additions of elements suppressing grey solidification promote white solidification.

    A mould developed by Georg Fischer Foundry, Schaffhausen CH, for sampling cast iron consist of a solid Cu block with openings on both sides Cu block with openings on both sides for the analysis sample which is 25 mm Ø and 5 mm thick.          The sample material is poured through an insulating sand core in order to reach the opening while still as hot as possible.        

   The riser can be knocked off. The underside of the opening must be kept metallically bright by regular use of brass brush.      

    The mould can easily be remachined.

    The cooling rate can be increased by casting pin samples. The samples is 40mm long, 6 mm in Ø, and the weight of the analysis sample is 8 g. Compared with this, the weight of the coin-type disc sample from Mould 3 is 30 g.

    A large number of cast iron samples cast in parallel as 4 mm coins and 6 mm pins with Ceq ＜5.0 show free C in the fracture for 65% of the coins. The pins were all graghite-free. The pin sample has the advantage that sample fracture is part of the sample preparation process and free C is detected. If the sample contains graphite, then either an analysis programme with a longer pre-spark time should be adopted; the sample should be discarded; or it should be homogenized in a further sample preparation stage. Sample preparation for the pin sample is simpler and faster than for the coin sample. As pin samples always have the same Ø, they can easily be fitted into a special holder. Except for sulphur, the same calibration curves are obtained as for coin samples, so that calibration can be carried out with coin reference samples. In precision and accuracy they are comparable with coin samples. Due to the poor thermal conductivity of these cast iron samples pre-sparking should be carried out with discharge parameters which do not lead to thermal overloading of the sample surface.

   In hypereutectic alloys precipitation is often already present in the melt and increasingly so as the temperature approaches the solidification temperature. This applies eg. to precipitated carbon in hypereutectic cast iron. This precipitated carbon is not detected during thermal analysis as it does not affect the holding temperature. When the sample is crushed for chemical analysis, its precipitates turn to powder, leading to low analysis values. ues. Pig iron which contains graphite foam in the run-off channel does not acquire a white structure even with rapid cooling. Coin and pin samples from such material contain graphite foam over their cross-section. Such samples should be prepared by remelting in separate furnaces, possibly with a pure iron addition in order to reduce the C eq.

   There are alloys, however, which show considerable segregation even at cooling rates >50°C/s. Examples are Al alloys with a few percent of Cu. In these cases pouring at a high cooling rate would lead to analysis samples which are far from homogeneous. For this reason they are poured into Mould in order to find a region which is homogeneous over some mm from the base of the sample outwards. This region is the analysis zone of the analysis sample which has to be stated and adhered to in subsequent analyses. The concentration in this zone need not correspond to the mean bulk concentration (that of the but it is proportional to it. For this reason, calibration curves plotted from calibration samples must be given a corresponding factor.

   Because of the concentration gradient a repeat analysis carried out after more material has been removed from the sample base (by turning, milling) leads to larger systematic deviations in the mushroom sample than in the cylindrical sample. If material is removed from the samples, the composition of the remaining portion of sample no longer corresponds to the spoon sample (molten metal), which should be borne in mind when carrying out checks by wet chemistry.

    Direct sampling from melts can also be carried out using submerged moulds which are dipped into the melt. The inner space is covered with wood or some other material so that slag cannot run in when it is penetrated. After a few s the mould is so hot that the excess pressure forces the wooden cover off and metal flows in. The sample space contains a deoxidising agent. With this sampling technique, oxidation is much less than when pouring from a spoon through air into the mould. Scrap rates with submerged samples are 3 times higher(about 10%)than with samples poured by spoon into a mould.

    Direct sampling is also carried out using submergible probes/114/.This type of sample is sometimes called a "lollipop”, because of its shape. Steelworks are trying to reduce sampling and preparation times because of expensive furnace time. It is difficult to deoxidise unkilled steel in probes. X-ray examinations show gas bubbles sometimes larger than the metal volume in many samples.

     After 3 s it is full. The rapidly solidifying coin sample is removed by breaking up the ceramic front part. There are similar versions for sampling from melting units, ladles and moulds.

    The times required for sampling, transport and preparation are often many times longer than the analysis time at the spectrometer. It is possible to shorten the overall time by appropriate organisation and the use of optimum working aids in this area. At many melting plants today analysis is carried out on site /229/, and further progress is being made here with automation.