Processing and demanding service environments

A powder compact must become a working solid

Powder preparation, forming and sintering influence density, grain growth and residual pores. Sintering develops bonds between particles and can densify the compact; it need not melt the whole compact. Not every transport path produces shrinkage: surface diffusion can grow a connecting neck without bringing particle centres closer. Densification, grain growth and their spatial variation must therefore be assessed separately.

Two ceramic pieces have the same dry bulk density. One has a porous surface region; the other has most of its pores nearer the centre. Can a density certificate alone decide how they respond when their surfaces cool faster than their interiors? We need both a processing history and a location-sensitive service model. First separate how particles bond from how the body changes volume.

A green compact is the shaped, unsintered particle assembly. Powder chemistry, particle sizes and agglomerates—clusters of particles—affect how it packs. Forming converts that feedstock into a shape; a binder may provide handling strength before ceramic contacts have developed. Where an organic binder is used, its removal is a separate stage: losing binder mass is not itself evidence that pores have disappeared.

A nonuniform fill or forming history can leave different amounts of solid in equal local volumes. Later, those regions need not shrink together. Powder conditioning and forming therefore establish an initial spatial state, not just an external outline. Record the preparation and inspect the interior as well as the accessible surface. Powder consolidation and polymer processing introduces how feedstock and route leave different evidence.

A wider contact does not tell us how much pore volume was removed

A neck is the solid bridge at a contact between particles. Surface energy gives a tendency to reduce energetically costly surface; atoms must still have an available transport route. In the ideal surface-diffusion neck-growth picture, atoms move along the free surface into the neck. This can strengthen a contact without moving the particle centres together. It changes the shape of the same solid rather than supplying a measurement of bulk densification.

Densifying transport instead reduces the space occupied by the compact while its solid is conserved. In an ideal solid-state picture, transport from a grain-boundary contact region to a pore/neck, through boundaries or the lattice, can allow particle centres to approach. Saying only “bulk diffusion” is incomplete: the source and sink matter. A lattice path fed from an outer particle surface can contribute to neck growth without that same shrinkage.

Grain growth is a third change: boundaries move and some grains consume others. It is not another name for decreasing pore volume. Large grains can coexist with retained pores. Grain and pore evolution interact, so a longer or hotter hold cannot be assumed to improve density and every service property together. Trapped gas can oppose pore shrinkage, and isolated pores may persist. Applied pressure changes the driving conditions; no pressure route guarantees that all defects disappear.

Solid-state sintering does not require melting the whole compact. Other routes can use a liquid fraction, but their phase chemistry and transport need their own account. Here the quantitative mass balance concerns one unchanged solid phase after binder removal. Diffusion and phase transformations develops transport paths; the present lesson does not supply a furnace schedule or a rate law.

Use a balance before interpreting a shrinking shape

Define relative density r = ρ_bulk/ρ_s, where ρ_s is the density of the pore-free solid at the comparison temperature. For a dry single-phase compact with negligible pore-gas mass, ρ_bulk = m/V and V_s = m/ρ_s. Thus r = V_s/V. Density and properties relative to mass explains why this includes pores in V.

Consider the invented records G, N and D below. They refer to the same dry compact after binder removal, cooled for every dimensional comparison to the same temperature. Its mass stays 15.36 g and its single solid phase has ρ_s = 3.20 g/cm³ throughout. The solid volume is therefore 4.80 cm³. G is a 20.00 mm cube of bulk volume 8.00 cm³ and relative density 0.600. N has wider contacts but the same outside volume and density. D has bulk volume 5.00 cm³, giving r = 4.80/5.00 = 0.960.

At conserved mass and solid density, r_iV_i = r_fV_f. Therefore V_f/V_i = r_i/r_f. If shrinkage is also uniform and isotropic, every length has the same ratio λ = (r_i/r_f)^(1/3). The exponent is one third because a cube has three shrinking dimensions: V_f/V_i = λ³. Linear shrinkage is 1 − λ, not 1 − r_i/r_f.

For G→D, V_f/V_i = 0.600/0.960 = 0.625. The bulk-volume reduction is 37.5%, but λ = 0.854988 and the linear shrinkage is only 14.50%. The final cube side is 17.0998 mm. The pores fall from 8.00 − 4.80 = 3.20 cm³ to 5.00 − 4.80 = 0.20 cm³. A 93.75% reduction of the initial pore volume is a different denominator again.

These invented records demonstrate an inference, not a sintering trajectory, measured material dataset or prediction of the time needed. A lost binder, oxidation, changed phase density or anisotropic shrinkage would require a different balance. A contact micrograph alone cannot select the N or D volume record.

Regions that would shrink differently are attached to one another

Suppose two separate, initially 20.00 mm cubes have relative green densities 0.600 and 0.660 after binder removal. If each could independently reach 0.960 at conserved mass and solid density, their free final sides would be 17.0998 and 17.6517 mm: free linear shrinkages of 14.50% and 11.74%. The initially denser cube needs less volume reduction to reach that same final relative density.

Now imagine regions with these differing free shrinkage tendencies joined within one compact. They cannot each attain their unconstrained shape while maintaining a perfect shared interface. Compatibility creates a tendency to distort or develop stresses; cracks can result if the evolving resistance is exceeded. The two-cube calculation diagnoses a mismatch tendency. It does not calculate the joined object’s final curvature, stress or crack probability: sintering rates, viscosity, geometry and the connection are also needed.

This provides a specific route from nonuniform packing to service: unequal densification can leave distortion or retained local flaws; final grinding to restore dimensions can introduce or expose surface damage; a later tensile thermal transient may act at that surface. No one link proves that this is an observed failure origin. Preserve local measurements, preparation history and fracture evidence before assigning a cause.

Derive a simple thermal-stress map before comparing local states

For an answerable location comparison, use a deliberately simplified assembly, separate from the cube records. Three straight bars of the same initial length connect two rigid endplates. Two identical outer bars represent a cooler surface group S; their combined area equals that of the warmer central bar I. The arrangement is symmetric. In this quasistatic model the plates remain parallel, move without an applied axial resultant, and impose one common axial strain ε. Each group has reference volume 10.0 mm³. End effects, transverse stresses, bending, buckling, joint failure and time-dependent strain are excluded.

The bars begin stress-free at 300°C. At one instant in a short cooling ramp, each outer bar is uniformly 240°C (ΔT_S = −60 K), and the central bar is uniformly 280°C (ΔT_I = −20 K). Temperatures are supplied, not calculated from a heat-flow model. Both states have linear expansion coefficient α = 5.00 × 10⁻⁶ K⁻¹. For small uniaxial linear-elastic response, total strain equals elastic plus free thermal strain: ε = σ_j/E_j + αΔT_j. Here j labels S or I, E is Young’s modulus, and positive σ means tension. Thus σ_j = E_j(ε − αΔT_j).

There is zero axial resultant, so Aσ_S + Aσ_I = 0 for the equal combined reference areas. Substitute the two stress expressions and solve: ε = α(E_SΔT_S + E_IΔT_I)/(E_S + E_I). This is a stiffness-weighted strain, not generally the arithmetic average of the two thermal strains. “No external axial force” requires opposite group forces; it does not require zero stress in each bar.

The fictional finished states L and H have separately supplied properties: L has r = 0.920, E = 100 GPa and tensile reference endpoint R = 14.0 MPa; H has r = 0.980, E = 120 GPa and R = 30.0 MPa. These are invented matched-condition inputs for the local comparison, not a law deriving modulus or strength from density. Assume E and α constant over this short ramp. R describes a stipulated tensile comparison at this ramp’s local temperature and preparation state; it is neither a population lower bound nor an allowable for a real ceramic.

Arrangement A puts L in the outer group and H in the centre. Its common strain is −190.909 microstrain, where one microstrain is 10⁻⁶. The outer free thermal strain would be −300 microstrain, so compatibility stretches that group by +109.091 microstrain relative to its preferred length. Multiplying by 100000 MPa gives +10.909 MPa tension. The warmer centre is compressed to −10.909 MPa. Arrangement B swaps L and H and gives ε = −209.091 microstrain. Its surface is also in +10.909 MPa tension, now carried by H.

The equal tensile magnitudes on swapping follow from this special two-group model: σ_S = [E_SE_I/(E_S + E_I)]α(ΔT_I − ΔT_S). A real swap can change heat transfer, geometry and other properties; the field would then need to be solved again. Thermal expansion and thermal stress supplies the strain decomposition; here equilibrium and compatibility determine the previously unknown movement.

An average cannot replace the state at the tensile location

At the common stress-free reference temperature, the equal-volume average is (0.920 + 0.980)/2 = 0.950 for either arrangement. With the same stipulated ρ_s = 3.20 g/cm³, each has total dry mass 0.06080 g in 20.0 mm³. These density and mass comparisons use reference volumes, before the small thermoelastic changes; we are not claiming identical local volumes during cooling.

For this teaching problem only, require 1.50σ_S ≤ R_S, using the positive surface tension and the endpoint of the state actually at S. The comparison demand is 1.50 × 10.909 = 16.364 MPa. A fails this screen because 16.364 > 14.0; B meets it because 16.364 ≤ 30.0. A misleading average endpoint, (14 + 30)/2 = 22 MPa, would hide the weak surface in A.

Meeting that one screen does not approve B as a component. The compression group, joints, flaw population, ramp reproducibility, model validity and other service mechanisms remain unassessed. The factor 1.50 is a declared classroom comparison, not a recommended ceramic design factor. Density also cannot identify sharp machining cracks or certify R: those require their own evidence. The purpose is to connect a local state to a local demand rather than to rank density in isolation.

Service tests must reflect the part

High temperature can promote creep or reaction; thermal gradients can generate stress. Evaluate joining, coatings, machining damage and the intended environment. A successful laboratory coupon does not establish an unlimited lifetime for a large, differently processed component.

A short cooling ramp and a long hot dwell ask different questions. During a dwell, thermally activated deformation can accumulate under load; in some ceramics a softened secondary phase at grain boundaries contributes. Under constrained movement, added time-dependent strain can also redistribute stress. The short, purely elastic model above contains no creep term and cannot predict either rate or rupture time.

Chemical exposure changes the state too. For example, silica-forming non-oxide ceramics can develop an oxide layer, yet water-vapour-containing environments can also remove silica through volatile silicon hydroxide species. The balance depends on composition and exposure. A coating must be assessed for its actual chemistry, continuity and connection to the substrate; the word “coated” does not supply a universal lifetime.

Final machining can change the surface flaw population after a density measurement; a joint can change restraint and thermal mismatch. Retain the finished route and inspect the exposed and connected state. To extend the example, first obtain relevant temperature histories and local properties for that state, then test time, gas composition, cycling and loading effects that the intended service introduces. Sampling, scale and extrapolation limits explains why several convenient coupons do not establish every location, production run or larger part.

If a changed-exposure test supplies a new H tensile endpoint of 12.0 MPa while a fresh analysis still gives the same 10.909 MPa surface tension, B no longer meets the same local screen: 16.364 > 12.0. That conclusion uses the stipulated new evidence. Neither unchanged average density nor a pass before exposure would justify keeping the old endpoint. The example gives no numerical rule for how fast the endpoint changed or how long the part can operate.

Worked examples

Try the reasoning

1. Must sintering melt the complete part?

   Show the answer and reasoning

   No.

   Solid-state transport can bond and densify particles. A solid-state route can form necks and densify. Identify whether the active source–path–sink combination reduces bulk pore volume; a neck alone does not prove that it does.

2. Why inspect after final machining?

   Show the answer and reasoning

   Machining can introduce new surface defects.

   The final state is the service state. A preceding density or shape measurement need not detect the new surface state. Compare relevant final preparation and local tensile response, rather than carrying the earlier evidence forward unchanged.

3. Two same-temperature dry records have unchanged mass and solid phase. The neck looks wider, but all three external dimensions remain unchanged. A report says relative density doubled. What balance contradicts that claim?

   Show the answer and reasoning

   With m, ρ_s and V unchanged, r = m/(ρ_sV) is unchanged.

   A local contact-shape change supplies no factor of two in that balance. Densification must be established independently; solid volume conservation alone also does not distinguish a densifying from a non-densifying path.

4. A different dry, single-phase compact changes from r_i = 0.640 to r_f = 0.800 with conserved mass and solid density. Its measured x and y length ratios are each 0.950. What z ratio would satisfy the balance? May all three ratios instead be replaced by (0.640/0.800)^(1/3)?

   Show the answer and reasoning

   The z ratio must be 0.800/(0.950²) = 0.886427. The measured anisotropy rules out imposing one common cube-root ratio.

   Volume ratio is 0.640/0.800 = 0.800 and equals λ_xλ_yλ_z. The isotropic cube-root ratio would be 0.928318, which is inconsistent with the two supplied 0.950 ratios.

5. Both groups in either thermal assembly now cool uniformly by 40 K, with the same α and no external axial resultant. Find the common strain and stresses. Does the result mean rapid cooling can never generate stress?

   Show the answer and reasoning

   ε = −200 microstrain and both axial stresses are zero in this uniform-temperature model.

   Each group has the same free thermal strain, so no elastic correction is required even though E differs. The original nonuniform cooling had incompatible preferred strains. Heat transfer, time history and constraints determine which temperature field actually occurs.

6. For the original arrangement A, a worksheet uses the arithmetic mean strain −200 microstrain. Calculate the resulting two group stresses and show why they do not satisfy the stated axial equilibrium.

   Show the answer and reasoning

   It gives σ_S = +10.0 MPa and σ_I = −12.0 MPa. Their equal-area resultant is not zero.

   For the problem’s equal combined areas, A(10 − 12) MPa leaves a net compressive resultant. The correct stiffness-weighted strain −190.909 microstrain gives equal and opposite 10.909 MPa stresses. Common strain does not mean equal elastic corrections when free strains and moduli differ.

7. After a different exposure, H has a newly supplied tensile endpoint of 12.0 MPa. A fresh analysis retains the original B surface tension +10.909 MPa. The average relative density is still 0.950. Does B still meet the same local screen, and can these numbers establish a service life?

   Show the answer and reasoning

   It fails: 1.50 × 10.909 = 16.364 MPa exceeds 12.0 MPa. No service life follows.

   Unchanged density does not preserve the previous tensile endpoint. The new evidence supports this changed screen only; it supplies neither a degradation-rate law nor a lifetime distribution.

What to carry forward

• Manufacturing sets defect and density distributions.
• Evaluate the finished component under relevant exposures.
• Neck growth, densification and grain growth answer different questions; identify the transport source, route and sink.
• Conserved mass and solid density constrain volume change; an isotropic length ratio additionally requires uniform equal shrinkage in all directions.
• Combine a local property/state map with local service demand. An average can conceal the vulnerable tensile location.
• A short elastic screen does not establish hot-dwell, chemical-exposure, joint or whole-component life.