Skip to main content

Microstructure, anisotropy and process–structure–property links

📍 Where we are: Part I · The Physics of the Melt — Chapter 3 of 19. Chapter 2 covered flaws and stress; this chapter is about the metal itself: the grains and sub-grain structure that the freezing pool leaves behind, and the properties that follow.

The simple version

Frozen water grows ice crystals as long needles in the direction heat leaves. Metal freezing under a laser does the same: crystals grow upward in columns, layer after layer, like the grain in a plank of wood. Wood is strong along the grain and weak across it, and AM metals can behave similarly. On top of that, 3D printing freezes metal so fast that it forms a fine internal "honeycomb" that conventional casting never makes. That honeycomb can make the metal stronger, which is a bonus, and the grain direction can make it uneven, which is a problem.

What this chapter covers​

We describe the typical microstructures of AM metals, how they differ from wrought material, which properties come out better or worse, and why the chain process → structure → property (PSP) is still more an aspiration than a calculator. Throughout, we label evidence as lab-scale, production, or contested.

Problem​

An engineer needs to know the strength, ductility, fatigue life and toughness of a printed part before it flies or is implanted. In wrought metal, decades of data link a rolling and heat-treatment schedule to those numbers. For AM, the same alloy can show different grain size, texture and phase content depending on machine, scan strategy and geometry, and therefore different properties. Reviews of AM metals stress the anisotropy and heterogeneity of microstructure and mechanical properties as a core issue [1].

State of the art​

What the process produces​

  • Columnar grains and epitaxy. Heat flows mainly down into the previous layers, so new solid grows from the existing grains (epitaxy) and the grains extend across several layers along the build direction. This is a standard finding summarised in comprehensive reviews of process, structure and properties [2].
  • Texture. Columnar growth often brings a preferred crystal orientation, so stiffness and strength differ with loading direction (anisotropy) [1].
  • Cellular substructure. In alloys such as 316L stainless steel, the rapid solidification forms sub-micrometre cells with dislocation-rich, solute-enriched walls. Work on additively manufactured stainless steel linked this hierarchical structure to a combination of high strength and ductility [3], and later work examined how the cells strengthen the steel and how they hold up thermally [4]. Demonstrated at lab scale, widely reproduced.
  • Non-equilibrium phases. Fast cooling and repeated reheating produce martensite in titanium alloys and steels and suppress or alter precipitation in others. Recent coupled flow-and-microstructure modelling of LPBF (2023) attributes the transition from ultra-fine to coarse cells and the intercellular segregation to solute trapping and melt convection, a modelling result rather than a direct measurement [5].

Controlling the structure​

LeverWhat it doesEvidence level
Scan strategy and beam parametersSteers local heat flow, giving columnar-to-equiaxed transitions or site-specific orientation. In electron beam melting of a nickel alloy, orientation was controlled region by region [6].Lab-scale; some production use
Nucleant particlesAdd nucleation sites so grains are fine and equiaxed, which also reduces cracking [7]. A 2025 study of Ti alloys deposited by directed energy deposition found a constitutional-supercooling parameter to be the most reliable composition guide for columnar-to-equiaxed transitions [8].Lab-scale
High-intensity ultrasoundDisrupts columnar growth in directed energy deposition [9].Lab-scale
Intrinsic heat treatmentRepeated reheating by later layers precipitation-hardens a maraging steel during the build [10].Lab-scale
Post-build heat treatment and HIPRecrystallise, relieve stress, close pores.Production

Properties: where AM metals stand​

As-built AM metals often show high yield strength from fine structure and, for some alloys, lower ductility or direction-dependent behaviour than wrought. Reviews of the mechanical properties of AM metals compile these comparisons [11]. Static properties are the easier case: they respond to grain structure and are fairly repeatable once the process is tuned.

Fatigue is the hard property. Fatigue life is set by the worst local feature, so it responds to the tail of the defect distribution, the surface condition and the microstructure all at once [12]. A state-of-the-art review explains that intrinsic defects (lack of fusion, pores, unmelted particles) act as local stress concentrators where fatigue cracks start, and that process optimisation and post-processing can reduce but not fully eliminate them [13]. A second review covers post-processing (heat treatment, HIP, shot or laser peening, finishing, machining) that targets porosity, surface finish and tensile residual stress, and reports significant fatigue improvements [14]. How closely treated AM material matches wrought fatigue performance is still contested: the data are scattered across machines, alloys, defect populations and test methods.

A process-structure-property-performance chain with example links and break points: parameters to melt pool, melt pool to defects and grains, grains and defects to static and fatigue properties, with solid arrows for links that are modelled or measured and a dashed arrow for the open fatigue link. The process–structure–property chain: the early links are modelled or measured, static properties are the easier case, and the route from structure to fatigue life remains open. Original diagram by the authors, created with AI assistance.

Why it is hard​

  • Many length scales. Cells (sub-micrometre), grains (tens of micrometres to millimetres), melt-track and layer features, and part-scale heterogeneity all matter, and tests on small coupons average over different scales from the part.
  • Position dependence. The same part can have different local cooling histories at a thin wall, a thick boss and an overhang, so properties vary within one part [1].
  • Testing and scatter. Fatigue and fracture tests need many specimens; defect-driven failures scatter widely.
  • Predicting the structure is only half the story. Even with perfectly known grains and defects, mapping them to fatigue life needs micromechanics that are not routinely validated. Data-driven approaches to PSP models are growing [15] but they share the data scarcity discussed in Chapter 5.

Open questions​

  1. Can microstructure be designed in space, for example fine and equiaxed where fatigue-critical and columnar where creep resistance matters, and delivered reliably on production machines? Site-specific control has been shown [6], but robustness is not established.
  2. What defect size, shape and position limits fatigue for each alloy and condition, and can that limit be set as an acceptance criterion?
  3. How should predictions of mechanical response from microstructure be validated? In the 2022 AM Bench macroscale tensile challenge at different orientations, blind predictions were compared against measured tensile behaviour of laser powder bed fusion material [16]. Such benchmarks are the proper yardstick.
  4. How stable is the as-built substructure in service at temperature? The cell-wall structures are thermally metastable [4].

Industry impact​

Industry has largely routed around the open PSP link with process control by recipe: freeze the machine, powder and parameters; qualify by testing (Chapter 12); then lock the recipe. This is demonstrated in production, but it means every change of machine, powder lot, or geometry class re-opens the question. The benefit of a closed PSP chain would be fewer tests per qualification, the ability to specify properties by location, and trustworthy lightweighting.

Why it matters​

The value of AM metals is not only geometry but performance: AM titanium implants, turbine parts and tooling are justified by properties. Fatigue in particular decides whether a printed part may be used in a load-bearing, safety-critical role.

In the real world​

  • Aerospace and medical producers commonly apply HIP and surface machining to fatigue-critical printed parts.
  • National benchmark programmes publish full measured data (feedstock, build, microstructure, tensile) so modellers can test their predictions against the same reference [16].
  • In research, alloys are now being designed around AM's rapid solidification rather than adapted from casting grades [10].

Key terms​

  • Epitaxial growth — new solid growing on existing crystals so that it continues their orientation.
  • Columnar grain — an elongated grain that grows along the direction of heat flow.
  • Texture — a non-random distribution of crystal orientations.
  • Anisotropy — dependence of a property on direction.
  • Cellular substructure — sub-micrometre cells formed by fast solidification, with dislocation-rich, solute-enriched walls.
  • Process–structure–property (PSP) chain — the causal link from process parameters to microstructure to properties.
  • Intrinsic heat treatment — heat treatment of earlier layers caused by the heat of later layers.
  • Fatigue life — number of load cycles before failure.
  • Wrought — metal processed by rolling, forging or extrusion, the usual baseline for comparison.
  • HIP — hot isostatic pressing, a heat-and-pressure treatment that closes internal pores.

Where this leads​

To predict any of this without printing hundreds of coupons, we need models that span scales: Simulation and integrated computational materials engineering.