A directly printed titanium alloy reached 1,501 megapascals of yield strength, 1,726 MPa of ultimate tensile strength and 7.5% uniform elongation. Machine learning helped locate the recipe and printing conditions.[1]
Yield strength marks the onset of permanent deformation; ultimate tensile strength is the maximum stress sustained in the pull test. Uniform elongation measures how far the sample stretches before deformation concentrates into a narrowing section.
Our calculation from the published stress–strain curves finds peak stress 49.7% above the Ti–6Al–4V reference, fabricated under identical processing conditions.[4][1]
Peak tensile stress in the published comparison
| Printed material | Peak stress |
|---|---|
| Commercially pure titanium | 485 MPaSource 4 |
| Ti–6Al–4V | 1,154 MPaSource 4 |
| Ti–7.5Al–3.3Cr | 1,727 MPaSource 4 |
The experimental starting point comprised 118 manufactured combinations of aluminium and chromium content, laser power and scanning speed. Treating them as 118 different chemical recipes would miss part of the experiment: the same mixture can be printed under several thermal conditions. The input library records both chemistry and processing beside each measured strength.[2]
XGBoost, which combines successive decision trees, screened 38,052 virtual combinations. These were predictions, not manufactured alloys.[1]
The virtual grid crossed three aluminium levels, 151 chromium levels, four laser powers and 21 scanning speeds. Multiplying those choices reproduces the reported 38,052 candidates. This is a search inside a defined experimental territory; the size of the list comes from recombining controlled variables.[5]
The useful boundaries inside the metal
Laser powder-bed printing melts successive layers of powder with a laser. Rapid cooling helps preserve the selected alloy’s bands, less than 10 nanometres thick, with different atomic stacking sequences. Microscopy and atomistic simulations link these boundaries to strength: they obstruct and divide moving dislocations, the defects through which metal deforms, while encouraging distributed slip and continued hardening.[1]
That combination addresses a familiar materials problem. Making deformation difficult can increase strength, but useful structural metal must also accommodate strain without concentrating it into one failing region. A boundary that interrupts motion yet helps distribute subsequent deformation can contribute to both properties.
Seven newly printed candidates tested the predictions. Their measured ultimate tensile strengths ranged from 1,625 to 1,726 MPa; six of the seven strength forecasts were within 5% of experiment, calculated relative to the measured values. The best observed alloy had been predicted at 1,676.15 MPa, rather than ranked as the strongest forecast in that follow-up set.[3]
Ductility supplied a separate piece of information. For the best alloy, the table’s uniform-elongation prediction was 4.4%, while the measured result was 7.5%. The successful output was therefore more than a high-scoring virtual recipe: printing and pulling the sample exposed a combination of properties the model had only partly anticipated.[3]
Sources & further reading
- Sub-10-nm stacking bands enable ultrastrong additively manufactured titanium Nature Materials · 25 Sep 2026 · Accessed 18 Oct 2026
- Supplementary Table 1: compositions, printing parameters and strength measurements for the 118-sample library Nature Materials / Bai et al. · 25 Sep 2026 · Accessed 18 Oct 2026
- Supplementary Table 3: predicted and experimentally measured properties of seven selected alloy-process combinations Nature Materials / Bai et al. · 25 Sep 2026 · Accessed 18 Oct 2026
- Source Data Fig. 3: mechanical properties of the printed titanium alloys Nature Materials / Bai et al. · 25 Sep 2026 · Accessed 18 Oct 2026
- Supplementary Table 2: predictions for 38,052 virtual alloy-process combinations Nature Materials / Bai et al. · 25 Sep 2026 · Accessed 18 Oct 2026