We studied nine treated (five Be-diffused yellow samples and two high-temperature (HT) treated pairs from pale-blue to blue colour) and 16 differently coloured synthetic corundum crystals grown by flame fusion (number n=8), Czochralski (n=1), and hydrothermal (n=7) methods. Synthetic samples were directly obtained from manufactures, whereas HT pairs were collected after colour change (colourless to blue) which was induced by heating experiments at 1700 °C in a chamber furnace in a non-controlled atmosphere (open corundum crucible in air). Be-diffused samples were treated in the same furnace at 1700 °C during 50 h in a closed crucible, and the samples were embedded in chrysoberyl (BeAl₂O₄) powder. Ruby from Karelia (Russia) from the collection of Johannes Gutenberg-Universität Mainz (JGU) was also analysed by SIMS in order to verify the values for the only known naturally occurring corundum strongly depleted in δ18O. Samples were cut to millimetre-wide wafers, arranged on adhesive tape and embedded into epoxy resin to minimize sample changes (Fig. 2), but we emphasize that cut gemstones can also be directly mounted in a SIMS sample holder (e.g. Giuliani et al., 2000; Wang et al., 2018). After sectioning and polishing with abrasives up to 1 µm diamond suspension, the mounts were ultrasonically cleaned with deionized water and ethanol.

To better constrain the causes of heterogeneous δ18O values for the two different Be-diffused materials studied here, we performed dedicated Be diffusion experiments under controlled conditions. For this, corundum samples originating from magmatic and metamorphic sources were selected since they have distinctly different oxygen isotopic compositions, with approximate δ18O values of +5.5 ‰ for magmatic and between +14 ‰ and +25 ‰ for metamorphic sources (Wong and Verdel, 2017). Two stones of magmatic (average δ18O = 6.45 ‰ and 6.28 ‰) and two stones of metamorphic (average δ18O = 13.28 ‰ and 15.80 ‰) origins were cut into two pieces, where one piece of each stone was further Be diffused by a gem manufacturer at 1700 °C over 50 h.

Before LA-ICP-MS trace element measurements, optical microscopic inspection was carried out in order to avoid areas with inclusions and to identify internal features such as colour zonation. LA-ICP-MS analysis of heat-treated, Be-diffused, and synthetic corundum was applied to determine their trace element composition by using an ESI NWR193 ArF Excimer Laser combined with an Agilent 7500ce quadrupole-ICP-MS at the Institut für Geowissenschaften, JGU, Germany. The samples were ablated with a spot size of 70 µm. The repetition rate was 10 Hz, and the energy density was approximately 3.3 Jcm-2 during the measurements. Every spot was analysed with 50 s warmup/background time, 30 s dwell time, and 20 s wash-out time. Trace elements of interest within the corundum samples included the following isotopes: ²⁴Mg, ⁴⁷Ti, ⁵¹V, ⁵²Cr, ⁵⁶Fe, ⁶⁰Ni, and ⁷¹Ga. Other masses including ⁶Li, ⁹Be, ²³Na, ²⁹Si, ³⁹K, ⁴³Ca, ⁵⁵Mn, ⁸⁸Sr, ⁹⁰Zr, ⁹³Nb, ¹³⁷Ba, ¹⁷⁹Hf, ¹⁸¹Ta, and ²⁰⁸Pb were added to monitor contamination with solid inclusions in time-resolved spectra but also to assess the impact of the Be-diffusing process as applied to some of the treated material. NIST SRM 610 glass was used as the primary reference material (Jochum et al., 2005, Jochum et al., 2011), and NIST SRM 612, USGS BCR-2G, USGS GSD-1G, and USGS GSE-1G glasses were used as quality control materials. Reference and quality control materials were measured after sets of 30 unknown corundum samples to monitor the accuracy and precision of calibration. The time-resolved signal spectra were processed in GLITTER 4.4.1 software (http://www.glitter-gemoc.com, last access: March 2026, Macquarie University, Sydney, Australia) using ²⁷Al as the internal standard, applying a theoretical value of Al₂O₃ at 100 wt % for the corundum samples and values given in the GeoReM database for the reference material. The measured concentrations for both QCM (quality-control material) agree for most trace elements of interest and were less than 15 % of the preferred values provided in the GeoReM database.

The oxygen isotope composition of the same corundum crystals analysed previously by LA-ICP-MS was determined using CAMECA IMS1280-HR and IMS 1300-HR3 ion microprobes at Heidelberg University (HIP) in Germany and at Curtin University (John de Laeter Center) in Australia, respectively. Before the SIMS measurements, BSE (back-scattered electron) images were taken using a LEO 440 scanning electron microscope coupled with an Oxford Instruments X-Max energy dispersive spectrometer in order to avoid areas with inclusions. The surface was then cleaned and coated with a ∼50–80 nm thick conductive Au layer. An ∼2 nA and 20 keV total impact energy Cs⁺ primary ion beam with a raster size of 10 µm (12 µm during pre-sputtering) was used for rim-to-rim measurements of corundum, avoiding direct proximity to the LA-ICP-MS craters. Sample charging was compensated for with a normal incidence electron gun, and negative secondary ions were accelerated to a kinetic energy of 10 keV. The ¹⁶O and ¹⁸O secondary ion beams were detected simultaneously in two Faraday cups with a mass resolving power of ∼2300 (m Δm⁻¹ at 10 % peak height). Prior to each analysis, the secondary beam was centred automatically in the field aperture (X and Y) and the entrance slit (X only). Including the time for beam centring, the analyses started after a total pre-sputtering time of 90 s, and each analysis had 20 cycles with 4 s integration time per cycle. The baseline of the FC amplifiers was determined with an integration time of 200 s or as a running average of 7×30 s integrations. The internal precision reported is the standard deviation of the mean value of the isotope ratios. The instrumental mass fractionation (IMF) was determined using HD-LR1, a synthetic laser ruby (δ18O=+18.40±0.14 ‰, 95 % confidence internal; Schmidt et al., 2024) placed on the same mount as the samples (all δ18O values relative to Vienna Standard Mean Ocean Water (SMOW) with 18O/16O=0.0020052; Baertschi, 1976). Bracketing analyses of HD-LR1 indicate a repeatability of the IMF of between 0.08 and 0.16 ‰ (1 SD) on individual mounts and 0.36 ‰ (1 SD) between different mounts. In the absence of a secondary corundum reference, reference zircon AS3 and 91500 were analysed under the same instrumental settings, and, when using 91500 as the primary reference (δ18O=9.86 ‰ ; Wiedenbeck et al., 2004), a value of +5.40±0.01 ‰ (n=7) was obtained, which agrees closely with the reported composition of +5.34 ‰ (Trail et al., 2007).

Trace elements commonly found in corundum (Be, Mg, Ti, Fe, Cr, V, Ga, Ni) and applied for the identification of synthetic corundum material from natural and/or treated counterparts (Muhlmeister et al., 1998) were determined in 86 spots; one spot affected by irregular ablation, likely due to the presence of solid inclusions, was excluded from further consideration (Tables 1–3). Five yellow-coloured Be-diffused corundum samples were homogeneously enriched in Fe (up to 3720 µgg-1) and also show a homogeneous distribution of Mg, V, Cr, and Ga with only minor fluctuations (Table 1). With the exception of one sample (no. 2 in Table 1), Ti abundances range from below the detection limit (bdl) to ∼100 µgg-1 likely due to ablation of micrometre-sized Ti-bearing minerals (ilmenite, rutile, etc.). Beryllium was also detected in these samples, reaching 13.7 µgg-1. As in the case of Be-diffused samples, two HT pairs (colourless and blue-coloured) show a homogeneous distribution of Mg, V, Cr, and Ga. For the same pairs, Fe and Ti are heterogeneously distributed with abundances of up to 2300 and 339 µgg-1, respectively. However, differences in trace element abundances between the colourless and HT-treated blue-colour parts of stone for a HT sample are insignificant.

Synthetic corundum samples are enriched by different chromophoric trace elements, depending on their colour. Accordingly, red- and orange-coloured varieties contain higher Cr values (max. 1.2 wt % in hydrothermally synthesized corundum) compared to other coloured or colourless samples. Chromium is mostly homogeneously distributed within flame fusion material, whereas abundances are heterogeneous within hydrothermally synthesized samples. Blue-coloured corundum samples yield higher Fe abundances (generally up to 536 µgg-1, with one anomalous spot at 1540 µgg-1 of Fe within a colourless sample, likely due to ablation of an inclusion) and Ti (up to 555 µgg-1) compared to the other coloured counterparts. Trace element variations indicate oscillatory zonation within hydrothermally synthesized samples, whereas such fluctuations are less obvious within flame fusion samples (Fig. 3). Magnesium and V are mostly homogeneously distributed within all studied synthetic materials, with minor variability. Gallium and Ni were also detected within synthetic sapphires. Although Ga was almost at the detection limit within flame fusion material, its content varies within the hydrothermally synthesized samples, reaching up to 63.2 µgg-1. Nickel was detected in two blue-coloured hydrothermally synthesized sapphires reaching 2507 µgg-1, but, for the rest of the hydrothermally synthesized material, it was below the detection limit. Czochralski-grown pink-coloured corundum is almost free of trace elements, except for Cr, which is present at abundances of 388–402 µgg-1.

In total, 450 points were analysed within the treated and synthetic corundum material, with all analyses showing identical secondary ion intensities as compared to reference corundum HD-LR1 (Tables 1–3). The results obtained on Be-diffused sapphires show both positive and negative oxygen isotopic compositions on the SMOW scale. Most Be-diffused sapphire samples (n=4) yielded values around zero, with slightly negative or positive δ18O variations from -0.90±0.13 ‰ to +0.58±0.11 ‰; only one sample deviates to higher δ18O values between +12.81±0.12 ‰ and +13.21±0.12 ‰.

Treated Be-diffused pieces of magmatic origin yielded δ18O averages of 6.42 ‰ (n=10) and 6.34 ‰ (n=10), whereas treated pieces of metamorphic origin averaged 13.47 ‰ (n=10) and 15.86 ‰ (n=10) (Fig. 5). Thus, δ18O values are indistinguishable between original and Be-diffused parts of these stones. The only exception is a yellow rim domain in a metamorphic Be-diffused crystal, where detailed SIMS mapping revealed a ∼0.5 ‰ increase in δ18O compared to the rest of the stone, whereas the untreated counterpart lacks this increase (Fig. 4).

The oxygen isotopic compositions of the HT-treated and HT-untreated pairs are similar: for the first pair, values range from +14.21±0.08 ‰ to +14.55±0.10 ‰, and, for the second pair, values range from +15.74±0.15 ‰ to 16.49±0.16 ‰. Importantly, the HT-treated corundum crystals lack any significant δ18O difference between the colourless and blue domains of the stone. One-way ANOVA tests performed on the two groups (original and HT-treated pieces) of the two stones analysed indicated p values of 0.4797 (samples 1 and 2, Table 1) and 0.0778 (samples 3 and 4, Table 1), which indicates that measured δ18O values for original and HT-treated pairs are indistinguishable.

In contrast to natural corundum (with and without Be-diffusing and HT treatment), synthetic corundum generally displays strongly negative δ18O values. This includes six hydrothermally synthesized samples with δ18O values ranging from -7.89±0.13 ‰ down to -14.54±0.13 ‰ and seven flame fusion samples with δ18O values between -6.73±0.13 ‰ and -17.46±0.13 ‰. The only exceptions are one red-coloured flame fusion sample and one Czochralski-grown sample: they yielded unusually high δ18O values ranging from +28.51±0.11 ‰ to +30.47±0.10 ‰.

Different trace elements (Mg, Ti, Fe, Cr, Ga, etc.) and their ratios including Ga/Mg, Fe/Ti, Fe/Mg, and Cr/Ga are commonly used as a proxy for corundum origins (e.g. Peucat et al., 2007; Sutherland et al., 2009). However, some of these classification proxies are known to produce incorrect results, where, for example, ruby and sapphire compositions are convoluted (e.g. Palke et al., 2018; Sorokina et al., 2019; Filina et al., 2019; Wong and Verdel, 2017). Currently, the most reliable discrimination is based on the FeO–Cr₂O₃–MgO–V₂O₃ vs. FeO+TiO2+ Ga₂O₃ diagram (Giuliani et al., 2014), which is applicable for both ruby and sapphire (Palke et al., 2018; Sorokina et al., 2019, etc.). This diagram is reproduced in Fig. 5 with new LA-ICP-MS data for treated and hydrothermally synthesized corundum of different colour varieties. Most of these data plot in the field of “metasomatic corundum” (Fig. 5). By contrast, ruby (red-colour varieties) data fall into the “ruby in marble” field. The reason why hydrothermally synthesized stones plot within the fields defined by naturally occurring corundum is likely due to the content of Ga in most of the hydrothermally synthesized corundum samples reaching up to 63.2 µgg-1 (Table 2). Although Ni is present as a chromophore mostly in synthetic stones grown by the hydrothermal method (Schmetzer and Peretti, 1999), this criterion is ambiguous as it is not always detectable in synthetic corundum and sometimes can even be found in natural blue sapphire (Sorokina et al., 2019). Therefore, solely relying on LA-ICP-MS trace element analysis for distinguishing between natural and synthetic corundum origins, especially when grown by the hydrothermal method, can lead to misidentifications.