A low-cost and high precision determination method of ⁸⁷Sr/⁸⁶Sr ratios for red wine using thermal ionization mass spectrometry without column separation

Abstract

Thermal ionization mass spectrometry is considered as the “gold standard” for obtaining precise ⁸⁷Sr/⁸⁶Sr isotopic ratios. However, the conventional TIMS technique is laborious because it requires the purification of Sr using expensive and complex resin column techniques to overcome severe ⁸⁷Rb isobaric interference on ⁸⁷Sr and inhibition on ionization of Sr originating from matrix elements. To overcome this issue, we have developed a hydrofluoric acid coprecipitation method specifically designed for red wine samples. This method effectively captures Sr while eliminating ⁸⁷Rb isobaric interference and removing the majority of matrix element inhibition for red wine samples. Notably, our method offers several advantages, including low cost, rapidity, simplicity, and high sample throughput. We have applied this method to analyze nine red wine samples sourced from China, Spain, Chile, Argentina, France, Italy, and Australia. Our method greatly simplified the sample preparation process for red wine samples without compromising the analytical accuracy and precision.

---

1. Introduction

The captivating taste and the enchanting aroma of red wines have been attracting people for thousands of years. Identifying geographical origins of red wine in specific region and manufacturers brand is crucial because the geographical origins of wine greatly influence its quality and price. The naturally occurring ⁸⁷Sr/⁸⁶Sr isotopic ratio combined with mineral elemental content analysis has proved to be a valuable tool for detecting trends in the soil-vegetation system and tracing the provenance of a variety of food objects.^1–3 Thermal ionization mass spectrometry (TIMS) and multiple-collector inductively coupled plasma mass spectrometry (MC-ICP-MS) are the main instruments widely used for the precise determination of the ⁸⁷Sr/⁸⁶Sr ratio for wine.^1,2,4–9

Usually, an internal precision of ≤0.003% (2SE) for the ⁸⁷Sr/⁸⁶Sr ratio is required in wine authentication studies.^1,2,4–9 A precise and accurate determination of the ⁸⁷Sr/⁸⁶Sr ratio by TIMS or MC-ICP-MS requires a separation of Sr from the sample matrix,^1,2,4–9 specifically from Rb, which can produce an isobaric interference of ⁸⁷Rb on ⁸⁷Sr. Compared to MC-ICP-MS, TIMS exhibits better sensitivity and greater tolerability for sample purity when measuring the Sr isotopic ratio.^10–12 First, regarding the subtraction of the ⁸⁷Rb isobaric interference on ⁸⁷Sr, the Rb/Sr ratio of the Sr analyte after purification should be lower than 0.015 when using MC-ICP-MS.¹³ In the case of TIMS, no influence was observed, even if the Rb/Sr ratio of the Sr analyte reached 8, such as loading mixture Sr (2 ng) + Rb (16 ng), because a small amount (<16 ng) of Rb can be easily burned off prior to measuring Sr.¹⁰ Second, as for matrix element Ca, Ca argide (⁴⁸Ca⁴⁰Ar and ⁴⁶Ca⁴⁰Ar) or Ca dimer species (⁴⁸Ca⁴⁰Ca and ⁴⁶Ca⁴⁰Ca) can significantly affect the Sr isotopic ratio, especially for samples with high Ca/Sr ratios.^14,15 In the case of TIMS, no complex polyatomic isobaric interferences except ⁸⁷Rb interference are found. This means the background of the mass spectrum of TIMS is simple. However, too much coexisting matrix elements (Ca and Mg) can severely inhibit the Sr signal and lead to poor sensitivity.^10,11 Third, TIMS has an excellent sensitivity for Sr. Even for the 200 pg sample size (NIST SRM-987), high precision of ⁸⁷Sr/⁸⁶Sr (<0.003%, 2SD, n = 8) can be easily achieved using a TIMS equipped with the conventional amplifiers (10¹¹ resistor).¹² This means the inhibition caused by matrix elements can be partially alleviated by diluting the Sr content of the analyte. In summary, TIMS exhibits great potential for samples with simple matrix compositions due to its excellent sensitivity and good tolerability with matrix elements.

Red wine is an alcoholic beverage, a product of yeast fermentation of natural sugars present in grape juice that consists of thousands of organic chemical compositions.² Fortunately, all organic materials in red wine can be easily digested using a mixture of HNO₃ and H₂O₂ before measurements by TIMS or MC-ICP-MS.^3–9 Generally speaking, after digesting organic matrix interferents, the mineral elemental composition of wine are relatively simple. The main elements that constitute approximately 99.5% of the total elemental composition of wine are K, Na, Ca, Mg, B and Mn. Notably, K alone accounts for more than 80% in the total elemental mass of wine. To the best of our knowledge, no comprehensive publication of average elemental content values for red wine has been published around the world so far.

Recently, Wu et al.,¹⁶ verified the geographical origin of red wines in China and those imported from seven main export countries (France, USA, Chile, South Africa, Australia, Italy, and Spain) around the world. They performed elemental analyses on large numbers of red wine bottles (600) and released the average values for 16 elemental contents, which can be considered as the background values of red wine elemental contents. Generally, the average elemental ratios of red wine¹⁶ exhibit a very high K/Sr (1480), moderate Mg/Sr (148), low Ca/Sr (85.5), Na/Sr (35.4) and very low Rb/Sr (2.26). These fundamental data¹⁶ indicate that it is possible to achieve high precision of the ⁸⁷Sr/⁸⁶Sr ratio for red wine if K and Rb can be removed prior to TIMS measurements.

Our previous work¹⁰ showed that good data (±0.005%, 2SE) were obtained for natural water samples with a moderate Ca/Sr ratio(<500) using a TIMS equipped with amplifiers with 10¹² resistors even after directly loading the sample on the filament without any sample purification. This means TIMS exhibits good tolerance for sample purity even if the analyte has imperfect purity. Coprecipitation is a classical method for separating matrix elements or enriching trace elements for natural water and geological samples. Hydrofluoric acid is a commonly used coprecipitation reagent known for its effective separation of alkali metals (Li, Na, K, and Rb) and alkaline earth metals (Ca, Mg, Ba, and Sr). However, the potential of hydrofluoric acid coprecipitation for food samples is completely neglected. In theory, HF coprecipitation can facilitate the separation between Rb–K and Sr for wine samples because wine has relatively simple matrix elemental compositions. Nevertheless, several crucial questions remain unanswered. It is unclear whether hydrofluoric acid co-precipitation causes fractionation of the ⁸⁷Sr/⁸⁶Sr ratio in food materials. If hydrofluoric acid co-precipitation doesn't cause fractionation of ⁸⁷Sr/⁸⁶Sr, how is the yield of Sr after co-precipitation? How many matrix elements and Rb can be removed by co-precipitation? What is the best Sr loading size due to the Ca–Mg matrix coexisting with Sr? How is the applicability of the method for red wine samples with different matrix compositions?

The aim of this study is to develop a simpler, rapid and low-cost method for analyzing Sr isotopes in red wine samples, eliminating the need for laborious and expensive Sr spec resin, Bio-Rad AG 50 or Dowex 50 resin column separation schemes. Here, we propose a straightforward hydrofluoric acid coprecipitation method that does not compromise the data quality of red wine samples. To verify the robustness and stability of the method for Sr analyses, nine red wine samples from China, Spain, France, Argentina, Italy, Australia, and Chile were analyzed in this study. Excellent precision of ⁸⁷Sr/⁸⁶Sr (<0.003%, 2RSE), most of them better than (<0.002%, 2RSE), is achieved for all red wine samples by using a TIMS equipped with conventional amplifiers (10¹¹ resistor). Hence, the proposed method significantly reduces the experimental cost and improves sample preparation efficiency.

2. Experimental section

2.1 Reagents and materials

Ultra-pure water (Milli-Q) with a resistivity of 18.2 MΩ cm⁻¹ was used. The used labware included 15.0 mL PFA vials with screw top lids (Savillex Corporation, USA). These vials were employed for sample digestion, solution collection and evaporation. H₂O₂ (30%, Ultrapure Reagent), HCl and HNO₃ acid were obtained from the China National Pharmaceutical Group Corporation. HCl and HNO₃ acid were further purified via sub-boiling distillation using Savillex Teflon distillers.

Eight dry-type and one half-dry-type wine samples from seven countries were bought from Chinese supermarkets and labeled W1-China, W2-China, W3-China, W4-Spain, W5-Argentina, W6-France, W7-Italy (half-dry-type), W8-Australia, and W9-Chile. A stock solution of 200 mg L⁻¹ of NIST SRM-987 Sr was used to monitor the Triton Plus TIMS.

2.2 Sample decomposition and hydrofluoric acid coprecipitation

All sample preparations were undertaken in a class – 1000 clean room. Approximately 3 mL of the wine sample was weighed. Fig. 1 shows the analytical flow chart of our hydrofluoric acid coprecipitation method. All wine samples were first evaporated to near dryness on a hot plate at 90 °C. Then, 3 mL of 14 mol L⁻¹ HNO₃ and 1 mL of 30% H₂O₂ were added to the vials to digest the wine. Wine samples were slowly evaporated to dryness at 90 °C again. Then, after addition of 5 mL 14 mol L⁻¹ HNO₃, the container was tightly capped and then heated on a hot plate at 140 °C to digest the sample for four hours at least. The vials were then opened and the resulting solution was evaporated to dryness at 160 °C. The wine residue was re-dissolved with 0.5 mL of 0.4 M HCl at 100 °C. Subsequently, 1 mL of 22 mol L⁻¹ HF acid was added to the closed vial to form a fluoride precipitate enriched with Sr–Ca–Mg and heated at 120 °C for 1.5 hours.

Fig. 1 Analytical flow chart of ⁸⁷Sr/⁸⁶Sr for red wine using TIMS after HF coprecipitation.

After the sample solution was cooled to room temperature, it was shaken and then transferred to a 2 mL centrifuge tube. The sample solution was centrifuged at 8000 rpm for 10 min. The supernatant contained the most (>90%) Rb, K, Na, Mn and B. Also, parts of Sr (20–40%), Ca (5–8%) and Mg (4–7%) can be detected in the supernatant. Most of Sr–Ca–Mg is enriched in the white fluoride precipitate. After opening the centrifuge tube and pipetting off the supernatant, 0.3 mL of 7 M HNO₃ was added to the centrifuge tube. The fluoride precipitate in the centrifuge tube was easily dissolved by using 0.3 mL of 7 M HNO₃ after shaking and waiting for 20–30 minutes. The total procedural blank of our method is 36 ± 5 pg (n = 3), mainly from HNO₃, H₂O₂ and HF acid during wine digestion. Finally, the sample solution was kept in a centrifuge tube. 0.5–2 μL of the sample solution is directly loaded onto the Re filament for TIMS measurements.

2.3 Elemental content measurements

Ten elements were determined in order to check and examine the stability of hydrofluoric acid coprecipitation for different red wine samples. Inductively coupled plasma optical emission spectroscopy (ICP-OES, IRIS Advantage, Thermo Fisher, Germany) was used to analyze eight major elements (K, Ca, Mg, Na, Al, B, Mn and Fe). Rb and Sr were measured by using an iCAP Q-ICP-MS (Thermo Fisher, Germany). The analytical uncertainty of the ten elements is better than 10%. Calibration standards with concentration levels ranging from 0.001 to 2000 mg L⁻¹ were used in order to calibrate the elemental abundances in diluted wine samples. To achieve low background noise and minimize memory effects, 1 minute rinsing with 3% v/v HNO₃ was performed.

All elemental content data for red wine samples are listed in Table 1.

Table 1 Ten elemental contents in red wine samples

Sample	Al (μg g^−1)	B (μg g^−1)	Ca (μg g^−1)	Fe (μg g^−1)	K (μg g^−1)	Mg (μg g^−1)	Mn (μg g^−1)	Na (μg g^−1)	Rb (μg g^−1)	Sr (μg g^−1)
W-1	0.575	5.19	49.5	1.34	913	96.9	1.17	21.0	0.812	1.76
W-2	0.775	4.44	48.9	3.15	1088	94.2	3.84	10.1	1.216	0.80
W-3	0.753	1.39	40.6	1.82	515	34.8	1.03	33.5	0.490	0.44
W-4	0.117	5.39	47.7	1.46	878	81.3	0.80	7.66	0.846	2.55
W-5	0.219	6.82	60.5	1.84	1044	80.2	1.08	30.4	0.718	1.17
W-6	0.289	1.98	36.6	1.66	538	46.6	0.92	45.8	1.261	0.48
W-7	1.12	3.57	75.9	2.83	854	54.1	1.03	17.3	0.996	0.57
W-8	0.295	6.05	47.9	2.16	646	114.5	2.06	29.5	0.487	2.47
W-9	0.265	4.37	48.4	2.77	913	76.7	1.56	7.31	3.33	0.90

---

2.4 TIMS measurements

The Sr measurements were performed on a Thermo Fisher Scientific Triton Plus TIMS at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). A single Re filament configuration was employed to determine the Sr isotope. The loading sample method was similar to that in our previous studies.^10,11 The loading blank was ∼1 pg for Sr.

The Sr isotopic data were acquired in the static collection mode, and the configuration of the Faraday cups is shown in Table 2. To eliminate all gain calibration errors, amplifier gains were calibrated at the start of each day. Before measuring Sr, a warm-up procedure (120 s at ∼1250 °C) was performed for all samples to eliminate any potential ⁸⁷Rb isobaric interference due to the huge temperature gap (∼700 °C) between the evaporation of Sr and Rb. Before the commencement of each analysis, a peak-center routine was run, and then the baseline was measured. The integration time per cycle was 4 seconds. The filament current was ramped up in 500 mA min⁻¹. In general, the optimized ionization temperature is 1450 ± 50 °C for Sr in TIMS measurements.^11,12

Table 2 Cup setting of Sr in a Triton Plus TIMS

Element	L2	L1	CC	H1	H2
Sr	^84Sr	^85Rb	^86Sr	^87Sr	^88Sr

---

Finally, for measuring Sr isotopes, the current of the filament was slowly increased until the ⁸⁸Sr signal reached 0.05 V. The beam of ⁸⁸Sr was roughly focused, and the filament was slowly heated to 1350–1450 °C to obtain a ⁸⁸Sr signal of ∼2.5 V. Sr data acquisition began when the signal intensity of ⁸⁸Sr reached approximately 2.8 V. The total measurement time lasts for approximately 20 min and comprises 8–10 blocks of 20 scans. The typical internal precision of the ⁸⁷Sr/⁸⁶Sr ratio was better than ±0.002% (2RSE). During Sr measurements, the ⁸⁵Rb signal was never detected, indicating no Rb isobaric interferences, and the ⁸⁵Rb/⁸⁶Sr ratios obtained during Sr analysis were ≤3 × 10⁻⁵. All ⁸⁷Sr/⁸⁶Sr ratio data were normalized to ⁸⁸Sr/⁸⁶Sr = 8.375209 (ref. 5–15 and 17–20) for the mass fractionation correction using the exponential law.

The NIST SRM987 standard was employed to verify the instrument stability during the analytical session. During the data collection period, the mean value of the NIST SRM987 was 0.710248 ± 16 (2SD, N = 6) for ⁸⁷Sr/⁸⁶Sr, in good agreement with the reported values of 0.710250 ± 20.^{5–15,17–20}

3. Results and discussion

3.1 Removal capability of matrix elements, Rb isobaric interference and Sr recovery after HF coprecipitation

Good sample purity is the prerequisite to obtain high intensity, a stable signal and excellent analytical precision of the ⁸⁷Sr/⁸⁶Sr ratio. The major matrix elements K, Ca, Na, Mg and B for the wine sample accounts for ∼99% of the total dissolved mineral elements (TDME) in red wine. Especially, K alone accounts for ∼84% of TDME for most red wines. This means eliminating K removes the highest matrix interference. Recently, the HF coprecipitation method has been proved to be an auxiliary way to separation alkali metal elements (Li, Na, K, and Rb) and alkaline earth elements (Ca, Mg, Ba, and Sr) for silicate materials with a high Rb/Sr ratio.¹⁹ The removal rate of Rb for silicate materials is unstable and shows big variation (20–80%) depending on the actual matrix composition of silicate materials.¹⁹ As aforementioned, the elemental composition of wine is relatively simple. In theory, the removal rate of Rb for wine is stable and higher than that of silicate materials.

However, it is unknown whether HF coprecipitation can be suitable for separation of alkali metal elements and alkaline earth elements for food materials. We give a modification based on our recent HF acid coprecipitation method¹⁹ for silicate samples. Following the aforementioned HF acid coprecipitation method, the yield of Sr and the removal capability of Rb and matrix elements are carefully checked for all wine samples. After coprecipitation and centrifugation at 8000 rpm for 10 min, all suspensions were carefully picked out and evaporated to dryness on a hot plate. Then, the sample was dissolved using 6 mL of 3 M HNO₃. Afterwards, the major elemental contents and Rb–Sr contents of the suspension were measured by ICP-OES and Q-ICP-MS following the aforementioned method description, respectively. Table 3 lists the contents of major elements and Rb–Sr of the suspension after coprecipitation for red wine samples. As shown in Table 4, the removal rate of an element was expressed and calculated as the ratio of the content of the element in the supernatant to the content of the element in the wine samples.

Table 3 Ten elemental contents of the suspension after coprecipitation in red wine samples

Sample	Al (μg g^−1)	B (μg g^−1)	Ca (μg g^−1)	Fe (μg g^−1)	K (μg g^−1)	Mg (μg g^−1)	Mn (μg g^−1)	Na (μg g^−1)	Rb (μg g^−1)	Sr (μg g^−1)
W-1C	0.033	4.34	4.01	0.377	911	9.52	1.01	20.35	0.804	0.62
W-2C	0.039	3.92	2.60	1.08	1083	5.38	3.49	9.78	1.203	0.27
W-3C	0.001	1.23	2.43	0.856	506	2.43	0.97	31.64	0.478	0.17
W-4C	0.001	4.62	2.96	0.435	868	6.17	0.70	7.15	0.834	0.94
W-5C	0.001	6.12	2.35	0.596	1005	3.16	0.99	28.08	0.699	0.25
W-6C	0.001	1.76	2.16	0.747	522	2.03	0.86	42.81	1.237	0.19
W-7C	0.031	3.12	2.24	1.08	814	2.95	0.94	15.58	0.964	0.11
W-8C	0.001	5.11	2.97	0.528	597	8.10	1.65	26.20	0.454	0.86
W-9C	0.001	3.89	2.22	1.00	898	3.85	1.40	6.87	3.15	0.30

---

Table 4 Removal rates of the ten elements after the coprecipitation reaction^a

Sample	Al (%)	B (%)	Ca (%)	Fe (%)	K (%)	Mg (%)	Mn (%)	Na (%)	Rb (%)	Sr (%)
a The removal rate of an element was expressed as the ratio of the amount of the element in the supernatant to the total amount of the element in the sample.
W-1	5.78	83.7	8.1	28.1	99.7	9.8	86.5	97.1	99.0	35.4
W-2	5.02	88.3	5.3	34.3	99.5	5.7	91.1	96.7	98.9	33.6
W-3	0.13	88.3	6.0	47.0	98.2	7.0	94.3	94.5	97.5	38.3
W-4	0.86	85.7	6.2	29.9	98.8	7.6	88.1	93.3	98.6	37.0
W-5	0.46	89.7	3.9	32.5	96.3	3.9	91.7	92.3	97.4	21.7
W-6	0.35	88.9	5.9	44.9	97.0	4.4	94.3	93.4	98.1	39.5
W-7	2.79	87.5	2.9	38.3	95.4	5.5	90.7	89.9	96.8	19.6
W-8	0.34	84.5	6.2	24.5	92.4	7.1	80.4	88.9	93.2	34.9
W-9	0.38	88.9	4.6	36.1	98.4	5.0	89.7	94.1	94.7	33.2

---

As shown in Table 4 and Fig. 2, most of K (92.4–99.7%), Na (88.9–97.1%), B (83.7–89.7%), Mn (80.4–94.3%) and Rb (93.2–99.0%) can be removed. In this process, part of Sr (19.6–39.5%), Ca (2.9–8.1%) and Mg (3.9–9.8%) are lost. Most of Sr (60.5–80.4%), Ca (91.9–97.1%), Mg (91.2–96.1%) and Al (94.2–99.9%) are found in the coprecipitation phase. The yield of Sr in the coprecipitation phase is imperfect which may be attributed to the low elemental content of Mg, Ca and Sr in wine, which leads to incomplete coprecipitation for Sr. As shown in Table 1, the variation range of Sr content is from 0.45 to 2.55 ppm. Considering that the conventional consumption of wine is 3 mL, the final captured amounts of Sr are thus approximately 799–4591 ng, and the yield of Sr is 60% in the coprecipitation phase. This means sufficient amounts of Sr can be obtained even though the yield of Sr is imperfect because the typical loading size of Sr is 2–4 ng for samples with imperfect purity.¹⁰ Low residual Rb (1–7%) is also found in the coprecipitation phase. The Rb/Sr ratio variation in the precipitation phase is 0.007 to 0.29 in this study. This means a minor amount of Rb (<1 ng) still coexisted with Sr in most cases when loading 10 ng Sr on the Re filament. Considering the extreme situation of Rb isobaric interference, the highest coexisting Rb amount is ∼3.6 ng when loading 12 ng Sr for W-9 in this study. No Rb isobaric interference is detected for all samples in this study after a pre-warming procedure. As for TIMS, residual Rb in the Sr fraction can be easily burned off by a preheating procedure before Sr data acquisition since the ionization temperature of Rb (600–700 °C) and Sr (1300–1450 °C) has a huge gap.¹⁰ Li et al.,^10,11 gave a detailed evaluation on the Rb interference on Sr for a suite of the mixture Rb + Sr. This study¹⁰ indicated that no Rb signal after filament warming was detected even if 16 ng of Rb was loaded on a single Re filament.

Fig. 2 Removal rates of 10 elements after the coprecipitation reaction for all red wine samples.

For a radiogenic Sr isotope system, an internal correction method is usually employed to correct the whole fractionation which existed in chemical separation and mass spectrometry measurements. Precise and good reproducibility of Sr data always can be obtained after internal correction using the ratio ⁸⁸Sr/⁸⁶Sr = 8.375209.^{5–15,17–20} We compared the data using HF coprecipitation and Sr spec resin column separation. As shown in Table 5 and Fig. 3, the comparison of data between HF coprecipitation and the conventional method indicates no detected Sr isotope fractionation during the HF coprecipitation step. Therefore, imperfect yield of Sr during HF coprecipitation doesn't give rise to a bias of ⁸⁷Sr/⁸⁶Sr ratios within analytical error (±0.003%, 2RSE).

Table 5 Comparison of the Sr ratio of red wine using TIMS between fluoride coprecipitation and Sr spec column purification^a

After HF coprecipitation				After purification
Sample no.	^87Sr/^86Sr	2SE	Load (ng)	Sample no.	^87Sr/^86Sr	2SE
a The fluoride sample containing 3 mL digested red wine is dissolved using 0.3 mL of 7 M HNO3. W-X-0.5, W-X-1 and W-X-2 represent loading 0.5 μL, 1 μL and 2 μL sample solution, respectively. The load represents the Sr loading size (ng) which is corresponding to 0.5 μL, 1 μL and 2 μL sample solution, respectively.
W-1-0.5	0.712995	0.000015	5.7
W-1-1	0.712975	0.000009	11.4	W-1-p1	0.712983	0.000008
W-1-2	0.712968	0.000009	22.8	W-1-p2	0.712984	0.000008
Mean	0.712979			Mean	0.712984
W-2-0.5	0.709437	0.000011	2.7
W-2-1	0.709425	0.000011	5.3	W-2-p1	0.709411	0.000008
W-2-2	0.709422	0.000012	10.7	W-2-p2	0.709422	0.000009
Mean	0.709425			Mean	0.709416
W-3-0.5	0.709940	0.000020	1.4
W-3-1	0.709928	0.000016	2.7	W-3-p1	0.709916	0.000008
W-3-2	0.709949	0.000024	5.5	W-3-p2	0.709913	0.000008
Mean	0.709939			Mean	0.709915
W-4-0.5	0.709295	0.000012	8.0
W-4-1	0.709279	0.000009	16.1	W-4-p1	0.709293	0.000009
W-4-2	0.709317	0.000010	32.2	W-4-p2	0.709278	0.000010
Mean	0.709297			Mean	0.709286
W-5-0.5	0.707904	0.000015	4.6
W-5-1	0.707929	0.000011	9.2	W-5-p1	0.707930	0.000008
W-5-2	0.707917	0.000014	18.4	W-5-p2	0.707920	0.000008
Mean	0.707917			Mean	0.707925
W-6-0.5	0.706314	0.000016	1.5
W-6-1	0.706286	0.000015	2.9	W-6-p1	0.706332	0.000009
W-6-2	0.706322	0.000014	5.9	W-6-p2	0.706309	0.000009
Mean	0.706307			Mean	0.706321
W-7-0.5	0.708349	0.000027	2.3
W-7-1	0.708369	0.000020	4.5	W-7-p1	0.708329	0.000008
W-7-2	0.708342	0.000044	9.1	W-7-p2	0.708320	0.000008
Mean	0.708353			Mean	0.708325
W-8-0.5	0.710300	0.000013	8.1
W-8-1	0.710331	0.000009	16.1	W-8-p1	0.710329	0.000010
W-8-2	0.710322	0.000009	32.2	W-8-p2	0.710301	0.000008
Mean	0.710318			Mean	0.710315
W-9-0.5	0.705056	0.000013	3.0
W-9-1	0.705030	0.000013	6.0	W-9-p1	0.705062	0.000008
W-9-2	0.705075	0.000015	12.1	W-9-p2	0.705048	0.000008
Mean	0.705054			Mean	0.705055

---

Fig. 3 Comparison of ⁸⁷Sr/⁸⁶Sr values of red wine sample fluoride coprecipitation and Sr spec resin column purification. Relative deviation (%) = (the mean value of ⁸⁷Sr/⁸⁶Sr_{coprecipitation} − ⁸⁷Sr/⁸⁶Sr_puried)/⁸⁷Sr/⁸⁶Sr_puried × 100%.

3.2 Accuracy and precision

The robustness of this method was assessed by performing analyses of nine red wine samples that covered a wide range of matrix compositions (K, Ca, Na and Mg), and Rb and Sr concentrations. As shown in Table 5, the typical internal precision of ⁸⁷Sr/⁸⁶Sr data for this suite of red wine samples except for W-7 is better than 0.002% (2SE) when loading 1 μL sample solution. The precision of ⁸⁷Sr/⁸⁶Sr for W-7 is decreased with increasing loading size. This is because W-7 has a relatively high Ca/Sr and Mg/Sr ratio and thus a higher loading size causes stronger matrix inhibition for the Sr signal.

To further verify the reliability of the HF acid coprecipitation method, all red wine samples, after digestion using HNO₃ and H₂O₂ following the aforementioned description, were also purified using the Sr spec resin.²⁰ The purification procedure is shown in Table 6 and gave a small modification based on our previous study.²⁰ The procedural blank was about 48 ± 9 pg (n = 3) for Sr. Amounts of Sr processed were 1.32–7.41 μg for this suite of red wine in the case of 3 mL digestion, and the contribution of analytical contamination to the separated analyte was thus negligible. BCR-2 (∼30 mg) basalt reference material was also digested^11,12,17 and employed to monitor the purification procedure and TIMS measurement. The ⁸⁷Sr/⁸⁶Sr value of BCR-2 in this study is 0.705018 ± 0.000013 and 0.705025 ± 0.000012 which were in good agreement with reported values for BCR-2.^{10–13,17–20} All purified red wine samples were measured using the same TIMS instrument. The typical intensity of ⁸⁸Sr was approximately 5000–7500 mV for purified samples. As indicated in Table 5 and Fig. 3, the results obtained with column chemical separation and HF acid coprecipitation clearly indicate good agreement within analytical error. In general, the data accuracy and precision of the proposed method are satisfactory and completely fit the demands of authentication tracing in red wine samples.

Table 6 Purification procedures for Sr separation using a small Sr spec column^a

Procedure	Eluting reagent	Eluting volume (mL)
a The Sr spec resin polyethylene column was 2.5 cm long with a 3 mm i.d. and 2 mL reservoir, packed with 0.15 mL of Sr spec resin.
Cleaning column	Ultrapure water	6 (1.5 × 4)
Cleaning column	7 M HNO3	4.5 (1.5 × 3)
Loading sample	7 M HNO3	0.5
Rinsing	7 M HNO3	3 (1 × 3)
Eluting Sr	Ultrapure water	3 (1.5 × 2)

---

3.3 The optimization Sr loading size for the coprecipitation method

As discussed above, the HF acid coprecipitation method provides a fast and good separation between Sr–Mg–Ca and K–Na–Rb. In contrast to complex silicate samples,¹⁹ the removal rate of K–Na–Rb for red wine is much better especially for Rb and K. Most of Ca and Mg generally coexisted with Sr. Previous studies^10–12 demonstrated that Ca and Mg have strong inhibitions on the Sr signal especially for Ca. Fortunately, TIMS has excellent sensitivity for Sr measurements.¹² Hence, we can dilute sample content and load relatively small size Sr on the Re filament to reduce the adverse effect originated from Ca and Mg.

Usually, for most Sr isotope applications in wine authenticity, the basic major and trace elemental content data are usually indispensable and measured preferentially by ICP-OES and ICP-MS. Thus, calculating Sr loading size is easy before TIMS measurements. In this study, the variation range of the capture rate of Sr in the coprecipitation phase is 61.7–80.4% which yields an average value of 67.4 ± 7.1% (n = 9). In future studies, we thus recommend calculating the final capture amounts of Sr according to our average value (∼70%) rather than checking every real wine sample that will lead to more additional experimental ICP-MS studies. This is because the typical loading size of Sr during the TIMS measurement is generally wide (1.5–6 ng). Hence, this has no significant influence on the final TIMS measurement even if the actual capture rate of Sr in the coprecipitation phase is deviated by around 20–30% compared to the true value.

The loading blank in our previous studies^10–12 is 0.5–1.5 pg. Hence, the minimum Sr load should be higher than 1.5 ng in order to obtain highly accurate data. According to the statistical data published by Wu et al.,¹⁶ for 600 bottles of red wine from China and the main export countries around the world, the average value of Ca/Sr and Mg/Sr is 85.5 and 148.3, respectively. This means the corresponding amount of Ca and Mg is ∼257 ng and ∼445 ng for most real red wine samples when loading 3 ng of Sr. In this study, all red wine samples were measured using a conventional amplifier equipped with a 10¹¹ resistor. The internal precision of ⁸⁷Sr/⁸⁶Sr is generally better than 0.002% (2RSE) when loading 1 μL of dissolved coprecipitation solution containing 1.4–32 ng of Sr. For wine samples with low Ca/Sr and Mg/Sr, 10 ng load of Sr is more suitable to easily obtain excellent precision of ⁸⁷Sr/⁸⁶Sr (<0.002%, 2RSE). For wine samples with high Ca/Sr and Mg/Sr, the load of Sr is recommended as 1.5–2 ng. Too much load of Sr will bring more Ca and Mg that leads to poor intensity and analytical precision of Sr. Generally, 3–6 ng load of Sr is preferential for most red wine samples.

In order to obtain excellent precision (<0.002%, 2RSE), employing a special amplifier with a 10¹² resistor is necessary for wine samples with high ratios (>200) of Ca/Sr and Mg/Sr. We also try to determine W-7, which has the highest ratios of Ca/Sr and Mg/Sr among this suite of red wine samples, using an amplifier with a 10¹² resistor following our previous method.¹⁰ The ratio of ⁸⁷Sr/⁸⁶Sr for W-7 using a 10¹² resistor is 0.708343 ± 0.000020 when loading 0.5 μL containing 1.4 ng of Sr. Clearly, in contrast to the data obtained by using an amplifier with a 10¹¹ resistor (Table 5), the internal precision can be further improved from ±0.000027 to ±0.000020 by using an amplifier with a 10¹² resistor.

3.4 Merits and limitations

The presented method has significant merits compared with previous methods.^4–9 It is low-cost, simple to operate, and has high sample-throughput. All traditional methods need chromatographic column separation based on the Sr spec resin, Bio-Rad AG 50 or Dowex 50 cation resin. All resin column schemes need complete pre-cleaning using HCl or HNO₃ acid before separation of Sr from wine since a large number of hazardous chemical reagents are consumed which is unfriendly for environment. In addition, chromatographic column schemes are generally complex including many eluting steps. Furthermore, after eluting the Sr fraction, the large volume of sample solution needs to be evaporated to dryness. Therefore, traditional resin column techniques are complex, time-consuming, and expensive although excellent sample purity can be achieved.

In contrast, our method circumvents the use of resin entirely, thereby eliminating the need for resin column pre-washing and the consumption of hazardous chemical reagents. The coprecipitation reaction employed in our method is rapid, taking only 90 minutes. The reaction generates only a small volume of hazardous reagent (<1.5 mL of HF acid), which is relatively friendly for the environment. After Sr–Mg–Ca fluoride was cooled to room temperature and the centrifugation process, the suspension was discarded. A small volume (0.3 mL) of 7 M HNO₃ is employed to dissolve Sr–Mg–Ca fluoride, which is quick (<20 min) and straightforward. The dissolved Sr–Mg–Ca fluoride solution can be directly loaded onto the Re filament for TIMS analysis. As a result, the cost of sample preparation is cut down ∼95% at least due to the absence of resin and the minimal amount of reagents used. Furthermore, the whole sample preparation process takes only 2.5 hours, and sample evaporation is not required, resulting in a two-fold improvement in sample preparation efficiency. Importantly, the method minimizes the potential contamination during complicated chromatographic column techniques. Although this method demonstrated low-cost and high efficiency, the coexisting Ca–Mg component still inhibited the Sr signal intensity. This effect is particularly noticeable in samples with high ratios of Ca/Sr and Mg/Sr, and the precision of ⁸⁷Sr/⁸⁶Sr is deteriorated even using the optimized loading size (3–6 ng of Sr). Therefore, we recommend employing a special amplifier with a 10¹² resistor to perform measurement for samples with high ratios (>200) of Ca/Sr and Mg/Sr. As for wine samples with extremely high ratios (>400) of Ca/Sr and Mg/Sr, the traditional resin column purification techniques^4–9 are still indispensable.

In the future, the method may have great application potential for other food materials with moderate ratios (<200) of Ca/Sr and Mg/Sr, such as spirit, beer, and fruit juice. However, further research is needed to evaluate these analytes.

4. Conclusions

High precision ⁸⁷Sr/⁸⁶Sr ratios in red wine samples can be obtained using TIMS after HF acid coprecipitation. Our data demonstrate that there is no ⁸⁷Rb isobaric interference on ⁸⁷Sr after a warm-up procedure for wine samples. The applicability of this method was rigorously examined using a suite of red wine samples from seven countries, and all data exhibited excellent precision for ⁸⁷Sr/⁸⁶Sr. Compared to traditional methods for red wine,^4–9 our method significantly reduces experimental costs and laboratory time without compromising the accuracy of the final results. Sample throughput for Sr isotope measurements by TIMS is thus significantly improved. The cost of sample preparation is almost negligible because no expensive resins are used. Furthermore, the time for sample preparation is reduced by at least 50%. Because of its low-cost and rapidity, the newly developed analytical method may be applied to not only red wine but a range of food samples with moderate ratios of Ca/Sr and Mg/Sr.

Author contributions

Chaofeng Li: conceptualization, methodology, validation, investigation, data creation, writing-original draft, writing-review & editing. Xuance Wang: writing-review & editing. Zhuyin Chu: writing-review & editing, funding acquisition. Peng Peng: funding acquisition.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We thank Dr Yanhong Liu for her help with the ICP-OES and ICP-Q-MS analyses. We are grateful to the editor for patience and Dr Ryan Ickert and one anonymous reviewer for critical and constructive comments that greatly improved this manuscript. This work was jointly supported by the Instrument Function Developing Project of the Chinese Academy of Sciences (No. GNKF202103) and by the National Natural Science Foundation of China (No. 42073050 and 42125206).

References

1. I. Coelho, I. Castanheira, J. M. Bordado, O. Donard and J. A. L. Silva, TrAC, Trends Anal. Chem., 2017, 90, 45–61.
2. P. Pohl, TrAC, Trends Anal. Chem., 2007, 26, 941–949.
3. E. I. Geană, C. Sandru, V. Stanciu and R. E. Ionete, Food Anal. Methods, 2017, 10, 63–73.
4. M. Barbaste, K. Robinson, S. Guilfoyle, B. Medina and R. Lobinski, J. Anal. At. Spectrom., 2002, 17, 135–137.
5. E. N. Epova, S. Bérail, F. Séby, V. Vacchina, G. Bareille, B. Médina, L. Sarthou and O. F. X. Donard, Food Chem., 2019, 294, 35–45.
6. C. Durante, L. Bertacchini, M. Cocchi, D. Manzini, A. Marchetti, M. C. Rossi, S. Sighinolfi and L. Tassi, Food Chem., 2018, 255, 139–146.
7. V. Vinciguerra, R. Stevenson, K. Pedneault, A. Poirier, J. Hélie and D. Widory, Food Chem., 2017, 210, 121–128.
8. S. Marchionni, A. Buccianti, A. Bollati, E. Braschi, F. Cifelli and P. Molin, Food Chem., 2016, 190, 777–785.
9. R. Cellier, S. Berail, J. Barre, E. Epova, F. Claverie, A. Ronzani, S. Milcent, P. Ors and O. F. X. Donard, Talanta, 2021, 234, 122433–122445.
10. C. F. Li, J. H. Guo, Z. Y. Chu, L. J. Feng and X. C. Wang, Anal. Chem., 2015, 87, 7426–7432.
11. C. F. Li, Z. Y. Chu, X. C. Wang, J. H. Guo and S. A. Wilde, Talanta, 2021, 233, 122537–122544.
12. C. F. Li, Z. Y. Chu, X. C. Wang, J. H. Guo and S. A. Wilde, Anal. Chem., 2019, 91, 7288–7294.
13. Y. H. Yang, F. Y. Wu, L. W. Xie, J. H. Yang and Y. B. Zhang, Spectrochim. Acta, Part B, 2011, 66, 656–660.
14. T. Waight, J. Baker and D. Peate, Int. J. Mass Spectrom., 2002, 221, 229–244.
15. J. Woodhead, S. Swearer, J. Hergta and R. Maas, J. Anal. At. Spectrom., 2005, 20, 22–27.
16. H. Wu, L. Tian, B. Chen, B. H. Jin, B. Tian, L. Q. Xie, K. M. Rogers and G. H. Lin, Food Chem., 2019, 301, 125137–125145.
17. C. F. Li, X. C. Wang, J. H. Guo, Z. Y. Chu and L. J. Feng, J. Anal. At. Spectrom., 2016, 31, 1150–1159.
18. C. F. Li, X. H. Li, Q. L. Li, J. H. Guo, X. H. Li and Y. H. Yang, Anal. Chim. Acta, 2012, 727, 54–60.
19. W. G. Liu, S. Wei, J. Zhang, C. Ao, F. T. Liu, B. Cai, H. Y. Zhou, J. L. Yang and C. F. Li, J. Anal. At. Spectrom., 2020, 35, 953–961.
20. C. F. Li, X. H. Li, J. H. Guo, X. H. Li, H. K. Li and G. Z. Li, Geochimica, 2011, 40, 399–406 , in Chinese with English abstract..

---