A rapid method for detecting trace Rb and Cs in high-salinity brines by ICP-MS equipped with an all-matrix sampling device

Abstract

This study presents a rapid and highly sensitive method for the determination of trace rubidium (Rb) and cesium (Cs) in high-salinity brines using inductively coupled plasma mass spectrometry (ICP-MS) equipped with an all-matrix sampling (AMS) device. The AMS system achieves online gas dilution by vertically introducing argon gas into the brine sample flow, effectively reducing the severe matrix suppression effect caused by 35 g·L⁻¹ salinity to an intermediate level. Experimental results demonstrated that the signal suppression induced by coexisting cations (K⁺, Na⁺, Ca²⁺, Mg²⁺) in actual brine samples was minimal (< 1.5%), thereby eliminating the need for conventional matrix matching or standard addition approaches. Accurate quantification was achieved through a straightforward calibration process based on standard curves (Rb: 5–400 μg·L⁻¹; Cs: 5–400 μg·L⁻¹; R² > 0.999), enhanced by dynamic internal standard correction using yttrium (Y) and rhodium (Rh), along with optimized instrument parameters (RF power and nebulizer gas flow rate). The method demonstrates excellent limits of detection (LOD: 0.039 μg·L⁻¹ for Rb; 0.005 μg·L⁻¹ for Cs), high precision (RSD < 5%), and acceptable recovery rates (85%–108%). The accuracy was further validated through comparison with AAS standard addition for high-concentration samples (> 200 μg·L⁻¹), yielding consistent recoveries (98.6% –114%) and inter-method deviations ≤ 12.2%. Additionally, the simplification of the sample pretreatment procedure—from a traditional multi-step dilution to a single-step dilution without acid-washed containers—enhances analytical efficiency by over 70%. This approach provides a robust, sensitive, and operationally efficient solution for the analysis of extreme high-salinity environmental samples.

Introduction

Rubidium (Rb) and cesium (Cs), prized for their distinctive chemical properties, play crucial roles in high-tech industries^1–4. Salt lake brines represent a significant source of Rb and Cs^5,6, however, their low concentrations^7,8 coupled with high salinity levels (ranging from 50 g·L⁻¹ to approximately 400 g·L⁻¹ or even more)^9,10 present substantial challenges for precise determination, which is critical for resource evaluation and extraction. Rb and Cs are typically quantified using instrumental analytical methods, such as atomic absorption spectrometry (AAS), inductively coupled plasma optical emission spectrometry (ICP-OES), ion chromatography (IC), and inductively coupled plasma mass spectrometry (ICP-MS). Most of these methods require salinity levels below 2 g L⁻¹, except for AAS, which can tolerate salinity levels up to 10 g·L⁻¹. For brines with salinities around 400 g·L⁻¹, extensive dilution by at least a two-step 200-fold procedure, or even more, is necessary for accurate determination, otherwise, insufficient dilution may lead to instrument clogging. However, such extensive dilution further reduces the already low concentrations of Rb and Cs, making their detection challenging due to the relatively high detection limits (~ 10⁻⁴ g·L⁻¹) of AAS, ICP-OES and IC. Consequently, studies employing these techniques for Rb and Cs determination in brines have often reported concentrations higher than actual values, and the salinity levels examined were generally low (< 160 g·L⁻¹)^11–17. In addition to extensive dilution, complex pre-treatment processes are required for the determination of Rb and Cs in high-salinity samples when using these instruments. For instance, potassium ion ionization buffers are added to suppress excessive ionization of Rb and Cs in AAS¹², while matrix-matching between standard solutions and samples is necessary for ICP-OES¹⁴. Although inductively coupled plasma mass spectrometry (ICP-MS) has a significant advantage in terms of low detection limits (~ 10⁻⁹ g·L⁻¹), there are still limitations when analyzing high-salinity brine samples. Firstly, the high-dilution pretreatment methods commonly adopted in existing studies (e.g., those described for the analysis of low-salinity brines and seawater^18,19) may result in the concentration of the target analyte approaching or falling below the limit of detection (LOD), thereby introducing significant quantitative errors. Secondly, high-concentration coexisting ions in complex matrices may exert an unclear matrix interference effect on the ICP-MS detection of Rb and Cs. This potential influencing factor requires systematic investigation.

The aim of this study is to establish and validate a method for the precise determination of trace Rb and Cs in high-salinity brines using ICP-MS equipped with an all matrix sampling (AMS) device, minimize offline dilution steps, simplify the sample preparation process, and investigate the influence of the matrix and coexisting ions on the determination.

Experimental section

Instruments

The study employed an ultrapure water meter (Millipore, US), an electronic balance (Mettler Toledo, CH), an Flame Atomic Absorption Spectrometry (AAS) and an ICP-MS system (Perkin Elmer, US) equipped with an all-matrix sampling (AMS) device. The AMS device, illustrated in Fig. 1, is designed with a salt tolerance of approximately 35 g·L⁻¹ and incorporates an argon-based gas dilution system. The red arrow indicates the path of the dilution gas, which is perpendicular to the sample solution flow. An independent argon channel regulates the dilution gas flow rate. The dilution effect intensifies with increasing flow rate, and the dilution gas exerts a more pronounced dilution impact on the matrix compared to the target element. This phenomenon occurs because, when the sample is diluted to 35 g·L⁻¹, the matrix-to-target-element content ratio ranges from 2×10⁴:1 to 2×10⁷:1 (calculated using the maximum sodium matrix concentration from Table 5 and the maximum rubidium/minimum cesium concentrations from Table 6). The substantial difference in these proportions results in significantly greater dilution of the matrix relative to the target element under identical dilution flow conditions, due to its higher absolute abundance in the aerosol. In addition, this phenomenon may also be attributed to the physical and chemical properties of the high-concentration matrix, such as viscosity and surface tension, which contribute to the formation of larger aerosol particles that exhibit reduced evaporation efficiency in inductively coupled plasma (ICP)^20,21. Upon activation of the dilution gas, to prevent excessive dispersion of the sample aerosol and maintain plasma stability, the nebulizer gas flow rate was reduced to compensate for the added dilution gas flow. This ensured that the combined flow rate (nebulizer + dilution gas) remained within a controlled range. The ratio of the two gases was optimized to balance dilution efficiency, sensitivity, and matrix interference mitigation.

Fig. 1.

Schematic diagram of all-matrix sampling (AMS) device.

Table 5.

Composition of actual brine samples.

Sample name	Sample number	Mass concentration (g·L−1)
		Na+	K+	Ca2+	Mg2+	Li+	B	Sr2+
Mahai	1	86.60	1.69	5.30	6.29	0.0090	0.024	0.10
Balun Mahai Sea	2	82.68	7.47	0.18	27.50	0.024	0.071	0.0053
Balun Mahai Sea	3	83.51	7.71	0.19	27.66	0.025	0.055	0.0058
Alkaline mountain deep brine	4	48.06	0.98	2.53	0.30	0.042	0.33	0.19

Table 6.

Rb and Cs measurement values and spiked recovery rates of actual brines at different dilution ratios.

Sample	TDS	Dilution factor	Diluted TDS	Rb measured	Spiked	Measured (Spiked)	Rb Recovery	Cs measured	Spiked	Measured (Spiked)	Cs Recovery
1	272.0	200	2.72	1.00 ± 0.03	2	3.00 ± 0.2	100 ± 10	0.0194 ± 0.002	0.05	0.0684 ± 0.004	98.0 ± 9
		50	5.44	4.38 ± 0.07	10	15.2 ± 0.3	108 ± 3	0.0429 ± 0.004	0.1	0.139 ± 0.002	96.1 ± 4
		10	27.2	25.2 ± 0.2	25	50.6 ± 0.2	102 ± 1	0.224 ± 0.003	0.2	0.409 ± 0.01	92.5 ± 6
2	348.9	200	1.74	4.68 ± 0.2	5	9.63 ± 0.8	99.0 ± 16	0.263 ± 0.03	0.5	0.692 ± 0.04	85.8 ± 10
		50	6.98	18.2 ± 0.5	25	42.0 ± 1	95.0 ± 6	0.670 ± 0.05	2	2.72 ± 0.005	102 ± 2
		10	34.89	90.8 ± 2	50	142 ± 9	102 ± 18	2.83 ± 0.02	5	7.34 ± 0.6	90.2 ± 12
3	352.7	200	1.76	3.71 ± 0.2	0.5	4.22 ± 0.08	102 ± 37	0.363 ± 0.04	0.5	0.873 ± 0.01	102 ± 8
		50	7.05	15.9 ± 0.05	25	40.5 ± 0.4	98.3 ± 2	0.665 ± 0.02	2	2.73 ± 0.08	103 ± 4
		10	35.27	73.6 ± 2	50	127 ± 3	106 ± 7	3.82 ± 0.1	5	8.80 ± 0.09	99.6 ± 3
4	170.0	200	0.85	10.2 ± 0.2	5	15.3 ± 0.3	101 ± 6	5.13 ± 0.2	5	10.2 ± 0.3	102 ± 6
		50	3.40	42.8 ± 1	25	67.3 ± 1	97.9 ± 7	20.9 ± 0.8	25	43.4 ± 1	90.0 ± 6
		10	17.0	206 ± 5	100	308 ± 10	101 ± 11	98.7 ± 2	100	188 ± 12	89.1 ± 12

Units: TDS (g·L⁻¹), Rb and Cs concentrations (μg·L⁻¹), Recovery (%).

Reagents and standard

All chemicals used in this study were of superior reagent grade and were used without further purification. Sodium chloride (NaCl, 99.99%) was purchased from Jinsui Bio-Tech (Shanghai, China), while potassium chloride (KCl, 99.98–100.02%) was obtained from Leyan Biomedical Tech (Shanghai, China). Magnesium chloride hexahydrate (MgCl₂·6H₂O, 99.99%) and calcium chloride dihydrate (CaCl₂·2H₂O, 99.99%) were supplied by Meridian Tech (Shanghai, China) and Mreda Tech (Beijing, China), respectively. Stock standard solutions of rubidium (Rb) and cesium (Cs) at a concentration of 100 mg·L⁻¹ were acquired from Guobiao Testing & Certification (Beijing, China). The mixed internal standard solution (yttrium (Y), rhodium (Rh), germanium (Ge), scandium (Sc), bismuth (Bi), 100 mg·L⁻¹) was obtained from PerkinElmer (US). Low-concentration calibration and internal standard solutions were prepared by diluting the stock solutions with ultrapure water.

Sample collection and preservation

Four actual brine samples were directly collected from salt lakes using plastic bottles without further pretreatment and were labeled as samples 1, 2, 3, and 4. The major ions present in these samples are summarized in Table 5. Subsequently, the samples were filtered through medium-speed filter paper with pore sizes ranging from 15 to 20 μm and diluted with ultrapure water prior to analysis. The brine samples were diluted with ultrapure water to achieve final salinity levels of approximately 35 g·L⁻¹ (10-fold dilution), 10 g·L⁻¹ (50-fold dilution), and 2 g·L⁻¹ (200-fold dilution). The 10-fold and 50-fold dilutions were directly diluted and used to evaluate the applicability and accuracy of the all-matrix sampling (AMS-ICP-MS) method. In contrast, the 200-fold dilution was performed in two steps: a 10-fold dilution followed by a 20-fold dilution. This approach represents the conventional high-dilution method and was used to compare the analytical results between matrix dilution (online gas dilution) and traditional offline dilution. Samples 1 and 4 were diluted 200-fold using volumetric flasks that had been soaked in 10% HNO₃ for 24 h to remove trace Rb/Cs residues. Samples 2 and 3 were diluted 200-fold as well but served as the unacid-washed control group for parallel experiments. Detailed experimental results are presented in Table 6.

ICP-MS analysis

Rb has two isotopes: Rb⁸⁵, which is primarily influenced by rare earth elements (Yb⁺⁺ and Er⁺⁺), and Rb⁸⁷, which is additionally interfered with by Sr⁸⁷. Considering the low abundance of rare earth elements and the relatively higher Sr concentrations in salt lake brines²², Rb⁸⁵ was chosen for determination. In contrast, Cs has only one isotope, Cs¹³³, with 100% natural abundance and no significant interferences, making it highly suitable for determination. The internal standard element was selected based on experimental evaluation, and single-factor and multi-factor optimization experiments were conducted to determine the optimal conditions for the dilution gas flow rate, RF power, and nebulizer gas flow rate.

Matrix effect and impact of major cations on Rb and Cs analysis in high-salinity brines

The matrix effect before and after activating the dilution gas was calculated. The primary cations in salt lake brines are Na⁺, followed by K⁺, Ca²⁺, and Mg²⁺. Due to their similar chemical properties and ionization energies compared to Rb and Cs, these cations may cause interference in the determination of these two elements. Notably, the AMS technique enables direct analysis of high-salinity samples at 35·g L⁻¹, thereby necessitating a systematic investigation into the interference effects of brine cations (K⁺, Na⁺, Ca²⁺, Mg²⁺) on Rb and Cs determinations under such elevated salt conditions. Consequently, the interference effects of Na⁺, K⁺, Ca²⁺, and Mg²⁺ on Rb and Cs determinations were evaluated under both non-dilution gas (without AMS) and dilution gas (with AMS) conditions. Recovery rates were calculated, with deviations exceeding ± 20% indicating significant interference.

Method validation

The method was validated for linearity, limits of detection (LOD), recovery (accuracy), and precision (repeatability)²³. Calibration curves for Rb and Cs were constructed using standard solutions at multiple concentrations to evaluate linearity. The LOD_solvent values were determined by analyzing ultrapure water (as a reagent blank) 11 times and calculated as three times the standard deviation of the blank divided by the slope of the calibration curve²⁴, respectively. Subsequently, matrix effects were corrected using the matrix discriminant factor, as described in²⁵. The accuracy of the measurements was validated by comparing the determined Rb and Cs concentrations with their respective certified reference values. For comparative analysis, Flame Atomic Absorption Spectrometry (AAS) was employed using the standard addition method. Considering the relatively high detection limits of AAS (~ 0.1 mg·L⁻¹ for Rb and Cs), two brine samples with significantly elevated Rb and Cs concentrations (> 200 μg·L⁻¹) were selected to ensure robust and reliable quantification. The standard addition approach, based on five-point calibration curves, was implemented to effectively account for matrix-related interferences, thereby obviating the need for chemical modifiers during the analytical procedure. To assess repeatability, 11 replicate analyses were performed on diluted brine solutions, with results expressed as relative standard deviation (RSD%).

Results and discussion

Optimization of instrument parameters

While increasing the dilution gas flow rate enhances dilution and thereby reduces matrix interference, it may simultaneously decrease sensitivity and lead to sample loss. To balance these effects, key instrument parameters were optimized through single-factor experiments using a 20 μg·L⁻¹ Rb and Cs mixed standard solution (Fig. 2).

Fig. 2.

Sentivity change with instrument parameters. (a) AMS gas flow; (b) Radio frequency power; (c) Nebulizer gas flow.

Prior to introducing dilution gas, the nebulizer gas flow rate was initially optimized to 0.74 mL·min⁻¹. Following this, with the dilution gas flow rate fixed at 0.2 mL·min⁻¹, the nebulizer gas flow rate was systematically reduced to minimize sample dispersion while preserving optimal sensitivity. Based on the experimental results presented in Fig. 2, the selected factor levels for the multi-factor experiment were as follows: dilution gas flow rate (0.20, 0.30, 0.40) mL·min⁻¹; radio frequency power (1200, 1300, 1400) W; and nebulizer gas flow rate (0.70, 0.72, 0.74) mL·min⁻¹. During parameter optimization under this activated condition, plasma stability and minimal sample diffusion were ensured by adhering to the principle that “the total gas flow rate (nebulizer gas + dilution gas) should remain approximately constant across all experimental groups.” Specifically, once the dilution gas flow rate (Factor A) was set, the nebulizer gas flow rate (Factor C) was correspondingly adjusted to maintain a consistent total flow rate. The objective of the orthogonal optimization was to identify the optimal combination of dilution gas flow rate (A), radio frequency power (B), and nebulizer gas flow rate (C) within the newly established dilution-based experimental system. The ultimate goal was to achieve the greatest possible reduction in matrix interference while maintaining an acceptable level of sensitivity, thereby enabling accurate and stable quantification of trace rubidium (Rb) and cesium (Cs) in high-salinity salt lake brines (35 g·L⁻¹).

Table 1 presents the orthogonal experimental design (L9(3⁴)) scheme and the measurement results (Rb sensitivity, Cs sensitivity, combined sensitivity). k1, k2, and k3 respectively represent the average values of the combined sensitivity of all experiments containing each level of each factor (A, B, C). For instance, the k1 value of factor A (dilution gas flow rate) at level 1 (0.20 mL·min⁻¹) is the average of the combined sensitivities of experiments 1, 2, and 3 ((8.87 + 12.57 + 9.61)/3 ≈ 10.35). The range R is the difference between the maximum and minimum values of k1, k2, and k3 for a certain factor (R = max(k) − min(k)), and its magnitude directly reflects the degree to which the level change of the factor affects the combined sensitivity (the larger the R, the more significant the influence of the factor). The intuitive analysis (Table 1) shows that the range of factor A (dilution gas flow rate) is the largest at 5.16, far exceeding that of factor B (RF power, 1.76) and factor C (nebulizer gas flow rate, 1.61), which clearly indicates that the dilution gas flow rate is the most dominant factor affecting the sensitivity.

Table 1.

Orthogonal experimental design and optimization.

No	A Dilution gas flow (mL·min−1)	B RF power (W)	C Nebulizer gas flow (mL·min−1)	Rb sensitivity (cps × 105)	Cs sensitivity (cps × 105)	Combined sensitivity (cps × 105)
1	0.20	1200	0.70	3.06	5.81	8.87
2	0.20	1300	0.74	5.64	6.93	12.57
3	0.20	1400	0.72	4.00	5.61	9.61
4	0.30	1200	0.72	4.26	5.64	9.90
5	0.30	1300	0.70	3.67	5.08	8.75
6	0.30	1400	0.74	3.66	4.13	7.79
7	0.40	1200	0.74	2.83	3.31	6.14
8	0.40	1300	0.72	2.59	2.81	5.40
9	0.40	1400	0.70	1.84	2.20	4.04
k1	10.35	8.30	7.22
k2	8.81	8.91	8.30
k3	5.19	7.15	8.83
Range (R)	5.16	1.76	1.61

k1/k2/k3: Average combined sensitivity of each factor at different levels. The range R: calculated as R = max(k) − min(k), indicates the extent of each factor’s influence. Impact order is: A > B > C, The optimal conditions: A1 (0.20), B2 (1300), and C3 (0.74).

Based on the k-value analysis, the optimal combination determined was A1B2C3, with a dilution gas flow rate of 0.20 mL·min⁻¹, a radio frequency power of 1300 W, and an atomizer gas flow rate of 0.74 mL·min⁻¹. Under these optimized conditions, the total sample gas flow rate (atomizer gas + dilution gas) reached 0.94 mL·min⁻¹. Although this value was slightly higher than the atomizer gas flow rate independently optimized without dilution gas (0.74 mL·min⁻¹), the orthogonal experiment data fully confirmed that this parameter combination exhibited the highest sensitivity under all test conditions. The auxiliary gas and plasma gas flow rates were maintained at constant values of 1.2 mL·min⁻¹ and 12 mL·min⁻¹, respectively. All subsequent experiments were conducted under the above-optimized conditions to ensure data consistency and result reproducibility.

Although reducing the nebulizer gas flow rate during orthogonal experiment may raise concerns about potentially compromised nebulization efficiency, the stability of sensitivity within the optimized range (Fig. 2c), together with the identical optimal nebulizer flow rate (0.74 mL·min⁻¹) selected before and after AMS activation, confirms that aerosol generation remained efficient. This conclusion is further supported by the method’s high precision (RSD < 5%, Table 7).

Table 7.

Precision tests of Rb and Cs measurements in brine samples.

Sample	Dilute factor	Rb measurement (μg·L−1)											RSD%
		1	2	3	4	5	6	7	8	9	10	11
1-Rb	10	24.7	23.2	24.2	25.2	24.8	24.9	25.34	25.1	25.2	24.0	25.1	2.64
1-Cs		0.251	0.209	0.225	0.230	0.230	0.222	0.225	0.226	0.228	0.214	0.219	4.77
2-Rb	10	92.8	89.8	88.9	91.0	93.7	93.6	91.2	94.8	85.9	87.8	89.2	3.05
2-Cs		2.92	2.89	2.88	2.72	2.74	2.90	2.89	2.87	2.75	2.76	2.79	2.68

Internal standard element selection for ICP-MS analysis

Due to the high sensitivity and susceptibility to signal drift in ICP-MS, the careful selection of internal standard elements is critical for compensating signal fluctuations and matrix interference. Ideal internal standards should not be present in the sample matrix and should exhibit minimal susceptibility to interfering factors. In this study, experiments (Fig. 3) were conducted using matrices with varying NaCl concentrations (the dominant component in brines²⁶) to evaluate the correction performance of potential internal standard elements (Rh, Y, In, Sc, Ge, Bi) for Rb and Cs analysis. Y and Rh exhibited the most stable correction curves in response to instrumental drift, thereby confirming their effectiveness in minimizing signal fluctuations during continuous analysis. However, while they contribute to correcting matrix suppression, their impact is secondary compared to the AMS dilution effect. This is because physical matrix interferences, such as salt deposition, require physical rather than analytical mitigation strategies²⁷.

Fig. 3.

Internal standard calibration curves under increasing NaCl matrix for Rb/Cs quantification.

Matrix effect

The matrix effect is defined as the suppression or enhancement of the analyte signal caused by co-existing matrix components, where a positive value indicates signal enhancement and a negative value indicates suppression. Soft matrix effects (suppression or enhancement of 0%-20%) are generally considered negligible. However, for medium matrix effects (suppression or enhancement of 20%-50%), these can be mitigated through optimization of mass spectrometry conditions and the use of internal standard correction methods, thereby minimizing interference from high-salt matrices and ensuring the acquisition of stable and reliable data. For strong matrix effects (suppression or enhancement > 50%), specific strategies, such as matrix matching or standard addition, must be implemented to overcome the influence of the matrix²⁷. To assess the impact of dilution gas on the matrix effect, this study employed the method described in reference²⁷. The matrix effect percentage (ME%) experienced by the target analytes (Rb and Cs) was calculated based on the response values of the internal standard elements (Rh) measured in both the matrix samples(I_matrix) and the blank solvent​ (I_solvent):

1

where ME% is the matrix effect and and are the response values of the internal standard element in the matrix sample and solvent, respectively.

Based on the four actual brine samples analyzed in this study, a single dilution step (approximately 35 g·L^-1) was applied, and the signal values of the internal standard were measured in both the matrix and the blank solvent. The average matrix effect percentage (ME%) was -71.2% before and -42.2% after the introduction of the dilution gas, respectively. To directly evaluate the behavior of the analytes and validate the accuracy of the proposed matrix effect suppression calculation method described in the reference²⁷, we further determined the matrix effect on rubidium (Rb) and cesium (Cs) by analyzing their response values in both matrix samples and blank solvents after the introduction of dilution gas (using Eq. 1). The results indicated that the matrix effect on Rb ranged from -44% to -47%, with an average of approximately -45.5%, while for Cs, the range was from -34% to -40%, with an average of approximately -37%. The matrix suppression effect calculated using the method from reference²⁷ after introducing dilution gas (-42.2%) fell within the moderate suppression range and exhibited a high degree of consistency with the suppression levels obtained directly from the analytes themselves. This close agreement provides strong evidence that the approach described in reference²⁷-which uses internal standard signals to estimate the matrix effect on target analytes-is valid and robust. Moreover, it demonstrates that the selected internal standards (Rh and Y) accurately reflect the matrix-induced suppression experienced by the target analytes (Rb and Cs).

Overall, the results demonstrate that high-salt matrices cause significant suppression of Rb and Cs signals. However, the introduction of dilution gas effectively mitigates this effect, reducing the suppression from a strong inhibition level to a moderate range. The decrease in salt load is primarily attributed to the dilution effect of the AMS gas (as shown in Fig. 1), while the internal standards (Y and Rh) efficiently compensate for signal instability caused by plasma fluctuations.

Interference effects of coexisting ions on Rb and Cs determinations

Figure 4 showed the interference effects of Na⁺, K⁺, Ca²⁺, and Mg²⁺ on Rb and Cs determinations under non-dilution gas (No AMS) and dilution gas (AMS) conditions. Recovery rates were calculated using Eq. (2).

Fig. 4.

Recovery rates of Rb and Cs under interference of Na⁺, K⁺, Ca²⁺, and Mg²⁺ with/without AMS dilution gas.

Figure 4 demonstrates that AMS notably extends the interference-free concentration range in the following analytical scenarios: (a) the determination of Rb and Cs in high NaCl matrices (> 30 g·L⁻¹);(b) the detection of Cs in high KCl (> 30 g·L⁻¹), MgCl₂ (> 20 g·L⁻¹), and CaCl₂ (> 5 g·L⁻¹) matrices. In contrast, the enhancement effect of AMS is relatively limited for other analyte-matrix combinations, such as Rb in KCl, MgCl₂, or CaCl₂ matrices.

2

In summary, K exhibits the most significant influence on the determination of Rb and Cs. This is primarily due to their analogous chemical properties^28,29. Specifically, they possess closely aligned ionization energies (Rb: 403 kJ·mol⁻¹, Cs: 375 kJ·mol⁻¹, K: 418 kJ·mol⁻¹). When the sample solution is diluted using the AMS device, K, which has a larger matrix effect compared to Rb and Cs, demonstrates a more pronounced dilution behavior. Consequently, Cs is preferentially ionized within the potassium matrix, leading to a notable impact. However, the similarity in ionization energies between Rb and K complicates their precise differentiation, resulting in relatively inferior performance. Similarly, when a substantial amount of Na (ionization energy: 496 kJ·mol⁻¹) is present, Rb and Cs are preferentially ionized, achieving a satisfactory dilution effect in the sodium matrix. The ionization energies of Ca and Mg are 589 kJ·mol⁻¹ and 737 kJ·mol⁻¹, respectively, both significantly higher than those of Rb and Cs. As such, there is minimal ionization interference either before or after the activation of the dilution gas. Fluctuations in recovery rates are predominantly attributed to physical clogging of the inlet system (atomizer, sampling cone, and interception cone) at relatively high concentrations³⁰.

Method selection

For the actual brine sample collected from Chinese salt lake regions and diluted to 35 g·L⁻¹ (see Table 2): (a) Na⁺ (the dominant matrix ion, concentration range: 10 g·L^-1-30 g·L⁻¹): AMS exhibits a highly effective suppression of Na⁺-induced interference, which is critical for achieving accurate and reliable quantification of target analytes; (b) K⁺ (concentration range: 0.13 g·L-1-2.1 g·L⁻¹): The concentration of K⁺ remains substantially below the threshold that would cause significant interference (as shown in Fig. 4, Rb recovery deviation < 5% when K⁺ < 2 g·L⁻¹); (c) Mg²⁺/Ca²⁺ (≤ 2.9 g·L⁻¹): The background interference from these ions is negligible and does not compromise analytical performance. In conclusion, AMS plays an essential role in enabling robust detection of Rb and Cs in high-sodium matrices. Although its influence on low-interference ions is relatively modest, it still contributes significantly to improving the method’s versatility and overall reliability across a wide range of sample types. Combined with the finding that the matrix effect was reduced from strong (- 71.2%) to medium level (- 42.2%) primarily by AMS gas dilution, it can be concluded that the concentrations of Rb and Cs in actual brines can be directly determined using a simple standard curve method without matrix matching²⁷. The optimized parameters^27,31 and Y/Rh internal standards compensate for instrumental fluctuations, but do not substitute for physical matrix reduction³².

Table 2.

K⁺, Na⁺, Ca²⁺, Mg²⁺ concentration in salt lakes in different regions of China³³.

Region	Ions (mg·L−1)				Salinity (g·L−1)
	Na+	K+	Ca2+	Mg2+
Tibet	61,607	6646	158.7	4658	204.24
Qinghai	65,638	6687	3407	28,397	340.75
Inner Mongolia	97,163	2638	129.3	3961	288.32
Xinjiang	98,483	1873	722.7	9848	289.85
Others	26,011	476.4	155.7	12,546	224.83

Method validation: calibration curvel linearity and sensitivity

Upon activation of the dilution gas, the standard curve was constructed by plotting the ratio of internal standard element intensity to the measured Rb or Cs intensity against their respective concentrations. The LDQ of the matrix were corrected from the LDQ of the solvent using Eqs. (3) and (4) from the signal strength of the spiked data of actual brines of Table 6²⁷.

3

4

Where : signal strength of the unspiked sample 1, 2, 3, 4 of tenfold dilution (approximately 35 g L⁻¹); : corresponding signal strength of the spiked sample ; : signal strength of the standard solution with the same concentration as the spiked sample; : signal strength of the solvent used for the standard solution.

The method demonstrated exceptional linearity, as summarized in Table 3, with high R² values (≥ 0.9998) and remarkable sensitivity.

Table 3.

Linearity, limit of detection, and limit of quantification of Rb/Cs.

Elemental	Standard solution range(ug·L−1)	Calibration curves	R2	LDQsolvent (ug·L−1)	LDQ matrix (ug·L−1)
Rb	5-400	y=0.0072x-0.013	0.9998	0.030	0.039
Cs	5-400	y=0.0030x-0.0050	0.9999	0.003	0.005

Accuracy: determination of standard samples

Ideally, brine-certified reference materials would have been utilized to assess the accuracy of the proposed method. However, due to the unavailability of certified reference materials (CRM) for Rb and Cs in salt lake brine, NaCl, which is the predominant component in brine, was added to simulate the brine matrix for method validation. The quantitative relationship obtained from the regression analysis was applied to determine the national standard substance (with 3.0% NaCl as the matrix), thereby validating the accuracy of the standard curve. The standard substance’s name, standard value, dilution factor, measured value and relative error (RE%) are presented in Table 4.

Table 4.

Accuracy of national certified reference materials (CRMs) determination (3.0% NaCl as the matrix).

Certified reference material	Dilution factor	Standard value (mg·L−1)	Measured value (μg·L−1)	RE %
GNM-M263963-2013	2000	Rb: 10 ± 0.2	Rb: 5.08 ± 0.4	1.6%
		Cs: 10 ± 0.2	Cs: 4.96 ± 0.3	0.7%
GNM-M022139-2013	2000	Rb: 100 ± 1.4	Rb: 53.2 ± 2.0	4.5%
		Cs: 100 ± 1.4	Cs: 52.4 ± 2.0	4.6%

As shown in the table, the measured values are closely aligned with the standard values, demonstrating a determination deviation of less than 5%. This indicates that the method possesses high accuracy.

Accuracy: determination of actual salt lake brine samples

The concentrations of Rb and Cs in four brine samples with varying salt contents were quantified using the standard curve method, and spike recovery experiments were performed to validate the accuracy. The major ions present in the samples are summarized in Table 5, while the analytical results are presented in Table 6.

Table 6 shows that for brine samples with a salt content of approximately 35 g·L⁻¹, the measured Rb values were consistent with those at 2 g·L⁻¹ and 5 g·L⁻¹. The spiked recoveries ranged from 95% to 108%, which further validates the reliability of Rb determination using the online gas dilution method.

For Cs determination, Sample 4 (high Cs concentration) exhibited consistent values across three salinity levels (2 g·L⁻¹, 5 g·L⁻¹, and 35 g·L⁻¹), indicating reliable measurements under varying salt conditions. Sample 1 (low Cs concentration) also demonstrated stable results, likely attributed to the use of acid-treated volumetric flasks for high-water dilution during the experiment. In contrast, Samples 2 and 3 (low Cs concentrations) were prepared using untreated glassware, leading to 57% and 118% concentration deviations at 2 g·L⁻¹compared to concentrations at 35 g·L⁻¹ and 10 g·L⁻¹, the recovery rate of Cs was 85%to103%, indicating that the deviation was not caused by the method, are likely attributable to systematic errors caused by trace residues in the untreated glassware.

Experiments demonstrate that under traditional high-dilution water conditions (200-fold or higher), in addition to errors inherently introduced by the dilution process, residual substances in containers significantly compromise the quantitative accuracy of Rb and Cs. Consequently, pickling pretreatment is indispensable for ensuring accurate determination at low concentrations. However, this step introduces additional operational complexity, increasing the overall procedure duration by more than 70% when combined with secondary dilution (10-fold × 20-fold). To address these limitations, the online gas dilution method proposed herein enables direct injection analysis of high-salinity brine (salinity of 35 g·L⁻¹) via a single dilution step (10-fold) coupled with dynamic argon matrix suppression (0.2 mL·min⁻¹). This approach entirely circumvents secondary dilution and acid washing pretreatment while maintaining high detection accuracy for Rb and Cs with recovery rate range from 85% to108%. By simplifying pretreatment steps and mitigating matrix interference, this method markedly enhances the efficiency and reliability of trace Rb and Cs analysis of high salinity brines, offering an innovative solution for rapid element detection in high-salinity complex matrices.

Precision of the method

The 10-fold diluted solutions of actual brine samples 1 and 2 (with TDS approximately 35 g·L⁻¹) were measured 11 times to evaluate the precision of the method. The results, as shown in Table 7, indicate outstanding method precision.

Validation by flame atomic absorption spectroscopy (AAS)

The AMS-ICP-MS results were validated using AAS with the standard addition method. Due to the relatively higher detection limit of AAS (~ 0.1 mg·L⁻¹ for Rb and Cs), two brine samples with elevated concentrations (Samples 5 and 6, > 200 μg·L⁻¹) were selected for analysis.

As summarized in Table 8, spike recoveries for both analytical methods ranged from 98.6% to 114%, which falls well within the acceptable recovery range of 80%-120%. The relative deviations between the two methods were ≤ 5.3% for Rb and ≤ 12.2% for Cs. The greater deviation observed for Cs in Sample 6 (12.2%) may be attributed to matrix-induced adsorption differences during sample pretreatment. Importantly, the recovery rates demonstrate a high degree of consistency (e.g., 103% for ICP-MS vs 99.6% for AAS in Cs Sample 6), thereby confirming the accuracy of the quantification achieved by the proposed method.

Table 8.

Comparison of Rb and Cs determination by AMS-ICP-MS and AAS.

Elements	Sample	Method	Baseline value(μg·L−1)	Spike (μg·L−1)	Spiked value (μg·L−1)	Recovery (%)
Rb	5	ICP-MS	428 ± 0.5	200	633 ± 3	102 ± 2
		AAS	398 ± 1	200	597 ± 4	99.6 ± 2
	6	ICP-MS	224 ± 3	200	453 ± 2	114 ± 2
		AAS	217 ± 6	200	414 ± 2	98.6 ± 3
Cs	5	ICP-MS	230 ± 0.4	200	458 ± 2	114 ± 1
		AAS	223 ± 4	200	424 ± 1	100 ± 2
	6	ICP-MS	233 ± 3	200	439 ± 1	103 ± 2
		AAS	205 ± 0.1	200	404 ± 2	99.6 ± 1

This work enables the accurate and rapid determination of trace Rb and Cs in high-salt and complex matrices. It recently provides analytical support for investigating the gradient enrichment and migration mechanisms of Rb and Cs during the evaporation process of salt lake brine³⁴, as well as their content and occurrence characteristics in the brine and sediments of Tibetan salt lakes⁵, but also extends its applicability to the determination of Rb and Cs in other high-salt matrices, such as hydrochloric acid, ammonium acetate, and acetic acid leaching phases⁵.

Conclusion

1. This paper presents a rapid and accurate determination method for quantifying trace Rb and Cs in high-salinity brines using ICP-MS equipped with an all-matrix sampling (AMS) device. The AMS device achieves online gas dilution by vertically injecting argon gas into the brine matrix sample flow, enabling direct injection analysis of brine samples with salt concentrations as high as 35 g·L⁻¹. Experimental results indicate that at a salt concentration of 35 g·L⁻¹, after activating the gas dilution, the strong matrix effect is significantly mitigated to middle level, and the cation (K⁺, Na⁺, Ca²⁺, Mg²⁺) content levels in brines from various regions do not interfere with the determination of Rb and Cs. Detection of Rb and Cs can be achieved through a simple standard curve method combined with internal standard correction using Y and Rh, along with instrument parameter optimization. The analysis of real salt lake samples demonstrate the applicability of the simple standard curve method. Compared to the traditional method requiring acid-washed pretreatment for dilution containers and a two-step (or even multi-step) dilution process, this method requires only single-step dilution (10-fold) in conjunction with online gas dilution to achieve precise determination of low-concentration Rb and Cs. This approach enhances analytical efficiency by more than 70% while eliminating the need for complex matrix matching or standard addition methods. The method exhibits excellent linearity (R² > 0.9998), with detection limits of 0.039 μg·L⁻¹ for Rb and 0.005 μg·L⁻¹ for Cs. Sample spike recovery rates range from 85% to 108%, and the relative standard deviation (RSD) is less than 5%, indicating high accuracy and precision. The accuracy was further verified by AAS standard addition for high-concentration samples (> 200 μg·L⁻¹), showing consistent recoveries (98.6%-114%) and inter-method deviations ≤ 12.2%.

2. Future research could focus on two key directions: First, for the accurate detection of rubidium and cesium in extremely high-potassium brine and extraction industries (K⁺ > 20 g·L⁻¹), breakthroughs can be achieved by optimizing parameters such as the flow rate and dilution ratio of the dilution gas, as well as improving matrix matching strategies. Second, it is essential to validate the applicability of this method for trace high-value elements (e.g., U and I). Given their low concentrations (at the (10⁻⁹-10⁻¹²) g·L⁻¹ level), these elements are susceptible to cross-contamination interference, necessitating further method optimization to enhance the universality and reliability of the approach.

Author contributions

Xiuzhen Ma: Conceptualization, Methodology, Writing – original draft. Guojing Zhu, Yubin Li, Dan Ma: Investigation, Formal Analysis. Zhe Ma, Xin Liu: Methodology, Validation, Writing – review and editing. Qi Wei: Investigation (design and execution of supplementary experiments), Validation, Writing – review and editing. All authors have read and approved the final version of the manuscript.

Funding

This work was funded by the Method Development and Innovation Project of the Chinese Academy of Sciences (lz2024g105), Qinghai Province Ten National Science and Technology Innovation Platform Training and Construction Project (2023-ZJ-J03).

Data availability

The data used to support the findings of this study are included within the article.

Declarations

Competing interests

All authors declare no competing financial or non-financial interests.