Highly Efficient Photochemical Vapor Generation for Sensitive Determination of Iridium by Inductively Coupled Plasma Mass Spectrometry

Herein, we describe the highly efficient photochemical vapor generation (PVG) of a volatile species of Ir (presumably iridium tetracarbonyl hydride) for subsequent detection by inductively coupled plasma mass spectrometry (ICPMS). A thin-film flow-through photoreactor, operated in flow injection mode, provided high efficiency following optimization of identified key PVG parameters, notably, irradiation time, pH of the reaction medium, and the presence of metal sensitizers. For routine use and analytical application, PVG conditions comprising 4 M formic acid as the reaction medium, the presence of 10 mg L–1 Co2+ and 25 mg L–1 Cd2+ as added sensitizers, and an irradiation time of 29 s were chosen. An almost 90% overall PVG efficiency for both Ir3+ and Ir4+ oxidation states was accompanied by excellent repeatability of 1.0% (n = 15) of the peak area response from a 50 ng L–1 Ir standard. Limits of detection ranged from 3 to 6 pg L–1 (1.5–3 fg absolute), dependent on use of the ICPMS reaction/collision cell. Interferences from several transition metals and metalloids as well as inorganic acids and their anions were investigated, and outstanding tolerance toward chloride was found. Accuracy of the developed methodology was verified by analysis of NIST SRM 2556 (Used Auto Catalyst) following peroxide fusion for sample preparation. Practical application was further demonstrated by the direct analysis of spring water, river water, lake water, and two seawater samples with around 100% spike recovery and no sample preparation except the addition of formic acid and the sensitizers.


EXPERIMENTAL SECTION
Instrumentation. Sample solutions were introduced in a flow-injection (FI) mode into a stream of the reaction medium with the aid of an injection valve (0.5 mL sample volume). Delivery at an arbitrary flow rate to the photoreactor was undertaken using a peristaltic pump (Reglo Digital, Ismatec) which was also used to evacuate waste from the gas-liquid separator (GLS). injection valve (0.5 mL sample loop volume) and this arrangement was exclusively utilized for estimation/determination of overall PVG efficiency (see Section "Procedure and Conventions"). A schematic of the PVG system coupled to ICPMS is depicted in Figure S1.
No special cleaning of the photoreactor was necessary between sequential measurements and the chemifold was typically only flushed with DIW at the end of the measurement day. From time to time, the quartz photoreactor was manually filled with concentrated HNO 3 via a syringe and the UV lamp was powered on to initiate decomposition of HNO 3 and to facilitate dissolution and removal of any deposited (metal) impurities.

S-4
Figure S1. PVG arrangement for FI coupling to ICPMS with simultaneous liquid nebulization.

GLS -gas-liquid separator, UHMI -ultra-high matrix introduction port.
Detection of generated volatile Ir species was achieved using an Agilent 8900 triple quadrupole ICPMS employing "wet plasma" conditions created by simultaneous pneumatic nebulization    Table S2.

RESULTS AND DISCUSSION
Release and Transport of Volatile Species. The effect of argon (chemifold) carrier flow on the release and transport of volatile species to the ICPMS was examined using a reaction medium of 10 M HCOOH and sample flow rate of 1.5 mL min -1 . The gas stream leaving the GLS was mixed with an additional flow of argon (not shown in Figure S1) before it was introduced to the ICPMS. Care was taken to keep the total gas flow to the ICP the same, so as not to influence conditions in the plasma or sampling depth. Although a slightly lower peak area sensitivity (by 9%) was obtained at 50 mL min -1 supplied to the GLS, no significant further effect of carrier Ar for PVG was observed in the range 100−600 mL min −1 , suggesting an efficient release and stability of the gaseous product. A flow rate of 200 mL min −1 was chosen as optimal in order to minimize any potential load of HCOOH vapor on the ICP.
The efficiency of release of volatile species from the liquid reaction medium was also investigated. The GLS was operated in such a way that no liquid comprising the sample was maintained in the GLS (the PTFE tube for waste removal was moved to the bottom of the GLS) and the peak area sensitivities were compared to those obtained with 1.5 mL of the reaction medium maintained inside the GLS. No significant changes in peak area sensitivity using 10 M HCOOH as the reaction medium were identified between either setup for sample flow rates of 1, 1.5 and 2.5 mL min -1 . This result suggests that the majority of the volatile Ir species is released to the gas phase prior to reaching the GLS, most probably after mixing with carrier Ar in the short transfer line to the GLS (see Figure S1) or the gas-liquid partitioning of the volatile species is intrinsically rapid. Chilling of the GLS in an ice-water bath also had no impact on peak area sensitivity but was used throughout as it significantly limited carryover of small droplets of the reaction medium formed in the GLS to the transport tube connected to the ICP and thus improved stability of measured signals. Figure S2. Influence of HCOOH concentration on peak area response from 200 ng L -1 Ir 3+ at a sample flow rate of 1.5 mL min -1 . Uncertainties expressed as SD (n ≥ 3) are sufficiently small that they cannot be discerned from the data points in some cases.  are sufficiently small that they cannot be discerned from the data points in some cases.

Re-optimization of the PVG System
S-9 Figure S6. Relative effects of added HNO 3 on PVG from 50 ng L -1 Ir 3+ in various HCOOH media containing 10 mg L -1 Co 2+ and 25 mg L -1 Cd 2+ as sensitizers. Combined uncertainty associated with individual data points is lower than 2% in all cases. Figure S7. Effect of sample flow rate using 10 mg L -1 Co 2+ and 25 mg L -1 Cd 2+ as sensitizers in 4 M HCOOH on peak area response from 50 ng L -1 Ir 3+ . Uncertainties expressed as SD (n ≥ 3) are sufficiently small that they cannot be discerned from the data points in some cases.