Experimental evaluation of micro-ion trap mass spectrometer geometries

Abstract

We describe a new fabrication method, simulations, and experimental results for micromachined cylindrical ion trap (μ-CIT) arrays for use in miniaturized mass spectrometers. The micromachined μ-CIT arrays were fabricated in a silicon-on-insulator (SOI) substrate, and a variety of trap geometries were incorporated into a single μ-CIT array chip to allow fast iterative measurements of the differences in the mass spectra from μ-CITs with different ratios of half-axial to half-radial dimensions (z0/r0). The chip dimensions were approximately 1.0cm×1.5cm×0.1cm. A series of z0/r0 were chosen in incremental steps of 3% for each array by changing r0 from 308 to 392μm while keeping z0 fixed at 355μm, resulting in a range of z0/r0 from 1.16 to 0.92 (nine geometries in total). Simulations were performed in SIMION 7.0 to determine the optimum range of μ-CIT z0/r0 to be fabricated and tested, by producing simulated mass spectra from μ-CITs with a variety of z0/r0 to evaluate predicted mass resolution. Following the simulations, we fabricated the arrays of μ-CIT geometries in SOI wafers using deep reactive ion etching (DRIE) to create the cylindrical structures and surface metallization to create ion trap electrodes. Symmetrical arrays of half-CITs were fabricated, diced, and bonded back-to-back to obtain complete μ-CIT array chips containing all nine geometrical ratios (z0/r0), which are referred to in this paper as the “gradient arrays”. The bonding process provided approximately 5-μm alignment accuracy between the two arrays of half-CITs, and the resulting arrays had flat μ-CIT endplate electrodes with <3μm upwards bow. We discuss several critical issues encountered during process development such as delamination of the buried oxide layer, excessive wafer bow, high capacitance, and ring-electrode wall verticality, along with solutions to mitigate these issues. Mass spectra were obtained experimentally from each trap geometry, and μ-CIT performance was found to follow the trend with respect to z0/r0 observed in the simulations. Experimental efforts indicated that axial modulation on one endplate electrode was required to remove spurious peaks in the mass spectra (caused by higher-order multipole contributions to the trapping electric field), and resulted in mass spectra with full-width-at-half-maximum peaks of 0.4amu.