A nanometer aerosol differential mobility analyzer, Nano-DMA, has been developed for measuring the size distribution of nanometer aerosols in the particle size range of 3-50 nm. The design is based on a cylindrical configuration and is optimized by means of the numerical model of Chen and Pui. Important design features include high particle penetration (low loss) through the Nano-DMA and high sizing resolution. For reducing particle loss in the aerosol transport passage, the aerosol residence time in the Nano-DMA is reduced by shortening the inlet transport passage. An optional feature of high inlet flow (up to 16.5 lpm) is designed in order to further reduce the residence time between the aerosol inlet and the slit in the classifying region of the Nano-DMA. A new entrance slit is designed to have optimal aerosol and sheath flow matching at a flow ratio of 1:10, and has a wide dynamic flow-ratio range (up to aerosol/sheath flow ratio of 1/70) compared with the TSI-standard DMA design. This slit improvement makes the Nano-DMA suitable for high resolution particle sizing and classification. For reducing the effect of Brownian diffusion broadening on the transfer function of the Nano-DMA, the collector tube length is shortened to 5.0 cm compared to the TSI-Standard DMA of 44.44 cm and TSI-Short DMA of 11.11 cm. At the design flow condition of 1.5 lpm aerosol (or 16.5 lpm aerosol high inlet how case) and 15.0 lpm sheath flow rates, the measurable size range is from 3-50 nm. The lower detection limit of 3.0 nm coincides with the lower detection limit of the TSI UCPC (Ultrafine Condensation Particle Counter). The base of the Nano-DMA is completely re-designed to avoid particle loss due to the undesirable electrosratic effect observed by Kousaka et al., and to obtain an uniform electric field in the entire classifying region. The overall performance of the Nano-DMA is then evaluated by the numerical model of Chen and Pui before its construction and experimental evaluation. By comparing with the experimental results obtained using the Tandem DMA technique described in Hummes et al., it is concluded that the Nano-DMA is performing well in the designed size range and its transfer function agrees well with the numerical prediction.