Charged impurities in semiconductor quantum dots comprise one of the main obstacles to achieving scalable fabrication and manipulation of singlet-triplet spin qubits. We theoretically show that using dots that contain several electrons each can help to overcome this problem through the screening of the rough and noisy impurity potential by the excess electrons. We demonstrate how the desired screening properties turn on as the number of electrons is increased, and we characterize the properties of a double quantum dot singlet-triplet qubit for small odd numbers of electrons per dot. We show that the sensitivity of the multi-electron qubit to charge noise may be an order of magnitude smaller than that of the two-electron qubit.One of the most promising paths to scalable quantum computation is to use laterally defined double quantum dots (DQDs) in semiconductor heterostructures. The qubit is formed by the spin states of the two-electron DQD with total spin projection zero along the z axis. 1,2 Such a qubit is insensitive to spatially uniform magnetic field fluctuations, and, most importantly, amenable to fast electrical manipulation. 2,3 Recent experiments have made tremendous advances along these lines, demonstrating single-qubit initialization, arbitrary manipulation, and single-shot readout, all within a fraction of the coherence time of the qubit. 3-7 Preliminary steps toward an entangling two-qubit gate have also been reported. 8 In principle, successful completion of that program leaves the (admittedly enormous) challenge of scaling up to large numbers of qubits as the last remaining hurdle in the fabrication of a practical quantum computer.However, a practical issue has emerged which threatens to be a crippling impediment to continued rapid progress. The semiconductor samples used to create quantum dots invariably contain a number of charge impurity centers, perhaps 10 10 cm −2 in GaAs systems. 9 Even if the charge on these centers can be frozen to avoid switching noise, their presence inhibits access to the oneelectron-per-dot regime since the lowest energy states of the dot may be fragmented due to the roughened potential landscape. [10][11][12][13] This makes it difficult to find samples suitable for spin qubit realization. Furthermore, typically the impurities do introduce some switching noise, [14][15][16][17] so that even in samples in which the impurities are all far enough from the DQD that a two-electron singlettriplet qubit can be accessed, the interdot exchange energy is still subject to random fluctuation, leading to gate errors and decoherence. [18][19][20] This necessitates operating in a parameter regime such that the sensitivity of the exchange energy to the charge noise is minimized, a so-called "sweet spot". 21 In general, the charge noise problem is even more pernicious when performing twoqubit operations directly mediated by the Coulomb interaction, and one must again seek a sweet spot. 22,23 However, in practice, this strategy may not be sufficient since one typically cannot optimize ov...