The thermodynamics and kinetics of binding to nanocrystalline TiO2 were investigated for five ruthenium complexes that differed structurally in the number of possible anchoring carboxy groups (one, two, four, or six) attached to coordinated bipyridyl ligands and in the number of auxiliary ligands (bipyridine, CN-, or SCN-). Diffuse reflectance infrared spectroscopic data indicated that the dyes predominantly bound to TiO2 in a bridging mode in which the oxygen atoms of an attached carboxy group were bound to separate titanium atoms on the TiO2 surface. Furthermore, in the dry state, complexes with only one monocarboxy or dicarboxy ligand used essentially all of their available carboxy groups to bind to the surface. However, complexes having two or three dicarboxy ligands used on average two carboxylato groups in binding to TiO2. The structural differences between the complexes were manifested chemically in that the five dyes yielded similar maximum coverages (> 100 nmol cm(-2)) on nanocrystalline TiO2 electrodes, but exhibited different binding constants (10(3)-10(5) M-1) and different adsorption and desorption kinetics (3-11) x 10(3) M-1 h(-1) and 1-100 h, respectively). The binding constant for the monocarboxy dye was significantly lower than the binding constants for dyes with dicarboxy ligands, correlating primarily with an increase in the desorption rate of the monocarboxy complex. The adsorption rate constants were similar for all of the dyes, suggesting that formation of the first bond to TiO2 was rate limiting. Binding of the dyes from an ethanolic solution that contained pyridine and pyridinium as an acidic proton activity buffer yielded lower coverages than binding from a nonbuffered ethanol solution, even though the binding constants were up to 100 times greater under buffered conditions. The lower equilibrium dye coverage in buffered ethanol did not correlate with changes in the protonation state of the dyes but rather indicated competition for, and/or deactivation of, TiO2 active sites in buffered ethanol. The more weakly bound monocarboxy dye displayed the lowest short-circuit current density and open-circuit voltage under simulated solar illumination in a photoelectrochemical cell containing 0.50 M LiI, 0.040 M I-2, 0.020 M pyridine, and 0.020 M pyridinium triflate in acetonitrile. Additionally, even at constant coverage, the integrated quantum yield for photocurrent flow was lowest for TiO2 sensitized with the monocarboxy dye. The potential required to drive 0.1 mA cm(-2) of cathodic current density in the dark on dye-sensitized TiO2 photoelectrodes was least negative for the monocarboxy dye, indicating more facile electron transfer between reduced TiO2 and the solution redox couple. Hence, in this series of ruthenium carboxy-bipyridyl dyes, the most weakly bound species (i.e., the monocarboxy dye) yielded inferior photoelectrode properties, whereas differences between the dyes that contained at least one dicarboxy ligand resulted primarily from differences in the light absorption and energetic properties of the metal complexes. These observations suggest an important role for the linkage to the TiO2 surface in achieving temporal stability as well as in tuning both the steady-state quantum yield and the magnitude of the predominant back-reaction rate in dye-sensitized TiO2-based photoelectrochemical solar cells.