As indicated by Liudmil Antonov, the spectral features you are referring to, arise from the vibrational structure of a single electronic transition (S0->S2, the bright HOMO->LUMO computed by TDDFT). To retrieve them, the transitions between the vibrational states in S0 to those in S2 need to be included in the calculation of the whole lineshape. This implies going beyond the calculation of the vertical energy and tackle the calculation of the frequencies in the initial and final states and, then, perform the sum over all the possible transitions (Time-Independent methods) or the Fourier transform of the dipole correlation function (Time Dependent methods). Some of these approaches are available in Gaussian09 (not G03) as well as in other packages and stand alone programs.

In particular, the simulation of the lineshapes of that transition for different carotenoids has already been attempted in the last years in different works, such as:

J. Chem. Phys. 137 (2012) 22A534

Chem. Phys. 415 (2013) 247-255

J. Chem. Theory Comput. 9 (2013) 4947-4958

TD-DFT gives only 0-0 electronic transitions, does not give vibrational structure of the bands. So the methodology at this level of theory is correct.

Maybe you can usedifferent functionals such as CAM-B3LYP which is quite used  to calculate the UV-visible spectra. This particular functional is good when long-interactions are demanded which is the case of carotene. 

If it is not enough. How do you calculate the initial structure? Sometimes not all the possible conformers give the same results. Check if you've close energy conformers.

Another issue is whether the measurements were in gas phase or a liquid. A liquid phase can perturb the electronic states, yielding small shifts of energy absorptions.

Mr. Albar,

Please pay attention to teh attachment. There is shown how to cfonduct those calculations with corresponding keywords as illustrated in an open gaussian window.

As indicated by Liudmil Antonov, the spectral features you are referring to, arise from the vibrational structure of a single electronic transition (S0->S2, the bright HOMO->LUMO computed by TDDFT). To retrieve them, the transitions between the vibrational states in S0 to those in S2 need to be included in the calculation of the whole lineshape. This implies going beyond the calculation of the vertical energy and tackle the calculation of the frequencies in the initial and final states and, then, perform the sum over all the possible transitions (Time-Independent methods) or the Fourier transform of the dipole correlation function (Time Dependent methods). Some of these approaches are available in Gaussian09 (not G03) as well as in other packages and stand alone programs.

In particular, the simulation of the lineshapes of that transition for different carotenoids has already been attempted in the last years in different works, such as:

J. Chem. Phys. 137 (2012) 22A534

Chem. Phys. 415 (2013) 247-255

J. Chem. Theory Comput. 9 (2013) 4947-4958

The vibrational contribution can be obtained by FCClasses, which is free of charge and can be found at: 

http://www.pi.iccom.cnr.it/fcclasses

Remains, however the questions about how are the bands within the discussed relatively broad region, because of as the question is posted, the computation should be at NStates=1, because of by default it is NS=3, and thus there should have 3 transitions. If this is particular state computation, then it is not representative to the system.

Secondly complex multi-band profile, experimentally, of beta-carotene is observed in acetates. In set of non-polar solvents, however, it has only a broad band at 450 nm. In this context a computations in solution are also important to be conducted as it has pointed out by Mr. Harczuk .

Generally it would be better if the computations involved optimization by TD of Sx states, using GDIIS, which is recommendale for large systems (Gaussian manual), involving Nstates=6 and using quadratically convergent to SCF.

In the case of beta-carotene and carotenoids of similar size, the bright broad band is assigned experimentally to the S0->S2 transition. The state S2 is characterized by a HOMO->LUMO character and it is normally the one corresponding to the first root in TDDFT. S1 state has multireference character and one would need to go beyond DFT (CASSCF, MRCI...) to catch it properly (e.g. Chem. Phys. 373 (2010) 98-103) but since S0->S1 is not allowed in one-photon absorption, this is not a problem to get the spectrum.

A single point TDDFT calculation would give you the vertical energy, i.e., a single line which is normally convoluted with a Gaussian to get an approximation of the spectrum. Depending on your goal, that can be good, and you should focus on finding a DFT functional that provides you with an accurate prediction of the vertical energy. As pointed out by Ignat Harczuk, at this point, the inclusion of the solvent is crucial in order to get a realistic estimation of the vertical energy.

However, as stated along this post, if you want to get the fine details of the band, you need to incorporate the vibrational structure. As pointed out by  Bojidarka B. Ivanova FCclasses is a very good tool to do it. However, note that beta-carotene is a challenging system due to:

1. The size of the system: this implies that the optimization and, mainly, the computation of the frequencies on the excited state is quite expensive. But with patient and enough computational resources, this can be solved. Note that it would also be preferable that the optimization/frequency calculation is done incorporating the effect of the solvent. 

2. There are a lot of vibrational states, which implies that to get the vibrational shape  you need to account for an extremely large number of vibrational transitions with TI methods. Moreover, since a lot of the modes have low energy, a lot of vibrational states of S0 will be populated at room temperature, which further increases the number of possible vibronic transitions. Indeed, the experimental spectrum already shows a very remarkable temperature effect (experiments at 77K and room temperature are available in the literature, and they are quite different). The best solution is going for a TD method (instead of TI), which do not need to compute the vibrational states, but rather compute the dipole correlation function, which is analytical in the harmonic approximation (J Chem Phys, 119 (2003) 7188-7196; Chem. Phys, 370 (2010) 215-222) (but you still need the frequencies).

3. For beta-carotene, the optimized structures in the S0 and S2 states reflect a large structural displacement concerning the dihedral that connects the beta-ionone ring with the conjugated chain. This challenges the description of normal modes in Cartesian coordinates and, indeed, it has been shown that one would need to go to internal coordinates (J Chem Phys, 115 (2001) 9103-9109) or use vertical approaches to perform a realistic simulation of the spectral lineshape of this molecule (J. Chem. Theory Comput. 9 (2013) 4947-4958)

http://pubs.acs.org/doi/abs/10.1021/ct4005849