Photoinduced charge generation plays a central role in a broad range of physical, chemical, and biological processes that underlie natural and engineered photocatalytic and photovoltaic systems. Organic donor–acceptor systems are particularly intriguing candidates for light-harvesting applications, as their properties can be readily modified using well-established chemical synthesis techniques. Improving the efficiency of the underlying light-harvesting and charge generation processes, however, requires detailed knowledge of all the steps from the initial light-induced excitation of the chromophore to the final state, where charges are separated in the donor and acceptor phases. Copper-phthalocyanine (CuPc):C₆₀ is a canonical model system for this class of devices, but despite a significant body of research, fundamental mechanisms for charge separation remain obscure. Even more concerning, partly contradicting interpretations and models have been presented regarding the question, which initial excitations contribute to charge generation, and which do not. Evidently, a better theoretical understanding and novel experimental approaches are needed to validate or dismiss fundamental assumptions, regarding the nature and fate of photoexcited states in organic heterojunctions.

Light harvesting in CuPc:C₆₀ is initiated through creation of an excitonic state at the chromophore (CuPc), while the desired final state consists of a separated electron–hole pair with a vacancy in the chromophore and a free electron in C₆₀. For many organic systems, C₆₀ is an excellent acceptor, capturing the electron and thus separating the charges. Adding a small amount of C₆₀, the fluorescence radiation from the recombination of the excitonic state of the chromophore is quenched^1,2, indicating a vastly improved efficiency of charge generation. Photoemission and inverse photoemission spectroscopy demonstrated that this process is energetically enabled by the electronic level alignment of the compounds forming the heterojunction, including CuPc and C₆₀ (ref. ³).

Let us review the knowledge about the energy landscape of the relevant electronic states involved. The energy level diagram of the CuPc:C₆₀ heterojunction, based on a combination of various spectroscopic data, is shown in Fig. 1. Singlet and triplet excitons in CuPc are located ~2 and 1.2 eV, respectively, above the ground state^3,4,5,6,7. This is shown on the left of Fig. 1. Photoemission measurements established an offset of 1.45 eV between the HOMO of CuPc and C₆₀ (ref. ⁸). This allows to determine the position of the HOMO of C₆₀ on the right side of Fig. 1 relative to the HOMO of CuPc. Combining photoemission and inverse photoemission spectroscopy, the onset of the HOMO–LUMO gap in solid C₆₀ was determined to 2.3 eV (ref. ⁹). Gas-phase spectra of negatively charged C₆₀⁻ indicate an energy difference of 2.0 eV between the X¹A_g ground state and the B¹S₁ lowest singlet excited state of neutral C₆₀ (ref. ¹⁰). Accordingly, the LUMO orbital of a fully occupied C₆₀ is located at an energy between 2.0 and 2.3 eV above the HOMO. This corresponds to the charge-separated state configuration on the C₆₀ side of the heterojunction, where C₆₀ is now negatively charged. Finally, from the open-circuit voltage of the CuPc:C₆₀ heterojunction solar cell, which corresponds to ~0.5 eV (ref. ¹¹), we can derive a boundary for the minimum energy difference between the HOMO of CuPc and the LUMO of C_60.

While the (static) energetics is quite well established, the dynamics of the charge generation processes and the states involved are much less defined, if not to say controversial. As far as CuPc as chromophore is concerned, the lowest optically allowed excitation is a singlet exciton (S₁) with ~2 eV excitation energy^3,4,5,6,7. This exciton converts by intersystem crossing with a lifetime of ~500 fs into a triplet exciton (T₁), which is located ~1.2 eV above the ground state^6,7 (cf. Fig. 1). When C₆₀ is in contact with CuPc, additional interface excitonic states (so-called interfacial charge transfer (ICT) states) with various electron and hole configurations exist⁷, which are located several 100 meV in energy above the triplet state. These interface excitons are characterized by electronic configurations with the electron predominantly located in the C₆₀, while the hole remains in the adjacent CuPc. These ICT states can be viewed as precursors to the charge-separated state.

While a consensus has largely been reached regarding the static picture of the various states and their energy positions, their dynamics and the details of the processes leading to the generation of separate charges remain controversial. For example, a common notion found in the literature on CuPc:C₆₀ is that fully relaxed ICT¹² and triplet^5,6,7 states are too low in energy to result in separated charges, while opposition to this picture is more rare⁹. However, we previously showed¹⁰ that T₁ states indeed dominate the generation of free charges on timescales of 100’s of picoseconds to nanoseconds. In contrast to these studies on Pc:C₆₀ systems, for a number of blends of C₆₀ with polymers or other small molecules, strong indications were found that low-energy charge-transfer excitons do in fact contribute to free-carrier generation, even for the fully relaxed ICT excitons^13,14,15. The direct, real-time observation of this charge generation channel, however, and its prevalence over competing loss channels is still outstanding.

Recently, we established time-resolved X-ray photoelectron spectroscopy (TR-XPS) as a unique tool to detect the presence of charge transfer electrons in C₆₀ (refs. ^16,17). To the best of our knowledge, this is the first femtosecond time-resolved core-level photoemission spectroscopy study of Pc:C₆₀ systems. Access to the C 1s core levels provides a unique perspective of the dynamic charge evolution in direct vicinity of the atom from which the core electron is emitted. This is the basis of electron spectroscopy for chemical analysis, a widely used standard tool for the study of the chemical and electronic structure of surfaces and interfaces. Ultrafast TR-XPS extends this capability into a new regime, allowing to monitor the femtosecond dynamics of charge generation and motion in complex interfacial systems with atomic specificity.

The technique is described in the methods section, as well as in more detail in our previous publications^16,17. Even though our earlier experiments were limited to a temporal resolution of ~70 ps, they clearly demonstrated that triplet states (T₁) substantially contribute to the generation of separated charges in the CuPc:C₆₀ system. Integrated over all timescales from picoseconds to nanoseconds, the triplet states even generate an order of magnitude more charges than any other state¹⁷. Their long lifetimes more than compensate for their small diffusivities, leading to large diffusion lengths and efficient charge generation at the interface to the C₆₀.

Here, we extend this new spectroscopic technique to the femtosecond regime, in order to capture the fast dynamics of the photoexcited states at the crucial moment when bound excitons dissociate into separated charges. We note that by carefully adjusting the fluence of the optical laser and the FEL, as well as the exposure time of each sample spot, we avoided complications, such as sample charging or space-charge effects that affected previous PES investigations at FEL facilities. Specifically, we are interested in the relaxation of ICT states, which are considered important “gateways” for the generation of separate charges^12,18. A common perception is that ICT states only dissociate into separate charges as hot states, but not when cooled down to the lowest energy of the corresponding state manifold^6,12. Thus, the cooling time is considered a temporal gate for charge generation.

We use femtosecond TR-XPS in a pump–probe scheme at the free-electron laser FLASH to study a planar heterojunction consisting of a CuPc donor and a C₆₀ acceptor phase. FLASH offers femtosecond X-ray pulses with sufficiently high photon energies to study C 1s photoelectrons. As we have established previously^16,17, the binding energy of the C 1s core electrons of C₆₀ directly reflects the charge transfer on a local atomic scale. A shift of the C 1s line to lower binding energy indicates that additional electronic charge is present near the carbon atom that is ionized. Accordingly, XPS enables direct, local, and quantitative insight into the critically important step of the decay of the excitonic state into separated charges. Using an excitation wavelength of 775 nm, the experiment specifically addresses the relaxation dynamics of photoinduced ICT states near the low-energy limit of the ICT state manifold. TR-XPS spectra are recorded on timescales between ~100 fs and several picoseconds.