1: \setcounter{chapter}{5}

2: \setcounter{section}{5}

3: 

4: \chapter{Applications - Experimental Spectral Weight}

5: \label{applica}

6: 

7: \section{The organic compound TTF-TCNQ}

8: 

9: We will in this section devote some attention to the organic material "Tetrathiafulvalene Tetracyanoquinodimethane", abbreviated TTF-TCNQ. For temperatures above the broken symmetry state (linked to a Peierls transition as further described in this section), it is characterized as a metallic "charge transfer salt" consisting of linear stacks of planar molecules. In the subsequent sections, some general properties of this material will be discussed conveying the reasons for why it constitutes a reasonable quasi 1D material allowing it to be compared with the theoretical results obtained so far. It is not our intention to explain in greater detail the rich physical literature that exists regarding TTF-TCNQ, nor to make a "from first principles" derivation of its physical properties. This is outside the scope of this thesis report. However, the interested reader could use the references given in this and in the subsequent section for a deeper study of the properties of TTF-TCNQ.

10: 

11: The charge transfer occurs between the two types of molecules, i.e. between the stacks: an approximate $0.59$ electrons per molecule is transferred from the TTF to the TCNQ molecule, which drives the stacks metallic. In the metallic phase, the electrical conductivity is about three orders of magnitude larger in the intra-strack direction (for both molecules) than in other directions \cite{ttftcnq1}, a property attributed to the crystal structure \cite{ttftcnq01} \cite{ttftcnq02}. This is manifested through $\pi$-type orbitals overlapping in the conduction direction, i.e. overlapping with neighboring molecules belonging to the same stack. The high conductivity, as compared to inorganic metals, has been a key motivator for scientific investigations since the 1970's. Actually, TTF-TCNQ is one of the most celebrated organic conductors which have been widely examined since its discovery \cite{ttftcnq11} mainly regarding its electronic conductivity and optical properties, which have been thoroughly investigated in for example Refs. \cite{ttftcnq2}-\cite{ttftcnq6}. For our purposes, the material makes a good candidate for a "one dimensional electronic system" regarding the one electron removal spectral properties of the metallic phase. The crystal structure of TTF-TCNQ, along with its Brillouin zone, is depicted in Fig. (\ref{figcrystal}).

12: 

13: The non Drude behavior of the metallic regime is described in Ref. \cite{ttftcnq7}, a regime reached above the critical temperature $T_P=54$ K. This critical temperature is about half the value of the predicted mean-field weak coupling value, a deviation attributed to strong 1D fluctuations \cite{claessen1}. Below this temperature a charge density wave builds up in the TCNQ chains, eventually turning the system into an insulator at 38 K (the corresponding temperature for the TTF stacks is 49 K \cite{ttftcnq7}). The critical temperature is linked with a Peierls transition, where an electronic gap opens up due to the molecule lattice displacements \cite{ttftcnq8}-\cite{ttftcnq10}. This means that even though the theoretical model developed here refers to zero temperature, we have to study the material at temperatures higher than the critical temperature in order to reach the metallic phase, for which our theory is valid.

14: 

15: \begin{figure}

16: \hspace{1.0cm}

17: \subfigure{\includegraphics[width=6cm,height=8cm]{crystal_structure.eps}}

18: \hspace{2.0cm}

19: \subfigure{\includegraphics[width=5cm,height=8cm]{fourier_TTFTCNQ.eps}}

20: \caption{\label{figcrystal} The crystal structure of TTF-TCNQ. The angles about the $\bm{a}$ axis, as measured with respect to the $\bm{c}$ axis, ensures maximum covalent bonding along the stack direction, hence making the electronic conductivity strongly anisotropic. As a result, the electrical conductivity in the $\bm{b}$ direction is three orders of magnitude larger than in the other directions. The Brillouin zone (right) shows the high symmetry points in the corresponding reciprocal space directions. Thus, we measure electronic momentum along the $\bm{b}^*$ direction, identifying the centre of the Brillouin zone $\Gamma$, as the zero momentum point. \vspace{0.5cm}}

21: \end{figure}

22: 

23: The charge transfer between the molecules shifts the intra-molecular density of electrons by $n_{TCNQ}=0.59$ for TCNQ and thus by $n_{TTF}=2-0.59=1.41$ for TTF, i.e. with a hole density of $0.59$. Thus, due to the particle-hole symmetry of our model, it suffices to study systems with a density of $0.59$. In this way, "removing an electron" (RHB) for the TCNQ translates into "adding a hole" for the TTF (LHB), and hence the spectral weight of both transitions can be mapped onto the {\it same} ($k,\omega$) region, with $\omega \leq 0$. We must note however, that the transfer integrals for the individual stacks are different from each other. 

24: 

25: The fact that Coloumb interaction plays a key role for the electronic structure of TTF-TCNQ is not a new claim \cite{ttftcnq02} \cite{torrance}, and from physical properties other than the photo emission spectrum, one would expect that $U \approx 4t$, with a slightly higher value for this ratio for TTF, than for TCNQ. By comparing with the photo emission spectrum, it is a simple fitting procedure (with the pseudofermion energy dispersions as the fitting functions) to deduce the values of ($U/t$) yielding the best match between experimental and theoretical results \cite{claessen1} (note that the $c0$ pseudofermion bandwidth is constant and equal to $4t$). In the following, we shall use the results of Refs. \cite{ourprl} and \cite{carmclaes}, where an almost perfect agreement between the theoretical and the experimental spectral weight is found for $t=0.40$ eV and $U=1.96$ eV for TCNQ, and $t=0.35$ eV and $U=1.96$ eV for TTF, respectively. We note that these values of the Hubbard parameters produce an exact fit with the pseudofermion energy bandwidths. We thus obtain effective values $(U/t)=4.9$ for TCNQ and $(U/t)=5.61$ for TTF, respectively, which furthermore coincide with the findings of Ref. \cite{ttftcnq7} (and will be further demonstrated below). 

26: 

27: One last note should be made on the subject of "dimensional crossover", i.e. the phenomenon that the quasi 1D material, for some reason and in some regime, changes its physical behavior to start acting like, for example, a 2D layer. In Ref. \cite{zwick6} it is confirmed that the intermolecular transfer integral is on the order of $5$ meV $\approx k_B T_P$ in the $\Gamma B$ direction.  This means that at sufficiently low temperatures, we will start to see hopping between identical molecules belonging to {\it different} stacks. Moreover, below the Peierls transition we will also have important contributions to the dynamics of the system from electron-phonon interactions. These considerations force us to only study the system at temperatures $T>T_P$.

28: 

29: \section{ARPES experiments on TTF-TCNQ}

30: 

31: ARPES is an abbreviation for "angle resolved photoelectron spectroscopy" and basically stands for the experimental procedure equivalent to what a theoretician would call "one electron removal" \cite{arpes}. The basic physical considerations regarding the ARPES technique will not be accounted for here, other than just stating that the energy and direction of the photoemitted electron defines the quantum state of the material sample. Thus, varying these parameters, it is possible to obtain a full energy and momentum map. The ARPES technique does not demand a certain environment or sample temperature {\it per se}, and the experiments can be conducted at zero magnetic fields. Usually, this technique is employed for probing the shape of the Fermi surface of the sample, however in our case this technique is used to study properties of the bulk material itself. 

32: 

33: To motivate this, we use the reasoning of F. Zwick {\it et al} \cite{zwick}. In this reference, special concern is taken regarding surface effects: surface sensitiveness to radiation damage, aging of cleaved surfaces, eventual perturbations on the bulk charge balance and lattice periodicity due to surface effects, possibility of an insulating surface, and so forth. The "passing" of these tests for TTF-TCNQ allows us to conclude that the ARPES measurements actually do give us information on the bulk properties of the material. Some of the references that F. Zwick {\it et al} use in this analysis include Refs. \cite{zwick1}-\cite{zwick7}, with the main conclusion that cleaved surfaces of TTF-TCNQ are "highly ordered and retain the periodicity of the bulk" up to the penetration depth used by the ARPES technique. By keeping the material sample at an ambient temperature of 150 K, F. Zwick {\it et al} \cite{zwick} measure the spectral weight along the $\bm{b}$ axis, as well as along the perpendicular $\bm{a}$ axis. The last measurement serves the purpose of excluding this direction as a conductive direction, showing no dispersive behavior. The $\Gamma$ point of the Brillouin zone is the natural zero momentum point, with the spectral weight in the $\bm{b}$ direction symmetrical around this point. Further on, Zwick goes on discussing the temperature dependence of the spectra, and how the size of the Peierls gap fits with various theoretical models. The interested reader should note that the spectral weight distribution of TTF-TCNQ indeed does have a very interesting temperature dependence, which is summarized for example in Refs. \cite{claessen1} \cite{zwick}. However, the conclusions are expected: the experimental results do not fit with a Fermi liquid description (an example of this is the absence of the usual metallic Fermi edges of the spectral weight; instead a complete supression of the quasi particle weight is found), neither do they fit with a strong electron-phonon interaction description. Due to the incoherency of the spectral weight and the interaction dependent singular lines in the mapped ($k,\omega$) plane, one should use 1D correlated fermion models in order to explain the experimentally obtained spectral features. 

34: 

35: But there is one feature not explicitly touched in that reference, a feature upon which our theory triumphs or fails completely: the separation of the charge-type $c0$ excitations and the spin-type $s1$ excitations, usually referred to as the "spin-charge separation" \cite{boson3}. In contrast to standard Luttinger liquid theory, where one studies the spin-charge separation for low lying excitations only, here this separation occurs for {\it finite} energy excitations. In recent studies of the spectral behavior of the 1D Hubbard model, such separate charge and spin excitations were identified \cite{spectralattempt1} \cite{spectralattempt2}. In our language, however, this separation is manifested through the different types of contributions which lead to the full spectral function, with the branch lines and the border lines as the most obvious features. 

36: 

37: \begin{figure}

38: \begin{center}

39: \includegraphics[width=11cm,height=8.5cm]{claessen_bands.eps}

40: \caption{\label{figklasband} A density plot of the obtained ARPES dispersions of TTF-TCNQ according to Ref. \cite{ttftcnq1}, together with a density functional band prediction regarding the singular line features (of the same reference). We see major quantitative and qualitative discrepancies, for example the mismatch between the lines $a$, $b$ and $c$ and the experimentally obtained data. Also, we see that the band theory is incapable to reproduce the experimental features labeled $d$ and $d'$, which is perhaps the most serious failure of this method.  \vspace{0.5cm}}

41: \end{center}

42: \end{figure}

43: 

44: From an experimental point of view, we find that Ref. \cite{ttftcnq1} presents ARPES measurements on TTF-TCNQ with details of the spectral features not previously reported. In that reference, the ARPES procedure involved a momentum and energy resolution of $0.07$ \AA$^{-1}$ and $60$ meV, respectively, and with an ambient temperature of $T=61$ K. Moreover, a comparison with the predictions of density functional band theory is made. Unfortunately, the line predictions of this method fails completely, producing non physical band dispersions (i.e. predicting spectral weight in the ($k,\omega$) plane along lines with the wrong bandwidth) and missing some experimentally proven spectral features all together. This is illustrated in Fig. (\ref{figklasband}). 

45: 

46: \begin{figure}

47: \begin{center}

48: \includegraphics[width=11cm,height=8.5cm]{claessen_pseudof.eps}

49: \caption{\label{figtheend} A density plot of the obtained ARPES dispersions of TTF-TCNQ according to Ref. \cite{ttftcnq1}, together with the line features of the pseudofermion representation of the 1D Hubbard model. The line between $-k_F$ and $+k_F$ encircling the symmetry point $\Gamma$, as well as the lines emerging from $\pm k_F$ and extending to higher values of $\vert \omega \vert$, find their exact equivalence in the RHB s-branch line and c-branch line theoretical features, respectively, with $(U/t)=4.9$, see Fig. (\ref{fig2DRHBU4p9}). The other line shapes, originating at $+k_F$ and fading away as we follow the dispersion towards the symmetry point $Z$, are exactly matched by the LHB s-branch line and c-branch line theoretical features, respectively, with $(U/t)=5.61$. Note that the latter s-branch vanishes quite rapidly as we depart from the Fermi momentum. The dashed line between the symmetry points $Z-k_F$ and $Z+k_F$ is nothing but the $(U/t)=5.61$ LHB border line. The LHB features are adopted from Fig. (\ref{fig2DLHBU5p61}). TTF has a hole concentration of $0.59$, and thus TCNQ has an electronic density of $0.59$. Due to the particle-hole symmetry of our model, the spectral weight of both transitions can be mapped onto the {\it same} ($k,\omega$) region, with $\omega \leq 0$. \vspace{0.5cm}}

50: \end{center}

51: \end{figure}

52: 

53: We have already discussed and shown in section (\ref{theorweight}), the various line features that the pseudofermion description of the Hubbard model produces. In Fig. (\ref{figtheend}) we show the same grey scale density plot of the experimental spectral weight of Ref. \cite{ttftcnq1}, but now fitted with the characteristic branch lines and border line features of our pseudofermion theory. These line features are exactly the same as those given in section (\ref{theorweight}) for the relevant values of ($U/t$), but with the TTF spectra folded onto negative energies, illustrating the donor-acceptor relationship between TTF (donor, holes) and TCNQ (acceptor, electrons). Note that a different choice of ($U/t$) for any of the two materials would produce line features not corresponding to the experimental line features. Now, it is straightforward to invert the spectral function of Fig. (\ref{figLHBU5p61}), from positive to negative energies, and add it to the spectral function of Fig. (\ref{figspecU4p9}), to produce the full theoretical spectral weight corresponding to the ARPES experimental result, Figs. (\ref{figfull}) and (\ref{figfull2}).

54: 

55: \begin{figure}

56: \begin{center}

57: \includegraphics[width=10cm,height=10cm]{spectral1.eps}

58: \caption{\label{figfull} Theoretical spectral weight for TTF-TCNQ, obtained by use of Eqs. (\ref{2P}), (\ref{borde1}), (\ref{borde2}), (\ref{alphabr}), (\ref{speclutt}), and (\ref{0P}), respectively. This figure is nothing but the superposition of the two spectral weight distributions already given in Figs. (\ref{figspecU4p9}) and (\ref{figLHBU5p61}), respectively. We expect the singular features of this spectral function to be visible in suitable photo emission studies of TTF-TCNQ. This figure is used to obtain Fig. (\ref{densityspec}), which indeed confirms the agreement between theoretical and experimental spectral weights. \vspace{0.5cm}}

59: \end{center}

60: \end{figure}

61: 

62: \begin{figure}

63: \begin{center}

64: \includegraphics[width=10cm,height=10cm]{spectral2.eps}

65: \caption{\label{figfull2} Theoretical spectral weight for TTF-TCNQ, obtained by use of Eqs. (\ref{2P}), (\ref{borde1}), (\ref{borde2}), (\ref{alphabr}), (\ref{speclutt}), and (\ref{0P}), respectively. This figure is identical to Fig. (\ref{figfull}) but shows the spectral weight distribution from a different angle. \vspace{0.5cm}}

66: \end{center}

67: \end{figure}

68: 

69: Lastly, we present the density plot corresponding to the spectral function provided in Figs. (\ref{figfull}) and (\ref{figfull2}), which is provided in Fig. (\ref{densityspec}). These plots lets us identify the types of excitations responsible for the total spectral weight of the system, hence chartering previously unknown territory. The dominant line shapes are thus due to separate charge type and spin type excitations for all relevant excitations, not just the low lying ones. Moreover, the relatively large spectral weight quite deep inside the band can here be identified with a "velocity resonance effect": the border line is the line where the two types of excitations propagate with the same group velocity.

70: 

71: \begin{figure}

72: \begin{center}

73: \includegraphics[width=10cm,height=10cm]{density_TTFTCNQ.eps}

74: \caption{\label{densityspec} Theoretical density plot of the spectral weight for TTF-TCNQ. Technically, it is this figure which should be compared with the experimentally obtained ARPES dispersions. We see that we have an almost complete agreement between this figure and Fig. (\ref{figtheend}). Important features include the two charge type excitations both originating at $k=k_F$, one related to the TCNQ stack and one to the TTF stack. They both fade away at higher momentum values. In contrast, the TCNQ spin type excitation carries much more spectral weight than its TTF counterpart, with the opposite relationship for the border lines. \vspace{0.5cm}}

75: \end{center}

76: \end{figure}

77: 

78: 

79: 

80: 

81: 

82: 

83: 

84: 

85: 

86: 

87: