1: \begin{abstract}

2: We introduce a 

3: multimodel  

4: approach

5: to solve 

6: coupled cluster equations, employing a quasi Newton algorithm for the ground state 

7: and an 

8: Olsen algorithm for the excited states.

9: In these algorithms, 

10: both of which can be viewed as Newton algorithms,

11: the Jacobian matrix of a lower level 

12: coupled cluster

13: model is used 

14: in 

15: Newton equations 

16: associated with

17: the target model.

18: Improvements in convergence

19: then implies 

20: savings for sufficiently large molecular 

21: systems, 

22: since

23: the  

24: computational cost of 

25: macroiterations  

26: scales more steeply

27: with system size

28: than the cost of 

29: microiterations. 

30: The 

31: multimodel

32: approach is 

33: suitable

34: when

35: there is a lower level Jacobian matrix 

36: that is 

37: much

38: more accurate 

39: than the zeroth order

40: approximation.

41: Applying

42: the approach to the CC3 equations, using the CCSD approximation of the Jacobian, 

43: we

44: show that the time spent to determine the ground and valence excited states can be significantly reduced. We also find 

45: improved convergence

46: for core excited states, indicating that similar savings 

47: will be obtained

48: with

49: an explicit implementation of the core-valence separated CCSD Jacobian transformation. 

50: \end{abstract}

51: