sentence-transformers How to use geethakurup/arxiv-finetuned with sentence-transformers:
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("geethakurup/arxiv-finetuned")
sentences = [
"<S> dimer vacancy ( dv ) defect complexes in the si(001)@xmath0 surface were investigated using high - resolution scanning tunneling microscopy and first principles calculations . </S> <S> we find that under low bias filled - state tunneling conditions , isolated ` split - off ' dimers in these defect complexes are imaged as pairs of protrusions while the surrounding si surface dimers appear as the usual `` bean - shaped '' protrusions . </S> <S> we attribute this to the formation of @xmath1-bonds between the two atoms of the split - off dimer and second layer atoms , and present charge density plots to support this assignment . </S> <S> we observe a local brightness enhancement due to strain for different dv complexes and provide the first experimental confirmation of an earlier prediction that the 1 + 2-dv induces less surface strain than other dv complexes </S> <S> . finally , we present a previously unreported triangular shaped split - off dimer defect complex that exists at s@xmath2-type step edges , and propose a structure for this defect involving a bound si monomer . </S>",
"any _ in vivo _ quantum system is in a contact with its environment . although typically weak , this interaction becomes relevant when studying the evolution of a system over long time scales . in particular , the asymptotic state of such an _ open _ system depends both on the unitary action induced by the system hamiltonian , and the action of the environment , conventionally called ` dissipation ' . a recent idea of `` engineering by dissipation '' @xcite , the creation of designated pure and highly entangled states of many - body quantum systems by using specially designed dissipative operators , has promoted dissipation to the same level of importance as the underlying unitary dynamics . the use of time - periodic modulations constitutes another channel to impact states of a quantum system . in the coherent limit , when the system is isolated from the environment , modulations imply an explicit time - periodicity of the system hamiltonian , @xmath3 . the dynamics of the system are determined by the basis of time - periodic _ floquet eigenstates _ the properties of the floquet states depend on various modulation parameters . modulations being resonant with intrinsic system frequencies can create a set of non - equilibrium eigenstates with properties drastically different from those with time - independent hamiltonians . modulations enrich the physics occurring in fields such as quantum optics , optomechanics , solid state and ultra - cold atom physics @xcite and disclose a spectrum of new phenomena @xcite . as an object of mathematical physics , ( [ lind ] ) has a specific structure and possesses a variety of important properties @xcite . in the case of a time - independent , stationary hamiltonian @xmath15 , the generator @xmath5 induces a continuous set of completely positive quantum maps @xmath16 @xcite . under some conditions ( ` quantum ergodicity ' ) , the system evolves from an initial state @xmath17 to a unique and time - independent asymptotic state @xmath18 , @xmath19 @xcite . when time - periodic modulations are present , eq . ( [ lind ] ) preserves the complete positivity of the evolution if all coupling rates are non - negative at any instance of time , @xmath20 @xcite . under some suitable , experimentally relevant assumptions , a set - up `` time - dependent hamiltonian and time - independent dissipation '' provides a valid approximation @xcite . here , we address the particular case of quench - like , time - periodic dependence of the hamiltonian , @xmath21 , corresponding to periodic switches between several constant hamiltonians @xcite . a popular choice is the set - up composed of two hamiltonians , @xmath22 where @xmath23 , @xmath24 $ ] . this minimal form has recently been used to investigate the connection between integrability and thermalization @xcite or , alike , for disorder - induced localization @xcite in _ isolated _ periodically modulated many - body systems . from a mathematical point of view , ( [ lind],[pc ] ) define a linear equation with a time - periodic generator @xmath25 . therefore , floquet theory applies and asymptotic solutions of the equation are all time - periodic with temporal period @xmath26 @xcite . @xmath25 is a dissipative operator and , in the absence of relevant symmetries @xcite , the system evolution in the asymptotic limit @xmath27 is determined by a unique ` quantum attractor ' , i.e. , by an asymptotic , time - periodic density operator , @xmath28 , @xmath24 $ ] and @xmath29 . the challenge here consists in explicit numerical evaluation of the matrix form of this operator . to use spectral methods ( complete / partial diagonalization and different kinds of iterative algorithms @xcite ) to calculate @xmath30 as eigen - element of a superoperator would mean to deal with @xmath31 scaling of computationally expensive operations . in the case of periodically modulated systems it restricts the use of spectral methods to @xmath32 , @xmath33 and the corresponding lindblad operators @xmath34,@xmath35 are sparse , the floquet map @xmath36 is a dense matrix . therefore , the numerical evaluation can not benefit from sparse - matrix methods . ] . a direct propagation of eq . ( [ lind ] ) for a time span long enough for @xmath37 to approach the attractor is not feasible for @xmath1 for two reasons : first , direct propagation requires to numerically propagate @xmath38 complex differential equations with time - dependent coefficients , and second , accuracy might become a problem for large times . although the accuracy may be improved by implementing high - order integration schemes @xcite or faber and newton polynomial integrators @xcite , this approach is hardly parallelizable ) by using the so - called time evolving block decimation ( tebd ) technique @xcite . the numerical effort scales as @xmath39 . however , this algorithm can only be used for lattice systems , i.e. , systems that can be partitioned into @xmath40 ` pieces ' coupled by next - neighbor interactions , both unitary and dissipative . in this case it gives a correct answer when the asymptotic state is characterized by a low entanglement . ] . systems with @xmath41 states may still be too small , for example , to explore mbl effects in open periodically - modulated systems . is it possible to go beyond this limit ? and if so how far ? we attempt to answer these questions by unraveling of the quantum master equation ( [ lind ] ) into a set of stochastic realizations , called `` quantum trajectories '' @xcite . this method allows us to transform the problem of the numerical solution of eqs . ( [ lind],[pc ] ) into a task of statistical sampling over quantum trajectories which form vectors of the size @xmath4 . the price to pay for the reduction from @xmath31 to @xmath4 is that we now have to sample over many realizations . this problem is very well suited for parallelization @xcite and we can definitely benefit from the use of a computation cluster . if the number of realizations @xmath42 becomes large , the sampling of the density operator @xmath37 with initial condition @xmath43 converges to the solution of eq . ( [ lind ] ) @xcite . this method is popular in the field of quantum optics where it adequately describes the physics of experiments @xcite . [ cols=\"^,^ \" , ] finally , we estimated the computation time to propagate a single trajectory on a single - core as a function of system size @xmath4 , see table [ table2 ] . for the model specified by eqs . ( [ eq : hamiltonian ] , [ eq : jump ] ) this time scales as @xmath44 ; this is due to the multiplication of the quadratic scaling of a dense matrix - vector multiplication and a linear scaling of the jump frequency . the latter scaling is , however , model specific and may differ for other physical systems so that the overall computation time may vary substantially with the type of hamiltonian and/or dissipators under study . on top , the numbers we present in table [ table1 ] depend on the values of the coupling constant @xmath45 and the period of modulations @xmath26 . this is so because these parameters control the rate of the jumps . therefore , the obtained estimates are specific and can not be used as universal quantifiers . we now report the results of our simulations obtained for the physical setup with eqs . ( [ eq : hamiltonian ] , [ eq : jump ] ) . we start with the performance of the algorithm , table [ table1 ] . the idea of the algorithm mimics a float : the algorithm constantly attempts to ` float to the surface ' , i.e. , to increase the time step of integration towards its maximal value @xmath46 while every next jump pulls it downwards to @xmath47 , see fig . [ fig : husimi ] . the average time between two consequent jumps is the mean of the local maxima in the depicted saw - like time sequence of @xmath48 . there is no problem in overestimating @xmath46 , simply because the time step will rarely reach its maximum . the shortest time step , @xmath47 , or , equivalently , the depth @xmath49 , is tuned to the values needed to reach the desired accuracy . next we turn to the averages @xmath50 over the realizations and the corresponding statistical variances @xmath51 $ ] of the matrix elements , as discussed in detail in section [ sampling ] . both quantities converge to `` limit cycles '' if the propagation time @xmath52 , @xmath53 , is much larger than all relaxation times . this means that for @xmath54 the density matrix converges to a time - periodic quantum attractor , i.e. , @xmath55 , and the variances also become time - periodic functions , @xmath56 \\backsimeq \\operatorname{var}[\\varrho^\\mathrm{att}_{kl}(\\tau)]$ ] , see fig . [ fig:3]b . the crumpled caustic - like shapes of the limit cycles is a result of the projection on a plane of a global limit - cycle living in @xmath57-dimensional space . this limit - cycle is not a topological product of @xmath31 two - dimensional limit cycles ; elements of the asymptotic density matrix do not evolve independently , they do interact so that their means and variances are coupled . in the asymptotic regime sampling can be performed stroboscopically , i.e. , after every period @xmath26 . in our simulations we used as the transient time @xmath58 and then performed the stroboscopic sampling of @xmath59 . the attractor density matrix at any other instant of time @xmath60 can be sampled by shifting the starting time of the sampling , @xmath61 , or simply by performing an extra - sampling at all needed intermediate points . with @xmath62 samples per trajectory ( that amounts to an additional propagation for the time @xmath63 ) it became possible to collect @xmath64 samples for the model of dimension @xmath65 ( i.e. , with @xmath66 indistinguishable bosons ) by running the program on @xmath67 cores during three days . for some choice of parameters the sample density matrix displays a standard diagonal - dominated structure , see fig . [ fig:3]a . the husimi distribution of the sample density matrix is depicted in fig . [ fig:4 ] . there is an intriguing similarity between the distribution of the quantum attractor and the phase - space structure of the classical attractor ( its stroboscopic section , to be more precise ) produced by the mean - field equations , eqs . ( [ eq : thetadot ] ) . this allows us to conjecture that the attractor density matrix is resolved with good accuracy . the @xmath68 cores allowed us to sample the same number of realizations for the model of the dimension @xmath69 during approximately one week , [ eq : jump ] ) involve summation over series of binomial coefficients of the order @xmath4 . it was not possible , however , to go beyond @xmath70 when evaluating the corresponding husimi distributions . ] . the primary aim of this study was to estimate the numerical horizon of a high - accuracy sampling of non - equilibrium dissipative quantum attractors , i.e. , the asymptotic solutions of periodically driven open systems , by using the quantum trajectory method . our main result is that it is possible to resolve asymptotic density matrices of systems of several thousand of states by using even a small cluster ( with @xmath71 cores ) on a time scale of a few days . the benefit of having access to the whole density matrix is the possibility to extract more detailed information about non - equilibrium regimes encoded in the matrix structure such as purity and many - body entanglement @xcite . naturally , the next step must be a systematic analysis of the errors , statistical and numerical ones . we reserve this objective for further studies . here , we would like to surmise on further optimization of the sampling procedure . an immediate idea is to use a reduced adaptive basis , constructed by using coherent states @xcite . this however demands a priori knowledge of the attractor s structure . this insight can be obtained by using models of small sizes , @xmath72 . another possibility is the choice of more optimal initial conditions which in turn will substantially reduce the transient time @xmath73 ( for example , one can use an extrapolation of the largest - eigenvalue eigenstate of the density matrix for a small system ) . altogether , we expect that such improvements could increase @xmath4 further by a factor of @xmath74 . research areas where quantum attractors are of potential interest were already mentioned in section [ introduction ] . here , we briefly recall them . first , this is many - body localization @xcite where the action of temporal modulations @xcite and dissipation @xcite so far have been considered separately . a combined action of both factors presents an intriguing challenge . a complementary task is to extend the idea of `` dissipative engineering '' @xcite to periodically - modulated quantum systems . finally , a survival of floquet topological insulators @xcite in the presence of dissipation is a timely question . the authors acknowledge support by the russian science foundation ( grant no . 15 - 12 - 20029 ) . we thank i. vakulchik for the help with preparing the figures . f. pastawski , l. clemente , and j. i. cirac , phys . a * 83 * ( 2011 ) 012304 ; m. j. kastoryano , m. m. wolf , and j. eisert , phys . rev * 110 * ( 2013 ) 110501 ; e. kapit , phys * 116 * ( 2016)150501 . k. j. strm and b. bernhardsson , _ comparison of periodic and event based sampling for first - order stochastic systems_. in preprints 14th world congress of ifac , beijing , p.r . china , july 1999 . ; k. j. strm and b. bernhardsson , lecture notes in control and information sciences * 286 * ( 2002 ) 1 . d. poletti , j .- s . bernier , a. georges , c. kollath , phys . * 109 * ( 2012 ) 045302 . c. gross , t. zibold , e. nicklas , j. esteve , m.k . oberthaler , nature * 464 * ( 2010 ) 1165 . j. tomkovic , w. muessel , h. strobel , s. lock , p. schlagheck , r. ketzmerick , m. k. oberthaler , _ observing the emergence of chaos in a many - particle quantum system _ ,",
"there are currently several exciting proposals to use the ( 001 ) surface of silicon for the construction of atomic - scale electronic devices , including single electron transistors @xcite , ultra - dense memories @xcite and quantum computers @xcite . however , since any random charge or spin defects in the vicinity of these devices could potentially destroy their operation , a thorough understanding of the nature of crystalline defects on this surface is essential . the si(001 ) surface was first observed in real space at atomic resolution using scanning tunneling microscopy ( stm ) by tromp _ _ et . al.__@xcite in 1985 . in this study they observed the surface consisted of rows of `` bean - shaped '' protrusions which were interpreted as tunneling from the @xmath1-bonds of surface si dimers , thereby establishing the dimer model as the correct model for this surface . since then , stm has been instrumental in further elucidating the characteristics of this surface , and in particular atomic - scale defects present on the surface@xcite . the simplest defect of the si(001 ) surface is the single dimer vacancy defect ( 1-dv ) , shown schematically in figs . [ def1](a ) and [ def1](b ) . this defect consists of the absence of a single dimer from the surface and can either expose four second - layer atoms ( fig . [ def1](a ) ) or form a more stable structure where rebonding of the second - layer atoms occurs @xcite as shown in fig . [ def1](b ) . while the rebonded 1-dv strains the bonds of its neighboring dimers it also results in a lowering of the number of surface dangling bonds and has been found to be more stable than the nonbonded structure . @xcite single dimer vacancy defects can also cluster to form larger defects such as the double dimer vacancy defect ( 2-dv ) and the triple dimer vacancy defect ( 3-dv ) . more complex clusters also form , the most commonly observed@xcite example is the 1 + 2-dv consisting of a 1-dv and a 2-dv separated by a single surface dimer , the so - called `` split - off dimer '' . the accepted structure of the 1 + 2-dv , as proposed by wang _ et . based on total energy calculations,@xcite is shown in fig . [ def1](c ) and consists of a rebonded 1-dv ( left ) , a split - off dimer , and a 2-dv with a rebonding atom ( right ) . recently we have observed another dv complex that contains a split - off dimer , called the 1 + 1-dv , which consists of a rebonded 1-dv and a nonbonded 1-dv separated by a split - off dimer , as shown in fig . [ def1](d ) . here we present a detailed investigation of dv defect complexes that contain split - off dimers . using high - resolution , low - bias stm we observe that split - off dimers appear as well - resolved pairs of protrusions under imaging conditions where normal si dimers appear as single `` bean - shaped '' protrusions . we show that this difference arises from an absence of the expected @xmath1-bonding between the two atoms of the split - off dimer but instead the formation of @xmath1-bonds between the split - off dimer atoms and second layer atoms . electron charge density plots obtained using first principles calculations support this interpretation . we observe an intensity enhancement surrounding some split - off dimer defect complexes in our stm images and thereby discuss the local strain induced in the formation of these defects . finally , we present a model for a previously unreported triangular - shaped split - off dimer defect complex that exists at s@xmath2-type step edges . experiments were performed in two separate but identical variable temperature stm systems ( omicron vt - stm ) . the base pressure of the ultra - high vacuum ( uhv ) chamber was @xmath3 mbar . phosphorus doped @xmath4 and @xmath5 @xmath6 wafers , orientated towards the [ 001 ] direction were used . these wafers were cleaved into @xmath7 mm@xmath8 sized samples , mounted in sample holders , and then transferred into the uhv chamber . wafers and samples were handled using ceramic tweezers and mounted in tantalum / molybdenum / ceramic sample holders to avoid contamination from metals such as ni and w. sample preparation@xcite was performed in vacuum without prior _ ex - situ _ treatment by outgassing overnight at 850 k using a resistive heater element , followed by flashing to 1400 k by passing a direct current through the sample . after flashing , the samples were cooled slowly ( @xmath9 k / s ) from 1150 k to room temperature . the sample preparation procedure outlined above routinely produced samples with very low surface defect densities . however , the density of defects , including split - off dimer defects , was found to increase over time with repeated sample preparation and stm imaging , as reported previously.@xcite it is known that split - off dimer defects are induced on the si(001 ) surface by the presence of metal contamination such as ni , @xcite and w @xcite . the appearance of these defects in our samples therefore points to a build up of metal contamination , either ni from in - vacuum stainless steel parts , or more likely w contamination from the stm tip . after using an old w stm tip to scratch a @xmath10 1 mm line on a si(001 ) sample in vacuum and then reflashing , the concentration of split - off dimer defects on the surface was found to have dramatically increased , confirming the stm tip as the source of the metal contamination . figure [ sods ] shows an stm image of a si(001 ) surface containing a @xmath10 10% coverage of split - off dimer defects . the majority of the defects in this image can be identified as 1 + 2-dvs , however , two 1 + 1-dvs are also present , as indicated . the most striking feature of this image is the difference in appearance of the split - off dimers in contrast to the surrounding normal surface dimers . each split - off dimer in this image appears as a double - lobed protrusion , while the surrounding normal si dimers each appear as a single `` bean - shaped '' protrusion , as expected at this tunneling bias . @xcite line profiles taken across a 1 + 2-dv both parallel and perpendicular to the dimer row direction are shown in fig . [ sods](b ) . the line profile parallel to the dimer row direction agrees with previously reported profiles over 1 + 2-dvs and fits well with the accepted structure , @xcite as shown by the overlayed ball and stick model . the line profile taken perpendicular to the dimer row direction , however , clearly shows that the split - off dimer of this defect is separated into two protrusions while the neighboring si dimers are single protrusions . this is the first recognition and explanation of split - off dimers appearing as double - lobed protrusions . 1 surface with split - off dimer defects is shown in ( a ) . tunneling conditions for this image were @xmath11 v sample bias and 0.8 na tunnel current . line profiles are taken across a single 1 + 2-dv both parallel , x x@xmath12 ( b ) , and perpendicular , y y@xmath12 ( c ) , to the dimer row direction , as indicated in ( a ) . the schematic ( d ) is a top view ball and stick model of a 1 + 2-dv with the approximate positions of @xmath1-bonds indicated by shaded ellipses . ] to understand why split - off dimers appear as double - lobed protrusions we must consider the structure of these defects shown in figs . [ def1](c ) and [ def1](d ) . normally si(001 ) surface dimers appear as `` bean - shaped '' protrusions in stm images because the dangling bonds of each si dimer atom mix to form a @xmath1-bond between the two dimer atoms . however , if we examine the split - off dimer structure closely ( figs . [ def1](c ) and [ def1](d ) ) we see that unlike normal surface dimers , the split - off dimer has two nearest neighbor second layer atoms that each have a dangling bond . the separation distance between the split - off dimer atoms and these second layer atoms is sufficiently close to allow the formation of @xmath1-bonds . the resulting four - atom structure can therefore be referred to as a _ tetramer_. we propose that the four dangling bonds of the split - off dimer tetramer interact primarily along the backbonds between the split - off dimer atoms and the second layer atoms to form @xmath1-bonds down the backbonds , as drawn schematically in fig . [ sods](c ) . these two spatially separated @xmath1-bonds therefore lead to the double - lobed appearance of the split - off dimers under low bias filled - state tunneling conditions , which we confirm in section [ theory1 ] with charge density calculations . in an attempt to fully characterize the appearance of these split - off dimers in stm images , we have performed a series of experiments observing split - off dimers with changing stm sample bias . figure [ sodv ] summarizes our results , showing images where a 1 + 2-dv and a 1 + 1-dv located next to each other are observed at four different sample biases two filled - state images and two empty - state images . in the filled - state image of fig . [ sodv](a ) we see that at @xmath13 v the split - off dimer of both the 1 + 2-dv and the 1 + 1-dv appear as double - lobed protrusions similar to those in fig . [ sods](a ) . however , when the filled - state bias is increased in magnitude to @xmath14 v , fig . [ sodv](b ) , the split - off dimers become single - protrusions and appear very similar to the surrounding normal si surface dimers . this is because as the bias magnitude is increased towards @xmath14 v , the dimer @xmath15-bond and bulk states contribute increasingly to the tunneling current @xcite and the image of the split - off dimer reverts to the bean - shaped protrusion in the same manner as normal surface si dimers . in both of the empty - state images , figs . [ sodv](c ) and [ sodv](d ) , acquired at + 0.8 v and + 2 v , respectively , the appearance of the split - off dimers is very similar to that of the surrounding normal surface dimers . this is because under empty - state tunneling conditions electrons tunnel into the @xmath16-antibonding orbitals of the dimers , resulting in the normal si dimers appearing as double - lobed protrusions . @xcite it is therefore only under low bias magnitude filled - state tunneling conditions that split - off dimers appear significantly different to the surrounding normal si surface dimers . v , ( b ) @xmath14 v , ( c ) @xmath17 v , ( d ) @xmath18 v. ] another noticeable feature of figs . [ sods](a ) and [ sodv](a ) is the enhanced brightness of the 1 + 1-dv compared to the 1 + 2-dv . this is a reproducible effect that we attribute to an increased amount of surface strain induced by the 1 + 1-dv . figure [ strain ] shows a series of adjacent defects forming a short vacancy line channel in the surface . this channel is composed of individual 1-dv , 3-dv , 1 + 2-dv , and 1 + 1-dv defects ( see figure caption ) . in the filled - state image , fig . [ strain](a ) , there is a clear brightening of the dimers on one end of the 1 + 1-dvs and the dimers on both ends of the 1-dv , which is not present for the 1 + 2-dvs . in the empty - state image of the line of defect complexes , fig . [ strain](b ) , we notice that there is a darkening of the same dimers that are enhanced in the filled - state image . v , @xmath19 v , 0.15 na ) of a short chain of dvs in a si(001 ) surface . the individual defects are ( from top left to bottom right ) : 1 + 1-dv , 1 + 1-dv , 1 + 2-dv , 3-dv , 1 + 2-dv , 1 + 2-dv , 1 + 1-dv , 1-dv , and 1 + 2-dv . note the strain - induced brightening of the 1-dv and 1 + 1-dvs in the filled - state ( a ) and the corresponding darkening in the empty - state ( b ) ] owen _ et . _ , @xcite have shown using low bias stm and first principles calculations , that the dimers neighboring a rebonded 1-dv are enhanced in low bias filled - state stm images due to the strain induced by the defect shifting the surface states upwards in energy toward the fermi energy . this effect can be seen for the 1-dv in fig . [ strain](a ) , where the neighboring dimers in the same row as the 1-dv are enhanced in intensity , with the magnitude of the enhancement decaying with distance from the 1-dv . a very similar enhancement can be seen around the 1 + 1-dv sites in this image , with the split - off dimer in particular appearing much brighter than the surrounding normal surface dimers . however , for the 1 + 1-dv only the dimers on one end of the defect are enhanced in intensity while the dimers on the other end of the defect are not . this observation can be readily explained since the 1 + 1-dv is composed of a rebonded 1-dv adjacent to a nonbonded 1-dv ( fig . [ def1](d ) ) and owen _ et . @xcite have shown that while the rebonded 1-dv results in strain - induced image enhancement , the nonbonded 1-dv does not . the observation of an asymmetric strain - induced enhancement of the 1 + 1-dv in fig . [ strain](a ) can therefore be taken as an experimental confirmation of the structure of this defect ( fig . [ def1](d ) ) and the first application of the method of owen _ et . @xcite for identifying strain in more complex surface defect structures . the fact that the 1 + 2-dv causes no enhancement of its neighboring dimers over the surrounding normal surface dimers suggests that the 1 + 2-dv , unlike the 1-dv and 1 + 1-dvs , does not increase the strain of the surface . this at first seems strange , since the 1 + 2-dv involves a rebonded 1-dv similar to the 1 + 1-dv structure . however , wang _ et . _ @xcite have shown , using total energy calculations , that the junction formed between the 1-dv and the 2-dv to create the 1 + 2-dv releases the surface strain that is present when these two defects exist separately from one another . the stm data that we have presented here is therefore the first experimental verification of this calculation . the fact that both the 1-dv and the 1 + 1-dv show local enhancement due to strain , while the 1 + 2-dv does not , indicates that the 1 + 2-dv structure induces less local strain than the 1-dv . in their paper , owen _ et . do not present empty state stm images , nor do they consider empty states in their tight binding calculations . in fig . [ strain](b ) , we show an empty state image of the same line of defects shown in fig . [ strain](a ) . interestingly , in this empty state image the dimers that were enhanced in brightness surrounding the 1-dv and 1 + 1-dvs in the filled - state image are less bright than the surrounding si dimers in the empty - state image . this suggests that the strain associated with these defects causes the lowest unoccupied molecular orbital ( lumo ) of the adjacent dimers to also shift higher in energy , away from the fermi energy . to confirm the interpretation of our stm images , we have performed first principles electronic structure calculations of both the 1 + 2-dv and 1 + 1-dv complexes using the car - parrinello molecular dynamics program . @xcite valence electrons were described using goedecker pseudopotentials @xcite expanded in a basis set of plane waves with an energy cutoff of 18 rydbergs and the exchange - correlation functional was of the blyp form . @xcite slab calculations contained between 124 and 128 si atoms in a @xmath20 @xmath21 supercell , corresponding to six layers of vacuum in the @xmath22-direction , and all calculations were performed with gamma point sampling of the brillouin zone only . a reference calculation was performed with no surface vacancies and assuming the @xmath23 structure in which the dimers buckled alternately along the row . a single 256 atom calculation with a duplication along the y - axis confirmed that the effect of dispersion across the rows is minor as has been noted elsewhere . @xcite both zero temperature geometry optimization and high temperature molecular dynamics calculations were used to explore a variety of surface and second - layer bonding configurations for the 1 + 2-dv and 1 + 1-dv . the results confirm the configurations in figs . [ def1](c ) and [ def1](d ) are the lowest energy geometries of both defect complexes . the dimers are drawn symmetric in these schematics , however , the true minimum energy structure at zero temperature involved charge - transfer buckling of the si dimers . it is well known that at room temperature the barrier is sufficiently small for the dimers to flip - flop between the two equivalent configurations . @xcite our calculations show that the split - off dimer tetramer also has two symmetrically equivalent buckling configurations , with charge transfer between the atoms of the tetramer buckling adjacent atoms in alternate directions . by analogy with the normal dimers we can expect room - temperature stm measurements of the tetramer to image the average of the two configurations . the chemical potential was determined from a 512-atom bulk calculation , which yielded a formation energy of 0.85 ev for the 1 + 2-dv , similar to the value of 0.65 ev computed by wang _ at . al_. @xcite the 1 + 1-dv formation energy has not been previously reported , and we found it to be 1.13 ev . we note that this value is high , but this is consistent with the rarity of observation of the 1 + 1-dv in stm experiments . in fig . [ 1 + 2-dv ] we present a series of calculated electron density slices through various regions of the 1 + 2-dv marked by ( a ) , ( b ) , ( c ) , and ( d ) in the ball and stick schematic . the charge density shown in the figure is the sum of the occupied kohn - sham orbitals within 0.25 ev of the highest occupied molecular orbital ( homo ) . taking into account the @xmath24 ev surface band gap of si(001 ) and the n - type doping of the experimental samples , these states correspond approximately to the accessible states for a @xmath25 v sample bias and can therefore be directly compared to the experimental data in fig . [ sodv](a ) , which was acquired with a @xmath13 v sample bias . -bonding as inferred from the electron density ( see text ) . each electron density plot is an average of both buckling configurations , and the atomic positions and bonds are shown as black balls and sticks . the slices are ( a ) rebonded 1-dv edge dimer , ( b ) split - off dimer , ( c ) split - off dimer backbonds , ( d ) 2-dv edge dimer . ] the four charge density slices in fig . [ 1 + 2-dv ] show : fig . [ 1 + 2-dv](a ) the 1-dv edge dimer , fig . [ 1 + 2-dv](b ) the split - off dimer , fig . [ 1 + 2-dv](c ) the backbond of the split - off dimer , and fig . [ 1 + 2-dv](d ) the 2-dv edge dimer , as indicated schematically in fig . [ 1 + 2-dv](e ) . the charge densities of both buckling configurations of the dimers and backbond atoms are averaged , and the positions of the dimer and tetramer atoms are shown superimposed in both buckling configurations . in the case of the backbonds , the two configurations are not coincident , and so the atoms and bonds are shown in projection onto the plane in fig . [ 1 + 2-dv](c ) . the 1-dv edge dimer in fig . [ 1 + 2-dv](a ) shows a clear three - lobed character with significant overlap between the up - atom charge density of the two buckling orientations , and a single lobe beneath the plane of the surface at the mid - point of the dimer . density functional calculations by hata _ et . al . _ @xcite and tight - binding green s function calculations by pollman _ et @xcite have separately identified this three - lobed feature as being characteristic of @xmath1-bonding in flip - flop dimers on the silicon surface , and we can therefore take this three - lobed feature as a signature of @xmath1-bonding in this work . the backbond of the split - off dimer in fig . [ 1 + 2-dv](c ) connects a first - layer atom to a second - layer atom and also shows a three - lobed structure . by analogy with the surface dimer in fig . [ 1 + 2-dv](a ) we characterize this bond as having @xmath1-character and have indicated this by the shaded ellipse ( c ) shown in fig . [ 1 + 2-dv](e ) . the split - off dimer itself in fig . [ 1 + 2-dv](b ) , however , does not exhibit three - lobed character . instead , the split - off dimer has four lobes ; two located above the up - atoms of the dimer in each buckling configuration , and a second pair of spatially separated lobes beneath the bond . the calculations thus show that @xmath1-bonding occurs down the backbonds of the split - off dimer , but not across the dimer itself . the absence of the @xmath1-bond across the split - off dimer correlates with the double - protrusions observed in the stm images . finally , we also consider the charge density of the 2-dv edge dimer , fig . [ 1 + 2-dv](d ) , and note that it also exhibits three - lobed character , indicative of @xmath1-bonding . this gives the 2-dv edge dimer a bean - shaped appearance in the stm image , as for the 1-dv dimer in fig . [ 1 + 2-dv](a ) . a similar situation exists for the 1 + 1-dv charge density slices shown in fig . [ 1 + 1-dv ] the first three charge density slices , figs . [ 1 + 1-dv](a ) [ 1 + 1-dv](c ) , are analogous to the slices for the 1 + 2-dv as was the case for the 1 + 2-dv , the rebonded 1-dv edge dimer , fig . [ 1 + 1-dv](a ) and the split - off dimer backbonds , fig . [ 1 + 1-dv](c ) exhibit three - lobed @xmath1-like character , while the split - off dimer , fig . [ 1 + 1-dv](b ) exhibits four - lobed character , consistent with an end - on view of @xmath1-bonding down the backbonds . finally , another slice is presented in fig . [ 1 + 1-dv](d ) , which is through the nonbonded 1-dv edge dimer as indicated schematically in fig . [ 1 + 1-dv](e ) . it can be seen that the nonbonded 1-dv edge dimer appears quite different to the charge density slices discussed so far . in particular , we notice that the nonbonded 1-dv edge dimer has a much reduced charge density compared to the other slices , fig . [ 1 + 1-dv](a ) [ 1 + 1-dv](c ) . examination of the structure identifies strain as the characteristic that differentiates the dimer in fig . [ 1 + 1-dv](d ) from the other dimers . since the dimer in fig . [ 1 + 1-dv](d ) is part of a tetramer , one might expect its appearance to resemble the split - off dimer which is also part of the tetramer shown figs . [ 1 + 1-dv](b ) and [ 1 + 1-dv](c ) . however , a detailed examination of the simulated structure reveals that the nonbonded 1-dv tetramer is relaxed , since there is one adjacent dimer present , while the split - off tetramer is highly strained because of the rebonding in the second - layer . since the nonbonded 1-dv tetramer is much less strained , its occupied states lie further from the fermi level , explaining the charge reduction observed in calculations in fig . [ 1 + 1-dv](d ) . as discussed in ref . , the minimum energy arrangement of the electrons in a tetramer is one where the @xmath1-states are delocalized across the four atoms , to form three bonding segments , as indicated by the ellipses in fig . [ 1 + 1-dv](e ) . the charge density slice of fig . [ 1 + 1-dv](d ) is consistent with such an arrangement where the charge density is shared between @xmath1-like bonds on both backbonds and across the dimer atoms . we conclude that this charge density arrangement forms for the nonbonded 1-dv tetramer because it is allowed to relax . in the case of the split - off dimer , the tetramer is constrained by the rebonding and instead forms a higher energy configuration in which the @xmath1-bonds conjugate to form two @xmath1-bonds down its backbonds . -bonding as inferred from the electron density ( see text ) . each electron density plot is an average of both buckling configurations , and the atomic positions and bonds are shown as black balls and sticks . the slices are ( a ) rebonded 1-dv edge dimer , ( b ) split - off dimer , ( c ) split - off dimer backbonds , ( d ) nonbonded 1-dv edge dimer . ] having presented a detailed understanding of the electronic structure of previously observed split - off dimer defects in the si(001 ) surface using both stm and first - principles calculations , we now turn our attention to elucidating the structure of a previously unreported split - off dimer defect . in figs . [ triangular](a ) and [ triangular](b ) we show filled- and empty - state stm images of dv defects at a single - layer s@xmath2-type step edge . at the top of these images white arrows indicate are three defects known as s@xmath2-dvs , which are rebonded 1-dvs at the step edge , which leave a single split - off dimer as the last dimer before the lower terrace begins . @xcite as was the case for the 1 + 1-dv and 1 + 2-dv , the split - off dimers in s@xmath2-dvs appear as double - lobed protrusions under low - bias filled - state imaging conditions , fig . [ triangular](a ) . at the bottom of fig . [ triangular](a ) two similar dv complexes can be observed , as indicated by black arrows , however these defects have a third protrusion giving them a triangular appearance . in empty - state imaging , fig . [ triangular](b ) , however , the additional third feature is not present . these triangular - shaped defects have not been reported on the si(001 ) surface before and most likely arise due to the presence of w contamination . 1.2 v ) of dv defects at an s@xmath2-type step edge . white arrows indicate s@xmath2-dvs , @xcite while black arrows point to a previously unreported defect that exhibits a third protrusion in the filled - state giving it a triangular appearance . we propose the structure ( c ) as a model for this defect . calculated charge density slices at a constant @xmath22-height for the dashed region of ( c ) are shown in ( d ) and ( e ) ( for kohn - sham orbitals summed over 0.45 ev below the homo and 0.45 ev above the lumo , respectively ) . these contour slices are in good agreement with the stm images in ( a ) and ( b ) , in particular predicting the correct spacing of 6.4 between the split - off dimer and third protrusion and also the disappearance of the third protrusion in the empty - state . the horizontal tic - marks in ( d ) and ( e ) indicate the dimer positions on the defect - free surface . ] our proposed structural model of the triangular - shaped defects in fig . [ triangular](a ) is shown in fig . [ triangular](c ) . this model consists of a nonbonded 1-dv defect at an s@xmath2-type step edge , followed by a rebonded split - off dimer and a bound si monomer . swartzentruber has previously observed si monomers on the si(001 ) surface using high - resolution stm after depositing a few percent of a monolayer of si atoms to the surface . @xcite these monomers were bound at rebonded s@xmath2-type step edges , confirming the minimum energy binding position predicted by first principles calculations . the binding position of the monomer in our proposed structure , fig . [ triangular](c ) , is essentially the same position observed by swartzentruber , with the difference being the presence of the dv defect adjacent to the step edge . swartzentruber also observed that the si monomers bound at s@xmath2-type step edges were visible in one bias polarity ( empty - state ) but invisible in the other ( filled - state ) . our images reveal a similar effect , however the feature we observe appears in filled - state images while being invisible in empty - state images . we have performed first - principles calculations to produce charge density contours for our proposed structure . figure 7(d ) shows a constant @xmath22-height contour slice taken 1.2 above the monomer for occupied kohn - sham orbitals within 0.45 ev of the homo . we see in this charge density contour slice the two lobes expected for the split - off dimer as well as a third lobe due to the bound monomer . moreover , the distance between the split - off dimer lobes and the monomer lobe is 6.4 in agreement with the separation seen in the stm image . in fig . [ triangular](e ) we show an empty - state slice taken at the same @xmath22-height and summed over kohn - sham orbitals up to 0.45 ev above the lumo . in this contour the double lobe of the split - off dimer is still present but the monomer lobe is significantly lessened in intensity . the results of our first - principles calculations therefore give good agreement between our proposed structure and the observed defect . the presence of the split - off dimer must therefore be responsible for the reversal of the filled- and empty - state monomer characteristics when compared to those observed for monomers bound to rebonded s@xmath2-type step edges . we have investigated split - off dimers on the si(001)2@xmath261 surface using high resolution stm and first principles calculations . we find that split - off dimers form @xmath1-bonds with second layer atoms which gives them a double - lobed appearance in low bias filled - state stm images . we apply the method of owen _ et . @xcite for identifying local areas of increased surface strain to dimer vacancy defect complexes and thereby present the first experimental confirmation of the predicted strain relief offered by the 1 + 2-dv . finally , we have presented a previously unreported triangular - shaped defect on the si(001 ) surface and a proposed model for this structure involving a bound si monomer .",
"recently , the study of the dynamics of @xmath3 on the @xmath2-character variety of @xmath5 for a compact hyperbolizable @xmath6-manifold @xmath0 with nonempty boundary and no toroidal boundary components , has led to the somewhat surprising discovery that often the dynamical and geometric decompositions of the @xmath2-character variety do not coincide . the @xmath2-character variety of @xmath5 denoted @xmath7 is the geometric quotient of @xmath8 by inner automorphisms of @xmath9 sitting inside @xmath10 is @xmath11 the set of discrete and faithful representations , which can equivalently be thought of as the deformation space of hyperbolic @xmath6-manifolds homotopy equivalent to @xmath12 the action of @xmath3 on the interior of @xmath13 is properly discontinuous ( see canary @xcite ) , but if @xmath0 has a primitive essential annulus , the action can not be properly discontinuous on all of @xmath13 ( canary - storm @xcite ) . minsky in @xcite was the first to observe that there exists a domain of discontinuity , called the set of primitive - stable representations , for @xmath3 containing the interior of @xmath13 as well as points on @xmath14 when @xmath0 is a handlebody . in particular , the set of primitive - stable representations contains both discrete and faithful representations and dense representations . jeon - kim - ohshika - lecuire @xcite gave a complete criterion for a discrete and faithful representation to be primitive - stable in terms of disc - busting properties of the ending lamination and parabolic loci of the associated hyperbolic @xmath6-manifold . in her thesis @xcite , the second author gave a generalization of the primitive - stable condition to representations of the fundamental group of a compression body into @xmath2 , called the separable - stable condition , and she showed that the set of separable - stable representations is a domain of discontinuity for the action of @xmath3 that contains the interior of @xmath13 as well as point on @xmath15 in this paper , we give a characterization of which discrete and faithful representations of @xmath1 into @xmath2 are separable - stable analogous to the jeon - kim - ohshika - lecuire criterion . a compression body @xmath0 is the boundary connect sum of a @xmath6-ball , a collection of trivial @xmath16-bundles over surfaces , and a handlebody , where the other components are connected to the @xmath6-ball along disjoint discs . its fundamental group is a free product of closed surface groups and a free group . if @xmath0 is not the boundary connect sum of two trivial @xmath16-bundles over closed surfaces , then an element in @xmath1 is separable if it lies in a proper factor of a free decomposition of @xmath4 if @xmath0 is the boundary connect sum of @xmath17 and @xmath18 where @xmath19 and @xmath20 are closed surfaces , then an element is separable if it lies in a factor of a decomposition @xmath21 where @xmath22 is homotopic to a simple closed curve on @xmath19 or @xmath20 . a homomorphism @xmath23 is separable - stable if every geodesic defined by a separable element in the cayley graph of @xmath24 is mapped to a uniform quasi - geodesic in @xmath25 . see section [ sep - stab ] for a precise definition . for any discrete and faithful representation @xmath26 we let @xmath27 denote the corresponding hyperbolic @xmath6-manifold obtained from taking the quotient @xmath28 the main theorem of this article is : [ thm : main ] let @xmath0 be a nontrivial hyperbolizable compression body without toroidal boundary components that is not the boundary connect sum of @xmath17 and @xmath18 where @xmath19 and @xmath20 are closed surfaces with genus at least two . let @xmath29 be a discrete and faithful representation of @xmath1 into @xmath2 . then , @xmath29 is separable - stable if and only if in @xmath30 each component of the parabolic loci and each ending lamination is disc - busting . a measured lamination @xmath31 is _ disc - busting _ if there exists @xmath32 such that @xmath33 for all unweighted meridians @xmath34 . otherwise it is _ disc - dodging_. we say that an ending lamination is disc - busting if it is the support of a measured lamination that is disc - busting . such a lamination is used by jeon - kim - ohshika - lecuire in @xcite for the characterization of primitive stable representations of free groups , and has a root in doubly incompressible laminations ( see kim - lecuire - oshika @xcite or lecuire @xcite ) . in @xcite , the second author showed that representations obtained by pinching masur domain curves or masur domain laminations on @xmath35 are separable - stable . for the case where @xmath0 is the boundary connect sum of @xmath36 and @xmath37 where @xmath19 and @xmath20 are closed surfaces , we establish the following analogous characterization of the set of discrete and faithful representations that are separable - stable . [ thm : special case ] let @xmath0 be the boundary connect sum of @xmath17 and @xmath18 where @xmath19 and @xmath20 are closed surfaces with genus at least two . let @xmath26 be a discrete and faithful representation . then , @xmath29 is separable - stable if and only if in @xmath30 each component of the parabolic loci and each ending lamination is annulus - busting . a measured lamination @xmath31 is _ annulus busting _ if there exists @xmath38 such that @xmath39 for any essential annulus @xmath40 a _ compression body _ is a compact , orientable , irreducible @xmath6-manifold @xmath0 with a boundary component , @xmath41 called the _ exterior boundary _ , such that the inclusion @xmath42 induces a surjection on the level of fundamental groups @xmath43 . the other boundary components are called _ interior boundary components_. equivalently , a compression body is the boundary connect sum of a @xmath6-ball , a collection of trivial @xmath16-bundles over surfaces , and a handlebody , where the other components are connected to the @xmath6-ball along disjoint discs . the fundamental group of a compression body can be decomposed as a free product , @xmath44 where @xmath45 is isomorphic to a closed surface group for @xmath46 and @xmath47 is infinite cyclic @xmath48 . by grushko s theorem ( @xcite ) and kurosh s subgroup theorem ( @xcite ) , any other decomposition of the fundamental group into a free product , @xmath49 where each factor is freely indecomposable , satisfies @xmath50 and @xmath51 , up to re - ordering . a compression body is _ trivial _ if @xmath52 is an isomorphism , i.e. if @xmath0 is a trivial @xmath16-bundle . we say that @xmath0 is _ uniquely freely decomposable _ if @xmath0 is the boundary connect sum of two trivial @xmath16-bundles over closed surfaces since , in this case , the decomposition of @xmath1 is essentially unique . an _ essential disc _ in @xmath0 is a properly embedded disc whose boundary is nontrivial in @xmath53 a compression body @xmath0 has the property that any splitting of @xmath1 as a free product or as an hnn - extension over the trivial group can be realized by an essential disc in @xmath0 in the following sense . suppose that @xmath54 is a separating essential disc and @xmath55 if @xmath56 for @xmath57 then @xmath58 where @xmath59 is inclusion and the base point is chosen to lie in @xmath60 in this case , we say that @xmath54 realizes the splitting of the fundamental group . as we can move the basepoint around , this is only well - defined up to conjugation . if @xmath54 is a non - separating essential disc and @xmath61 then @xmath62 we say that @xmath54 realizes this hnn - extension . [ lem : freeprod ] let @xmath0 be a compression body and let @xmath63 be a nontrivial splitting of @xmath1 into a free product . then , the splitting is realizable by an essential disc . [ lem : hnn ] let @xmath0 be a compression body . any splitting of the form @xmath64 is realizable by an essential disc . recall that @xmath65 is isomorphic to @xmath66 by lemma [ lem : freeprod ] , we can realize the latter splitting by an essential disc @xmath67 . then , @xmath67 separates @xmath68 into two components @xmath69 and @xmath70 where @xmath71 is conjugate to @xmath72 and @xmath70 is a compact , oriented , irreducible @xmath6-manifold with infinite cyclic fundamental group , i.e. a solid torus . then , the essential disc @xmath54 in the solid torus , realizes the original splitting . in a free group , an element is called _ primitive _ if it can be completed to a free generating set and an element is _ separable _ if it lies in a free factor . the notion of separability is originally due to stallings ( @xcite ) in his study of separable sets in free groups that generalized whitehead s ( @xcite ) study of primitive elements of free groups . although the notion of a primitive element does not generalize to other freely decomposable groups , the notion of a separable element does . if @xmath0 is a compression body that is not uniquely freely decomposable , then an element in @xmath1 is _ separable _ if it lies in a proper factor of a decomposition of @xmath1 into a nontrivial free product . if @xmath0 is the boundary connect sum of two trivial @xmath16-bundles over closed surfaces @xmath19 and @xmath20 , then an element in @xmath1 is _ separable _ if it lies in a factor of a decomposition @xmath73 where @xmath22 is homotopic to a simple closed curve on @xmath19 or @xmath20 one can think of separability in the following geometric way . if @xmath0 is not uniquely freely decomposable , by lemma [ lem : freeprod ] , an element @xmath74 of @xmath1 lies in a proper factor of a free decomposition if and only if its associated curve in @xmath0 is homotopic to one that misses an essential disc . similarly , lying in a factor of a decomposition @xmath73 where @xmath22 is homotopic to a simple closed curve on @xmath19 or @xmath20 is equivalent to missing the corresponding essential annulus @xmath75 up to homotopy . essential annulus _ in @xmath0 is a properly embedded incompressible annulus that is not homotopic , relative to its boundary , into the boundary of @xmath12 let @xmath24 denote the fundamental group of @xmath0 and let @xmath76 be a finite symmetric generating set for @xmath77 the _ cayley graph of @xmath24 with respect to @xmath76 _ , denoted @xmath78 , is a graph where the vertices are in one - to - one correspondence with the elements of @xmath24 and there is an edge between @xmath74 and @xmath79 if there exists an element @xmath80 in the generating set @xmath76 such that @xmath81 . the group @xmath24 acts on @xmath78 by left multiplication on the vertices . since @xmath24 is torsion - free and gromov hyperbolic each element @xmath74 of @xmath24 has two fixed points on @xmath82 . for each @xmath74 in @xmath24 , let @xmath83 and @xmath84 denote the repelling and attracting fixed points of @xmath74 acting on @xmath82 . let @xmath85 denote the set of geodesics connecting @xmath83 and @xmath84 and @xmath86 denote the set of geodesics @xmath87 in @xmath78 such that @xmath87 is contained in @xmath85 for some separable element @xmath74 . given a representation @xmath88 and a basepoint @xmath89 in @xmath25 , there exists a unique @xmath29-equivariant map @xmath90 taking the identity to @xmath89 and edges to geodesic segments . a representation @xmath88 is called _ @xmath91-separable - stable _ if there exists a basepoint @xmath89 in @xmath25 such that @xmath92 is a @xmath91-quasi - geodesic for any @xmath87 in @xmath93 we will call @xmath29 , a representation , _ separable - stable _ if there exists @xmath91 such that @xmath29 is @xmath91-separable - stable . separable - stability is independent of the choice of basepoint in @xmath25 and the choice of generators @xmath76 for @xmath24 , and separable - stability is invariant under conjugation ( see lemma 16 in lee @xcite ) . the set of separable - stable representations forms a domain of discontinuity for the action of @xmath3 strictly larger than the interior of @xmath13 . namely the following is true . let @xmath0 be a nontrivial compression body without toroidal boundary components . the outer automorphism group @xmath3 acts properly discontinuously on the set of separable - stable representations , which contains the interior of @xmath13 as well as points on the boundary of @xmath94 in this section , we describe otal s ( @xcite ) generalization of the whitehead graph for a free group to the whitehead graph for a compression body , which we will use in proving the sufficient condition . let @xmath0 be a nontrivial compression body without toroidal boundary components . a _ system of meridians _ @xmath95 on @xmath0 is a collection of pairwise nonisotopic and pairwise disjoint meridians @xmath96 bounding discs @xmath97 such that @xmath98 is a disjoint union of trivial @xmath16-bundles over closed surfaces , where @xmath99 is a regular neighborhood of @xmath97 . by thurston s hyperbolization theorem ( see kapovich @xcite ) and marden ( @xcite ) , we can fix @xmath100 a convex cocompact representation of @xmath1 such that @xmath101 is homeomorphic to @xmath0 , where @xmath102 is the conformal boundary of @xmath103 let @xmath104 be the limit set of @xmath105 let @xmath106 be a closed subset of @xmath107 that is @xmath108-invariant and also invariant under switching the two factors . in this paper , @xmath106 will be one of the following two examples . * if @xmath109 is a closed geodesic in @xmath110 , then @xmath111 is the set of endpoints of all the lifts of @xmath112 * if @xmath31 is a lamination on @xmath35 that is in tight position with respect to @xmath95 , then @xmath113 is the set of endpoints of all lifts of all leaves of @xmath114 see the discussion before proposition [ discbusting ] for the definition of a lamination in tight position and lemma [ endpoints ] to see why @xmath113 is well - defined . identify @xmath35 with @xmath102 . let @xmath95 be a system of meridians bounding the discs @xmath115 and suppose @xmath116 is the disjoint union of @xmath117 where each @xmath118 is a closed surface of genus at least two . the _ whitehead graph for @xmath0 of @xmath106 with respect to @xmath95 _ , denoted @xmath119 is a collection of not necessarily connected graphs , @xmath120 where the elements in the collection are in one - to - one correspondence with the components of @xmath116 . in @xmath116 , there are two copies @xmath121 and @xmath122 of each @xmath97 in @xmath54 . given a component @xmath123 of @xmath116 , the vertices in the corresponding graph @xmath124 are in one - to - one correspondence with the components of @xmath125 in the frontier of @xmath123 . abusing notation , relabel the vertices @xmath126 fix a jordan curve @xmath127 in @xmath128 that is invariant under a conjugate of @xmath129 , which we will continue to denote @xmath129 . let @xmath130 denote the boundary component of @xmath123 coming from @xmath131 . fix a lift @xmath132 of each @xmath133 in @xmath134 such that @xmath135 lies in the component of the preimage of @xmath130 in @xmath134 containing @xmath127 on its boundary . let @xmath136 be the open set in @xmath134 , bounded by @xmath132 not containing @xmath127 . the edges from @xmath137 to @xmath138 will be in one - to - one correspondence with elements @xmath74 in @xmath129 such that @xmath139 is nonempty . we will denote such an edge @xmath140 . although these edges are directed , for each edge from @xmath136 to @xmath141 labeled @xmath74 , there is an edge from @xmath141 to @xmath136 labeled @xmath142 . [ otal ] a connected component of @xmath143 is _ strongly connected _ if there exists a cycle that represents a nontrivial element of @xmath129 . a connected component of @xmath143 has a _ strong cutpoint _ if we can express the graph as the union of two graphs @xmath144 and @xmath145 that intersect in a single vertex such that either @xmath144 or @xmath145 is not strongly connected . we take the convention that a cycle @xmath146 corresponds to the group element @xmath147 . although we made several choices when defining @xmath148 , the two properties defined above , strong connectedness and the presence of a strong cutpoint are independent of these choices ( see section 3.1.2 in lee @xcite ) . otal in @xcite , also , gave a dichotomy between the whitehead graph of a separable element and the whitehead graph of a disc - busting lamination in tight position with respect to a system of meridians . in this section , we define and collect the relevant properties of a lamination in tight position . the results in this section are essentially all due to otal in @xcite . since @xcite is difficult to obtain for each result we either give an alternate source containing a proof or we provide a proof . a lamination @xmath31 is in _ tight position with respect to a system of meridians @xmath95 _ if there are no waves disjoint from @xmath31 where a _ wave _ is an arc @xmath149 satisfying the following two conditions : 1 . @xmath149 has endpoints on @xmath95 , but its interior is disjoint from @xmath95 and , 2 . @xmath149 is homotopic , relative to its endpoints , in @xmath0 but not in @xmath35 into @xmath150 for any disc - busting lamination @xmath31 we can find a system of meridians @xmath95 such that @xmath31 is in tight position with respect to @xmath150 otal proved this for masur domain laminations but his proof applies in this more general case . [ discbusting ] if @xmath31 is a disc - busting lamination , there exists a system of meridians @xmath95 with respect to which @xmath31 is in tight position . suppose that @xmath31 is not in tight position with respect to a system of meridians @xmath95 . let @xmath149 be a wave and suppose its endpoints lie on @xmath151 then the endpoints of @xmath149 split @xmath96 into two arcs @xmath152 and @xmath153 such that @xmath154 and @xmath155 are the endpoints of @xmath156 since @xmath149 is homotopic into @xmath96 through @xmath0 but not in @xmath157 we have that @xmath158 and @xmath159 are both meridians . since @xmath31 is disc - busting , @xmath160 and @xmath161 by construction both @xmath162 and @xmath163 are disjoint from @xmath164 for any @xmath165 replacing @xmath96 with either @xmath162 or @xmath163 will produce a new collection of meridians @xmath166 and @xmath167 respectively , whose intersection number with @xmath31 is at least @xmath168 less than @xmath169 we claim that at least one of these new collections of meridians @xmath166 or @xmath167 is a system of meridians . by construction @xmath170 and @xmath163 bound a pair of pants on @xmath53 since they are all meridians and @xmath0 is irreducible , the discs they bound , @xmath171 and @xmath172 bound a ball in @xmath0 . so @xmath173 contains a component that is a ball @xmath174 . we will use a @xmath175 superscript to denote the copy of @xmath97 ( or @xmath176 ) that lies on the boundary of the ball , i.e. @xmath177 lie on the boundary of @xmath178 since @xmath179 is a collection of trivial @xmath16-bundles over closed surfaces , @xmath180 must consist of a collection of trivial @xmath16-bundles over closed surfaces and two balls , one of which is the ball @xmath174 described above . the other ball must contain either @xmath181 or @xmath182 on its boundary ( possibly both ) . if it contains @xmath183 then @xmath167 will be a system of meridians , and vice versa . so , replacing @xmath96 with one of @xmath162 and @xmath163 results in a system of meridians whose intersection number with @xmath31 is at least @xmath168 less than @xmath169 if @xmath31 is not in tight position with respect to this new system of meridians , repeat this process . since @xmath184 for any system of meridians @xmath185 , eventually , this process must terminate . an important property of disc - busting , minimal laminations in tight position with respect to a system of meridians @xmath95 is that any lift of a leaf has well - defined endpoints . [ endpoints ] let @xmath0 be a nontrivial compression body . let @xmath186 be a convex cocompact representation uniformizing @xmath0 . let @xmath31 be a lamination on @xmath102 in tight position with respect to @xmath150 let @xmath87 be a leaf @xmath31 such that any half leaf of @xmath87 intersects @xmath150 then , any lift @xmath187 of @xmath87 to @xmath188 has two well - defined endpoints in @xmath189 propositions [ discbusting ] and lemma [ endpoints ] imply that if @xmath31 is a disc - busting ending lamination , then @xmath113 , the set of endpoints of lifts of leaves of @xmath190 is well - defined . in particular , we will refer to the whitehead graph of @xmath31 as the whitehead graph of @xmath191 otal proved the following two results , which provide a dichotomy between separable elements and disc - busting laminations . [ separableelement ] let @xmath0 be a nontrivial compression body that is not the boundary connect sum of two trivial @xmath16-bundles over closed surfaces . if @xmath74 is separable , then for any system of meridians @xmath95 some connected component of @xmath192 is either not strongly connected , or has a strong cut - point . [ otal ] if @xmath31 is a disc - busting lamination in tight position with respect to a system of meridians @xmath193 then the whitehead graph of @xmath31 with respect to @xmath95 is strongly connected and without strong cutpoints . in @xcite , otal states proposition [ otal ] only for masur domain laminations , but in fact his proof applies in this more general case . in general , for @xmath194 a discrete and faithful representation , the associated hyperbolic @xmath6-manifold @xmath27 will be homotopy equivalent but not necessarily homeomorphic to @xmath12 since otal s results only apply in the case when @xmath0 is a compression body , in this section we associate a whitehead graph to any hyperbolizable @xmath6-manifold @xmath195 homotopy equivalent , but not necessarily homeomorphic to @xmath0 . we can do this by using the characteristic compression body introduced by bonahon in @xcite . we also describe how to compare whitehead graphs for @xmath0 and @xmath195 . let @xmath196 denote the set of compact , orientable , irreducible , atoroidal ( unmarked ) @xmath6-manifolds homotopy equivalent to @xmath0 , up to homeomorphism . let @xmath197 , \\ldots , [ m_n]$ ] be the elements in @xmath198 where @xmath199 is homeomorphic to @xmath12 for each @xmath200 fix a convex cocompact representation @xmath201 such that @xmath202 is homeomorphic to the interior of @xmath203 let @xmath204 be an orbit map for @xmath201 as in the definition of separable - stable representations where @xmath205 . then , @xmath206 is a quasi - isometric embedding and there is a continuous extension , @xmath207 which is a homeomorphism . we suppressed the basepoint in @xmath206 since @xmath208 is independent of the choice of basepoint . then , let @xmath209 be the homeomorphism from @xmath210 to @xmath211 in order to compare the whitehead graph of @xmath199 and @xmath200 we need a way to compare systems of meridians on the two manifolds . define a map @xmath212 from the set @xmath213 of unweighted meridians on @xmath200 to the set @xmath214 of unweighted meridians on @xmath199 in the following way ( see figure [ pic ] ) . on the characteristic compression body @xmath215 of @xmath200 is mapped to the system of meridians @xmath216 on @xmath217,width=491 ] let @xmath218 denote the essential disc bounded by @xmath150 first suppose that @xmath95 is a meridian in @xmath200 such that @xmath218 separates @xmath203 then , @xmath218 realizes a splitting @xmath219 up to conjugation . in @xmath199 we can realize the splitting by an essential disc @xmath67 . let @xmath220 be the boundary of @xmath221 this is well - defined up to isotopy . indeed if we consider @xmath220 on the conformal boundary of @xmath222 then some lift @xmath223 separates @xmath224 into two components one of which contains all the attracting fixed points of any element whose reduced word representation starts with an element in @xmath72 and the other contains all the attracting fixed points of any element whose reduced word representation starts with an element in @xmath178 since such points are dense in @xmath225 this determines @xmath226 up to isotopy on @xmath227 now suppose that @xmath218 is non - separating . then , @xmath218 realizes an hnn extension , @xmath228 by lemma [ lem : hnn ] , we can realize such a splitting in @xmath199 by an essential disc @xmath67 . as before , let @xmath220 be the boundary of @xmath221 again , to see that @xmath220 is well - defined notice that some lift @xmath229 separates @xmath134 into two components one of which contains all the attracting fixed points of any element whose reduced word representation starts with an element of @xmath72 or @xmath230 and the other contains all the attracting fixed points of any element whose reduced word representation starts with @xmath231 where @xmath230 is the new generator associated to the hnn - extension . any manifold @xmath0 with compressible boundary has a _ characteristic compression body _ @xmath232 , unique up to isotopy that has the following properties ( see bonahon @xcite or mccullough - miller @xcite ) . the exterior boundary of @xmath232 coincides with @xmath35 . the closure of @xmath233 contains no essential compression discs and no component of @xmath234 is a 3-ball . let @xmath215 denote the characteristic compression body of @xmath235 with any trivial compression body components removed . [ lem : mermap ] let @xmath236 be a system of meridians on @xmath237 then , the collection of meridians @xmath238 is a system of meridians on @xmath217 first we will show that the meridians @xmath238 can be realized such that they are pairwise disjoint and pairwise nonisotopic . to see that we can realize @xmath239 and @xmath240 disjointly , observe that since @xmath164 and @xmath241 are disjoint , any two lifts @xmath242 and @xmath243 , have the property that there exist @xmath141 a component of @xmath244 and @xmath245 a component of @xmath246 such that @xmath247 and @xmath248 are each non - empty and disjoint from each other . there exists lifts @xmath249 and @xmath250 and components @xmath251 and @xmath252 of @xmath253 and @xmath254 respectively , such that @xmath255 and @xmath256 then , we can homotope , @xmath239 and @xmath240 to be disjoint . to see that @xmath239 and @xmath240 are not isotopic , notice that if they were isotopic , then there exists lifts @xmath249 and @xmath250 that bound a region in @xmath134 containing no limit points . then , the same would be true about the corresponding lifts @xmath242 and @xmath257 which contradicts that they are not isotopic . let @xmath258 be the disc bounded by @xmath239 in @xmath217 finally , we want to see that @xmath259 is a collection of trivial @xmath16-bundles over closed surfaces . we will first show that @xmath260 is a collection of trivial @xmath16-bundles over closed surfaces . recall that @xmath261 where @xmath215 is the characteristic compression body and @xmath262 has incompressible boundary . by definition , @xmath263 is a collection of trivial @xmath16-bundles over closed surfaces . we claim that the components of @xmath262 are all trivial @xmath16-bundles over closed surfaces . let @xmath127 be a component of @xmath264 and let @xmath265 be a component of @xmath266 since @xmath127 has incompressible boundary and can not be a @xmath6-ball , we have that @xmath267 lies in a surface group factor of some maximal free decomposition of @xmath1 i.e. @xmath267 is a subgroup of some @xmath268 . the image of @xmath269 must be finite index in @xmath268 since they are both closed surface groups . this implies that @xmath269 is finite index in @xmath270 which in turn implies that @xmath127 is an @xmath16-bundle ( theorem 10.5 in hempel @xcite ) . so @xmath127 is either a trivial @xmath16-bundle or twisted @xmath16-bundle . it remains to check that @xmath127 can not be a twisted @xmath16-bundle . if @xmath127 were a twisted @xmath16-bundle , then the fundamental group of @xmath127 would be the fundamental group of a non - orientable surface , but this is impossible as @xmath267 is finite index in @xmath271 so , we have shown that the components of @xmath262 are all trivial @xmath16-bundles over surfaces . now @xmath272 is formed by gluing @xmath262 and @xmath263 along their boundaries , so the result will be a collection of trivial @xmath16-bundles over closed surfaces . now , consider @xmath273 a component of @xmath274 . we want to see that @xmath273 is a trivial @xmath16-bundle over a closed surface . if not , then either @xmath273 is a @xmath6-ball , or @xmath273 has compressible boundary . in the former case , let @xmath238 be the components of @xmath220 that lie on the boundary of @xmath273 ( notice that we can have two copies of @xmath164 on the boundary of @xmath273 , relabel accordingly ) . fix a lift @xmath275 of @xmath276 then , the corresponding lifts @xmath277 bound a connected region in @xmath134 containing no limit points . the corresponding lifts of @xmath278 of @xmath95 would have the same property . this implies that there is a @xmath6-ball component of @xmath279 a contradiction . if @xmath273 has compressible boundary , then there exists a closed curve @xmath109 in @xmath273 such that @xmath109 does not lie in any surface group factor of a free decomposition of @xmath4 moreover , any lift @xmath280 of @xmath109 has the property that both endpoints lie in the same component of @xmath253 for any lift of any @xmath281 since @xmath109 misses any @xmath282 then , we would also have that @xmath280 has both endpoints in the same component of @xmath283 so , @xmath109 misses all the meridians @xmath281 a contradiction , since @xmath284 consists of trivial @xmath16-bundles over closed surfaces . using lemma [ lem : mermap ] we can compare the whitehead graphs of @xmath215 and @xmath217 recall from section [ white ] the definition of the whitehead graph of a compression body . for @xmath285 we will abuse notation , and let @xmath286 denote the image of @xmath106 under @xmath287 [ whotherhomeo ] let @xmath200 lie in @xmath196 and let @xmath215 be the characteristic compression body of @xmath203 if @xmath288 is strongly connected and without strong cutpoint , then @xmath289 is strongly connected and without strong cutpoint . recall that we have a @xmath24-equivariant homeomorphism @xmath290 although @xmath291 is only defined on the limit set of @xmath292 it induces a natural map on lifts of @xmath293 to lifts of @xmath294 for example , given a lift @xmath242 of @xmath164 , there exists a corresponding lift @xmath249 of @xmath295 that partitions @xmath296 in the same way as @xmath242 via @xmath297 we ll abuse notation to denote such a lift @xmath298 as @xmath299 more precisely , suppose that @xmath300 then , @xmath301 is the lift @xmath249 of @xmath239 such that if @xmath302 then @xmath303 where @xmath304 up to switching @xmath305 and @xmath306 suppose that @xmath123 is a component of @xmath307 recall that when defining the whitehead graph , we fixed @xmath127 a jordan curve in @xmath134 invariant under @xmath129 . we let @xmath130 denote the boundary component of @xmath123 intersecting @xmath95 and we fixed the lift @xmath308 of @xmath130 whose boundary is @xmath309 if @xmath242 corresponds to a vertex of @xmath310 then map it to @xmath299 we want to show that this is a well - defined map on the vertices . first notice that since @xmath311 is a system of meridians , @xmath312 consists of trivial @xmath16-bundles over surfaces . one of these components @xmath313 must have fundamental group conjugate to @xmath314 by the uniqueness of the free decomposition of @xmath1 ( see section [ sec : compbody ] ) . let @xmath315 be the jordan curve invariant under @xmath129 . note that @xmath316 let @xmath317 be the component of @xmath123 intersecting @xmath220 and let @xmath318 denote the component of the preimage of @xmath317 with @xmath315 on its boundary . then , we claim that @xmath301 lies on @xmath319 indeed , this is clear from the fact that @xmath301 partitions @xmath296 in the same way as @xmath242 via @xmath320 as described above . so we see that the map on the vertices is well - defined . now suppose that @xmath321 is an edge in @xmath322 this implies that the intersection of @xmath106 and @xmath323 is nonempty . by construction , this is true if and only if the same is true for @xmath286 and @xmath324 so , we have a bijection between the whitehead graph of @xmath288 and that of @xmath325 where an edge in @xmath288 labeled @xmath74 is mapped to an edge in @xmath325 with the same label . in particular , we see that @xmath288 is strongly connected and without cutpoint if and only if the same is true for @xmath326 recently , mj ( @xcite ) has proved the existence of cannon - thurston maps for general kleinian groups . in particular , given a kleinian group @xmath24 and a discrete and faithful representation @xmath327 , he showed that there exists a continuous extension @xmath328 of @xmath329 moreover , he gives a characterization , which we describe below , of which points are mapped non - injectively by @xmath330 . before stating mj s result we recall some facts from hyperbolic geometry that we will need . we will restrict to the case that @xmath24 has no non - cyclic abelian subgroups since this is the case in this paper . there exists a constant @xmath331 called the margulis constant , such that for any hyperbolic @xmath6-manfiold , and any @xmath332 each component of the @xmath333-thin part of @xmath334 is a metric neighborhood of a closed geodesic or a parabolic cusp homeomorphic to @xmath335 given @xmath336 the _ non - cuspidal part of @xmath27 _ , denoted @xmath337 is the submanifold of @xmath27 obtained by removing the noncompact components of the @xmath333-thin part of @xmath338 when @xmath339 has parabolics , @xmath340 has a _ relative compact core , _ @xmath341 that is a compact submanifold whose inclusion is a homotopy equivalence and that intersects each cusp neighborhood along an annulus whose core curve we call a _ parabolic curve _ ( mccullough @xcite ) . let @xmath342 denote the union of these annuli . moreover , by the tameness ( agol @xcite or calegari - gabai @xcite ) and the uniqueness of cores ( mccullough - miller - swarup @xcite ) @xmath27 is homeomorphic to the interior of @xmath127 . each frontier component @xmath343 of @xmath344 in @xmath345 faces an end @xmath346 with a well - defined ending lamination @xmath347 obtained from taking the support of a projective limit of any sequence of simple closed geodesics exiting the end @xmath348 the ending lamination @xmath349 is a filling minimal lamination on @xmath343 ( canary @xcite ) . let @xmath350 be the union of @xmath349 and the parabolic curves , and assume that each is disc - busting . if we fix a convex cocompact structure @xmath110 on @xmath127 , then we can identify @xmath351 with a geodesic lamination on @xmath102 . by lemma [ endpoints ] any lift @xmath187 of a leaf @xmath87 of @xmath349 for any @xmath59 has two well defined endpoints at infinity on @xmath128 . since @xmath100 is convex cocompact , we can identify @xmath82 with @xmath128 . we introduce a relation @xmath352 for points in @xmath128 ( and so also on @xmath82 ) such that @xmath353 if and only if either @xmath354 and @xmath355 are the endpoints of a leaf of @xmath356 or ideal vertices of a complementary region of @xmath356 . in contrast to the case of groups without parabolics , there are complementary regions of @xmath357 which are ideal polygons with infinitely many sides . these are exactly regions touching lifts of parabolic curves , which are isolated leaves . the relation @xmath358 may not be transitive . we define @xmath359 to be the transitive closure of @xmath352 , which is an equivalence relation . mj s theorem about the identified points of a cannon - thurston map can be adapted to our case as follows . [ mj2 ] let @xmath23 be a discrete and faithful representation . the identification of @xmath78 with its image under @xmath360 extends continuously to a map @xmath361 . moreover , for @xmath362 , we have @xmath363 if and only if @xmath364 . in this section , we use the dichotomy between the whitehead graph of a separable element and a disc - busting lamination to show the sufficient condition of theorem [ thm : main ] , namely that when @xmath1 is not uniquely freely decomposable , a representation is separable - stable if each end invariant is disc - busting . for the sufficient condition , we ll use mj s result described in the previous section . the following lemma , whose proof can be found in lee @xcite lemma 21 , sets the stage for using mj s result . [ notss ] let @xmath26 be discrete and faithful . if @xmath29 is not separable - stable , then there exists a sequence of separable elements @xmath365 such that if @xmath366 and @xmath367 are the endpoints of @xmath368 then @xmath369 and @xmath370 converge , up to subsequence , to the same point in @xmath371 but @xmath366 and @xmath367 converge to @xmath372 and @xmath373 in @xmath374 where @xmath375 [ sufficient ] let @xmath0 be a nontrivial compression body , without toroidal boundary components , that is not the boundary connect sum of two trivial @xmath16-bundles over closed surfaces . let @xmath26 be a discrete and faithful representation . if each parabolic locus and ending lamination of @xmath27 is disc - busting , then @xmath29 is separable - stable . notice that the statement of the above proposition does not make any assumption about the homeomorphism type of @xmath338 suppose that @xmath27 is homeomorphic to the interior of @xmath376 where @xmath377 \\in a(m).$ ] let @xmath378 be the union of the parabolic loci and ending laminations on @xmath379 if @xmath29 is not separable - stable , then by lemma [ notss ] there exists a sequence of separable elements @xmath365 such that the fixed points of @xmath365 in @xmath82 converge to @xmath372 and @xmath380 but the endpoints of @xmath381 converge to the same point in @xmath382 by theorem [ mj2 ] we have that @xmath383 and @xmath384 are each an endpoint of a leaf of one of the end invariants , although not necessarily the same end - invariant . suppose that @xmath383 contains an endpoint of a leaf of @xmath385 let @xmath386 be the set of limit points of @xmath387 then @xmath388 is @xmath389-invariant and also invariant under switching the two factors . moreover , @xmath390 lies in @xmath391 this implies that @xmath392 contains @xmath393 for any system of meridians @xmath95 ( see the proof of proposition 22 in lee @xcite ) . fix a system of meridians @xmath95 such that @xmath31 is in tight position with respect to @xmath150 then , @xmath393 is strongly connected and without any strong cutpoints by proposition [ otal ] . by lemma [ whotherhomeo ] , the whitehead graph @xmath394 is strongly connected and without strong cutpoint . observe that @xmath393 is a finite graph since edges correspond to homotopy classes of arcs in @xmath190 so @xmath394 is also finite . since the graph @xmath392 contains @xmath393 , for @xmath59 large enough @xmath395 contains the whitehead graph @xmath396 this implies @xmath397 contains @xmath398 for @xmath59 large enough . but then the whitehead graph @xmath397 is strongly connected and without strong cutpoint . notice , that @xmath399 is the same as @xmath400 this contradicts proposition [ separableelement ] . in this section , we show that when @xmath1 is not uniquely freely decomposable , if @xmath26 is discrete and faithful , then the condition that each end invariant of @xmath27 is disc - busting is also necessary . notice that if each end invariant is disc - busting , then , in particular , @xmath27 contains only one compressible boundary component . we will make use of the following characterization of which discrete and faithful representations are separable - stable . [ compact ] let @xmath26 be discrete and faithful . then , @xmath29 is separable - stable if and only if every separable curve is homotopic to a closed geodesic and the set of separable geodesics is contained in a compact set of @xmath338 if @xmath31 is a minimal lamination that is not a simple closed curve , then @xmath401 the _ supporting surface of _ @xmath190 is the unique minimal compact subsurface with geodesic boundary containing @xmath114 let @xmath402 be the unique , up to isotopy , maximum , simple , multi - curve of @xmath403 disjoint from @xmath114 the multi - curve @xmath402 contains @xmath404 . [ necessary ] let @xmath0 be a nontrivial compression body , without toroidal boundary components , that is not the boundary connect sum of two trivial @xmath16-bundles over closed surfaces . suppose @xmath405 is a discrete and faithful representation . if @xmath29 is separable - stable , then each end invariant of @xmath27 is disc - busting . the proof of the necessary condition follows almost exactly the proof in the case when @xmath0 is a handlebody from jeon - kim - ohshika - lecuire @xcite ( see section 5 ) . we include it here for completeness . by lemma [ compact ] , if a representation maps a separable element of @xmath24 to a parabolic element in @xmath406 then the representation is not separable - stable . since any disc - dodging , closed curve is separable , if there is a disc - dodging parabolic curve in @xmath27 , then @xmath29 is not separable - stable . so we may assume that each parabolic curve is disc - busting . similarly , if there exists an ending lamination @xmath31 that misses an essential disc @xmath54 , then @xmath31 is the limit of separable elements . in particular , there exists a sequence of separable elements exiting every compact set of @xmath338 by lemma [ compact ] , @xmath29 is not separable - stable . so , we may also assume that @xmath407 for any essential disc @xmath54 and any ending lamination @xmath31 of @xmath338 then , proposition [ necessary ] follows from the following two lemmas whose proofs we include below . we first explain how the necessary condition follows from the following two lemmas , and we include the proofs of the lemmas , afterwards . [ esan ] let @xmath0 have compressible boundary . for any essential annulus @xmath72 in @xmath0 , there is a meridian which is disjoint from @xmath72 . [ limitannuli ] let @xmath26 be discrete and faithful . suppose that @xmath31 is a disc - dodging ending lamination of @xmath27 . if some component @xmath22 of @xmath402 is disc - busting , then @xmath31 is the unique minimal component in the hausdorff limit of a sequence @xmath408 where @xmath409 is an essential annulus . indeed , by lemma [ limitannuli ] , if @xmath27 has an ending lamination that is disc - dodging , then there exists a sequence of simple closed curves @xmath410 , that are homotopic to core curves of essential annuli , such that @xmath411 exits every compact set . by lemma [ esan ] these curves are separable . hence , there exists a sequence of separable curves exiting every compact set . by lemma [ compact ] , the representation is not separable - stable . let @xmath72 be an essential annulus in @xmath12 since @xmath0 has compressible boundary there exists an essential disc @xmath60 we can isotope @xmath54 so that there are no inessential intersections between @xmath54 and @xmath72 . if @xmath412 , then the boundary of @xmath54 is a meridian disjoint from @xmath40 if not , consider an outermost arc @xmath149 in @xmath413 then , @xmath149 together with an arc in @xmath414 bounds a disc @xmath415 . if both of the endpoints of @xmath149 lie on the same component of @xmath416 , then together with an arc in @xmath417 the arc @xmath149 cuts off a disc @xmath418 from @xmath419 and @xmath420 is an essential disc which can be isotoped off @xmath72 . if @xmath149 connects two components of @xmath416 , then we can boundary - compress @xmath72 along @xmath415 , and to get an essential disc disjoint from @xmath40 the proof of lemma [ limitannuli ] makes use of a result of lecuire in @xcite that states the following . if @xmath0 is a compact hyperbolizable @xmath6-manifold , and there exists a lamination @xmath421 such that * each closed leaf has weight at most @xmath422 and * for any compressing disc @xmath54 , we have @xmath423 then , the following two conditions are equivalent . 1 . there exists @xmath168 such that @xmath424 for any essential annulus @xmath40 2 . if @xmath425 and @xmath426 are disjoint geodesic rays on @xmath427 such that two lifts @xmath428 and @xmath429 share the same endpoint in @xmath371 then they are asymptotic . the first step is finding a lamination containing @xmath31 that satisfies the first two conditions ( a ) and ( b ) . let @xmath430 be the union of @xmath31 and @xmath431 where we place a weight of @xmath422 on each closed curve . if there exists an essential annulus @xmath72 that is disjoint from @xmath432 then cut @xmath0 along @xmath72 and let @xmath433 be the component containing @xmath114 we claim that @xmath31 is still disc - dodging on @xmath434 indeed , suppose that @xmath435 is a sequence of meridians on @xmath0 such that @xmath436 then , as in the proof of lemma [ esan ] we can find a sequence of meridians @xmath437 disjoint from @xmath419 such that @xmath438 hence , @xmath31 is still disc dodging on @xmath434 let @xmath439 be the union of @xmath430 and the core curve or curves of @xmath419 depending on whether @xmath72 separates @xmath0 . put a weight of @xmath422 on each core curve that is added . notice that this new sequence of meridians also has the property that @xmath440 if there exists an essential annulus on @xmath433 disjoint from @xmath441 we repeat the procedure above . since there exists at most finitely many disjoint non - isotopic essential annuli on @xmath68 this procedure must terminate . let @xmath442 be the manifold obtained at the end of this procedure , and let @xmath443 be the corresponding lamination . then , any closed leaf has weight at most @xmath444 by construction . to see that @xmath445 for any essential disc @xmath446 it suffices to show that @xmath414 intersects an element in @xmath402 twice . indeed , this implies that @xmath447 and since @xmath448 we have that @xmath449 since there is a disc - busting component @xmath22 of @xmath431 we know that @xmath22 intersects @xmath54 at least once . if @xmath22 intersects @xmath54 exactly once , then a regular neighborhood of @xmath450 is a solid torus , @xmath451 we can form an essential disc missing @xmath22 by taking the closure of @xmath452 , a contradiction since @xmath22 is disc - busting . so @xmath443 satisfies conditions @xmath453 and @xmath454 above . secondly , we want to find two geodesic rays satisfying the hypothesis of condition @xmath455 above . namely , we want to find two disjoint , geodesic rays on @xmath456 that have lifts with a common endpoint in @xmath382 recall that there exists a sequence of meridians @xmath457 such that @xmath458 and @xmath459 up to subsequence , @xmath457 converges to a geodesic lamination @xmath460 in the hausdorff topology . by casson s criterion ( see casson - long @xcite or theoreme b1 in lecuire @xcite ) , @xmath106 contains a homoclinic leaf @xmath461 we want to show that @xmath79 has two half - leaves @xmath462 and @xmath463 that are disjoint from @xmath464 then , if @xmath22 is the disc - busting component in @xmath431 we have that @xmath462 and @xmath463 will lie on @xmath465 since @xmath466 is an incompressible subsurface , any lift @xmath467 has a well - defined endpoint in @xmath468 where @xmath469 is the subgroup of @xmath470 corresponding to the subsurface @xmath471 since @xmath79 is homoclinic , @xmath462 and @xmath463 must have the same endpoint . to see why two such half - leaves exist , first note that if @xmath79 intersects @xmath31 transversely , then it contains an arc @xmath149 such that @xmath472 this implies that for @xmath59 large enough , @xmath473 which is a contradiction since @xmath474 so @xmath79 does not intersect @xmath31 transversely . it follows that @xmath79 has two half - leaves @xmath475 that are disjoint from @xmath476 similarly , if @xmath79 intersects a leaf @xmath109 of @xmath443 outside of @xmath401 then @xmath477 for @xmath59 large enough , a contradiction since @xmath459 so @xmath79 has two half - leaves @xmath475 that are disjoint from @xmath464 next , we want to see that @xmath443 does not satisfy condition ( c ) above . indeed we want to show that if we take lifts @xmath478 and @xmath479 with the same endpoint , then they are not asymptotic . suppose they are asymptotic and take a sequence of geodesic arcs @xmath480 joining @xmath478 and @xmath479 such that the length of @xmath480 goes to zero . let @xmath481 be the projection of @xmath480 to @xmath482 . then @xmath483 since @xmath31 is minimal and not a simple closed curve . since @xmath478 and @xmath479 are asymptotic and do not transversely intersect @xmath31 we can assume that @xmath481 are homotopic relative to their endpoints and that the homotopy is through arcs transverse to @xmath114 then , we must have @xmath484 for all @xmath485 . in particular , the arc @xmath486 is disjoint from @xmath114 then , adjusting @xmath487 so that it begins at @xmath488 , if we take @xmath489 this projects to a nontrivial curve @xmath34 on @xmath482 disjoint from @xmath31 that is trivial in @xmath442 . it follows from the proof of dehn s lemma that there is a compressing disc disjoint from @xmath190 a contradiction . since condition ( c ) is not true , we have that condition ( c ) is also not true . in particular , there exists a sequence @xmath409 of essential annuli such that @xmath490 since each component of @xmath491 is a closed curve with weight @xmath422 , we have that @xmath492 for @xmath59 large enough . up to subsequence , @xmath493 converges , in the hausdorff topology , to a geodesic lamination @xmath494 then , @xmath495 and @xmath31 can not intersect transversely . since @xmath496 we can not have that @xmath495 and @xmath31 are disjoint . since @xmath492 for @xmath59 large enough , we have that @xmath497 in particular , @xmath495 lies on @xmath498 since @xmath31 is filling on @xmath499 , we have that @xmath31 and @xmath495 coincide as geodesic laminations . since @xmath493 is disjoint from all the essential annuli used to construct @xmath442 , we can view @xmath495 as a lamination on @xmath35 and we are done . this section deals with the missing case when @xmath1 is uniquely freely decomposable , i.e. when @xmath0 is the boundary connect sum of @xmath17 and @xmath18 where @xmath343 is a closed surfaces of genus at least two for @xmath500 here an element of @xmath1 is separable if and only if it misses an essential annulus in one of the two surface group factors . by lemma [ esan ] for any essential annulus @xmath419 there exists an essential disc disjoint from @xmath40 since there is only one essential disc , @xmath446 up to isotopy , in fact , @xmath72 lies in one of the two surface group factors . if @xmath72 is contained in @xmath501 then @xmath72 is either an essential annulus in @xmath502 or homotopic ( relative to its boundary ) into the boundary of @xmath503 if the latter is true , since @xmath72 is essential in @xmath68 if @xmath504 and @xmath72 lies in @xmath505 then @xmath416 bounds an annulus in @xmath506 containing @xmath60 in either case , if a curve or lamination misses @xmath419 it misses an essential annulus in @xmath17 of @xmath507 we will consider annulus - busting laminations on @xmath35 . recall that a measured lamination @xmath31 is annulus - busting if there exists an @xmath168 such that @xmath508 for any essential annulus @xmath40 let @xmath0 be the boundary connect sum of @xmath17 and @xmath18 where @xmath19 and @xmath20 are closed surfaces of genus at least two . let @xmath26 be discrete and faithful . then , @xmath29 is separable - stable if and only if each end invariant of @xmath27 is annulus - busting . we will start with the sufficient condition , and we will follow the general outline of proposition [ sufficient ] . in this case , we have the following simplifications . firstly , @xmath196 has only one element , i.e. any manifold homotopy equivalent to @xmath0 is homeomorphic to @xmath0 . secondly , we have only one meridian @xmath95 . if @xmath31 is annulus - busting , then it must intersect @xmath193 and so it is automatically in tight position with respect to @xmath150 now suppose that @xmath26 is a discrete and faithful representation such that each end invariant and parabolic locus of @xmath27 is annulus - busting . if @xmath29 is not separable - stable , by lemma [ notss ] , there exists a sequence of separable elements @xmath365 with endpoints @xmath509 such that @xmath510 and @xmath511 then , by theorem [ mj2 ] , the image @xmath512 is an endpoint of one of the end invariants @xmath31 . here @xmath31 can either be an ending lamination or a simple closed curve that is mapped to a parabolic element under @xmath513 as before , this implies that for @xmath59 large enough @xmath514 contains @xmath515 if @xmath31 is annulus - busting , then @xmath516 intersects each essential annulus in the following sense . suppose that @xmath72 is an annulus in @xmath503 let @xmath517 . in defining the whitehead graph , we fixed lifts @xmath518 of @xmath343 . if we take a lift @xmath519 of @xmath520 and the lift @xmath521 of @xmath522 with the same endpoints as @xmath519 , then @xmath523 forms a loop in @xmath134 . we will say that an edge @xmath524 intersects @xmath72 if there exists lifts @xmath519 and @xmath521 in @xmath518 as above such that @xmath136 and @xmath525 lie in different components of @xmath526 . since @xmath514 contains @xmath516 for @xmath59 large enough , we have that @xmath514 `` intersects '' any essential annulus . this implies that the geodesic representative @xmath410 of @xmath365 intersects any essential annulus , a contradiction . for the necessary condition , suppose that @xmath527 is a discrete , faithful and separable - stable representation and some end invariant of @xmath27 is annulus - dodging . if a parabolic curve @xmath22 is annulus - dodging , then @xmath22 is separable , a contradiction to lemma [ compact ] . so suppose that all parabolic curves are annulus - busting , but there is an ending lamination @xmath31 that is annulus - dodging . if @xmath31 is disjoint from an essential annulus @xmath419 then we can find a sequence of separable elements approaching @xmath190 a contradiction to lemma [ compact ] . so we can assume that @xmath528 for any essential annulus . since @xmath31 is annulus - dodging , there exists a sequence @xmath409 of essential annuli such that @xmath529 since each essential annulus misses the essential disc , @xmath446 up to a subsequence , we can assume that @xmath409 is contained in one of the two surface group factors for all @xmath530 without loss of generality , assume that @xmath409 is contained in @xmath531 for all @xmath530 let @xmath351 be the limit , up to subsequence , of @xmath493 in @xmath532 the space of projective measured laminations . since @xmath493 is disjoint from the meridian for all @xmath59 , so is @xmath351 . in particular , @xmath351 lies on @xmath533 since @xmath534 either @xmath31 and @xmath351 are disjoint or @xmath535 since @xmath31 is minimal . if @xmath31 and @xmath351 are disjoint , then we can find a simple closed curve @xmath22 on @xmath19 that is disjoint from @xmath114 then , @xmath22 is the boundary of an essential annulus @xmath536 that misses @xmath190 a contradiction . if @xmath537 then @xmath31 lies on @xmath19 and misses any essential annulus in @xmath18 a contradiction we thank the organizers of the workshop on ergodic decompositions of representation varieties \" held in princeton in october 2011 , where the layout of the paper was conceived . the second author thanks dick canary for several useful conversations . 99 i. agol , _ tameness of hyperbolic 3-manifolds _ , preprint available at arxiv : math\\0405568v1 f. bonahon , _ cobordism of automorphisms of surfaces _ , ann . sci . cole norm . ( 4 ) * 16 * ( 1983 ) , no . 2 , 237270 . d. calegari and d. gabai , _ shrinkwrapping and the taming of hyperbolic 3-manifolds _ , j. amer . * 19 * ( 2006 ) , no . 2 , 385446 . r. d. canary , _ ends of hyperbolic @xmath6-manifolds _ , j. amer . math . * 6 * ( 1993 ) , no . 1 , 135 . r.d . canary , dynamics on character varieties : a survey , 2013 , preprint available at arxiv:1306.5832 [ math.gt ] . canary and p. storm , moduli spaces of hyperbolic 3-manifolds and dynamics on character varieties , to appear in commentarii mathematici helvetici . a. casson and d. long , _ algorithmic compression of surface automorphisms _ , invent . math . 81 ( 1985 ) , no . 2 , 295303 . t. gelander , _ on deformations of fn in compact lie groups _ , israel j. math . 167 ( 2008 ) , 1526 . w. goldman , _ an ergodic action of the outer automorphism group of a free group _ , gafa , 17 ( 2007 ) , 793805 . i. a. grushko , on the bases of a free product of groups . matematicheskii sbornik , * 8 * ( 1940 ) , 169182 . j. hempel , _ @xmath6-manifolds _ , princeton univ . press , princeton , nj , 1976 . w. jeon , i. kim and k. ohshika with c. lecuire , _ primitive stable representations of free kleinian groups _ , to appear in israel jour math . m. kapovich , _ hyperbolic manifolds and discrete groups _ , progress in mathematics , 183 , birkhuser boston , boston , ma , 2001 . i. kim , c. lecuire and k. ohshika , _ convergence of freely decomposable kleinian groups _ , submitted . g. kleineidam and j. souto , _ algebraic convergence of function groups _ , comment.math . 77 ( 2002 ) , 244269 . a. kurosch , die untergruppen der freien produkte von beliebigen gruppen , math . * 109 * ( 1934 ) , no . 1 , 647660 . c. lecuire , _ plissage des varit hyperboliques de dimension 3 _ , invent . 164 ( 2006 ) , no . 1 , 85141 . c. lecuire , _ an extension of the masur domain . spaces of kleinian groups _ , london math . lecture note ser . 329 , ( 2006 ) , 4973 . m. lee , dynamics on @xmath2-character varieties of certain hyperbolic @xmath6-manifolds , ph.d . thesis , university of michigan , 2012 , available at http://hdl.handle.net/2027.42/94050 m. lee , _ dynamics on the @xmath2-character variety of a compression body _ , to appear in algebraic and geometric topology . a. marden , the geometry of finitely generated kleinian groups , ann . of math . ( 2 ) * 99 * ( 1974 ) , 383462 . d. mccullough , _ compact submanifolds of @xmath6-manifolds with boundary _ , quart . oxford ser . ( 2 ) * 37 * ( 1986 ) , no . 147 , 299307 . d. mccullough and a. miller , _ homeomorphisms of @xmath6-manifolds with compressible boundary _ , mem . * 61 * ( 1986 ) , no . 344 d. mccullough , a. miller and g. a. swarup , _ uniqueness of cores of noncompact @xmath6-manifolds _ , j. london math . soc . ( 2 ) * 32 * ( 1985 ) , y. n. minsky , on dynamics of @xmath538 on @xmath2 characters , israel j. math . * 193 * ( 2013 ) , no . 1 , 4770 . m. mj , _ cannon - thurston maps for kleinian groups _ , preprint 2010 available at arxiv:1002.0996 [ math.gt ] . j. p. otal , _ courants godsiques et produits libres _ , thse detat , universit de paris - sud , orsay ( 1988 ) . f. palesi , ergodic actions of mapping class groups on moduli spaces of representations of non - orientable surfaces , geom . dedicata * 151 * ( 2011 ) , 107140 . j. r. stallings , whitehead graphs on handlebodies , in _ geometric group theory down under ( canberra , 1996 ) _ , 317330 , de gruyter , berlin . d. sullivan , _ on the ergodic theory at infinity of an arbitrary discrete group of hyperbolic motions _ , riemann surfaces and related topics : proceedings of the 1978 stony brook conference , ann . of math . 97 , princeton , 1981 . d. pickrell and e. z. xia , _ ergodicity of mapping class group actions on representation varieties . i. closed surfaces , _ comment . * 77 * ( 2002 ) , no . 2 , 339362 . j. h. c. whitehead , _ on certain sets of elements in a free group _ , london math . 41 ( 1936 ) , 4856 . 782 - 800 ."
]
embeddings = model.encode(sentences)
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4] 
