Vorticity alignment results for the three-dimensional Euler and Navier-Stokes equations

合集下载

纵向涡发生器

纵向涡发生器

Numerical study on laminar convection heat transfer in a channel with longitudinal vortex generator.Part B:Parametric studyof major influence factorsJ.M.Wu a,b ,W.Q.Tao a,*aState Key Laboratory of Multiphase Flow in Power Engineering,School of Energy and Power Engineering,Xi’an Jiaotong University,Xi’an City 710049,ChinabSchool of Environment and Chemical Engineering,Xi’an Polytechnic University,Xi’an City 710048,ChinaReceived 12October 2006;received in revised form 17March 2007Available online 24May 2008AbstractThis paper presents the influences of main parameters of longitudinal vortex generator (LVG)on the heat transfer enhancement and flow resistance in a rectangular channel.The parameters include the location of LVG in the channel,geometric sizes and shape of LVG.Numerical results show that the overall Nusselt number of channel will decrease with the LVGs’location away from the inlet of the channel,and decrease too with the space between the LVG pair decreased.The location of LVG has no significant influence on the total pressure drop of channel.With the area of LVG increased,the average Nusselt number and the flow loss penalty of channel,especially when b =45°will increase.With the area of LVG fixed,increasing the length of rectangular winglet pair vortex generator will bring about more heat transfer enhancement and less flow loss increase than that increasing the height of rectangular winglet pair vortex gen-erator.With the same area of LVG,delta winglet pair is more effective than rectangular winglet pair on heat transfer enhancement of channel,and delta winglet pair-b is more effective than delta winglet pair-a.Delta winglet pair-a results in a higher pressure drop,the next is rectangular winglet pair and the last is delta winglet-b.The increase of heat transfer enhancement is always accompanied with the decrease of field synergy angle between the velocity and temperature gradient when the parameters of LVG are changed.This confirms again that the field synergy is the fundamental mechanism of heat transfer by longitudinal vortex.The laminar heat transfer of the chan-nel with punched delta winglet pair is experimentally and numerically studied in the present paper.The numerical result for the average heat transfer coefficient of the channel agrees well with the experimental result,indicating the reliability of the present numerical predictions.Ó2007Published by Elsevier Ltd.Keywords:Vortex generator;Heat transfer enhancement;Flow loss;Field synergy principle1.IntroductionIn the companion paper [1]of the present article,the longitudinal vortex generator was introduced and the major related studies [2–14]were reviewed in detail.Thus,for the simplicity of presentation,the present article willnot repeat such contents.From the review of [1],it is con-cluded that even though a large amount of researches,both numerical and experimental,have been conducted,follow-ing two aspects need further study for a deeper understand-ing of the LVG performance and the essence of heat transfer enhancement by LVG.First,in engineering prac-tice the LVG is punched from the base sheet with a finite thickness,hence,there is a corresponding hole under the LVG.However,in most of the existing literatures,either the LVG thickness or the hole was not the taken into account.And it is also very limited for such parametric0017-9310/$-see front matter Ó2007Published by Elsevier Ltd.doi:10.1016/j.ijheatmasstransfer.2007.03.031*Corresponding author.E-mail address:wqtao@ (W.Q.Tao)./locate/ijhmtAvailable online at International Journal of Heat and Mass Transfer 51(2008)3683–3692study,which took both the LVG thickness and the hole into account.Second,the heat transfer enhancement by LVG was usually explained by the traditional enhancement mechanisms,including the re-development of a boundary layer,the mixing caused by swirling of the LVG and the LVG-induced disturbances.From our preliminary study shown in[1],the fundamental mechanism of heat transfer enhancement by the LVG is the improvement of the syn-ergy between velocity and temperature gradient.The major purpose of this paper is to present our numerical results of a parametric study for the influences of the major parame-ters and reveal the fundamental mechanism of the LVG enhancement.To describe the performance of LVG,we take the comparison of the heat transfer and friction factor results between the duct with LVG and that without LVG at the same other conditions.It should be noted that there are a number of comparison criteria for the enhanced heat transfer surfaces as can be found in[16].For such compre-hensive comparison it needs another full paper,and is not the task of the present study.It may be useful to indicate that thefield synergy principle is a new idea or concept of the heat transfer enhancement mechanism,it is not the comparison criterion for which a number of comparison criteria can be found in[15].From the results presented later,it can be found that any heat transfer enhancement by the LVG is always accompanied by a better synergy between velocity and temperature gradient,thus un-doubt-fully demonstrated that the fundamental reason for LVG enhancement is in the improvement of thefield synergy.For the readers’convenience,the computed channel is shown in Fig.1.The governing equations and boundary conditions have been described in detail in[14],and for the simplicity of presentation they are not shown in this paper.The numerical results will be presented in detail inthis paper.The mechanism of heat transfer by LVG will be discussed from thefield synergy principle,and the domain average synergy angle will be used to show the degree of synergy between velocity and temperature gradi-ent.A brief introduction to thefield synergy principle has been provided in the companion paper[14],and for the simplicity of presentation,it will not be re-stated here.In the following,the numerical predicted effects of the LVG’s location,size and shape willfirst be presented,followed by a comparison between experimental measurement and numerical prediction.Finally,some conclusions will be drawn.Nomenclaturea transverse space between the winglet pair de-fined in Fig.1(m)b thickness of vortex generator(m)B width of channel(m)h height of vortex generator(m)H height of channel(m)f fanning frictional factorl chord length of vortex generator(m)Nu Nusselt numbers streamwise coordinate of LVG defined in Fig.1 (m)Greek symbolb attack angle(°)Subscriptsm average0channel without vortex generator3684J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–36922.Effects of the RWLVG’s location on heat transfer enhancement,field synergy and flow lossAs we know,LV mainly influences the heat transfer in the downstream of the RWLVG.Thus,we can be sure that the location of the RWLVG in the channel will influence the global heat transfer for a given channel.In this section,the influences of the location of RWLVG pair on the heat transfer enhancement and the synergy angle of the channel is numerically computed.The location of the RWLVG is described by the parameters of a and s (see Fig.1).For a convenient presentation the relative values of the ratios of s to H and a to H are adopted.In the following compu-tations Reynolds number of channel is set to 1600.When a /H =0.5,the variations of average Nusselt num-ber and average synergy angle in the whole field vs s /H are shown in Fig.2.1It shows that when the RWLVG pair is moved away from the inlet of channel (increasing the value of s ),the channel average synergy angle will increase and the heat transfer enhancement will decrease.Because with the RWLVG moving to the channel inlet,the influence range of the LV is getting larger in streamwise direction,the synergy between the velocity and the temperature fields is improved in larger region.As a result,the average syn-ergy angle in entire field will decrease and the heat transfer will be enhanced.When s /H =4.0,the variations of average Nusselt num-ber and average synergy angle in the whole field vs a /H are shown in Fig.3.With decreasing the transverse space between the RWLVG pair (decreasing the value of a ),aver-age synergy angle increases and heat transfer enhancement decreases.Especially,when a /H =0.1,the average Nusselt number for the case of b =45°will drop to the value for the case of b =30°at the same condition.The reason is that the LV pair generated by RWLVG pair in the channel rotates in the opposite directions.With decreasing thetransverse space of a ,the interaction between the LV pair in their inner side is getting stronger.As a result,each vor-tex will be weakened by the other,hence,heat transfer is deteriorated and the synergy between velocity and temper-ature becomes worse.Fig.4shows the comparison of theFig. parison of temperature profiles at different cross section (b =45°).J.M.Wu,W.Q.Tao /International Journal of Heat and Mass Transfer 51(2008)3683–36923685temperature profiles at the different cross sections under the conditions of b=45°between the cases of a/H=0.5 with a/H=0.1.We can clearly observe that only the cen-tral region of the channel is disturbed by LVs for the case of a/H=0.1,and temperature profiles at the both side regions of the channel are still layered.The influence range by LVs for the case of a/H=0.5is wider than that for the case of a/H=0.1,and of course the convection heat trans-fer in the former case is enhanced more than in the later case.Therefore,to a given width of the channel,the trans-verse space between the RWLVG pair should not be too small so that the two LVs generated will not be interacted each other and the spanwise LV’s influences may reach a wider distance.As far as the pressure drop is concerned,numericalresults revealed that for a given channel with the same LVG,the value of s almost has no effect on the totalflow loss of the channel,while decreasing the value of a,will bring about a slight increase of totalflow loss of channel because of the stronger interaction between the LV pair.3.Effects of the geometric sizes of RWLVG on heat transfer enhancement,field synergy andflow lossAs an example,the rectangular winglet pair is still used as the longitudinal vortex generator.Due to that the rect-angular winglet pair is punched out from thefin wall,the thickness of RWLVG is taken as constant.Therefore, under the condition of a given attack angle,geometric sizes of RWLVG include its chord length of l and height of h. The location of RWLVG is given as:a/H=0.5and s/H=4.0,Reynolds number is still1600.As we know, the size of h will decide the shielded part of theflow cross section in the height direction.Thus,a dimensionless ratio of h to H is employed to express the relative height of RWLVG.Similarly,the shielded part of theflow cross sec-tion in the spanwise direction is decided by the size of chord length l,therefore,another dimensionless ratio of l to B/2is used to express the relative chord length of the RWLVG.When l/0.5B=0.5,the influences of the height of the RWLVG on the heat transfer enhancement,average syn-ergy angle and increase of pressure loss in the channel are shown in Figs.5and6.Fig.5shows that with increas-ing the height of the RWLVG,the synergy angle decreases, and the heat transfer enhancement increases.As we know, the cross-sectionalflow area of the channel decreases when h is increased,and thefluid velocity at the side edge of RWLVG increases at the same volumeflow rate.There-fore,the strength of the generated LVs increases,thus improving the heat transfer performance.However,the associated penalty of pressure drop is also increased greatly with increasing h,especially,for the case of b=45°.For example,when h/H=0.7,the average friction factor of the channel with RWLVG in condition of b=45°will increase by88.7%compared with the plain channel without LVG,which is about28.3%higher than that of h/H=0.5 (see Fig.6).Therefore,the suitable height of RWLVG pair in the channel should be h/H=0.5.When h/H=0.5,the effects of the chord length of the RWLVG on the heat transfer enhancement,synergy angle3686J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–3692and penalty of pressure loss are shown in Figs.7and8.The LVs are generated whenfluidflows over the side-edge of the RWLVG because of inertia and boundary layer separa-tion.The number of generated LVs will increase with increasing the chord length of l.This leads to two positive results.One is that the total strength of the LV will increase due to the superposition of the LVs,the other is that the transverse influence range of LVs is getting wider.There-fore,thefield synergy between the velocity and temperature fields will be better,and the heat transfer enhancement is improved.Meanwhile,with the increasing the chord length,the area of the RWLVG also increases.This leads to an increase of the form drag of RWLVG and more pres-sure drop of the channel.However,the increase rate of pressure drop with increasing l for the case of b=30°is much lower than that for the case of b=45°(see Fig.8). So,it is feasible to enhance the heat transfer by properly increasing the chord length of the RWLVG for the case of b=30°.From the above results,we can know that the increase of the area of RWLVG will bring about more heat transfer enhancement either by increasing the height or length of LVG.To investigate which one is more effective to heat transfer enhancement,the computed results for the cases with the same area but different height and length of RWLVG pair are compared in Table1.The area of the RWLVG for cases1and2is the same(150mm2),the length of RWLVG for case1is larger than that for case 2,and of course the height of RWLVG for case1is smaller than that for case2.The area of the RWLVG for cases3 and4is the same(250mm2),the length of RWLVG for case3is larger than that for case4,and the height of RWLVG for case3is smaller than that for case4.The attack angle of RWLVG pair for every case can be30°or45°.From the data in Table1,we canfind that the RWLVG pair with a larger length and smaller height can cause the synergy angle of the channel to be smaller,heat transfer enhancement to be more and the pressure drop increase to be smaller.Therefore,when the area of RWLVG is given,properly increasing the length of RWLVG and decreasing its height is a good idea to get more heat transfer enhancement and to preventflow loss from increasing too fast.4.Effects of the LVG’s shape on heat transfer enhancement,field synergy andflow lossBecause the influence range in the transverse direction (called transverse influence range hereafter)of the LV gen-erated by winglet pair type vortex generator is larger than that of LV generated by wing type vortex generator,so the winglet pair vortex generator is more acceptable in the practical application.The laminarflows and heat transfer in the channels with two kinds of delta winglet vortex gen-erators(DWLVG)(called DWLVG-a and DWLVG-b) were numerically calculated,respectively,and the results are compared with those of the channel with RWLVG pre-sented in the above sections.For the comparisons being meaningful,the following conditions are taken.The geo-metric sizes of the channels for the three cases are kept the same as the one in the above computations.The height of DWLVG-b is half of channel height,which is same as that of RWLVG,and the height of DWLVG-a is the same as the channel height.The locations of RWLVG and DWLVG-a in the channels are s/H=4.0,and a/H=0.5. The location DWLVG-b in the channel is s/H=4.0,and a/H=0.4.Taking a/H=0.4in the channel with DWLVG-b is to have enough area to punch the DWLVG out of thefin wall.The numerical predictions in Section2 show that this will not bring much effect on the result.The areas of RWLVG,DWLVG-a and DWLVG-b are the same,200mm2.The attack angle of the three LVGs is set to30°.The comparisons of heat transfer enhancement andfield synergy between the three cases are shown in Fig.9.It is found that for different Reynolds numbers the heat transfer is enhanced more by DWLVG than by RWLVG.The aver-age synergy angles between the velocity and temperature gradient for the cases with DWLVG are smaller than that for the case with RWLVG.Moreover,the DWLVG-b is more advantageous for heat transfer enhancement than the DWLVG-a under higher Reynolds number condition. These can be explained as follows.The LVs will be gener-ated at the side edge RWLVG due to the inertia and boundary layer separation of rectangular winglet,but forTable1Comparison between the influences of the length and the height of the RWLVGCase RWLVG sizes(mm):lÂh Area(mm2)Attackangle(°)Nu m/Nu0Synergyangle(°)f/f0120Â7.515030 1.13286.78 1.18045 1.15086.51 1.294215Â1030 1.09886.84 1.19145 1.12286.56 1.322320Â12.525030 1.19886.24 1.36845 1.24385.99 1.700425Â1030 1.24786.19 1.35045 1.28985.87 1.681J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–36923687delta winglet LVG,however,the LVs are generated at the leading edge.Under the condition of same area,the longer the side edge (the long side of the rectangular winglet)or the leading edge(the bevel side of the triangle winglet)the more the LVs generated and the stronger the LVs’strength,and of course,the stronger the disturbance of LVs to the flow,therefore,the more the heat transfer enhancement.The length of the leading edge of DWLVG-a is longer than that of the side edge of RWLVG,and the length of the leading edge of DWLVG-b is longer than that of DWLVG-a.Thus,DWLVG-b is more advantageous than DWLVG-a,and DWLVG-a is superior to the RWLVG for enhancing heat transfer.Fig.10shows the differences of the flow loss between the three channels with different kinds of LVGs.From Fig.10,we can find that the pressure loss for the channel with DWLVG-a is the largest.This is because the height of DWLVG-a is equal to that of the channel so that the shielding of DWLVG-a to the cross sectional flow area in the vertical direction is severe.The flow loss for the channel with DWLVG-b is the smallest due to its small height.Although the height of DWLVG-b is the same as that ofRWLVG,which is half of the channel height,its influence to the flow is mild,not so bining the effects of the LVG’s shape to heat transfer enhancement and flow loss,DWLVG-b may be taken as a kind of LVG with bet-ter performance.Again the conclusion that properly increasing the length of LVG and decreasing the height of can obtain more heat transfer enhancement and prevent flow loss from increasing too fast is verifiparisons with experiment 5.1.Experimental apparatusThe test was conducted in a small tunnel shown in Fig.11.The wall of the tunnel was made of organic glass with thickness of 10mm.The cross section of the test sec-tion is 162mm in width and 62mm in height.A compound aluminum plate is mounted in the middle of the test section by two small bakelite supporting posts as shown in Fig.12.So actually the test section is composed of two channels through which air flows over the two surfaces of the com-pound aluminum plate.The average convection heat trans-fer on the up and down surfaces of the compound aluminum plate was tested.The compound aluminum plate with a pair of punched delta winglets,as shown in Fig.13,was made of two layers of aluminum plate with thickness of 1mm and a thin electric heating film (0.15mm)used to heat air through the aluminum plates.The sizes of the aluminum plates and the position of delta winglet are shown in Fig.14.To avoid the heat transfer through the walls of the test section,the out surface of the walls are thermally isolated with the foam plates of thickness 100mm.The inlet temperature measuring mesh contains 8ther-mocouples,and the outlet temperature measuring mesh contains 16thermocouples.The surface temperature of the up and down aluminum plates was measured by 8ther-3688J.M.Wu,W.Q.Tao /International Journal of Heat and Mass Transfer 51(2008)3683–3692mocouples,respectively.The measurement uncertainty ofthese thermocouples was within ±0.2°C.The air volumet-ric flow rate was metered by a rotor flow meter with a rel-ative error of 2.5%.To keep the flow in the channels in laminar state,the air volumetric flow rate was controlled within 9–33m 3h À1,which corresponds to the frontal velocities from 0.25to 1m/s.The power of the heating film is adjusted by a booster and controlled within 35±1W.The energy un-balance between the heat gained by air and heating power was within 5%.The uncertainty in the reported experimental value of the surface convection heat transfer coefficient was estimated by the method suggested by Moffat [16],and is less 5%in the present experimental conditions.5.2.Experimental results and comparisonThe surface average convection heat transfer for the five compound aluminum plates,including the plain plate with-Fig.12.Structure of the test section.Fig.13.Structure of the compound aluminumplate.J.M.Wu,W.Q.Tao /International Journal of Heat and Mass Transfer 51(2008)3683–36923689out LVG and the ones with punched delta winglet pair at four different attack angles,in a channelflow were tested. Fig.15shows the experimental results.In comparison with the plain plate,the heat transfer coefficient of the plate with delta winglet pair at b=15°is increased by8–11%,those at b=30°,45°,60°is increased by15–20%,21–29%,21–34%, respectively.The heat transfer enhancement is increased with the increase of the average velocity in the channel. We canfind that the heat transfer coefficient for the case of b=45°is just slightly higher than that of b=60°.The result is coincident with the one given by Fiebig[7]for delta winglet pair,even though the tested channel in our experi-ment has some differences with the one by Fiebig[7].To validate the present models and methods used in the present numerical solutions,the heat transfer in the tested channel is numerically simulated.Due to the symmetry,only half of the tested channel is computed as shown in Fig.16.As a bell-type inlet and a straighter–equalizer were adopted in the test tunnel(see Fig.11),a uniform inlet boundary condi-tion is employed.To use the developed boundary condition at outlet,the computed domain is extended20times the height of channel in downstream.The walls of tested section are taken as adiabatic.The surfaces of two bakelite support-ing post are taken as adiabatic walls too.The heatingfilm is treated as solid with inner heat source.Aluminum plates including delta winglet pair are treated as solid conjugated with air in the computation.Fig.17shows the comparisons between the numerical and experimental results for thefive cases.Deviations between tested and simulated results of heat transfer for every case is less than10%.The agreement between the numerical and experimental results proves that the models and methods used in the present study are feasi-ble and the numerical results are reliable.6.ConclusionsExtending our study of part A[14],the influences of main parameters of longitudinal vortex generator(LVG) on heat transfer enhancement andflow resistance are numerically computed and analyzed in the present paper. The majorfindings are as follows:(1)The increase of heat transfer enhancement is alwayscompanied by the decrease offield synergy angle between the velocity and temperature gradient when the parameters of LVG are changed.This confirms again that thefield synergy is the fundamental mech-anism of heat transfer by longitudinal vortex.(2)The heat transfer enhancement of channel willdecrease with the increase of LVGs’location from the channel inlet,and with the decrease of the trans-verse space between the LVG pair.The location of LVG has no significant influence on theflow loss of channel.Fig.16.The schematic diagram of computed domain.3690J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–3692(3)With the increasing area of LVG,both the heat trans-fer enhancement of the channel and the pressure loss will increase.At afixed area of LVG,properly increasing the length of RWLVG and decreasing the height of RWLVG is a good idea to get more heat transfer enhancement and to prevent from a signifi-cantflow loss increase.(4)With the same area of LVG,DWLVG is more effec-tive than RWLVG on the channel heat transfer enhancement,and DWLVG-b is more effective thanDWLVG-a.DWLVG-b leads to less pressure loss in comparison with DWLVG-a and EWLVG.DWLVG-b may be regarded as the best one among LVGs compared.(5)The surface average convection heat transfer for thefive compound aluminum plates,including the plain plate without LVG and the ones with punched delta winglet pair at four different attack angles,in a chan-nelflow were tested as well as simulated.Deviations of heat transfer between the numerical and experi-J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–36923691mental results for every case is less than10%.The agreement between the numerical and experimental results proves that the models and methods used in the present study are feasible and the numerical results are reliable.AcknowledgementsSupports from the National Basic Research of China (973Program)(2007CB206902),and the National Natural Science Foundation of China(No.50476046)are greatly appreciated.References[1]J.M.Wu,W.Q.Tao,Numerical study on laminar convection heattransfer in a rectangular channel with longitudinal vortex generator, Part A:verification offield synergy principle,Int.J.Heat Mass Transfer51(2008)1179–1191.[2]M.Fiebig,U.Brockmeier,N.K.Mitra,T.Guntermann,Structure ofvelocity and temperaturefields in laminar channelflows with vortex generators,Numer.Heat Transfer,Part A15(1989)281–302. [3]G.Biswas,H.Chattopadhyay,Heat transfer in a channel with built-in wing type vortex generators,Int.J.Heat Mass Transfer35(1992) 803–914.[4]M.Fiebig,P.Kallweit,N.Mitra,S.Tiggelbeck,Heat transferenhancement and drag by longitudinal vortex generators in channel flow,Exp.Therm.Fluid Sci.4(1991)103–114.[5]S.Tiggelbeck,N.K.Mitra,M.Fiebig,Experimental investigations ofheat transfer enhancement andflow losses in a channel with double rows of longitudinal vortex generators,Int.J.Heat Mass Transfer36 (1993)2327–2337.[6]G.Biswas,P.Deb,S.Biswas,Generation of longitudinal streamwisevortices–a device for improving heat exchanger design,ASME J.Heat Transfer116(1994)558–597.[7]M.Fiebig,Embedded vortices in internalflow:heat transfer andpressure loss enhancement,Int.J.Heat Fluid Flow16(1995)376–388.[8]G.Biswas,K.Torii,D.Fuji,K.Nishino,Numerical and experimentaldetermination offlow structure and heat transfer effects of longitu-dinal vortices in a channelflow,Int.J.Heat Mass Transfer39(1996) 3441–3451.[9]Y.Chen,M.Fiebig,N.K.Mitra,Heat transfer enhancement of afinned oval tube with staggered punched longitudinal vortex gener-ators,Int.J.Heat Mass Transfer43(1998)417–435.[10]C.C.Wang,J.Jo,Y.T.Lin,C.S.Wei,Flow visualization of annularand delta winglet vortex generators infin-and-tube heat exchanger application,Int.J.Heat Mass Transfer45(2002)3803–3815. [11]J.S.Leu,Y.H.Wu,J.Y.Jang,Heat transfer andfluidflow analysis inplate-fin and tube heat exchangers with a pair of block shape vortex generators,Int.J.Heat Mass Transfer47(2004)4327–4338. [12]S.M.Pesteei,P.M.V.Subbarao,R.S.Agarwal,Experimental study ofthe effect of winglet location on heat transfer enhancement and pressure drop infin-tube heat exchangers,Appl.Therm.Eng.25 (2005)1684–1696.[13]J.E.O’Brien,M.S.Sohal,Heat transfer enhancement forfinned-tubeheat exchangers with winglets,ASME J Heat Transfer127(2005) 171–178.[14]K.M.Kwak,K.Torri,K.Nishino,Simultaneous heat transferenhancement and pressure loss reduction forfinned-tube bundles with first or two transverse rows of built-in winglets,Exp.Therm.Fluid Sci.29(2005)625–632.[15]R.L.Webb,N.H.Kim,Principles of Enhanced Heat Transfer,seconded.,Taylor&Francis,Boca Raton,2005(Chapter3).[16]R.J.Moffat,Describing the uncertainties in experimental results,Exp.Therm.Fluid Sci.1(1988)3–17.3692J.M.Wu,W.Q.Tao/International Journal of Heat and Mass Transfer51(2008)3683–3692。

三圆幅值法优化某中巴车传动系动平衡研究

三圆幅值法优化某中巴车传动系动平衡研究

doi:10.3969/j.issn.1005-2550.2023.05.002 收搞日期:2023-07-28三圆幅值法优化某中巴车传动系动平衡研究赵建1,2,宋雨1,2,王明正1,2,仲崇发1,2,陈曦1,2(1.中国第一汽车集团有限公司 研发总院,长春130022;2. 高端汽车集成与控制全国重点实验室,长春130022)摘 要:某中巴车在80km/h以上车内有明显的轰鸣和振动,经试验测试分析确定为传动系一阶不平衡导致。

根据现有技术水平,难以对整个传动系统组件进行整体动平衡下线检测和控制,本文首先对传动系动平衡技术进行了总结,对三圆动平衡进行了理论分析。

并采用三圆幅值法在传动轴和后主减连接位置传动轴侧对整个传动系统进行了重新动平衡,试验结果表明,优化前后驾驶员座椅导轨传动系一阶振动RSS峰值降低了73%,车内传动系一阶噪声峰值由降低了15%,高车速车内轰鸣和振动消失,问题得以解决,说明三圆幅值法解决类似工程实际问题快速有效。

关键词:传动系;阶次分析;三圆幅值;动平衡;振动中图分类号:U463.2 文献标识码:A 文章编号:1005-2550(2023)05-0009-07Study on the Optimization of the Dynamic Balance of theTransmission System of a Medium Bus by the Three-circleAmplitude MethodZHAO Jian1,2, SONG Yu1,2, WANG Ming-zheng1,2, ZHONG Chong-fa1,2, CHEN Xi1,2(1. China First Automobile Group Co., Ltd. Research and Development Institute, Array Changchun 130022, China;2.National Key Laboratory of Advanced Vehicle Integrationand Control, Changchun 130022, China)Abstract: A minibus has obvious roaring and vibration in the vehicle when the speed is more than 80km/h, which is determined to be caused by the first order imbalance of thetransmission system through test and analysis. According to the existing technical level, it isdifficult to carry out offline detection and control of the overall dynamic balance of the entiretransmission system components. First, this paper summarizes the dynamic balance technologyof the transmission system, and makes a theoretical analysis of the three circle dynamic balance.The three circle amplitude method is used to re balance the whole transmission system atthe transmission shaft and the transmission shaft side at the connection position of the rearmain reducer. The test results show that the peak value of the first order vibration RSS of theoptimized front and rear driver seat rail transmission system is reduced from 4.12g to 1.12g,and the peak value of the first order noise of the transmission system in the vehicle is reducedfrom 101dB to 86 dB. The rumble and vibration in the vehicle at high speed disappear, and theproblem is solved, It shows that the three circle amplitude method is fast and effective in solvingpractical problems of similar projects.Key Words: Drive train; Order analysis; Amplitude of three circles; Unbalance; Vibration衡后,问题得以解决,为类似车辆市场抱怨问题的解决提供了重要的工程实践参考。

Statistical mechanics of two-dimensional vortices and stellar systems

Statistical mechanics of two-dimensional vortices and stellar systems

a r X iv:c o n d -ma t/212223v1[c ond-m at.stat-m ec h]1D ec22Statistical mechanics of two-dimensional vortices and stellar systems Pierre-Henri Chavanis Laboratoire de Physique Quantique,Universit´e Paul Sabatier,118,route de Narbonne 31062Toulouse,France Abstract.The formation of large-scale vortices is an intriguing phenomenon in two-dimensional turbulence.Such organization is observed in large-scale oceanic or atmo-spheric flows,and can be reproduced in laboratory experiments and numerical simula-tions.A general explanation of this organization was first proposed by Onsager (1949)by considering the statistical mechanics for a set of point vortices in two-dimensional hydrodynamics.Similarly,the structure and the organization of stellar systems (glob-ular clusters,elliptical galaxies,...)in astrophysics can be understood by developing a statistical mechanics for a system of particles in gravitational interaction as initiated by Chandrasekhar (1942).These statistical mechanics turn out to be relatively similar and present the same difficulties due to the unshielded long-range nature of the in-teraction.This analogy concerns not only the equilibrium states,i.e.the formation of large-scale structures,but also the relaxation towards equilibrium and the statistics of fluctuations.We will discuss these analogies in detail and also point out the specificities of each system.1Introduction Two-dimensional flows with high Reynolds numbers have the striking property of organizing spontaneously into coherent structures (the vortices)which dominate the dynamics [93](see Fig.1).The robustness of Jupiter’s Great Red Spot,a huge vortex persisting for more than three centuries in a turbulent shear between two zonal jets,is probably related to this general property.Some other coherent structures like dipoles (pairs of cyclone/anticyclone)and sometimes tripoles have been found in atmospheric or oceanic systems and can persist during several days or weeks responsible for atmospheric blocking.Some astrophysicists invoke the existence of organized vortices in the gaseous component of disk galaxies in relation with the emission of spiral density waves [99].It has also been proposed that planetary formation might have begun inside persistent gaseous vortices born out of the protoplanetary nebula [5,121,15,60,33](see Fig.2).As a result,hydrodynamical vortices occur in a wide of geophysical or astrophysical situations and their robustness demands a general understanding.Similarly,it is striking to observe that self-gravitating systems follow a kindof organization despite the diversity of their initial conditions and their en-vironement [9](see Fig.3).This organization is illustrated by morphological classification schemes such as the Hubble sequence for galaxies and by simple2Pierre-Henri Chavanisrules which govern the structure of individual self-gravitating systems.For ex-ample,elliptical galaxies display a quasi-universal luminosity profile described by de Vaucouleur’s R1/4law and most of globular clusters are well-fitted by the Michie-King model.On the other hand,theflat rotation curves of spiral galaxies can be explained by the presence of a dark matter halo with a density profile decreasing as r−2at large distances.The fractal nature of the interstellar medium and the large scale structures of the universe also display some form of organization.Fig.1.Self-organization of two-dimensional turbulentflows into large-scale vortices [93].These vortices are long-lived and dominate the dynamics.The question that naturally emerges is what determines the particular con-figuration to which a self-gravitating system or a large-scale vortex settles.It is possible that their actual configuration crucially depends on the conditions that prevail at their birth and on the details of their evolution.However,in view of their apparent regularity,it is tempting to investigate whether their organi-zation can be favoured by some fundamental physical principles like those of thermodynamics and statistical physics.We ask therefore if the actual states of self-gravitating systems in the universe and coherent vortices in two-dimensional2D vortices and stellar systems3Fig.2.A scenario of planet formation inside large-scale vortices presumably present in the Keplerian gaseous disk surrounding a star at its birth.Starting from a random vorticityfield,a series of anticyclonic vortices appears spontaneously(upper panel). Due to the Coriolis force and to the friction with the gas,these vortices can efficiently trap dust particles passing nearby(lower pannel).The local increase of dust concen-tration inside the vortices can initiate the formation of planetesimals and planets by gravitational instability.This numerical simulation is taken from[15].turbulentflows are not simply more probable than any other possible configura-tion,i.e.if they cannot be considered as maximum entropy states.This statistical mechanics approach has been initiated by Onsager[101]for a system of point vortices and by Chandrasekhar[21]in the case of self-gravitating systems.It turns out that the statistical mechanics of two-dimensional vortices and self-gravitating systems present a deep analogy despite the very different physical nature of these systems.This analogy was pointed out by Chavanis in[29,32,35] and further developed in[54,30,34,36,47,48].In the following,we will essentially discuss the statistical mechanics of2D vortices and refer to the review of Pad-manabhan[103](and his contribution in this book)for more details about the4Pierre-Henri Chavanisrge-scale structures in the universe as observed with the Hubble space tele-scope.The analogy with Fig.1is striking and will be discussed in detail in this paper. statistical mechanics of self-gravitating systems.We will see that the analogy be-tween two-dimensional vortices and(three-dimensional)self-gravitating systems concerns not only the prediction of the equilibrium state,i.e.the formation of large-scale structures,but also the statistics offluctuations and the relaxation towards equilibrium.This paper is organized as follows.In Sec.2,we discuss the statistical me-chanics of point vortices introduced by Onsager[101]and further developed by Joyce&Montgomery[70]and Pointin&Lundgren[107]among others(see a complete list of references in the book of Newton[98]).We discuss the existence of a thermodynamic limit in Sec.2.7and make the connexion withfield theory. Statistical equilibrium states of axisymmetricflows are obtained analytically in Sec.2.8-2.9.The relation with equilibrium states of self-gravitating systems is shown in Sec.2.10.In Sec.3,we discuss the statistics of velocityfluctuations pro-duced by a random distribution of point vortices and use this stochastic approach to obtain an estimate of the diffusion coefficient of point vortices.Application to2D decaying turbulence is considered in Sec.3.4.In Sec.4,we describe the relaxation of a point vortex in a thermal bath and analyze this relaxation in terms of a Fokker-Planck equation involving a diffusion and a drift.In Sec.5,we develop a more general kinetic theory of point vortices.A new kinetic equation is obtained which satisfies all conservation laws of the point vortex system and increases the Boltzmann entropy(H-theorem).We mention the connexion with2D vortices and stellar systems5 the kinetic theory of stars developed by Chandrasekhar[21].In Sec.6,we dis-cuss the violent relaxation of2D vortices and stellar systems.We mention the analogy between the Vlasov and the Euler equations and between the statistical approach developed by Lynden-Bell[90]for collisionless stellar systems and by Kuz’min[83],Miller[95]and Robert&Sommeria[111]for continuous vorticity fields.The concepts of“chaotic mixing”and“incomplete relaxation”are dis-cussed in the light of a relaxation theory in Sec.6.3.Application of statistical mechanics to geophysicalflows and Jupiter’s Great Red Spot are evocated in Sec.6.4.2Statistical mechanics of point vortices intwo-dimensional hydrodynamics2.1Two-dimensional perfectflowsThe equations governing the dynamics of an invisicidflow are the equation of continuity and the Euler equation:∂ρ∂t +(u∇)u=−16Pierre-Henri Chavaniswhere ∆=∂2xx +∂2yy is the Laplacian operator.In an unbounded domain,thisequation can be written in integral form asψ(r ,t )=−12πz × ω(r ′,t )r −r ′∂t +u ∇ω=0.(10)This corresponds to the transport of the vorticity ωby the velocity field u .It is easy to show that the flow conserves the kinetic energyE =u 22 (∇ψ)2d 2r =12 ωψd 2r ,(12)where the second equality is obtained by a part integration with the condition ψ=0on the boundary.Therefore,E can be interpreted either as the kinetic energy of the flow (see Eq.(11))or as a potential energy of interaction between vortices (see Eq.(12)).2.2The point vortex gasWe shall consider the situation in which the velocity is created by a collection of N point vortices.In that case,the vorticity field can be expressed as a sum of δ-functions in the formω(r ,t )=N i =1γi δ(r −r i (t )),(13)where r i (t )denotes the position of point vortex i at time t and γi is its circulation.According to Eqs.(9)(13),the velocity of a point vortex is equal to the sum of the velocities V (j →i )produced by the N −1other vortices,i.e.V i = j =i V (j →i )with V (j →i )=−γj |r j −r i |2.(14)2D vortices and stellar systems7 As emphasized by Kirchhoff[79],the above dynamics can be cast in a Hamil-tonian formγi dx i∂y i,γidy i∂x i,(15) H=−12m).Thisis related to the particular circumstance that a point vortex is not a material particle.Indeed,an isolated vortex remains at rest contrary to a material particle which has a rectilinear motion due to its inertia.Point vortices form therefore a very peculiar Hamiltonian system.Note also that the Hamiltonian of point vortices can be either positive or negative(in the case of vortices of different signs)whereas the kinetic energy of theflow is necessarily positive.This is clearlya drawback of the point vortex model.2.3The microcanonical approach of Onsager(1949)The statistical mechanics of point vortices wasfirst considered by Onsager[101] who showed the existence of negative temperature states at which point vortices of the same sign cluster into“supervortices”.He could therefore explain the formation of large,isolated vortices in nature.This was a remarkable anticipation since observations were very scarce at that time.Let us consider a liquid enclosed by a boundary,so that the vortices are confined to an area A.Since the coordinates(x,y)of the point vortices are canonically conjugate,the phase space coincides with the configuration space and isfinite:dx1dy1...dx N dy N= dxdy N=A N.(17)This striking property contrasts with most classical Hamiltonian systems con-sidered in statistical mechanics which have unbounded phase spaces due to the presence of a kinetic term in the Hamiltonian.As is usual in the microcanonical description of a system of N particles,we introduce the density of statesg(E)= dx1dy1...dx N dy Nδ E−H(x1,y1,...,x N,y N) ,(18)8Pierre-Henri Chavaniswhich gives the phase space volume per unit interaction energy E.The equilib-rium N-body distribution of the system,satisfying the normalization condition µ(r1,...,r N)d2r1...d2r N=1,is given byµ(r1,...,r N)=1T =dS2π i<jγiγj ln|r i−r j|.(22)2D vortices and stellar systems9 Making the change of variable x=r/R,wefind that g(E,V)=V N g(E′,1) with E′=E+1βV 1+ββV 1+γ2(N−1)γ2.(25)We shall see in Sec.2.8that this negative critical inverse temperature is the minimum inverse temperature that the system can achieve.If,on the other hand,we consider a neutral system consisting of N/2vortices of circulationγand N/2vortices of circulation−γ,wefindP=N8π .(26)This result is well-known is plasma physics[113].The critical temperature at which the pressure vanishes is now positiveβc=8π∂t =−Ni=1γ∇δ(r−r i(t))V i.(28)Since V i=u(r i(t),t),we can rewrite the foregoing equation in the form∂ω10Pierre-Henri ChavanisSince the velocity is divergenceless,we obtain∂ω4π i=jγ2 ln|r i−r j|1=−2D vortices and stellar systems 11whereg (r 1,r 2,t )=µ(r 1,...,r N ,t )d 3r 2...d 2r N ,(36)is the two-body distribution function.In the mean-field approximation,whichis exact in a properly defined thermodynamic limit with N →+∞(see Sec.2.7),we haveg (r 1,r 2,t )=P (r 1,t )P (r 2,t ).(37)Accounting that N (N −1)≃N 2for large N ,the average energy takes the formE =−12ω ψd 2r =u 2n i !.(40)The logarithm of this number defines the Boltzmann ing Stirlingformula and considering the continuum limit in which ∆,ν→0,we get the classical formula S =−NP (r )ln P (r )d 2r ,(41)where P (r )is the density probability that a point vortex be found in the sur-face element centered on r .At equilibrium,the system is in the most probablemacroscopic state,i.e.the state that is the most represented at the microscopic level.This optimal state is obtained by maximizing the Boltzmann entropy (41)at fixed energy (39)and vortex number N ,or total circulationΓ=Nγ=ω d 2r .(42)12Pierre-Henri ChavanisWriting the variational principle in the formδS−βδE−αδΓ=0,(43) whereβandαare Lagrange multipliers,it is readily found that the maximum entropy state corresponds to the Boltzmann distributionω =Ae−βγψ,(44) with inverse temperatureβ.We can account for the conservation of angular momentum L= ω r2d2r(in a circular domain)and impulse P= ω yd2r (in a channel)by introducing appropriate Lagrange multipliersΩand U foreach of these constraints.In that case,Eq.(44)remains valid provided that we replace the streamfunctionψby the relative streamfunctionψ′=ψ+Ω2π i<j ln|r i−r j|2D vortices and stellar systems13 where the potential of interaction has been normalized by R.For simplicity, we have ignored the contribution of the images but this shall not affect the final results.We now introduce the change of variables x=r/R and define the functionu(x1,...,x N)=1N2γ2.(49) In terms of these quantities,the density of states can be rewritteng(E)=2πV NNu(x1,...,x N) .(50)The proper thermodynamic limit for a system of point vortices with equal cir-culation in the microcanonical ensemble is such that N→+∞withfixedΛ.Wesee that the box size R does not enter in the normalized energyΛ.Therefore,the thermodynamic limit corresponds to N→+∞withγ∼N−1→0and E∼1. This is a very unusual thermodynamic limit due to the non-extensivity of thesystem.Note that the total circulationΓ=Nγremainsfixed in this process.For sufficiently large N,the density of states can be writteng(E)≃ Dρe NS[ρ]δ Λ−E[ρ]) δ 1− ρ(r)d2r ,(51) withS[ρ]=− ρ(r)lnρ(r)d2r,(52) E[ρ]=−114Pierre-Henri Chavaniswhich is the normalization factor in the Gibbs measure1µ(r1,...,r N)=ln Z.(57)βUsing the notations introduced previously,we can rewrite the integral(55)in the formZ(β)=V N N0... N0N i=1d2x i eηu(x1,...,x N),(58)whereβNγ2η=2D vortices and stellar systems15 and phase transitions occur.This is the case,in particular,for the gravitational problem(see Sec.2.10).Finally,the grand canonical partition function is defined byZ GC=+∞ N=0z N2(∇ξ)2−ρ(r)ξ(r)}d2r=e−12(∇ξ)2d2r+√−βγξ(r),we can easily carry out the summation on N to obtainZ GC= Dφe−12(∇φ)2−µ2eφ(r)},(67)T eff=−βγ2,µ2=−zβγ2.(68) Therefore,the grand partition function of the point vortex gas corresponds to a Liouvillefield theory with an actionS[φ]=12(∇φ)2−µ2eφ(r)}.(69)While the previous description is formally correct if we define Z and Z GC by Eqs.(55)and(64),it must be noted however that the canonical and grand canon-ical ensembles may not have a physical meaning for point vortices.In particular, it is not clear how one can impose a thermal bath at negative temperature. On the other hand,the usual procedure to derive the canonical ensemble from the microcanonical ensemble rests on a condition of additivity which is clearly lacking in the present case.2.8Axisymmetric equilibrium states in a diskLet us consider a collection of N point vortices with circulationγconfined within a disk of radius R.At statistical equilibrium,the streamfunctionψis solution of the Boltzmann-Poisson equation(45).If we work in a circular domain,we must in principle account for the conservation of angular momentum.This can lead to bifurcations between axisymmetric and off-axis solutions[117].We shall,16Pierre-Henri Chavanishowever,ignore this constraint for the moment in order to obtain analytical ex-pressions for the thermodynamical parameters.This is a sufficient approximation to illustrate the structure of the problem,which is our main concern here.If we confine our attention to axisymmetric solutions,the Boltzmann-Poisson equation(45)can be written1dr r dψξddξ =λe−φ,(72)φ(0)=φ′(0)=0,(73)withλ=1ifβ<0andλ=−1ifβ>0.It turns out that this equation can be solved analytically as noticed by a number of authors.With the change of variables t=lnξandφ=2lnξ−z,Eq.(72)can be rewrittend2zdz(λe z).(74)This corresponds to the motion of aficticious particle in a potential V(z)=λe z. This equation is readily integrated and,returning to original variables,wefinally obtaine−φ=18ξ2)2.(75)From the circulation theorem(6)applied to an axisymmetricflow,we have−dψ2πr.(76) whereΓ(r)= r0ω(r′)2πr′dr′is the circulation within r.Taking r=R and introducing the dimensionless variables previously defined,we obtainη≡βγΓπR2(η+4)1η+4r22D vortices and stellar systems 17At positive temperatures (η>0),the vorticity is an increasing function of the distance and the vortices tend to accumulate on the boundary of the domain (Fig.4).On the contrary at negative temperatures (η<0),the vorticity is a decreasing function of the distance and the vortices tend to group themselves in the core of the domain to form a “supervortex”(Fig.5).These results are consistent with Onsager’s prediction [101].We also confirm that statistical equilibrium states only exist for η>ηc =−4,as previously discussed.At this critical temperature,the central vorticity becomes infinite and the solution tends to a Dirac peak:ω (r )→Γδ(r ),for η→ηc =−4.(79)00.20.40.60.81r/R1234567<ω>/ω∗η>0η=0η=20Fig.4.Statistical equilibrium states of point vortices at positive temperatures (ω∗=Γ/πR 2).The vortices are preferentially localized near the wall.The energy defined by Eq.(39)can be written in the dimensionless formΛ≡2πE2η2 α0φ′(ξ)2ξdξ.(80)The integral can be carried out explicitly using Eq.(75).Eliminating αbetweenEqs.(80)and (77),we find that the temperature is related to the energy by the equation of state Λ=1ηln418Pierre-Henri Chavanis00.20.40.60.81r/R246810<ω>/ω∗η<0η=−2η=−3η=−3.5Fig.5.Statistical equilibrium states of point vortices at negative temperatures showing a clustering.For η=−4,the vortices collapse at the center of the domain and the vorticity profile is a Dirac peak.0.250.50.751Λ=2πE/Γ2−10−8−6−4−20246810η=βγΓ/2πΛ0=1/8ηc =−4Fig.6.Equilibrium phase diagram (caloric curve)for point vortices with equal circula-tion confined within a disk.For simplicity,the angular momentum has not been taken into account (Ω=0).2D vortices and stellar systems1900.250.50.751Λ=2πE/Γ2−7−6−5−4−3S /NΛ0=1/8η=0η>0η<0ηc =−4Fig.7.Entropy vs energy plot for a system of point vortices with equal circulation confined within a disk.The entropy (41)can also be calculated easily from the above results.Within an unimportant additive constant,it is given byS ηln 4−1+820Pierre-Henri Chavanisaccount,the density of states is given byg (E,L )=δ E −H (r 1,...,r N ) δ L −N i =1γr 2i N i =1d 2r i ,(83)and the angular velocity of the flow byΩ=2T∂SL ,onefinds the exact resultΩ=2N8π.(85)Therefore,the vorticity field is determined by the Boltzmann distributionω =Ae −βγψ′,(86)where ψ′is the relative streamfunctionψ′≡ψ+Ω4βL(4+η)r 2.(87)For η=0,one hasω =γNΓLr 2.(88)For large r ,the asymptotic behavior of Eq.(86)isω ∼14L(4+η)r 2(r →+∞),(89)where we have used ψ∼−(Γ/2π)ln r at large distances.From Eq.(89),one sees that η≥−4is required for the existence of an integrable solution.Inserting the relation (86)in the Poisson equation (7),we get−∆ψ′=Ae −2πη2πLη(4+η).(90)With the change of variablesξ=γNN 2γ2−2πηdξ2+1dξ=2πηe φ−(4+η),(92)2D vortices and stellar systems21 and the vorticity(86)becomesω =N2γ22πL,if r≤(2L/Γ)1/2,(94)and ω =0otherwise.This vortex patch is the state of minimum energy at fixed circulation and angular momentum.Forη→−4,one has approximatelyω =N2γ2(1−AπηγN4L(4+η)r2.(95)Thefirst factor is an exact solution of Eq.(92)with the second term on the right hand side neglected(see Sec.2.8).The second factor is a correction for large r, in agreement with the asymptotic result expressed by Eq.(89).The parameter A tends to infinity asη→−4and is determined from the condition ω d2r=Γby the formulaπA+ln(πA)=−C−ln 1+η|r j−r i|3,(97)where F(j→i)is the force created by star j on star i.The force can be written as the gradient F=−∇Φof a gravitational potentialΦwhich is related to the stellar densityρ(r,t)=Ni=1mδ(r−r i),(98)by the Poisson equation∆Φ=4πGρ.(99)22Pierre-Henri ChavanisFurthermore,the equations of motion(Newton’s equations)can be put in the Hamiltonian formm d r i∂v i,md v i∂r i,H=1|r i−r j|.(100)In the analogy between stellar systems and two-dimensional vortices,the star densityρplays the role of the vorticityω,the force F the role of the velocity V and the gravitational potentialΦthe role of the streamfunctionψ.The crucial point to realize is that,for the two systems,the interaction is a long-range un-shielded Coulombian interaction(in D=3or D=2dimensions).This makes the connexion between point vortices and stellar systems deeper than between point vortices and electric charges for example.In particular,point vortices can organize into large scale clusters,like stars in galaxies,while the distribution of electric charges in a neutral plasma is uniform.There are,on the other hand, fundamental differences between stars and vortices.In particular,a star creates an acceleration while a vortex creates a velocity.On the other hand,the gravi-tational interaction is attractive and directed along the line joining the particles while the interaction between vortices is rotational and perpendicular to the line joining the vortices.Despite these important differences,the statistical mechanics of2D vortices and stellar systems are relatively similar.Like the point vortex gas,the self-gravitating gas is described at statistical equilibrium by the Boltzmann distri-butionρ =Ae−βΦ,(101) obtained by maximizing the Boltzmann entropy atfixed mass M and energy E. Its structure is therefore determined by solving the Boltzmann-Poisson equation∆Φ=4πGAe−βΦ,(102) where A andβ>0have to be related to M and E.This statistical mechanics approach has been developed principally for globular clusters relaxing towards equilibrium via two-body encounters[9].It is clear that the Boltzmann-Poisson equation(102)is similar to the Boltzmann-Poisson equation(45)for point vor-tices at negative temperatures.The density profile determined by these equations is a decreasing function of the distance,which corresponds to a situation of clus-tering(see Figs.5and8).The similarity of the maximum entropy problem for stars and vortices,and the Boltzmann-Poisson equations(102)(45),is afirst manifestation of the formal analogy existing between these two systems.However,due to the different dimension of space(D=3for stars instead of D=2for vortices),the mathematical problems differ in the details.First of all,the density profile determined by the Boltzmann-Poisson equation(102)in D=3decreases like r−2at large distances leading to the so-called infinite mass2D vortices and stellar systems 23−50510ln(ξ)−20−15−10−50ln (e −ψ)singularsphereξ−2Fig.8.Density profile of the self-gravitating gas at statistical equilibrium.The dashed line corresponds to the singular solution ρ=1/2πGβr 2.problem since M = +∞0ρ4πr 2dr →+∞[19].There is no such problem forpoint vortices in two dimensions:the vorticity decreases like r −4,or even more rapidly if the conservation of angular momentum is accounted for,and the totalcirculation Γ= +∞0ω2πrdr is finite.The infinite mass problem implies thatno statistical equilibrium state exists for open star clusters,even in theory.A system of particles in gravitational interaction tends to evaporate so that the final state is just two stars in Keplerian orbit.This evaporation process has been clearly identified in the case of globular clusters which gradually lose stars to the benefit of a neighboring galaxy.In fact,the evaporation is so slow that we can consider in a first approximation that the system passes by a succession of quasiequilibrium states described by a truncated isothermal distribution function (Michie-King model)[9].This justifies the statistical mechanics approach in that sense.Another way of solving the infinite mass problem is to confine the system within a box of radius R .However,even in that case,the notion of equilibrium poses problem regarding what now happens at the center of the configuration.The equilibrium phase diagram (E,T )for bounded self-gravitating systems is represented in Fig.9.The caloric curve has a striking spiral behavior parametrized by the density contrast R =ρ(0)/ρ(R )going from 1(homogeneous system)to +∞(singular sphere)as we proceed along the spiral.There is no equilibrium state below E c =−0.335GM 2/R or T c =GMm24Pierre-Henri ChavanisΛ=−ER/GM 20.51.52.5η=βG M /R Fig.9.Equilibrium phase diagram for self-gravitating systems confined within a box.For sufficiently low energy or temperature,there is no equilibrium state and the system undergoes gravitational collapse.−1−0.500.51 1.5Λ=−ER/GM 200.511.522.533.5η=βG M /R µ=105µ=104µ=103µ=102µ=10Fig.10.Equilibrium phase diagram for self-gravitating fermions [41].The degeneracy parameter µplays the role of a small-scale cut-offǫ∼1/µ.For ǫ→0,the classical spiral of Fig.9is recovered.2D vortices and stellar systems25 mass goes to zero.Therefore,the singularity contains no mass and this process alone cannot lead to a black hole.Since the T(E)curve has turning points,this implies that the microcanon-ical and canonical ensembles are not equivalent and that phase transitions will occur[103].In the microcanonical ensemble,the series of equilibria becomes un-stable after thefirst turning point of energy(MCE)corresponding to a density contrast of709.At that point,the solutions pass from local entropy maxima to saddle points.In the canonical ensemble,the series of equilibria becomes unsta-ble after thefirst turning point of temperature(CE)corresponding to a density contrast of32.1.At that point,the solutions pass from minima of free energy (F=E−T S)to saddle points.It can be noted that the region of negative specific heats between(CE)and(MCE)is stable in the microcanonical en-semble but unstable in the canonical ensemble,as expected on general physical grounds.The thermodynamical stability of isothermal spheres can be deduced from the topology of theβ−E curve by using the turning point criterion of Katz[75]who has extended Poincar´e’s theory of linear series of equilibria.The stability problem can also be reduced to the study of an eigenvalue equation associated with the second order variations of entropy or free energy as studied by Padmanabhan[102]in the microcanonical ensemble and by Chavanis[37]in the canonical ensemble.This study has been recently extended to other statisti-cal ensembles[44]:grand canonical,grand microcanonical,isobaric....The same stability limits as Katz are obtained but this method provides in addition the form of the density perturbation profiles that trigger the instability at the critical points.It also enables one to show a clear equivalence between thermodynamical stability in the canonical ensemble and dynamical stability with respect to the Navier-Stokes equations(Jeans problem)[37,44].These analytical methods can be extended to general relativity[38].It must be stressed,however,that the statistical equilibrium states of self-gravitating systems are at most metastable: there is no global maximum of entropy or free energy for a classical system of point masses in gravitational interaction[2].Phase transitions in self-gravitating systems can be studied in detail by in-troducing a small-scale cut-offǫin order to regularize the potential.This can be achieved for example by considering a system of self-gravitating fermions (in which case an effective repulsion is played by the Pauli exclusion principle) [62,8,52,41,43]or a hard spheres gas[3,103,120].Other forms of regularization are possible[59,128,45].For these systems,there can still be gravitational collapse but the core will cease to shrink when it feels the influence of the cut-off.The result is the formation of a compact object with a large mass:a“fermion ball”or a hard spheres“condensate”.The equilibrium phase diagram of self-gravitating fermions is represented in Fig.10and has been discussed at length by Chavanis [41]in the light of an analytical model.The introduction of a small-scale cut-offhas the effect of unwinding the classical spiral of Fig.9.For a small cut-offǫ≪1,the trace of the spiral is still visible and the T(E)curve is multivalued (Fig.11).This can lead to a gravitationalfirst order phase transition between a gaseous phase with an almost homogeneous density profile(upper branch)and a。

Bose-Einstein condensates with vortices in rotating traps

Bose-Einstein condensates with vortices in rotating traps

a r X i v :c o n d -m a t /9906144v211J un1999Bose-Einstein condensates with vortices in rotating trapsY.Castin 1and R.Dum 1,21Laboratoire Kastler Brossel ∗,´Ecole normale sup´e rieure,24rue Lhomond,F-75231Paris Cedex 05,France2Institut d’optique,BP 147,F-91403Orsay Cedex,France (16March 1999)We investigate minimal energy solutions with vortices for an interacting Bose-Einstein condensate in a rotating trap.The atoms are strongly confined along the axis of rotation z ,leading to an effective 2D situation in the x −y plane.We first use a simple numerical algorithm converging to local minima of energy.Inspired by the numerical results we present a variational Ansatz in the regime where the interaction energy per particle is stronger than the quantum of vibration in the harmonic trap in the x −y plane,the so-called Thomas-Fermi regime.This Ansatz allows an easy calculation of the energy of the vortices as function of the rotation frequency of the trap;it gives a physical understanding of the stabilisation of vortices by rotation of the trap and of the spatial arrangement of vortex cores.We also present analytical results concerning the possibility of detecting vortices by a time-of-flight measurement or by interference effects.In the final section we give numerical results for a 3D configuration.I.INTRODUCTION After the achievement of Bose-Einstein condensates in trapped atomic gases [1]many properties of these systems have been studied experimentally and theoretically [2].However a striking feature of superfluid helium,quantized vortices [3],[4],has not yet been observed in trapped atomic gases.There is an abundant literature on vortices in helium II,an overview is given in [4].The atomic gases have interesting properties which justify efforts to generate vortices in these systems:the core size of the vortices is adjustable,as in contrast to helium the strength of the interaction can be adjusted through the density;the number of vortices in atomic gases can be in principle well controlled;for a small number of particles in the gas metastability of the vortices can be studied,that is one can watch spontaneous transitions between configurations with different number of vortices.Several ways to create vortices in atomic gases have been suggested.A method inspired from liquid helium consists in rotating the trap confining the atoms [5];at a large enough rotation frequency it becomes energetically favorable at low temperatures to produce vortices;two different paths could be in principle followed:(1)producing first a condensate then rotating the trap,or (2)cooling the gas directly in a rotating trap.It has been recently proposed in [6]to use quantum topological effects to obtain a vortex.Other methods that do not rely on thermal equilibrium have been suggested [7],[8].Here we study theoretically the minimal energy configurations of vortices in a rotating trap [9].The model is defined in section II;in sections III to VI we assume a strong confinement of the atoms along the rotation axis z so that we face an effective 2D problem in the transverse plane x −y .We present numerical results for solutions with vortices that are local minima of the Gross-Pitaevskii energy functional (section III).These solutions contain only vortices with a charge ±1,the vortices with a charge larger than or equal to 2are thermodynamically unstable (section IV).We discuss possibilities to get experimental evidence of vortices in atomic gases in section V.Finally,we concentrate on the regime where the interaction energy is much larger than the trap frequenciesωx,y ,the so-called Thomas-Fermi limit [2].This is complementary to the work of [10].We obtain inthis “strong interacting”regime analytical predictions based on a variational Ansatz that reproducesatisfactorily the numerical results (section VI).In section VII we present results for vortices in 3D,that is in a trap with a weak confinement along the rotation axis.II.MODEL CONSIDERED IN THIS PAPERThe atoms are trapped in a potential rotating at angular velocity Ω.In the laboratory frame theHamiltonian of the gas is therefore time dependent.To eliminate this time dependence we introducea rotating frame at the angular velocity Ωso that the trapping potential becomes time independent;this change of frame is achieved by the single-atom unitary transform:U(t)=e i Ω· Lt/¯h(1) where L is the angular momentum operator of a single atom.As the unitary transform is timedependent the Hamiltonian in the rotating frame contains an extra inertial term,given for eachatom byi¯h U†(t)dφ|φ +1φ|φ 2.(3)In this energy functional H0contains the kinetic energy and the trapping potential energy of theparticles:H0=−¯h22mω2αr2α.(5)Furthermore in all but in section VII we will assume that the trapping potential is much strongeralong the z axis than along the x,y axis,with an oscillation frequency much larger than the typicalinteraction energy Ng3D|φ|2per particle.This situation,although not realized experimentally yet,is not out of reach,in particular when one uses optical traps rather than magnetic traps[11].Inthis strong confining regime the motion of the particles along z is frozen in the ground state of thestrong harmonic potential:φ(x,y,z)≃ψ(x,y) mωz2¯hωz:E[ψ,ψ∗]= d2 rψ∗( r)[H⊥−ΩL z]ψ( r)2Ng|ψ|42m ∆x,y+12π¯h 1/2.(9)Most of the results of the paper are dealing with the2D energy functional;a numerical result for a local minimum of the full3D energy functional will be given in the section VII.We concentrate on the so-called Thomas-Fermi regime,where the interaction energy per particle is much larger than ¯hωx,y.The opposite regime has already been studied in[10].III.LOCAL MINIMA OF ENERGY WITH VORTICES In this section we briefly discuss the general problem of minimizing energy functionals of the type Eq.(7).We present the numerical algorithm that we have used and we give numerical results for the2D problem.A.A numerical algorithm tofind local minimaThe algorithm in our numerical calculations is commonly used in the literature to minize energy functionals E[ψ,ψ∗]of the form Eq.(7).The intuitive idea is to start from a randomψand move it opposite to the local gradient of E[ψ,ψ∗]that is along the local downhill slope of the energy. Numerically this is implemented by an evolution ofψparametrized by afictitious timeτ:−dδψ∗[ψ,ψ∗].(10)Assuming aψnormalized to unity we get the following equation of motion forψ:−ddτE[ψ,ψ∗]=−2 d2 r δE2mΩ2r2cannot exceed the trapping potential.Therefore E has to convergeto afinite value forτ→∞.Asymptotically dE/dτ=0andψsatisfiesδEωx=ω/(1+ǫ)(14)ωy=ω(1+ǫ).(15) In Fig.1we show different local minima configurations obtained forǫ=0.3and a rotation frequencyΩ=0.2ω;each configuration has been obtained for different random initialψ’s.The holes observedin the spatial density correspond to the vortex cores.We have always found that the phase ofψchanges by2πaround a vortex core;we have not found vortices with a charge±q,where the integerq is strictly larger than one;this fact will be explained in the next section.Furthermore the senseof circulation is the same for all vortices.To quantify the effect of the non-axisymmetry of the trap we have plotted in Fig.2the dependenceof energy of different vortex configurations onωx/ωy for afixedω;we measure the energies fromE iso,the energy of the zero-vortex solution in the axisymmetric caseǫ=0.The zero-vortex solutionexhibits a significant variation of energy withǫ;for a non-zeroǫthe wavefunctionψdevelops a phase proportional toΩfor weakΩ’s,which accounts for the energy change as explained in section VI B.The solutions with vortices experience quasi the same energy shift as function ofǫ.As only theenergy difference between the various local minima matters we will from now on only consider the axisymmetric caseǫ=0to identify the solution with the absolute minimal energy.Note that the solutionsψwith several vortices obtained in the limiting caseǫ=0are not eigen-vectors of L z;this reflects a general property of non-linear equations such as the Gross-Pitaevskiiequations to have symmetry broken solutions;it is explained in[10]how to reconcile this symmetrybreaking with the fact that eigenvectors of the full N-atom Hamiltonian are of well defined angular momentum.IV.STABILITY PROPERTIES OF VORTICESIn this section we recall that a(normalized)wavefunctionψsuch that E[ψ,ψ∗]has a local min-imum inψ,describes a condensate having all the desired properties of stability,that is dynamicaland thermodynamical stability.We then show that a vortex centered at r= 0with an angular mo-mentum strictly larger than¯h is not a local minimum of energy and is therefore thermodynamicallyunstable.A.Stability properties of local minimaLet us express the fact thatψcorresponds to a local minimum of the energy.Afirst condition isthat the energy functional is stationary forψ,that isψsolves the Gross-Pitaevskii equation Eq.(13).To get the second condition,we consider a small variation ofψ,ψ→ψ+δψ(16) preserving the normalization of the condensate wavefunction to unity:||ψ+δψ||2−||ψ||2=0= ψ|δψ + δψ|ψ + δψ|δψ .(17) We expand the energy functional E[ψ,ψ∗]in powers ofδψ,neglecting terms of orderδψ3or higher.Using Eq.(17)and Eq.(13)wefind that terms linear inδψvanish so that1δE=1.Dynamical stabilityConsiderfirst the problem of so-called“dynamical stability”:to be a physically acceptable con-densate wavefunction,ψhas to be a stable solution of the time dependent Gross-Pitaevskii equationi¯h∂tψ=H GPψ(21) otherwise any small perturbation ofψ,e.g.the effect of quantumfluctuations or experimental noise,may lead to an evolution ofψfar from its initial value.To determine the evolution of a smalldeviationδψas in Eq.(16)we linearize Eq.(21):i¯h∂t |δψ |δψ∗ =L |δψ |δψ∗ (22) where the operator L is related to L c byL c= 100−1 L.(23)Asψis time independent,so is L and dynamical stability is equivalent to the requirement that theeigenvalues of L have all a negative or vanishing imaginary part.As we now show the positivity ofL c leads to a purely real spectrum for L.Consider an eigenvector(u,v)of L with the eigenvalueε.Contracting Eq.(23)between the ket(|u ,|v )and the bra( u|, v|)we getε[ u|u − v|v ]=( u|, v|)L c |u |v .(24)Note that the matrix element of L c is real positive as L c is a positive hermitian operator.We nowface two possible cases for the real quantity u|u − v|v :• u|u − v|v =0.In this case L c has a vanishing expectation value in(|u ,|v );as L c is positive(|u ,|v )has to be an eigenvector of L c with the eigenvalue zero;from Eq.(23)and the fact that 100−1 is invertible wefind that(|u ,|v ) is also an eigenvalue of L with the eigenvalue0,so thatε=0is a real number.• u|u − v|v >0:we getεas the ratio of two real numbers,so thatεis real.2.Thermodynamical stabilityA second criterion of stability is the so-called“thermodynamical”stability.For zero temperature,this condition can be formulated in the Bogoliubov approach[2],where the particles out of thecondensate,which always exist because of the interactions,are described by a set of uncoupledharmonic oscillators with frequenciesεsign[ u|u − v|v ]/¯h,where(u,v)is an eigenvector of L withthe eigenvalueε.In order for a thermal equilibrium to exist for these oscillators,their frequenciesshould be strictly positive,which is the case here in virtue of Eq.(24)[13].If a mode with a negativefrequency were present thermalization by collisions would transfer particles from the condensateψto this mode,leading to a possible evolution of the system far from the initial stateψ[14].What happens for solutionsψof the Gross-Pitaevskii equations that are not local minima of energy?The operator L c has at least an eigenvector with a strictly negative eigenvalue.In thiscase one cannot have thermodynamical stability,that is one cannot haveε[ u|u − v|v ]>0forall modes[13].From the non-positivity of L c one cannot however distinguish between a simplethermodynamically instability or a more dramatic dynamical instability.B.Why not a vortex of angular momentum larger than¯h?For simplicity we consider only a single vortex in the center of an axi-symmetric trap.We show that vortices with a change of phase of2qπare not local minima of energy,that is are(at leastthermodynamically)unstable.We have found numerically a solution of the Gross-Pitaevskii equationEq.(13)by an evolution in complex time,starting from a wavefunctionψwith an angular momentumq¯h along z,as already done in[15];our solution of the Gross-Pitaevskii equation with imposedsymmetry is a local mimimum of energy in the subspace of functions with angular momentum q¯halong z,but not necessarily a local minimum in the whole functional space,as we will see for|q|>1.In the Thomas-Fermi regimeµ≫¯hωwefind that the solutions can be well reproduced bya variational Ansatz of the formψ(x,y)=e iqθ[tanhκq r]|q| ˜µ−1Ng 1/2(25) whereθis the polar angle in the x−y plane and where˜µ,the chemical potential in the lab frame˜µ=µ+q¯hΩ(26) does not depend onΩ.In this Ansatz the vortex core is accounted for by tanh|q|,a function thatvanishes as r|q|in zero as it should,and the condensate density outside the core coincides withthe Thomas-Fermi approximation commonly used for the zero-vortex solution[2].We calculate themean energy Eq.(7)of the variational Ansatz and we minimize it with respect to the variationalparameterκq;we getκq= ˜µmq2 +∞0du u tanh2|q|(u)−1 2(28) is a number(c1=0.7687,c2=0.5349,...).In order for the vortex of charge q to be a local minimum of energy,the operator L c of Eq.(19)has to be positive.This implies that the operator on thefirst line,first column of L c,the so-calledHartree-Fock Hamiltonian,be positive:H HF=H⊥+2Ng|ψ|2−˜µ+q¯hΩ−ΩL z≥0.(29) To show that this is not the case it is sufficient tofind a wavefunction f(x,y)leading to a negative expectation value for H HF.As the potential appearing in H HF has a dip at r=0we have taken fof a form localized around r=0:1f(x,y)=¯h2 1/2r (30) whereγis adjusted to minimize the expectation value.For e.g.q=2we takeγ=1leading tof|H HF|fsection the trap is axi-symmetry with a time dependent frequencyω(t).We consider the evolutionin the laboratory frame,as the detection is performed in this frame:i¯h∂tψlab= −¯h22mω2(t)r2+Ng|ψlab|2 ψlab.(32)As shown in[19,20]the effect of the time dependence ofω(t)can be absorbed by a scaling and gaugetransform of the wavefunction:1ψlab( r,t)=−ω2(t)λ(34)λ3with initial conditionsλ(0)=1,˙λ(0)=0;if the trap in the x−y plane is abruptly switched offatt=0+the scaling parameter is given byλ(t)==dτ(36)λ2(t)wefind that˜ψsolves the same equation asψlab with a constant trap frequency equal toω(0):i¯h∂τ˜ψ= −¯h22mω2(0)r2+Ng|˜ψ|2 ˜ψ.(37)Asψlab rotates in the trap at the frequencyΩin the lab frame,so does˜ψin terms of the renormalizedtimeτ.In the limit of t→∞,τtends to afinite valueτmax,so that˜ψis rotated by afinite angleduring the ballistic expansion:Ωτmax=Ω ∞0dt2ΩVI.INTUITIVE V ARIATIONAL CALCULATIONTo get a better understanding of the numerical results we now proceed to an intuitive Ansatzfor the wavefunction with several vortices.It coincides very well with the numerical results andallows an easy construction of the minimal energy configurations with vortices.It gives a physical understanding of the stability conditions and of the structure of the solutions:a set of n vortices isequivalent to a gas of interacting particles in presence of an external potential adjusted by the rota-tion frequency of the trap.We restrict to the case of an axi-symmetric trap,a good approximationfor weak(<10%)non-axisymmetries(see section III B).A.Ansatz for the densityTo construct the Ansatz we splitψin a modulus and a phase:ψ(x,y)=|ψ|e iS.(40) In the Thomas-Fermi regime,the modulus in presence of n vortices appears as a slowly varyingenvelope given by the Thomas-Fermi approximation used in the0-vortex case:ψslow= µ−1Ng 1/2(41)with narrow holes digged by the vortices with charge q=±1,represented by tanh functions ofadjustable widths and with zeros at adjustable positions:|ψ|=ψslow×nk=1tanh[κk| r− αk R|].(42)The positions of the vortex cores αk are expressed in units of the Thomas-Fermi radius R of the condensate:R= mω2.(43)From section IV B we expect as typical values for the inverse width of the vortex coresκk≃(mµ/¯h2)1/2.The chemical potential is not an independent variable but is expressed as a functionof the other parameters from the normalization condition ψ|ψ =1;neglecting overlap integralsbetween the holes we getµ=µ0 1+2n k=1(1−α2k)ln2(κR)4) (44)where1/(κR)4∼(¯hω/µ)4≪1and whereµ0is the Thomas-Fermi approximation for the condensatechemical potential without vortices:µ0= mω2Ngcentered on the vortex core and S 0is the single-valued part of the phase.The function S 0can in principle be determined from the modulus of ψfrom the continuity equation:div[|ψ|2 v ]=0.(47)The local velocity field v is related to the phase S byv =¯h ¯h [x∂y −y∂x ]|ψ|2=0.(49)This can be turned into an equation for the single-valued part S 0of the phase;because the density |ψ|2in a trap vanishes at the border of the condensate S 0is uniquely determined (up to a constant)by the resulting equation (see Appendix);this is to be contrasted to the case of superfluid helium in a container,where the flux,not the density,vanishes at the border,which requires a boundary condition on the gradient of the phase.Eq.(49)can be solved for a non-axisymmetric trap in the absence of vortices.The solution is given byS (x,y )=−m Ωω2x +ω2yxy (50)which leads to a change in the energy per particleδE =−1(ω2x +ω2y )ω2x ω2y(51)where µT F is the Thomas-Fermi approximation for the chemical potential for Ω=0,µT F =(mωx ωy Ng/π)1/2[23].As can be seen in Fig.2this prediction is in good agreement with our numerical results.In presence of vortices the equation for S is more difficult to solve analytically.From now on we consider the case of an axi-symmetric trap,as the energy ordering of the vortices solutions is not affected for weak (<10%)non-axisymmetries (see section III B).For a single vortex at the center of the trap one can see that S 0=0solves Eq.(49).From the spatial dependence of the phase obtained numerically (section III B)for a displaced vortex or several vortices we have identified the following heuristic Ansatz,obtained in setting ωx =ωy =ωin Eq.(50):S 0(x,y )≡0(52)that we will use in the remaining part of the section.C.Further approximations for the mean energyIn the calculation of the mean energy,we make some further approximations in the spirit of the Ansatz Eq.(42).The reader not interested by these more technical considerations can proceed to the next subsection.The kinetic energy involves an integral of the gradient squared of the wavefunction:| ∇ψ|2=|ψ|2 ( ∇ln |ψ|)2+( ∇S )2 .(53)For the gradient of the modulus of ψwe neglect the variation of the slow envelope ψslow :∇ln |ψ|≃nk =1κktanh ′of ψinvolves holes with a density varying as 1−tanh 2=sech 2.In the following we keep the sech 2for the vortex k only if it is multiplied by ( ∇θk )2,a quantity diverging in the center of the core;the other terms lead toconverging integrals smaller by a factor (µ/¯h ω)2,which is the inverse surface of a vortex core ( d 3 r |ψslow |2sech 2κr ∝1/(κR )2).This finally leads toE kin ≃¯h 22mω2r 2+13µ0+nk =1W ( αk ,κk )+1µ012ln(1−α2)−q ¯h Ω (1−α2)2 +µ0(κR )2(60)where C =0.495063.The lines inEq.(60)correspond successively to E kin ,E rot and E pot .This can be seen as an effective potential for the vortices.One can check that the part of W independent of Ωexpells the vortex core from the trap center,whereas the part proportional to Ωprovides a confinement of the vortex core (see the following subsection).The vortex interaction potential is given by12m q αq β d 2 r |ψslow |2 ∇θ αR · ∇θ βR .(61)This interaction term is equivalent to the one found in the homogeneous case and describes a repulsive interaction for vortices turning in the same direction (q αq β>0)and is attractive for vortices with opposite charges [9].An attractive interaction will lead to the coalescence and consequently annihilation of vortices with opposite charges.Therefore we find in stationary systems always vortices with equal charges.As the interaction potential V ( α, β)does not depend on the parameters κwe can optimize sepa-rately the self-energy part with respect to κand find(κR)2=ξ2(1−α2) µ03(4ln2−1) 1/2≃1.08707.By rewriting the above equation as¯h2κ22ξ2 µ−1µ0 13+lnνµ0¯hω2(1−α2) (64) whereν=0.49312.E.Case of a single vortex:critical frequenciesIn Fig.4we have plotted the self-energy of a vortex as a function of the displacement of the corefrom the trap center,for different values of the rotation frequencyΩ.The analytical predictioncoincides very well with the numerical value[24].ForΩ=0the position of the vortex at the trap center gives an energy maximum.ForΩ>0therotation of the trap provides an effective confinement of the vortex core at the center of the trap forpositive charges q(see the term proportional toΩin Eq.(64));from now one we therefore take allthe charges q k to be equal to+1.For a large enoughΩwe reach a situation where a vortex at thetrap center corresponds to a local energy minimum,by further increasingΩthe vortex state at thetrap center becomes a global minimum with energy less than the condensate without vortex.The above suggests that we have to distinguish two critical rotation frequencies:Thefirst onedefines the frequencyΩstab above which the vortex is a local minimum of energy.Above the frequencyΩc the single vortex solution has an energy lower than the condensate without vortex.We calculateΩstab from the condition d2W/dα2=0atα=0andΩc from the condition W=0atα=0:Ωc=¯hω2¯hω (65)Ωstab=¯hω2¯hω (66)where C′=e(2ln2+1)/3+1/2ν≃1.8011.As we are in the regimeµ0≫¯hωΩc is approximately twice Ωstab[25].Our prediction forΩc scale as(logµ0)/µ0as in[15],with a coefficient C′leading to better agreement with the numerics.F.Case of several vorticesBy integrating Eq.(61)we get an explicit form for the vortex interaction potential for vortices with equal charges:V( α, β)=(¯hω)2(1− α· β)+1| α− β|4 (67)At short distances between the two vortex cores the logarithmic term in the above expression dominates,leading to a repulsive potential∼−2(1− α· β)log| α− β|(¯hω)2/µ0.In Fig.5we plot the interaction energy between a vortex at the center of the trap and one of equal charge displaced by αR;the interaction is purely repulsive.A conclusion which essentially holds as well for arbitrary vortex positions.In Fig.6we show the total(interaction+self-energy)for two vortices symmetrically displaced from the trap center,as function of the displacement;the analytical prediction coincides again very well with the numerical results[24].To obtain the equilibrium distance between the two vortex cores one minimizes the total energy overαin Fig.6.To get the minimal energy configurations as function of the rotation frequency of the trap,we minimize our analytical prediction for the energy over the positions of the n=1,2,...vortex cores.The result is shown in Fig.7.Each curve corresponds to afixed value of n;it starts atΩ=Ωstab(n)(forΩ<Ωstab(n)there is no local minima of energy with n vortices);it becomes the global energyminimum forΩ=Ωc(n).We have plotted these two critical frequencies as function of n in Fig.8.We have also given numerical results(circles)in Fig.7.Even if there is good agreement between analytical and numerical results,we still need a numerical calculation to check the stability of thesolutions;our simple analytical Ansatz is indeed not sufficient to predict the destabilization of agiven vortex configuration at highΩ,a phenomenon studied with a numerical calculation of theBogoliubov spectrum for a single vortex in[26].For afixed value of the number of vortices n there may exist local minima of energy,in addition to the global minimum plotted in Fig.7,a situation known from superfluid helium[4].E.g.for n=6(see Fig.9)the global minimum of energy is given by a configuration with six vortex cores on acircle;there exists also a local minimum of energy with one vortex core at the center of the trap andfive vortex cores on a circle.The energy difference per particle between the two configurations isvery small,δE≃0.002¯hωfor the parameters of thefigure and probably beyond the accuracy of ourvariational Ansatz.For relatively large rotation frequenciesΩone canfind local minima of energyconfigurations with many vortices(see[4]for superfluid helium);we plot two configurations with18vortices in Fig.10,with an energy differenceδE=0.0034¯hω.In estimating the physical relevance of these energy differences one should keep in mind that NδE matters,rather thanδE,where N is the number of particles in the condensate: e.g.at afinite temperature T the ground energy configuration is statistically favored as compared to themetastable one when NδE≫k B T.VII.VORTICES IN A3D CONFIGURATIONWe have extended the numerical calculation to the case of a3D cigar-shaped trap,that is witha confinement weaker along the rotation axis than in the x−y plane.Even in this case rotationof the trap can stabilize the vortex.We show in Fig.11density cuts of a solution with5vortices;the vortex cores are almost straight lines in the considered Thomas-Fermi regime,except at vicinityof the borders of the condensate.As in section VI the core diameter is determined by the localchemical potential in the gas.This suggests that our2D Ansatz(section VI)can be generalized to3D situations,with αk and κk depending on z.VIII.CONCLUSION AND PERSPECTIVESWe have presented in this paper an efficient numerical algorithm and a heuristic variational Ansatz to determine the local minima energy configurations for a Bose-Einstein condensate stronglyconfined along z and subject to a rotating harmonic trap in the x−y plane.Our results can be used as afirst step towardsfinite temperature calculations.Interesting prob-lems are e.g.the critical temperature for the vortex formation and the Magnus forces induced bythe non-condensed particles on the vortex core[27].Acknowledgement:We acknowledge useful discussions with Sandro Stringari and Dan Rokhsar.We thank J.Dalibard for useful comments on the manuscript.We thank the ITP at Santa Barbarafor its hospitality and the NSF for support under grant No.PHY94-07194.This work was partiallysupported by the TMR Network“Coherent Matter Wave Interactions”,FMRX-CT96-0002.∗L.K.B.is an unit´e de recherche de l’Ecole Normale Sup´e rieure et de l’Universit´e Pierre et Marie Curie,associ´e e au CNRS.APPENDIX A:UNIQUENESS OF THE PHASE FROM THE CONTINUITY EQUATION IN A TRAPConsider two solutions S1and S2of the continuity equation:mΩdiv[|ψ|2 ∇S]=。

三角翼涡发生器强化传热

三角翼涡发生器强化传热

Flow structures and heat transfer characteristics in fan flows withand without delta-wing vortex generatorsT.Y.Chen *,H.T.ShuDepartment of Aerospace Engineering,Tamkang University,Taiwan,ROCReceived 28April 2002;accepted 20May 2003AbstractEffects of an external delta-wing vortex generator on the flow and heat transfer characteristics in fan flows and uniform flows were experimentally investigated and compared.A heated plate,installed on the bottom wall of a duct,was used as the heat transfer surface.Three-component mean and fluctuating velocity measurements were conducted using a laser Doppler velocimetry to characterize the flow structures and to obtain the near-wall flow parameters,including the axial mean velocity,axial vorticity and turbulent kinetic energy.Temperatures on the heat transfer surface were measured using thermocouples to obtain the Nusselt numbers.Results show that the external delta-wing vortex generator in fan flows has little overall effect on the near-wall averaged axial mean velocity and axial vorticity,but increases the turbulent kinetic energy,in the investigated X =D ranges.The increase in the turbulent kinetic energy by the delta-wing has little effect on heat transfer in the inherently vortical fan flows.Consequently,the delta-wing vortex generator in fan flows has little effect on the heat transfer augmentation.Ó2003Elsevier Inc.All rights reserved.Keywords:Fan flow;Vortex generator;Heat transfer1.IntroductionExtensive studies have been conducted on the heat transfer enhancement using turbulators in simulated heat exchangers,turbine blades and so on [1–14].The turbulator usually generates high turbulence vortical motion in uniform flows,which may change the mean velocity fields,modify the flow turbulence properties and the structures of the near wall layers in the velocity boundary layer.These,in turn,cause modifications in the heat transfer characteristics.Edwards and Alker [1]investigated convective heat transfer augmentation using three-dimensional surface protrusions in the form of cubes and vortex generators.The cube was found to produce the highest local improvement,but the vortex generator effect extended further downstream.Russell et al.[2]studied the heat transfer enhancement on plate fin surfaces on which local vortex generators produced aset of contra-rotating vortices.Their study showed that considerable heat transfer enhancement occurred,and the pressure loss penalty associated with vortex gener-ators was modest.Eibeck and Eaton [3]experimentally examined the heat transfer effects of an isolated longi-tudinal vortex embedded in a turbulent boundary layer.They found that the longitudinal vortex effects on the Stanton number were attributed largely to the distortion in the mean velocity field.Fiebig and his group [5–8]conducted systematic experimental investigations on heat transfer enhancement and induced drag using delta wings,rectangular wings,delta winglets and rectangular winglets in channel uniform flows.They found that the heat transfer was increased up to 60°angle of attack.The heat transfer enhancement per unit vortex generator area was the highest for delta wings followed by delta winglets and rectangular winglets [5].However,the re-sults from Ref.[7]showed that delta winglets produced better heat transfer performance than wings.No expla-nations were given on these contradictory results.Their studies also showed that the vortices in the wake of the second row were more unsteadiness than the first-row vortices [6].A pair of delta winglets performed slightly*Corresponding author.Tel.:+886-2-2629-4728;fax:+886-2-2620-9746.E-mail address:yuan@.tw (T.Y.Chen).0894-1777/$-see front matter Ó2003Elsevier Inc.All rights reserved.doi:10.1016/S0894-1777(03)00107-9Experimental Thermal and Fluid Science 28(2004)273–282/locate/etfsbetter than a pair of rectangular winglets at higher angles of attack and Reynolds numbers[8].The results from studies by Biswas et al.[9,10]indicated that the flow loss due to the winglet-pair was less than that due to the wing,and the zone of poor heat transfer as it was observed with the wing could be avoided by using win-glet-pair.The use of winglet-pair appeared to be a more attractive augmentation technique.Hwang et al.[12] and Liou et al.[13]investigated the effects of ridge shapes and rib-arrays on heat transfer.Their studies showed that various ridge shapes have comparable thermal performance and the composite-ribbed per-formed best in their ribbed configurations.Liou et al.[14]also studied the effect of divider thickness on the local heat transfer distributions.Their results showed that the direction and strength of the secondaryflow are the most importantfluid dynamic factors affecting the heat transfer distributions,followed by the convective mean velocity and then the turbulent kinetic energy.The previous studies focused on the investigations of turbulator effects on heat transfer in uniformflows,and indicated that the turbulators have substantial effects on heat transfer rates.Fans have been widely used in elec-tronics and computers to inducefluid motions for heat transfer applications.The fan itself is a good turbulator, which generates vorticalflow motions.The studies of an external turbulator effect on heat transfer andflow characteristics in such inherently vortical fanflows have not been seen in literatures,which motivates this research.Cross-sectionalflow velocities werefirst mea-sured to characterize theflow structures.Three near-wallflow parameters,including the axial mean velocity, axial vorticity and turbulent kinetic energy,were then obtained from three-component mean andfluctuating velocities.Temperature distributions on a heated alu-minum plate,acting as the heat transfer surface,were measured to obtain the Nusselt numbers.The turbulator effects on theflow and heat transfer characteristics were discussed.2.Experimental setup and methods2.1.Experimental setupThe primary experimental setup utilized in this re-search is shown in Fig.1.An open-circuit wind tunnel system was used to develop uniformflows with turbu-lence intensities less than1%inside a7·7cm2duct.Air was supplied to the50cm long test section by a blower, and passed through a turbulence-management section that includes a diffuser,honeycombs,screens and a contraction section.An axial DC fan was placed be-tween the contraction section and the test section to develop fanflows inside the duct,where the diffuser,NomenclatureA surface area of the heated plate(0.008m2) D duct hydraulic diameter(0.07m)h heat transfer coefficient(W/m2K)k thermal conductivity of the air,evaluated at film temperature(W/m K)k2near-wall turbulent kinetic energy(m2/s2) Nu Nusselt numberQ in power input to the heated plate(W)Q loss conduction and radiation losses(W)Re D Reynolds number based on the duct hy-draulic diameterT w surface temperature of the heated plate(K) T o inlet air temperature(K)U av near-wall averaged axial mean velocity(m/s)U c duct center velocity(m/s)U d duct average velocity(m/s)U0X-componentfluctuating velocity(m/s)U near-wall axial mean velocity(m/s)V0Y-componentfluctuating velocity(m/s)v local instant velocity(m/s)v local mean velocity(m/s)W0Z-componentfluctuating velocity(m/s)X spatial coordinate,axial directionY spatial coordinate,normal directionZ spatial coordinate,spanwise directiong axial vorticity(sÀ1)m kinematic viscosity of the air(m2/s)Fig.1.The experimental setup for the uniform-flow and fan-flow systems.274T.Y.Chen,H.T.Shu/Experimental Thermal and Fluid Science28(2004)273–282honeycombs and screens were removed from the wind tunnel system.The 8·8cm 2fan has seven blades with the maximum airflow of 45.2CFM,produced by Yen-Sun Tech,Taiwan.A delta-wing with 40°angle of attack was used as a turbulator,which is placed on the bottom wall of the duct,10cm downstream of the turbulence-management section or the fan,see Figs.1and 2.The base width and tip height of the delta-wing are 7and 3.5cm,respectively.The origin of the coordinate system used for data presentation is at the delta-wing location.An aluminum plate,4·20cm 2,was fixed to the bottom wall of the test section,20cm downstream of the delta-wing between X =D ¼2:856and 5.712to serve as the heat transfer surface.The 4cm width of this aluminum plate is smaller than the duct width to reduce the corner-effect on heat transfer,where secondary vortices or corner vortices possibly occur at duct corners.The in-vestigated flows include the fan flows and uniform flows with and without the delta-wing vortex generator.2.2.Velocity measurementThree-component,mean and fluctuating velocity measurements were performed using a TSI laser Doppler velocimetry (LDV).A new test section,made of plexiglass to provide optical access,was used for the velocity measurements.The duct average velocities were 1, 1.62, 2.13and 2.67m/s,where the corresponding Reynolds numbers are 4430,7200,9440and 11820,re-spectively,based on the duct hydraulic diameter.These measurements were conducted at 121locations in each Y –Z plane,see Fig.3,at four X =D stations of 1.428,2.856,4.284and 5.712,and 140locations in each X –Y plane at two Z =D stations of )0.286and 0.286to characterize the flow structures.Each velocity was cal-culated from 4096collected velocity samples.The mean velocity ( v )is the averaged value of the 4096velocitysamples and the fluctuating velocity is obtained from ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiP v 2À v 2p ,where v is the local instant velocity.Thecross-sectional axial component of vorticity (axial vor-ticity)at each Y –Z plane was obtained from the 242Y -and Z -component mean velocities.The circulation of a small block,see Fig.3,was first calculated using the velocities at the block corners.The cross-sectional axial vorticity was the 100-block circulations divided by the 100-block area.Three-component velocity measure-ments near the heat transfer surface were also conducted because these velocities play an important role in heat transfer characteristics.These near-wall velocity mea-surements were conducted at 72locations between Y ¼0:2and 1.6cm (Y =D ¼0:0286and 0.2286),see Fig.3,at two X =D stations of 2.856and 5.712,which cover the uniform-flow velocity boundary layers conducted in this research.The near-wall axial vorticity was obtained from the mean Y -and Z -components velocities,calcu-lated similar to that of the cross-sectional axial vorticity.The particles for the LDV measurements were generated by a smoke wire located upstream of the apparatus.Particle diameters in the sub-micron range have been shown to be able to follow turbulence frequencies ex-ceeding 1kHz [15].The uncertainty in the velocity measurements was estimated to be less than 5%.2.3.Heat transfer measurementThe heat transfer coefficient,h ¼ðQ in ÀQ loss Þ=ðA ðT w ÀT 0ÞÞ,was measured on the heat transfer surface (aluminum plate).The heat transfer plate consists of the 4·20cm 2,1mm thick,aluminum plate (heat transfer surface),a heating plate,two electrical insulation plates and a fiberglass plate,see Fig.2.Electrical power was continuously supplied to the heating plate from a DC power supply,resulting in a constant-heat-flux heat transfer surface.The air inlet temperature,T 0,is the reference temperature,obtained by a type-T,0.1mm-diameter thermocouple placed at the contractionsectionFig.2.The bottom wall of the test section,showing the delta-wing,the coordinate system,the heat transfer plate and the temperature mea-surement locations (unit:cm).Fig.3.The cross-sectional view of the duct,showing the velocity measurement locations and the small blocks for the calculation of axial vorticities (unit:cm).T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–282275of the wind tunnel.The temperatures were read using a TempScan/1100system produced by IOtech,Inc.,with the uncertainty within 0.5°C.The surface temperatures,T w ,of the heated aluminum plate were measured using 25thermocouples.The junction beads of thermocouples were cemented,using thermally conducting epoxy,in small holes on the back of the aluminum plate approx-imately 0.5mm from the top surface.The obtained temperatures were negligibly different from the surface temperatures due to the thin and highly conductive aluminum plate.The averaged temperature from the 25thermocouples is used to calculate the average heat transfer coefficient on the heat transfer surface.The power input,Q in ,to the heating plate is 12.4W with 3%uncertainty,resulting in the surface temperatures between 50and 85°C with the uncertainty of 1°C from repeatability tests.Eight more thermocouples were placed in the insulation fiberglass plate,six on the heating side and two on the ambient side of the plate.The averaged temperatures from the six and two ther-mocouples,respectively,are used to determine the temperature drop through the plate.Heat conduction loss through the back of the heating plate is calculated using the Fourier’s law,which is estimated to be less than 3%of the total heat generated.Radiation loss is estimated to be less than 3.6%of the total heat flux.Conduction loss along the sides of the heat transfer plate accounted for a small fraction and was neglected.The total uncertainty in the measurement of the heat transfer coefficient was estimated to be within10%.Fig.4.The secondary flow velocity vectors in Y –Z plane for (a)the uniform flow of Re D ¼11820across the delta-wing at X =D ¼2:856;(b)the fan flow of Re D ¼11820at X =D ¼2:856and (c)the fan flow of Re D ¼11820across the delta-wing at X =D ¼2:856.276T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–2823.Results and discussion3.1.Cross-sectional flow characteristicsThe cross-sectional flow structures in the Y –Z and X –Y planes for the investigated flows were first dis-cussed.Fig.4(a)–(c)present typical secondary flow ve-locity vectors in Y –Z planes looking in downstream flow direction.The uniform flow across the delta-wing gen-erates two vortices of different directions,and the fan generates a symmetrical,counter-clockwise vortical flow.Also,the secondary flow velocities for the fan flow are much larger than those for the uniform flow across the delta-wing,see Fig.4(a)and (b).It is noted from Fig.4(c)that the secondary flow structure for the fan flow across the delta-wing is generally similar to that for the fan flow though the vortex center was slightly lifted upward.This is possibly due to the flow blockage by the bottom-wall delta-wing that diverts the flow a little up-ward.The secondary flow structures for the investigated flows at different X =D stations are qualitatively similar to those shown in Fig.4(a)–(c),respectively,but with small velocities as the vortices develop downstream.Fig.5(a)–(f)present typically mean flow velocity vectors in X –Y planes.Note that the delta-wing has 40°angle of attack.Its base is at X =D ¼0,and its tip point is at Y =D ¼0:5and Z =D ¼0.Fig.5(a)and (b)show that a close recirculation zone occurred in the immediate neighborhood behind the delta-wing at Z =D ¼À0:286,while no reversed flow was observed at Z =D ¼0,for the uniform flow across the delta-wing.As the flow develops downstream,the delta-wing effect on the flow structures become less.The flows were vortical at the inlets of the heat transfer surface (X =D ¼2:856)with the flow ve-locities slightly smaller and larger than the duct aver-aged velocity (2.13m/s)at Z =D ¼À0:286and 0,respectively,and with the velocity components slightly toward the heat transfer surface.Fig.5(c)and (d)show that the fan flow has downward and upward velocity components at Z =D ¼À0:286and 0.286,respectively,due to counter-clockwise rotation of the vortical flow motion.The delta-wing effects on the fan-flow structures are presented in Fig.5(e)and (f).A very small reversed flow region was observed at Z =D ¼0:286,while no re-versed flows occurred at Z =D ¼À0:286.It is noted that the fan flow is counter-clockwise rotated as it develops downstream.On the up-wash side of the vortex (Z =D ¼0:286),the flows are blocked by the delta-wing that causes the decrease in flow velocities downstream of the delta-wing.The flows are accelerated at Z =D ¼À0:286to maintain the constant flow rate at each cross section of the confined duct flow.The delta-wing effect on the fan-flow structures becomes less as the vortex develops downstream.Fig.5(c)–(f)show that the flow characteristics at the inlets of the heat transfer surface are generally similar between the fan flows with and without the delta-wing.It can be expected that the complicated flow structures in the immediate neighbor-hood behind the delta-wing should have different effects on heat transfer,which will be discussed in other paper.The cross-sectional normalized axial vorticity,ðg D =U d ÞCS ,distributions for the investigated flows of Re D ¼11820are shown in Fig.6,where the axial vor-ticity of the uniform flow across the delta-wing is the magnitude sum of the two opposite vortices.Fig.6shows that the fan generates stronger vortices than the uniform flow across the delta-wing,and the vortex strengths decrease as the vortices develop downstream.The values of ðg D =U d ÞCS are 4.76and 3.19at X =D ¼1:428and 5.712,respectively,for the fanflow,Fig.5.The mean flow velocity vectors in X –Y plane for (a)the uniform flow of Re D ¼9440across the delta-wing at Z =D ¼À0:286;(b)the uniform flow of Re D ¼9440across the delta-wing at Z =D ¼0;(c)the fan flow of Re D ¼9440at Z =D ¼À0:286;(d)the fan flow of Re D ¼9440at Z =D ¼0:286;(e)the fan flow of Re D ¼9440across the delta-wing at Z =D ¼À0:286and (f)the fan flow of Re D ¼9440across the delta-wing at Z =D ¼0:286.T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–282277and are 2.31and 0.77at X =D ¼1:428and 5.712,re-spectively,for the uniform flow across the delta-wing.The loss in the axial vorticity for the fan flow (33%)is less than that for the uniform flow across the delta-wing (67%)from X =D ¼1:428to 5.712.The vortex generated by the fan persists further downstream than that gen-erated by the uniform flow across the delta-wing.Fig.6also shows that the delta-wing in fan flows has little effect on the axial vorticity,which causes a little increase and decrease at X =D ¼1:428and 5.712,respectively.This result should be due that the fan-generated vortex is strong enough such that the vortex structure is only slightly disturbed by the delta-wing,as discussed in Fig.4(b)and (c).It is also possibly caused by the interaction of vortex systems generated by the fan and the delta-wing.It is noted from Fig.4(a)that the uniform flow across the delta-wing produced a counter-rotating vor-tex system,where the two vortices have different direc-tion of rotation.The interactions of these two vortices may result in additive and subtractive effects in vortex strength and,thus,the net effect becomes insignificant.3.2.Flow characteristics near the heat transfer surface The axial mean velocities,the axial vorticities and the turbulent kinetic energy near the heat transfer surface were investigated next because the near-wall flow char-acteristics should have major effects on heat transfer.Measurements showed that the boundary-layer thick-nesses for the uniform flows of the investigated Rey-nolds numbers over the heat transfer surface were between 9and 13mm.Typical measured boundary-layer velocity profiles at X =D ¼2:856,Z =D ¼0are shown in Fig.7.The exact Blasius solution for a laminar flow over a flat plate is also presented forcomparison,Fig. 6.The cross-sectional normalized axial vorticity,ðg D =U d ÞCS ,distributions for the investigated flows of Re D ¼11820(U d ¼2:67m/s).Fig.7.The measured boundary-layer velocity profiles for the uniform flows of the investigated Reynolds numbers and the theoretical Blasiussolution.Fig.5(continued )278T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–282which shows good agreements between the theory and experiments.The velocity data between Y ¼2and 12mm for the flows of Re D ¼9440and 11820,and between Y ¼2and 14mm for the flows of Re D ¼4430and 7200were,thus,used to calculate the near-wall flow param-eters.3.2.1.Axial mean velocity (convective effect)The axial mean velocity is an indication of the con-vective effect on heat transfer.Its magnitude is expected to be proportional to the convective heat transfer rate.Fig.8(a)–(c)present some results of the near-wall axial mean velocity contours for the investigated flows of Re D ¼11820at X =D ¼2:856,showing the profound effects of vortices on the velocity distributions.The near-wall axial mean velocity distributions are completely different from the uniform-flow boundary-layer velocity distribution.Fig.8(a)shows that the velocities on the left and right sides of the flow region are smaller than those on the central region for the uniform flow across the delta-wing.Fig.8(b)and (c)indicate that the ve-locities on the left sides of the flow region are larger than those on the right sides of the flow region for both the fan flows with and without the delta-wing,due to the vortex characteristics as presented in Fig.4(a)–(c).On the up-wash side of the vortex,low-speed fluid is lifted away from the wall,while on the downwash side of the vortex,high-speed fluid is swept into the near wall re-gion.These findings are consistent with the results byEibeck et al [3].These figures also show that the near-wall axial mean velocities are mostly larger than the duct average velocity.The averaged axial mean velocity,U av ,averaged form the near-wall axial mean velocities,is used to represent the overall convective effect on heat transfer.Fig.9presents the variations of the normalized averaged axial mean velocity,U av =U d ,with Reynolds number for the investigated flows at X =D ¼2:856and 5.712,showing the weak dependence of U av =U d on Reynolds number.The averaged axial mean velocities slightly decrease from X =D ¼2:856to X =D ¼5:712for the fan flows,while there is up to 22%decrease for both fan and uniform flows across the delta-wing.Because of the constant flow rate in each duct cross-section,this result reveals that the delta-wing causes the near-wall flow deceleration as the flows develop downstream.3.2.2.Axial vorticity (secondary-flow effect)Large strength of the secondary flow is expected to have a large effect on heat transfer.Fig.10presents the variations of the near-wall normalized axial vorticity,ðg D =U d ÞNW ,with Reynolds number for the investigated flows at two X =D stations of 2.856and 5.712.The normalized axial vorticities generally remain constant with Reynolds numbers,and decrease as the flows de-velop rger axial vorticities are generated in fan flows than in uniform flows across the delta-wing,as expected.It has been shown in Fig.9thattheFig.8.The near-wall axial mean velocity contours for (a)the uniform flow of Re D ¼11820across the delta-wing at X =D ¼2:856;(b)the fan flow of Re D ¼11820at X =D ¼2:856and (c)the fan flow of Re D ¼11820across the delta-wing at X =D ¼2:856.T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–282279delta-wing in the investigated flows causes the near-wall flow deceleration from X =D ¼2:856to 5.712,which possibly results in the near-wall flow relaminarization.As a result,the fan flows across the delta-wing have larger and smaller axial vorticity than the fan flows at X =D ¼2:856and X =D ¼5:712,respectively.The delta-wing vortex generator has,thus,little overall effect on the near-wall axial vorticity over the heat transfer surface.3.2.3.Turbulent kinetic energy (turbulence effect)The turbulent kinetic energy,k 2¼ðU 02þV 02þW 02Þ=2,was obtained from the X -,Y -and Z -component fluctuating rge turbulent kinetic energy is expected to have a large effect on heat transfer rate.Fig.11presents the variations of the near-wall normalizedturbulent kinetic energy,k 2=U 2d ,with Reynolds number for the investigated flows at X =D ¼2:856and 5.712,showing the weak dependence of k 2=U 2d on Reynoldsnumber.The delta-wing increases the turbulent kinetic energy up to 80%both in fan flows and uniform flows.However,it causes the turbulent kinetic energy decrease fast as the vortices develop downstream.Fig.11indi-cates that the turbulent kinetic energy at X =D ¼5:712for the fan flows across the delta-wing is approximately 10%less than that for the fan flows.The fast decrease in the turbulent kinetic energy caused by the delta-wing is also possibly due to the near-wall flow relaminarization as discussed above.The overall effect of the delta-wing vortex generator is to increase the turbulent kinetic en-ergy over the heat transfer surface in fan flows and uniform flows.3.3.Heat transfer characteristicsThe heat transfer characteristics on the heat transfer surface are presented in a dimensionless form as the Nusselt number,Nu ¼hD =k .The Nusselt numbers on the heat transfer surface for the uniform flows of the investigated Reynolds numbers were also measured and compared with the analytical unheated starting length solutions presented in Kays and Crawford [16]to check the accuracy of the measured heat transfer data.Fig.12shows that the Nusselt number distribution is approxi-mately 6%higher than the predictions,which is under the uncertainty estimate.Due to the vortical flow mo-tions,the fan flows and the uniform flows across the delta-wing have much better heat transfer performance than the uniform flows.Also,the Nusselt numbers in-crease as the Reynolds numbers increase,and the Nus-selt numbers are higher for the fan flows than the uniform flows across the delta-wing.The delta-wing in uniform flows causes large increase in the Nusselt number,while it has little effect on heat transfer rates in fan flows in the investigated X =Dranges.Fig.9.The near-wall averaged axial mean velocities,U av =U d ,for the investigated flows at X =D ¼2:856and 5.712vs.Reynoldsnumber.Fig.10.The near-wall normalized axial vorticities,ðg D =U d ÞNW ,for the investigated flows at X =D ¼2:856and 5.712vs.Reynoldsnumber.Fig.11.The near-wall normalized turbulent kinetic energy,k 2=U 2d ,for the investigated flows at X =D ¼2:856and 5.712vs.Reynolds number.280T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–2823.4.Relations between heat transfer and flow character-isticsIt has been shown above that larger near-wall aver-aged axial mean velocity,axial vorticity and turbulent kinetic energy occurred in fan flows and uniform flows across the delta-wing than in uniform flows.As a result,the vortical flows have higher heat transfer rates than the uniform flows.Fig.13presents the Nusselt number ra-tios between the investigated flows and the uniform flows,Nu x =Nu 0,and between the fan flows with and without the delta-wing,Nu 3=Nu 2.The increases in Nus-selt numbers for the fan flows and uniform flows across the delta-wing were up to 115%and 65%,respectively.The delta-wing has large effect on heat transfer aug-mentation in uniform flows.However,the delta-wing ininherently vortical fan flows has little effect on heat transfer in the investigated X =D ranges.The values of Nu 3=Nu 2are approximately 1.02for all of the investi-gated Reynolds numbers.To elucidate the relations be-tween the near-wall flow parameters and the heat transfer in fan flows,two 4·5cm 2heat transfer plates were placed between X =D ¼2:5and 3.214,and between X =D ¼5:357and 6.071,respectively,for heat transfer measurements.Nine thermocouples were embedded on each heat transfer plate,and the averaged temperature from these nine thermocouples was used to calculate the Nusselt number on that plate.Fig.13shows that the Nusselt number ratios between the fan flows with and without the delta-wing for the heat transfer plate placed between X =D ¼2:5and 3.214,ðNu 3=Nu 2ÞX =D ¼2:856,are approximately 1.05for all of the investigated Reynolds numbers.Figs.9–11have shown that the near-wall av-eraged axial mean velocities and axial vorticities in fan flows were a little increased,while the turbulent kinetic energy was largely increased,by the delta-wing at X =D ¼2:856.If the turbulent kinetic energy induced by the delta-wing plays an important role in heat transfer,the Nusselt number for the fan flows with the delta-wing should,accordingly,be largely increased.These results suggest that the turbulent kinetic energy in fan flows should not have a dominated effect on heat transfer rate.The near-wall averaged axial mean velocities,axial vor-ticities and turbulent kinetic energy for the fan flows across the delta-wing were slightly smaller than those for the fan flows without the delta-wing at X =D ¼5:712.Thus,the delta-wing has little effect on heat transfer rate at X =D ¼5:712,where the Nusselt number ratios for the heat transfer plate placed between X =D ¼5:357and 6.071,ðNu 3=Nu 2ÞX =D ¼5:712,are approximately 0.97.3.5.Practical significanceFans have been widely used in electronics and com-puters as the pumping sources to induce flow motions for heat transfer applications.The fan generates vortical flow motions that result in higher heat transfer rates than the uniform flows.The needs for even faster heat transfer rates are urgent,for example,in CPU cooling of modern computers.An understanding of the fan-flow charac-teristics and their relations with the heat transfer would be helpful to improve the heat transfer rate for engi-neering applications.The delta-wing vortex generator in uniform flows augments the heat transfer.However,it has negligible effect on heat transfer augmentation in fan flows in the investigated X =D ranges.One of the possible reasons is due to the cancellation of the vortex systems generated by the fan and the delta-wing vortex generator.Other types of vortex generator may have different effect on heat transfer.The other possible reason is that the delta-wing effect on the near-wall flow characteristics does not persist downstream enough.This researchhasFig.12.The Nusselt numbers,Nu ,on the heat transfer surface for the investigated flows vs.Reynolds number and the analytical unheated starting length (USL)solutions.Fig.13.The Nusselt number ratios between the investigated flows and the uniform flows,Nu x =Nu 0,and between the fan flows with and without the delta-wing,Nu 3=Nu 2,where Nu 0,Nu 1,Nu 2,Nu 3are the Nusselt numbers for the uniform flows without and with the delta-wing,the fan flows without and with the delta-wing,respectively.T.Y.Chen,H.T.Shu /Experimental Thermal and Fluid Science 28(2004)273–282281。

新型无机闪烁体的能量分辨率(中英文对照)

新型无机闪烁体的能量分辨率(中英文对照)

成都理工大学学生毕业设计(论文)外文译文极,(b)光电子是后来ηNph,(c)这些∝ηNph电子在第一倍增极和到达(d)倍增极的k(k = 1,2…)放大后为δk 并且我们假设δ1=δ2=δ3=δk=δ的,并且δ/δ1≈1的。

我们可以得出:R2=Rlid2=5.56δ/[∝ηNph(δ-1)] ≈5.56/Nel (3)Nel表示第一次到达光电倍增管的数目。

在试验中,δ1≈10>δ2=δ3=δk,因此,在实际情况下,我们可以通过(3)看出R2的值比实际测得大。

请注意,对于一个半导体二极管(不倍增极结构)(3)也适用。

那么Nel就是是在二极管产生电子空穴对的数目。

在物质不均匀,光收集不完整,不相称和偏差的影响从光电子生产过程中的二项式分布及电子收集在第一倍增极不理想的情况下,例如由于阴极不均匀性和不完善的重点,我们有:R2=Rsci2+Rlid2≈5.56[(νN-1/Nel)+1/Nel] (4)νN光子的产生包括所有非理想情况下的收集和1/Nel的理想情况。

为了说明,我们在图上显示,如图1所示。

ΔE/E的作为伽玛射线能量E的函数,为碘化钠:铊闪烁耦合到光电倍增管图。

1。

对ΔE/E的示意图(全曲线)作为伽玛射线能量E功能的碘化钠:铊晶体耦合到光电倍增管。

虚线/虚线代表了主要贡献。

例如见[9,10]。

对于Rsci除了1/(Nel)1/2的组成部分,我们看到有两个组成部分,代表在0-4%的不均匀性,不完整的光收集水平线,等等,并与在0-400代表非相称keV的最大曲线。

表1给出了E=662Kev时的数值(137Cs)在传统的闪烁体资料可见。

从图一我们可以清楚的看到在低能量E<100Kev,如果Nel,也就是Nph增大的话,是可以提高能量分辨率的。

这是很难达到的,因为光额产量已经很高了(见表1)在能量E>300Kev时,Rsci主要由能量支配其能量分辨率,这是没办法减小Rsci 的。

然而,在下一节我们将会讲到,可以用闪烁体在高能量一样有高的分辨率。

物理实验报告英文版7

物理实验报告英文版7

iv
Table of Contents
Title Page Authorization Page Signature Page Acknowledgements Table of Contents List of Figures List of Tables Abstract Chapter1 Introduction 1.1 Structure of Carbon Nanotubes . . . . . . . . . . . . . . . . . . . . 1.2 Electronic properties of Carbon Nanotubes . . . . . . . . . . . . . . Chapter2 Superconductivity in 0.4nm Carbon Nanotubes array 2.1 The band structure of 0.4nm Carbon Nanotubes . . . . . . . . . . 2.2 Meissner effect in 0.4nm Carbon Nanotubes array . . . . . . . . . 2.3 The model of coupled one-dimensional superconducting wires . . . 2.4 Motivation and scope of the thesis . . . . . . . . . . . . . . . . . . i ii iii iv v vii xi xii 1 3 4 8 9 9 12 13
July 2008, Hong Kong
HKUST Library Reproduction is prohibited without the author’s prior written consent

Vortices in Schwinger-Boson Mean-Field Theory of Two-Dimensional Quantum Antiferromagnets

Vortices in Schwinger-Boson Mean-Field Theory of Two-Dimensional Quantum Antiferromagnets

a rXiv:c ond-ma t/95471v119Apr1995Vortices in Schwinger-Boson Mean-Field Theory of Two-Dimensional Quantum Antiferromagnets Tai-Kai Ng department of Physics,Hong Kong University of Science and Technology,Clear Water Bay Road,Kowloon,Hong Kong (February 6,2008)Abstract In this paper we study the properties of vortices in two dimensional quantum antiferromagnets with spin magmitude S on a square lattice within the frame-work of Schwinger-boson mean-field theory.Based on a continuum descrip-tion,we show that vortices are stable topological excitations in the disordered state of quantum antiferromagnets.Furthermore,we argue that vortices can be divided into two kinds:the first kind always carries zero angular momen-tum and are bosons,whereas the second kind carries angular momentum S under favourable conditions and are fermions if S is half-integer.A plausible consequence of our results relating to the RVB theories of High-T c supercon-ductors is pointed out.PACS Nos,75.10.Jm,75.30.Ds,79.60.-iTypeset using REVT E XI.INTRODUCTIONIn the past few years there has been a lot of interests in the study of quantum anti-ferromagnets in two dimension on a square lattice,stimulated by the discovery of High-T c superconductors[1,2].Among others,one approximated approach to the quantum antiferro-magnet problems is the Schwinger-Boson Mean-field Theory(SBMFT).[2].In this approach, the quantum spins are represented as Schwinger bosons and an approximated ground state is constructed by bose condensation of spin-singlet pairs.The mean-field theory can also be formulated in a large-N expansion as the saddle point solution to a generalized SU(N) quantum spin model[2,3].In the limit N→∞,the mean-field theory becomes exact. SBMFT predicts that in one dimension,Heisenberg antiferromagnet is always disordered, with a non-zero spin gap in the excitation spectrum.The theory is found to give an adequate description for integer spin chains,while it fails to describe the massless state of half-integer spin chains.This is because the topological Berry phase term which plays crucial role in the latter case is not taken into account correctly in SBMFT.[4]However,in two dimension topological term does not exist and SBMFT is more reliable.In particular,the theory pre-dicts that the disordered(spin gap)phase exists only when the spin magnitude S is small enough,when S<S c∼0.19.It turns out that SBMFT offers a fairly accurate description for the magnetic properties of the High-T c compounds in the low doping regime[5].However, at large doping levels,the generalization of SBMFT which includes holes as fermions[6], does not seem to describe the High-T c compounds correctly.For example,photo-emission experiments[7]indicate that the compounds have”large”Fermi-surfaces which satisfy Lut-tinger theorem,whereas the generalization of SBMFT to include holes predicts a”small”Fermi surface with area proportional to concentration of holes.In the large doping regime, it turns out that the slave-boson mean-field theory[8]which treats the spins as fermions and holes as bosons produces a lot of properties of High-T c compounds correctly[9]and the SBMFT is only superier in describing the low-doping,antiferromagnetic state.Thus a natural question is:what is the relation between the two theories?Can one understand thetwo theories within a single framework?How does nature crossover from one description tothe other as concentration of holes increases?In SBMFT,the spin operator S i on site i is expressed in terms of Schwinger bosons S=¯Z i σZ i,where σis the Pauli matrix,Z i(¯Z i)are two component spinors¯Z i=(¯Z i↑,¯Z i↓), ietc.Notice that in order to represent a spin with magnitude S,there should be2S bosonsper site[2].The Hamiltonian can then be represented in terms of Schwinger bosons,and amean-field theory can be formulated by introducing order parameters∆i,i+η=<Z i↑Z i+η↓−Z i↓Z i+η↑>(η=±ˆx,±ˆy)[2].Alternatively,SBMFT can also be formulated in a large-N expansion as the saddle point solution to a generalized SU(N)quantum spin model.To derive the large-N theory,we divide the square lattice into A and B sublattices and consider the following transformation for the Schwinger bosons on B-sublattice,Z B j↑(¯Z B j↑)→−Z B j↓(¯Z B j↓),Z B j↓(¯Z B j↓)→Z B j↑(¯Z B j↑),for all sites j on the B sublattice.The Schwinger bosons on the A sublattice remainsunchanged in the above transformation.The Lagrangian of the Heisenberg model can thenbe represented in the transformed boson coordinates as[3]L= iσ[¯Z A iσ(d dτ+iλj)Z B jσ−i2Sλj]+J i,η=±ˆx,±ˆy|∆i,i+η|2−J iσ,η=±ˆx,±ˆy[∆∗i,i+ηZ A iσZ B i+ησ+H.C.](1) whereλi’s are Lagrange multiplierfields enforcing the constraint that there are2S bosons per site,∆i,i+η’s are Hubbard-Stratonovichfields introduced in decoupling of H.σis the spin index.A SU(N)spin theory can be formulated with the above Lagrangian if the usual SU(2)spins↑and↓are extented to the SU(N)case where N-spin species are introduced. SBMFT can be considered as the saddle point solution of the path integral over L which becomes exact in the limit N→∞[3].For infinite system with no defects,the saddle point solution has position independent∆i,i+η’s andλ’s and the solution is formally verysimilar to the BCS solution for superconductors,except that in present case the pairing objects are Schwinger bosons but not electrons.Another important difference is that in the present case,the pairing bosons are always located on different sublattices,whereas no such distinction is found in the case of superconductors.The similarity between SBMFT and BCS theory leads us to ask the natural question of whether vortex-like solutions can be found in SBMFT as in the case of superconductors. More precisely,one may ask the question of whether one can construct a stable solution in SBMFT where∆i,i+ηhas a phase structure∆i,i+η∼|∆i,i+η|e iθ(i+η/2),whereθ( x)is a smooth function of x and has a singular point at x′∼0such that for distance| x|>>0,θ( x)→θ( x)+2πif one moves vector x around a close loop C which enclosed the point x′=0.As in the case of superconductors,the simplest way to address this question is to construct a continuum description for SBMFT(analogous to Ginsburg-Landau theory for superconductors)and study the possibility of having vortex solution in the continuum approximation.In the following we shall perform such a study in the disordered state in SBMFT where the Schwinger bosons are not bose-condensed and spin is a good quantum number.In realistic High-T c cuprates long range antiferromagnetic order is destroyed by introduction of charge carriers(holes).We shall not consider complications introduced by holes here and shall assume simply a disordered magnetic state with realistic spin magnitudes which can be described by the spin gap phase of SBMFT.We shall examine the properties of vortices within the continuum description.In section II we shall derive the continuum theory.Based on the continuum equations,we shallfirst study the properties of a single,non-magnetic impurity in the disordered state of SBMFT,where we shall point out the important difference between perturbations which are symmetric to the A and B sublattices,and perturbations which distinguish the two sublattices.In section III,we shall study static vortex excitations in our model.Based on the continuum theory,we shall argue that vortices are stable topological excitations in the disordered state of2D quantumantiferromagnets.We shall then show that there exists two kind of vortices in the continuum theory,corresponding to vortices centered on the mid-point of a plaquette,and vortices centered on a lattice point.We shall show that the properties of these two kinds of vortices are very different,because of the different symmetry with respect to the two sublattices.The first kind of vortex which is symmetric to the two sublattices can carry only zero angular momentum and is a boson,whereas the second kind of vortex distinguishes between the two sublattices can carry angular momentum S under favourble conditions and is a fermion if S is half-integer.In section IV we shall concentrate ourselves at the case S=1/2which is the physical case of interests and shall examine properties of a”liquid”of fermionic vortices. Based on simple symmetry arguments,we shall argue that the effective theory of a liquid of fermionic vortices has precisely the same form as the slave-boson mean-field theory for spins in the undoped limit.The case withfinite concentration of holes will also be discussed.The findings in this paper will be summarized in section V,where we shall discuss a plausible scenerio of how the system crossover from a state described by SBMFT to a state described by slave-boson meanfield theory upon doping.II.CONTINUUM DESCRIPTION FOR SBMFTTo derive the continuum theory we following Read and Sachdev[3]and consider the representation where our system has two sites per unit cell and introduce the’uniform’and ’staggered’components in the∆andλfields,where∆i,i±η=12)+q±η(i±η2)+A±η(i±η2[λA( k)+λB( k)],(2b) Aτ( k)=1Theφ( x)field describes uniform antiferromagnetic correlations whereas q( x)field de-scribes spin dimerization(spin-Peierls)effects[3].θ( x)and Aη( x)arefields describing the corresponding uniform and staggered phasefluctuations,respectively.Allfields are slowly varying on the scale of a unit cell.Notice that the reason why four independent(real)fields are needed to describefluctuations in one unit cell is precisely because we have divided the system into A and B sublattices.In the case of usual superconductors such a distinction is not present and only two realfields(or one complex scalarfield)enters into the Ginsburg-Landau equation.Notice that Vµ( x)=∂µθ( x)and Aµ( x)can be considered as U(1)gauge fields coupling to the Schwinger boson Z’s.The’uniform’gaugefield Vµ( x)couples to the bosons on the two sublattices with same gauge charge,whereas the gauge charges for the ’staggered’gaugefield Aµ( x)are opposite on the two sublattices.A single vortex solution centered at x′=0corresponds to a solution of SBMFT with boundary conditionVµ( x)dxµ=2nπ,or in terms of the gaugefield,there is a’uniform’gaugeflux of n/2flux quantum passing through the origin,similar to the case of superconductors.Notice that in SBMFT where<∆i,i+η>=0,the gauge symmetry associated with the’uniform’gaugefield Vµ( x) (Z A→Z A e iΓ,Z B→Z B e iΓ,Vµ→Vµ−2∂µΓ)is broken,whereas the’staggered’gauge symmetry(Z A→Z A e iΓ,Z B→Z B e−iΓ,Aµ→Aµ−∂µΓ)remains intact in SBMFT.The existence of vortex solution is tied with the broken symmetry gaugefield Vµ( x)as in usual superconductors.At distance much larger than lattice spacing,fluctuations associated with the’uniform’variablesφandθτare unimportant.Thus we shall neglect the∇φ(x)and∇θτ(x)terms in the following and derive a continuum theory for the rest of the variables.Notice that the Vµvariable is kept in our continuum theory since the term is essential for studying of vortices. In the continuum limit,the Lagrangian L becomesL→dτd 2x σ{¯Z A σ(x )∂∂τZ B σ(x )−µ=ˆx ,ˆy φ(x )(1−12φ(x )(D µ¯Z A σ(x ))(D ∗µ¯Z B σ(x ))+12 µ[φ(x )2+q µ(x )2]−4Sθτ(x ) ,(3)where D µ=∂µ+iA µand D ∗µ=∂µ−iA µ.The continuum Lagrangian Eq.(3)can befurther simplified by introducing new boson fieldsψσ=12(Z A σ−¯Z B σ).The πσfield can be integrated out safely at large distance and low energy [3],leaving effective Lagrangian for ψσfield,L =σ dτd 2x φ(x ) |D ∗µψσ|2+m (x )2|ψσ|2 +(φ1(x ))−1(D τ−q µD µ)ψ+σ(D ∗τ+q µD ∗µ)ψσ+18(V µ(x )V µ(x )))+m (x )2 (4)where φ( x )m ( x )2=2θτ( x )−4[1−(1/8)(V µ( x )V µ( x )]φ( x )and and φ1( x )=(1/2){θτ( x )+2[1−(1/8)(V µ( x )V µ( x )]φ( x )}.In the limit q µ=0and A µ=0,and neglecting gradient terms ∇φ,∇θτ,the above Lagrangian can be easily diagonalized resulting in an effective Lagrangian L effin terms of the φ( x )and V µ( x )fields.The dynamical effect of the remaining terms can be obtained by looking at Gaussian fluctuations of the fields around the saddle point solution A µ,q µ=0.[3]In particular,in the small m ( x )∼m →0limit,we obtain (see Appendix)L eff=dτ d2x (a−4(2S+1))φ+φ2+(1φ)qµqµ+(2S+1)φF2µν+icFµτqµ+O(m2) ,(5) 2e2where e2∼m,a(<4(2S+1)),b and c are constants of order O(1).The precise values of a,b,c depend on the underlying lattice structure and cannot be obtained in a continuum theory.We have consider the realistic case N=2in deriving the above expression.Notice that similar effective Lagrangian has been obtained by Read and Sachdev previously in studying effect of instantons[3].The only difference here is that the VµVµterm which was not considered by Read and Sachdev is now retained.Recall that Vµis the’uniform’U(1) gaugefield coupling to the bosons and the VµVµterm in Eq.(5)just represents the Meissner effect associated with nonzero order parameter∆in SBMFT.Notice that as in case of usual superconductors,we have choosen the London gauge∇. V=0in our derivation.The term can also be written in a gauge invariant way by replacing Vµby the gauge invariant object Vµ−2∂µΓin Eq.(5).This term does not contribute to the instanton effect discussed by Read and Sachdev[3]but is crucial to the study of vortices.To understand more bout the continuum theory wefirst consider the properties of a singlenon-magnetic impurity in the disordered state of SBMFT.[10]We shall assume that a non-magnetic impurity simply replaces a spin at site i by a non-magnetic object.The simplestway to model this in our formalism is to is replace the constraint equation¯ZZ=2S onsite i by¯ZZ=2S(1−n i)=0,where n i is the number of non-magnetic impurity on site i.It is than easy to see that the effect of nonzero n i in our effective Lagrangian Eq.(5)is tointroduce in the Coulomb gauge an extra term−(2Sn i e i)Aτ,where e i=±1,depending onwhether the impurity is located on the A-or B-sublattices,[10]i.e.,non-magnetic impuritiesappear as effective(staggered)gauge charges of magnitude2Se localized at the impuritysites.A corresponding electrostatic potential V(r)∼(2Se i)ln(r/ξ)(whereξ∼m−1)will beinduced around the impurity.To lower the electrostatic energy,2S bosons will be nucleatedout from the vacuum to screen the electricfield,resulting in formation of local magneticmoment of magnitude S around a non-magnetic impurity[10].Notice that this effect hasbeen observed experimentally in the High-T c compounds in the underdoped region upon substitution of Cu by Zn in the conducting planes[11].This result can also be understood in an alternative way by observing that the same physical effect should have occurred if instead of replacing a spin on site i by an non-magnetic impurity,we let the Heisenberg coupling J i to go to zero for those bonds joining to site i. In this case the’impurity’site i behaves as an non-magnetic impurity as far as the rest of the system is concerned.The only difference is that a free spin of magnitude S now remains on site i.However,the effect of this approach on our effective Lagrangian Eq.(5)looks rather different compared with our previous approach.Instead of introducing a’staggered’electric charge of magnitude2Se on site i,a boundary condition∆i,i+η=0(η=±ˆx,±ˆy)is now imposed on our ing Eq.(2a),the corresponding boundary condition in the continuum theory becomesqˆn(x i)=±ˆnφ(x i)where i→x i in the continuum limit.ˆn is a unit vector.±depends on whether site i is on A-or B-sublattices.The properties of the system around the non-magnetic impurity is obtained by minimizing the free energy Eq.(5)under this boundary condition.Performing the calculation,wefindfirst of all thatφ(x i)=0and correspondingly q(x i)= 0.The reason for this behaviour can be understood easily from Eq.(5)by noticing that the coefficient in front of the qµqµterm becomes negative asφbecomes smaller than2b, implying that spontaneously dimerization will occur whenφbecomes small enough.Because of stability requirement,it can be shown that dimmerization| q|cannot have value largerthan q o=(√2)φ(see Appendix).Anyway bothφ(xi)and q(x i)∼ q o should be nonzeroin the continuum solution.The reason why q should be nonzero can also be understood from symmetry consideration.Notice that theφ(x)field is symmetric under exchange of A-and B-sublattices whereas qµ(x)field is antisymmetric.Thus a solution with qµ(x)field identically zero would not distinguish between the two sublattices.However,the single non-magnetic impurity problem certainly distinguishes the two sublattices since the impuritycan only be placed on either one of the two sublattices.Thus we expect qµ(x)to be nonzero around a single impurity.Minimizing the free energy with respect to the qµfield wefind also that E(x)∼ q(x),where Eµ=iFµτis the’staggered’electricfield in the system.With the boundary condition q∼(±ˆn)φaround the impurity wefind that electricfield radiating out (or in)from the impurity is obtained in our solution,i.e.,depending on whether the impurity is located on the A-or B-sublattices,it behaves as source or sink for the electricfield,in exact agreement with our previous approach which predicts that non-magnetic impurity behaves as a single staggered gauge charge added to the system.One may also extend our approach to discuss the case when the Heisenberg coupling J′joining to the site i is small but nonzero,so that the spin at site i still couples weakly with the surrounding environment.The previous discussions still apply,except that the spin at site i appears now as a localized spinon state occupied by2S spinons in the system (notice that at two dimension,an arbitrary small attractive interaction is enough to generate a bound state[12]).The spinon state again behaves as a(staggered)gauge charge(with magnitude2S),and will bind2S other spinons with opposite gauge charge(or localized on opposite sublattice)next to itself as in the case with J′=0.However for nonzero J′the localized spinons on the two sublattices will interact antiferromagnetically because of the underlying antiferromagnetic correlation in the system,[13]forming at the end a localized spin singlet around site i.It is only in the limit J′→0that the two spinons are decoupled and free local magnetic moments appear.Our discussion can also be extended to study other forms of local defects.For example, one may study the situation when the Heisenberg coupling J i,i+ˆx between sites i and i+ˆx is set to zero,i.e.,a single bond is being removed from the system.Notice that the perturbation is symmetric with respect to interchange of A-and B-sublattices.It is straightforward to generalize our previous discussion to this case where wefind that the perturbation now appears as an effective electric dipole moment in L eff.No local moment is expected to form around the impurity bond in this case because of the follow reason:because of the symmetry between A-and B-sublattices bosons(spinons)can be nucleated out from vaccuum only inpairs,with one localized on the A sublattice and one on the B sublattice.The interaction between the nucleated spinon pairs will again be antiferromagnetic and has a magnitude ∼Je−x/ξ,where x is the distance between the two spinons andξ∼m−1is the’size’of the spinon wavefunction.For the single bond defect,the distance x between the two spinons is of order of lattice spacing<<ξ.Thus the effective interaction between the two spinons is of order J,i.e.,they will form a strong local spin singlet and no isolated magnetic moment will appear.Experimentally,the Cu ion in the conducting plane of high-T c cupartes can be substituted by Zn(non-magnetic impurity)[11]and magnitude of Heisenberg exchange J can be modified locally by introducing impurities out of conduction planes or substitution of O ions.It is observed that while substitution of Cu by Zn in the conducting plane introduces local magnetic moments in the underdoped(spin gap)phases of High-T c cuprates[11],no such effect is observed in other ways of introducing defects.Our theoretical investigation on disordered state of quantum antiferromagnets based on continuum description of SBMFT is in satisfying agreement with experimental results.[10]III.STATIC VORTICES IN DISORDERED STATE OF SBMFTIn the continuum theory,a unit vortex centered as x=0corresponds to a solution ofδL eff=0,withθ(r,Ω)=Ω,where r andΩare the distance and angle in the(2D) polar co-ordinate.The nonzero vorticity introduces a diverging kinetic energy through the VµVµ∼(∇θ)2term in L eff,which supresses the magnitude ofφaround the vortex core r→0,as in the case of usual superconductors.Minimizing the free energy with respect√2)φ,wefind thatφ=| q|=0for toφ,and keeping in mind that for smallφ,| q|∼(r<r o∼[(8(2S+1)+4b−2a)/(2S+1)]1/2,but not going to zero smoothly as r→0, as in the case of usual superconductors.This result is an artifact of our continuum theory where the(∇φ)2term is not included in L eff.Nevertheless,the qualitative effect whereφis being suppressed around the vortex core is clear.For large r,φ(r)→φo=(4(2S+1)−a)/2and q→0,indicating that vortices are stable topological excitations in SBMFT.The behaviour of vortices in SBMFT is very similar to vortices in usual superconductors at large r.However,the small r behaviours are very different.The suppression ofφand qfields around vortex core reflects the fact that around the center of vortex,the bond amplitude ∆i,i+µ’s are being suppressed.As has been discussed in the previous section,suppression of local bond amplitudes may lead to generation of localized magnetic moments,depending on the detailed bond configuration.To understand the properties of a vortex,it is thus important to understand the underlying bond structure at the vortex core.Before studying the bond structure at vortex core,let us explainfirst how local magnetic moment binding to vortex core modifies the properties of a vortex.For a unit vortex the magneticflux seen by a Schwinger boson is halfflux quantum.(recall that Vµis an’uniform’gaugefield which has same gauge charge for bosons on both sublattices)Therefore the orbital angular momentum of a Schwinger boson around a unit vortex is1the VµVµterm.Following similar analysis as in previous section wefind that the sup-pressed bonds introduce in the continuum theory an effective electric quadrapole structure surrounding the vortex center.Notice that as in the case of single broken bond the structure is symmetric with respect to the A-and B-sublattices.Thus we expect that localized spinon pairs may form around the center of vortex but isolated magnetic moments cannot occur. Thus this kind of vortex carries zero angular momentum and is a boson.The second kind of vortex is centered at a lattice point and as a result,it is expected that the four bonds joining to the vortex center(see Fig.1b)will be largely suppressed. This situation is very similar to the case of single non-magnetic impurity discussed in the last section.First of all,2S spinons will be found localized at the vortex center because of the suppressed bonds.The spinons behaves as a’staggered’gauge charge with magnitide 2S and2S other spinons with opposite gauge charge will be nucleated from the vacuum to screen the’staggered’electricfield generated by the spinons localized at the vortex cen-ter.The spinons on opposite sublattices interact antiferromagnetically through an effective Heisenberg exchange of order J′∼J|∆c|2,where∆c is the bond amplitude of the four bonds joining to the vortex center.As a result,a spin singlet will be formed.The vortex carries zero angular momentum and is again a boson.Notice that this is true even in the limit J′→0since integer number of bosons=4S are found binding to the vortex.The situation is however quite different if we consider afinite density of the second kind of vortices.We shall show that when density of vortices is large enough,it may become energetically unfavourable to nucleate spinons from vacuum to screen the staggered gauge field and as a result,only2S spinons will be found binding to vortex center resulting in fermionic vortices when S is half-integer(odd number of bound bosons).For simplicity we shallfirst consider two vortices seperated by distance l,with one vortex on each sublattice. First let us consider the energy when spinons are nucleated from the vacuum to screen the staggered gauge charge on each vortex.The total energy will be sum of three terms:(i) the electrostatic energy which is of order2(2Se)2ln(l o/ξ),where l o∼ξis the’size’of the nucleated spinon wavefunction,(ii)The exchange energy between the spinons localized asvortex center and nucleated spinons,which is of order−2S(S+1)J′e−l o/ξ∼−2S(S+1)J′, and(iii)The energy needed to nucleate the spinons from the vacuum,which is of order 4Sm.The sum of the three terms is of order2S(2m−(S+1)J′).The energy for the second case when no spinons are nucleated from vacuum consists of two terms:(i)the electrostatics energy which is of order(2Se)2ln(l/ξ)and(ii)the exchange energy between spinons located at the center of the two vortices which is or order−S(S+1)J′e−l/ξ.In2D,e2∼m and the electrostatic energy is of order4S2m.ln(l/ξ).For l>>ξ,it is certainly energtically more favourable to nucleate bosons from vacuum to screen the vortex gauge charge.However, when l≤ξ,the energy cost of the second case is of order−S(S+1)J′,and is energetically more favourable if S(S+1)J′<4Sm.Notice that J′∼J|∆c|2is expected to be very small because of suppresion effect around vortex core.Thus the condition can be satisfied easily even with a relatively small m.Forfinite density of vorticesδ,l∼δ1which are rather easy to acheive.We cannot,however,be absolutely sure whether’fermionic’vortices exist because of the natural limitation of a continuum theory which gives only or-der of magnitude estimates.Notice also one important distinction between the vortices we study in this paper and usual vortices in superfluids.In the present case,vortices cannot be distingished by sign of vorticity since vortices carrying1/2(orπ)and−1/2flux quanta should be considered as identical.On a lattice where the order parameterfield∆i,i+µis a link-variable,πand−πvortices can be related by a pure gauge transformation.The usual singularity encountered in the gauge transformation in continous space does not arise here because of the discretized lattice structure.It is important to clarify the different roles played by the’uniform’and’staggered’gaugefields in deciding the properties of vortices.The gaugefields arise fromfluctuation in phases of the order parameter∆i,i+µ’s.The uniform gaugefield couples to bosons on the two sublattices with same gauge charge,and the corresponding gauge symmetry is broken in SBMFT.Vortices are stable because of this broken gauge symmetry,and corresponds to solutions with”uniform”magneticflux penetrating center of vortices.The’staggered’gaugefield couples to bosons on opposite sublattices with opposite gauge charge.It plays no role in generating a stable vortex solution,but has strong effects in determining the precise properties of the vortex core.Without the’staggered’gaugefield,spinons need not be nucleated from vacuum to screen the’staggered’charges generated at,for example,the center of the second kind of vortex.In this case,for half-integer spin systems,the second kind of vortices will always be fermions independent of density,since only odd number of bosons are found binding to center of each vortex(corresponding to the spin S at center of vortex).Similarly,local magnetic moments will not be generated by non-magnetic impurities if’staggered’gaugefield does not exist.It is also interesting to point out that similar vortex excitations have been considered by Read and Chakraborty[15]in a short-ranged RVB wavefunction for S=1/2quantum antiferromagnets.They considered also the two kinds of vortices we discussed here.The statistics of the vortices were examined by direct Berry phase computations where similar。

  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

equations with a corresponding angle φ = 0. Under certain assumptions for more general flows it is shown that there is an attracting fixed point of the (α, χ) equations which corresponds to positive vortex stretching and for which the cosine of the corresponding angle is close to unity. This indicates that near alignment is an attracting state of the system and is consistent with the formation of Burgers-like structures. email addresses: galanti@dvir.weizmann.ac.il; j.gibbon@ PACS: 47.27.Jv, , 47.15.Ki AMS: 76F05, 76C05
χ α
Weizmann Institute of Science, 76100 Rehovot, Israel. Imperial College of Science, Technoloφ(x, t) = arctan
, which lies between ω and S ω. For the Euler equations
† Department of Chemical Physics, ∗ Department of Mathematics, London SW7 2BZ, UK.
Abstract We address the problem in Navier-Stokes isotropic turbulence of why the vorticity accumulates on thin sets such as quasi-one-dimensional tubes and quasi-two-dimensional sheets. Taking our motivation from the work of Ashurst, Kerstein, Kerr and Gibson, who observed that the vorticity vector ω aligns with the intermediate eigenvector of the strain matrix S , we study this problem in the context of both the three-dimensional Euler and Navier-Stokes ˆ · Sξ ˆ and χ = ξ ˆ × Sξ ˆ where ξ ˆ = ω /ω . This introduces the equations using the variables α = ξ
This observation was based on a study of the probablility density function (PDF) of the cosine of the angle between their directions. This important alignment process has also been confirmed in other numerical simulations [15, 16, 18, 20, 26, 28] as well as in turbulent grid flow experiments conducted by Tsinober et al [29, 30, 31, 32] and it has been suggested that it is a kinematic process [33, 34]. In fact preferential vorticity alignment appears to be a universal feature of smaller scale structures in such flows even though larger scale features may vary from flow to flow. A common feature seems to be that vorticity is concentrated in tubes of width intermediate between the Taylor and Kolmogorov microscales [14, 16, 18, 20, 28] with viscous dissipation 1
1
Introduction
As early as 1938 G. I. Taylor [1] showed that in isotropic Navier-Stokes turbulence vortex stretching has a major effect on vorticity production and dissipation. One of the many interesting features of high Reynolds number turbulent flows, illustrated beautifully by modern flow visualization methods, is the fact that vorticity is not evenly distributed throughout a flow but has a tendency to accumulate on ‘thin sets’. The morphology of these sets is characterized by the predominance of quasi one-dimensional tubes or filaments and quasi two-dimensional sheets although, being fractal in nature [2, 3, 4], many individual vortical structures may be neither precisely one nor the other. This morphology makes it clear that the direction of vorticity as well as its production plays a significant role in the self-organization processes through which apparently intense regions of vorticity appear to metamorphose into approximate thin coherent structures. The challenge facing the theorist working with the three-dimensional Navier-Stokes equations is to explain these geometric accumulation processes and their consequences even though a detailed description of their precise topology is beyond our reach at present. The subtle fine spatial structure of these sets indicates that an alternative approach is needed to the conventional one of Navier-Stokes analysis where L2 -spatial averages are taken over the whole domain when estimating the effect of the vortex stretching term. In 1951, Townsend [5, 6] indicated that something like the thin sets described above might be relevant in turbulent flows and in the last two decades their occurrence has been a regular theme of the literature [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]. They would also seem to be the visual manifestation of the vorticity alignment phenomenon first reported by Ashurst et al [14] who observed that in driven simulated isotropic Navier-Stokes turbulence with a Taylor microscale Reynolds number of 83 the vorticity vector ω has a tendency to align with the intermediate eigenvector of the rate of strain matrix Sij = 1 (ui,j + uj,i ) . 2 (1)
Vorticity alignment results for the three-dimensional Euler and Navier-Stokes equations
相关文档
最新文档