D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

Embed Size (px)

Citation preview

  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    1/6

    This content has been downloaded from IOPscience. Please scroll down to see the full text.

    Download details:

    IP Address: 157.253.50.10

    This content was downloaded on 02/02/2014 at 21:34

    Please note that terms and conditions apply.

    D'Alembert's principle of zero virtual power in classical mechanics revisited

    View the table of contents for this issue, or go to thejournal homepagefor more

    1993 Eur. J. Phys. 14 217

    (http://iopscience.iop.org/0143-0807/14/5/005)

    Home Search Collections Journals About Contact us My IOPscience

    http://localhost/var/www/apps/conversion/tmp/scratch_4/iopscience.iop.org/page/termshttp://iopscience.iop.org/0143-0807/14/5http://iopscience.iop.org/0143-0807http://iopscience.iop.org/http://iopscience.iop.org/searchhttp://iopscience.iop.org/collectionshttp://iopscience.iop.org/journalshttp://iopscience.iop.org/page/aboutioppublishinghttp://iopscience.iop.org/contacthttp://iopscience.iop.org/myiopsciencehttp://iopscience.iop.org/myiopsciencehttp://iopscience.iop.org/contacthttp://iopscience.iop.org/page/aboutioppublishinghttp://iopscience.iop.org/journalshttp://iopscience.iop.org/collectionshttp://iopscience.iop.org/searchhttp://iopscience.iop.org/http://iopscience.iop.org/0143-0807http://iopscience.iop.org/0143-0807/14/5http://localhost/var/www/apps/conversion/tmp/scratch_4/iopscience.iop.org/page/terms
  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    2/6

    Eur.

    J. P h p

    14

    (19931

    211-221

    Printed /n the UK

    217

    DAlemberts principle

    of

    zero

    virtual power in classical

    mechanics revisited

    F

    Verheestt and

    N

    Van den Bergh

    tsterren kun dig Observatorium, Vakgroep Wiskundige Natu urk und e en Sterrenkunde, Universiteit Gent,

    Krijgslaan 281, B-9000 Gent, Belgium

    A+Dienst van de Studie-inspecteur, Fakulteit Toegepaste Wetenschappen, Universiteit Gent, Jozef

    Plateaustraat

    22,

    B-9000 Gen t, Belgium

    Received 24 May 1993

    A b m a c r D l e m b e n s pnnciplc of anal)ucal mechmcs,

    that the

    vinual power of

    he cansuaint forces

    vanishes, c3n

    be uwd to

    determine

    rclction, expleitl}

    and umquel}. The

    u l l equrralencc b e i w c n the Sewtoman, dAlcmbeman

    and

    Lsgrangiao

    dcscnption;

    is also

    pro\&

    1. Introduction

    Most of the textbooks of classical mechanics, after

    having exposed the Newtonian framework, use the

    method of virtual work or power to introduce a

    Lagrangian description of classical mechanics

    with

    constraints. In what follows we will give the logical

    sequence, as we see it, and indicate some weaknesses

    in the existing treatments.

    Since the basic information

    for

    a given mechanical

    problem

    is

    the knowledge abo ut the given forces and

    the prescribed constraints, a Newtonian description

    requires

    the introduction

    of

    reaction

    or

    constraint

    forces,

    so

    that the system obeys the constraints. That

    constraints must give rise to forces is obvious,

    because from a Newtonian point of view forces are

    the only agents which can change the motion

    of

    the

    system. However, the reaction forces are not given

    in the same priori way as the external forces are,

    but rather have to be treated as unknowns of the

    problem. From this working definition the unique-

    ness of the reactions is

    not

    evident.

    Next comes the key notion of

    virtual

    velocities (or

    displacements, to use the better known way of pre-

    senting what ar e in reality aspects of calculus

    of

    vari-

    ations). The observation, that in problems involving

    equality constraints the total m echanical power asso-

    ciated with the reactions usually vanishes for virtual

    ResumS.

    Le principe dAlembert en mkniquc classique,

    a m m e

    quoi la

    puissance virtuelle

    des forces de

    contrainte

    est nulle, pennet de determiner les &actions

    de

    fqon

    explicite et unique. En outre, I uivalence totale

    entre les

    descriptions Newtonienne, dAlembertieone et

    Lagrangienne est aussi prouvee.

    velocities @ut no t fo r real motions in general), is then

    erected into a principle. This principle sometimes

    car

    ries the name ofdAlembert (Rosenberg 1977, p 122),

    although in many cases something different is meant

    by dAlemberts principle. It is then claimed that the

    said principle of zero virtual power determines the

    reactions uniquely, with the help of Lag rangian m ul-

    tipliers (pars 1965, p 24). although this assertion is

    seldom fully proved, in the sense that ways ar e indi-

    cated to determine the multipliers explicitly as func-

    tions of positions, velocities and possibly time.

    Furthermore,

    th is

    principle

    of

    zero virtual power

    of

    the reactions is mostly given, together with the

    associated necessary fundamental equation, in the

    case of statics. The transition to dynamics is then

    done, more often than not, by the principle

    of

    educ-

    ing dynamics to statics (Lanczos 1949, pp 88-92,

    Goldstein 1980, pp 16-18). The latter principle is

    also very often called dAlemberts principle. This

    notion that one could somehow equate dynamics to

    statics,

    however,

    is

    confusing because

    it

    overlooks the

    inherent difficulties in determining the

    reactions

    together with the motion of the system. In statics the

    equilibrium positions are determined and the reaG

    tions then merely balance the given forces, which is

    an altogether different problem.

    So

    this part of the

    usual textbook treatment is not satisfactory.

  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    3/6

    218

    F Verheest

    and

    N

    Van den

    Bergh

    I n addition, the fundamental equation (see (22)

    below) following from the principle of zero virtual

    reaction power (Pars 1965, p

    28,

    Rosenberg 1977,

    p 149)

    is

    then used by many to introduce the Lagran-

    gian description of motion with constraints, rather

    than invoking a more abstract variational principle

    like Hamilton's.

    So,

    before going

    on

    to

    the Lagran-

    gian description, one should prove that this funda-

    mental equation is also sufficient to determine the

    motion of the system, making afterwards the New to-

    nian and Lagrangian descriptions totally equivalent.

    This way back is also lacking in almost any treat-

    ment involving virtual power.

    In view of the sometimes critical remarks made in

    the preceding paragraphs, we propose to revisit the

    whole area around d'Alembert's principle, trying to

    the only agents which can change the motion of the

    particles (Rosenberg 1977, p 122).

    The equations for the constraints themselves are

    given in non-Pfaffian fom as

    (3)

    where A t and

    E

    are known functions of position

    and time:

    A t = A ; ( r l , , ,rN,f)

    E m= B o ( r I , . , r ~ v ,) .

    4)

    Some of the constraints could be holonomic,of the

    form

    fa r l , .

    ..

    r N , ) = o

    5)

    and their differentiated expressions are then under-

    stood included in (3). Furthermore, the set (3) is SUP-

    posed independent, i.e. reduced to the least possible

    num ber (Pars 1965, pp 22-7). More precisely, the

    rank of the matrix which one can construct from

    (0= I , . . . ,L, 5

    ive a systematically structured and complete

    description of the transition from Newtonian to

    ~ ~ ~ ~ ~ ~ g i ~hanicr this route, including the

    proofs for the uniqueness of the under

    $Memb&'s p,.inhple an d the ull of

    the different frameworks.

    2. React ions and d'Alem bert 's pr in c iple

    In

    what follows we will only deal with equality con-

    straints for systems of point masses, in order to keep

    the notations and OUT l i e of approach sufficiently

    transparent. When dealing with mechanical prob-

    lems involving constrained motions, we are usually

    given information about the constraints, as elabo-

    rated further

    on,

    and about the external forces

    Fh = F k ( r l , , . r N , l , . ., N,

    )

    k = 1,..., )

    ( 1 )

    acting

    on

    point particles witb masses mk nd position

    vectors rk with respect to an inertial reference system.

    In

    what follows,

    k

    will always be assumed to vary

    from 1 to

    N.

    As a typical example to which we will

    return later, the classical spherical pendulum is a

    point mass moving under the influence of its weight

    on the surface of

    a

    smooth, fixed sphere.

    In order to fit the description of problems with

    constraints into a Newtonian framework, originally

    devised for one unconstrained particle, reaction

    forces

    Rk

    are introduced, such that Newton's second

    law reads as

    mhrh Fh Rk.

    (2)

    The reactions

    Rh

    are not given in the same

    a

    priori

    way as the external forces

    Fk

    are, however, but

    rather have to be treated as unknowns of the pro-

    blem. So in writing

    Z),

    one really defines the reac-

    tions as those force terms which have to be added

    to the equations of motion so as to make sure that

    the components of

    A ;

    is

    L.

    Once the constraints are given, it becomes possible

    to introduce at each point compatible with the holo-

    nomic constraints

    5 )

    the set of virtual velocities w h,

    which by definition obey

    N

    A t ,

    wk

    =

    0 . (6)

    Possible velocities on the other hand, being compat-

    ible witb the constraints as they evolve in time, obey

    3)

    as already written down, Admissible trajectories

    r h ( t )

    satisfy 3 ) and

    5).

    The next step is to note that for all standard prob-

    lems,

    dealt with in the textbooks and involving equal-

    ity constraints, the reactions globally do no work

    under virtual displacements of the system. We will

    call this d'Alembert's principle of virtual power,

    and it states that at a given time t along trajectories

    r h ( f ) he total m echanical power associated with the

    reactions vanishes for any set

    off

    virtual velocities

    ( 'h).

    k = l

    (7)

    Since the principle expressed in

    7)

    annot constitute

    an additional constraint in view of the independence

    of

    (3),

    the

    Rk

    have to be linear combinations of the

    A t via Lagrangian multipliers A, giving

    So

    we see that the Newtonian equations of motion

    can be rewritten as

    the particles are obeying the constraints at all rel-

    evant times. And the constraints must give rise to mhFh= F~ A , A ~ . (9)

    forces, since in a Newtonian approach forces are

    L

    0 = 1

  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    4/6

    O AIembert s

    principle

    of zero

    vir tual

    power in ciassical mechanics

    revisited 219

    The com plete motion, i.e. the

    N

    unknowns rk(f)and

    the newly introduced Lagrangian multipliers A e f ) , is

    now to be determined from the N vector equations

    (9) together with the L scalar equations 3,tarting

    from initial conditions are adm itted by 3) and 5).

    So far we have not stated any thing new o r exciting,

    but have trodden common ground in recalling the

    different notions involved (see e.g. Rosenberg 1977,

    pp 119-48). We will now show what is usually

    omitted, namely that it is possible to reduce the

    above description further, by proving that the differ-

    ent A,, and hence the reactions, can be determined

    uniquely indeed as functions of positions, velocities

    and time. T o

    that

    effect we arrange (9) along a given

    trajectory as

    which after scalar multiplication with Af and

    s u m -

    mation over

    k

    gives

    (@= , ... L) . (11)

    If we take the total time derivative of the constraints

    (3), we find equations of the form

    where the scalar expressions CQ re functions

    C ( f )

    C ' ( r ~ ( r )..., N ( t ) , L I ( r )...

    N t ) ; f )

    (13)

    which we can explicitly compute. With the help of

    (12) one can replace the first sum on the RHS of

    (11). As the second s u m on the RHS of I

    1)

    similarly

    is a function which follows from the given forces

    and the constraints, (11) ultimately turns out to be

    of the form

    L

    M @& = 09

    o . = l

    where

    are the elements of a symmetric square matrix

    M.

    The expressions

    D (f) =D (r l ( f ) ...,

    N(f ) , r l ( f )

    . . . ,

    N f ) , f ) 16)

    can be computed explicitly, as indicated above. N ow

    the matrix M is positive definite, since

    and the vanishing

    of

    the LHS

    would

    imply that

    which in view of the independence of the set of con-

    straint equations 3) is only possible for all U =

    0.

    Hence M is non-singular, so that (14) can be

    inverted and the different A, uniquely determined,

    although their expressions may be cumbersome

    to

    write down explicitly.

    However, the reasoning in going from

    ( 6 )

    to (18)

    was done along trajectories, so that really we should

    write

    A )

    =A&I ( .. . , r N ( f h ( 4 , .

    . r N ( 4 , 4

    (19)

    for the fidlexpressions of the

    A,

    Ifw e now define new

    functions

    A.

    through

    = A a ( r l , .. . , N , l . . ,rN,

    (20)

    and write instead of (9) that

    L

    mkFk= F~ i ~;

    or

    we obtain N coupled ordinary differential equations

    for the determination of the motion rk(t) of the

    dynamical system. One can prove that the solutions

    rk(f)of (21) are the same as of the set (3) and (9),

    for given initial positions and velocities which are

    admissible by the constraints. The preceding way of

    presenting the equations

    of

    motion does not imply

    that

    i t

    would be easier to work with (21) than with

    (3) and (9) together, only that i t is possible concep-

    tually to write the equations of motion as differential

    equations for the determination of the trajectories

    with given RHS.

    Having dealt with the reactions, we deduce from

    (2) and (7) or from (6) and (9) or from (21) a neces-

    sary condition, sometimes called the fundamental

    equation pars 1965, p 28, Rosenberg 1977, p 149),

    valid for all admissible virtual velocities in points of

    the trajectories considered. We have arrived at the

    fundamental equation (22) without invoking the

    principle whereby one is supposed to reduce

    dynamics to statics, by calling the acceleration terms

    the inertia forces (Lanczos 1949, pp 88-93). Th is

    principle is often quoted as d'Alembert's principle,

    although the notion that o ne could somehow equate

    dynamics to statics is insulting to the genius of

    d'illembert (Ham el 1949, p 220), since it totally

    obfuscates the inherent difficulties in determining

    the reactions together with the trajectories. In statics

    the positions are known and the reactions then

    merely balance the external or given forces. Noth ing

    fancy there.

    The fundamental equation (22) is

    used

    by many,

    starting from a Newtonian point of view, to intro-

    duce the Lagrangian description of motion with con-

    straints, instead of invoking a more abstract

    variational principle.

    So

    one also has to prove that

    (22) is sufficient to determine the motion and hence

    (18)

  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    5/6

    220 F

    Verheest

    and N Van

    den

    Bergh

    the reactions, and it is precisely this way back which

    is lacking in almost any treatment involving virtual

    power or work, or deriving the Lagrangian from

    the Newtonian formalism.

    If

    22)

    is

    not only neces-

    sary but also sufficient to determine the motion,

    then ultimately the Newtonian and Lagrangian

    descriptions

    will

    be totally equivalent. This

    is

    briefly

    addressed in the next section.

    3. Sufficient conditions and equivalence

    Here we start from a curve

    rL(t) ,

    obeying the con-

    straints and such that

    is valid for

    all

    virtual velocities (wh}. By

    F;

    and sub-

    sequent an alogou s expressions with

    a

    prime we mean

    tha t the relevant quantity

    is

    expressed along the curve

    rL(t):

    F; r) = F k ( r i ( f )

    . . . ,

    k( t ) , ? ' ( ( i ) . . . , h i),i). 24)

    If we simply

    call

    then U) s in shorthand

    S k = mti:

    -

    FL

    U)

    s,. Wk = 0

    h = l

    for all virtual velocities {wk} obeying

    N

    CAP.

    h

    =

    0. (27)

    k = 1

    If at

    this

    stage one is happy to call Sk he reactions,

    then the curve rL(t) corresponds to a motion of the

    set of oarticles and the circle

    is

    triviallv closed

    but also

    A ? ( t i ) = ( t i )

    B (rl)

    =

    P W (30)

    (t i)

    = f i h ) .

    Since in the direct part we proved that the Lagran-

    gian multipliers only depend on positions and veloc-

    ities, there immediately follows that

    &PI)

    =

    &(id

    = P 11

    (31)

    Rk(fi)=Sk fd.

    (32)

    leading

    to

    the equality of

    We can repeat the argument at other times and see

    that

    a

    curve such that (23) holds for

    all

    admissible

    virtual velocities indeed corresponds to

    a

    motion of

    the dynamical system with given admissible initial

    positions and velocities.

    Once we have the

    full

    equivalence of the motions,

    determined either from the reformulated Newtonian

    equations of motion (21) or from the fundamental

    equation

    22),

    the transformation

    of

    the latter into

    Lagrange's equations of motion is standard and can

    be found in the literature (see e.g. Pars 1965, pp

    73-6, Rosenberg 1977, pp 201-6, Goldstein 1980.

    pp 18-21, Gr iff iths 1985, pp 261-4).

    4.

    Spherical pendulum

    As an illustration of the above remarks we will very

    briefly recall the problem of the spherical pendu-

    lum, treated

    as a

    point mass moving under the influ-

    ence of gravity

    on

    the surface of

    a smooth

    fixed

    sphere

    wt

    centre

    in

    and radius

    e

    Since the given

    force is the constant weight of the particle, Newton's

    second law reads as

    m i =

    mg

    R. (33)

    r . r = P

    (34)

    (Amold'1980, pp 91-2). However, we havd detiued

    the reactions Rh

    as

    those forces which make the mov-

    ing

    system obey the constraints, and so there might

    be subtle differences between the notions Rh along

    a trajectory and Sk long a curve

    so

    far merely deter-

    mined from (23).

    The constraint is holonomic and SC~erOnOmic,

    and in differentiated form

    r . i . = O .

    (35)

    n

    the latter &se, we repeat the argument exposed

    in the previous section with Lagrangian multipliers

    p m

    and referring to the curve I;(?) to get

    Along a trajectory r t ) the virtual velocities

    w

    have

    to

    obey

    L

    Sk = p e A P .

    (28) r . w = O (36)

    e.=I

    indicating that they lie in the tangent plane to the

    sphere, and d'Alembert's principle of virtual power

    states that

    Now we sta rt from ac urve with given initial positions

    r l t J

    and velocities

    ? ; t o ) ,

    admissible in the sense of

    R.w=O, (37)

    g

    compatible wit h &e constraints and obeying

    (23)

    at

    all

    times

    i

    to.

    Then,

    at

    an arbitrary time

    tl 4 ,

    we use r;(t,) and

    ;;(ti) as

    initid conditions

    for traiectories

    r J d .

    solutions

    of (21).

    and see that

    There being only on e vector constraint, we immedi-

    ately find that the reaction

  • 7/25/2019 D'Alembert Principle of Zero Virtual Power in Classical Mechanics Revisited

    6/6

    DAlemberts principle

    of

    zero virtual power

    in

    CIasSicaI mechanics revisited

    22

    equation of motion (33) is rewritten as

    After scalar multiplication with r we can solve for X

    and get

    mi: =

    mg+

    Xr. (39)

    (40)

    r . r + i * = o (41)

    (42)

    m

    r

    X

    =?(r.r-

    r . g ) .

    Taking the total time derivative of

    (35)

    gives

    so

    that 40) ecomes

    A

    = --(

    2

    i * + r . g ) .

    Inserting this in (39) yields the differential equation

    for the determination

    of

    the trajectory

    r(f) .

    One can

    prove that the solutions of (43) are the same as of

    the original set (35) and (39), for given admissible

    initial positions and velocities. Moreover, the solu-

    tions obey the constraint, as can be

    seen

    by tracing

    the steps back from

    (43) to (35).

    5.

    Conclusions

    We have given and discussed in the preceding sec-

    tions a logical sequence which leads from the Newto-

    nian to the Lagrangian description of mechanical

    problems with equality constraints, with the help of

    the principle of dAlembert that the reaction forces

    have no virtual power.

    This

    principle allows an

    explicit and unique determination of the reactions,

    with the full equivalence of the descriptions as a con-

    sequence. We have indicated some of the weaknesses

    in the existing treatments and try

    to

    remedy these.

    In

    all fairness it should be pointed out that the matter

    under consideration here has recently been treated

    using the methods of and couched in the language

    of modem differential-geometric approaches to

    mechanics (Cardin and Zinzotto

    1989,

    Giachetta

    1992).

    Acknowledgments

    It is

    a

    pleasure to acknowledge the essential part

    played by Professor

    A

    Vanderbauwhede almost a

    decade ago in looking critically

    at

    the subject. We

    would also like

    to

    thank Professors W Sarlet and

    F

    Cantrijn for their many incisive remarks which

    undoubtedly improved our presentation beyond

    recognition.

    References

    Arnold

    V I 1980

    Mathemtical Methodr

    of Classical

    Cardin

    F and

    Zanzotto G

    1989 J .

    Math. Phys.

    30 1473-9

    Giachetta G

    1992

    J. Math. Phys. 33

    1652-65

    Goldstein H

    1980

    Ch si co l Mechanics

    2nd

    edn (Reading,

    Griffiths I B

    1985

    The Theory of/Clmsicol Dynamics

    Hamel 949 Theorerisehe Mechanik Berlin: Springer)

    Lanczos C

    1949

    The Variational Principles

    of

    Mechanics

    Pars L A 1965 A Treatise on Anolyticnl Dynomics

    Rosenberg R 1977

    Analylicol Dynamics

    of

    Discrete

    Mechanics 2nd edn (Berlin: Springer)

    M A Addison-Wesley)

    (Cambridge: Cambridge

    University

    Press)

    (Toronto: University of Toronto

    Press)

    (London: Heinemann)

    Systems ( N e w York: Plenum)