Quantum Mechanics and Experience David Z Albert Harvard Universily Press Cambri u =, Massachusents mn, EnglandContents Preface 1. Superposition 2, The Mathematical Formalism and the Standard ‘Way of Thinking about I 3. Nonlocaiy 4. The Measurement Problem 5. The Collapse of the Wave Funston 6. The Dynamics by tself 7. Bots Theory 8, Se Measurement Appendix: The Kochen-Healy-Dieks Interpretations Bibliography Index 6 B 80 134 180 19t 199 203Preface ‘This book was written hoth a¢ an elementary text and as an ateempt fo add to what we presently understand, at the me Advanced level, about what seems to™me to be the central difficulty at the foundations of quantum mechanics, which is che diffculry bout measurement ‘The ist four chapeers are a more or les stesightforward intco- Hardness - Figwe 12 color box, then that electron will with certainty emerge from that second color box through the b aperture as well. The same goes for white electrons, and the same goes (with hardness boxes) for hard and soft electrons too. All that can be (and has been) con- firmed by means of texts with those boxes [Nov suppose that it occurs tou to be curious about the possi= bility chat the color and hardness properties of electrons might somehow be related to one another. One way to look for such a relation might be to check for correlations berween the values of the hardness and color properties of electrons. Ie easy to check for correlations like that with our boxes; and it turns out (once the checking is done) thar no such coreelations exist, OF any large collection of, say, white electrons, all of which are fed into the left aperture of a hardness box, precisely half emerge through the hard aperture, and precisely half emerge through the soft one. The same joes for black electrons fed into the lee aperture ofa hardness bony and the same for hard or soft ones fed into the left apertures of color boxes. The color (hardness) ofan electron apparently entails nothing whatever about its hardness (colo). Suppose we set up a sequence of three boxes. First a color box, say, thena hardness box, and then another calor box. Consider snSUFCRPOSITION clectron which emerges through the white aperture ofthe fst color box, and thereafter (without having heen tampered with) is fed ingo the left aperture ofthe hardness box, and which happens to ‘emerge from that box through the soft aperture (as half of such clectrons wll, and thereafter (once again with no tampering) is fed int the left aperture ofthe second coloe box. That electron, a8 i enters that third box, is presumably known to be both white and soft. As there has boen no tampering between boxes here, we should expect that the electron will emerge from this third box ‘through the white aperture, confirming the resalt of the ist mea surement, As a matter of fact that isn’t what happens. Precisely hall of sch electrons emerge from the white aperture ofthat third box, and the other half (the other hal, that i, of those electrons ‘which have been measured to be white and soft by the previous ‘wo hoxes) emerge from the black aperture. The same goes for any ‘other pair of results inthe frst two Boxes, andthe same goes ifthe color boxes inthe above example are replaced with hardness hoxes and the hardaess box with 3 color one. Apparently (in che example We considered) the presence ofthe hardness box between the two color boxes self constitutes some sort of color tampering, Indeed, that hardness box must be what's to blame for changing half of ‘those white electrons to black ones, since te aeeady know that ‘wo color measurements, without tampering berween the boxes and without aa intervening hardness measurement, will invariably produce identical resulst Pechaps the hardness box is poorly built, crudely built. Ke seems +0 do its job of measuring hardness (without disturbing the hard nese in that process) well but in the course of doing that job it apparently docs disrupt cofor. That raises two questions. Fist whether hardness boxes can be built less crudely; whether the job fof measuring hardness can be accomplished more delicately Whether it can be accomplished without disrupting colo. Second, inthe ease ofthis “crude” apparatus, this apparatus which changes the colors of fully half of the electrons whose hardnesses it mea 5. Excl what consis “uanpedng” atl what des is (couse some hing oe er, ir, experience sures: what i it that determines precisely rohich electrons have their colors changed and which don't? Let's talk about the second question first. The right way 0 discover precisely what i is that determines which electrons change «olor in passing through that intermediate hardiness box and which don't would seem to be to monicor very caefuly all of the mea surable properties of al ofthe electrons which at fed into that fst color box in the course of some particule experiment and which are at that point Found eo be, say, white; and to make very certain thatthe physical states of those thece boxes ae held perfectly and constantly fxed throughout that experiment; and to look for cor relations berween the measurable physical properties of those i coming electrons and ther final postions 35 they emerge from the second color box. Well, it turns out that, in s0 far as we are presently able to tell, absolutely no such correlations exist. As a matter of fact, when we take whatever pains we know how to take 1 insure tha all ofthe electrons in some particulat experiment are fed into that first color box with precisely identical ses of physical properties, and to insuee thatthe physical states of those boxes are indeed held precisely and constantly fixed throughout that experi ment, the statistics of final outcomes remain peossely as they were described above. Insofar as we are now able ro determine, then, this second question has no answer. That iin 0 far a we are now able to determine, those electrons whose color is changed by pa ‘sige through a hardness box and those electrons whose color isnt changed by passage through a hardness box need not intially difer from one another in any way whatever. Le’s ty the first question. Can hardness boxes be buil lst ‘rudely? Wel, hardness boxes can be builtin a number of entcely different ways. We can try each one. Ie turns out that they all produce the statistics described above, All of those boxes change the color of (statsially) precisely half ofthe electrons which pass ‘through them. We ean ty to be much more carefal and much more precise in constructing our hardness boxes, bu it turns out that that doesn't change anything ether. What's striking here isn't that we af unable o build hardness boxes which don’ disturb the color of electrons at all but rather that we are unable to move the6 statistics of color disruption even so much as one millionth of one percentage point away from fifty-iyin ether direction, no matter ‘what we ty. So long as the device at hand fulfills che definitional requirements of a hardness box (that i: so long as ta device with ‘which the hardnesss of electrons can be determined, repeatably, then the color randomization produced by that device has always, inour experienc, been coals and all ofthe same goes for the effects fof color boxes on hardness. Suppose we wanted to build a colorsnd-hardness box that is, device with which both the color and the hardness of elects ‘could be determined, That box would need fve apertures (sce Figure 1.3i; one (on the left) where the electrons are taken in, one where white and hard electrons emerge, one where white and sofe ones ‘merge, one where black and hard ones emerge, and one where black and sofe ones emerge Consider how we could build a box like that. A box like thar ‘would seem to need to consist ofa hardness box and a color box. ‘But if the incoming electrons are made to pass frst through, sa, the hardness box, then theic hardnesses might subsequently be —> and Hardness Fw 1 SUPERPOSITION fonds changed when they passthrough the color box, and we would end up with reliable information only about the colors of che emergent clectrons. If we por the color box first, we would end up with ‘eliable information only abou che hardaesses. Nobody's been able to think of any other ways to build a color-and-hardness box, and it’s hard to imagine how, in principle, there could be other ways (other, that is, than building dhem out of color boxes and hardness boxes. So the task of putting ourselves in a position to say “the color ofthis electron is now such-and:-such and the hardness of his clectron is now sueh-and-such” seems tobe fundamentally beyond ‘That facts an example of the uncertainty principle. Measurable physical properties like color and hardness are said to be “in compatible” with one another, since measurements af one will (0 far as we know) always necessarily disrupt the other. (XK. Let's get in deeper Consider the rather complicated device8 shown in Figure 1.4. In one corner theres 2 hardness box. Hard clectrons emerge from that box, along route b, and at a certain point on route b there's a “micron” oF a “reflecting wall,” which ‘Changes the diection of motion ofthe electron but doesn't change lanything else (more particularly, it doesn’ change the hardness of fn electron that hounces off it), as shown. Similarly, soft electrons fmerge along routes, and they run into the same sort of mitrog tnd finally routes band s converge atthe “black box.” That black box is another device For changing the diections of ‘tion of electrons on either of these wo routes without changing their hardnesses, What it does isto make those ewo routes coincide after they passthrough i, The fact cha a soft electron entering the back box along « will emerge along hand s asa soft electron, and that @ hard electron entering the black box along b will emerge along P and sas 2 hard electron, can, of couse, be tested indepen- dently before we set everything up; and the same goes for the ‘Lets do some experiments with this device. Suppose, fst, that we feed a white electron into che hardness box, and then, once sha passed all through the setup drawn in figure 14, along path and s, we measure its hardness, What will swe expect to find? Well 50 percent of such electrons wil ake route Jy out of the hardness box, and 50 percent will ke routes. And 0.50 percent will end up at P and s a5 hard electrons, and 50 percent will end up there as soft ones (since nothing between the hardness box and b ands, on either root, can change hardnesse). On repeating the experiment many times, we find that chat is precisely what happens Suppose that we feed a hard electron into the hardness box and then measure its color at band s, Well, every electron like chat wil take route b through the apparatus and will emerge at and s a8 4 hard electron, and, a8 we know, $0 percent of such electrons will be found to be black by a color measurement, and 50 percent will be found to be white Ifa soft electron were fed into the hardness box at che beginning it would take route + and sill he soft at b And s, and so the final colorsmeasurement statistics of electrons ike that ought tobe fifty-ilty 100. And all that, as before, i indeed just SUPERPOSITION 9 what happens when we actually eacry outa large number of such experiments Ler’ try one more. Tis one isthe surprise. Suppose that we fed 1 white electron into the hardness box: a the begining, and mea sure color atthe end. What should swe expect then? Well, just as above, 50 percent of such electrons wil urn out to be hard and will eke ehrough the apparatus; and 50 percent will tun out eo be sof and wil takes. Consider the frst half. They will all emerge ath and sas hard electrons, and consequently 50 percent will wen ‘ut tobe black, ona color measurement, and 50 percent will wen ‘ut to be white. The second half, on the other hand, wil all emerge as soft electrons, bur of course theit color statistics wl be precisely the same. Putting all that together, ic follows that of any large set fof white electrons fed (one at 2 time) into this apparatus half should be found atthe end to be white, and the ther haf should be found to be black; and of course that makes very good sense, since (aside from afew armless mirrors and a harmless black box) this apparatus is really justa hardness box, which i already known to randomize the color! ‘The fonny thing is that when you try this last experiment, that isn’ what happens a all Ie ucns out that exactly 100 percent of white electrons fed into this appasatus come out white t the end. Thats very odd. It's hard to imagine what can possibly be going ‘on, Maybe another experiment will help clea things up. Les try this: Rig up a small, movable, lecton-stopping wall that can be slid a wil, in and out of, say, routes ee Bgure 1.5), When the wal is “out,” we have precisely out easier apparatus; but when the wal is “in,” all electrons moving along s gee stopped and only those moving along h get through to and s, What should we expect to happen when we slide the wall in? ‘Well to begin with the overall output of electrons at hand sought to go down by 50 percent, since the input whit electcons ought to be half hard and half soft and the latter shouldn't now be getting through. What about the color statistics of that remaining 50 percent? Wel, when the wall is out, 100 percent of white electrons fed in end up white. That means that all the electrons that take « tend up white and all the electrons that take b end up white; and0 Figure 1S since we can easily verify that whether the wall isin or out of scan Ihave no effect on the colors of electrons traveling along that implies that those remaining 50 percent should al be white ‘What acwally happens when we do che experiment? Well the ‘output is dawn by $0 percent, 35 we expect, But the remaining 50 pescent is aot all white. Ie’ half white and half black. The same thing happens, similarly contrary to out expectations if we insert ‘wall in the haed path instead. Now we're in real trouble. ‘Consider an electron which passes through our apparatus when the wall is out. Consider the possiilities as to which route that electron can have taken, Can ic have taken 6? Apparently not, because electrons which take (as we've just seen agai) are known teohave the property that thee color satis are Bilty-flty, whereas an electron passing through our device with the wall out is known to have the property of being white at b and s Can it have taken «then? No, forthe same reasons. Cant somehow have taken both ‘routes? Well, suppose chat when a certain electron ie in the midst ‘of passing through this apparatus, we stop the experiment and look to see where itis. Ie eurns ov that half the time we find i on b, and half the time we find it on s, We never find two electrons in ther, of two halves ofa single, split electron, one on each route, for anything lke that. There isnt any sense in which the eletron seems ro be taking both routes. Can it have taken neither route? SuFERPOSITION ii Certainly not. IE we wall up both routes, nothing gets dhrough at all ‘So what we're faced with is tis: Electrons passing through thie so far as we ate able to fathom the matter, do not tke couteb and do not take route # and do not take both of those routes and do nor take neither of those routes, and che trouble is that those four possiblities are simply all of the logical possiblities that we have aay notion whatever of how to entertain! ‘What can such electrons be doing? Ie seems they mast be doing something which has simply never been dreamt of before (if our experiments are valid, and if our arguments are right). Electrons seem 10 have modes of being, o¢ modes of moving, available to them which are quite unlike what we know how to think aboot The name of that new made (which i ust a name for something wwe don't understand) is superposition. What we say about an initially white electron whichis now passing through our apparatus ‘with the wall out that isnot on and not ons and not on both and not on ncither, at, rather, that it's ina superposition of being ‘on band being on s. And what that means (other than “none of the above") we don’ know. And some of what this book is going to be about are a number of attempts o as it were) say something ‘more about superposition than that. Let's make the main poin which is that superpositions are extraor dinarily mysterious situations) one oF two more times. Here's a second example. I's possible ro construct boxes which Td like ro eal “totaof-nothing” boxes. A toxalof-nothing box is a box with two apertures. An electcon which i fed into one aper ture emerges from the other with all of its measurable properties (color, hardness, velocity, whatever) unchanged; and the time of passage through the box for any electron is equal to the time it ‘would have taken for that electron ta traverse an empty space the size ofthe box, Those are the defining properties of total-of nothing boxes Cleary, there ae lots of ways to build total-of thing boxes. A complecely empty box with two holes in itis total-of nothing box: ‘We can also imagine building boxes which do all sorts of violent2 things tothe electrons fed into them but which subsequently undo all those things and finally ejet those electrons, atthe right times, a the eight speeds, with all oftheir original physical properties. Every box like that will be a total-of nothing box too. Now, recall the ewo-path apparatus of figure 1. White electrons fed into that apparatus always come out white. I urns out to be posible to build a cotal-o- nothing box which, when inserted into either one of those two paths, will make all of those outgoing electrons black instead of white. If dhe box is removed from the path, che outgoing electrons will ll go back to being whites So, inserting such a box into one of chose paths will change the color cof an electron passing through this two-paths apparatus. But total ‘of-nothing boxes, by definition, change none of the properties of «electrons which passthrough them; and, ofcourse, total-of nothing boxes change none ofthe properties of electrons which don’ pass ‘through them! So i isn’t possible that these clectcons pass through the total-oF-nothing box on one of the paths, since in that event thee colors couldn't have been changed from white to black by che presence of the box; and it isn’ possible that those electrons pass ‘outside ofthe box, since in thar event their colors couldn't have ‘been changed either! And i isnt possible (by our eaeie arguments) that those eletrons pass both inside and outside of the box, and it isn't possible that they pass neither inside nor outside ofthe box. Here's one final example, a very well-known one, Consider an arrangement such as is depicted in figure 16, On the left isa source ‘of lectrons. Elestcons emerge fom that source a whole spectrum of possible directions, as show. Slightly farther o the right is a screen which electrons eannot pass through, and that screen has two holes int. Sil farther othe ight isa fluorescent screen, much like a television screen, which lights up, at the poine of impact, whenever itis struck by an eletron (that i this faorescen screen iva measuring device forthe postions of electrons) Suppose, first, thatthe top hole in the frst screen is closed up, asin figure 1.6A. Electrons emerge, one by one, from the source, 4. Thar there cam be bones ie hat was fe ped in fone paper of Figure 16SuPEEPOSITION 4 and move toward the fist screen. Most of them ran into the screen and are stopped here. Some ger chrough the hole. Those latter land at various points on the fluorescent sereen, The statistics of those landings (that is: how many land in any particular region) are shown in the figure. Figure 1.65 gives the same informacion for the cease when the bottom hole is closed instead of the top one ‘What sor of landing satisics should we expect when both hoes ate open? Well all ofthe electrons which arrive ae the fluorescent screen will have passed ether through the top hole or through the brtom one. Those which passthrough the bortom hole are known (by our first experiment with this apparatus) to give rise to the statistical landing pattern of figure 1.6A; and those which pass through the top hole are known so give ise othe staisical landing pattern of figure 1.68; and so, inthe event that both holes are open (and inthe even that only one electron is allowed ro passthrough this apparatus at 2 time), we should expect a statistical landing pattern which is che direct sam of those two (as shown in figure 1.6C). Bur tha i wil be no surprise by now) i not what happens, ‘The statistical landing pattern which emerges on the fluorescent screen when both holes are open is markedly diferent from the dlitece sum of patterns A and B. So, is inconsscent with these ‘experimental results t© suppose that an electron passing theough this apparatus passes through the upper hole when both hoes are ‘open; and itis inconsistent with these experimental results to sup- pose cha an electron passing through this apparatus passes through the lower hole when both holes are open. And the same sorts of ‘experiments and arguments as were described above will ental that it also can't be maincained that such electrons pass through both holes, and chat it also can't be maintained that they passthrough either hol, ‘These electrons are (then) in superpositions of passing through the upper hole and passing through the lower one; but (once again) we have no idea, or rather only a negative dea, of what that means, All his stuff about superpositions, by he way, sheds avery curious light on the phenomena of uncertainyy and incompatibility (be- SUPERPOSITION is ‘sween color and hardness, say) that we ran Inco atthe beginning ofthis chapcer Consider ths: We know, by experiment, hat electrons emerge from the hard aperture ofa hardness bos ifand only if they're hard electrons when they enter that box (char asa matter of fact, isthe sole property in virue of which anything ever deserves to be caled 1 hardness box). Similarly for soft ones. Now, when a white elec- tron is fed into a hardness box, it invariably emerges (as we've just seen) neither through the haed apercure nor through the soft one nor theough both nor chrough neither. So, it follows chat a white clectron can't bea hard one, oF 2 soft one, or (somehow) both, or either, To say that an clectcon is white must be just the same as to say that ie’ in a superposition of being hard and soft. ‘And consider (in ight of thar) why iis that we find thar we can ‘ever put ourselves in a poscion to say, “The color ofthis electron is now such-and-such and its hardness ie now such-ande-such.” Ie isn't chat out color and hardness boxes are built (somehow) crudely (whichis whar we suspected at fist). And in fact ie isnt at all a ‘matter of our being unable ro simultaneously know what che color and the hardness ofa certain electron is (chat isi isn'ta matter of lgmorance), les deeper than that, I's that any electron’ even having any definite coloe apparently entails that is neither hard nor soft ‘not both nor neither, and that any elecron's even having any efinite hardness apparently entails that it's neither black nor white ror both nor neither, ‘And consider (while we're at it) why itis chat the rules for predicting the outcome of a measurement of (say the hardness of 2 white electron rum out (in so far as we're now able ro determine) ta be probabilistic rules rather than deterministic ones, lt’ like thst if ir could ever be said of a white electron that a measurement of its hardness will with certainty produce the outcome (say) “soft or if ic could ever be said of a white electron that a messorement of ics hardness will wth certainty come out “han,” that would apparently be inconsistent with what we now know tobe the ease, ‘which is that such an electron (a white one is in a superposition ‘of being hard and bein sft. And on the other hand, our experinceSureRPosinion is that every hardness measurement whatsoever either comes out “hard” oF it comes oot “soft.” And so apparently the outcome of athardness measurement on a white electron has go to be a matter of probability. But of course the business of talking more carefully about al his stuff (which is what I want to do hers) will equite an appropriate ‘mathematical apparatus, And so the next chapter wil lay an apps- ‘atu like that out, see Dave The Mathematical Formalism and the Standard Way of Thinking about It ‘There is an algorithm (and the name of that algorithm, of course, is quantum mechanics) for predicting the behaviors of physical systems, which correctly predicts all ofthe unfathomable-looking behaviors ofthe electron in the story in Chapter 1, and there is standard way of interpreting that algorithm (hat iy way attempt ing to fathom those behaviors, a way of attemping ta confront the fact of superposition) which can more or less be traced back to some sayings of Niels Bohr This chapter will describe tha algo- ‘thm and eehearse that standard way of alking about it, and then ie will apply chem both, in some detail, to that story. Mobematical Preliminaries Let me say afew things, co begin with, about the particular math- ematical language in which it is most convenient to verite the algorithm down, Let's start with something about vectors. A good way to think about vectors i to think about arzows. A vector is 2 mathematical ‘object, an abstract object, which (ike an arrow) is characterized by 1. Theory eh elton ofthis nda way of thinking i very leg and competed eee ani lhe cmpletay ied hse. Theft eect (qotin a wha Horie rly thought sons hn url bere a ‘hoe wll mater fora he legacy wich Bob and Followers hae by tshzever oa, a whatever they hemslen ay have igi shy ‘Poem pyc Tha ley at rd my canbe shared fh sere, The ame a hat gaye Coenen tet of guna shanTHe MatHeManical Fevaat 8 hich the arrow is pointing) and a a direction (the dicection in ‘magnitude (the length ofthe arrow. "Think of a coordinate system with a specified origin point, Every distinct geometrical point in the space mapped out by such a ‘coordinate system can be associated with some paticular (and listinct) vector, as follows: the vector associated with any given Point (in that given coocdinate system) is the one whose tip lies at ‘the given point and whose tl ies atthe origin, The length ofthat vector isthe distance between these two points, and the direction ‘of that vector isthe direction from the origin tothe given point [sce figure 2.1) The infinite colletion of vectors associated with all the poines in such a space is referred to asa vector space. ‘Spaces of points can be characterized by (among other things) their dimensionality, and spaces of vectors can too. The dimension of a given vector space is just the dimension ofthe associated space ‘of points, That later dimension, of cours, is equal to the number ‘of magnitudes, the aumber of coordinates, that need tobe specified in order to pick out (given a coordinate system) some particular geometrical point. THE MAIIMATICAt FORA 19 Figure 2.2, for example, shows a two-dimensional space a plane of points, wherein (given the indicated coordinate system) two coordinates need to be specified the x-coordinate and the y-coor- dinate) in order to pick out a point. The reader can convince herself that a line of points forms a one-dimensional space, and that the space we move around in has three dimensions. Spaces of points with moce dimensions than that are hard co visualize, but the formal handling (that is: the mathematical handling) of such spaces is nota problem Let's introduce a notation for vectors: the symbols |) around some expression will henceforth indicat that that expression isthe name of the vector; so that, fr example, |A) will denote the vector called A. That’ the notation mose commonly used in the literature fof quantum mechanics. ‘Vectors can he added to one another. Here's how: To add JA) to |B), move the rl of |B) tothe cp of |A) (without altering the length ‘or the direction of ether in the process). The sum of jA) and [B) (oehich is writen |A) + |B) is defined to be that vector ((C)) whose tail now coincides with the tail of A) and whose tip now coincidesTHe MATHEMAT AL Fonmauisa, 20 withthe tip of B) (sce igure 2). The sum of any wo veetors in fny particule vector space is always another vector in tha same Space (tha, indeed, ix part of the dition of a vector space. Think, for example, ofthe spaces discussed above “That fats going tobe important. Vector in quaneam mechan ies, are going to represent phyial sates of afar. The addition of vectors wil turn out t have something 10 do wit the sperpo- Son of physical states of af. The fat that vo vets can be fcdded together to form 4 third will tun out to accommodate, wwithin the algoiten the fact that certain physical states of afar, Stes ike being white, re superpostions of certain other stares of Mais, stats Ike being hard and being sof but of all his more inter ‘Vectors can be muliplied too, Ther are two ways to multiply thom. ist ofall they can be muliphed by numbers. The vector 5A), say, defined to be that vector whose direction isthe same a thedreton of A) and whos length is 5 times the length of A). Sia) = |) + 1A) + [AD + |A) + A). OF course, if [A isa element ofa certain vector spac, any number tes A) wil be an element ofthat space too. The other way to muliply vector isco multiply thm by other vectors, The mulipliation of «vector by another vetr yields a fhumber (noe a vector) |A) times [B) (which is weiter (AID) is defined to be the flowing number: the length of [) times the length of 8 mes the cosine ofthe angle, 0, berween A) and |B). fe> >, rs Bre 23 THe Matnematic ~ 2 “The length of [A) (also called the norm of |A), which is writen IAD is obviously equal the square root ofthe number (Ald, since the cosine of O° (0" i the angle between |A) and itelf i equal to 1 So, vectors plus vectors are vectors, and vectors times numbers are vectors, and vectors times vectors are numbers, Here's slightly moce sophisticated way of defining a vector space: a vector space is a collection of vectors such thatthe sum of any two vectors inthe collection is alzo a vector in the collection, and such that any vector inthe colletion times any (eal) number is also a vector in the collection. Such eolletions (by the way) clearly have to be infinite. Think, again, ofthe examples of spaces described above IF [Al = 0 and [By v0 and yer (AIB) = 0 (that i ifthe angle between {A) and |B) 90°, since cos 90° = 0), then lA) and |B) are said to be orthogonal to one another. Orthogonal jst means pet pendicalae Here's another definition of dimension: The dimension ofa vec tor space is equal (by definition) to the maximum number (all that ‘number N) of vectors JA, JA. |As) which can be chosen i the space such that forall values of i and jfrom 1 theough N such that # j, (AJA) ~ 0. That ithe dimension of a space is equal to the cumber of mutually perpendicular directions in which vectors within that space ean pon. Given a space, there are generally lots of ways to pick out those directions Pick a vector, at random, from an Nedimensional space 1k will always be possible to ind a see of N ~ 1 other vectors in that space which are all orthogonal t that original vector and to ‘one another. In most cases, given that original vector, there wil il bbe many such orthogonal ses (or, rather, an infinity of such sets) to choose from. Figize 24 shows some examples, ‘Think of an Nedimensional space. Think of any collection of N mutually orthogonal vectors in that space, and suppose that the ‘norm, che length ofeach of those vectors happens fo be 1. Such a set of vectors is said to form an orthonormal basis of that N-di mensional space. Ortho is for orthogonal, normal is for norm, and here's why sets of vectors like that are called bases of their spaces: Suppose that the set Ai), [As)« |An) forms a basis of aTHE MATHEMATICAL FORMMAUsN Ieee ewe 24 certain N-dimensional vector space; it turns out that any vector whatever in that space (call it |B) an be expressed asthe following 1) |B) = BAD + BJA) +. bxAN) where the b, ae all simply aumbers—more particularly, simply the following numbers a lay So any vector in a vector space can be "buile up” (as in (2-1) out of the elements of any basis ofthat space. All hat sillustrated, for 1 two-dimensional space, in figure 2.5. Bases end up amounting to precisely the same things coordinate systems: given a coordinate system for an N-dimensional point space, N numbers the coordinate values) will suffice to pick out a point; given a basis of an N-dimensional vector space, N numbers the by of equation (2.1)) will suffice to pick out a vector. Vectors Which are of norm I and shich point along the pespendicular coordinate axes ofan N-dimensional point space will constitute an Fomaaw Figure 25 ‘orthonormal basis of the associated Nimensional vector space, and vice versa, For any space of more than a single dimension, there will be an infinity of equivalently good orthornormal bases 0 choose from. ‘Any vector in that space will be writable, la (2.1), in terms of any of those bass, but ofcourse, fr a given vector [B), the numbers by in (21) (which, by the way are called expansion coefSciens) will dlfer from bass to bass, Figure 2.6 shows how that works, "Now, it happens to be the ease that for any three vectors /A},[B), and |O), the product |A) times the vector (JB) + [C) is equal to the product |A) times [B) pls the product [) ims |C) (23) (AB) + (0) = lB) + 10, and tha can be shown to entail, for any two vectors |M) and [O), that (24a) 1M) + 10) = (me + and + (me Had + + (on guliAsdTHe AMATHEANANICAL FORBAUSAL 2a 4 gine 28 and that (2.46) (AQ) = mgs + mmgy = + me wherein the m, and g. are the expansion coefficients of [M) and |), respectively in any particular bass A). The numbers q and my, wll, fof course, depend on the choice of basis, but noc thatthe sum of theit products in (2-4b) (which is equal to (MIQ), which depends ‘only on which vectors [M) and (Q) happen tobe, and not on which basis we happen to map them out in} will not. That sum, rather, will be inriant under changes of basis. Suppose thar we have agreed to setle on some particular basis, for some particular vector space. Once that's done, all chat willbe equited for us to pick ont some particular vector (jO), say) willbe tospecfythe numbers the expansion coefciets) of |O) for that particular basis. Those N numbers chen (once the basis is chosen} THe MATHEMATICAL FORMALISM 25 can serv to represent the vector. Those numbers are usually written down in a column; for example: as 1 fo=| 5 | [oan = 1 = the three~dimensional vector for which OIA.) = 5 lay = (see equation (2.2)), where the |A) are dhe chosen bass vectors. I follows fom (2.4b} that the norm (the length) of any vector |O) vill be equal to the square root of the sum of the squares ofits expansion coefcients. That number, too, must obviously be inva: ant under changes of basis. ‘That’ all chat will concern us about vectors. The other sorts of| mathematical objects which we shall need to know something about are operators Operators are mechanisms for making new vectors out of old ‘nes. An operator on @ vector space, mote particularly, is some definite prescription for taking every vector in that space into some other vector; itis a mapping (for those readers who know the ‘mathematical meaning of that word) ofa vector space into isl. Let's introduce a notation. Suppose that the operator called O is applied to the vector |B) (that is: suppose that the prescription called © is cartied out on the vector |B). The result of tha oper ation, ofthat procedure, i writen: 25 of) “Then what was just sid about operators can be expressed like this 2.7) OB) = |B) for any vector [B) in the sector space on which (0 isan operator. where [Bis some vector in the same space as [8 Here are some examples. One example is the “unit” operatorAlical FoR MAUS (that’s the prescription which insructs ws to multiply every vector inthe space by the number 1, to transform every vector into itself). The unit operator is the one for which 2.8) O48) = [B) = 18 Anosher example isthe operator “multiply evry vector by the ‘number 5.” Anotber example ithe operator “otate every vector locks by 90° abou some particular vector [)” (ee Figure 27). Another example isthe operator “map every vector in the space into some particular vector [A “The particular sets of operator which wll ply a vital coe in the quantum-mechanicl algorithm are linear operators. Lineat operators ate, by definition operators which have the folowing properties: (2.93) O14) + [8)) = O4> + O18) and 298) fea) = aO'AN) 2 Thence ponent weet ce bw becomes al Le THe Mars HAATICAL FORMALSH D for any vectors |) and |B) and any number ¢, Hese’s what (2.9a) says take that vector which isthe sum of two other vectors |A) and |B) (such sums, remember, are always vectors) and operate on that sum with any linear operator. The resultant (new) vector wil be that vector which isthe sum ofthe new vector produced by oper ating on |A) with O and the new vector produced by operating on |B) with O. Whae (2.96) says is thatthe vector produced by oper ating on times [A) with O is the same as ¢ times the vector Produced by operating on |A) itself with O, for any number c. [Nov the two conditions in (2.9) pick out a very particular sort ‘of operator. They are by ao means properties of operators in general, Let me leave it as an exercise for the reader t0 show, for example, that of the four operators just now described, the frst three ate lineae and the last one ist. Lineat operators are very conveniently representable by atrays ‘of numbers. Weleaened it was possible, remember, to represent any N-imensional vector, given a choice of basis, By N numbers a la «equation (2.5); and it similarly urns out o be posible to eepresent any linear operator (the linearity is crucial here) on an Nien sional yector space by N numbers. Those N? numbers are tadi sionally arranged notin a column (as in equation (2.5), for vectors), but ina matrix, as (fora two-dimensional operator, ay) follows On, Ou] on[on or] The numbers O, in (2.10) are defined to be 11} Oy = (AJOIA) That is: the number O, isthe vector OA) multiplied bythe vector {AD (such products of vectors, remember, ace always members, 2 Theo part of 2.9) aren complelyinapedet of one sneteby the vay Note foreman th ere tat ema ape (298 red hyTHe MATHEMATICAL FoRmausa, 28 where the Ay) ace the chosen basis vectors ofthe space. There's a rule for muleplying operator matrices by vector columns, which is 212) [Om Ou) [bi], [lub + nbs) [On On} “[o} " (Ono: + Ont) [Note thatthe right-hand side of (2.12) ie a vector column s0 this rl stipulates thatthe product of an operator matrix and a vector column is a new vector columa, Here's why ll cis notation susefal: crarns out (we won't prove it here) that any linear operator whatever can be uniquely specified (given a basis choice) by specifying the N* O, of equations (2.10) nd (2.11) fist as any vector can he uniquely’ specified by specify- ingthe N b,of equations (2-1) and (2.2) and (2.5); and iteurns out that for any linear operator O, we can calculate O's effect on any vector [B) simply by muliplying the O-mateix by the (8)-columa [given, as always, a basis choice) asin (2.12). Thais, for any linear ‘operator O and any vector |B): any is On Ou], [hr]. [Oubs + Oba) Ou, On}* [bu] (On, + Oba = (Oubs + OnbylAnd + (Onbs + OnbylA eee oe ae “ehese are number? where [A) are the chosen basis vectors. (The nexe-colast equality follows fom equations (2.1) and (2.2) and 2.5) §.ecape is worth ying altho wor: node cluae he effet of any lace opertor © om sy rector) it coon + ay then esate he [ncaa seco in tho ac by mean of femal (2.2) then acl he rato a ih es yan formas (2.11 hen ptt clan coy ha operator mata by meso els [2 12} a the es of hat Icaton wil be be column vector tat sue ass of the ew veo I) (lat te vestor otal by pestig wih O on 1B. fe MATHEMATICAL Fo) @ One more definition willbe wsefal I it happens to be the case for some particular operator O and some pasticlae vector |B) that 2.14) OB) = @|8) where @ is some number--that i, ifthe new vector generated by ‘operating on |B) with O happens be a vector pointing inthe same alirction as [B)—then [B) is said to be an eigenvector of O, with cjgenvalue @ (where @ is the length of that new vector relative 0 the length of). Certain vectors will in general be eigenvectors of some operators ad not of some others certain operators wil in general have some vectors, and not others, a8 eigenvectors, and other operators will have other vectors as eigenvectors. The operator eigenwector rela sion, however, depends only on the vector and the operator in question, and not at all on the basis in which we choose to write those objects down. In other words, if the eigenvector operator relation obtains berween the vector column and the operator ma trix ofa certain vector anda certain operator ina certain paticwlar hss, chen it can be shown that che same relation, with the same cigeavalu, wil obtain between the vector column and the operator ‘matric in any basis whatever of that space. “Here are some examples all vectors are eigenvectors ofthe wit operat, and al have eigenvalue 1y and simlary [but with eigen value 5} for the operator “multiply every vector by 5." All vectors of the form @|C), where @ is any number, are eigenvectors of the foperator “rotate every vector about (C) by 90°; all those vectors hhave eigenvalue 1, and there are no other igenvectors of that ‘operator The four-dimensional space operator {writen down in some particular basis) eas) 500 oj ooo O-lo 02 al 000-9HE MATHEMATICAL FORMMAusst 30 has eigenvectors oritten in that same basis) (2.16) ny 0} 0) fo) y-|2} am=|8} 1o=[2] wy [9 of fol fo} with eigenvalus 5,922, and ~7 spectively Any number times 14) of |B) of |C) oF |D) will be an eigenvector of O too, with the same eigenvalue; but vectors lke |A) + [B) won't be eigenvectors of O. ‘Quontum Mechonies [Now we're in a positon co write out the algorithm. It pretty much all boils down to five principles. (A) Physical States. Physical situations, physical states of affairs ae represented in this algorithm by vectors. They're called state vectors. Here's how that works: Every physical system (that is every physical object, and every collection of such object), to begin with is associated in the algorithm with some particular vector space; and the various posible physical sates of any such system ate stipulated by or coztespond to vectors, and more particularly {o vectors of legeh 1, in that systems associated epace; and every such vectors taken co pick out some particular such state; and the states picked out by all those vectors are taken to comprise all of the possible physical situations ofthat rystem (the correspondence jan’ precisely one-to-one, however: we shall soon diseoves, for example, that for any vector |A) of length 1, ~[4) must necessaely pick out the same pisicl state as |A) doe). This will rura out to bea very apt way to represent sates, since (as mentioned before) the possibilty of “superposing” two sates to form another gets reflected inthe algorithm by the possibilty of Adding (or subtracting) wo vectors to form another, (B) Measurable Properties. Measurable properties of physical sys- tems (such properties are referred to as observables, in the quan: Te MATHEMATICAL FORMALISM 31 ‘um-mechanical literature) are represented ia the algorithm by lin cat operators on the vector spaces associated with those systems. ‘There's a rule that connects those operators {and their properties) and those vectors (and thes physical states), which runs as follows: If the vector associated with some particular physical state happens to be an eigenvector, with eigenvalue (say) a, of am operator assor ciated with some particular measurable property ofthe system in ‘question (in such circumstances, the sate i sid to be an “eigen state” of the property in question}, then tha state has the value a ‘of tha particular measurable property. Ler’ ty all that out. Lets construct a vector space in which the state of being hard and the sate of being soft can be represented. Suppose we ler the following two two-dimensional column vectors stand for hardness and softness en tard) =[}] bo = v0 lo 1 Notice that if we adopt (2.17), (hardsoft) = 0 (see equations (2.4), and (2.5)) Asa matter of fac, the wo vectors in (2.17} constitute 2 basis of the two-dimensional space which they inhabit. That particular basis, by the way, i precisely the one in which he vector columns in (2.17) have been written down (that ist the felevant basis vectors [AY and JA.) of equation (2.5) are, in the case of (2.17), precisely hard) and fof). ‘What operator should represent the hardness property? Lat’ try thi an f 8) haadnes operate = [3 ‘] ‘where we stipulate that “hardness = +1 means “ha ‘and that hardness = ~1" means “sof” So far allthis works out sight: [hard and soft) of equation (2.17) tar, indeed, a5 the reader can now easily confim, eigenvectors ofTHe MarHemanicat Format Tar put hi example ahr. Remember ht i seed 0 in Chart tne the hs ee te a bth Be Sees ofthe “har and ot trance ber ham epee caps) fate tpepnon of pps {Rison ofc opeine oat veo emembe hat he tumor he ity of ty two eon ny pero er {poe ces eater er i ane pce Al ht Stes har eas f beng whe ard eigen ee terol by vs pace oan hee oh ihe bi gun 2% va [a] (3) 219) [back = “That works out right too: The reader can show that the various stipulations of (2.19) are all consistent with one another as ae the stipulations of (2.17) and (2.18). Furthermore black|white) = 0 too; and [black and jvhite)constcute another basis of this space [Now it follows from (2.43) that 220 ie wolf) ane im =[5] in some particular bass, then woim-[et3] in that same basis (and the same applies, of course, to vector columns of any dimension) Notice, then, how beautifully (2.17) and (2.18) and (2.19) reflect the principles of superposition and incompatibility. First ofall, it follows from (2.17) and (2.19) thar: ean etfbard) + also) ‘arthard) ~ Wiisof) Uhard) = aback) + Wxwhite) bof) = Varblack) ~ Watwhite> So sums and differences of vectors in the algorithm, do denote superpositions of physical states; and {just as we concluded in the last chapter) states of definite color are superpositions of diferent hardness states, and states of definite hardness are superpositions of different color states, Moreover, look how well the forms of the hardness and color ‘operators confirm al this: It's easy to verify thatthe “black” and “white” vectors aren eigenvectors ofthe hardness operator and thatthe “hard” and “soft” vectors aren't eigenvectors of the color ‘operator. The hardness and color operators are (ust as they ought to be) incompatible with one another, in the sense that sates of definite hardness (that i: states whose vectors ate eigenvectors of the hardness operator) apparently have no assignable color value (since those vectors aren’ eigenvectors of the color operator) and So it turns out thatthe descriptions of color and of hardness and of all the relations between them can be subsumed within a single, ‘wo-dimensional vector space. That space is ceferced to within the (quantum-mechanical literature as the spin space, and color and hardness ate refered to as spin properties, Ler’s get back to the enumeration ofthe five principles. (C) Dynamics. Given the state of any physical sytem at any “ini- tial” time (given, that i, the vector which represents the state of that system at that time}, and given the forees and constraints «0THe Mar Which that sytem is subject, there isa prescription whereby the state of that system at any later time (thats, the vetor at any later time) can, in principle, be caleulaced. There is, in other words, a dynamics of the state vector; there are deterministic laws about hhow the stae vector of any given system, subject co given forces and constraints, changes with time, Those laes are generally cast tn the form of an equation of motion, and the name ofthat equa tion, for nonrelativistic systems, is the Srédinger equation. ‘Since every state vector must, by definition, bea vector of length one, the changes in state vectors dictated by the dynamical laws are ‘exclusively changes of direction, and never of length Here’ an impoctant property of che quantuin-mechanical namical laws: Suppose that a certain system, subject to cert specified forces and constrains and whose state vector at time fi. {A), evolves, in accordance with che laws, into the state [A’) at time ‘4; and suppose chat chat same system, subject fo those same forces fand constraints if ts state vector at iy rather, |B), evolves, in Accordance with those laws, into the state [Bat time fs. Then, the laws diceae chat if chat same system, subject Co those same forces and constraints, were, rather ithe sta alA) + BIB) at time 4, then is state at time f will be aA” + BiB’) (where [A) and |B) can be any state vectors at all). Ths property of che laws will concern usa good deal later on. The name of this property is linearity (and rote that there is indeed a resemblance between “linearity” a8 applied t dynamical laws, hers, and “linearity” as applied to ‘operators, asin the two equations in (2.9) (D) The Connection with Experiment. So far, almost nothing in ‘these principles has touched upon the results of measurements. All wwe have isa stipulation in (B) chat the physical state whose sate vector is an eigenvector, with eigenvalue a, of the operator associ sted with some particular measurable property will have the value 2 for that property; and presumably it follows that a measurement ‘of thar propery, carried out on 2 system which happens to be in that state, wil produce the result a. But much more needs co be said about the results of measurements than that! What if we measure a certain property ofa certain physical sytem at amoment The MaTnemaricat FoRMausas ‘when (as must happen in the vast majority of cases] the state vector of that system does not happen to be an eigenvector ofthat prop- erty operator? What if, say, we measure the color of a hard electron, an electron in a superposition of being white and being black? ‘What happens then? Principle (B) is of no help here. Anew pein ple shall have to be introduced fo settle the question, which runs a8 follows Suppose we have before usa system whose state vector isla), and we catry out a measurement of the value of property B on that system, where the eigenvectors of the property operator for 8 are [B= &), with eigenvalues b, fe, BIB = b) = |B = b) for all. According to quantum mechanics, the outcome of such a measure iment is a matter of probability; and (more particularly) quantum mechanics stipulates chat the probability that the outcome ofthis measurement will be B = b, is equal ro: oy [Note that (as must be the ease for probability) the mumber denoted by the above formula wil always be less than or equal to I; and note that in che special case af eigenvectors covered by principle (B), (2.22) yields (asic should) che probability 1. And note that it follows from (2.17) and (2.19) and (2.22) thatthe probability chat a black electcon will be found by a hardness measurement to be, say, sof, is precisely as we have learned to expect) 12. ‘And this is where i emerges that the correspondence berween states and vectors of length | isn’t precisely one-to-one. First ofall, it follows from equation (2.3) that (for any vectors la) and |b) and any number @)(al@is) ~ @ialb). Now, since che probability (2.22) ‘Sepends only on the uae ofthe product ofthe vectors involved, and since (Ix)! = (x) i follows that the probability of any ‘esulk of any measurement caried out on a system ia the tae [a) willbe identical to the probability ofthat same result ofthat same ‘measurement carried out on a system in the state ~la). Vectors a) and ja), then, have precisely the same observable consequences; which isto say as is customary inthe quantum-nechanieal iteraTHE MATHEMATICAL FORA! 36 ture) that the vectors fz) and ~ physical state. ) represent precisely the same (E) Collapse. Measurements (as I remarked in Chapter 1) are ‘ways, in principle, repeatable. Once a measurement is carried out and 2 resueis obtained, the state ofthe measured system must be such as to guarantee tha if that measurement i repeated, the same tesule will be abeained.* ‘Consider what that entails about the state vector ofthe measured system. Something happens to that state vector when the measute- ment occurs. If, say, a measurement of an observable called O is catried out on @ system called S, and if the outcome ofthat mea surement is © = @, then, whatever the state vector of S was just prior to the measurement of O, the state vector ofS just afer that ‘measurement must necessarily be an eigenvector of O with eigen value @. The effet of measuring an observable muse necessarily be to change the stave vector ofthe measured system, to “collapse” iy to make it “jump” from whatever it may have been just prior £9 the measurement into some eigenvector ofthe measured observable ‘operator. Which particular such eigenvector ic ges changed into is cof course determined by the eutcome of the measurement and note that that outcome, in accordance with principle (D), ie a matter of probability. Tes ar this point, then, and at no point other than this ‘one, that an element of pare chance enters into the evolution ofthe ‘Those are the principles of quantum mechanics. They are the most precise mechanism for predicting the outcomes of experi iments on physical systems ever devised. No exceptions t© them Ihave ever heen discovered, Nobody expects any. Suppose chat we should lke to predict the behavior of some par ticular physical system by means ofthis algorithm, How, exactly, 4. Seppo ou, that heres een 90 “amperng ia the imei nd suposnythat ot enough tne has eps forthe par mums of Be ‘eemeed ye sit bingabout caps the ao of he mead chy ae MATHEMATICAL FORMALS 9 ddo-we go about that? The frst thing to do is to identify the vector Space associated wit that system: the space wherein all the possible physical sates of tha system can be represented. Given a precise physical description ofthe system, there are systematic techniques for doing that. Then the operators associated with the various smeasucable properties of that system need to be identified, There ate techniques for doing that too. With that done, the specific ‘correspondences between individual physical states and individual vectors can be mapped out (the vector which corresponds t0 the state wherein a cetain measurable property has a certain value, for cxample, will be the one which isan eigenvector, with that cigen ‘value, of the operator associated with that property). Then the Present state vector ofthe system can be ascereained by means of measurements, and then (given the various forces and constraints to which the system will be subject) che state vector of any furure time canbe calculated by means af the prescription of principle (C)y and then the probabilities of particular outcomes of a measurement aried out at some such future time ean be cakulated hy means of principle {D), and the effect of such 4 measurement on the state vector can be taken into account by means of principle (E). And then principle (C) can be applied yer again, ro char new state vector {che state vector which emerges from the measurement} 0 calculate the state vector of this system yet farther inthe future, up tothe ‘moment when the next measurement occurs, whereupon principles (D) and (E) can be reapplied, and so on. Notice, by the way, that principle (6) stipulates that under certain parccula circumstances (namely, when a measurement occurs) the state vector evolves in a certain particular way it “collapses” onto an eigenvector of the measured observable operator). Notice, 0, that principle (C) is supposed to be a completely general account ‘of how the state vector evolves under any circumstances If thats allso, a question of consistency necessaly arses: it seems like (E) ‘ought tobe jst a special ease of (C), tha (E} ought 0 be dedacible from (C). But it isnt easy to see how that could be 30, since the changes in the state vector stipulated by (E} are probabilistic, whereas those stpolated by (C) ate invariably, deterministic, This38 's going to require some worrying about, br lets not start that just yet thae worrying will commence in earnest in Chapter 4 ‘As I mentioned before, theresa standard way of talking, which students of physics are waditionally required to master along with this algorithms, about wha superpositions are. Tha ling that way of dealing with the apparent contradiction of Chapter 1, boils down to this: the right way to think about supespositions of, sa, being black and being white is to think of them as situations wherein color predicates cannot be applied situations aherein color talk is unineligible. Talking and inquiring about the color fof an electron sn such circumstances is (on this view) like talking for inguiting about, say, whether or not the number 5 is stil a bachelor. On this view, then, the contradictions of Chapter 1 go away. On this view i jst isn't so that hard electrons are not black and not white and not both and not nether, since color talk of any kind, about haed electrons, simpy has no meaninygat all. And tha’ the way things are, on this view, forall sorts of superposition: superpositions are situations whercin the superposed predicates just don’ apply: ‘OF course, once an electron has been measured co be white or black, chen it white or black (then, in other words, color predi «ates surely do apply). Measuring the color ofa hard electron, then, isn't a matter of ascertaining what the color of tha haed electron is; rather, itis matter of Bist changing che sate of the measured electron into one to which the color predicate applies, and ro which the hardness predicate cannot apply (this is the “collapse” of prin- ciple (E)) and then of ascertaining the color ofthat newly created, coloc-appicable state. Measurements in quantum mechanics (and parcicularly within this interpretation of quantum mechanics) are very active processes. They aren't processes of merely learning something; they are invariably processes which drastically change the measured system, “That's what's atch heart ofthe standard view. The test (of which I shall have much more to say later on) is details THe MATHEMATICAL FORMAUIS 30 Here (before we move on to particular cases) ate afew more general technicalities Fist, the vector spaces which are made use of in quantum me. chanics are complex vector spaces. A complex vector space is one in which its permissible to multiply vectors not merely by real numbers but by complex (te, eeal oF imaginary or both) numbers in order to produce new veciors. In complex ector spaces, the expansion coefficients of vectors in given bases (the D, of equation {2.1)) may be complex numbers too. That will necessitate a few refinements of what's been introduced thus far, In complex vector spaces, the formula forthe product of ewo vectors, written in terms of their expansion coefficients in some particular basis (that i, the formula (2.4b)), needs to be changed, very slightly (what, precisely, i gets changed into need not concern us here), im oeder ro guarante shat che norm of any vector that i, its length: AVA) cemains, under all circumstances postive eal number. Formula (2.22) for probabilities needs to be altered very slightly too, since, i complex spaces, (AJB) and, hence, (2.22) may be complex numbers (and yet probabilities must necessarily be real positive numbers berween 0 and 1). The soliton sto change (2.22) 2) (iB oF where the vertical bars denore absolute value (or “distance from 2210," which is invariably a real, positive number). Equation (2.23) stipulates that the probability that a measurement of B on a system in the stare la) will produce the outcome B= is equal to the square of the distance from 0 of the complex number (alB = bs and chat probability, so defined, wil invariably be areal and pos itive number. Formula (2.22), by the way, will entail not only that |A) and ~/A) represent the same physical state (we've already scen that to be the case}, but, more generally, that |) and @#A) repeesent the same stare, whete @ may be anyone ofthe infinity of complex ‘numbers of absolute value “The elements of the operator matrices of linear operators onTHe MarHEMaTicas FoReAAUsat 20 complex vector spaces (hati he numbers O, of (2.10) and (2.11) tan be complex numbers too, Nonetheles, it may happen to some such operators that all of their eigenvectors are associated only ‘vith real eigenvalues (albeit, perhaps, nor all of thee matrix ele ‘ments ©, and perhaps even none of them, are real). Linear oper ators like that are called Hermitian: operators; and i's clear from Principle (B) (since, of course, the values of physically’ measur able quantities are always real numbers} tha the operators assoc ated with measurable properties must necessarily be Hermitian ‘operators. ere are some facts about Hermitian operators (1) Ifo vectors are both eigenvectors of the same Hermitian ‘operator and ifthe eigenvalues associated with those ewo eigenvec- tors ace two diferent (real) sumbers, then the ewo Vectors in question are necessarily orthogonal to each other. Tat prety much had to be so, if this algorithm i going to work fut right; otherwise, measurements wouldn't be repeatable. The differen eigenvalues of a property operator, afterall, correspond to different values of that property; and (if measurements of Property are to be repeatable) having a certain value of a certain Property must entail that subsequent measurements of tha peop: ‘erty wil certainly not fied any other value of i? and that (given principle (D)} will requite thar sate vectors connected with differ nt values of the same measurable property (black) and (white), say, o hard) and [sof be orthogonal to one another. (2) Any Hermitian operator on an N-dimensional space will always have at lease one ser of N mutually orthogonal cigenvectors ‘Which isco say: it will slways be possible to form a basis of che space out ofthe eigenvectors of any Hermitian operator; diferent bases, ofcourse, for diferent operators. Consider, for example, the hardness operator of equation (2.18) and the coloe operator of equation (2.19) 3) The reader ought to be able to persuade herself, now ofthe following: if a Hermitian operator on an N-dimensional space 5. Supposing onc against wo tampering, and vo dynamical esi, bs 41 happens to have N diferent eigenvalues, then there is a unique vector in the space (or, rather, unique modulo multiplication by numbers) associated with each different one of those eigenvalues and ofcourse the set ofall eigenvectors of length I of tha operator will form a unique bass ofthat space (or, rather, unique modulo ‘multiplication by numbers of absolute value 1), Operators lke that are called complete or nondegenerate operators (4) Any Hermitian operator on 2 given space will invariably be associated with some measurable property ofthe physical system ‘connected with that space this is just a somewhat more informative ‘version ofthe first part of principle (B)). (5) Any vector whatever in a given space will invariably be an cigenvector of some complete Hermitian operator on that space, That, combined with fact (4) and principle (B), will entail that any {quantum state whatever of given physical system will invariably be associated with some definite valve of some measurable property of that system. Allehis turns out to entail (among other things) tum-mechanical system necessarily has an infinity of mutually in- compatible measurable properties. Thnk {just to have something concrete totale about) of the space of possible spin states of an lectron. There are to begin with, a continuous infinity of different Such stats (since there area continuous infinity of vectors of length {Vin a two-dimensional space); moreover, given any one of those state, there are clearly 2 continuous infinity of different possible states which are nor orthogonal to it. And, by face (3) and (S) above, every state in this space is necessarily the only eigenstate associated with a certain particular eigenvalue of a certain partc- lar complete operator, and, by fact (1), none of the continuous infinity of stares which aren't orthogonal to the state in question can possibly be eigenstates ofthe same complete operator: What's ‘more, the complete operators of which those other states are eigen: states clearly can't even be compatible with the operator in ques tion, And so since allthis applies to every state in the space) there must necessarily bea continaous infinity of mutually incompatible ‘complete measurable properties, of which color and hardness are only ewo.Ie will be useful, for what comes late, to give wo more of those properties names. ‘The vectors izhlack) + VSi2bwhite) and VSi2|black) ~ 1/2}hice are both of length 1 and are orthogonal to one another (and aren't ‘orthogonal 10 any ofthe eigenvectors of color or hardness), and so it follows that there must bea compete observable of which they are both eigenstates, with diferent eigenvalues (which can always heset at +1 and ~1, respectively). Ler call hat observable “gle, And the vectors 1u2[black) ~ V3Ajwhite) and V3I2[back) + W2}white) are both of length 1 and are orthogonal to one another (and aren’ ‘orthogonal 10 any ofthe eigenvectors of color or hardness or seb), and 50 i follows tha there must bea complete observable of which they ace both eigenstates, with different eigenvalues (which can always be set at +1 and ~1, respectively). L's call thar observable “scrad.” Ofcourse, the eigenstates of lb and sera (us lke those of color and hardness) both form diferent bases of the spn space. Finally, there ate cules (never mind what those rules are, precisely} for adding and subtracting matrices to or fom one another, and for multiplying them by one another. The commutator of two ‘matrices A and B, whichis denoted by the symbol [A,B], is defined to be the object AB ~ BA the rules for mulkiplying matrices by ‘one another ena thatthe order of multiplication counts: AB isn’ necessarily the same as BA) ‘Now, it can be shown that inthe event that [4,B] = 0 (chat is, inthe event that AB # equal to BA], A and B sbare a least one set of eigenvectors which form a hasis of the space. A lite reflection ‘will confirm thatthe operator matrices of incompatible observables ‘can’t posubly shace any such complete basis of eigenvectors (since such eigenvectors would correspond to definite vale states of hoch observables at the same time). Ir must be the case, then, that she ‘commutators of incompatible observable matrices are nonzero, So THE MATHEAMTICAL FORMA a3 the property of commutativity (that i, the condition [4,8] ® 0) turns out to he a convenient mathematical test for compatibility Moreover, in cases of incompatible observables, the commutator of the two observables in question turns out ro be extremely useful for assessing the degree of their incompatibility.’ Cooidinate Space Let's begin to apply al his. Lvs see, in some detail, how ro setup a quaneum-mechanical representation, and a quantum-mechanical ‘dynamics, of some simple physical system. Forget about color and hardness for the moment. Think of a familie sor of particle, one with only the familie soets of physical properties: positon and ‘eocity and momentum and energy and things like that Here's a way to get started: We know, from hundreds of years of| experience, tht the behaviors of relatively big particles, with rela- tively big masses (particles you can see lke rocks and baseballs and planets) are very well described by the classical mechanics of Newton. That entails something about the quantum theory of particles: whatever that theory ends up predicting about the stcange, tiny particles of Chapter 1, it oughe to predict that every day particles, subject to everyday circumstances, will behave in the 6 hap the notion of her ing varios erent dere of nc relies Some dustin, Hers wit tee or heres hat ay Indhe simples cat whe the obserabes cred ate Bath comple Conder re compe andncompaiie ober al tem Aan fe plc sytem. fy when ay prc een of Acai the acon ot Deane of Bean be peed (snes of msl (2.2) with cen syproschig cra heh focachegenector ochre some parle ‘lense of B ach thatthe prods hae two vests ie smehing spec tigen then Aand Bari be ony very gly ncmpaile, Bu f(t the ‘ther exec, when any paar ngraae of Alea the pobleeo ‘he various pable atome oft meester Baz lhe seh lowing the vale ofS pes arn inoematon whsterer boat he out fan ‘gconing manieent Btn & and 8 ae and ob moma cepa “Solo exampe color andbardnets which eral compte her able) ates goad del more incapable wth one nate ha el and srTHe Mariemar aa everyday, Newtonian way turns out thet that requirement (which {is refered to inthe quantum-mechanical literawure a the principle ‘of correspondence) can be parlayed into a prescription for calcu lating the commutators of the quantum observable operators from ‘mathematical relations among the corresponding measurable prop- teties ofthe classial theory. [Now it happens that thie prescription implies thatthe momen- tum and the postion of a particle are incompatible observables. The commutator of p and x (p is the traditional symbol for the momentum of a particle, and = stands for postion) is 2.29 pal where is a number, a physical constant, called Planck’ constant, and iis the imaginary number Y=1, The important thing about (2.24), of course, is just that i isn zero. So, there will hea basis ofthe space of possible states of such particles consisting entirely of eigenvectors ofthe x operator, and there will a basis of that space consisting entitely of eigenvectors of the p operator and [since x and p are incompatible) those wo, bases won't consist of the same vectors. A state of some definite momentum will be a superposition of variows diferent states of definite position, and a state of some definite postion will be a superposition of various diferent staes of definite momentum (jst 48 happened with color and hardiness) Let's start 0 explore tis space. Ler's Took at the x basis (the basis 7 Acualy the loge reltionsip bene tat ne pestipson (which rate the Pinan Balas ofthe casa they 10 te commons of te “eatom deny andthe coespondence pice has long ben 2 mabe of “pute, The pescrigion ulimate jsifeaton is ha i sees to wod Whar aoe te consiteey ofa ht with he known airs of, 8% Sascala? Baba afer a do have gute de postions and gue dente ‘lcs be mare amc! Ad wl, afer latest moment ied by ‘he mus fhe prac hw tht work ice loc ae oe momen “ddd ty ma ange encras)of vlc aes range of momen “ed by So gna ean sso fren whe the ‘raretueancrsimy gl o tse hoe ase ately ai) Cn behave at hey alae do aod yet ily aconance with he a MATHEMATICAL FORMAL: as of postion eigenvectors}. Lat’ call the postion operator X- Con sider a particle which is confined (to Keep things simple for the moment) 10 one-dimensional coordinate space; a particle which isconstrained co move along a line. Let [X = 5) represent the state in which thar particle s located at the point 5. Then, in accordance ‘with principle (8) 22s) XK = 5)= SK = 5) Note thatthe possible eigenvalues of X (unlike those of color and hardness) wil frm a continuum extending from ~~ to += (since the points on a line, the possible locations of such a patil, are ‘continuous and infinitely extended), "Now, since (in accordance with the facts about Hermitian oper ators that were recited at the end of the last section) the various dhferenr eigenvectors of X must necessarily form a bass ofthe state space of this parle, and since X has an infinity of different eigenvalues, and since the eigenvectors [X= @) associated with those diferent eigenvalues must necessarily all be orthogonal 0 fone another, i follows thatthe state space of this particle must necessarily be infinite dimensional! And here, by the wa, a partic: larly dangerous confusion isto be scrupulously avoided. Thee are rsvo "spaces" coming into play here: the one-dimensional coordi nate space, which is the space of locations, the familar, ordinary, physical space in which the particle is fre to move around and the much more abstract vector space of states, which is here infinite: ‘dimensional and of which the locations (which constitute the entire «coordinate space) metely form a basis. The two shoulda’ get mixed up. “The fact tha the X eigenvectors form a basis ofthe state space also entails thar any vector whatever (ly), say) in thaeinfnite-d- ‘mensional space can be expanded {in accordance with (2.1) and (2.2) in terms of X eigenstates like this: (2.26) W)= aX = 5) + aX = 7) + anglX = 72.93) + where a, = (WiX = =)THE MATHEMATICAL FORMALISM ro Let me introduce a notation now which will serve precisely che same purpose for infinite dimensional vector spaces as the column- vector notation of equation (2.5) serves for finite dimensional ve tor spaces. Think of a, in (2.26) asa function of x: 227) a= (yk = 9) = vis) Just as che NV aumbers in (2.5), ina given bass, pick out a unique vector, the function w(x} {the infinite list of correspondences be ‘ween @ values and x values), in the X basis which i implicit here, serves to pick the unique vector |y) out ofthe infnite-dimensional space. Given a bass choice (which, a8 I said, is implicit here) a. function (a eave function, as i's called in the lieratue) caries precisely as mach information as does |y itself. As a matter of fac, le} constitutes a blueprint from which (ala (2.26) [y) ean be xplicily constructed, and vice ves Tt follows from (2.26), for example, tha the function yle) for the state [X= 5) isthe funetion with value 1 ar the point = = Sand 0 elsewhere and that the function y{x) for the stare WAX = 3) ~ WX = 7) is the function with value WE at the point x = 3 and the value ~12 a the poinex = 7, and 0 elsewhere. [Location probabilities can be read off from the wave Functions 00, It follows from (2.27) and (2.22) that ifthe wave function of 4 certain particle ata certain time is y(x), and if position mea- surement ds carried out at that time on that particle, then the probability that chat measurement will find thae particle to be located at the point x = x1, say, will be equal ro [yi (that is the square of che magnitude of y(x) atthe point x) ‘Moreover, any measurable property of particles (momentum, 9 llth» he bt overs. The way Ie been alin over he as few prageapis and heey IT bean thoughout hee of as book, except tape aca) is an ifthe psn space chan elon an poealy ‘ccnp form 2 crete t-Bu wht the pot ely fry ofcourse 3 “nau, Ade sms ike the omen (2.26 really ght be epson ‘ons of postion probes one scaly cleats uanr mechani ae relly probes ocean parc ing ou in eta fe ep sd ot Protas of ing hem sera paral pote THE MatHemaric 7 energy, whatever) tums out to be representable as an operator on the wave function (namely, as some prescription for taking one ‘wave function into another, rather than on the state vectors; and eigenfunetion-igenvalue-operator relations precisely analogous to ‘those stipulated in equation (2.14) and in principle (B) apply heres and there are rules for adding and multiplying wave functions {analogous to (24a) and (24) whereby the sums and products ‘of state vectors can be calculated; and even the equations of motion ofthe state vector can be recast as equations of motion forthe wave function. To make a long story short, anything whatever that can be said about the state vectors of particles can be translated into the language of wave functions. And tha’ pretty mach all there is to the quantum mechanics of single, structureless particles Systems Consising of More Thon @ Single Paticle We shall need to know something about the quantum mechanics of systems consisting of more than one particle too. A two-particle system will suffice as an example. Let's setup the state space of a system lke that. Here's how to start: Imagine a pair of particles, one of which (umber 1) i in the state y,) and the other of which (number 2) isin che sate [ys. The quantum state vector of a pait of particles Tike that is traditionally written down like this: (2.28) Iwahlvos or ke this 229) Wh wd These are taken to represent # vector in the state space of the ‘wo-particle system. Let's reason out some of the properties of such systems. The Bur none of that ic going to make aay ciference whatever to the sort of wens Twn tk shout beTHE MATHEMATICAL FORA ~ ry probability caleuls, to hegin with is going to get more compl «ated: we shall have to deal, now, with joint probabilities involving the outcomes of one sort of measurement carried Out On, sf particle 1 and the outcomes of another sort of measurement carried ‘ou, atthe same time, on particle 2. Suppose thatthe two particles described above don’ interact with one another. Then the familiar Jaws of composition for independent probabilities ought ro apply: The probabily tha the outcome of a measurement of property A fon particle I is A = a and the outcome of a measurement of property B on particle 2is B = b ought ro be equal tothe proba- biley ofthe former outcome times the probability ofthe latter one. ‘Those two probabilities, of couse, are ones that we know how 10 calculate from the one-particle theory. Now, if the ewo-paricle Probability calculations ate to proceed in accordance with (2.22), and ifthe laws of composition are to apply then is easy to show thatthe rule for multiplying vectors like (2.28) or (2.28) by one another is going to have to be (2.30) (YA, Wily wD = (ull) lve Lets go on.” Suppose that the vectors [yl) yl) constitute a basis ofthe state space of particle 1, and that jy). yA) const ue a basis of the sta space of particle 2 (ofcourse, we've just learned that each of those bases must i fact coms of a continuous infinity of vectors; but pretend, for the moment, to make things simpler, that that is’ so). Then (2.30) entails that 23) vA viv wi and j=! 10, Weve dig things kind af ackwar hereof core If we knew recy wha sot a recor space we exe dei with apd we ken pecly wat ocr of wet are represen by (2.2) sl (229) thon (2.30) nl py be ‘dred from ay (24D, alten the empin aw for pbabies would {low tom 250) a (222) Areas and hoes we doo wha srt scons (228) and (2.29) epeent Our ssn for Bing tat or came ‘he composton wo erie [30 om tha ether wh (223), a theo ‘se [2.0] 10 serie wha otf etre an ete aces we desing ah THE MATHEMATICA ao and shat entails that the dimensionality of the two-particle erate space will be equal to the dimensionality of the state space of particle 1 times the dimensionality of the state space of particle 2 ‘The dimensionality ofthe rwo- particle space, in other words, N “The entre set of N" vectors ofthe form 232) Why fori N will ll be orthogonal tone another, ad they will form «basis of the eworparticle space; and (as always) any linear combination of those basis veetors ofthe form 2.38) fy auivle vi * anivl wD + will be another vector another possible sae, in that space, Now something interesting comes up. Consider a state of the ‘wo-particle space of the form: 2) IQ) = rly, wi) + Wnty vl) le can be shown that the state JQ) cannot possibly be recast (no matter what bases we choose within the spaces of the two individ tual particles) inthe form [fg Thar is, states like (2.34) cannot possibly be decomposed into a well-defined state of particle 1 and A well-defined state of particle 2; states like that eanmor possibly be described by propositions of the form “the state of parce 1 is such-and-such” and “the state of particle 2 is such-and-such.” In states lke (2.34 (and this is just another way of saying the same ‘hing, no measurable property of particle 1 alone, and no measur able properey of particle 2 alone, has any definite value. The small est system t0 which any sate can be assigned here, che smallest system which can be assigned a definite value of any measurable property" isthe rworpartcle system. States ike that are called nnonseparable; and the phenomenon of nonseparabilty (ike those 1. Wen aig abou properties ie mats or hare of ous the ae spy among the dng chars of pss tsb corinse ce profes ke poston ot moment o elas ot ee.THE MATHEMATICAL FORMALISH 30 ‘of superposition and incompatibility is widely thought to be one fof the most profound differences there is between quantum me: chanics and the lasial picture of the world “Here's an example: Consider the state (235) [P= MaN! = Spx! = 7) 4 WTR! = oe = 11) which I've written down in the notation of (2.28) (X" and X* are the X operators of particles 1 and 2, expoctively). Neither the position of pacicle 1, nor the postion of particle 2, nor anything lsc about ether of ther separately has any definite vale here; but the difference in thee postions does have a defini value; tha is (236) (XQ) =2 Let me say something more about the calculations of probabilities of experimental results in the two-particle case, Forma (2.22) i, ‘of course, sil the cue; bur the application of (2.22) ro rwo-paricle systems will require some elaborating, There are ewo interesting Suppose, frst, that the state ofthe wo-partcl system isk), and that A" and B® (hich are complete observables of particles ! and 2, respectively) are measured (the example discussed just above ‘equation (2.30) ia case ike that in chat ease, the state &) happens to be separable). The probability thatthe outcomes of those exper- iments will be A = 4, and B® = bis (237) KAT = ay BP = bly Now suppose thatthe two-particle state is |) and that only At is measured. The probability thatthe outcome of that measurement will be A MIP + KA! = ay LP = Alb + 12 Psa woth saying expla hie junc hate wy on pins ore prise eran to pair calthe ay one woul ik one (hee Ia! =i BY =a) am cgaveton ofA wh geval ye of Bath Semler stndok A= Bh ogee 238) KA = ay 1 THe MATHEMATICAL FoRMausta a where 1? is some complete observable of particle 2, and where the sum eanges over all the eigenvalues fof L. Ian be any complete observable of particle 2 at all no matter which one we pick, the answer will come out the same, The intuition connected with (2.38) is something lke this: the probability that A! comes out to be ais ‘equal tothe sum of the probabilities ofall ofthe various different ‘ways in which it might come to pass that A" comes out t0 be ay ‘The principle of collapse for two-particle systems will need some claboratng too. Principle () is sil the rue, but we shall need ro say more precisely whae (E) means for wwo-particle systems. Sup pose, then, thatthe state of a certain two-particle system just prior to the time ts | and suppose that at 1; the observable A’ jof particle 1) is measured, and suppose that the outcome of that measurement is A' = a). Heres how to calculate the state of the ‘worparticle system just aftr fy (here, that is, i how ro calculate ‘what state the state of the two-particle system gets changed into by the A! measurement} Start with the state [D) expressed in terms of eigenstates of A' and L? (where L?is any complete observable of particle 2; no matter which one we pick, cis calculation will come fut the same) 239) ID) = dyfA = ay LE = 1) # dill = ay P= 8) + + ddA = aL In) + dal = ay, = 14 where 240) dy Then, throw away all ofthe terms in (2.39) other than the ones for which A? = a. Then, muleiply all the remaining d's (that is, the <4) by some number (the same number for all the dy $0 as not ‘alter their relative magnitudes), so as 10 make the remaining pare, ‘of 2.38) a vector of norm 1. And that new vectors the state vector ‘ofthis two-particle system just after the measurement, Here ace some examples. For one-particle systems (for which, of cours, there is 20 L? property) this generalized prescription for collapse will ental tha the effets of measurements ae just as theyTHE MATHEMATICAL FORMALISM 52 were described in principle (E). For two-particle systems in sepa ble states, this principle will entail that the measurement only affects the state of the measured particle. Here's how that works Suppose that the premeasurement state [D) is the separable state 241) |D)=jO'= uM = 2) suppose that A’ is measured, with the outcome A* = as. Then, let Lin (2.39) be ME (remember that we can choose any operator we like for Land (2.39) wil eake the form: (2.42) [Dye dla! ay, Mo =a + dla =a Mao + with 243) (a! = ajo! = uw) -we = a! = And so, following the instructions below (2.40), we end up with the postmeasurement (pm) state: Me 244) Dye = 1A The interesting cases, though, ae the nonseparable ones. In those cases (nike in the separable ones) the measurement brings about changes inthe quantum mechanical description of the smeasured parcicle as well. Suppose, for example, thatthe premeasurement 245) |D) = ARIAT = ay LE = by + DAT = 04, EF = bd where, once again, A; is eventually measured, withthe result A! = 4, Formula (2.45) has already been writen down in the form of (2.39), so we can proceed diveely to carrying out the instructions below’ (2.40). Once those instructions have been carried out, we end up with (2.46) Dow = [AY ay? = bo) MATHEMATICAL FORMAUSAL 53 (One more thing needs to be described before we can talc about the story in chapter 1. Ifthe values of M separate and independent Physical properties of a certain physical system need to be specified in order to uniquely pick out that system's physical state, ehen that system is said ro have M degrees of freedom. A single, structureless particle confined to a one-dimensional coordina space has 1 de sree of freedom; and a system consisting of two such properties (such as we have just now been considering) has 2. degrees of freedom; and a single particle free to move in a three-dimensional «coordinate space has 3: and a particle that has color hardness properties and is fee to move in a three-dimensional coordinate space has 4. The quanturm-mechanical treatment of systems with multiple degrees of frecorn is precisely analogons to that of mul tiple particle systems just described. The states of such systems, a5 for muliple-particle systems, are written down, degree of freedom hy degree of freedom, side by side asin (2.28) and (2.29), For example the state of a white electron whose eoordinace-space wave fanction is win) i writen (in the hardness basis of equation (2.17)) And everything I've said here about mukiple-partile systems ap- Plies straightforwardly wo multiple-degree-of-reedom systems to The Twwo-Path Experiments Now we're ready to retell the crucial stories of Chapter 1, The thd ‘worpath experiment (the one thats sa perplexing) is mapped ut carefully, with the help of a coordinate system, in figure 2.8. The ‘ims at which the various different stages ofthe experiment unfold ae indicated there too. ‘At time 1, when the particle is about to enter the apparatus itsTus MATHEMATICAL Fonmaushy u gure 28 (2.48) white, Xm, Y= 99) rua] =[ W]ixea, rey wal] vet} sve wahard)X = 2, ¥ = yi) ~ AaofOX = as, Y= 9 ——eEe lo » where, as usual I've written out che spin vectors inthe hardness basis. Here's how to caleulate what happens next. Consider this: If the state att, weren't 2.48) bat, rather, just a), and ifthe hardness box really is hardness box, then the state at time f; would be (2.49) fhandIX = xy Y= 99) Tue Manienan aanauiso 35 And if the stare at ¢, were just [b}, them the state att; would be 12.50) boFIX = a5 Y= 90) However, asthe state at fis in fact neither la) nor |b) but, rather, tla) ~ i), i follows from (2.49) and (2.50) and from the linearity ofthe quantum dynamics (which was spelled out i prin- ciple (D} that che state at 8 ceally $1) WithardX = 25) ¥ 99) = WEBOAIX = 5 = 90 This tae, by the way, involves nonseparable corelations beeween spin and coordinatespace properties ofthe electron: No spin prop: ety of the electron in this state (nether hardness nor color nor anything ese) nor any of its coordinatespace properties (position, ‘momentum, et.) has any definite value her, just as no property of wither particle 1 or particle 2is separately definite in the sate of ‘xquaton (2.34). The only properties which are definite in (2.51) involve combinations of spin-space and coordinate-space variables ofthe pacile. Well talk about those later on. Formula (2.51) represen a superposition of states, in one of Which the electron is traveling along the haed path and inthe other ‘of which the electon is traveling along. the soft path. Ic is of the state in formula (2.51) that we were compelled to condlade, in Chapter 1, tha it's ale that the electron tales the hard path, and false that it takes the sofe one, and false hati takes both, and false that it akes neither; and the problem was thar those four claims together amount toa contradiction. On the standard way of think ing, a8 Tve mentioned already, those hypotheses (hard path, soft path, both, neither) aren’ fale, they'ze meaningless, they're eate sorieal mistakes. Let's go on. The same sort of reasoning as led from (2.48) to (2.51) wl imply, starting from (2.51), tha the sate ofthe electron 2152) UsphandIX = 24, ¥ = ys) ~ WbOfOIK = a4 Y=)MaTHEManicat FORMALISM 56 and the same sor of reasoning applied in turn, to (2.52) wil imply ‘thatthe state at time fis 12.53) Wafhard)X = as, Y= 99) — WalsofO}X = #4, Y= 30 = \rthard) ~ fof)X = x, Y= 90) IwhitolX = 2s, Y= 90 A this point, the spin state and the coordinate-space state have become separable again. The position of the electron once again has a definite value now and that (as can be seen from the calew lation in (2.53)) renders its color definite too. So, the fac that & hard electron fed into this device will come out hard, and chat 2 soft eletron will come out (at that stm point) sof, coger with the linearity of the quantum dynamics, entails chat a white electron {ed into this device wll come out (ust as we found it did} white. ‘What if we were to stop the experiment inthe middle by mex suring the position of the electrom at, say, 1,? Then the superposition would go away; a collapse (much like the collapse described in equations (2.45}+(2.46)) would oscut, and the state just afer the measurement would be ether (2.54) [hank =, Y= 99) oF ofdIN = xy ¥= yD cach with a probability of 1/2 (in accordance with (222) and (2.38}), and the state att, would be (respectively) 288) fhanbX = x6 Y= yd oF PORN = ay Y= 9 ‘What if we put a wall into the soft path at y, 92 Then the state at te would be: 2.56) Wathard)X = as, ¥ = ya) ~ WapoMIX = x, ¥= Jn this instance, then, the state remains noaseparable between the spin and coordinate space properties at tu If measurement of the position a this electron were co be carried out att if say, we were Tae MATHEMATICAL FORMAUSAN o to fook and see whether the electron had emerged from the black box), the probability of finding i a (ys) would be 12, and if t were found there it would be hard, and its color were measured, it would be as likely to be found biack as white; and all tha, once again, is precisely in accord with the results of the experiments described in Chapter 1 The state described in (2.56), as Ijust mentioned, i nonseparable between spin and coordinate space, so it isn't associated with any Aefnite values for hardness or color or position of momentum oF anything lke that. Nonetheless, (2.56) is quantum state (that is its a vector inthe state space ofthis electron), so fn virtue of fact (5) about Hermitian operators) it mast be associated with definite values of some measurable properties, Let's find out what those properties are. Fist le’ simplify our ‘notation a bit. The electron in (2.56) has only two possible pos tions, s0 let’s replace the position operator (which has an inbinity fof possible eigenvalues, most of which ate irrelevant here) with something we'll call a'“where" operator {which has only two possible eigenvalues). “Where = +1" means X = xn Y= yu “whete = ~1" means X = xy, ¥ = ys. IF we represent the eigen vectors of where like this ? wemrero=[i) = 2 en [5] So, (2.56) can be rewrinen like this: > BL BL CL BL. ‘The subscripts to he column vectors above indicate what degrees of freedom those vectors refer to. Now let's define one more Her-Tae Maticwat 38 ov-lty i 2a) yes x i} = Fy] and ic follows from (2.4b) cha the square ofthe product of any column vector with (2.62) will beequal tothe square of the product ‘of tha same vector (whatever vector that} with (2.61) as) [A = hana) = THE MATHEMATICAL 59 So, a box which changes the state of any incoming electron into Times tha incoming state will be a toxalof nothing box, since ic will change none ofthe values, or any probabilities of values, ‘of any ofthe observables of any clectron which passes through i ‘needless sy, such a ox wil have no effect whatever on the sate fof an electron which pases outside of it Bur the effects of such a box om an electron which is in a superposition of passing through it and outside of ie may be quite a different matter. Suppose, for example, that such a box is inserted in the soft path of our troxpath device at (83s, 9). Then, if the initially inparelectcon was whit, the state att, will be, as above, (2.51), and the state a aftr the “passage” through the box} will be not (2.52) bt, rather, 12.63) Wtthard}X = x5, ¥ = 9 + tof = x, » (The sign of the second term is changed, relative to (2.52), by the passage through the box.) The same reasoning as led from (2.52) (2.53) will now imply thatthe color of the electron att iy with certainty, black Field Theory Theres just one more thing that it will turn out eo be convenient (2 good deal later on) to have mentioned here I has to do with adjusting the quantum-mechanical formalism s0.as make it consistent with special relativity Ie turns out that this adjustmene doesn’t require any change at all in principles (A}-(D). What needs to be changed isthe funda ‘mental ontology ofthe world. What you have to dois give up the ‘dea thatthe material world consists of particles (since it ruens out that a relativistic quantum theory of particles, a theory which satis (A}-{D), just ean be cooked up) and adopt the idea that ‘it consists of something else. Here's the general iden ‘What goes on in relativistic quantum theories is thar one in agines that there is an infinitely tiny physical system permanently located at every single mathematical point in the entirety of spacesTHe MATHEMATICAL FORMALISM 20 ‘ome imagines (that e) hat theres literally an finite array of such systems, one for each such point. And each one of those infinitely tiny systems i stipulated to be quancum-mechanical system. And ‘ach one of them is stipulated to interact in a particular way with cach of its neighbors.” And the complete array of them is called the fl ‘And i tens out that a relativistic quantum theory of the ied (2 ‘theory which satisfies (AJ4(D)) cam be cooked ups moreover, chat theory ean be cooked up in such a way as to guarantee that the familiar quantum-mechanical observables of the material world (ahar is: the observables of “particles”) can for the most part be reinterpeeted there as observables of the field. For example, state: ‘ments about the number of particles in a given region of space, or bout the kinds of particles in thar region, or about the physical states of the particles in that zegion, will all turn into statements bout the quantum states of the infinitely tiny fed systems in that region. And az a matte of fact, it can be shown that what the ‘nonrelativistic limit ofthis theory amounts to is precisely the non- felativistc quantum theory of particles which was outlined inthe section shove om coordinate space and which most of this book is sing to be about. 13. 0r yu cl par shi way Wha goes elvis quota heres ‘thi ope apne that every noe mathemati pin ia pace il t ‘uty mechan! sys, hat each oe ofthc pareaar ‘ty wah ts noe eo ee Nonlocality |A famous attempt to escape from the standard way of thinking about quantum mechanics was initiated in the 1930s by Einstein and Podolsky and Rosen, and had a surprising aftermath, in the siates, in the work of Boil, and that (Bell's work) is what this chapter will mostiy be about. Bot fist ler me describe the escape attempt itself. In 1935, Ein- stein, Podolsky, and Rosen (who have since then come to be known 2s “EPR") produced an argument, which was supposed to open the ‘way to that eseape, tha ifthe predictions of quantunn mechanics about the outcomes of experiments are correct) the quantum: ‘mechanical description of the world is necessarily incomplete Here's wha they meant by “completeness a description ofthe world is complete, for them, justin case nothing thats true about the work, nothing that's an “element ofthe eeality” of the world, set left our of that description, Of course, thar entail that if we want to find ovt whether or not a certain description ofthe wold i complete, we need first o find ‘our what all th elements ofthe celity of the world ae; and it turns fut (not surprisingly) that EPR had nothing whatever ro offer in the way of a general prescription for doing that. What they did Iwwhich is something much narrower, but which turns out ro be enough for their purposes) isto write down a merely sufficient «condition fora measurable property of a certain system ata certain moment to be an element of the reality of that system at chat ‘moment. The condition is that “if, without in any way disturbing «system, we ean predict with certainty (ie, with probability equal2 to unity) the value of a physical quantity, chen there exists an clement of reality corresponding co this physical quanti.” Las sce what this condition amounts to, Consider 9 question like this fa measurement of a certain particular observable (call it O) of a certain particular physical system (calli $) were o be carried out ata certain particular future time (callie T), what would the outcome be? Suppose thar there is method whereby I can put myself in position, prio co 7, to answer that question, with cer tainty, And suppose thatthe method whereby Tcan put myself in that position invalves ao physical disturbance of S whatsoever Then (according to EPR) there must now already be some matter of fact about what the outcome of a future O measurement on § would be; there must now already be some faet about 5 (since the facts abour § aren't going to get tampered wich from the outside in ‘the course of my putting myself na postion ro answer the question about the O measurement) in virtue of which cha future measure ‘ment wold come out in that particular 3! So, what EPR wane to argue (ance again) is that ifthe empirical predictions of quantum mechanics are correct, then there must be tlements of the relty ofthe world sehich have no corresponding slement in the quanturm-mechanical desertion ofthe worl. They ‘holt, meeeert ete to be cared oa four aking ck 1 precion ned not involve ay farther seracon with he etre sal Soe EPR reality condom ens net color mt at rset bea elmet of he ety ‘ofthis elas ado one ha abo prec wha the eanurvaecani ‘nesurment ofthe clo of ta elton wold Ref ah a meatremen were {ote cared ou) lwooldneed menor te cao tht ron (wold ee, Noniocau 3 want fo use quantum mechanics, sore of paradoxically, against itself. The argument goes something lke this: Consider a system consisting of rwo electtons. Electron 1 is located at position 1, and electron 2 is located at position 2, The spin-space state of these two electrons i the following (nonsepars ble) one: 1B.) [A)= blac white» ~ Vjwhie blake |A), like any quantam state, is necessarily an eigenstate of some complete observable ofthis pair of electrons. Call that observable (Ox, and suppose that Os) = |A). Now) has been weitten dover in equation (3.1) in the color bass, but it happens to be an extraoe inary mathematical fact about this particular state that (as the reader can explicitly verify for herself, by means of equation (2.21)) this state has precisely the same form iF it written down in the Inardness basis. That is 18.2) |A) = ‘4afhardh bof), ~ Wafsof hard And asa matter of fact it turns out tha [A)eetains precisely the same mathematica form if color (in equation (3.1)) or hardness (in ‘equation (3.2}} is teplaced by gleb oF by sera (which were defined in Chapter 2) or by any one ofthe continuous infinity of electron spin-space observables whatsoever Focus first on equation (3.1), Suppose that we were to carry out 1 measurement of the color of electron 1. The outeome of that ‘measurement wil be ether “black” of “white,” with equal probs- bility (that follows feom the rules for ealeulating probabilities for ‘to-partice systems) Moreover, quantum mechanics entails (and itis experimentally known to be true) that inthe event that the ‘outcome ofthat measurement is “black,” then the outcome of any subsequent measurement of the color of electron 2 will necessarily 2. Se equation 238bbe “white" and in the event cha the outcome of thar measurement fon particle 1 is “white,” then che outcome of any subsequent measurement of particle 2 will necessarily be “black” (all of that fellows from the collapse postulate for two-particle systems)” EPR assumed (and this i the omly assumption that enters into theie argoment other than the assumption that the predictions of {quantum mechanics about the esults of experiments are correct and the name of this other assumption is locality chat things could Jn principle be set up in such a way as eo guarantee that the measurement of the color of electron 1 produces no physical dis: turbance whatsoever in electron 2. Tha seemed almost self-evident “There seemed to be any number of ways co do i: you could, for example, separate the two electrons hy come immense distance (since quantum mechanics predicts that none of what’ been said here depends on how far apart the two electrons happen 0 bel, oF you could insegt an impeneteable wall betwen them or Build im Penetrable shields around them (ince quantum mechanics predicts that none of whars been said depends on what happens ro lie between oF around those two electrons), of you could set up any array of detectors you like in order to verify that no measurable signals ever pass from one of the eleccons to the other in the course of the experiment (since quantum mechanies predicts chat no such array, in such circamstances, whatever sorts of signals it may be designed co dec, will ever register anything).* So, returning to (3.1): whenever |A) obtains there is a means of predicting, with certainty, and (in principle if locality is true, and :F you ser things up sight) withour disturbing electron 2, what the ‘outcome of any subsequent measurement of the color of electron 4 Sein parca, quations (2.4) and 246) 4 Lerbsay lee more abot pesly what the lealyasrption amounts tothe amp sy tht Tc puch yo inthe no aes ny Bt to ‘Of cause, smthing I So wih my fs far fom whace your ose i an cae sci ater ar ich near where Yor pach Fou nthe noe oe "hag do wath ny Fst mig beso somebody ee op ou he nose, foramen amp a tha ny ar hve nye ee oa te the ea ph you the mane rectly thn cant ny hat phe ono he none. hd something wid ny Bt a fam where or one ‘he cue four eng paced athe pve, he fon tha mpi) ere mas eo 2 willbe. The way to do that i to measure the color of electron 1; Since t's known thatthe outcome of any measurement ofthe color of electron 2 will invariably be the opposite of the outcome of any ‘measurement ofthe color of eleetton 1. And so it follows from the reality condition that color must necessarily be an clement of the reality of electron 2, that there must necessarily bea mater of fct ahout what the value of the color of electron 2 4, whenever [A) obtains. Now focus on equation (3.2). Repeat precisely the same argu ment, wth coor replaced by hardness. follows that whenever) ‘obcains, there isa means of predicting, with certainty, and without listing electon 2, what the outcome of any subsequent mea- sarement of the hardness of electron 2 willbe (the way to da that is to measure the hardness of electron 1). And so it follows from the realty condition thatthe hardness of electron 2 is alo neces sarily an element ofthe reality ofthat paticle whenever [A) obtains, And so the standard way of thinking must simply be false, since there ca be circumstances (like [A)) in wehich there are simula neously matters of fact about the valves of both the color and the hardness of a single electron, even though those rwo observables are supposed to be incompatible. ‘ces be some cl sequence of eve at angsou poi in space nd at congo manens nine the propagation ofa spa, yl sete al te sro with break fom weer wa tht dah my iv you beng Pinched the nove. And of cose the capacity of an ach ence feet 0 ee he capaci that fy ft oem ou tobe mcd in sh oe) ll ‘ecratly depend on what sano physical condition anne pce tween ‘hy ead your noe i yf eum, depend on th sec of page al (raf aomavesbvorterso what hve you And tat woe vet ‘eof Beng een, nvr of big ng canes hone es ae ‘ew cates wow ft ae mew coe") mu col ee se aie fie tocol oni Ad a mae fat he spi hry of retny et ta the elie at which phys infunce canbe peopupted rough space by eat of Sh eguenn of neighboring evens cannot poubl) ace the vi of igh he atic others thre couldnt be any mater of objective lac south the Sequence of howe ever ad hay ofcourse ea nt hte er ae rth hich he eer an row ely conc oe ene) ein hig p Inach «way at gus that he meouremetof theslerof ecco That tele haere onthe oucme of the menace he ler of econ 26 Moreover, the formalism itself (quite apart from any particular way of thinking abovt i) must necessarily be incomplete, since there ate necessarily certain fats, certain elements of the physical realty ofthe world, which have no corresponding elements in the formalism, There are facts about the color and the hardness of clectron 2, for example, when iA) obsains, but there isn anything in che mathematical description ofthe state /), in this formalism, from which the values of the color or the hardness of electron 2 can be read off, ‘And since (A) eetains its mathematical form in every basis, the same argument can be schearsed in all of them, and so, when |A) ‘obtains, there must necessarily simultaneously be matters of fact bout the values of every spin-space observable of electron 2; and ‘0 the standard way of thinking must be (a it were) ininizely Els, and the formalism mast be infinitely incomplete And of course precisely the same arguments can he made about electron 1 If allthis is right, then whenever [A) obtains all ofthe continuous infinity of spin observables are simultaneously elements of the ‘zai of both of the electrons in question. And if that's so, then the statement that [A) obtains necessaily constitutes avery incom- plete description ofthe state ofa pair of electrons. The statement that |A) obtains mast be teve ofa gigantic collection of different “eve” states ofthe pair in some of which, say, eletron 1 is black and soft and scrad = ~1 (and so on) and electron 2 is white and hard and scrad = +1 (and so on), and i others of which electron Tis black and soft and scrad = +1 (and so on) and electron 2 is white and hard and serad = —1 (and s0 on), and #0 on, 'Nonetheles, the information that |4) obtains must certainly constrain the “teve” state of a pair of electrons in a number of ‘ways, since the outcomes of spin measurements on such pairs of clectrons are (afterall determined by what dheie “true” sates are, and since we're assuming thatthe quantur-meckanical predictions bout the statistics of the outcomes of such meamurements are Les sce what sorts of constraints arse. First ofall if|A) obtains lif tha is, the outcome of a measurement of Oy is +1), then the Nona: 7 ‘outcome of a measurement of any spin-space observable of electeon 1 will necessarily be the opposite of the outcome of any measure- ment ofthe same observable on electron 2. Whenever /A) obtains, then, the “tue” stare of the pai of electrons in question is con strained, with cersiny, to be one in which the value of every spin-space observable of electron 1 isthe opposite of the value of that same observable of electron 2 “There are statistical sorts of constraints, too. Those are a bit more complicated: Suppose that |) obtains, and suppose, for ex Ample, that we were to measure the color of electron 1 and chen the srad of electron 2, There will be four possible outcomes: black. (color = +1) and scrad =~ 1, white (color = —1) and scrad = +1, black and serad = +1, and white and serad = ~1. Consider the frst wo, in both of which the valve of the color of electron 1 {isthe opposite ofthe value ofthe srad of electron 2. Le’ calculate (in the way that we're instructed to by the quantum-mecharical formalism) the probability that the outcome of such an expertment would be either one of those, Initially |4) obtain. Thea a measure ment of the color of electron 1 is carried out, That measurement {in accordance with the probability rules and the collapse postulate for two-particle systems, which were both desenibed in Chapter 2) will change the quantum state of this pair of electrons to either black) white, or white black). with equal probabilities, In the first case (in the event, that is, that the outcome of the color measurement on electron 1 is “Black, the probability that the outcome of the serad measurement on electron 2 is 1 is (Gwhitelserad = —1)F = A. In the second case (inthe event, that is, that the outcome of the color measurement on electron 1 is “white"), che probability that the outcome of the scrad measure- ment on electron 2is +4 is |blackscrad = +1) = VA. Since the two possible outcomes ofthe color measurement of electron 1 are cqually probable, it follows char the overall probability that the ‘tcome ofthe color measurement of electron I isthe opporite of the outcome of the scrad measurement of electron 2, when [A) obiains, is Us ‘And 35 4 matter of fact, it turns out that if |A) obtains, and if either color or serad or gleb are measured on both electcons, and ifthe observable that gets measured on electron 1 isnt the SameNoniocanny 68 ‘observable as the one that gets measured on clecton 2, then the ‘guantum-mechanical probability thatthe outcome of the measure ‘ment on electron 1 is the opposite of the outcome ofthe measure ‘ment on electon 2 is ahways exactly VA ‘And that will amount to a constrain on the relative fequencies of various different “true” states of pats of |A)-type electrons, which runs as follows: Consider a large collection of pairs of clectrons, each of which (each pas, cha is) is known to have the property that Os = #1. Pick any two ofthe three observables color, Serad gle, Call one P and the other Q. One-quarter (statistically) (of che pairs of electrons in any such large collection will have to have the property that the vale of P for eletron 1 isthe opposite ofthe value of Q for electron 2, and the remaining ehree- quarters ‘of the pairs will ave ro have the property thatthe value of P for clectron 1 the same as the value of O for electron 2 ‘And now there comes the punch line) a well-defined question can be posed as to whether these two constraints (the deterministic ‘onitaint aboot the values of identical observables for the two ‘electrons, and the statistical constrain about the values of different observables for the two electrons) are mathematically consistent With one another, fe was Bell who frst clearly posed and answered ‘that question, tvency-nine years after the publication of the EPR argument (Bell, 1964); and w curns out that the answer to that ‘question is 90." And so the conclusion of the EPR argument is logically impos ble; and so cither locality must be false or the predictions of quan S. Her whys Comer to begin wth) callin of pio eators which snes the determine oneal ach of he asin ay calleton ie at, {heolor alc seo {Ha ob he opponte the est sale of decron arth wad valor elon That 0 ete opante of he send tle of Slaton andthe lcbsalae of earn I ar 0 be the opponte fe gc abe Sf clcrs 2 Nowy ono on he rene ca xn ey oe heel that those ne exalyce dilereor way an whch vals of color and and ead { posiy sip tom po of sesrne such hat eerie lt eesplimapeton of thse ih ons asaigmets rhe rent can stn ey scope wn wil now that every ag oe of the (which so sty eey ale one of te peo decom an) cllcton of pars of Nontocaury ‘um mechanis about the outcomes of spin measurements on |A)- states must be faie (since those are the only two assumptions on ‘which the argument depends}s and ithappens that those predictions are now experimentally known to be tue; snd so the assumption thatthe physical workings of the world are invariably local must {astonishingly} be false, Here's another way t0 tll the story Einstein, Podolsky, and Rosen noticed that cher was something ‘odd about the collapse postulate for two-particle systems, They nnoiced that it was nonlocal: if the two particles are initially in a ronseparabe state, then a measurement eaceed out on one of them can bring about changes, instaneancously, in the quantun-mechan. eal description ofthe other one, no matter how far apart those two particles may happen to be oc what might lie between them, Consider, for example, a pair of electrons, and suppose that |A) initially obtains and that a measurement of the color of elect 1 is carried out. The outcome of that measurement (as we've seen} will be either “black” or “white,” with equal probabilies. The collapse postulate for ewo-partile systems entails that as of the instant that that measurement is ovet, the state of eletton 2 will be either jrhite) inthe event thatthe outcome ofthe measurement ‘oneectron { was “black”) or [black) fn the event chat the outcome ‘of the measurement on electron 1 was “white”) leone whanvr which tise the determin sont ther ho the pop fy tha the colo alu of electron tthe ppt th ae of et ow eas he propery thatthe clr valu olsen one te oppo of ‘heverad valve of etn 2 athe hs the papery tha the ge alc of leon tthe ppt of he cd vale of eto 2 Bat note atm any alec fps felons whick nse the atti ensrim, the fraction ofthe pit which have the Br ne she pepe ‘deserted in thea paragraph ot Us adh rch othe pc wich hove he second ane a tore poper a ls tt be othe fon ‘he pis whch have hehe one of tise popes ae ab got he a ‘othe rater of heptane whch ht oy neo the thie popes to be than oval 09 ‘Ad 0 nocolltion af aro elec whatever in poly sty th heEPR suspected that this nonlocalty must merely be a disposable artfaet of tis particular mathemasical formalism, ofthis particular procedure for calculating the statistics of the outcomes of exper ‘ments; and that there must be other (as yet undiscovered) such procedures, which give ise co precisely che same statistical predic tions bue which are entirely foal ‘And it emerged thirty years later, in the work of Bell, that that suspicion was demonsteably wrong, Bell’ work is sometimes taken to amount to a proof that any attempt to eseape from the standard way of thinking, any attempt to be realistic about the values ofthe spin observables of a pait of clectrons for which |A) obtains, must necessarily turn our to be nonlocal, But (and this isthe point of telling the story this way) things are actually a good deal more serious than that. What Bel has given us is a proof that there is a a matter of facta genuine nonlocality in the actual workings of nature, however we attempe to describe it, period. That nonlocality is, to begin with, a feature ‘of quantum mechanics itself, and it earns out (via Bells theorem) that tis necessarily also a Teature of every possble manner of calculating (without or with superpositions) which produces the ‘ame statistical predictions as quantum mechanics does; and those predictions are now experimentally known to be correct. Lets be somewhat more precive about ust what sort of nonlocality {quantum mechanics exhibits First of all, when [A) obtains, the sats spin measurements on eleeton 2 depend nonlocally (as we've seen) fon the outcomes of spin measurements on electzon 1, and vice versa. Bur consider whether oF not the statistics ofthe outcomes of spin measurements oa electron 2, when |A) obtains, depend non locally on whether a spin measurement is first carried out on electron 1 (and vice versa). Let's gure it ut. Suppose that |A) obtain, and suppose to begin with, chata measurement of color is earied out on eleeton 2. Well it follows from equation (3.1), and from the standard quantum ‘mechanical cules for calculating the probabilities of measurement outcomes, that the outcome of that measurement is equally likely to be “black” of “white.” And now suppose that [A) obtains and is ofthe outcomes of that a measurement of the color of electron 1 is carried our, and there a measurement ofthe color of electron 2 is carried out. Well the measurement ofthe color of electron 1 is now (Vor precisely the same reasons) equally likely 0 come out “black” or “white.” In the event that it comes out “black” it follows from the collapse postulate for eworparcicle systems that the subsequent meastre- ment ofthe color of electron 2 will come out “white,” and i the event that it comes our “white” it follows from the collapse post late for tworpartile systems thac the subsequent measurement of electron 2 will come oat “Black.” So, when [A) obtains, the out come of a measurement of the color of eletron 2 is equally likely to be “black™ or “white” whether of not a measurement of the color of eletrom 1 is carted out fist. Now suppose that |A) obtains and that a measurement of the hardness of electron 1 is carted out, and then @ measurement of the color of electron 2iscaried out. I follows from equation (3.2) and fom he probability rule, thae the outcome ofthe hardness ‘measurement on electron Tis equally likely to be “hard” of “sof.” Now, in the event that the outcome of hat fist measurement is “sof.” i follows from the collapse posculate and from the proba bility rales thatthe outeome of the second measutement (the color measurement on electron 2} 6 equally likely to be “black” or “white.” And the same thing is tue in the event that the outcome ‘of the fist measurement is “hard.” And s0 here's where we are so far: when [A) obtains, the outcome of a measurement ofthe color ‘of electron 2 is equally likely ro be “black” or “white,” whether a measurement of the color of electron 1 is carried out Fist, or a measurement of the hardness of electron I i eatried out fist, or no measurement on electron 1 is carried out ist. And everything that's been said here is also teue if “eoloe” and hardness” are interchanged, or if either or both of them are replaced by “pleb" or by "scrad” or by any of the other spin ‘observables; and everything tha's beea said here i of course also ‘eve if “electron 1” i exchanged with “electron 2.” You can extend this kind of argument to other sorts of observables and to other sort of physical systems, too, And if you follow this sor of thing out a8 far a8 you can, here’ where you end up: Take any composite physical system S. Divide72 ic up, any way you like, into ra (possibly sill composite) subsys teats. Call those two subsystems 5; and s. Take any quantum state whatsoever, |Q), of S. Choose any observable O, of s, and any ‘observable Or of. I's possible co prove the following completly general theorem: Whenever |Q) obtains, the probabilities of the ‘anious possible outcomes of a measurement of O, don't depend at all‘on whether or nota measurement of O; is caried out fst. So, there are (for the mb time) nonlocal influences ia nacure (if the relevant predictions of quantum mechanics are eight; and they are right); bur those influenees ace invariably ofa partiularly subsle kind, The outcomes of measurements do sometimes depend non: Iocally on the outcomes of other, distant, measurements; bur the ‘outcomes of measurements invariably do nor depend nonlocally on whether any other, distant, measurements get carried out! Let me put it another way (and his is really the punch line of this little section). The subilety ofthese influences is such tha even though they surely exist, even chough the statistics ofthe outcomes of experiments can't be understood without them) they cannot possibly be exploited to transmit a detctable signa, they carvot possibly be mide to carry information, nonlocaly, between any ‘two distant points. The problem i that it just won't work ro encode the information you want to transmit in a decision ro make a measurement or not 10 make doe, oF in a decision about what to make a measurement of since (a8 we've just seen) no such decisions can ever have detectable nonlocal effects; and there just isn't any 1ugyto encode the information you want ro transmit in the outcome ‘of ap appropriate sort of experiment {in, say, the outcome of a ‘measurement of the color of electron 1, when |A) obtains), since, fas matter of principle, the outcomes of such measurements are ‘necessarily entirely beyond our control; and all ofthe rest of quae tum mechanics, the parts that deal with the reactions of physical systems to everything other than measurements thats: the dynam: ‘eal equations of motion), are, 50 far as we know, absolutly local from start ta finish “There will be a good deal mote to say ofall this later on, we de The Measurement Problem "Now I want 0 begin to worry in earnest about whether of not the «dynamics says the same thing asthe postulae of collapse says about what happens to the state vector of a physical system when the system gets measured. Here's what looked worrisome back in ‘Chapter 2: the dynamics (which is supposed to be about how the state vectors of physical systems evolve it genera) is fully deter ‘minstic, bt the collapse postulace (swhich i supposed to be about how the state veetor ofa system evolves when i comes in contact with a measuring device) isnt; and so it isn’t clear precisely how the two can be consistent Let's figure out what the dynamics says about what happens when things get measured, Suppose that everything inthe world always evolves in accoe dance with the dynamical equations of motion. And suppose that we have a device (which operates in accordance with those equa tions, just like everything ese does) for measuring the hardness of anelecton; and suppose tht that deviee works lke this: The device has a dial on the fron, with a pointer; and the pointer has three possible postion. In the first position the pointer points to the ‘word “ready.” and in the second positon it points to the word hard,” and inthe thd it points tothe word “sof,” Electrons are fed into one side ofthe device and come out the other and in the ‘colise of passing through (if the device is set up right, with the Pointer incally in ts “ready” postion) they get their hardnesses measured, and the outcomes of those measurements get recorded inthe final postion of che pointer (figure 4.1, ”Tue Measurement Posie 74 igure 43 If the device is set up ight, and ifthe dynamics is always crue, then (Go put all this another way) the dynamical equations of ‘motion ental thae it behaves ike this: (4) freadylhard), > [*hard) hard, and 4.2) ready), > [sof sof ‘That is: if the device (whose state vector is labeled with subscript 's initially in the ready state, and if an electron (whose state wector is labeled with subscript) that is ard gets fed chrough i, then the device ends up inthe state whetein che pointer is pointing at hard"; and if the device is initially inthe ready sate, and ifa ‘soft electon gets fed through it, then the device ends up i the stare ‘wherein che pointer is pointing at “soft. Thar’s what it means for 4 measuring device for hardness to bea good one and to beset up right ‘Now (sill supposing thatthe dynamics is always tue), consider what happens if this device the one for measuring hardness) is set Up tight and isin its ready state, and a black electron is fed into it-Tetuens out that (4.1) and (4.2), and the face chat the dynamical {equations of motion are invariably linea, suffice by themselves to THE MEAUREMENT PROBLEM figure that out: The inival state of the electron and the measuring device is (43) eady)black, = ready WaThardh, + Vzhof) = Uteeady)qiaed), + Vitreadylof Which is precisely WA times the initial state in (4.1) plas V4 times the intial tate in (42). So, since (by hypothesis) the dynamical ‘equations of motion entail chat [esdy),lhard), evolves asin (4.1) and that |ready)lbo), evolves asin (4.2 that is: since this device is set upright, and since i's 2 good measuring device for hardness), is follows from the linearity of those equations thatthe state in (4.3), when the measuring device gets switched on, will necessarily valve into (44) rlhard") hard), + Wn)"sof of. That's how things end up, with certainty, acording tothe dh ‘And the way things end up according tothe postulate of collapse lovhen you stare with (4.3)) 6 M45} either |*hard”)lhara, (ith probability 2) or [sofa (with probability 12) smearing devices for harder be the ote devel re man be eee Combed contraptions especially you look ahem ot eel st of hs onsen ston) snd mut ave exeeely complicated qari of meio te Soltion of wich mst bean exert eompcsed mater Al of tha ‘hat simplifies hing bere the fc that however complexed tore ests tmay be, 1) a [2 ar suey elns of hm ance he Sntepon nee ‘hal mth by hypotesi a whee ley be, food measuring dee fee the hard of am elton) ad eer he eget te se std thon wo Tact a crap by temas nue tha if 2 ack een ed tothe dei he he equstins wl eal th things wil cd wp nthe teTHe MEASUREMENT And the trouble is that (44) and (4.8) ate measurably different The tate desrited in (4.5) isthe one that's right itis (asa matter of empirical fact) how things do end up when you sare with (4.3) The state described in (4.4) is not how things end upy? (4.4) is something very strange. I°8 a superposition of one state in which the pointer is pointing at “hard” and another state in which the pointer is pointing at “soft it’ a stae in which (onthe standard way of thinking] there is no mater of fact about where the pointer is poining* Lets make this somewhat sharper. Suppose that a human observer tnters the picture, and looks atthe measuring instrament (when the measurement is all done) and sees where the pointer is pointing. Lees figure out what the dynamics wil say about tha. Suppose, then {just as we did before, that terally every physical system in the wotld (and this now includes human beings; and it 2. esha this ought 0 be expand on Kt. The pin tha ee are i sscondane withthe pone gut nc that ee oui Chae SJemeanymensuaie popes of hr ste} whey can mpage 1e'Spmenalydingaced fom cer of he te saa mae ff fom any ae ste hanes Tete wl Bes pod deat ay, er, Shout precy hat thse proprio are they compete ome tod the ‘rear legal be ceemel kh, hate mnt far the Tiina any sesso roth cntensonl wim abou these mats bathe wil be mah more oy abo eae om " Somerhngsugt ta be mentioned poseng here, something tha wil em ‘Whar weve jn dhcveed tha the i cain andament ec ofthe eying of snare a he emergence of sme Jee nme Sliberensemeny the emergence of ere mater fac about py wht he oh renueent whic ot peded by he dal quate ic conie anche fs f the crying ou ofa mesavree one which we finden the coun of Suto at ards nd clr Chae ‘The cong ou a menuement epee ev Inensed system wach ve compte th de orale hae gets ese Tun ot har the dynamin equations of ton dpe tat THe MeAUREMENT PROBE includes she brains of human beings) always evolves in accordance ‘withthe dynamical equations of morion; and suppose that a black lectron is fed through a measuring device for hardness thats set up right and that starts out in its ceady state (so thatthe state of the electron and the device is now the one in (4.4); and suppose that somebody named Martha comes along and looks a the device, and suppose that Martha isa competent abserver ofthe postions of poiners Being 2 “competent observer” is something ike being a measur- ing device thav's set up rights Whar ie means for Martha to be a ‘competent observer ofthe position of a pointer is that whenever Martha looks ata pointer that’s pointing to “hard,” se eventally comes to believe thatthe pointer is pointing to "hard": and that whenever Martha looks ata pointer that's pointing {0 “sof,” she eventually comes to believe chat the pointer is pointing to “sof {and so on, in whatever direction the pointer may be pointing What it means {to put it somewhat more precisely) is that the dynamical equations of motion entail that Martha (who isa phys ieal system, subject tothe physical laws) behaves lke this 46 Jready.|rendyie > [ready”,eeady). and lready)thard™pe > Phar) hare and Iready),|*50f") eof "s0f”) In hese expressions, rend. that physical tate of Martha's bain in which she i alert and in which she is intent on looking at the Look or example atthe evltion fo (4.3 10 (44 Equation (4.3) an signs fhe color a thecton osha show te meses (4 (ich the tae flowing the ieracton of the ton wth 9 Bod ‘nesriag denice for he hd, earn othe eguason fein ie (4) is rates ot aterm of erties the colo the atone he reser can now ely dl stn ou ob #speestion of ea sae (with ‘ul confers ane of whch he len lack ad othe eke which the etn fheTHe MEASUREMENT PROBLEM 78 pointer and finding out where ifs pointing; ready"). thar phys ical state of Martha's brain in which she beiewes that the pointer js pointing to the word “ready” on the dia; [*hard”), is that physical state of Martha's beain in which she believes that the pointer is pointing ta the word “hard” on the dial and I"sot”), that physical sate of Martha's brain in which she believes that the pointer is pointing to the word *sofe" on the dial’ Lets get back to the story. The stare of the electron and the measuring device (a the point where we left off isthe strange one in (4). And now in comes Martha, and Martha is a competent ‘observer ofthe postion of the pointer, and Martha i in her ready ‘tate, and Martha looks a the device, Ie follows from the linearity fof the dynamical equations of motion (i those equations are right), nd from what it means to be a competent observer ofthe position ff the pointer, thatthe state when Martha's done is with certainty going to be soe", 2) Watrbare” rd" hard), + Ey sot Thar’s what the dynamics entails, And of course what the postulate of collapse entail is that when Martha's all done, then 8) either [Phard)J*hard™)ahacd, (with probability 12) or [Paoft)/*sofof), (with probability 72) is going to obtain. ‘And (4.7) and (4.8) are empirically different. The state described in (4.8) isthe one thats right; (4.7) is unspeakably strange. The state described in (4.7) i at odds with what we know of ourselves S.A hry ede sang sha hi a barely oneal escrpion of| Maro’ ban an at his a aud overampiel accu ofthe Ways it tic mena aes ve socal suppose! to pervene on alo sres fu all The trn ot ptt make ay direc ea hi age of ge ‘Neca il ede whenever we win, 0 whuteet extent we wan They THe MEAUREMENT PROBLEM 79 by direct trospection. I's a superposition of one state in which ‘Martha thinks tha the poiner is pointing to "hard" and another ‘tae in which Martha thinks that the pointe s pointing to “sof; isa state in which there is no matter of fact about whether or not Martha thinks the pointer is pointing in any particular direction ‘And so things are taming out badly. The dynamics and the posts late of collapse ate flatly in contradiction with one another (just as we had feared they might be); and the postulate of collapse seems tobe right about what happens when we make measurements, and the dynamics seems to be bizarrely wrong about what happens when we make measurements; and yet the dynamics seems to be right about what happens whenever we aren't making measate ments; and so the whole thing i very confusing and the problem ‘of what to do about allthis has come to be called “the problem of ‘We shall be thinking about that fr the rest af this book: iui putea i uo inal se thames onde pene soci ayo Tiles ta wes aap we en Chop 1) gh oy ‘ara eles tthe per pg “hrs dt {Se Marks se ep eng oo da ih a ‘Sst thot tbo tet mg al nd ‘say that she bas neither of thene beliefs) The Collapse of the Wave Function The Idee of the Collapse The measurement problem was frst put in its sharpest possible form in the 1930s, by John von Neumann, in an extraordinary book called Mathematical Foundations of Onantunt Mechanics (von Neumann, 1955). I looked to von Newmann as though the ‘only thing that could possibly be done about the measurement problem was to bite the bullet, and admit that the dynamics is ‘simply wrong about what happens when measurements occu, and honetheless right about everything cst. And so what he concluded ‘was that there must be rwo fundamental lass abour how the states ff quantum-mechanical systems evolve: |. When no measurements ae going on, the states ofall physical systems invariably evolve in accordance with the dynamical ‘equations of motion I When there are measurements going on, the states of the measured systems evolve in accordance with the postulate of collapse, notin accordance with the dynamical equations of Bus this clearly won’ do, Here's the trouble: What chese laws actually amount to (tha is: what they actally say) will depend on the precise meaning ofthe word measurement (because these 10 laws entail that which one of them is being obeyed at any given moment depends on whether or gor a “measurement” is being carried out 3¢ that moment). And it happens that the word me oy surement simply doesn't have any absolutely precise mesning in ‘ordinary languages and it happens (moreover) that von Neumann didn't make any attempr to cook up a meaning for it either, ‘And so those laws, as vor Neumann wrote them dove, simpy don't determine exactly how the world behaves (which isto sy: they don' really amount to prospective fundamental “ives” at ll) And there has consequently been along tradition of atempts to figure out how to write them down in such a way that they do, Here's where things stand: Suppose tha a certain system i initally in an eigenstate of observable A, and that a measurement of ob: servable B is carried out on that system, and that A and # ate incompatible with one another. What we know with absolue cer tainty, by put intcospection, is that by the time thae messurement is all done, and a seatient observer has looked atthe measoring device and formed a conscious impression of how that device presently appears and what it presently indicates, then some wave function must already have violated the dynamical equations of ‘motion and collapsed. What we need todo isto figure ou precisely ‘when tha collapse occurs Lats uy to guess Pechaps the collapse always occurs precisely atthe last posible moment: perhaps (that is) it always occurs precisely atthe level of consciousness,’ and pethaps, moreover, consciousness ie always the ‘agent tha brings i abou Pat off che temptation to dismiss this as nonsense just foe long enough to see what it amounts to. (On this proposal (whichis due to Wigner, 1961), the correct laws of the evolution ofthe states of physical systems look something like this: All physical objects almost always evolve in strict accor dance with the dynamical equations of motion. But every now and ‘then, in the course of some such dynamical evolutions {in the course 1 Thiet 2 wy of saying thar hee’ anything sory abo te collate 16s ust away of sping a peeiy wt po he cole whch i pl owes ces, a way of say precy what href procare pepeTHe COUAPSE OF THE Wave FUNCTION ‘of measurements, for example), the brain of 2 sentient being may center a state wherein (as we've sen) staes connected with various different conscious experiences are superposed: and at such mo- nents, the mind connected with that rain (as it were) opens its Inner eye, and gazes on that brain, and that causes the entice system (brain, measuring instrument, measured system, everything) t0 col lapse, with the usual quantum-mechanical probabilities, onto one for another of those states! and then the eye closes, and everything proceeds again in accordance with the dynamical equations of ‘motion until the next such superposition arises, and then that mind's eye opens up again, and So on. “This proposal entails that there are gwo fundamentally diferent sors of physical systems inthe world A. Purely physical systems (that is: systems which dont contain sentient observers), These systems, so long as they remain ‘isolated from outside influences, always evolve in accordance with the dynamical equations of motion, B. Conscious ystems (that is: systems which do contain sentient observers). These systems evolve in accordance with the more ‘complicated rales described above. Bur the trouble here is pretty obvious too: How the physical state fof a certain sytem evolves (on tis proposal} depends on whether ‘9 not that system is conscious and 80 in order to know precisely how things physically behave, we need to know precisely what is 2. Hage an example An oben ie 06 mentee of he Rado of Hak ton Bret (hon the mesurng deve ae de ork ond ‘he obser oe athe cee) igs get bei the item esto 4.7), tthe he mid’ ee he cure opens, ats upon herbs ances = ‘Sle wih seal potable oot ee he Bo he cond of the terme 1 ele oy amon fal wn ada ineracove mand body lon Wage thot tah sr f una aes oot Grey) wo Be 2 erry coseuonc pce be thapht ha thee wae a lyse fb 0 be oe inthe wold he bel clap wave unions) which coal ely be done by noe purely physical tng THe COuArst oF THE Wave Func 88 ‘conscious and what isnt. What this “theory” predicts (that is: what “theory” it 1s) will hinge on the precise meaning of the word a fevese white white) ad sha ashing loko thi wy oti a8 fete oe wl in wich fects hand andthe amie in whch 2 Ontbeandard quantum mecha formalin, ae all, the choi of se cot be vestor inch o wake water down as no pyc sence wha. THe ynanes ay INSEE 15 ment of a hard electron, che “probability” thar that measurement ‘will come out white is 12, The trouble is that that sort of» mea surement (on this way of talking) will with certainty give rise co ‘vo worlds in one of which theres an b who ses thatthe outcome (of the measurement is white, and in he other of which there's an Jb who sees that the outcome ofthe measurement is black; and there isn’ going eo be any mater of fact about which one of those two worlds isthe real one, or about which one of those two B's is the original. ‘And there are myriad other difcuties with talking the many worlds talk too (see, for example, Bareett, 1992}, but I guess we need not Fehearse any more of thern here. There are ways of making many-sworlds alk sound les vulgar {which is to says there ace ways of making it sound less literal), But they don’t get at what the real problems are Sometimes i gets proposed [for example) that there is exactly fone physical world bur that (when states ike (6.1) obran) there are two incompatible stories about that world, or mayhe about how b +325 that world, which are both somehow simultaneously true.) Te seems 10 me that thas eeally hard to understand. But one of the things that's obvious abour i is chat it runs into exactly the same sort of puzzle about what probabilities mean as the many’ ‘worlds talk does. Suppose, for example, that an observer named b carries outa measurement (us lke the one we talked about above) ofthe color ofa hard electron, Tey to figure out what it might mean to say of an experiment like that {if you try total like this} that the probability tha its outcome will be black is Ya. The trouble is thar this sort of tall i going to ental, with certainty, that there are 1ivo stories about what happens in an experiment lke that; and there isnt going to be any mater of fact about which ome of those 3. The mont intringatenge know ofa aking ke sat it Micha Lockwood ia Mind, Br ond he Quam Lockwood, 1989) Lockwood ts harder hes anphey ale dacs ih the publ ection of ackwou co league David Desc, whose eso pnts numberof cai poms Mind Bray andthe Quanta) to thnk abo what mean ha otk sot ‘sha i ie} or thereto Be mee ane te tory, when sate He (6.1),Tue Dynamics ay ITSEIF 16 stories isthe true one, and there ise going to be any matter of face labout which one of those stories isthe one that’s about the orig- inal b I think it tums out to be a good deal more interesting to read Evert in a rather different way. ‘Suppose that there i only one world, and suppose that there is ‘only one fll story about char world eha's ere, and suppose that the linear quantum-mechanical equations of morion are the true and complete equations of motion of the world, and suppose that the standard way of thinking about whae is means to be in a superposition isthe eight way of thinking about what it means to bein a superposition, and consider the question of what it would fel like to bein astae like the one in (6.1) (that is: the question fof what it would fee ike to be the experimenter in a state like the fone in (6.1) "That question wasn't confronted in Chapter 4. There didn't seem to be much ofa point (back then} in confronting it. What seemed important was just that whatever i might feel like co be ina state like the one in (6.1, it cereinly would mot fel like what we fel like when we're al done measuring the color ofa hard electron. ‘Bt (since it sur out noe to he easy fo cook up 2 good-Jooking, theory ofthe collapse, and since it urns out that no theory of the ‘collapse whatsoever is going to beable zo preclude the development ff stares like the one in (6.1) in people who undergo the kind of ‘surgery described at the end of Chapter 5, and since there aren't any normal experimental reasons for believing that there are Any such things as collapses) things are diferent now. Here's a way t0 get seared: ‘Suppose tha the linear quantum-mechanical equations of mo- tion were invariably trae and (consequently) that observers ike the fone described above feesuenty did end up, a the conclusions of color measurements, in states like the one in (6.1). Let's sce if we ean igre out what chose equations would en {Thar wha seremgortne was ast that i cole estased fy cans ofthe spam on pe 112 tha as marr te ama expsmenes do Sad wp tes hes, a the coco hows of esse. about how an observer like thar, ina state ike the one in (6.1) ‘would respond to questions about how she feels (that is: about ‘har her meatal state is. Maybe that wil tells something ‘The most obvious question toask is: "Whats your present belie bout the color ofthe electron?” But chat question turns out not to be of mach wse here. Heres why: Suppose hat the observer in {question {the one thar's now i the sate in (6.1) gives honest responses to such questions; suppose, that that when her Brain states believes e black) she invariably responds ro such a question by saying the woed “black,” and when her brain sat is believes e white) she invariably responds to such a question by saying the word “whic.” The problem is tha precisely the same lineasty of ‘the equations of motion which brought aboot the superposition of alifferent brain states inthe state in (6.1) i the fist place will ow ental that if we were to addese this sor of a question to this sort ‘of an observer, when (6.1) obtains, then the state ofthe world after she responds to the question will bea superposition of one in which she says “black” and another in which she says “white; and of ‘course it won't be any easier o interpreta “response” ike that than it-was to interpret the superposition of brain states in (6.1) that thar response was intended to he a description oft Bar there are other sorts of questions that turn out to be more informative Note, co hegin with, that it follows from the linearity of the ‘operators that represent observables of quancum-mechanical sy tems (the sort of linearity that was defined in equation (2.8) chat if any observable O of any quantum-mechanical sytem S has some particular determinate value in the tae |4). and if © also has chat same determinate value in some other stae |B), then O will neces- sarily also have precisely that same determinate value in any linear superposition of those «wo states! 5. Thats an ently commonsenicl ma Sor obese tobehiee if you ik ietrogh Suppo, for eal ht thre sare whic inn ouperpsion ‘tbe laste nee ge a a inthe ee alo etn be What he linen ofthe observa ot «parle that gg to etal or ey ane st the things ha 1 going to et tat ht pres agent “nae "the pate snore ine bow eal?” with eigen yes”ie Let's apply that to he superposition of states in (6.1) Suppose that we were to say thie to hs “Dont tell me whether you believe the election to be black or you believe it to be white, bor cell me merely whether oF not one of those two isthe cas; tll sme (in other words) merely whether or not you now have any parccular definite belief (not uncertain and not confused and not ‘vague and not superposed) about the value of the color of this electron.” 'Nows if (when we ask b that) the state believes ¢ blacks x [Pblack") black), obtains, and if is indeed an honest and compe tent eeporer of hee mental states, then she will presumably answer, Yes, Ihave some definite belief at present, one of those rwo isthe ase"; and of course she will answer in precisely the same way in the event thar [believes e white)" white”) white), obrains. ‘And so responding to this particular question in this particulae way (by saying “yes") is an observable propery of b in both of those states, and consequently (and ths i che punch lin) ie will aleo be an observable property of her in any superpesitio of those two brain states, and consequently (in pacticula) ic will be an ‘observable property of her in (6.1) That's odd. Look what we've found out: On the one hand, the dynamical equations of motion predict that is ong to end up, tthe conclusion of a measurement like the one we've been talking bout n the state in (6.1), and not in either one of the bea states ‘associated with any definite particular belief about the color of the electron; on the other hand, we have just now discovered tha those same equations also predice that when a state like (6.1) obtains, b ‘is necessarily going to be convinced (or at any rate she is necessaily going to report that she does havea definite particular belief about the color of the eletron, And so when a stae ike (6.1) obtains, b 's apparently going to he radically deceived even about what ber own occurrent mental sate is. ‘And so it turns out that there was a hell of a for too much being, taken for granted wen we gor convinced (back in Chapter 4) that there i some particular point inthe course ofthe sort of measure- meat we've been talking about by which a collapse of the wave function must necessarily already have taken place, some particular Dynamics ay ne point (that i) at which the dynamical equations of motion together with the standard way of thinking about what it means to be in 2 Superpesition somehow flatly contradics what we unmistakably know to be true of our own mental ives Lets go on, Suppose that carries our a measurement ofthe color ‘of a hard of electron with a color measuring device called ml, and suppose thar when that's done (that is: when a. state like (6.1) ‘obtans) carries out a second measuremant of the color of that ‘lectron, with a second color measuring device called m2, When thar’ all done, the state of band of the two measuring devices and the electron (ifthe measuring devices are good, and if bis compe ‘ent, and if everything evolves in accordance withthe linear dynam eal equations of motion is going to look like this (52) \erbelieves outcome of fst measurement is “black” and believes outcome of second measurement i¢ “black")|*back"" black”) lack, 4 eleres outcome of fst measurement is “white™ tnd belicves outcome of second measurement is white" white") a*hite”Pwhite)) And suppose that at that poi (when (6.2) obtains) we were t0 say toh: “Don tll me what the outcomes of ether of those two color measurements wer; just tll me whether or not you now believe that those two measurements both had definite outcomes, and ‘whether or not chose two outcomes were the same. Tewill follow from the same sors of arguments as we gave above that Bs response to a question ke that (even though, as a matter of fact, on the standard way of thinking, neither of those expe ‘ments had any dente outcome) will nccesarily be: "Yes, they both had definite outcomes, and both of those outcomes were the same.” ‘And it wil also follow from the same sorts of arguments that if wo observers were both to carry out measurements ofthe color of some particular initially hard electron, and if they were subse- ‘quently eo talk to one another about the outcomes oftheir respec tive experiments {i they were both, tha i, to ebeck xp on one25 another), then both of those observers will eport falsely, that the ‘other observer has reported some definite particular outcome of her feasurement, and both of them will report that that reported ‘outcome is completely in agreement with her own, [evs make up a name forall that. Let’ say that when a state ike (6.1) obtains, shen (eventhough there isnt any matter of fact about what the color ofthe electron i, and even though there ist even ‘any matter of fact about what b's belief about the color of the electron is} what the dynamics entails is that “effectively knows” what the color of the electron is Let’sgo on some more. Suppose that his confronted with an finite collection of electrons, all of which are initially hard, and hat b undertakes to measute the color of each one of those dleetrons. Before those measurements star, the state of b and of those electrons (whose names are 1, 2,» ) and afb’ color measuring devices (whose names ae, respectively, mr Dy. (63) readyhlreadyfhardhiready)lhardyreadybahardy ‘Once the measurement ofthe color of electron 1 isdone the tate believes 1 black)black”)a lack), + [heieves 1 white /*hite” sara shied) % feady ard ead [And once the measurement ofthe color of electzon 2 is done, the (6.5) Wabeleves 1 black and 2 black)*black")a: I black*),ablack) black 4 lbelicves 1 black and 2 whit))*lack")as FFwhite" [black jvhices) + (leleves 1 white and 2 black){"white™ir PPblackj white) fleck,) 6 ay Inseir 121 _ + letieves 1 white and 2 whiteh/*white”)a !white”aelwhite white) % lreadybefhand, And s0.0n. The numberof separate mathematical terms in the state vector of the world (if you write ie out in the sort of buss that's used here) will increase geometrically (ike the numbers of the branches in the diagram in Bgute 6.1, as you work your way up) asthe number of color measurements increases, [Now suppose that once the frst N of those measurements are ‘complete we say this toh: "Don't tll me what the color of dletron or electon 2 of any particular one ofthe fist N electrons eurned ‘out ro be; tll me merely whether or not you believe that each one of those electrons now has a definite color, and tell me aso (if the ansier to that Fist question is yes) what fraction of those first N electrons tamed out to be Black.” ‘ite <——— ——> dick od andics ay ITSEtF 122 ‘Thar won't ell us much, as i stands. The answer to the frst {question (as we've already seen) is going tobe “yes” (and moreover, ar that point, is going co be a physical fact about the world that b effectively knows the color of exch of those ist N electrons). But of course b ist going to produce any coherent answer t0 the second question; once h has responded to that question (if b isa competent converser on these matters), the state of the world is ‘going to be a superposition of states flike the supeeposiion that arises in the event that we ask fr what the color of the electron is ‘when (6.1) obtains) in which Fr answers that question in various diferent ways. But here’s something curious: Ie happens that inthe limit as N _g08s to infinity (that is: inthe limi asthe numberof coloe mes- surements which # has so far performed goes to infinity), che state ‘of che world will, with certainty, approach a stare in which b will answer that question ina perfectly determinate way, and in which the answer b gives wl with certainty be "12" (whichis, of course, precisely what ordinary quantummechanies will predict, with cer tary, about thar eespons, in that hit ‘And thar turns our to be an instance of something @ good deal {6s hats why that ot of hg uo be te. Here how oa lt Canie forexagle) annie cal of clesrns es call hem 12, Sy allo whch ret ther state; a onder he owing bur tf han cleton: Ox (0) heuer of black ctor ang the ot N "Now ardoary quanto mechanics ht: ane hans wth cls) cra tar he enor of ech neo hoe earn were oe mess the Ise thee ae aiely may of tne ero st allen) pect al of hse eastern would with erat come oa "Hack," nd precy hall ‘them woul with een came ot hte” Moreover Oy fo any vale of 1 iscompunbl with the clar of eery nae fowler Ard! 0 flows ‘hat ory gouoamrscans nade that N approche infinite poe Aig bata Mesure of he value of Oy on the ellen of serene Sc above wil id he le 12 wil approach I. And no lions a hat lc of toes (ie to meses be ie an eget of wher erator fi hat Ow approaches aN apposchs ny, with Cael Ye “ANd wil follow fc a that ht wheter or mat ther ae ever any sich ios scallops teste a compan stem comenga a collcoon analy N THe Dynamics ay fr5e1r 123 more general which runs a follows: Suppose that an observer bis confronted with an infaite ensemble of identical systems in ident- cal staes and that she caries out a certain identical measurement fon each one of them. Thea, even though there will actualy be n0 ‘matter of fact about what b rakes the outcomes of any of those measurements to be, nonetheless as the number of those measure ments which have already been extried out goes to infinity, the stare of the world will approach (as a well-defined mathematical iit) a state in which the reports of b about the staistical frequency of any particular outcome of those measurements will be perfectly definite, and also perfectly ip accord with the standard qcantent ‘mechanical predictions about what that frequency Ought to be So itcurns out noe to be altogether imposible (even ifthe standaed sway of thinking about whae it means to be in a superposition isthe ight way of thinking about i) thatthe state we end up in at the ‘conclusion of a measurement ofthe color of a hard electron isthe fone in (61). And so everything we've been thinking about the measurement problem vp till ow isnt right. ‘And wharall chs obviowsy suggests is that maybe there justin any such thing as 2 measurement problem. ‘That is: maybe (even if the standard way of thinking abour what is means to ben a superposition is the right way of thinking about ‘what it means to be in a superposition) the linear dynamical laws are nonetheless the complete Iaws of the evolution of the entire world, and maybe all of the appearances to the conteay tke the appearance thar experiments have outcomes, and the appearance thatthe world doesn't evolve determinisially) turn out 0 be just the sorts of delusions which those laws themselves can be shown to bring on! ‘rlons ad of compet hsover whos jae cared out mesnirrets tthe calor of al the eer wil, wth sary, bea eget a hat ‘hieers porn i sb ska) tha the al of whatever Operators at Dn arproathes a Nspyoachen nity ‘ete mathe dss fal hs stent uch ie po han {te ore abe can Be fou athe dtl dsenaton of Jl Bate 1990,Tue Dynamics ay inser ina ‘This is an amazingly coo! idea (er allie “the bare theory"), and this is the idea that it strikes me as interesting to read into Everett's paper” [Notwithstanding al the stuff we've just learned, however, ie seems ‘to me that the bare theory can’t be quite rghe either Note, for example, that ifthe bare theory struc, then there will bbe matters of face aboue whar we think about (s3)) the frequencies of “black” outcomes of measurements ofthe color of hard clec- trons only (if at all) in the limit asthe number of those measure- ments goes to infinity. And so if the bare theory is true (and since only a finite number of such measurements has ever actually been carried out by any one of us, or even in the entre history of the ‘worla) then there can't now be any mater of fat (notwithstanding ‘our delusion that thee i one) aboot what we take those fequen- «ics to be. Ano, if the bare theory i true then there can be any matter of fact (notwithstanding our delusion thar there is one) about whether or not we take those frequencies tobe in accordance swith the standard quantummechanical predictions about them, And so, ifthe bare theory is tru, isn clear what sorts of reasons ‘we ean possibly have for Believing in anything like quantum me: chanics (which is whar che bare theory is supposed t0 bea way of ‘making sense of in the frst place ‘And as a matter of fac, ifthe are theory is true, then it seems | ‘extraordinarily unlikely thatthe present quantum state ofthe world ‘ean possibly be one of those in which there's even a matter of fact about whether oF not any sentient experimenters exist at al. And ‘of course in the event that there ten any matter of fact about 17 OF couse he hypothesis thatthe auntons of mote ae always macy ih ato the Sr hig that Dane Linge and Presper and al hse the bys moto of Caper 3 took tenses to beaches “The ost tha antonsingy s er nest ave ocr those pe hat lows rom hat hypothe thst experenent ast never havea, ‘nds noe ofthe ever wore sat haw come ots with at 50 ae of thm ever eed it he sot of comida that wee the nde other 5 This xy lc way of puting the pole i ew Joshua Newman whether or not any sentient experimenters exist, then it becomes ‘intelligible even to ingure (as we've been doing bere) about what sorts of things such experimenters wil report ‘And then (as far as Ican cell all bets ae off And so it seems to me not toe entertainabl, in any of the ways we've talked about so far, or in any other way U know of, that he linear quanturn-mechanical equations of motion are the teue and ‘complete equations of motion ofthe whole world. And thats shat. ‘Bur there are nonetheless interesting things co be learned about the measurement problem (things that it will be wel to beatin ind in connection withthe problems we ran into at the end of the lst chapter and in connection with problems we will run into at che ‘end of the next chapter) in this suff about what superposiions fel like. ‘What that stuff shows, I chink, is that precisely tha featre of ‘those equations which makes it clear that they cannot possibly be the we and complete equations of morion of the whole world {that is: heir linearity) alo makes it radically wncleae bow mich of the world and sehich parts of the world those equations possibly cam bye the true and complete equations of motion of. ‘What I think i shows (0 put it another way) is that there can be no such thing as a definitive lst of what there have absolutely got to be matters of fact about which i scientifically St to serve as fin “observational basis from which ll attempts at ving quancar mechanics up must stat ot ‘What I think it shows is that what there are and what there arent determinate matters of fact abous, even in connection with the most ‘mundane and everyday macroscopic fearures of the external phys ical world, and even in connection wich the most mundane and everyday eatutes of our own mental lives is something which we Shall ulimately have to learn (in some part from whatever turns ‘ut to be the best way of xing quantum mechanics wp? 9 But motto ht nig wl be no sihformad at sine oe of he hing ha all is aie diffe qusone ous he ery bots ong athe et way ing gunn mechanThe Dynamics Almas by tll Lets stare again ‘Suppose that there's just one world. And suppose that cere’ just fone complete story of the world tha’ true ‘And suppose that quantum-mechanical state vectors are com plere descriptions of physical systems. And suppose thatthe dynam- feal equations of motion are always exactly right. ‘And suppose that we should like to insist, as a mater of princi pi, that healthy people can correctly report whether oF not they themselves have any determinate belie about (say) the position of some particular point, ‘Then (since the dynamical equations of motion entail that heathy people in superpositions of brain states corresponding to dlfferent beliefs about the position of some particular pointer will ‘with certainty report that they have some determinate belief about the postion of that pointer) there's going to have co be something, funny about how mental staces supervene on brain states.” Lets se ifwe ean cook up something funny like that ‘Think of b when she’s about to measure the color of the hard electron, when she's in her “ready” state. When the measurement is over, the physical stare of # and her measuring device and the electzon is going robe the one in (6.1). Tha’ what's dictated, with ‘certainty, bythe deterministic equations of motion Suppose, however that all shar tue, but thatthe evolution of ‘is mental stare in the course of a measurement like this one is explicitly probablistic. Here's how things would goin this particu lar cas: b starts out (with certainty) in the mental state associated with ready), and she ends up (with equal probabilities) either in the mental state associated with [believes e black), or in che mental sate associated with believes e white) What’ certain about how she ends up, though, s chat she ends up (ust as she testifies she 10, That 8 ping to Rave 0 be he cate ha ames living tat chan sich mot enc with ome parle ae ha peri rae te we've ro cling beer cha suc) being 1k oon: how hs opt ro be eral fm he een tat he in sate ofthe cern is hte) + Hac ai he eet tha esses that ‘seo’ col the thre} li ot weth etait, the mot sae does) with some perfetly determinate belief about shat the color of the lecton is. So far so good. Let's ty wo rake it lle further ‘Whatever belief # does end up with, when (6.1) obtains, is necessarily going tobe a false belief. Bu here are very natural ways ‘of cooking things up so as ro guarantee that that belief will none- theless have an important kind of effective validity, at leas in 50 far as bis concerned! there are ways of cooking things up (that is) so as to guarantee tha che future evolution of b's mental state ‘will proceed, in general, exactly as if B's beliefs were true Here’ what I mean Suppose that the mental state chat fends up in when (6.1) ‘obtains (all that time #) happens to be the one associated with [believes e black), and suppose that she subsequent repeats that color measurement (with another color measuring device) on that same electron. When that’s done, che physical state of things is soing to be the one in (6.2), and b will with ceetainy (on this proposal) end up in the mental state associated with [hlieves out: come of frst measurement is “black” and believes outcome of second measurement is “black").'* And that’s precisely how B's mental state would have ended up, sith certainty (on this pro- posal), in the event that her belief that the electron was black at time ¢ (which was fale) had been true. ‘And Suppose just as above) that a state ike the one in (6.1) ssoied with fey ashen up in the mil ste ascited with ites white wih raasty and abel rp in the eal se !sscined with es¢ Dc with pay 12. Ticsr of hing was fog git aang ine go, bt osreb Aileen for ess which had wong co. wth ha the cg dion date abou what i fel the to be In superposition) by Bead Pespa 1971 1, Of com, thte sense in which it seemed igh ony ha fete nw what the color he elton whee (1) obs, om the bare ery too, Bue what were king shot now wil aman 0 sathing od! Seat cope tanta. Ad of cos in he vet tha PS al ate 3 appt be te be ssid with bles ent hen al with ety ed up the metal tte secede once of it nenurnent ahi” and er ‘some of end estrone hiTue Dynamics av InsetF 128 ‘obtains, and suppose that b's mental sate happens to be the one associated with [believes blacks, and suppose that she subse- {query carsies out a measurement ofthe hardness of that same electron; then, when sha’ done, the physical sare of things i going, tobe (66) Vllfbeeves outcome of frst measurement is “black” and believes outcome of second measurement is “hard”)Black")athaed”blbard)) + Iibsieves outcome of ist measurement is “black” and believes outcome of second measurement is soft *Back”) "SOR" lof,) + Ubelieves outcome of fist measurement is “white” and ‘peeves outcome of second measurement i “hard” oehite”)*hard”plhard),) ~ ibsieves outcome of fist measurement is “white” and believes outcome of second measurement is “soi, hite")n "804" )asoF (where mit ie the color measuring device and m2 isthe hardness measuring device), and the probability thar b will end up in the mental state associated with [believes outcome of ist measurement is “black” and believes outcome of second one is “hard” wil be “a, and the probabilry that she will end up in the mental state associated with [believes outcome of fist measurement is “black” tnd believes outcome of second measurement is “sft”), will be Ya, and the probability of her ending up in the mental states associated with either [believes outcome of first measurement is “white” and believes outcome of second measurement is "hard”). or [believes outcome of frst measurement is “white” and believes ‘outcome of second measurement is "sof", will be 0. And suppose that astae like the one in (64) obtains and that ‘h% menal stare happens to be the one associated with believes 1 black), and suppose that now catres our a measurement of the color of elecon 2 (in which case the physical state of the world ‘will become the one in (6.5). Then (on this proposal) the proba bility of 8 ending up in the mental state associated with [believes 1 black and 2 black), will be 12, che probability of F's ending up in the mental state associated with [believes 1 black and 2 white) willbe 2, and the peobabilices oF Fs ending up inthe mental states associated with ether believes 1 white and 2 black), or [believes 1 ‘white and 2 white), will both be 0. ‘And #0 on." ‘That (technically) will do the trick. On this proposal, quantum ‘mechanical wave functions ae complete descriptions of the physi cal states of things, and thote wave functions invariably evolve in peefect accordance with the dynamical equations of motion, and it ‘makes no physical difference at all what basis we choose to wite those wave fonctions dos i," and measurements carried out by sentient observers thats by observers with minds) invariably have determinate outcomes in the minds of those observers, and the “statistical distributions of those outcomes will be the usual quan- tum-mechanical ones, and there isn't anything mysterious about how probabilities come up in this theory,” and the reports of seotient observers about chir own mental states will invariably on this propos, be correct. 1S, Wha be sd fa sly ee dee acount which, hy he ses cam br found in Alber sn Locus 1988) doe amon» compe foal ser lawe ofthe elon of mental tater boas iethatcan be ake (ted thoy cn feck op ch 3 wy seo ane that eeyting Pre ‘Set abarthos ar wil be te 1, OF ca, there wil (on ths pcwre, and on every way of acemping rake we of quem mechani ke some parc ae Sain sae ich rrp oor mere) "eigenstates frais ba whet Bac hat a ‘mans bea mater of convention coc, wast ha that wiley Spend Inte) om the prc! rare of the bran sp gueton. The bai ae tat Cnrepents oie that crn econ Hac, for example il res Sly be tbe one which puro vt ofthe dynarical equations of moto] “possi ose oespond to an ute ke "Whar coo do vou eve tenn 0 bet" with an utterance ike Neen he elaron oe lace” And of ure hat rn se tha wi ei bea cmp banded gh forwany posal question’ 17 he way that probate cme up ths theory, afer al hae hey ge tino by a a tht Bt ples thet thw plese wo be ander ‘ood in prey the convenTHE Dynamics ay INSEL 130 [And of course this view ofthe woeld isa thoroughly realist one (chat is: this view entails that there is invariably a single correct objective description of the entire physical and mental universe, even if nobody happens to know what char desc view (even though i's an explicitly dualist view) entails that the mental parts ofthe world have no effects whatever on is physical ‘pats (that iss this view n't at all ike any of the dal of collapse, sis view entails that the physical world is causally closed). Bur the dualism of this sort ofa picture is nonetheless pretty had On this proposal (for example all but one of the terms in a superposition fike the one in (61) represent (as it were) mindless Jus; and which one of those terms is not 4 mindless bull can’t bbe deduced from the physical state of the world, or from the ‘outcome of any sort of an experiment; and i will follow from this proposal that most of the people we take ourselves co have met in ‘our lives have as a mater of fact beer such hulk, and not realy people (not really animate, chat is) a all Here’s a way to partly fix thar up: ‘Suppose that every sentient physical system there isis associated fot with a single mind but eather with 2 comtimous infinity of ‘minds; and suppose (his is part of the proposal too) that the measure of the iniite subset of those minds which happen to be ‘in some particular mental state at any particular time is equal to the squace of the absolute value ofthe coefficient of the brain state associated with that mental stat, in the wave function of the world, at that particular time (so that, for example, when states lke (6.1) ‘brain, haf of Hs continuous infinity of minds will believe that the electron is lack, and half of them will believe thatthe electron is white. The time evolution ofeach individual mind, on this proposal, is precisely the probabilistic one described above (the one that we cooked up for the single-mind proposal, but since (on this pro- posal) there are always a continuous infinity of minds (or ese 90 ‘minds ar all) in any particular mental state the evolution of the ids of any particular sentient observer asset is invariably (that 's: with probability 1) going to be deterministic. Moreover, at any Tat Dynamics ay Hse 131 particular instant, the mental states ofthe minds of any particular observer will necessarily be distibuted in accordance with the prescription ofthe last paragraph, So this proposals going oental that what you might call che “global” mental sate of exer sentient being s uniquely fixed by the physical state ofthe world" ‘And there's something ese about thiskind ofa picture that's nice: thiskind of a pictures local. That's surprising, That's precisely the sort of thing that Bell’ theorem was thought to have ruled out, Let's see how it works, ‘Consider an EPR.type state: (6.7) [lack ites ~ hoi black and suppose that electron 1 is located at point 1 and that electron 2 is located at point 2 and that an observer named bt (located at point 1) measures some spin observable of electron I and that a ‘observer named h2 (located at point 2) measures some spin obsct able (not necessarily che same one} of electron 2. ‘What Bel proved is that there can't be any local way of account- ing forthe observed correlations between the outcomes of measure ‘ments like that; but of course (and this is the crux of the whole business) the idea thar there ever are matters of fact about the outcomes” of a pai of measurements like tha i ust what this sort of a picture denies! Lets go through it careflly. At the conclusion of a pair of measurements like che one just described, on this picture, the state ofthe world is going to be a superposition of state, in each of which each of those two mes- surements have one oF the other of theic «wo different possible ‘outcomes. And at that point, on tis picture, no matter whae spin ‘observable of electron 1 gets measuted by bi and no matter what spin observable of electzon 2 gets measured by #2, haf of t's ‘minds are going to believe thatthe outcome of whatever measure ment she did was +1, and che other half of her minds are going £0 1. This isthe walled many. mies imerpetation of guar mechs, ‘which ass propened by Bay Leewer alms Ales aad Laces 1988,