You are on page 1of 5

2015 12th International Conference on Information Technology - New Generations

Qubits or Symbolic Substitutions for General-purpose


Quantum Computing?
C.H. Wu
Dept. of Electrical and Computer Engineering
Missouri University of Science and Technology
Rolla, MO, USA
chw@mst.edu
one step than that one processor can perform in four
consecutive steps. That is the origin, or at least one of the
origins, for higher-order intelligence. Larger quantity
(processors) means higher quality (the artificial intelligence).
But towards that goal of massive parallel computing, the
use of qubit concept is fundamentally flawed as we elaborate
more in Section II. The appropriate solution of quantum
computing is thus to use symbolic substitutions. This is argued
strongly with the fundamental quantum processor presented in
Section III. The quantum computing architecture is then
naturally resided in the use of cellular automata. This is
explained in Section IV. The conclusions are described in
Section V.

AbstractThe fundamental flaws of qubit concept for generalpurpose quantum computing are elaborated here. We show that
from implementing of the addition operation of two bits, only
four symbolic substitution rules are needed. Superposition of
four states from the two qubits is then irrelevant for the addition
operation of two bits. The fundamental quantum processor
needed is further described here against the quibit concept. The
quantum Turing Machine is thus settled in the cellular automata
architecture. Therefore, quantum computing must be rulebased, rather than logic-gate based.
Index TermsQuantum Computing, Symbolic Substitutions,
Cellular Automata, Quantum Turing Machine, Aharonov-Bohm
Effect, Quantum Networks.

I. INTRODUCTION
The concept of qubits for quantum computing is well
known for decades now [1]. Yet for general-purpose quantum
computing, this qubit concept has been challenged recently [25]. Any general-purpose quantum computer must be able to
compute anything that is computable. But the computable
things can range from those which are purely sequential in
nature, such as addition operation, to those which are purely
parallel, such as Fourier transform. A true quantum Turing
machine is a massive parallel computing machine that must be
able to handle both extreme computations and anything that are
in between the two extremes. Yet our understanding of
quantum computing rely on those computable things that are
purely parallel, such as Shor algorithm [6], which are specialpurpose computing. Any general-purpose parallel computer
must show how addition operation is implemented first.
The general strategy of parallel computing is to trade space
with time. That means, with millions of processors used, it can
save the number of steps (timewise) needed generally if the
nature of computing is not purely sequential. But the number of
steps required for a purely sequential- nature computing will be
the same for any machine, from a quantum parallel machine
or from a classical electronic machine. There will be no step
advantage (since each machine requires different time to
accomplish one step, there can be no comparison on the
absolute time scale needed). For example, for the addition of
3+5 operation in digital operation, the number of steps (which
is 4 steps) is fixed from any machine because of the ripple
carry requirement. But when four identical processors are
connected together, it can perform higher quality computing in

978-1-4799-8828-0/15 $31.00 2015 IEEE


DOI 10.1109/ITNG.2015.117

II. WHY QUBIT CONCEPT IS FUNDAMENTALLY


FLAWED?
The fundamental flaws of qubit concept for generalpurpose computing has been pointed out recently [3-5] Here
we would like to challenge this qubit concept for generalpurpose computing from another point of view. It is well
established that any parallel computing machine must consist
of a huge number of identical processors with some form of
interconnections between them. Each processor can execute
certain number of instructions and a storage capability of data.
The qubit concept centered on the property of quantum
superposition principle. A large electron wave packet
containing billion superposition of 0s and 1s can be
manipulated to perform space and time saving computing, as
many scientists have believed generally. When we use many
qubits for parallel computing purpose, we can use one qubit as
one processor or we can use two qubits as one processor and so
on. If one-qubit-one-processor scheme is used, then each
processor will have the smallest capability of computing
truth/false only. If two qubits are used as one processor, then
there are four states in superposition. Let us designate those
four states as

S1

for the (0,0) pair,

for the (1,0) pair and

S4

S2

for the (0,1) pair, S3

for the (1,1) pair. In this case, this

two-qubit processor is in principle capable of computing the


addition operation of two bits and can function as a half-adder
processor since all four combinations of the operand pairs are
available and in superposition. However, in parallel computing,
698

all processors must be interconnected in certain way,


depending on the computer architecture employed. But no
matter what kind of parallel-computing architecture one
chooses, there are always processor-to-processor spatial
relations from the imposed interconnection scheme. For
example, for the addition operation of two long bit-strings, the
ripple carry must move from the least significant bit-pair
location to that of the most significant bit-pair. Therefore, the
processor at the least-significant-bit location must be connected
only to its left side (and not to the right side) processor where
the second-least-significant-bit is located. The proper
processor-to-processor interconnections, or the spatial
relations, are entirely missing in the qubit approach and is thus
the first fundamental flaw of qubit concept for a generalpurpose quantum computing. For the entangled electron wave
packet of qubits, there is no distinction of interconnections of a
given processor with respect to other processors which is from
the left side, from the right side, from the up side or from the
down side.
The second fundamental flaw of qubit concept follows
directly from the first. In the qubit approach, the manipulations
of qubit states, or the proper algorithms, are through the use of
some basic quantum logic gates. This is fundamentally wrong
[3-5]. Any parallel computing is rule-based because there are
processor-to-processor spatial relations imposed in any
computer architecture. Logic gates are truth-table based and
there are no spatial relations explicitly imposed. Rule-based
computing means the use of symbolic substitution rules, as
shown in a Turing Machine or in cellular automata architecture
where the proper spatial relations must be explicitly imposed.
The third fundamental flaw of qubit concept will be addressed
in Section III.

Fig.1

Figure 1: (a) N flux qubits as N point-contacted AB


rings. (b) A pair of qubits is grouped together and
used as one processor with two flux inputs and the
readout probes. (c) The two qubits must be
collapsed to form a diatomic molecule with three
readout probes, S, C and D terminals. A test pulse
is from terminal S. This is the magnetic AB version
with two fluxes as two input operands. (d) The
electric version of the AB effect. The charging test
pulse is scattered out to S or C terminals and
deposit the test charge on metal cylinders V1 and

III. A FUNDAMENTAL QUANTUM PROCESSOR WITH FOUR


SYMBOLIC SUBSTITUTION RULES.
We will use Aharonov-Bohm-effect (AB effect) based
quantum networks with two flux qubits to illustrate that it is
the four symbolic substitution rules that are relevant for the
addition operation of two bits, rather than the four
superposition of states. Each AB ring is like an artificial atom
with two possible directions of the angular momentum. The
half-adder quantum processor (US Patent #8,525,544) is
constructed from two coupled AB rings as shown in Fig.1c.
That means the two entangled flux qubits that possess the
superposition of four states, as shown in Fig.1b, must be
collapsed first to form a strongly coupled diatomic molecule
with three attached probing terminals, S, C and D. In other
words, the internal coupling (the entanglement) of qubits and
the external coupling (the readouts) must be integrated as one
complete system and there can be no separation. As shown in
Fig.1b, the two flux qubits are in point-contacted situation in
order to maintain the superposition. But each AB ring is a
closed harmonic oscillator ring. Such a ring (a closed system)
can only couple to external probes made of the harmonic
chains (the terminals, S,C and D) of the same strength. Any
weaker or stronger terminals will not provide proper readouts.

V2 or scattered to D terminal and dumped. QC is a


quantum circulator.
Because two point-contacted AB rings are very weakly
coupled, this forces the external probes to be just as weak. The
result is that such weak probes will provide inconsistent or
undecidable readouts [2]. In other words, internal coupling
between the two qubits is not consistent or comparable with the
required external couplings. We show that the external probes

699

have to be strong, therefore the internal coupling has to be


adjusted to cooperate with the external coupling and it is not
the other way around as many researchers have attempted.
What we showed in Fig.1c is that the two qubits must be
collapsed to form a diatomic molecules first and with three
external probing terminals, C,S and D , attached. With the two

can be somehow manipulated successfully to perform the


addition of two bits, the results cannot be better than what the
four symbolic substitution rules can provide. Our symbolic
substitution rules are derived purely from the gauge invariance
of quantum mechanics and quantum interference of the phase
of electron wave function from the multiply-connected space of
quantum network provides the results. Thus the superposition
of two qubits is totally irrelevant for the addition of two bits.
Therefore more qubits are not even needed for general purpose
computing. In the special situation when the nature of the
computing is pure parallel, such as Fourier transform, all qubits
are independent of each other, very much like a bundle of
photons passing through a lens and some filters, a specialpurpose quantum computing may be possible.
In Fig.1d, we further convert the processor from the use of
the magnetic AB effect into that of the equivalent electric AB
effect. The two coupled AB rings are drawn in two squares

fluxes, ( 1 , 2 ), as inputs for the operands, the four


addition rule- based symbolic substitution rules can be
obtained. That is to say that when we provide a test pulse of an
electron packet originates from terminal S, such a pulse will be
scattered to terminal C if the two fluxes correspond to (1,1)
bit pair, or

S4

state. The same test pulse will be scattered to

terminal S if the two fluxes correspond to (1,0) bit pair ( that


state), or correspond to (0,1) bit pair ( that is S 2 state).
Finally, if the two fluxes correspond to (0, 0) bit pair (that is

is

S3

with the use of two charging cylinders, V1 , V2 as shown. The


use of electric AB effect allows the charge of the test pulse
from terminal S be scattered and deposited directly on one of
the two metal cylinders or be dumped to terminal D. Before the
scattered charge arrives at any of the three terminals, the
previous charge at each cylinder is removed (or refreshed). If
the charge at one of the two cylinders is saved during the
refresh, then the computing is reversible through Buttiker
symmetry rule. Thus the reversible quantum computing is less
stringent than what some unitary operations on an initial state
can imply. However, at each new pulse, the energy of the
previous pulse and the energy used in the refresh are
consumed. Therefore it is thermodynamically irreversible in
any case. In other words, quantum computing dissipates energy
through the use of test and fresh pulses.

S1 state), the test pulse will be scattered out to terminal D. The


four flux pairings can be mapped into four bit-pairings
properly. Thus four symbolic substitution rules are thus
derived. In Fig. 2 [4], we show the four rules of symbolic
substitutions with the four states of the operand pairings and
the dual-rail notations later used for cellular automata. Note
that the symbolic substitution rules in Fig. 2 distinguish left
and right, the spatial relations, while all logic gates are truthtable based. This is a great departure of our approach to the
quantum computing as compared to other researchers.
The reason that such four symbolic substation rules can be
established in Fig.1c is originated from the function of a
quantum circulator [6] in a single AB ring. A single AB ring
with three terminals can have a scattering event such that at
certain flux value, an entire test pulse from one terminal can be
scattered to one of the two other terminals. That means if the

ring is removed from Fig.1c, a test pulse from terminal S


IV. CELLULAR AUTOMATA AS THE NATURAL ARCHITECTURE
FOR QUANTUM COMPUTING

can be scattered entirely to terminal C at certain flux value of

. However, if the direction of this flux is reversed, then

Referring to Fig.1d, the quantum processor is capable of


executing four instructions with the storage of two data. Note
that terminal C of the processor is always connected to the next
processor to its left side, while terminal S is connected back to
the same processor through the use of a QC. The scattered
electron pulse to terminal D is dumped and unused. This is the
canonical interconnections. Thus an array of the quantum
processors connected in this fashion forms a one-dimensional
cellular automaton. Addition operation of any two long bitstrings can be performed in the cellular automaton as we have
shown [3,5,7]. In cellular automata computing architecture,
each cell contains one processor as we have shown in Fig.1d
and the processor-to-processor interconnections are the
minimum level. Namely, the number of interconnections is
equal to the number of processors used. Thus each processor is
only connected directly to its left processor due to the ripple
carry requirement. In other words, the sequential nature of the
addition operation imposes such a computing architecture. But
in such architecture there is no need to construct a full adder
that is commonly used in the sequential classical electronic

the same test pulse will be scattered to terminal D instead. That


is the function of a quantum circulator (QC). Now we extend
the concept of a QC further by attaching a second AB ring with
a second input flux, 2 to modulate the phase of the electron
wave function along the center common path with the first AB
ring. Now we have four flux pairings to map with the four
operand pairings when two bits are to be added. The four
scattering events can be mapped into the four addition rules for
two bits.
Now we would like to point out here the third flaw of the
qubit concept for general quantum computing. The internal
coupling of two qubits does not cooperate with the external
coupling needed as one complete system. The internal coupling
has to be strong and thus the superposition has to be destroyed.
In addition, the locations to place those external couplings (the
readout probes) are themselves part of the computing and
cannot be separated. All those above considerations are not
implemented in the qubit approach. We note if the two qubits

700

computing circuits. This is because whenever a full adder is


needed, one can always employ the quantum processor two
times, since space and time are traded in parallel computing.
Let us illustrate the addition operation for 5+7 in cellular
automaton operation. The two operands in four digits are 0101
and 0111 respectively. The least significant bit pair is thus
(1,1) or
or

S2

S4

state, the second least significant bit pair is (0,1)

Figure 3: The iteration sequence for the addition


operation of 5+7. The operands are located from
cell locations from 1 through 4 on the top row. The
configuration changes from initial at 0 and finishes
at the third iteration, as labelled on the left vertical
axis. The result of 12 is read from the charging of
V1 cylinders from cells 1 through cell 4.

and so on for the rest of bit pairs. So the initial

configuration of the operands is thus S1 S 4 S 2 S 4 . They are


located at the cell positions, labelled as cell 1, cell 2, cell 3 and
cell 4 on the top row of Fig 3. The rest of the cells are filled
with

S1

states (from cell 5, 6 up or from 0, -1 down). The

setting of ( V1 , V2 ) at each cell determines the state at each


cell.

of
This configuration is to go through four iterations of pulses
and refresh cycles at each cell (labelled as 0,1,2,3 on the
vertical axis on the left side) to arrive at the final configuration

S2 S2 S1S1

at the third iteration . The lower sections of

this configuration, which are located at the

V1

cylinders, is

the sequence of (1100) which is the digital value of 12, the


result. The upper sections of this configuration, which are

Fig 2

located at V2 cylinders , is the sequence of (0000). Outside of


the four cells, all the states stay at

S1 state.

The iteration processes of Fig.3 can be explained through


the 16 cellular automata transition rules that transform from a
parent cell to its child cell after each iteration. This is shown in
Figure 4. Those 16 rules are totally equivalent to the four

Fig 4.

Figure 4: The four symbolic substitution rules in


Figure 2 are equivalently transformed into a set of
16 parent-child transition rules. Each cells parent
state is transformed into its child state according to
the state of its right neighbor only. This set of
addition-rule compatible cellular automata rules is
the only one set out of 4.3 billion sets available.

Figure 2: The four symbolic substitution rules by


the quantum processor shown in Fig. 1d. The four
rules correspond to the four states with dual-rail
notations shown.

Fig.3

symbolic substitutions shown in Figure 2. For each cell the


transition rules involve its left neighbor only. In other words,

701

the wiring is connected to its left cell at its

V2

computing, classical or quantum, requires a description,


through the use of a computing architecture, of how the billions
of identical processors are to be interconnected. Those spatial
relations are not included in the qubit concept. This is the first
flaw. The need of specifying those spatial relations in quantum
computing clearly requires that quantum computing is rulebased and not logic-gate or truth-table based. Therefore
general-purpose quantum computing cannot be a sequence of
unitary manipulations of initial state through some quantum
gates. This is the second flaw. The superposition of states in
two qubits can be maintained only through very weak internal
coupling. But the external coupling for the readouts requires a
strong coupling and cannot be the other way around. This
means internal coupling has to be strong to destroy the
superposition. When the four symbolic substitution rule are
established, we show that superposition of four states from two
qubits are totally irrelevant for the addition operation. It is the
four symbolic substitutions that are needed. That is the third
flaw of the qubit concept.
When the symbolic substitution rules are established,
addition operation is in one-dimensional cellular automaton
with a correct quantum processor inserted in each cell and the
proper interconnections as shown. The processor executes four
instructions and stores two data. The cellular automata
architecture provides the minimum interconnections between
the processors. The exponential growth of the computing
power, or the growth of parallel instructions, is thus through
the number of processors used.

cylinder. It is

worth noting that this is the only set of the transition rules, out
16

of a total of 4,294,967,296 (= 4 ) possible sets of rules


available. To find such a set of transition rules in Fig.4 directly
from all possible transition rules would be like to find a needle
in a huge pile of haystack.
Thus addition-rule compatible cellular automaton has to be
wired in the canonical manner as we have described. Any
deviation from such connections will result in a new kind of
science as phrased by Wolfram [8], which are not additionrule-compatible cellular automata. For example, if terminal C
and terminal S are interchanged, locally the processor in each
cell still functions as a half adder. But the long range relation is
distorted with respect to its neighboring cells so that the
iteration results will show the oscillations between two
configurations or in the form of a moving signal [5]. Those
results are attributed to the Non-Euclidean interconnections
because geometry and cellular automata for computing are
intertwined and are imposed by the use of Euclidean space
only. Any deviation is thus becomes Non-Euclidean.
It is worth noting that in the Turing machine, each cell
contains a one-bit processor and there are eight symbolic
substitution rules used for the addition operation. However, the
proper processor for such eight instructions has not been
shown. Instead, here we show that each cell must contain a
two-bit processor and there are four symbolic substitution rules
associated with such a processor, which is explicitly shown in
Fig.1d. Once addition operation is established, multiplication
operation consists of many addition operations in parallel that
can be performed in two-dimensional cellular automata.
Therefore all algebraic operations can be executed in a two
dimensional cellular automata. Thus we have a quantum
Turing machine established through the use of the fundamental
quantum processor shown in Fig.1d, because anything that is
computable can be transformed into some proper algebraic
operations. .

REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

V. CONCLUSIONS
We show that the use of qubit concepts for general-purpose
computing is fundamentally flawed. Massive parallel

702

A. Childs and W. van Dam. Rev. Mod. Physics, 82,22(2010)


C.A. Cain and C.H. Wu. J. Appl. Phys. 113,154309 (2013).
C.H. Wu and C.A. Cain. Physica E,59, 243 (2014).
C.A. Cain and C.H. Wu. J.Appl. Phys. 110,054315 (2011). Also
US Patent #8,525,544 .
C.H. Wu. J. Cellular Atuomata, 9,271-286(2014).
C.H. Wu and D. Ramamurthy. Phys. Rev. B,65,075313 (2012).
C.H. Wu, Int. J. Engr. Sci. Invention 3, 59-70, (2014)
S. Wolfram. A New Kind of Science, Champion, ILh,
Wolfram Media, (2002).

You might also like