Quantum Types
Once initialized, Qmod variables reference a quantum object in some state. Quantum types determine the overall number of qubits used to store the object, as well as the interpretation of its state. For example, a quantum object stored on 4 qubits can represent an array of 4 bits, an integer number in the domain 0 to 15, or an array of two fixed-point numbers in the domain [-1.0, -0.5, 0, -0.5]. The type determines which interpretation is the intended one, for example, when evaluating quantum operators.
Qmod has two categories of quantum scalar types - bits and numbers. Qmod also supports quantum array and struct types, which can be arbitrarily nested.
Quantum Scalar Types
In Qmod, there are two kinds of scalar quantum types:
qbitrepresents the states \(|0\rangle\), \(|1\rangle\), or a superposition of the twoqnumrepresents numbers in some discrete domain - integers or fixed-point reals
When declaring a qnum variable, you can optionally specify its numeric attributes -
overall size in bits, whether it is signed, and the number of binary fraction digits.
Syntax
qbit
qnum [ < size-int-expr [ , sign-bool-expr , frac-digits-int-expr ] > ]
In Python the classes QBit and QNum are used as type hints in the declaration of arguments:
name : QBit
name : QNum [ [ size-int-expr [ , sign-bool-expr , frac-digits-int-expr ] ] ]
The same classes are used to declare local variables:
name = QBit ( " local_name " )
name = QBit ( " name " , [ [ size = ] size-int-expr , [ [ is_signed = ] sign-bool-expr , [ [ fraction_digits = ] frac-digits-int-expr ] )
It is recommended to use the SIGNED and UNSIGNED built-in constants instead
of True and False respectively when specifying the sign-bool-expr qnum
property.
Semantics
- Computational-basis encoding of numeric types is big-endian (the most significant bit has the highest index).
- size-int-expr determines the overall number of qubits used to store the number, including sign and fraction where applicable.
- If sign-bool-expr is
True(SIGNED), two's complement is used to represent signed numbers, utilizing the most-significant bit for sign. - frac-digits-int-expr determines the number of least-significant bits representing binary fraction digits.
- When only size-int-expr is specified and sign-bool-expr and
frac-digits-int-expr are left out, the later two are set to
UNSIGNEDand0(integer) respectively.
Examples
In the following example, two 4-qubit numeric variables, x and y, are prepared to store the
bit string 1101. x is declared with no sign bit and no fraction digits, and therefore
its state represents the number 13. y is declared to be signed and have one fraction-digit, and
thus, the same bit-level state represents the number -1.5.
qfunc prepare_1101(output qba: qbit[]) {
allocate(4, qba);
X(qba[0]);
X(qba[2]);
X(qba[3]);
}
qfunc main(output x: qnum<4>, output y: qnum<4, SIGNED, 1>) {
prepare_1101(x);
prepare_1101(y);
}
from classiq import qfunc, Output, QNum, QArray, allocate, QBit, SIGNED, X
@qfunc
def prepare_1101(qba: Output[QArray[QBit]]):
allocate(4, qba)
X(qba[0])
X(qba[2])
X(qba[3])
@qfunc
def main(x: Output[QNum[4]], y: Output[QNum[4, SIGNED, 1]]):
prepare_1101(x)
prepare_1101(y)
Numeric Inference Rules
Numeric representation modifiers are optional in the declaration. When left out, the
representation attributes of a qnum variable are determined upon its first initialization.
Following are the inference rules for these cases:
- When the variable is passed to a function as its output argument with declared type
qbit[], the size is determined by the actual array size, while sign and fraction-digits default toFalse(UNSIGNED) and 0 respectively. Specifically, this is the rule when usingallocate. - When the variable is passed to a function as its output argument with declared type
qnum, the size is determined by the actual size, sign, and fraction-digits of the function's output. - When the variable is initialized on the left of an out-of-place assignment
=, the domain of the expression determines its representation properties. - Variables retain their type, including the representation attributes, even after being
un-initialized (for example, when occurring on the left side of a bind statement).
Subsequent initializations must agree with the specific
qnumtype. - On the right side of a bind statement (
->) the representation attributes ofqnumvariables must already be known either through declaration, or by previous initialization (and subsequent un-initialization).
Examples
The following example demonstrates the default and explicit numeric interpretation of
quantum states. Two variables, a and b, are initialized to some quantum state.
a is left with the default unsigned integer interpretation. b is initialized to a
superposition of the bit strings 01 and 10 interpreted with a sign bit and one
fraction digit. This implies that its domain is [-1.0, -0.5, 0, 0.5] and its value
is in a superposition of -1.0 and 0.5. res is accordingly uniformly distributed on the 8
possible addition values.
qfunc main(output a: qnum, output b: qnum<2, SIGNED, 1>, output res: qnum) {
allocate(2, a);
hadamard_transform(a);
prepare_state([0, 0.5, 0.5, 0], 0, b);
res = a + b;
}
from classiq import (
Output,
QNum,
SIGNED,
qfunc,
prepare_state,
allocate,
hadamard_transform,
)
@qfunc
def main(a: Output[QNum], b: Output[QNum[2, SIGNED, 1]], res: Output[QNum]) -> None:
allocate(2, a) # 'a' is a 3 qubit unsigned int in the domain [0, 1, 2, 3]
hadamard_transform(a) # 'a' is in a superposition of all values in its domain
prepare_state([0, 0.5, 0.5, 0], 0, b) # 'b' is in superposition of 01 and 10
res |= a + b
Allocate-num
To allocate a qnum in the zero state with a specific numeric interpretation, use the
function allocate_num().
qfunc allocate_num<num_qubits: int, is_signed: bool, fraction_digits: int>(output out: qnum<num_qubits, is_signed, fraction_digits>);
def allocate_num(
num_qubits: CInt,
is_signed: CBool,
fraction_digits: CInt,
out: Output[QNum["num_qubits", "is_signed", "fraction_digits"]],
) -> None:
pass
Quantum arrays
A quantum array is an object that supports indexed access to parts of its state - its elements. Elements are interpreted as values of the specified array element type. A quantum array is one object with respect to its lifetime. Elements of an array cannot be initialized separately or bound separately to other variables. Also, an array's length (the number of elements it represents) is fixed at the time of its initialization and remains constant throughout its lifetime.
Syntax
element-type [ [ _length-expr- ]
In Python the class QArray is used as type hints in the declaration of arguments:
name : QArray [ [ element-type [ , length-expr ] ] ]
The same class is used to declare local variables:
name = QArray ( " name " [ , [ element_type =** ] element-type ][ **,** [ **length =** ] length-expr ] )**
Semantics
- element_type optionally determines the type of the array elements. Arrays are homogenous,
that is, all elements are of the same type. When left unspecified, the type defaults to
qbit. - length-expr optionally determines the number of elements in the array. The overall size of the array is its length multiplied by the size of the element type. When the length is unspecified, it is determined upon initialization based on the element size. Similarly, when the size of the element type is not specified, it is inferred upon initialization based on the length. Either the length, or the size of the element type, must be specified in the declaration
- The length cannot change throughout the lifetime of an array.
Expressions of quantum array type support the following operations:
- Subscript: array-expression [ index-expression ]
- Slice: array-expression [ from-index-expression : to-index-expression ]
- Length: array-expression . len
Examples
In the following example, a Boolean expression of a 3-SAT formula is evaluated over
the elements of a qubit array, which is prepared in the state of uniform superposition.
Note that bitwise operators are used in this case, but equivalent logical operators
and, or, and not (and their respective Python counterparts in package qmod.symbolic)
are also supported.
qfunc main(output x: qbit[3], output res: qbit) {
allocate(3, x);
hadamard_transform(x);
res = (x[0] | ~x[1] | ~x[2]) & (~x[0] | x[1] | ~x[2]);
}
from classiq import qfunc, Output, QArray, QBit, hadamard_transform, bind
@qfunc
def main(x: Output[QArray[QBit, 3]], res: Output[QBit]) -> None:
allocate(3, x)
hadamard_transform(x)
res |= (x[0] | ~x[1] | ~x[2]) & (~x[0] | x[1] | ~x[2])
The next example demonstrates the initialization of a numeric array using the bind
statement (->). Two numeric variables are declared and initialized separately and
subsequently bound together to initialize the array. The declared type of these
variables is an unsigned integer, but the declared element type of the array is signed. Hence,
the values 6 and 7 are interpreted as -2 and -1, respectively. When executing the resulting
quantum program, res is sampled with the value -3 (with probability 1).
qfunc main(output res: qnum) {
n0: qnum<3>;
prepare_int(6, n0);
n1: qnum<3>;
prepare_int(7, n1);
n_arr: qnum<3, SIGNED, 0>[];
{n0, n1} -> n_arr;
res = n_arr[0] + n_arr[1];
}
from classiq import qfunc, Output, QArray, QNum, SIGNED, prepare_int, bind
@qfunc
def main(res: Output[QNum]) -> None:
n0 = QNum("n0", 3)
prepare_int(6, n0)
n1 = QNum("n1", 3)
prepare_int(7, n1)
n_arr = QArray("n_arr", QNum[3, SIGNED, 0])
bind([n0, n1], n_arr)
res |= n_arr[0] + n_arr[1]
Quantum structs
A quantum struct is an object that supports named access to parts of its state - its fields. Each field corresponds to a slice of the overall object, interpreted according to its declared type. A quantum struct is one object with respect to its lifetime. Fields of a struct cannot be initialized separately or bound separately to other variables.
Quantum structs are typically used to pack and unpack multiple variables, that is, to switch between contexts that treat the object in a generic way (as a qubit array) and in a problem-specific way (to capture expressions over fields).
Syntax
The following syntax is used to define a quantum struct type -
qstruct name { field_declarations }
field-declarations is a list of one or more field declarations in the form - name : quantum-type ;.
A quantum struct type in Python is defined using a Python class derived from the
class QStruct. Fields are declared with type hints, similar to how member variables
are declared in a Python dataclass.
Semantics
- Only quantum types are allowed as field types in a quantum struct.
- Quantum structs may be arbitrarily nested, that is, a field of a struct may itself be a struct or a struct array. However, recursive struct types are not allowed.
- The overall size of a struct (the number of qubits used to store it) must be known upon declaration. This means that the size of all fields, except at most one, must be fully specified.
Expressions of quantum struct type support field-access operation in the form - struct-expression . field-name.
Examples
In the following example, quantum struct MyQStruct is defined and subsequently initialized
and prepared in a specific state in function main.
qstruct MyQStruct {
a: qbit;
b: qnum;
}
qfunc main(output s: MyQStruct) {
allocate(4, s);
H(s.a);
inplace_prepare_int(6, s.b);
}
from classiq import qfunc, Output, QBit, allocate, QStruct, QNum, inplace_prepare_int, H
class MyQStruct(QStruct):
a: QBit
b: QNum
@qfunc
def main(s: Output[MyQStruct]) -> None:
allocate(4, s)
H(s.a)
inplace_prepare_int(6, s.b)
The example below demonstrates the common situation where an algorithm alternates between
the two views of a quantum state - the structured view with partition into problem variables,
and the unstructured view as an array of qubits. The example defines a constraint over
two variables, a and b, of different numeric types. It uses Grover-search to find a
solution. In the Grover-search algorithm, encapsulated by the function grover_search, the
oracle application uses the structured view of the state to evaluate the constraint, while the
diffuser is defined in a generic way and uses the qubit array view of the state.
qstruct MyProblem {
a: qnum<2, UNSIGNED, 2>;
b: qnum<3, UNSIGNED, 3>;
}
qfunc my_problem_constraint(p: MyProblem, res: qbit) {
res ^= (p.a + p.b) == 0.625;
}
qfunc main(output p: MyProblem) {
allocate(5, p);
grover_search(2, lambda(p) {
phase_oracle(my_problem_constraint, p);
}, p);
}
from classiq import (
qfunc,
Output,
QBit,
allocate,
QStruct,
QNum,
UNSIGNED,
X,
grover_search,
phase_oracle,
)
class MyProblem(QStruct):
a: QNum[2, UNSIGNED, 2]
b: QNum[3, UNSIGNED, 3]
@qfunc
def my_problem_constraint(p: MyProblem, res: QBit) -> None:
res ^= p.a + p.b == 0.625
@qfunc
def main(p: Output[MyProblem]) -> None:
allocate(5, p)
grover_search(2, lambda p: phase_oracle(my_problem_constraint, p), p)
Executing this model will sample a state representing a solution to the problem in very high probability. This is an example of an output. Here is an output example:
state={'p': {'a': 0.0, 'b': 0.625}} shots=350
state={'p': {'a': 0.25, 'b': 0.375}} shots=344
state={'p': {'a': 0.5, 'b': 0.125}} shots=306